Update aom to v1.0.0

Update aom to commit id d14c5bb4f336ef1842046089849dee4a301fbbf0.

third_party/aom/test/accounting_test.cc

@@ -7,7 +7,7 @@
  * obtain it at www.aomedia.org/license/software. If the Alliance for Open
  * Media Patent License 1.0 was not distributed with this source code in the
  * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
+ */
 
 #include <math.h>
 #include <stdlib.h>
@@ -35,10 +35,7 @@ TEST(AV1, TestAccounting) {
   }
   aom_stop_encode(&bw);
   aom_reader br;
-#if CONFIG_ANS && ANS_MAX_SYMBOLS
-  br.window_size = 1 << 16;
-#endif
-  aom_reader_init(&br, bw_buffer, bw.pos, NULL, NULL);
+  aom_reader_init(&br, bw_buffer, bw.pos);
 
   Accounting accounting;
   aom_accounting_init(&accounting);
@@ -54,7 +51,7 @@ TEST(AV1, TestAccounting) {
   GTEST_ASSERT_EQ(accounting.syms.num_syms, 0);
 
   // Should record 2 * kSymbols accounting symbols.
-  aom_reader_init(&br, bw_buffer, bw.pos, NULL, NULL);
+  aom_reader_init(&br, bw_buffer, bw.pos);
   br.accounting = &accounting;
   for (int i = 0; i < kSymbols; i++) {
     aom_read(&br, 32, "A");

third_party/aom/test/acm_random.h

@@ -36,6 +36,19 @@ class ACMRandom {
     return (value >> 15) & 0xffff;
   }
 
+  int16_t Rand15Signed(void) {
+    const uint32_t value =
+        random_.Generate(testing::internal::Random::kMaxRange);
+    return (value >> 17) & 0xffff;
+  }
+
+  uint16_t Rand12(void) {
+    const uint32_t value =
+        random_.Generate(testing::internal::Random::kMaxRange);
+    // There's a bit more entropy in the upper bits of this implementation.
+    return (value >> 19) & 0xfff;
+  }
+
   int16_t Rand9Signed(void) {
     // Use 9 bits: values between 255 (0x0FF) and -256 (0x100).
     const uint32_t value = random_.Generate(512);

third_party/aom/test/active_map_refresh_test.cc

@@ -1,129 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
-
-#include <algorithm>
-#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
-#include "test/codec_factory.h"
-#include "test/encode_test_driver.h"
-#include "test/util.h"
-#include "test/y4m_video_source.h"
-
-namespace {
-
-// Check if any pixel in a 16x16 macroblock varies between frames.
-int CheckMb(const aom_image_t &current, const aom_image_t &previous, int mb_r,
-            int mb_c) {
-  for (int plane = 0; plane < 3; plane++) {
-    int r = 16 * mb_r;
-    int c0 = 16 * mb_c;
-    int r_top = std::min(r + 16, static_cast<int>(current.d_h));
-    int c_top = std::min(c0 + 16, static_cast<int>(current.d_w));
-    r = std::max(r, 0);
-    c0 = std::max(c0, 0);
-    if (plane > 0 && current.x_chroma_shift) {
-      c_top = (c_top + 1) >> 1;
-      c0 >>= 1;
-    }
-    if (plane > 0 && current.y_chroma_shift) {
-      r_top = (r_top + 1) >> 1;
-      r >>= 1;
-    }
-    for (; r < r_top; ++r) {
-      for (int c = c0; c < c_top; ++c) {
-        if (current.planes[plane][current.stride[plane] * r + c] !=
-            previous.planes[plane][previous.stride[plane] * r + c])
-          return 1;
-      }
-    }
-  }
-  return 0;
-}
-
-void GenerateMap(int mb_rows, int mb_cols, const aom_image_t &current,
-                 const aom_image_t &previous, uint8_t *map) {
-  for (int mb_r = 0; mb_r < mb_rows; ++mb_r) {
-    for (int mb_c = 0; mb_c < mb_cols; ++mb_c) {
-      map[mb_r * mb_cols + mb_c] = CheckMb(current, previous, mb_r, mb_c);
-    }
-  }
-}
-
-const int kAqModeCyclicRefresh = 3;
-
-class ActiveMapRefreshTest
-    : public ::libaom_test::CodecTestWith2Params<libaom_test::TestMode, int>,
-      public ::libaom_test::EncoderTest {
- protected:
-  ActiveMapRefreshTest() : EncoderTest(GET_PARAM(0)) {}
-  virtual ~ActiveMapRefreshTest() {}
-
-  virtual void SetUp() {
-    InitializeConfig();
-    SetMode(GET_PARAM(1));
-    cpu_used_ = GET_PARAM(2);
-  }
-
-  virtual void PreEncodeFrameHook(::libaom_test::VideoSource *video,
-                                  ::libaom_test::Encoder *encoder) {
-    ::libaom_test::Y4mVideoSource *y4m_video =
-        static_cast<libaom_test::Y4mVideoSource *>(video);
-    if (video->frame() == 1) {
-      encoder->Control(AOME_SET_CPUUSED, cpu_used_);
-      encoder->Control(AV1E_SET_AQ_MODE, kAqModeCyclicRefresh);
-    } else if (video->frame() >= 2 && video->img()) {
-      aom_image_t *current = video->img();
-      aom_image_t *previous = y4m_holder_->img();
-      ASSERT_TRUE(previous != NULL);
-      aom_active_map_t map = aom_active_map_t();
-      const int width = static_cast<int>(current->d_w);
-      const int height = static_cast<int>(current->d_h);
-      const int mb_width = (width + 15) / 16;
-      const int mb_height = (height + 15) / 16;
-      uint8_t *active_map = new uint8_t[mb_width * mb_height];
-      GenerateMap(mb_height, mb_width, *current, *previous, active_map);
-      map.cols = mb_width;
-      map.rows = mb_height;
-      map.active_map = active_map;
-      encoder->Control(AOME_SET_ACTIVEMAP, &map);
-      delete[] active_map;
-    }
-    if (video->img()) {
-      y4m_video->SwapBuffers(y4m_holder_);
-    }
-  }
-
-  int cpu_used_;
-  ::libaom_test::Y4mVideoSource *y4m_holder_;
-};
-
-TEST_P(ActiveMapRefreshTest, Test) {
-  cfg_.g_lag_in_frames = 0;
-  cfg_.g_profile = 1;
-  cfg_.rc_target_bitrate = 600;
-  cfg_.rc_resize_mode = 0;
-  cfg_.rc_min_quantizer = 8;
-  cfg_.rc_max_quantizer = 30;
-  cfg_.g_pass = AOM_RC_ONE_PASS;
-  cfg_.rc_end_usage = AOM_CBR;
-  cfg_.kf_max_dist = 90000;
-
-  ::libaom_test::Y4mVideoSource video("desktop_credits.y4m", 0, 10);
-  ::libaom_test::Y4mVideoSource video_holder("desktop_credits.y4m", 0, 10);
-  video_holder.Begin();
-  y4m_holder_ = &video_holder;
-
-  ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
-}
-
-AV1_INSTANTIATE_TEST_CASE(ActiveMapRefreshTest,
-                          ::testing::Values(::libaom_test::kRealTime),
-                          ::testing::Range(5, 6));
-}  // namespace

third_party/aom/test/active_map_test.cc

@@ -7,7 +7,7 @@
  * obtain it at www.aomedia.org/license/software. If the Alliance for Open
  * Media Patent License 1.0 was not distributed with this source code in the
  * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
+ */
 
 #include <climits>
 #include <vector>

third_party/aom/test/android/Android.mk

@@ -1,58 +0,0 @@
-#
-# Copyright (c) 2016, Alliance for Open Media. All rights reserved
-#
-# This source code is subject to the terms of the BSD 2 Clause License and
-# the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
-# was not distributed with this source code in the LICENSE file, you can
-# obtain it at www.aomedia.org/license/software. If the Alliance for Open
-# Media Patent License 1.0 was not distributed with this source code in the
-# PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-#
-# This make file builds aom_test app for android.
-# The test app itself runs on the command line through adb shell
-# The paths are really messed up as the libaom make file
-# expects to be made from a parent directory.
-CUR_WD := $(call my-dir)
-BINDINGS_DIR := $(CUR_WD)/../../..
-LOCAL_PATH := $(CUR_WD)/../../..
-
-#libwebm
-include $(CLEAR_VARS)
-include $(BINDINGS_DIR)/libaom/third_party/libwebm/Android.mk
-LOCAL_PATH := $(CUR_WD)/../../..
-
-#libaom
-include $(CLEAR_VARS)
-LOCAL_STATIC_LIBRARIES := libwebm
-include $(BINDINGS_DIR)/libaom/build/make/Android.mk
-LOCAL_PATH := $(CUR_WD)/../..
-
-#libgtest
-include $(CLEAR_VARS)
-LOCAL_ARM_MODE := arm
-LOCAL_CPP_EXTENSION := .cc
-LOCAL_MODULE := gtest
-LOCAL_C_INCLUDES := $(LOCAL_PATH)/third_party/googletest/src/googletest/src
-LOCAL_C_INCLUDES += $(LOCAL_PATH)/third_party/googletest/src/googletest/include
-LOCAL_SRC_FILES := ./third_party/googletest/src/googletest/src/gtest-all.cc
-include $(BUILD_STATIC_LIBRARY)
-
-#libaom_test
-include $(CLEAR_VARS)
-LOCAL_ARM_MODE := arm
-LOCAL_MODULE := libaom_test
-LOCAL_STATIC_LIBRARIES := gtest libwebm
-
-ifeq ($(ENABLE_SHARED),1)
-  LOCAL_SHARED_LIBRARIES := aom
-else
-  LOCAL_STATIC_LIBRARIES += aom
-endif
-
-include $(LOCAL_PATH)/test/test.mk
-LOCAL_C_INCLUDES := $(BINDINGS_DIR)
-FILTERED_SRC := $(sort $(filter %.cc %.c, $(LIBAOM_TEST_SRCS-yes)))
-LOCAL_SRC_FILES := $(addprefix ./test/, $(FILTERED_SRC))
-# some test files depend on *_rtcd.h, ensure they're generated first.
-$(eval $(call rtcd_dep_template))
-include $(BUILD_EXECUTABLE)

third_party/aom/test/android/README

@@ -1,32 +0,0 @@
-Android.mk will build aom unittests on android.
-1) Configure libaom from the parent directory:
-./libaom/configure --target=armv7-android-gcc --enable-external-build \
-  --enable-postproc --disable-install-srcs --enable-multi-res-encoding \
-  --enable-temporal-denoising --disable-unit-tests --disable-install-docs \
-  --disable-examples --disable-runtime-cpu-detect --sdk-path=$NDK
-
-2) From the parent directory, invoke ndk-build:
-NDK_PROJECT_PATH=. ndk-build APP_BUILD_SCRIPT=./libaom/test/android/Android.mk \
-  APP_ABI=armeabi-v7a APP_PLATFORM=android-18 APP_OPTIM=release \
-  APP_STL=gnustl_static
-
-Note: Both adb and ndk-build are available prebuilt at:
-  https://chromium.googlesource.com/android_tools
-
-3) Run get_files.py to download the test files:
-python get_files.py -i /path/to/test-data.sha1 -o /path/to/put/files \
-  -u http://downloads.webmproject.org/test_data/libaom
-
-4) Transfer files to device using adb. Ensure you have proper permissions for
-the target
-
-adb push /path/to/test_files /data/local/tmp
-adb push /path/to/built_libs /data/local/tmp
-
-NOTE: Built_libs defaults to parent_dir/libs/armeabi-v7a
-
-5) Run tests:
-adb shell
-(on device)
-cd /data/local/tmp
-LD_LIBRARY_PATH=. ./aom_test

third_party/aom/test/android/get_files.py

@@ -1,120 +0,0 @@
-#
-# Copyright (c) 2016, Alliance for Open Media. All rights reserved
-#
-# This source code is subject to the terms of the BSD 2 Clause License and
-# the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
-# was not distributed with this source code in the LICENSE file, you can
-# obtain it at www.aomedia.org/license/software. If the Alliance for Open
-# Media Patent License 1.0 was not distributed with this source code in the
-# PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-#
-# This simple script pulls test files from the webm homepage
-# It is intelligent enough to only pull files if
-#   1) File / test_data folder does not exist
-#   2) SHA mismatch
-
-import pycurl
-import csv
-import hashlib
-import re
-import os.path
-import time
-import itertools
-import sys
-import getopt
-
-#globals
-url = ''
-file_list_path = ''
-local_resource_path = ''
-
-# Helper functions:
-# A simple function which returns the sha hash of a file in hex
-def get_file_sha(filename):
-  try:
-    sha_hash = hashlib.sha1()
-    with open(filename, 'rb') as file:
-      buf = file.read(HASH_CHUNK)
-      while len(buf) > 0:
-        sha_hash.update(buf)
-        buf = file.read(HASH_CHUNK)
-      return sha_hash.hexdigest()
-  except IOError:
-    print "Error reading " + filename
-
-# Downloads a file from a url, and then checks the sha against the passed
-# in sha
-def download_and_check_sha(url, filename, sha):
-  path = os.path.join(local_resource_path, filename)
-  fp = open(path, "wb")
-  curl = pycurl.Curl()
-  curl.setopt(pycurl.URL, url + "/" + filename)
-  curl.setopt(pycurl.WRITEDATA, fp)
-  curl.perform()
-  curl.close()
-  fp.close()
-  return get_file_sha(path) == sha
-
-#constants
-ftp_retries = 3
-
-SHA_COL = 0
-NAME_COL = 1
-EXPECTED_COL = 2
-HASH_CHUNK = 65536
-
-# Main script
-try:
-  opts, args = \
-      getopt.getopt(sys.argv[1:], \
-                    "u:i:o:", ["url=", "input_csv=", "output_dir="])
-except:
-  print 'get_files.py -u <url> -i <input_csv> -o <output_dir>'
-  sys.exit(2)
-
-for opt, arg in opts:
-  if opt == '-u':
-    url = arg
-  elif opt in ("-i", "--input_csv"):
-    file_list_path = os.path.join(arg)
-  elif opt in ("-o", "--output_dir"):
-    local_resource_path = os.path.join(arg)
-
-if len(sys.argv) != 7:
-  print "Expects two paths and a url!"
-  exit(1)
-
-if not os.path.isdir(local_resource_path):
-  os.makedirs(local_resource_path)
-
-file_list_csv = open(file_list_path, "rb")
-
-# Our 'csv' file uses multiple spaces as a delimiter, python's
-# csv class only uses single character delimiters, so we convert them below
-file_list_reader = csv.reader((re.sub(' +', ' ', line) \
-    for line in file_list_csv), delimiter = ' ')
-
-file_shas = []
-file_names = []
-
-for row in file_list_reader:
-  if len(row) != EXPECTED_COL:
-      continue
-  file_shas.append(row[SHA_COL])
-  file_names.append(row[NAME_COL])
-
-file_list_csv.close()
-
-# Download files, only if they don't already exist and have correct shas
-for filename, sha in itertools.izip(file_names, file_shas):
-  path = os.path.join(local_resource_path, filename)
-  if os.path.isfile(path) \
-      and get_file_sha(path) == sha:
-    print path + ' exists, skipping'
-    continue
-  for retry in range(0, ftp_retries):
-    print "Downloading " + path
-    if not download_and_check_sha(url, filename, sha):
-      print "Sha does not match, retrying..."
-    else:
-      break

third_party/aom/test/android/scrape_gtest_log.py

@@ -1,60 +0,0 @@
-#
-# Copyright (c) 2016, Alliance for Open Media. All rights reserved
-#
-# This source code is subject to the terms of the BSD 2 Clause License and
-# the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
-# was not distributed with this source code in the LICENSE file, you can
-# obtain it at www.aomedia.org/license/software. If the Alliance for Open
-# Media Patent License 1.0 was not distributed with this source code in the
-# PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-#
-
-"""Standalone script which parses a gtest log for json.
-
-Json is returned returns as an array.  This script is used by the libaom
-waterfall to gather json results mixed in with gtest logs.  This is
-dubious software engineering.
-"""
-
-import getopt
-import json
-import os
-import re
-import sys
-
-
-def main():
-  if len(sys.argv) != 3:
-    print "Expects a file to write json to!"
-    exit(1)
-
-  try:
-    opts, _ = \
-        getopt.getopt(sys.argv[1:], \
-                      'o:', ['output-json='])
-  except getopt.GetOptError:
-    print 'scrape_gtest_log.py -o <output_json>'
-    sys.exit(2)
-
-  output_json = ''
-  for opt, arg in opts:
-    if opt in ('-o', '--output-json'):
-      output_json = os.path.join(arg)
-
-  blob = sys.stdin.read()
-  json_string = '[' + ','.join('{' + x + '}' for x in
-                               re.findall(r'{([^}]*.?)}', blob)) + ']'
-  print blob
-
-  output = json.dumps(json.loads(json_string), indent=4, sort_keys=True)
-  print output
-
-  path = os.path.dirname(output_json)
-  if path and not os.path.exists(path):
-    os.makedirs(path)
-
-  outfile = open(output_json, 'w')
-  outfile.write(output)
-
-if __name__ == '__main__':
-  sys.exit(main())

third_party/aom/test/ans_codec_test.cc

@@ -1,97 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
-
-#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
-
-#include "test/codec_factory.h"
-#include "test/encode_test_driver.h"
-#include "test/util.h"
-#include "test/y4m_video_source.h"
-#include "aom_dsp/ans.h"
-#include "av1/av1_dx_iface.c"
-
-// A note on ANS_MAX_SYMBOLS == 0:
-// Fused gtest doesn't work with EXPECT_FATAL_FAILURE [1]. Just run with a
-// single iteration and don't try to check the window size if we are unwindowed.
-// [1] https://github.com/google/googletest/issues/356
-
-namespace {
-
-const char kTestVideoName[] = "niklas_1280_720_30.y4m";
-const int kTestVideoFrames = 10;
-
-class AnsCodecTest : public ::libaom_test::CodecTestWithParam<int>,
-                     public ::libaom_test::EncoderTest {
- protected:
-  AnsCodecTest()
-      : EncoderTest(GET_PARAM(0)), ans_window_size_log2_(GET_PARAM(1)) {}
-
-  virtual ~AnsCodecTest() {}
-
-  virtual void SetUp() {
-    InitializeConfig();
-    SetMode(::libaom_test::kOnePassGood);
-    cfg_.g_lag_in_frames = 25;
-    cfg_.rc_end_usage = AOM_CQ;
-  }
-
-  virtual void PreEncodeFrameHook(::libaom_test::VideoSource *video,
-                                  ::libaom_test::Encoder *encoder) {
-    if (video->frame() == 1) {
-#if ANS_MAX_SYMBOLS
-      encoder->Control(AV1E_SET_ANS_WINDOW_SIZE_LOG2, ans_window_size_log2_);
-#endif
-      // Try to push a high symbol count through the codec
-      encoder->Control(AOME_SET_CQ_LEVEL, 8);
-      encoder->Control(AOME_SET_CPUUSED, 2);
-      encoder->Control(AOME_SET_ENABLEAUTOALTREF, 1);
-      encoder->Control(AOME_SET_ARNR_MAXFRAMES, 7);
-      encoder->Control(AOME_SET_ARNR_STRENGTH, 5);
-      encoder->Control(AV1E_SET_TILE_COLUMNS, 0);
-      encoder->Control(AV1E_SET_TILE_ROWS, 0);
-    }
-  }
-
-  virtual bool HandleDecodeResult(const aom_codec_err_t res_dec,
-                                  libaom_test::Decoder *decoder) {
-    aom_codec_ctx_t *const av1_decoder = decoder->GetDecoder();
-#if ANS_MAX_SYMBOLS
-    aom_codec_alg_priv_t *const priv =
-        reinterpret_cast<aom_codec_alg_priv_t *>(av1_decoder->priv);
-    FrameWorkerData *const worker_data =
-        reinterpret_cast<FrameWorkerData *>(priv->frame_workers[0].data1);
-    AV1_COMMON *const common = &worker_data->pbi->common;
-
-    EXPECT_EQ(ans_window_size_log2_, common->ans_window_size_log2);
-#endif
-
-    EXPECT_EQ(AOM_CODEC_OK, res_dec) << decoder->DecodeError();
-    return AOM_CODEC_OK == res_dec;
-  }
-
- private:
-  int ans_window_size_log2_;
-};
-
-TEST_P(AnsCodecTest, BitstreamParms) {
-  testing::internal::scoped_ptr<libaom_test::VideoSource> video(
-      new libaom_test::Y4mVideoSource(kTestVideoName, 0, kTestVideoFrames));
-  ASSERT_TRUE(video.get() != NULL);
-
-  ASSERT_NO_FATAL_FAILURE(RunLoop(video.get()));
-}
-
-#if ANS_MAX_SYMBOLS
-AV1_INSTANTIATE_TEST_CASE(AnsCodecTest, ::testing::Range(8, 24));
-#else
-AV1_INSTANTIATE_TEST_CASE(AnsCodecTest, ::testing::Range(0, 1));
-#endif
-}  // namespace

third_party/aom/test/ans_test.cc

@@ -1,213 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <assert.h>
-#include <math.h>
-#include <stdio.h>
-#include <ctime>
-#include <utility>
-#include <vector>
-
-#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
-
-#include "test/acm_random.h"
-#include "aom_dsp/ansreader.h"
-#include "aom_dsp/buf_ans.h"
-
-namespace {
-typedef std::vector<std::pair<uint8_t, bool> > PvVec;
-
-const int kPrintStats = 0;
-// Use a small buffer size to exercise ANS window spills or buffer growth
-const int kBufAnsSize = 1 << 8;
-
-PvVec abs_encode_build_vals(int iters) {
-  PvVec ret;
-  libaom_test::ACMRandom gen(0x30317076);
-  double entropy = 0;
-  for (int i = 0; i < iters; ++i) {
-    uint8_t p;
-    do {
-      p = gen.Rand8();
-    } while (p == 0);  // zero is not a valid coding probability
-    bool b = gen.Rand8() < p;
-    ret.push_back(std::make_pair(static_cast<uint8_t>(p), b));
-    if (kPrintStats) {
-      double d = p / 256.;
-      entropy += -d * log2(d) - (1 - d) * log2(1 - d);
-    }
-  }
-  if (kPrintStats) printf("entropy %f\n", entropy);
-  return ret;
-}
-
-bool check_rabs(const PvVec &pv_vec, uint8_t *buf) {
-  BufAnsCoder a;
-  a.size = kBufAnsSize;
-  aom_buf_ans_alloc(&a, NULL);
-  buf_ans_write_init(&a, buf);
-
-  std::clock_t start = std::clock();
-  for (PvVec::const_iterator it = pv_vec.begin(); it != pv_vec.end(); ++it) {
-    buf_rabs_write(&a, it->second, 256 - it->first);
-  }
-  aom_buf_ans_flush(&a);
-  std::clock_t enc_time = std::clock() - start;
-  int offset = buf_ans_write_end(&a);
-  aom_buf_ans_free(&a);
-  bool okay = true;
-  AnsDecoder d;
-#if ANS_MAX_SYMBOLS
-  d.window_size = kBufAnsSize;
-#endif
-  if (ans_read_init(&d, buf, offset)) return false;
-  start = std::clock();
-  for (PvVec::const_iterator it = pv_vec.begin(); it != pv_vec.end(); ++it) {
-    okay = okay && (rabs_read(&d, 256 - it->first) != 0) == it->second;
-  }
-  std::clock_t dec_time = std::clock() - start;
-  if (!okay) return false;
-  if (kPrintStats)
-    printf("uABS size %d enc_time %f dec_time %f\n", offset,
-           static_cast<float>(enc_time) / CLOCKS_PER_SEC,
-           static_cast<float>(dec_time) / CLOCKS_PER_SEC);
-  return ans_read_end(&d) != 0;
-}
-
-const aom_cdf_prob spareto65[] = { 8320, 6018, 4402, 3254, 4259,
-                                   3919, 2057, 492,  45,   2 };
-
-const int kRansSymbols =
-    static_cast<int>(sizeof(spareto65) / sizeof(spareto65[0]));
-
-struct rans_sym {
-  aom_cdf_prob prob;
-  aom_cdf_prob cum_prob;  // not-inclusive
-};
-
-std::vector<int> ans_encode_build_vals(rans_sym *const tab, int iters) {
-  aom_cdf_prob sum = 0;
-  for (int i = 0; i < kRansSymbols; ++i) {
-    tab[i].cum_prob = sum;
-    tab[i].prob = spareto65[i];
-    sum += spareto65[i];
-  }
-  std::vector<int> p_to_sym;
-  for (int i = 0; i < kRansSymbols; ++i) {
-    p_to_sym.insert(p_to_sym.end(), tab[i].prob, i);
-  }
-  assert(p_to_sym.size() == RANS_PRECISION);
-  std::vector<int> ret;
-  libaom_test::ACMRandom gen(18543637);
-  for (int i = 0; i < iters; ++i) {
-    int sym =
-        p_to_sym[((gen.Rand8() << 8) + gen.Rand8()) & (RANS_PRECISION - 1)];
-    ret.push_back(sym);
-  }
-  return ret;
-}
-
-void rans_build_dec_tab(const struct rans_sym sym_tab[],
-                        aom_cdf_prob *dec_tab) {
-  unsigned int sum = 0;
-  for (int i = 0; sum < RANS_PRECISION; ++i) {
-    dec_tab[i] = sum += sym_tab[i].prob;
-  }
-}
-
-bool check_rans(const std::vector<int> &sym_vec, const rans_sym *const tab,
-                uint8_t *buf) {
-  BufAnsCoder a;
-  a.size = kBufAnsSize;
-  aom_buf_ans_alloc(&a, NULL);
-  buf_ans_write_init(&a, buf);
-  aom_cdf_prob dec_tab[kRansSymbols];
-  rans_build_dec_tab(tab, dec_tab);
-
-  std::clock_t start = std::clock();
-  for (std::vector<int>::const_iterator it = sym_vec.begin();
-       it != sym_vec.end(); ++it) {
-    buf_rans_write(&a, tab[*it].cum_prob, tab[*it].prob);
-  }
-  aom_buf_ans_flush(&a);
-  std::clock_t enc_time = std::clock() - start;
-  int offset = buf_ans_write_end(&a);
-  aom_buf_ans_free(&a);
-  bool okay = true;
-  AnsDecoder d;
-#if ANS_MAX_SYMBOLS
-  d.window_size = kBufAnsSize;
-#endif
-  if (ans_read_init(&d, buf, offset)) return false;
-  start = std::clock();
-  for (std::vector<int>::const_iterator it = sym_vec.begin();
-       it != sym_vec.end(); ++it) {
-    okay &= rans_read(&d, dec_tab) == *it;
-  }
-  std::clock_t dec_time = std::clock() - start;
-  if (!okay) return false;
-  if (kPrintStats)
-    printf("rANS size %d enc_time %f dec_time %f\n", offset,
-           static_cast<float>(enc_time) / CLOCKS_PER_SEC,
-           static_cast<float>(dec_time) / CLOCKS_PER_SEC);
-  return ans_read_end(&d) != 0;
-}
-
-class AbsTestFix : public ::testing::Test {
- protected:
-  static void SetUpTestCase() { pv_vec_ = abs_encode_build_vals(kNumBools); }
-  virtual void SetUp() { buf_ = new uint8_t[kNumBools / 8]; }
-  virtual void TearDown() { delete[] buf_; }
-  static const int kNumBools = 100000000;
-  static PvVec pv_vec_;
-  uint8_t *buf_;
-};
-PvVec AbsTestFix::pv_vec_;
-
-class AnsTestFix : public ::testing::Test {
- protected:
-  static void SetUpTestCase() {
-    sym_vec_ = ans_encode_build_vals(rans_sym_tab_, kNumSyms);
-  }
-  virtual void SetUp() { buf_ = new uint8_t[kNumSyms / 2]; }
-  virtual void TearDown() { delete[] buf_; }
-  static const int kNumSyms = 25000000;
-  static std::vector<int> sym_vec_;
-  static rans_sym rans_sym_tab_[kRansSymbols];
-  uint8_t *buf_;
-};
-std::vector<int> AnsTestFix::sym_vec_;
-rans_sym AnsTestFix::rans_sym_tab_[kRansSymbols];
-
-TEST_F(AbsTestFix, Rabs) { EXPECT_TRUE(check_rabs(pv_vec_, buf_)); }
-TEST_F(AnsTestFix, Rans) {
-  EXPECT_TRUE(check_rans(sym_vec_, rans_sym_tab_, buf_));
-}
-TEST(AnsTest, FinalStateSerialization) {
-  for (unsigned i = L_BASE; i < L_BASE * IO_BASE; ++i) {
-    uint8_t buf[8];
-    AnsCoder c;
-    ans_write_init(&c, buf);
-    c.state = i;
-    const int written_size = ans_write_end(&c);
-    ASSERT_LT(static_cast<size_t>(written_size), sizeof(buf));
-    AnsDecoder d;
-#if ANS_MAX_SYMBOLS
-    // There is no real data window here because no symbols are sent through
-    // ans (only synthetic states), so use a dummy value
-    d.window_size = 1024;
-#endif
-    const int read_init_status = ans_read_init(&d, buf, written_size);
-    EXPECT_EQ(read_init_status, 0);
-    EXPECT_EQ(d.state, i);
-  }
-}
-}  // namespace

third_party/aom/test/aom_integer_test.cc

@@ -0,0 +1,177 @@
+/*
+ * Copyright (c) 2018, Alliance for Open Media. All rights reserved
+ *
+ * This source code is subject to the terms of the BSD 2 Clause License and
+ * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
+ * was not distributed with this source code in the LICENSE file, you can
+ * obtain it at www.aomedia.org/license/software. If the Alliance for Open
+ * Media Patent License 1.0 was not distributed with this source code in the
+ * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
+ */
+
+#include "aom/aom_integer.h"
+#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
+
+namespace {
+const uint64_t kMaximumLeb128CodedSize = 8;
+const uint8_t kLeb128PadByte = 0x80;  // Binary: 10000000
+const uint64_t kMaximumLeb128Value = UINT32_MAX;
+const uint32_t kSizeTestNumValues = 6;
+const uint32_t kSizeTestExpectedSizes[kSizeTestNumValues] = {
+  1, 1, 2, 3, 4, 5
+};
+const uint64_t kSizeTestInputs[kSizeTestNumValues] = {
+  0, 0x7f, 0x3fff, 0x1fffff, 0xffffff, 0x10000000
+};
+
+const uint8_t kOutOfRangeLeb128Value[5] = { 0x80, 0x80, 0x80, 0x80,
+                                            0x10 };  // UINT32_MAX + 1
+}  // namespace
+
+TEST(AomLeb128, DecodeTest) {
+  const size_t num_leb128_bytes = 3;
+  const uint8_t leb128_bytes[num_leb128_bytes] = { 0xE5, 0x8E, 0x26 };
+  const uint64_t expected_value = 0x98765;  // 624485
+  const size_t expected_length = 3;
+  uint64_t value = ~0ULL;  // make sure value is cleared by the function
+  size_t length;
+  ASSERT_EQ(
+      aom_uleb_decode(&leb128_bytes[0], num_leb128_bytes, &value, &length), 0);
+  ASSERT_EQ(expected_value, value);
+  ASSERT_EQ(expected_length, length);
+
+  // Make sure the decoder stops on the last marked LEB128 byte.
+  aom_uleb_decode(&leb128_bytes[0], num_leb128_bytes + 1, &value, &length);
+  ASSERT_EQ(expected_value, value);
+  ASSERT_EQ(expected_length, length);
+}
+
+TEST(AomLeb128, EncodeTest) {
+  const uint32_t test_value = 0x98765;  // 624485
+  const uint8_t expected_bytes[3] = { 0xE5, 0x8E, 0x26 };
+  const size_t kWriteBufferSize = 4;
+  uint8_t write_buffer[kWriteBufferSize] = { 0 };
+  size_t bytes_written = 0;
+  ASSERT_EQ(aom_uleb_encode(test_value, kWriteBufferSize, &write_buffer[0],
+                            &bytes_written),
+            0);
+  ASSERT_EQ(bytes_written, 3u);
+  for (size_t i = 0; i < bytes_written; ++i) {
+    ASSERT_EQ(write_buffer[i], expected_bytes[i]);
+  }
+}
+
+TEST(AomLeb128, EncodeDecodeTest) {
+  const uint32_t value = 0x98765;  // 624485
+  const size_t kWriteBufferSize = 4;
+  uint8_t write_buffer[kWriteBufferSize] = { 0 };
+  size_t bytes_written = 0;
+  ASSERT_EQ(aom_uleb_encode(value, kWriteBufferSize, &write_buffer[0],
+                            &bytes_written),
+            0);
+  ASSERT_EQ(bytes_written, 3u);
+  uint64_t decoded_value;
+  size_t decoded_length;
+  aom_uleb_decode(&write_buffer[0], bytes_written, &decoded_value,
+                  &decoded_length);
+  ASSERT_EQ(value, decoded_value);
+  ASSERT_EQ(bytes_written, decoded_length);
+}
+
+TEST(AomLeb128, FixedSizeEncodeTest) {
+  const uint32_t test_value = 0x123;
+  const uint8_t expected_bytes[4] = { 0xa3, 0x82, 0x80, 0x00 };
+  const size_t kWriteBufferSize = 4;
+  uint8_t write_buffer[kWriteBufferSize] = { 0 };
+  size_t bytes_written = 0;
+  ASSERT_EQ(0, aom_uleb_encode_fixed_size(test_value, kWriteBufferSize,
+                                          kWriteBufferSize, &write_buffer[0],
+                                          &bytes_written));
+  ASSERT_EQ(kWriteBufferSize, bytes_written);
+  for (size_t i = 0; i < bytes_written; ++i) {
+    ASSERT_EQ(write_buffer[i], expected_bytes[i]);
+  }
+}
+
+TEST(AomLeb128, FixedSizeEncodeDecodeTest) {
+  const uint32_t value = 0x1;
+  const size_t kWriteBufferSize = 4;
+  uint8_t write_buffer[kWriteBufferSize] = { 0 };
+  size_t bytes_written = 0;
+  ASSERT_EQ(
+      aom_uleb_encode_fixed_size(value, kWriteBufferSize, kWriteBufferSize,
+                                 &write_buffer[0], &bytes_written),
+      0);
+  ASSERT_EQ(bytes_written, 4u);
+  uint64_t decoded_value;
+  size_t decoded_length;
+  aom_uleb_decode(&write_buffer[0], bytes_written, &decoded_value,
+                  &decoded_length);
+  ASSERT_EQ(value, decoded_value);
+  ASSERT_EQ(bytes_written, decoded_length);
+}
+
+TEST(AomLeb128, SizeTest) {
+  for (size_t i = 0; i < kSizeTestNumValues; ++i) {
+    ASSERT_EQ(kSizeTestExpectedSizes[i],
+              aom_uleb_size_in_bytes(kSizeTestInputs[i]));
+  }
+}
+
+TEST(AomLeb128, DecodeFailTest) {
+  // Input buffer containing what would be a valid 9 byte LEB128 encoded
+  // unsigned integer.
+  const uint8_t kAllPadBytesBuffer[kMaximumLeb128CodedSize + 1] = {
+    kLeb128PadByte, kLeb128PadByte, kLeb128PadByte,
+    kLeb128PadByte, kLeb128PadByte, kLeb128PadByte,
+    kLeb128PadByte, kLeb128PadByte, 0
+  };
+  uint64_t decoded_value;
+
+  // Test that decode fails when result would be valid 9 byte integer.
+  ASSERT_EQ(aom_uleb_decode(&kAllPadBytesBuffer[0], kMaximumLeb128CodedSize + 1,
+                            &decoded_value, NULL),
+            -1);
+
+  // Test that encoded value missing terminator byte within available buffer
+  // range causes decode error.
+  ASSERT_EQ(aom_uleb_decode(&kAllPadBytesBuffer[0], kMaximumLeb128CodedSize,
+                            &decoded_value, NULL),
+            -1);
+
+  // Test that LEB128 input that decodes to a value larger than 32-bits fails.
+  size_t value_size = 0;
+  ASSERT_EQ(aom_uleb_decode(&kOutOfRangeLeb128Value[0],
+                            sizeof(kOutOfRangeLeb128Value), &decoded_value,
+                            &value_size),
+            -1);
+}
+
+TEST(AomLeb128, EncodeFailTest) {
+  const size_t kWriteBufferSize = 4;
+  const uint32_t kValidTestValue = 1;
+  uint8_t write_buffer[kWriteBufferSize] = { 0 };
+  size_t coded_size = 0;
+  ASSERT_EQ(
+      aom_uleb_encode(kValidTestValue, kWriteBufferSize, NULL, &coded_size),
+      -1);
+  ASSERT_EQ(aom_uleb_encode(kValidTestValue, kWriteBufferSize, &write_buffer[0],
+                            NULL),
+            -1);
+
+  const uint32_t kValueOutOfRangeForBuffer = 0xFFFFFFFF;
+  ASSERT_EQ(aom_uleb_encode(kValueOutOfRangeForBuffer, kWriteBufferSize,
+                            &write_buffer[0], &coded_size),
+            -1);
+
+  const uint64_t kValueOutOfRange = kMaximumLeb128Value + 1;
+  ASSERT_EQ(aom_uleb_encode(kValueOutOfRange, kWriteBufferSize,
+                            &write_buffer[0], &coded_size),
+            -1);
+
+  const size_t kPadSizeOutOfRange = 5;
+  ASSERT_EQ(aom_uleb_encode_fixed_size(kValidTestValue, kWriteBufferSize,
+                                       kPadSizeOutOfRange, &write_buffer[0],
+                                       &coded_size),
+            -1);
+}

third_party/aom/test/aomdec.sh

@@ -17,10 +17,12 @@
 # Environment check: Make sure input is available.
 aomdec_verify_environment() {
   if [ "$(av1_encode_available)" != "yes" ] ; then
-    if [ ! -e "${AV1_WEBM_FILE}" ] || \
-      [ ! -e "${AV1_FPM_WEBM_FILE}" ] || \
-      [ ! -e "${AV1_LT_50_FRAMES_WEBM_FILE}" ] ; then
-      elog "Libaom test data must exist in LIBAOM_TEST_DATA_PATH."
+    if [ ! -e "${AV1_IVF_FILE}" ] || \
+       [ ! -e "${AV1_OBU_ANNEXB_FILE}" ] || \
+       [ ! -e "${AV1_OBU_SEC5_FILE}" ] || \
+       [ ! -e "${AV1_WEBM_FILE}" ]; then
+      elog "Libaom test data must exist before running this test script when " \
+           " encoding is disabled. "
       return 1
     fi
   fi
@@ -38,10 +40,8 @@ aomdec_pipe() {
   local readonly input="$1"
   shift
   if [ ! -e "${input}" ]; then
-    local file="${AOM_TEST_OUTPUT_DIR}/test_encode.ivf"
-    encode_yuv_raw_input_av1 "${file}" --ivf
-  else
-    local file="${input}"
+    elog "Input file ($input) missing in aomdec_pipe()"
+    return 1
   fi
   cat "${file}" | aomdec - "$@" ${devnull}
 }
@@ -63,62 +63,85 @@ aomdec_can_decode_av1() {
   fi
 }
 
+aomdec_av1_ivf() {
+  if [ "$(aomdec_can_decode_av1)" = "yes" ]; then
+    local readonly file="${AV1_IVF_FILE}"
+    if [ ! -e "${file}" ]; then
+      encode_yuv_raw_input_av1 "${file}" --ivf
+    fi
+    aomdec "${AV1_IVF_FILE}" --summary --noblit
+  fi
+}
+
+aomdec_av1_ivf_error_resilient() {
+  if [ "$(aomdec_can_decode_av1)" = "yes" ]; then
+    local readonly file="av1.error-resilient.ivf"
+    if [ ! -e "${file}" ]; then
+      encode_yuv_raw_input_av1 "${file}" --ivf --error-resilient=1
+    fi
+    aomdec "${file}" --summary --noblit
+  fi
+}
+
+aomdec_av1_ivf_multithread() {
+  if [ "$(aomdec_can_decode_av1)" = "yes" ]; then
+    local readonly file="${AV1_IVF_FILE}"
+    if [ ! -e "${file}" ]; then
+      encode_yuv_raw_input_av1 "${file}" --ivf
+    fi
+    for threads in 2 3 4 5 6 7 8; do
+      aomdec "${file}" --summary --noblit --threads=$threads
+    done
+  fi
+}
+
 aomdec_aom_ivf_pipe_input() {
   if [ "$(aomdec_can_decode_av1)" = "yes" ]; then
-    aomdec_pipe "${AOM_IVF_FILE}" --summary --noblit
+    local readonly file="${AV1_IVF_FILE}"
+    if [ ! -e "${file}" ]; then
+      encode_yuv_raw_input_av1 "${file}" --ivf
+    fi
+    aomdec_pipe "${AV1_IVF_FILE}" --summary --noblit
+  fi
+}
+
+aomdec_av1_obu_annexb() {
+  if [ "$(aomdec_can_decode_av1)" = "yes" ]; then
+    local readonly file="${AV1_OBU_ANNEXB_FILE}"
+    if [ ! -e "${file}" ]; then
+      encode_yuv_raw_input_av1 "${file}" --obu --annexb=1
+    fi
+    aomdec "${file}" --summary --noblit --annexb
+  fi
+}
+
+aomdec_av1_obu_section5() {
+  if [ "$(aomdec_can_decode_av1)" = "yes" ]; then
+    local readonly file="${AV1_OBU_SEC5_FILE}"
+    if [ ! -e "${file}" ]; then
+      encode_yuv_raw_input_av1 "${file}" --obu
+    fi
+    aomdec "${file}" --summary --noblit
   fi
 }
 
 aomdec_av1_webm() {
   if [ "$(aomdec_can_decode_av1)" = "yes" ] && \
      [ "$(webm_io_available)" = "yes" ]; then
-    if [ ! -e "${AV1_WEBM_FILE}" ]; then
-      local file="${AOM_TEST_OUTPUT_DIR}/test_encode.webm"
+    local readonly file="${AV1_WEBM_FILE}"
+    if [ ! -e "${file}" ]; then
       encode_yuv_raw_input_av1 "${file}"
-    else
-      aomdec "${AV1_WEBM_FILE}" --summary --noblit
     fi
+    aomdec "${AV1_WEBM_FILE}" --summary --noblit
   fi
 }
 
-aomdec_av1_webm_frame_parallel() {
-  if [ "$(aomdec_can_decode_av1)" = "yes" ] && \
-     [ "$(webm_io_available)" = "yes" ]; then
-    local file
-    if [ ! -e "${AV1_WEBM_FILE}" ]; then
-      file="${AOM_TEST_OUTPUT_DIR}/test_encode.webm"
-      encode_yuv_raw_input_av1 "${file}" "--ivf --error-resilient=1 "
-    else
-      file="${AV1_FPM_WEBM_FILE}"
-    fi
-    for threads in 2 3 4 5 6 7 8; do
-      aomdec "${file}" --summary --noblit --threads=$threads \
-        --frame-parallel
-    done
-  fi
-}
-
-# TODO(vigneshv): Enable or remove this test and associated code.
-DISABLED_aomdec_av1_webm_less_than_50_frames() {
-  # ensure that reaching eof in webm_guess_framerate doesn't result in invalid
-  # frames in actual webm_read_frame calls.
-  if [ "$(aomdec_can_decode_av1)" = "yes" ] && \
-     [ "$(webm_io_available)" = "yes" ]; then
-    local readonly decoder="$(aom_tool_path aomdec)"
-    local readonly expected=10
-    local readonly num_frames=$(${AOM_TEST_PREFIX} "${decoder}" \
-      "${AV1_LT_50_FRAMES_WEBM_FILE}" --summary --noblit 2>&1 \
-      | awk '/^[0-9]+ decoded frames/ { print $1 }')
-    if [ "$num_frames" -ne "$expected" ]; then
-      elog "Output frames ($num_frames) != expected ($expected)"
-      return 1
-    fi
-  fi
-}
-
-aomdec_tests="aomdec_av1_webm
-              aomdec_av1_webm_frame_parallel
+aomdec_tests="aomdec_av1_ivf
+              aomdec_av1_ivf_error_resilient
+              aomdec_av1_ivf_multithread
               aomdec_aom_ivf_pipe_input
-              DISABLED_aomdec_av1_webm_less_than_50_frames"
+              aomdec_av1_obu_annexb
+              aomdec_av1_obu_section5
+              aomdec_av1_webm"
 
 run_tests aomdec_verify_environment "${aomdec_tests}"

third_party/aom/test/aomenc.sh

@@ -15,8 +15,6 @@
 ##
 . $(dirname $0)/tools_common.sh
 
-readonly TEST_FRAMES=5
-
 # Environment check: Make sure input is available.
 aomenc_verify_environment() {
   if [ ! -e "${YUV_RAW_INPUT}" ]; then
@@ -57,32 +55,6 @@ y4m_input_720p() {
   echo ""${Y4M_720P_INPUT}""
 }
 
-# Echo default aomenc real time encoding params. $1 is the codec, which defaults
-# to av1 if unspecified.
-aomenc_rt_params() {
-  local readonly codec="${1:-av1}"
-  echo "--codec=${codec}
-    --buf-initial-sz=500
-    --buf-optimal-sz=600
-    --buf-sz=1000
-    --cpu-used=-6
-    --end-usage=cbr
-    --error-resilient=1
-    --kf-max-dist=90000
-    --lag-in-frames=0
-    --max-intra-rate=300
-    --max-q=56
-    --min-q=2
-    --noise-sensitivity=0
-    --overshoot-pct=50
-    --passes=1
-    --profile=0
-    --resize-allowed=0
-    --rt
-    --static-thresh=0
-    --undershoot-pct=50"
-}
-
 # Wrapper function for running aomenc with pipe input. Requires that
 # LIBAOM_BIN_PATH points to the directory containing aomenc. $1 is used as the
 # input file path and shifted away. All remaining parameters are passed through
@@ -110,10 +82,12 @@ aomenc() {
 
 aomenc_av1_ivf() {
   if [ "$(aomenc_can_encode_av1)" = "yes" ]; then
-    local readonly output="${AOM_TEST_OUTPUT_DIR}/av1.ivf"
+    local output="${AV1_IVF_FILE}"
+    if [ -e "${AV1_IVF_FILE}" ]; then
+      output="${AOM_TEST_OUTPUT_DIR}/av1_test.ivf"
+    fi
     aomenc $(yuv_raw_input) \
-      --codec=av1 \
-      --limit="${TEST_FRAMES}" \
+      $(aomenc_encode_test_fast_params) \
       --ivf \
       --output="${output}"
 
@@ -124,13 +98,52 @@ aomenc_av1_ivf() {
   fi
 }
 
+aomenc_av1_obu_annexb() {
+   if [ "$(aomenc_can_encode_av1)" = "yes" ]; then
+    local output="${AV1_OBU_ANNEXB_FILE}"
+    if [ -e "${AV1_OBU_ANNEXB_FILE}" ]; then
+      output="${AOM_TEST_OUTPUT_DIR}/av1_test.annexb.obu"
+    fi
+    aomenc $(yuv_raw_input) \
+      $(aomenc_encode_test_fast_params) \
+      --obu \
+      --annexb=1 \
+      --output="${output}"
+
+    if [ ! -e "${output}" ]; then
+      elog "Output file does not exist."
+      return 1
+    fi
+  fi
+}
+
+aomenc_av1_obu_section5() {
+   if [ "$(aomenc_can_encode_av1)" = "yes" ]; then
+    local output="${AV1_OBU_SEC5_FILE}"
+    if [ -e "${AV1_OBU_SEC5_FILE}" ]; then
+      output="${AOM_TEST_OUTPUT_DIR}/av1_test.section5.obu"
+    fi
+    aomenc $(yuv_raw_input) \
+      $(aomenc_encode_test_fast_params) \
+      --obu \
+      --output="${output}"
+
+    if [ ! -e "${output}" ]; then
+      elog "Output file does not exist."
+      return 1
+    fi
+  fi
+}
+
 aomenc_av1_webm() {
   if [ "$(aomenc_can_encode_av1)" = "yes" ] && \
      [ "$(webm_io_available)" = "yes" ]; then
-    local readonly output="${AOM_TEST_OUTPUT_DIR}/av1.webm"
+    local output="${AV1_WEBM_FILE}"
+    if [ -e "${AV1_WEBM_FILE}" ]; then
+      output="${AOM_TEST_OUTPUT_DIR}/av1_test.webm"
+    fi
     aomenc $(yuv_raw_input) \
-      --codec=av1 \
-      --limit="${TEST_FRAMES}" \
+      $(aomenc_encode_test_fast_params) \
       --output="${output}"
 
     if [ ! -e "${output}" ]; then
@@ -140,15 +153,14 @@ aomenc_av1_webm() {
   fi
 }
 
-aomenc_av1_webm_2pass() {
+aomenc_av1_webm_1pass() {
   if [ "$(aomenc_can_encode_av1)" = "yes" ] && \
      [ "$(webm_io_available)" = "yes" ]; then
-    local readonly output="${AOM_TEST_OUTPUT_DIR}/av1.webm"
+    local readonly output="${AOM_TEST_OUTPUT_DIR}/av1_test.webm"
     aomenc $(yuv_raw_input) \
-      --codec=av1 \
-      --limit="${TEST_FRAMES}" \
-      --output="${output}" \
-      --passes=2
+      $(aomenc_encode_test_fast_params) \
+      --passes=1 \
+      --output="${output}"
 
     if [ ! -e "${output}" ]; then
       elog "Output file does not exist."
@@ -161,8 +173,7 @@ aomenc_av1_ivf_lossless() {
   if [ "$(aomenc_can_encode_av1)" = "yes" ]; then
     local readonly output="${AOM_TEST_OUTPUT_DIR}/av1_lossless.ivf"
     aomenc $(yuv_raw_input) \
-      --codec=av1 \
-      --limit="${TEST_FRAMES}" \
+      $(aomenc_encode_test_fast_params) \
       --ivf \
       --output="${output}" \
       --lossless=1
@@ -178,8 +189,7 @@ aomenc_av1_ivf_minq0_maxq0() {
   if [ "$(aomenc_can_encode_av1)" = "yes" ]; then
     local readonly output="${AOM_TEST_OUTPUT_DIR}/av1_lossless_minq0_maxq0.ivf"
     aomenc $(yuv_raw_input) \
-      --codec=av1 \
-      --limit="${TEST_FRAMES}" \
+      $(aomenc_encode_test_fast_params) \
       --ivf \
       --output="${output}" \
       --min-q=0 \
@@ -199,12 +209,10 @@ aomenc_av1_webm_lag5_frames10() {
     local readonly lag_frames=5
     local readonly output="${AOM_TEST_OUTPUT_DIR}/av1_lag5_frames10.webm"
     aomenc $(yuv_raw_input) \
-      --codec=av1 \
-      --limit="${lag_total_frames}" \
-      --lag-in-frames="${lag_frames}" \
-      --output="${output}" \
-      --passes=2 \
-      --auto-alt-ref=1
+      $(aomenc_encode_test_fast_params) \
+      --limit=${lag_total_frames} \
+      --lag-in-frames=${lag_frames} \
+      --output="${output}"
 
     if [ ! -e "${output}" ]; then
       elog "Output file does not exist."
@@ -219,8 +227,7 @@ aomenc_av1_webm_non_square_par() {
      [ "$(webm_io_available)" = "yes" ]; then
     local readonly output="${AOM_TEST_OUTPUT_DIR}/av1_non_square_par.webm"
     aomenc $(y4m_input_non_square_par) \
-      --codec=av1 \
-      --limit="${TEST_FRAMES}" \
+      $(aomenc_encode_test_fast_params) \
       --output="${output}"
 
     if [ ! -e "${output}" ]; then
@@ -230,12 +237,33 @@ aomenc_av1_webm_non_square_par() {
   fi
 }
 
+aomenc_av1_webm_cdf_update_mode() {
+  if [ "$(aomenc_can_encode_av1)" = "yes" ] && \
+     [ "$(webm_io_available)" = "yes" ]; then
+    for mode in 0 1 2; do
+      local readonly output="${AOM_TEST_OUTPUT_DIR}/cdf_mode_${mode}.webm"
+      aomenc $(yuv_raw_input) \
+        $(aomenc_encode_test_fast_params) \
+        --cdf-update-mode=${mode} \
+        --output="${output}"
+
+      if [ ! -e "${output}" ]; then
+        elog "Output file does not exist."
+        return 1
+      fi
+    done
+  fi
+}
+
 aomenc_tests="aomenc_av1_ivf
+              aomenc_av1_obu_annexb
+              aomenc_av1_obu_section5
               aomenc_av1_webm
-              aomenc_av1_webm_2pass
+              aomenc_av1_webm_1pass
               aomenc_av1_ivf_lossless
               aomenc_av1_ivf_minq0_maxq0
               aomenc_av1_webm_lag5_frames10
-              aomenc_av1_webm_non_square_par"
+              aomenc_av1_webm_non_square_par
+              aomenc_av1_webm_cdf_update_mode"
 
 run_tests aomenc_verify_environment "${aomenc_tests}"

third_party/aom/test/aq_segment_test.cc

@@ -7,9 +7,10 @@
  * obtain it at www.aomedia.org/license/software. If the Alliance for Open
  * Media Patent License 1.0 was not distributed with this source code in the
  * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
+ */
+
+#include "config/aom_config.h"
 
-#include "./aom_config.h"
 #include "third_party/googletest/src/googletest/include/gtest/gtest.h"
 #include "test/codec_factory.h"
 #include "test/encode_test_driver.h"
@@ -37,18 +38,14 @@ class AqSegmentTest
     if (video->frame() == 1) {
       encoder->Control(AOME_SET_CPUUSED, set_cpu_used_);
       encoder->Control(AV1E_SET_AQ_MODE, aq_mode_);
-#if CONFIG_EXT_DELTA_Q
       encoder->Control(AV1E_SET_DELTAQ_MODE, deltaq_mode_);
-#endif
       encoder->Control(AOME_SET_MAX_INTRA_BITRATE_PCT, 100);
     }
   }
 
   void DoTest(int aq_mode) {
     aq_mode_ = aq_mode;
-#if CONFIG_EXT_DELTA_Q
     deltaq_mode_ = 0;
-#endif
     cfg_.kf_max_dist = 12;
     cfg_.rc_min_quantizer = 8;
     cfg_.rc_max_quantizer = 56;
@@ -65,9 +62,7 @@ class AqSegmentTest
 
   int set_cpu_used_;
   int aq_mode_;
-#if CONFIG_EXT_DELTA_Q
   int deltaq_mode_;
-#endif
 };
 
 // Validate that this AQ segmentation mode (AQ=1, variance_ap)
@@ -90,21 +85,6 @@ TEST_P(AqSegmentTestLarge, TestNoMisMatchAQ2) { DoTest(2); }
 
 TEST_P(AqSegmentTestLarge, TestNoMisMatchAQ3) { DoTest(3); }
 
-#if !CONFIG_EXT_DELTA_Q
-// Validate that this AQ mode (AQ=4, delta q)
-// encodes and decodes without a mismatch.
-TEST_P(AqSegmentTest, TestNoMisMatchAQ4) {
-  cfg_.rc_end_usage = AOM_CQ;
-  aq_mode_ = 4;
-
-  ::libaom_test::I420VideoSource video("hantro_collage_w352h288.yuv", 352, 288,
-                                       30, 1, 0, 15);
-
-  ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
-}
-#endif
-
-#if CONFIG_EXT_DELTA_Q
 // Validate that this delta q mode
 // encodes and decodes without a mismatch.
 TEST_P(AqSegmentTest, TestNoMisMatchExtDeltaQ) {
@@ -116,7 +96,6 @@ TEST_P(AqSegmentTest, TestNoMisMatchExtDeltaQ) {
 
   ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
 }
-#endif
 
 AV1_INSTANTIATE_TEST_CASE(AqSegmentTest,
                           ::testing::Values(::libaom_test::kRealTime,

third_party/aom/test/arf_freq_test.cc

@@ -7,7 +7,7 @@
  * obtain it at www.aomedia.org/license/software. If the Alliance for Open
  * Media Patent License 1.0 was not distributed with this source code in the
  * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
+ */
 
 #include "third_party/googletest/src/googletest/include/gtest/gtest.h"
 
@@ -50,9 +50,7 @@ const TestVideoParam kTestVectors[] = {
   { "hantro_collage_w352h288.yuv", 352, 288, 30, 1, 8, AOM_IMG_FMT_I420,
     AOM_BITS_8, 0 },
   { "rush_hour_444.y4m", 352, 288, 30, 1, 8, AOM_IMG_FMT_I444, AOM_BITS_8, 1 },
-#if CONFIG_HIGHBITDEPTH
-// Add list of profile 2/3 test videos here ...
-#endif  // CONFIG_HIGHBITDEPTH
+  // Add list of profile 2/3 test videos here ...
 };
 
 const TestEncodeParam kEncodeVectors[] = {
@@ -208,7 +206,6 @@ TEST_P(ArfFreqTestLarge, MinArfFreqTest) {
   }
 }
 
-#if CONFIG_HIGHBITDEPTH || CONFIG_EXT_REFS
 #if CONFIG_AV1_ENCODER
 // TODO(angiebird): 25-29 fail in high bitdepth mode.
 // TODO(zoeliu): This ArfFreqTest does not work with BWDREF_FRAME, as
@@ -223,9 +220,4 @@ INSTANTIATE_TEST_CASE_P(
         ::testing::ValuesIn(kTestVectors), ::testing::ValuesIn(kEncodeVectors),
         ::testing::ValuesIn(kMinArfVectors)));
 #endif  // CONFIG_AV1_ENCODER
-#else
-AV1_INSTANTIATE_TEST_CASE(ArfFreqTestLarge, ::testing::ValuesIn(kTestVectors),
-                          ::testing::ValuesIn(kEncodeVectors),
-                          ::testing::ValuesIn(kMinArfVectors));
-#endif  // CONFIG_HIGHBITDEPTH || CONFIG_EXT_REFS
 }  // namespace

third_party/aom/test/av1_convolve_2d_test.cc

@@ -12,29 +12,238 @@
 #include "third_party/googletest/src/googletest/include/gtest/gtest.h"
 #include "test/av1_convolve_2d_test_util.h"
 
-using std::tr1::tuple;
-using std::tr1::make_tuple;
+using ::testing::make_tuple;
+using ::testing::tuple;
 using libaom_test::ACMRandom;
-using libaom_test::AV1Convolve2D::AV1Convolve2DTest;
-#if CONFIG_HIGHBITDEPTH
-using libaom_test::AV1HighbdConvolve2D::AV1HighbdConvolve2DTest;
-#endif
-
+using libaom_test::AV1Convolve2D::AV1Convolve2DSrTest;
+using libaom_test::AV1Convolve2D::AV1JntConvolve2DTest;
+using libaom_test::AV1HighbdConvolve2D::AV1HighbdConvolve2DSrTest;
+using libaom_test::AV1HighbdConvolve2D::AV1HighbdJntConvolve2DTest;
 namespace {
 
-TEST_P(AV1Convolve2DTest, CheckOutput) { RunCheckOutput(GET_PARAM(2)); }
+TEST_P(AV1Convolve2DSrTest, DISABLED_Speed) { RunSpeedTest(GET_PARAM(0)); }
+
+TEST_P(AV1Convolve2DSrTest, CheckOutput) { RunCheckOutput(GET_PARAM(0)); }
 
 INSTANTIATE_TEST_CASE_P(
-    SSE2, AV1Convolve2DTest,
-    libaom_test::AV1Convolve2D::BuildParams(av1_convolve_2d_sse2));
+    C_COPY, AV1Convolve2DSrTest,
+    libaom_test::AV1Convolve2D::BuildParams(av1_convolve_2d_copy_sr_c, 0, 0));
+INSTANTIATE_TEST_CASE_P(
+    C_X, AV1Convolve2DSrTest,
+    libaom_test::AV1Convolve2D::BuildParams(av1_convolve_x_sr_c, 1, 0));
+INSTANTIATE_TEST_CASE_P(
+    C_Y, AV1Convolve2DSrTest,
+    libaom_test::AV1Convolve2D::BuildParams(av1_convolve_y_sr_c, 0, 1));
+INSTANTIATE_TEST_CASE_P(
+    C, AV1Convolve2DSrTest,
+    libaom_test::AV1Convolve2D::BuildParams(av1_convolve_2d_sr_c, 1, 1));
+#if HAVE_SSE2
+INSTANTIATE_TEST_CASE_P(SSE2_COPY, AV1Convolve2DSrTest,
+                        libaom_test::AV1Convolve2D::BuildParams(
+                            av1_convolve_2d_copy_sr_sse2, 0, 0));
+INSTANTIATE_TEST_CASE_P(
+    SSE2_X, AV1Convolve2DSrTest,
+    libaom_test::AV1Convolve2D::BuildParams(av1_convolve_x_sr_sse2, 1, 0));
+INSTANTIATE_TEST_CASE_P(
+    SSE2_Y, AV1Convolve2DSrTest,
+    libaom_test::AV1Convolve2D::BuildParams(av1_convolve_y_sr_sse2, 0, 1));
+INSTANTIATE_TEST_CASE_P(
+    SSE2, AV1Convolve2DSrTest,
+    libaom_test::AV1Convolve2D::BuildParams(av1_convolve_2d_sr_sse2, 1, 1));
+#if HAVE_AVX2
+INSTANTIATE_TEST_CASE_P(AVX2_COPY, AV1Convolve2DSrTest,
+                        libaom_test::AV1Convolve2D::BuildParams(
+                            av1_convolve_2d_copy_sr_avx2, 0, 0));
+INSTANTIATE_TEST_CASE_P(
+    AVX2_X, AV1Convolve2DSrTest,
+    libaom_test::AV1Convolve2D::BuildParams(av1_convolve_x_sr_avx2, 1, 0));
 
-#if CONFIG_HIGHBITDEPTH && HAVE_SSSE3
-TEST_P(AV1HighbdConvolve2DTest, CheckOutput) { RunCheckOutput(GET_PARAM(3)); }
+INSTANTIATE_TEST_CASE_P(
+    AVX2_Y, AV1Convolve2DSrTest,
+    libaom_test::AV1Convolve2D::BuildParams(av1_convolve_y_sr_avx2, 0, 1));
 
-INSTANTIATE_TEST_CASE_P(SSSE3, AV1HighbdConvolve2DTest,
+INSTANTIATE_TEST_CASE_P(
+    AVX2, AV1Convolve2DSrTest,
+    libaom_test::AV1Convolve2D::BuildParams(av1_convolve_2d_sr_avx2, 1, 1));
+#endif  // HAVE_AVX2
+#endif  // HAVE_SSE2
+
+#if HAVE_NEON
+INSTANTIATE_TEST_CASE_P(
+    NEON_X, AV1Convolve2DSrTest,
+    libaom_test::AV1Convolve2D::BuildParams(av1_convolve_x_sr_neon, 1, 0));
+
+INSTANTIATE_TEST_CASE_P(
+    NEON_Y, AV1Convolve2DSrTest,
+    libaom_test::AV1Convolve2D::BuildParams(av1_convolve_y_sr_neon, 0, 1));
+
+INSTANTIATE_TEST_CASE_P(
+    NEON, AV1Convolve2DSrTest,
+    libaom_test::AV1Convolve2D::BuildParams(av1_convolve_2d_sr_neon, 1, 1));
+
+INSTANTIATE_TEST_CASE_P(NEON_COPY, AV1Convolve2DSrTest,
+                        libaom_test::AV1Convolve2D::BuildParams(
+                            av1_convolve_2d_copy_sr_neon, 0, 0));
+#endif  // HAVE_NEON
+
+TEST_P(AV1JntConvolve2DTest, CheckOutput) { RunCheckOutput(GET_PARAM(0)); }
+TEST_P(AV1JntConvolve2DTest, DISABLED_Speed) { RunSpeedTest(GET_PARAM(0)); }
+
+INSTANTIATE_TEST_CASE_P(
+    C_COPY, AV1JntConvolve2DTest,
+    libaom_test::AV1Convolve2D::BuildParams(av1_jnt_convolve_2d_copy_c, 0, 0));
+
+INSTANTIATE_TEST_CASE_P(
+    C_X, AV1JntConvolve2DTest,
+    libaom_test::AV1Convolve2D::BuildParams(av1_jnt_convolve_x_c, 1, 0));
+
+INSTANTIATE_TEST_CASE_P(
+    C_Y, AV1JntConvolve2DTest,
+    libaom_test::AV1Convolve2D::BuildParams(av1_jnt_convolve_y_c, 0, 1));
+
+#if HAVE_SSE2
+INSTANTIATE_TEST_CASE_P(SSE2_COPY, AV1JntConvolve2DTest,
+                        libaom_test::AV1Convolve2D::BuildParams(
+                            av1_jnt_convolve_2d_copy_sse2, 0, 0));
+#if HAVE_SSE4_1
+INSTANTIATE_TEST_CASE_P(
+    SSE2_X, AV1JntConvolve2DTest,
+    libaom_test::AV1Convolve2D::BuildParams(av1_jnt_convolve_x_sse2, 1, 0));
+
+INSTANTIATE_TEST_CASE_P(
+    SSE2_Y, AV1JntConvolve2DTest,
+    libaom_test::AV1Convolve2D::BuildParams(av1_jnt_convolve_y_sse2, 0, 1));
+
+INSTANTIATE_TEST_CASE_P(
+    SSSE3, AV1JntConvolve2DTest,
+    libaom_test::AV1Convolve2D::BuildParams(av1_jnt_convolve_2d_ssse3, 1, 1));
+
+#if HAVE_AVX2
+INSTANTIATE_TEST_CASE_P(AVX2_COPY, AV1JntConvolve2DTest,
+                        libaom_test::AV1Convolve2D::BuildParams(
+                            av1_jnt_convolve_2d_copy_avx2, 0, 0));
+INSTANTIATE_TEST_CASE_P(
+    AVX2_X, AV1JntConvolve2DTest,
+    libaom_test::AV1Convolve2D::BuildParams(av1_jnt_convolve_x_avx2, 1, 0));
+
+INSTANTIATE_TEST_CASE_P(
+    AVX2_Y, AV1JntConvolve2DTest,
+    libaom_test::AV1Convolve2D::BuildParams(av1_jnt_convolve_y_avx2, 0, 1));
+
+INSTANTIATE_TEST_CASE_P(
+    AVX2, AV1JntConvolve2DTest,
+    libaom_test::AV1Convolve2D::BuildParams(av1_jnt_convolve_2d_avx2, 1, 1));
+#endif  // HAVE_AVX2
+#endif  // HAVE_SSE4_1
+#endif  // HAVE_SSE2
+#if HAVE_NEON
+INSTANTIATE_TEST_CASE_P(NEON_COPY, AV1JntConvolve2DTest,
+                        libaom_test::AV1Convolve2D::BuildParams(
+                            av1_jnt_convolve_2d_copy_neon, 0, 0));
+
+INSTANTIATE_TEST_CASE_P(
+    NEON, AV1JntConvolve2DTest,
+    libaom_test::AV1Convolve2D::BuildParams(av1_jnt_convolve_2d_neon, 1, 1));
+INSTANTIATE_TEST_CASE_P(
+    NEON_X, AV1JntConvolve2DTest,
+    libaom_test::AV1Convolve2D::BuildParams(av1_jnt_convolve_x_neon, 1, 0));
+
+INSTANTIATE_TEST_CASE_P(
+    NEON_Y, AV1JntConvolve2DTest,
+    libaom_test::AV1Convolve2D::BuildParams(av1_jnt_convolve_y_neon, 0, 1));
+#endif  // HAVE_NEON
+
+TEST_P(AV1HighbdConvolve2DSrTest, CheckOutput) { RunCheckOutput(GET_PARAM(1)); }
+TEST_P(AV1HighbdConvolve2DSrTest, DISABLED_Speed) {
+  RunSpeedTest(GET_PARAM(1));
+}
+
+INSTANTIATE_TEST_CASE_P(C_X, AV1HighbdConvolve2DSrTest,
                         libaom_test::AV1HighbdConvolve2D::BuildParams(
-                            av1_highbd_convolve_2d_ssse3));
+                            av1_highbd_convolve_x_sr_c, 1, 0));
 
-#endif
+INSTANTIATE_TEST_CASE_P(C_Y, AV1HighbdConvolve2DSrTest,
+                        libaom_test::AV1HighbdConvolve2D::BuildParams(
+                            av1_highbd_convolve_y_sr_c, 0, 1));
 
+INSTANTIATE_TEST_CASE_P(C_COPY, AV1HighbdConvolve2DSrTest,
+                        libaom_test::AV1HighbdConvolve2D::BuildParams(
+                            av1_highbd_convolve_2d_copy_sr_c, 0, 0));
+#if HAVE_SSE2
+INSTANTIATE_TEST_CASE_P(SSE2_COPY, AV1HighbdConvolve2DSrTest,
+                        libaom_test::AV1HighbdConvolve2D::BuildParams(
+                            av1_highbd_convolve_2d_copy_sr_sse2, 0, 0));
+#if HAVE_SSSE3
+INSTANTIATE_TEST_CASE_P(SSSE3, AV1HighbdConvolve2DSrTest,
+                        libaom_test::AV1HighbdConvolve2D::BuildParams(
+                            av1_highbd_convolve_2d_sr_ssse3, 1, 1));
+INSTANTIATE_TEST_CASE_P(SSSE3_X, AV1HighbdConvolve2DSrTest,
+                        libaom_test::AV1HighbdConvolve2D::BuildParams(
+                            av1_highbd_convolve_x_sr_ssse3, 1, 0));
+INSTANTIATE_TEST_CASE_P(SSSE3_Y, AV1HighbdConvolve2DSrTest,
+                        libaom_test::AV1HighbdConvolve2D::BuildParams(
+                            av1_highbd_convolve_y_sr_ssse3, 0, 1));
+#if HAVE_AVX2
+INSTANTIATE_TEST_CASE_P(AVX2, AV1HighbdConvolve2DSrTest,
+                        libaom_test::AV1HighbdConvolve2D::BuildParams(
+                            av1_highbd_convolve_2d_sr_avx2, 1, 1));
+INSTANTIATE_TEST_CASE_P(AVX2_X, AV1HighbdConvolve2DSrTest,
+                        libaom_test::AV1HighbdConvolve2D::BuildParams(
+                            av1_highbd_convolve_x_sr_avx2, 1, 0));
+INSTANTIATE_TEST_CASE_P(AVX2_Y, AV1HighbdConvolve2DSrTest,
+                        libaom_test::AV1HighbdConvolve2D::BuildParams(
+                            av1_highbd_convolve_y_sr_avx2, 0, 1));
+INSTANTIATE_TEST_CASE_P(AVX2_COPY, AV1HighbdConvolve2DSrTest,
+                        libaom_test::AV1HighbdConvolve2D::BuildParams(
+                            av1_highbd_convolve_2d_copy_sr_avx2, 0, 0));
+#endif  // HAVE_AVX2
+#endif  // HAVE_SSSE3
+#endif  // HAVE_SSE2
+TEST_P(AV1HighbdJntConvolve2DTest, CheckOutput) {
+  RunCheckOutput(GET_PARAM(1));
+}
+
+TEST_P(AV1HighbdJntConvolve2DTest, DISABLED_Speed) {
+  RunSpeedTest(GET_PARAM(1));
+}
+
+INSTANTIATE_TEST_CASE_P(C_X, AV1HighbdJntConvolve2DTest,
+                        libaom_test::AV1HighbdConvolve2D::BuildParams(
+                            av1_highbd_jnt_convolve_x_c, 1, 0));
+
+INSTANTIATE_TEST_CASE_P(C_Y, AV1HighbdJntConvolve2DTest,
+                        libaom_test::AV1HighbdConvolve2D::BuildParams(
+                            av1_highbd_jnt_convolve_y_c, 0, 1));
+
+INSTANTIATE_TEST_CASE_P(C_COPY, AV1HighbdJntConvolve2DTest,
+                        libaom_test::AV1HighbdConvolve2D::BuildParams(
+                            av1_highbd_jnt_convolve_2d_copy_c, 0, 0));
+#if HAVE_SSE4_1
+INSTANTIATE_TEST_CASE_P(SSE4_1_COPY, AV1HighbdJntConvolve2DTest,
+                        libaom_test::AV1HighbdConvolve2D::BuildParams(
+                            av1_highbd_jnt_convolve_2d_copy_sse4_1, 0, 0));
+INSTANTIATE_TEST_CASE_P(SSE4_1, AV1HighbdJntConvolve2DTest,
+                        libaom_test::AV1HighbdConvolve2D::BuildParams(
+                            av1_highbd_jnt_convolve_2d_sse4_1, 1, 1));
+INSTANTIATE_TEST_CASE_P(SSE4_1_X, AV1HighbdJntConvolve2DTest,
+                        libaom_test::AV1HighbdConvolve2D::BuildParams(
+                            av1_highbd_jnt_convolve_x_sse4_1, 1, 0));
+INSTANTIATE_TEST_CASE_P(SSE4_1_Y, AV1HighbdJntConvolve2DTest,
+                        libaom_test::AV1HighbdConvolve2D::BuildParams(
+                            av1_highbd_jnt_convolve_y_sse4_1, 0, 1));
+#if HAVE_AVX2
+INSTANTIATE_TEST_CASE_P(AVX2_COPY, AV1HighbdJntConvolve2DTest,
+                        libaom_test::AV1HighbdConvolve2D::BuildParams(
+                            av1_highbd_jnt_convolve_2d_copy_avx2, 0, 0));
+INSTANTIATE_TEST_CASE_P(AVX2, AV1HighbdJntConvolve2DTest,
+                        libaom_test::AV1HighbdConvolve2D::BuildParams(
+                            av1_highbd_jnt_convolve_2d_avx2, 1, 1));
+INSTANTIATE_TEST_CASE_P(AVX2_X, AV1HighbdJntConvolve2DTest,
+                        libaom_test::AV1HighbdConvolve2D::BuildParams(
+                            av1_highbd_jnt_convolve_x_avx2, 1, 0));
+INSTANTIATE_TEST_CASE_P(AVX2_Y, AV1HighbdJntConvolve2DTest,
+                        libaom_test::AV1HighbdConvolve2D::BuildParams(
+                            av1_highbd_jnt_convolve_y_avx2, 0, 1));
+#endif  // HAVE_AVX2
+#endif  // HAVE_SSE4_1
 }  // namespace

third_party/aom/test/av1_convolve_2d_test_util.cc

@@ -11,183 +11,695 @@
 
 #include "test/av1_convolve_2d_test_util.h"
 
+#include "aom_ports/aom_timer.h"
+#include "av1/common/common_data.h"
 #include "av1/common/convolve.h"
 
-using std::tr1::tuple;
-using std::tr1::make_tuple;
+using ::testing::make_tuple;
+using ::testing::tuple;
 
 namespace libaom_test {
 
+const int kMaxSize = 128 + 32;  // padding
 namespace AV1Convolve2D {
 
 ::testing::internal::ParamGenerator<Convolve2DParam> BuildParams(
-    convolve_2d_func filter) {
-  const Convolve2DParam params[] = {
-    make_tuple(4, 4, filter),   make_tuple(8, 8, filter),
-    make_tuple(64, 64, filter), make_tuple(4, 16, filter),
-    make_tuple(32, 8, filter),
-  };
-  return ::testing::ValuesIn(params);
+    convolve_2d_func filter, int has_subx, int has_suby) {
+  return ::testing::Combine(::testing::Values(filter),
+                            ::testing::Values(has_subx),
+                            ::testing::Values(has_suby),
+                            ::testing::Range(BLOCK_4X4, BLOCK_SIZES_ALL));
 }
 
-AV1Convolve2DTest::~AV1Convolve2DTest() {}
-void AV1Convolve2DTest::SetUp() { rnd_.Reset(ACMRandom::DeterministicSeed()); }
-
-void AV1Convolve2DTest::TearDown() { libaom_test::ClearSystemState(); }
-
-void AV1Convolve2DTest::RunCheckOutput(convolve_2d_func test_impl) {
-  const int w = 128, h = 128;
-  const int out_w = GET_PARAM(0), out_h = GET_PARAM(1);
-  int i, j, k;
-
-  uint8_t *input = new uint8_t[h * w];
-
-  int output_n = out_h * MAX_SB_SIZE;
-  CONV_BUF_TYPE *output = new CONV_BUF_TYPE[output_n];
-  CONV_BUF_TYPE *output2 = new CONV_BUF_TYPE[output_n];
-
-  for (i = 0; i < h; ++i)
-    for (j = 0; j < w; ++j) input[i * w + j] = rnd_.Rand8();
-
-  int hfilter, vfilter, subx, suby;
-  for (hfilter = EIGHTTAP_REGULAR; hfilter < INTERP_FILTERS_ALL; ++hfilter) {
-    for (vfilter = EIGHTTAP_REGULAR; vfilter < INTERP_FILTERS_ALL; ++vfilter) {
-      InterpFilterParams filter_params_x =
-          av1_get_interp_filter_params((InterpFilter)hfilter);
-      InterpFilterParams filter_params_y =
-          av1_get_interp_filter_params((InterpFilter)vfilter);
-      const int do_average = rnd_.Rand8() & 1;
-      ConvolveParams conv_params1 =
-          get_conv_params_no_round(0, do_average, 0, output, MAX_SB_SIZE);
-      ConvolveParams conv_params2 =
-          get_conv_params_no_round(0, do_average, 0, output2, MAX_SB_SIZE);
-
-      for (subx = 0; subx < 16; ++subx)
-        for (suby = 0; suby < 16; ++suby) {
-          // av1_convolve_2d is designed for accumulate two predicted blocks for
-          // compound mode, so we set num_iter to two here.
-          // A larger number may introduce overflow
-          const int num_iters = 2;
-          memset(output, 0, output_n * sizeof(*output));
-          memset(output2, 0, output_n * sizeof(*output2));
-          for (i = 0; i < num_iters; ++i) {
-            // Choose random locations within the source block
-            int offset_r = 3 + rnd_.PseudoUniform(h - out_h - 7);
-            int offset_c = 3 + rnd_.PseudoUniform(w - out_w - 7);
-            av1_convolve_2d_c(input + offset_r * w + offset_c, w, output,
-                              MAX_SB_SIZE, out_w, out_h, &filter_params_x,
-                              &filter_params_y, subx, suby, &conv_params1);
-            test_impl(input + offset_r * w + offset_c, w, output2, MAX_SB_SIZE,
-                      out_w, out_h, &filter_params_x, &filter_params_y, subx,
-                      suby, &conv_params2);
-
-            for (j = 0; j < out_h; ++j)
-              for (k = 0; k < out_w; ++k) {
-                int idx = j * MAX_SB_SIZE + k;
-                ASSERT_EQ(output[idx], output2[idx])
-                    << "Pixel mismatch at index " << idx << " = (" << j << ", "
-                    << k << "), sub pixel offset = (" << suby << ", " << subx
-                    << ")";
-              }
-          }
-        }
-    }
-  }
-  delete[] input;
-  delete[] output;
-  delete[] output2;
-}
-}  // namespace AV1Convolve2D
-
-#if CONFIG_HIGHBITDEPTH
-namespace AV1HighbdConvolve2D {
-
-::testing::internal::ParamGenerator<HighbdConvolve2DParam> BuildParams(
-    highbd_convolve_2d_func filter) {
-  const HighbdConvolve2DParam params[] = {
-    make_tuple(4, 4, 8, filter),    make_tuple(8, 8, 8, filter),
-    make_tuple(64, 64, 8, filter),  make_tuple(4, 16, 8, filter),
-    make_tuple(32, 8, 8, filter),   make_tuple(4, 4, 10, filter),
-    make_tuple(8, 8, 10, filter),   make_tuple(64, 64, 10, filter),
-    make_tuple(4, 16, 10, filter),  make_tuple(32, 8, 10, filter),
-    make_tuple(4, 4, 12, filter),   make_tuple(8, 8, 12, filter),
-    make_tuple(64, 64, 12, filter), make_tuple(4, 16, 12, filter),
-    make_tuple(32, 8, 12, filter),
-  };
-  return ::testing::ValuesIn(params);
-}
-
-AV1HighbdConvolve2DTest::~AV1HighbdConvolve2DTest() {}
-void AV1HighbdConvolve2DTest::SetUp() {
+AV1Convolve2DSrTest::~AV1Convolve2DSrTest() {}
+void AV1Convolve2DSrTest::SetUp() {
   rnd_.Reset(ACMRandom::DeterministicSeed());
 }
 
-void AV1HighbdConvolve2DTest::TearDown() { libaom_test::ClearSystemState(); }
-
-void AV1HighbdConvolve2DTest::RunCheckOutput(
-    highbd_convolve_2d_func test_impl) {
-  const int w = 128, h = 128;
-  const int out_w = GET_PARAM(0), out_h = GET_PARAM(1);
-  const int bd = GET_PARAM(2);
-  int i, j, k;
-
-  uint16_t *input = new uint16_t[h * w];
-
-  int output_n = out_h * MAX_SB_SIZE;
-  CONV_BUF_TYPE *output = new CONV_BUF_TYPE[output_n];
-  CONV_BUF_TYPE *output2 = new CONV_BUF_TYPE[output_n];
-
-  for (i = 0; i < h; ++i)
-    for (j = 0; j < w; ++j) input[i * w + j] = rnd_.Rand16() & ((1 << bd) - 1);
+void AV1Convolve2DSrTest::TearDown() { libaom_test::ClearSystemState(); }
 
+void AV1Convolve2DSrTest::RunCheckOutput(convolve_2d_func test_impl) {
+  const int w = kMaxSize, h = kMaxSize;
+  const int has_subx = GET_PARAM(1);
+  const int has_suby = GET_PARAM(2);
+  const int block_idx = GET_PARAM(3);
   int hfilter, vfilter, subx, suby;
+  uint8_t input[kMaxSize * kMaxSize];
+  DECLARE_ALIGNED(32, uint8_t, output[MAX_SB_SQUARE]);
+  DECLARE_ALIGNED(32, uint8_t, output2[MAX_SB_SQUARE]);
+
+  for (int i = 0; i < h; ++i)
+    for (int j = 0; j < w; ++j) input[i * w + j] = rnd_.Rand8();
+  for (int i = 0; i < MAX_SB_SQUARE; ++i)
+    output[i] = output2[i] = rnd_.Rand31();
+
+  // Make sure that sizes 2xN and Nx2 are also tested for chroma.
+  const int num_sizes =
+      (block_size_wide[block_idx] == 4 || block_size_high[block_idx] == 4) ? 2
+                                                                           : 1;
+  for (int shift = 0; shift < num_sizes; ++shift) {  // luma and chroma
+    const int out_w = block_size_wide[block_idx] >> shift;
+    const int out_h = block_size_high[block_idx] >> shift;
+    for (hfilter = EIGHTTAP_REGULAR; hfilter < INTERP_FILTERS_ALL; ++hfilter) {
+      for (vfilter = EIGHTTAP_REGULAR; vfilter < INTERP_FILTERS_ALL;
+           ++vfilter) {
+        InterpFilterParams filter_params_x =
+            av1_get_interp_filter_params_with_block_size((InterpFilter)hfilter,
+                                                         out_w);
+        InterpFilterParams filter_params_y =
+            av1_get_interp_filter_params_with_block_size((InterpFilter)vfilter,
+                                                         out_h);
+        for (int do_average = 0; do_average < 1; ++do_average) {
+          ConvolveParams conv_params1 =
+              get_conv_params_no_round(0, do_average, 0, NULL, 0, 0, 8);
+          ConvolveParams conv_params2 =
+              get_conv_params_no_round(0, do_average, 0, NULL, 0, 0, 8);
+
+          const int subx_range = has_subx ? 16 : 1;
+          const int suby_range = has_suby ? 16 : 1;
+          for (subx = 0; subx < subx_range; ++subx) {
+            for (suby = 0; suby < suby_range; ++suby) {
+              // Choose random locations within the source block
+              const int offset_r = 3 + rnd_.PseudoUniform(h - out_h - 7);
+              const int offset_c = 3 + rnd_.PseudoUniform(w - out_w - 7);
+              av1_convolve_2d_sr_c(input + offset_r * w + offset_c, w, output,
+                                   MAX_SB_SIZE, out_w, out_h, &filter_params_x,
+                                   &filter_params_y, subx, suby, &conv_params1);
+              test_impl(input + offset_r * w + offset_c, w, output2,
+                        MAX_SB_SIZE, out_w, out_h, &filter_params_x,
+                        &filter_params_y, subx, suby, &conv_params2);
+
+              if (memcmp(output, output2, sizeof(output))) {
+                for (int i = 0; i < MAX_SB_SIZE; ++i) {
+                  for (int j = 0; j < MAX_SB_SIZE; ++j) {
+                    int idx = i * MAX_SB_SIZE + j;
+                    ASSERT_EQ(output[idx], output2[idx])
+                        << out_w << "x" << out_h << " Pixel mismatch at index "
+                        << idx << " = (" << i << ", " << j
+                        << "), sub pixel offset = (" << suby << ", " << subx
+                        << ")";
+                  }
+                }
+              }
+            }
+          }
+        }
+      }
+    }
+  }
+}
+
+void AV1Convolve2DSrTest::RunSpeedTest(convolve_2d_func test_impl) {
+  const int w = kMaxSize, h = kMaxSize;
+  const int has_subx = GET_PARAM(1);
+  const int has_suby = GET_PARAM(2);
+  const int block_idx = GET_PARAM(3);
+
+  uint8_t input[kMaxSize * kMaxSize];
+  DECLARE_ALIGNED(32, uint8_t, output[MAX_SB_SQUARE]);
+
+  for (int i = 0; i < h; ++i)
+    for (int j = 0; j < w; ++j) input[i * w + j] = rnd_.Rand8();
+
+  int hfilter = EIGHTTAP_REGULAR, vfilter = EIGHTTAP_REGULAR;
+  int subx = 0, suby = 0;
+
+  const int do_average = 0;
+  ConvolveParams conv_params2 =
+      get_conv_params_no_round(0, do_average, 0, NULL, 0, 0, 8);
+
+  // Make sure that sizes 2xN and Nx2 are also tested for chroma.
+  const int num_sizes =
+      (block_size_wide[block_idx] == 4 || block_size_high[block_idx] == 4) ? 2
+                                                                           : 1;
+  for (int shift = 0; shift < num_sizes; ++shift) {  // luma and chroma
+    const int out_w = block_size_wide[block_idx] >> shift;
+    const int out_h = block_size_high[block_idx] >> shift;
+    const int num_loops = 1000000000 / (out_w + out_h);
+
+    InterpFilterParams filter_params_x =
+        av1_get_interp_filter_params_with_block_size((InterpFilter)hfilter,
+                                                     out_w);
+    InterpFilterParams filter_params_y =
+        av1_get_interp_filter_params_with_block_size((InterpFilter)vfilter,
+                                                     out_h);
+
+    aom_usec_timer timer;
+    aom_usec_timer_start(&timer);
+
+    for (int i = 0; i < num_loops; ++i)
+      test_impl(input, w, output, MAX_SB_SIZE, out_w, out_h, &filter_params_x,
+                &filter_params_y, subx, suby, &conv_params2);
+
+    aom_usec_timer_mark(&timer);
+    const int elapsed_time = static_cast<int>(aom_usec_timer_elapsed(&timer));
+    printf("%d,%d convolve %3dx%-3d: %7.2f us\n", has_subx, has_suby, out_w,
+           out_h, 1000.0 * elapsed_time / num_loops);
+  }
+}
+
+AV1JntConvolve2DTest::~AV1JntConvolve2DTest() {}
+void AV1JntConvolve2DTest::SetUp() {
+  rnd_.Reset(ACMRandom::DeterministicSeed());
+}
+
+void AV1JntConvolve2DTest::TearDown() { libaom_test::ClearSystemState(); }
+
+void AV1JntConvolve2DTest::RunCheckOutput(convolve_2d_func test_impl) {
+  const int w = kMaxSize, h = kMaxSize;
+  const int has_subx = GET_PARAM(1);
+  const int has_suby = GET_PARAM(2);
+  const int block_idx = GET_PARAM(3);
+  int hfilter, vfilter, subx, suby;
+  uint8_t input[kMaxSize * kMaxSize];
+  DECLARE_ALIGNED(32, CONV_BUF_TYPE, output1[MAX_SB_SQUARE]);
+  DECLARE_ALIGNED(32, CONV_BUF_TYPE, output2[MAX_SB_SQUARE]);
+  DECLARE_ALIGNED(16, uint8_t, output8_1[MAX_SB_SQUARE]);
+  DECLARE_ALIGNED(16, uint8_t, output8_2[MAX_SB_SQUARE]);
+
+  for (int i = 0; i < h; ++i)
+    for (int j = 0; j < w; ++j) input[i * w + j] = rnd_.Rand8();
+  for (int i = 0; i < MAX_SB_SQUARE; ++i) {
+    output1[i] = output2[i] = rnd_.Rand16();
+    output8_1[i] = output8_2[i] = rnd_.Rand8();
+  }
+
+  const int out_w = block_size_wide[block_idx];
+  const int out_h = block_size_high[block_idx];
   for (hfilter = EIGHTTAP_REGULAR; hfilter < INTERP_FILTERS_ALL; ++hfilter) {
     for (vfilter = EIGHTTAP_REGULAR; vfilter < INTERP_FILTERS_ALL; ++vfilter) {
       InterpFilterParams filter_params_x =
-          av1_get_interp_filter_params((InterpFilter)hfilter);
+          av1_get_interp_filter_params_with_block_size((InterpFilter)hfilter,
+                                                       out_w);
       InterpFilterParams filter_params_y =
-          av1_get_interp_filter_params((InterpFilter)vfilter);
-      ConvolveParams conv_params1 =
-          get_conv_params_no_round(0, 0, 0, output, MAX_SB_SIZE);
-      ConvolveParams conv_params2 =
-          get_conv_params_no_round(0, 0, 0, output2, MAX_SB_SIZE);
+          av1_get_interp_filter_params_with_block_size((InterpFilter)vfilter,
+                                                       out_h);
+      for (int do_average = 0; do_average <= 1; ++do_average) {
+        ConvolveParams conv_params1 = get_conv_params_no_round(
+            0, do_average, 0, output1, MAX_SB_SIZE, 1, 8);
+        ConvolveParams conv_params2 = get_conv_params_no_round(
+            0, do_average, 0, output2, MAX_SB_SIZE, 1, 8);
 
-      for (subx = 0; subx < 16; ++subx)
-        for (suby = 0; suby < 16; ++suby) {
-          // av1_convolve_2d is designed for accumulate two predicted blocks for
-          // compound mode, so we set num_iter to two here.
-          // A larger number may introduce overflow
-          const int num_iters = 2;
-          memset(output, 0, output_n * sizeof(*output));
-          memset(output2, 0, output_n * sizeof(*output2));
-          for (i = 0; i < num_iters; ++i) {
+        // Test special case where jnt_comp_avg is not used
+        conv_params1.use_jnt_comp_avg = 0;
+        conv_params2.use_jnt_comp_avg = 0;
+
+        const int subx_range = has_subx ? 16 : 1;
+        const int suby_range = has_suby ? 16 : 1;
+        for (subx = 0; subx < subx_range; ++subx) {
+          for (suby = 0; suby < suby_range; ++suby) {
             // Choose random locations within the source block
-            int offset_r = 3 + rnd_.PseudoUniform(h - out_h - 7);
-            int offset_c = 3 + rnd_.PseudoUniform(w - out_w - 7);
-            av1_highbd_convolve_2d_c(input + offset_r * w + offset_c, w, output,
-                                     MAX_SB_SIZE, out_w, out_h,
-                                     &filter_params_x, &filter_params_y, subx,
-                                     suby, &conv_params1, bd);
-            test_impl(input + offset_r * w + offset_c, w, output2, MAX_SB_SIZE,
-                      out_w, out_h, &filter_params_x, &filter_params_y, subx,
-                      suby, &conv_params2, bd);
+            const int offset_r = 3 + rnd_.PseudoUniform(h - out_h - 7);
+            const int offset_c = 3 + rnd_.PseudoUniform(w - out_w - 7);
+            av1_jnt_convolve_2d_c(input + offset_r * w + offset_c, w, output8_1,
+                                  MAX_SB_SIZE, out_w, out_h, &filter_params_x,
+                                  &filter_params_y, subx, suby, &conv_params1);
+            test_impl(input + offset_r * w + offset_c, w, output8_2,
+                      MAX_SB_SIZE, out_w, out_h, &filter_params_x,
+                      &filter_params_y, subx, suby, &conv_params2);
 
-            for (j = 0; j < out_h; ++j)
-              for (k = 0; k < out_w; ++k) {
-                int idx = j * MAX_SB_SIZE + k;
-                ASSERT_EQ(output[idx], output2[idx])
-                    << "Pixel mismatch at index " << idx << " = (" << j << ", "
-                    << k << "), sub pixel offset = (" << suby << ", " << subx
-                    << ")";
+            for (int i = 0; i < out_h; ++i) {
+              for (int j = 0; j < out_w; ++j) {
+                int idx = i * MAX_SB_SIZE + j;
+                ASSERT_EQ(output1[idx], output2[idx])
+                    << "Mismatch at unit tests for av1_jnt_convolve_2d\n"
+                    << out_w << "x" << out_h << " Pixel mismatch at index "
+                    << idx << " = (" << i << ", " << j
+                    << "), sub pixel offset = (" << suby << ", " << subx << ")";
               }
+            }
+
+            if (memcmp(output8_1, output8_2, sizeof(output8_1))) {
+              for (int i = 0; i < MAX_SB_SIZE; ++i) {
+                for (int j = 0; j < MAX_SB_SIZE; ++j) {
+                  int idx = i * MAX_SB_SIZE + j;
+                  ASSERT_EQ(output8_1[idx], output8_2[idx])
+                      << out_w << "x" << out_h << " Pixel mismatch at index "
+                      << idx << " = (" << i << ", " << j
+                      << "), sub pixel offset = (" << suby << ", " << subx
+                      << ")";
+                }
+              }
+            }
           }
         }
+
+        // Test different combination of fwd and bck offset weights
+        for (int k = 0; k < 2; ++k) {
+          for (int l = 0; l < 4; ++l) {
+            conv_params1.use_jnt_comp_avg = 1;
+            conv_params2.use_jnt_comp_avg = 1;
+            conv_params1.fwd_offset = quant_dist_lookup_table[k][l][0];
+            conv_params1.bck_offset = quant_dist_lookup_table[k][l][1];
+            conv_params2.fwd_offset = quant_dist_lookup_table[k][l][0];
+            conv_params2.bck_offset = quant_dist_lookup_table[k][l][1];
+
+            for (subx = 0; subx < subx_range; ++subx) {
+              for (suby = 0; suby < suby_range; ++suby) {
+                // Choose random locations within the source block
+                const int offset_r = 3 + rnd_.PseudoUniform(h - out_h - 7);
+                const int offset_c = 3 + rnd_.PseudoUniform(w - out_w - 7);
+                av1_jnt_convolve_2d_c(input + offset_r * w + offset_c, w,
+                                      output8_1, MAX_SB_SIZE, out_w, out_h,
+                                      &filter_params_x, &filter_params_y, subx,
+                                      suby, &conv_params1);
+                test_impl(input + offset_r * w + offset_c, w, output8_2,
+                          MAX_SB_SIZE, out_w, out_h, &filter_params_x,
+                          &filter_params_y, subx, suby, &conv_params2);
+
+                for (int i = 0; i < out_h; ++i) {
+                  for (int j = 0; j < out_w; ++j) {
+                    int idx = i * MAX_SB_SIZE + j;
+                    ASSERT_EQ(output1[idx], output2[idx])
+                        << "Mismatch at unit tests for "
+                           "av1_jnt_convolve_2d\n"
+                        << out_w << "x" << out_h << " Pixel mismatch at index "
+                        << idx << " = (" << i << ", " << j
+                        << "), sub pixel offset = (" << suby << ", " << subx
+                        << ")";
+                  }
+                }
+                if (memcmp(output8_1, output8_2, sizeof(output8_1))) {
+                  for (int i = 0; i < MAX_SB_SIZE; ++i) {
+                    for (int j = 0; j < MAX_SB_SIZE; ++j) {
+                      int idx = i * MAX_SB_SIZE + j;
+                      ASSERT_EQ(output8_1[idx], output8_2[idx])
+                          << out_w << "x" << out_h
+                          << " Pixel mismatch at index " << idx << " = (" << i
+                          << ", " << j << "), sub pixel offset = (" << suby
+                          << ", " << subx << ")";
+                    }
+                  }
+                }
+              }
+            }
+          }
+        }
+      }
+    }
+  }
+}
+
+void AV1JntConvolve2DTest::RunSpeedTest(convolve_2d_func test_impl) {
+  const int w = kMaxSize, h = kMaxSize;
+  const int has_subx = GET_PARAM(1);
+  const int has_suby = GET_PARAM(2);
+  const int block_idx = GET_PARAM(3);
+
+  int subx = 0, suby = 0;
+  uint8_t input[kMaxSize * kMaxSize];
+  DECLARE_ALIGNED(32, CONV_BUF_TYPE, output[MAX_SB_SQUARE]);
+  DECLARE_ALIGNED(16, uint8_t, output8[MAX_SB_SQUARE]);
+  int hfilter = EIGHTTAP_REGULAR, vfilter = EIGHTTAP_REGULAR;
+  for (int i = 0; i < h; ++i)
+    for (int j = 0; j < w; ++j) input[i * w + j] = rnd_.Rand8();
+  for (int i = 0; i < MAX_SB_SQUARE; ++i) {
+    output[i] = rnd_.Rand16();
+    output8[i] = rnd_.Rand8();
+  }
+
+  const int out_w = block_size_wide[block_idx];
+  const int out_h = block_size_high[block_idx];
+  const int num_loops = 1000000000 / (out_w + out_h);
+  const int do_average = 0;
+
+  InterpFilterParams filter_params_x =
+      av1_get_interp_filter_params_with_block_size((InterpFilter)hfilter,
+                                                   out_w);
+  InterpFilterParams filter_params_y =
+      av1_get_interp_filter_params_with_block_size((InterpFilter)vfilter,
+                                                   out_h);
+
+  ConvolveParams conv_params =
+      get_conv_params_no_round(0, do_average, 0, output, MAX_SB_SIZE, 1, 8);
+
+  conv_params.use_jnt_comp_avg = 0;
+
+  // Choose random locations within the source block
+  const int offset_r = 3 + rnd_.PseudoUniform(h - out_h - 7);
+  const int offset_c = 3 + rnd_.PseudoUniform(w - out_w - 7);
+
+  aom_usec_timer timer;
+  aom_usec_timer_start(&timer);
+
+  for (int i = 0; i < num_loops; ++i)
+    test_impl(input + offset_r * w + offset_c, w, output8, MAX_SB_SIZE, out_w,
+              out_h, &filter_params_x, &filter_params_y, subx, suby,
+              &conv_params);
+
+  aom_usec_timer_mark(&timer);
+  const int elapsed_time = static_cast<int>(aom_usec_timer_elapsed(&timer));
+  printf("%d,%d convolve %3dx%-3d: %7.2f us\n", has_subx, has_suby, out_w,
+         out_h, 1000.0 * elapsed_time / num_loops);
+}
+}  // namespace AV1Convolve2D
+
+namespace AV1HighbdConvolve2D {
+::testing::internal::ParamGenerator<HighbdConvolve2DParam> BuildParams(
+    highbd_convolve_2d_func filter, int has_subx, int has_suby) {
+  return ::testing::Combine(
+      ::testing::Range(8, 13, 2), ::testing::Values(filter),
+      ::testing::Values(has_subx), ::testing::Values(has_suby),
+      ::testing::Range(BLOCK_4X4, BLOCK_SIZES_ALL));
+}
+
+AV1HighbdConvolve2DSrTest::~AV1HighbdConvolve2DSrTest() {}
+void AV1HighbdConvolve2DSrTest::SetUp() {
+  rnd_.Reset(ACMRandom::DeterministicSeed());
+}
+
+void AV1HighbdConvolve2DSrTest::TearDown() { libaom_test::ClearSystemState(); }
+
+void AV1HighbdConvolve2DSrTest::RunSpeedTest(
+    highbd_convolve_2d_func test_impl) {
+  const int w = kMaxSize, h = kMaxSize;
+  const int bd = GET_PARAM(0);
+  const int has_subx = GET_PARAM(2);
+  const int has_suby = GET_PARAM(3);
+  const int block_idx = GET_PARAM(4);
+  int hfilter, vfilter, subx, suby;
+  uint16_t input[kMaxSize * kMaxSize];
+  DECLARE_ALIGNED(32, uint16_t, output[MAX_SB_SQUARE]);
+
+  for (int i = 0; i < h; ++i)
+    for (int j = 0; j < w; ++j)
+      input[i * w + j] = rnd_.Rand16() & ((1 << bd) - 1);
+
+  hfilter = EIGHTTAP_REGULAR;
+  vfilter = EIGHTTAP_REGULAR;
+  int do_average = 0;
+
+  const int offset_r = 3;
+  const int offset_c = 3;
+  subx = 0;
+  suby = 0;
+
+  ConvolveParams conv_params =
+      get_conv_params_no_round(0, do_average, 0, NULL, 0, 0, bd);
+
+  // Make sure that sizes 2xN and Nx2 are also tested for chroma.
+  const int num_sizes =
+      (block_size_wide[block_idx] == 4 || block_size_high[block_idx] == 4) ? 2
+                                                                           : 1;
+
+  for (int shift = 0; shift < num_sizes; ++shift) {  // luma and chroma
+    const int out_w = block_size_wide[block_idx] >> shift;
+    const int out_h = block_size_high[block_idx] >> shift;
+    const int num_loops = 1000000000 / (out_w + out_h);
+
+    InterpFilterParams filter_params_x =
+        av1_get_interp_filter_params_with_block_size((InterpFilter)hfilter,
+                                                     out_w);
+    InterpFilterParams filter_params_y =
+        av1_get_interp_filter_params_with_block_size((InterpFilter)vfilter,
+                                                     out_h);
+
+    aom_usec_timer timer;
+    aom_usec_timer_start(&timer);
+    for (int i = 0; i < num_loops; ++i)
+      test_impl(input + offset_r * w + offset_c, w, output, MAX_SB_SIZE, out_w,
+                out_h, &filter_params_x, &filter_params_y, subx, suby,
+                &conv_params, bd);
+
+    aom_usec_timer_mark(&timer);
+    const int elapsed_time = static_cast<int>(aom_usec_timer_elapsed(&timer));
+    printf("%d,%d convolve %3dx%-3d: %7.2f us\n", has_subx, has_suby, out_w,
+           out_h, 1000.0 * elapsed_time / num_loops);
+  }
+}
+
+void AV1HighbdConvolve2DSrTest::RunCheckOutput(
+    highbd_convolve_2d_func test_impl) {
+  const int w = kMaxSize, h = kMaxSize;
+  const int bd = GET_PARAM(0);
+  const int has_subx = GET_PARAM(2);
+  const int has_suby = GET_PARAM(3);
+  const int block_idx = GET_PARAM(4);
+  int hfilter, vfilter, subx, suby;
+  uint16_t input[kMaxSize * kMaxSize];
+  DECLARE_ALIGNED(32, uint16_t, output[MAX_SB_SQUARE]);
+  DECLARE_ALIGNED(32, uint16_t, output2[MAX_SB_SQUARE]);
+
+  for (int i = 0; i < h; ++i)
+    for (int j = 0; j < w; ++j)
+      input[i * w + j] = rnd_.Rand16() & ((1 << bd) - 1);
+  for (int i = 0; i < MAX_SB_SQUARE; ++i)
+    output[i] = output2[i] = rnd_.Rand31();
+
+  // Make sure that sizes 2xN and Nx2 are also tested for chroma.
+  const int num_sizes =
+      (block_size_wide[block_idx] == 4 || block_size_high[block_idx] == 4) ? 2
+                                                                           : 1;
+  for (int shift = 0; shift < num_sizes; ++shift) {  // luma and chroma
+    const int out_w = block_size_wide[block_idx] >> shift;
+    const int out_h = block_size_high[block_idx] >> shift;
+    for (hfilter = EIGHTTAP_REGULAR; hfilter < INTERP_FILTERS_ALL; ++hfilter) {
+      for (vfilter = EIGHTTAP_REGULAR; vfilter < INTERP_FILTERS_ALL;
+           ++vfilter) {
+        InterpFilterParams filter_params_x =
+            av1_get_interp_filter_params_with_block_size((InterpFilter)hfilter,
+                                                         out_w);
+        InterpFilterParams filter_params_y =
+            av1_get_interp_filter_params_with_block_size((InterpFilter)vfilter,
+                                                         out_h);
+        for (int do_average = 0; do_average < 1; ++do_average) {
+          ConvolveParams conv_params1 =
+              get_conv_params_no_round(0, do_average, 0, NULL, 0, 0, bd);
+          ConvolveParams conv_params2 =
+              get_conv_params_no_round(0, do_average, 0, NULL, 0, 0, bd);
+
+          const int subx_range = has_subx ? 16 : 1;
+          const int suby_range = has_suby ? 16 : 1;
+          for (subx = 0; subx < subx_range; ++subx) {
+            for (suby = 0; suby < suby_range; ++suby) {
+              // Choose random locations within the source block
+              const int offset_r = 3 + rnd_.PseudoUniform(h - out_h - 7);
+              const int offset_c = 3 + rnd_.PseudoUniform(w - out_w - 7);
+              av1_highbd_convolve_2d_sr_c(input + offset_r * w + offset_c, w,
+                                          output, MAX_SB_SIZE, out_w, out_h,
+                                          &filter_params_x, &filter_params_y,
+                                          subx, suby, &conv_params1, bd);
+              test_impl(input + offset_r * w + offset_c, w, output2,
+                        MAX_SB_SIZE, out_w, out_h, &filter_params_x,
+                        &filter_params_y, subx, suby, &conv_params2, bd);
+
+              if (memcmp(output, output2, sizeof(output))) {
+                for (int i = 0; i < MAX_SB_SIZE; ++i) {
+                  for (int j = 0; j < MAX_SB_SIZE; ++j) {
+                    int idx = i * MAX_SB_SIZE + j;
+                    ASSERT_EQ(output[idx], output2[idx])
+                        << out_w << "x" << out_h << " Pixel mismatch at index "
+                        << idx << " = (" << i << ", " << j
+                        << "), sub pixel offset = (" << suby << ", " << subx
+                        << ")";
+                  }
+                }
+              }
+            }
+          }
+        }
+      }
+    }
+  }
+}
+
+AV1HighbdJntConvolve2DTest::~AV1HighbdJntConvolve2DTest() {}
+void AV1HighbdJntConvolve2DTest::SetUp() {
+  rnd_.Reset(ACMRandom::DeterministicSeed());
+}
+
+void AV1HighbdJntConvolve2DTest::TearDown() { libaom_test::ClearSystemState(); }
+
+void AV1HighbdJntConvolve2DTest::RunSpeedTest(
+    highbd_convolve_2d_func test_impl) {
+  const int w = kMaxSize, h = kMaxSize;
+  const int bd = GET_PARAM(0);
+  const int block_idx = GET_PARAM(4);
+  int hfilter, vfilter, subx, suby;
+  uint16_t input[kMaxSize * kMaxSize];
+  DECLARE_ALIGNED(32, CONV_BUF_TYPE, output[MAX_SB_SQUARE]);
+  DECLARE_ALIGNED(32, uint16_t, output16[MAX_SB_SQUARE]);
+
+  for (int i = 0; i < h; ++i)
+    for (int j = 0; j < w; ++j)
+      input[i * w + j] = rnd_.Rand16() & ((1 << bd) - 1);
+  for (int i = 0; i < MAX_SB_SQUARE; ++i) output[i] = rnd_.Rand16();
+  hfilter = EIGHTTAP_REGULAR;
+  vfilter = EIGHTTAP_REGULAR;
+  int do_average = 0;
+  const int out_w = block_size_wide[block_idx];
+  const int out_h = block_size_high[block_idx];
+
+  InterpFilterParams filter_params_x =
+      av1_get_interp_filter_params_with_block_size((InterpFilter)hfilter,
+                                                   out_w);
+  InterpFilterParams filter_params_y =
+      av1_get_interp_filter_params_with_block_size((InterpFilter)vfilter,
+                                                   out_h);
+
+  ConvolveParams conv_params =
+      get_conv_params_no_round(0, do_average, 0, output, MAX_SB_SIZE, 1, bd);
+
+  // Test special case where jnt_comp_avg is not used
+  conv_params.use_jnt_comp_avg = 0;
+
+  subx = 0;
+  suby = 0;
+  // Choose random locations within the source block
+  const int offset_r = 3;
+  const int offset_c = 3;
+
+  const int num_loops = 1000000000 / (out_w + out_h);
+  aom_usec_timer timer;
+  aom_usec_timer_start(&timer);
+  for (int i = 0; i < num_loops; ++i)
+    test_impl(input + offset_r * w + offset_c, w, output16, MAX_SB_SIZE, out_w,
+              out_h, &filter_params_x, &filter_params_y, subx, suby,
+              &conv_params, bd);
+
+  aom_usec_timer_mark(&timer);
+  const int elapsed_time = static_cast<int>(aom_usec_timer_elapsed(&timer));
+  printf("convolve %3dx%-3d: %7.2f us\n", out_w, out_h,
+         1000.0 * elapsed_time / num_loops);
+}
+
+void AV1HighbdJntConvolve2DTest::RunCheckOutput(
+    highbd_convolve_2d_func test_impl) {
+  const int w = kMaxSize, h = kMaxSize;
+  const int bd = GET_PARAM(0);
+  const int has_subx = GET_PARAM(2);
+  const int has_suby = GET_PARAM(3);
+  const int block_idx = GET_PARAM(4);
+  int hfilter, vfilter, subx, suby;
+  uint16_t input[kMaxSize * kMaxSize];
+  DECLARE_ALIGNED(32, CONV_BUF_TYPE, output1[MAX_SB_SQUARE]);
+  DECLARE_ALIGNED(32, CONV_BUF_TYPE, output2[MAX_SB_SQUARE]);
+  DECLARE_ALIGNED(32, uint16_t, output16_1[MAX_SB_SQUARE]);
+  DECLARE_ALIGNED(32, uint16_t, output16_2[MAX_SB_SQUARE]);
+
+  for (int i = 0; i < h; ++i)
+    for (int j = 0; j < w; ++j)
+      input[i * w + j] = rnd_.Rand16() & ((1 << bd) - 1);
+  for (int i = 0; i < MAX_SB_SQUARE; ++i) {
+    output1[i] = output2[i] = rnd_.Rand16();
+    output16_1[i] = output16_2[i] = rnd_.Rand16();
+  }
+
+  const int out_w = block_size_wide[block_idx];
+  const int out_h = block_size_high[block_idx];
+  for (hfilter = EIGHTTAP_REGULAR; hfilter < INTERP_FILTERS_ALL; ++hfilter) {
+    for (vfilter = EIGHTTAP_REGULAR; vfilter < INTERP_FILTERS_ALL; ++vfilter) {
+      InterpFilterParams filter_params_x =
+          av1_get_interp_filter_params_with_block_size((InterpFilter)hfilter,
+                                                       out_w);
+      InterpFilterParams filter_params_y =
+          av1_get_interp_filter_params_with_block_size((InterpFilter)vfilter,
+                                                       out_h);
+      for (int do_average = 0; do_average <= 1; ++do_average) {
+        ConvolveParams conv_params1 = get_conv_params_no_round(
+            0, do_average, 0, output1, MAX_SB_SIZE, 1, bd);
+        ConvolveParams conv_params2 = get_conv_params_no_round(
+            0, do_average, 0, output2, MAX_SB_SIZE, 1, bd);
+
+        // Test special case where jnt_comp_avg is not used
+        conv_params1.use_jnt_comp_avg = 0;
+        conv_params2.use_jnt_comp_avg = 0;
+
+        const int subx_range = has_subx ? 16 : 1;
+        const int suby_range = has_suby ? 16 : 1;
+        for (subx = 0; subx < subx_range; ++subx) {
+          for (suby = 0; suby < suby_range; ++suby) {
+            // Choose random locations within the source block
+            const int offset_r = 3 + rnd_.PseudoUniform(h - out_h - 7);
+            const int offset_c = 3 + rnd_.PseudoUniform(w - out_w - 7);
+            av1_highbd_jnt_convolve_2d_c(input + offset_r * w + offset_c, w,
+                                         output16_1, MAX_SB_SIZE, out_w, out_h,
+                                         &filter_params_x, &filter_params_y,
+                                         subx, suby, &conv_params1, bd);
+            test_impl(input + offset_r * w + offset_c, w, output16_2,
+                      MAX_SB_SIZE, out_w, out_h, &filter_params_x,
+                      &filter_params_y, subx, suby, &conv_params2, bd);
+
+            for (int i = 0; i < out_h; ++i) {
+              for (int j = 0; j < out_w; ++j) {
+                int idx = i * MAX_SB_SIZE + j;
+                ASSERT_EQ(output1[idx], output2[idx])
+                    << out_w << "x" << out_h << " Pixel mismatch at index "
+                    << idx << " = (" << i << ", " << j
+                    << "), sub pixel offset = (" << suby << ", " << subx << ")";
+              }
+            }
+
+            if (memcmp(output16_1, output16_2, sizeof(output16_1))) {
+              for (int i = 0; i < MAX_SB_SIZE; ++i) {
+                for (int j = 0; j < MAX_SB_SIZE; ++j) {
+                  int idx = i * MAX_SB_SIZE + j;
+                  ASSERT_EQ(output16_1[idx], output16_2[idx])
+                      << out_w << "x" << out_h << " Pixel mismatch at index "
+                      << idx << " = (" << i << ", " << j
+                      << "), sub pixel offset = (" << suby << ", " << subx
+                      << ")";
+                }
+              }
+            }
+          }
+        }
+
+        // Test different combination of fwd and bck offset weights
+        for (int k = 0; k < 2; ++k) {
+          for (int l = 0; l < 4; ++l) {
+            conv_params1.use_jnt_comp_avg = 1;
+            conv_params2.use_jnt_comp_avg = 1;
+            conv_params1.fwd_offset = quant_dist_lookup_table[k][l][0];
+            conv_params1.bck_offset = quant_dist_lookup_table[k][l][1];
+            conv_params2.fwd_offset = quant_dist_lookup_table[k][l][0];
+            conv_params2.bck_offset = quant_dist_lookup_table[k][l][1];
+
+            const int subx_range = has_subx ? 16 : 1;
+            const int suby_range = has_suby ? 16 : 1;
+            for (subx = 0; subx < subx_range; ++subx) {
+              for (suby = 0; suby < suby_range; ++suby) {
+                // Choose random locations within the source block
+                const int offset_r = 3 + rnd_.PseudoUniform(h - out_h - 7);
+                const int offset_c = 3 + rnd_.PseudoUniform(w - out_w - 7);
+                av1_highbd_jnt_convolve_2d_c(
+                    input + offset_r * w + offset_c, w, output16_1, MAX_SB_SIZE,
+                    out_w, out_h, &filter_params_x, &filter_params_y, subx,
+                    suby, &conv_params1, bd);
+                test_impl(input + offset_r * w + offset_c, w, output16_2,
+                          MAX_SB_SIZE, out_w, out_h, &filter_params_x,
+                          &filter_params_y, subx, suby, &conv_params2, bd);
+
+                for (int i = 0; i < out_h; ++i) {
+                  for (int j = 0; j < out_w; ++j) {
+                    int idx = i * MAX_SB_SIZE + j;
+                    ASSERT_EQ(output1[idx], output2[idx])
+                        << out_w << "x" << out_h << " Pixel mismatch at index "
+                        << idx << " = (" << i << ", " << j
+                        << "), sub pixel offset = (" << suby << ", " << subx
+                        << ")";
+                  }
+                }
+
+                if (memcmp(output16_1, output16_2, sizeof(output16_1))) {
+                  for (int i = 0; i < MAX_SB_SIZE; ++i) {
+                    for (int j = 0; j < MAX_SB_SIZE; ++j) {
+                      int idx = i * MAX_SB_SIZE + j;
+                      ASSERT_EQ(output16_1[idx], output16_2[idx])
+                          << out_w << "x" << out_h
+                          << " Pixel mismatch at index " << idx << " = (" << i
+                          << ", " << j << "), sub pixel offset = (" << suby
+                          << ", " << subx << ")";
+                    }
+                  }
+                }
+              }
+            }
+          }
+        }
+      }
     }
   }
-  delete[] input;
-  delete[] output;
-  delete[] output2;
 }
 }  // namespace AV1HighbdConvolve2D
-#endif  // CONFIG_HIGHBITDEPTH
 }  // namespace libaom_test

third_party/aom/test/av1_convolve_2d_test_util.h

@@ -12,11 +12,13 @@
 #ifndef TEST_HIPREC_CONVOLVE_TEST_UTIL_H_
 #define TEST_HIPREC_CONVOLVE_TEST_UTIL_H_
 
+#include "config/av1_rtcd.h"
+#include "config/aom_dsp_rtcd.h"
+
 #include "third_party/googletest/src/googletest/include/gtest/gtest.h"
 #include "test/acm_random.h"
 #include "test/util.h"
-#include "./av1_rtcd.h"
-#include "./aom_dsp_rtcd.h"
+
 #include "test/clear_system_state.h"
 #include "test/register_state_check.h"
 
@@ -25,62 +27,90 @@ namespace libaom_test {
 namespace AV1Convolve2D {
 
 typedef void (*convolve_2d_func)(const uint8_t *src, int src_stride,
-                                 CONV_BUF_TYPE *dst, int dst_stride, int w,
-                                 int h, InterpFilterParams *filter_params_x,
+                                 uint8_t *dst, int dst_stride, int w, int h,
+                                 InterpFilterParams *filter_params_x,
                                  InterpFilterParams *filter_params_y,
                                  const int subpel_x_q4, const int subpel_y_q4,
                                  ConvolveParams *conv_params);
 
-typedef std::tr1::tuple<int, int, convolve_2d_func> Convolve2DParam;
+typedef ::testing::tuple<convolve_2d_func, int, int, BLOCK_SIZE>
+    Convolve2DParam;
 
 ::testing::internal::ParamGenerator<Convolve2DParam> BuildParams(
-    convolve_2d_func filter);
+    convolve_2d_func filter, int subx_exist, int suby_exist);
 
-class AV1Convolve2DTest : public ::testing::TestWithParam<Convolve2DParam> {
+class AV1Convolve2DSrTest : public ::testing::TestWithParam<Convolve2DParam> {
  public:
-  virtual ~AV1Convolve2DTest();
+  virtual ~AV1Convolve2DSrTest();
   virtual void SetUp();
 
   virtual void TearDown();
 
  protected:
   void RunCheckOutput(convolve_2d_func test_impl);
+  void RunSpeedTest(convolve_2d_func test_impl);
 
   libaom_test::ACMRandom rnd_;
 };
 
+class AV1JntConvolve2DTest : public ::testing::TestWithParam<Convolve2DParam> {
+ public:
+  virtual ~AV1JntConvolve2DTest();
+  virtual void SetUp();
+
+  virtual void TearDown();
+
+ protected:
+  void RunCheckOutput(convolve_2d_func test_impl);
+  void RunSpeedTest(convolve_2d_func test_impl);
+
+  libaom_test::ACMRandom rnd_;
+};
 }  // namespace AV1Convolve2D
 
-#if CONFIG_HIGHBITDEPTH
 namespace AV1HighbdConvolve2D {
 typedef void (*highbd_convolve_2d_func)(
-    const uint16_t *src, int src_stride, CONV_BUF_TYPE *dst, int dst_stride,
-    int w, int h, InterpFilterParams *filter_params_x,
+    const uint16_t *src, int src_stride, uint16_t *dst, int dst_stride, int w,
+    int h, InterpFilterParams *filter_params_x,
     InterpFilterParams *filter_params_y, const int subpel_x_q4,
     const int subpel_y_q4, ConvolveParams *conv_params, int bd);
 
-typedef std::tr1::tuple<int, int, int, highbd_convolve_2d_func>
+typedef ::testing::tuple<int, highbd_convolve_2d_func, int, int, BLOCK_SIZE>
     HighbdConvolve2DParam;
 
 ::testing::internal::ParamGenerator<HighbdConvolve2DParam> BuildParams(
-    highbd_convolve_2d_func filter);
+    highbd_convolve_2d_func filter, int subx_exist, int suby_exist);
 
-class AV1HighbdConvolve2DTest
+class AV1HighbdConvolve2DSrTest
     : public ::testing::TestWithParam<HighbdConvolve2DParam> {
  public:
-  virtual ~AV1HighbdConvolve2DTest();
+  virtual ~AV1HighbdConvolve2DSrTest();
   virtual void SetUp();
 
   virtual void TearDown();
 
  protected:
   void RunCheckOutput(highbd_convolve_2d_func test_impl);
+  void RunSpeedTest(highbd_convolve_2d_func test_impl);
 
   libaom_test::ACMRandom rnd_;
 };
 
+class AV1HighbdJntConvolve2DTest
+    : public ::testing::TestWithParam<HighbdConvolve2DParam> {
+ public:
+  virtual ~AV1HighbdJntConvolve2DTest();
+  virtual void SetUp();
+
+  virtual void TearDown();
+
+ protected:
+  void RunCheckOutput(highbd_convolve_2d_func test_impl);
+  void RunSpeedTest(highbd_convolve_2d_func test_impl);
+
+  libaom_test::ACMRandom rnd_;
+};
 }  // namespace AV1HighbdConvolve2D
-#endif  // CONFIG_HIGHBITDEPTH
 
 }  // namespace libaom_test
 

third_party/aom/test/av1_convolve_optimz_test.cc

@@ -1,405 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
-
-#include "./av1_rtcd.h"
-#include "test/acm_random.h"
-#include "test/clear_system_state.h"
-#include "test/register_state_check.h"
-#include "test/util.h"
-
-namespace {
-
-using std::tr1::tuple;
-using libaom_test::ACMRandom;
-
-typedef void (*ConvInit)();
-typedef void (*conv_filter_t)(const uint8_t *, int, uint8_t *, int, int, int,
-                              const InterpFilterParams, int, int,
-                              ConvolveParams *);
-#if CONFIG_HIGHBITDEPTH
-typedef void (*hbd_conv_filter_t)(const uint16_t *, int, uint16_t *, int, int,
-                                  int, const InterpFilterParams, int, int, int,
-                                  int);
-#endif
-
-// Test parameter list:
-//  <convolve_horiz_func, convolve_vert_func,
-//  <width, height>, filter_params, subpel_x_q4, avg>
-typedef tuple<int, int> BlockDimension;
-typedef tuple<ConvInit, conv_filter_t, conv_filter_t, BlockDimension,
-              InterpFilter, int, int>
-    ConvParams;
-#if CONFIG_HIGHBITDEPTH
-// Test parameter list:
-//  <convolve_horiz_func, convolve_vert_func,
-//  <width, height>, filter_params, subpel_x_q4, avg, bit_dpeth>
-typedef tuple<ConvInit, hbd_conv_filter_t, hbd_conv_filter_t, BlockDimension,
-              InterpFilter, int, int, int>
-    HbdConvParams;
-#endif
-
-// Note:
-//  src_ and src_ref_ have special boundary requirement
-//  dst_ and dst_ref_ don't
-const size_t maxWidth = 256;
-const size_t maxHeight = 256;
-const size_t maxBlockSize = maxWidth * maxHeight;
-const int horizOffset = 32;
-const int vertiOffset = 32;
-const int stride = 128;
-const int x_step_q4 = 16;
-
-class AV1ConvolveOptimzTest : public ::testing::TestWithParam<ConvParams> {
- public:
-  virtual ~AV1ConvolveOptimzTest() {}
-  virtual void SetUp() {
-    ConvInit conv_init = GET_PARAM(0);
-    conv_init();
-    conv_horiz_ = GET_PARAM(1);
-    conv_vert_ = GET_PARAM(2);
-    BlockDimension block = GET_PARAM(3);
-    width_ = std::tr1::get<0>(block);
-    height_ = std::tr1::get<1>(block);
-    filter_ = GET_PARAM(4);
-    subpel_ = GET_PARAM(5);
-    int ref = GET_PARAM(6);
-    const int plane = 0;
-    conv_params_ = get_conv_params(ref, ref, plane);
-
-    alloc_ = new uint8_t[maxBlockSize * 4];
-    src_ = alloc_ + (vertiOffset * maxWidth);
-    src_ += horizOffset;
-    src_ref_ = src_ + maxBlockSize;
-
-    dst_ = alloc_ + 2 * maxBlockSize;
-    dst_ref_ = alloc_ + 3 * maxBlockSize;
-  }
-
-  virtual void TearDown() {
-    delete[] alloc_;
-    libaom_test::ClearSystemState();
-  }
-
- protected:
-  void RunHorizFilterBitExactCheck();
-  void RunVertFilterBitExactCheck();
-
- private:
-  void PrepFilterBuffer();
-  void DiffFilterBuffer();
-  conv_filter_t conv_horiz_;
-  conv_filter_t conv_vert_;
-  uint8_t *alloc_;
-  uint8_t *src_;
-  uint8_t *dst_;
-  uint8_t *src_ref_;
-  uint8_t *dst_ref_;
-  int width_;
-  int height_;
-  InterpFilter filter_;
-  int subpel_;
-  ConvolveParams conv_params_;
-};
-
-void AV1ConvolveOptimzTest::PrepFilterBuffer() {
-  int r, c;
-  ACMRandom rnd(ACMRandom::DeterministicSeed());
-
-  memset(alloc_, 0, 4 * maxBlockSize * sizeof(alloc_[0]));
-
-  uint8_t *src_ptr = src_;
-  uint8_t *dst_ptr = dst_;
-  uint8_t *src_ref_ptr = src_ref_;
-  uint8_t *dst_ref_ptr = dst_ref_;
-
-  for (r = 0; r < height_; ++r) {
-    for (c = 0; c < width_; ++c) {
-      src_ptr[c] = rnd.Rand8();
-      src_ref_ptr[c] = src_ptr[c];
-      dst_ptr[c] = rnd.Rand8();
-      dst_ref_ptr[c] = dst_ptr[c];
-    }
-    src_ptr += stride;
-    src_ref_ptr += stride;
-    dst_ptr += stride;
-    dst_ref_ptr += stride;
-  }
-}
-
-void AV1ConvolveOptimzTest::DiffFilterBuffer() {
-  int r, c;
-  const uint8_t *dst_ptr = dst_;
-  const uint8_t *dst_ref_ptr = dst_ref_;
-  for (r = 0; r < height_; ++r) {
-    for (c = 0; c < width_; ++c) {
-      EXPECT_EQ((uint8_t)dst_ref_ptr[c], (uint8_t)dst_ptr[c])
-          << "Error at row: " << r << " col: " << c << " "
-          << "w = " << width_ << " "
-          << "h = " << height_ << " "
-          << "filter group index = " << filter_ << " "
-          << "filter index = " << subpel_;
-    }
-    dst_ptr += stride;
-    dst_ref_ptr += stride;
-  }
-}
-
-void AV1ConvolveOptimzTest::RunHorizFilterBitExactCheck() {
-  PrepFilterBuffer();
-
-  InterpFilterParams filter_params = av1_get_interp_filter_params(filter_);
-
-  av1_convolve_horiz_c(src_ref_, stride, dst_ref_, stride, width_, height_,
-                       filter_params, subpel_, x_step_q4, &conv_params_);
-
-  conv_horiz_(src_, stride, dst_, stride, width_, height_, filter_params,
-              subpel_, x_step_q4, &conv_params_);
-
-  DiffFilterBuffer();
-
-  // Note:
-  // Here we need calculate a height which is different from the specified one
-  // and test again.
-  int intermediate_height =
-      (((height_ - 1) * 16 + subpel_) >> SUBPEL_BITS) + filter_params.taps;
-  PrepFilterBuffer();
-
-  av1_convolve_horiz_c(src_ref_, stride, dst_ref_, stride, width_,
-                       intermediate_height, filter_params, subpel_, x_step_q4,
-                       &conv_params_);
-
-  conv_horiz_(src_, stride, dst_, stride, width_, intermediate_height,
-              filter_params, subpel_, x_step_q4, &conv_params_);
-
-  DiffFilterBuffer();
-}
-
-void AV1ConvolveOptimzTest::RunVertFilterBitExactCheck() {
-  PrepFilterBuffer();
-
-  InterpFilterParams filter_params = av1_get_interp_filter_params(filter_);
-
-  av1_convolve_vert_c(src_ref_, stride, dst_ref_, stride, width_, height_,
-                      filter_params, subpel_, x_step_q4, &conv_params_);
-
-  conv_vert_(src_, stride, dst_, stride, width_, height_, filter_params,
-             subpel_, x_step_q4, &conv_params_);
-
-  DiffFilterBuffer();
-}
-
-TEST_P(AV1ConvolveOptimzTest, HorizBitExactCheck) {
-  RunHorizFilterBitExactCheck();
-}
-TEST_P(AV1ConvolveOptimzTest, VerticalBitExactCheck) {
-  RunVertFilterBitExactCheck();
-}
-
-using std::tr1::make_tuple;
-
-#if (HAVE_SSSE3 || HAVE_SSE4_1) && CONFIG_DUAL_FILTER
-const BlockDimension kBlockDim[] = {
-  make_tuple(2, 2),    make_tuple(2, 4),    make_tuple(4, 4),
-  make_tuple(4, 8),    make_tuple(8, 4),    make_tuple(8, 8),
-  make_tuple(8, 16),   make_tuple(16, 8),   make_tuple(16, 16),
-  make_tuple(16, 32),  make_tuple(32, 16),  make_tuple(32, 32),
-  make_tuple(32, 64),  make_tuple(64, 32),  make_tuple(64, 64),
-  make_tuple(64, 128), make_tuple(128, 64), make_tuple(128, 128),
-};
-
-// 10/12-tap filters
-const InterpFilter kFilter[] = { EIGHTTAP_REGULAR, BILINEAR, MULTITAP_SHARP };
-
-const int kSubpelQ4[] = { 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 };
-
-const int kAvg[] = { 0, 1 };
-#endif
-
-#if HAVE_SSSE3 && CONFIG_DUAL_FILTER
-INSTANTIATE_TEST_CASE_P(
-    SSSE3, AV1ConvolveOptimzTest,
-    ::testing::Combine(::testing::Values(av1_lowbd_convolve_init_ssse3),
-                       ::testing::Values(av1_convolve_horiz_ssse3),
-                       ::testing::Values(av1_convolve_vert_ssse3),
-                       ::testing::ValuesIn(kBlockDim),
-                       ::testing::ValuesIn(kFilter),
-                       ::testing::ValuesIn(kSubpelQ4),
-                       ::testing::ValuesIn(kAvg)));
-#endif  // HAVE_SSSE3 && CONFIG_DUAL_FILTER
-
-#if CONFIG_HIGHBITDEPTH
-typedef ::testing::TestWithParam<HbdConvParams> TestWithHbdConvParams;
-class AV1HbdConvolveOptimzTest : public TestWithHbdConvParams {
- public:
-  virtual ~AV1HbdConvolveOptimzTest() {}
-  virtual void SetUp() {
-    ConvInit conv_init = GET_PARAM(0);
-    conv_init();
-    conv_horiz_ = GET_PARAM(1);
-    conv_vert_ = GET_PARAM(2);
-    BlockDimension block = GET_PARAM(3);
-    width_ = std::tr1::get<0>(block);
-    height_ = std::tr1::get<1>(block);
-    filter_ = GET_PARAM(4);
-    subpel_ = GET_PARAM(5);
-    avg_ = GET_PARAM(6);
-    bit_depth_ = GET_PARAM(7);
-
-    alloc_ = new uint16_t[maxBlockSize * 4];
-    src_ = alloc_ + (vertiOffset * maxWidth);
-    src_ += horizOffset;
-    src_ref_ = src_ + maxBlockSize;
-
-    dst_ = alloc_ + 2 * maxBlockSize;
-    dst_ref_ = alloc_ + 3 * maxBlockSize;
-  }
-
-  virtual void TearDown() {
-    delete[] alloc_;
-    libaom_test::ClearSystemState();
-  }
-
- protected:
-  void RunHorizFilterBitExactCheck();
-  void RunVertFilterBitExactCheck();
-
- private:
-  void PrepFilterBuffer();
-  void DiffFilterBuffer();
-  hbd_conv_filter_t conv_horiz_;
-  hbd_conv_filter_t conv_vert_;
-  uint16_t *alloc_;
-  uint16_t *src_;
-  uint16_t *dst_;
-  uint16_t *src_ref_;
-  uint16_t *dst_ref_;
-  int width_;
-  int height_;
-  InterpFilter filter_;
-  int subpel_;
-  int avg_;
-  int bit_depth_;
-};
-
-void AV1HbdConvolveOptimzTest::PrepFilterBuffer() {
-  int r, c;
-  ACMRandom rnd(ACMRandom::DeterministicSeed());
-
-  memset(alloc_, 0, 4 * maxBlockSize * sizeof(alloc_[0]));
-
-  uint16_t *src_ptr = src_;
-  uint16_t *dst_ptr = dst_;
-  uint16_t *dst_ref_ptr = dst_ref_;
-  uint16_t hbd_mask = (1 << bit_depth_) - 1;
-
-  for (r = 0; r < height_; ++r) {
-    for (c = 0; c < width_; ++c) {
-      src_ptr[c] = rnd.Rand16() & hbd_mask;
-      dst_ptr[c] = rnd.Rand16() & hbd_mask;
-      dst_ref_ptr[c] = dst_ptr[c];
-    }
-    src_ptr += stride;
-    dst_ptr += stride;
-    dst_ref_ptr += stride;
-  }
-}
-
-void AV1HbdConvolveOptimzTest::DiffFilterBuffer() {
-  int r, c;
-  const uint16_t *dst_ptr = dst_;
-  const uint16_t *dst_ref_ptr = dst_ref_;
-  for (r = 0; r < height_; ++r) {
-    for (c = 0; c < width_; ++c) {
-      EXPECT_EQ((uint16_t)dst_ref_ptr[c], (uint16_t)dst_ptr[c])
-          << "Error at row: " << r << " col: " << c << " "
-          << "w = " << width_ << " "
-          << "h = " << height_ << " "
-          << "filter group index = " << filter_ << " "
-          << "filter index = " << subpel_ << " "
-          << "bit depth = " << bit_depth_;
-    }
-    dst_ptr += stride;
-    dst_ref_ptr += stride;
-  }
-}
-
-void AV1HbdConvolveOptimzTest::RunHorizFilterBitExactCheck() {
-  PrepFilterBuffer();
-
-  InterpFilterParams filter_params = av1_get_interp_filter_params(filter_);
-
-  av1_highbd_convolve_horiz_c(src_, stride, dst_ref_, stride, width_, height_,
-                              filter_params, subpel_, x_step_q4, avg_,
-                              bit_depth_);
-
-  conv_horiz_(src_, stride, dst_, stride, width_, height_, filter_params,
-              subpel_, x_step_q4, avg_, bit_depth_);
-
-  DiffFilterBuffer();
-
-  // Note:
-  // Here we need calculate a height which is different from the specified one
-  // and test again.
-  int intermediate_height =
-      (((height_ - 1) * 16 + subpel_) >> SUBPEL_BITS) + filter_params.taps;
-  PrepFilterBuffer();
-
-  av1_highbd_convolve_horiz_c(src_, stride, dst_ref_, stride, width_,
-                              intermediate_height, filter_params, subpel_,
-                              x_step_q4, avg_, bit_depth_);
-
-  conv_horiz_(src_, stride, dst_, stride, width_, intermediate_height,
-              filter_params, subpel_, x_step_q4, avg_, bit_depth_);
-
-  DiffFilterBuffer();
-}
-
-void AV1HbdConvolveOptimzTest::RunVertFilterBitExactCheck() {
-  PrepFilterBuffer();
-
-  InterpFilterParams filter_params = av1_get_interp_filter_params(filter_);
-
-  av1_highbd_convolve_vert_c(src_, stride, dst_ref_, stride, width_, height_,
-                             filter_params, subpel_, x_step_q4, avg_,
-                             bit_depth_);
-
-  conv_vert_(src_, stride, dst_, stride, width_, height_, filter_params,
-             subpel_, x_step_q4, avg_, bit_depth_);
-
-  DiffFilterBuffer();
-}
-
-TEST_P(AV1HbdConvolveOptimzTest, HorizBitExactCheck) {
-  RunHorizFilterBitExactCheck();
-}
-TEST_P(AV1HbdConvolveOptimzTest, VertBitExactCheck) {
-  RunVertFilterBitExactCheck();
-}
-
-#if HAVE_SSE4_1 && CONFIG_DUAL_FILTER
-
-const int kBitdepth[] = { 10, 12 };
-
-INSTANTIATE_TEST_CASE_P(
-    SSE4_1, AV1HbdConvolveOptimzTest,
-    ::testing::Combine(::testing::Values(av1_highbd_convolve_init_sse4_1),
-                       ::testing::Values(av1_highbd_convolve_horiz_sse4_1),
-                       ::testing::Values(av1_highbd_convolve_vert_sse4_1),
-                       ::testing::ValuesIn(kBlockDim),
-                       ::testing::ValuesIn(kFilter),
-                       ::testing::ValuesIn(kSubpelQ4),
-                       ::testing::ValuesIn(kAvg),
-                       ::testing::ValuesIn(kBitdepth)));
-#endif  // HAVE_SSE4_1 && CONFIG_DUAL_FILTER
-#endif  // CONFIG_HIGHBITDEPTH
-}  // namespace

third_party/aom/test/av1_convolve_scale_test.cc

@@ -13,13 +13,16 @@
 
 #include "third_party/googletest/src/googletest/include/gtest/gtest.h"
 
-#include "./av1_rtcd.h"
+#include "config/av1_rtcd.h"
+
 #include "aom_ports/aom_timer.h"
 #include "test/acm_random.h"
 #include "test/clear_system_state.h"
 #include "test/register_state_check.h"
 #include "test/util.h"
 
+#include "av1/common/common_data.h"
+
 namespace {
 const int kTestIters = 10;
 const int kPerfIters = 1000;
@@ -29,8 +32,8 @@ const int kHPad = 32;
 const int kXStepQn = 16;
 const int kYStepQn = 20;
 
-using std::tr1::tuple;
-using std::tr1::make_tuple;
+using ::testing::make_tuple;
+using ::testing::tuple;
 using libaom_test::ACMRandom;
 
 enum NTaps { EIGHT_TAP, TEN_TAP, TWELVE_TAP };
@@ -120,6 +123,7 @@ class TestImage {
     // Allocate image data
     src_data_.resize(2 * src_block_size());
     dst_data_.resize(2 * dst_block_size());
+    dst_16_data_.resize(2 * dst_block_size());
   }
 
   void Initialize(ACMRandom *rnd);
@@ -136,8 +140,13 @@ class TestImage {
     return borders ? block : block + kHPad + src_stride_ * kVPad;
   }
 
-  int32_t *GetDstData(bool ref, bool borders) {
-    int32_t *block = &dst_data_[ref ? 0 : dst_block_size()];
+  SrcPixel *GetDstData(bool ref, bool borders) {
+    SrcPixel *block = &dst_data_[ref ? 0 : dst_block_size()];
+    return borders ? block : block + kHPad + dst_stride_ * kVPad;
+  }
+
+  CONV_BUF_TYPE *GetDst16Data(bool ref, bool borders) {
+    CONV_BUF_TYPE *block = &dst_16_data_[ref ? 0 : dst_block_size()];
     return borders ? block : block + kHPad + dst_stride_ * kVPad;
   }
 
@@ -146,7 +155,8 @@ class TestImage {
   int src_stride_, dst_stride_;
 
   std::vector<SrcPixel> src_data_;
-  std::vector<int32_t> dst_data_;
+  std::vector<SrcPixel> dst_data_;
+  std::vector<CONV_BUF_TYPE> dst_16_data_;
 };
 
 template <typename Pixel>
@@ -190,17 +200,23 @@ template <typename SrcPixel>
 void TestImage<SrcPixel>::Initialize(ACMRandom *rnd) {
   PrepBuffers(rnd, w_, h_, src_stride_, bd_, false, &src_data_[0]);
   PrepBuffers(rnd, w_, h_, dst_stride_, bd_, true, &dst_data_[0]);
+  PrepBuffers(rnd, w_, h_, dst_stride_, bd_, true, &dst_16_data_[0]);
 }
 
 template <typename SrcPixel>
 void TestImage<SrcPixel>::Check() const {
   // If memcmp returns 0, there's nothing to do.
   const int num_pixels = dst_block_size();
-  const int32_t *ref_dst = &dst_data_[0];
-  const int32_t *tst_dst = &dst_data_[num_pixels];
+  const SrcPixel *ref_dst = &dst_data_[0];
+  const SrcPixel *tst_dst = &dst_data_[num_pixels];
 
-  if (0 == memcmp(ref_dst, tst_dst, sizeof(*ref_dst) * num_pixels)) return;
+  const CONV_BUF_TYPE *ref_16_dst = &dst_16_data_[0];
+  const CONV_BUF_TYPE *tst_16_dst = &dst_16_data_[num_pixels];
 
+  if (0 == memcmp(ref_dst, tst_dst, sizeof(*ref_dst) * num_pixels)) {
+    if (0 == memcmp(ref_16_dst, tst_16_dst, sizeof(*ref_16_dst) * num_pixels))
+      return;
+  }
   // Otherwise, iterate through the buffer looking for differences (including
   // the edges)
   const int stride = dst_stride_;
@@ -213,6 +229,17 @@ void TestImage<SrcPixel>::Check() const {
           << "Error at row: " << (r - kVPad) << ", col: " << (c - kHPad);
     }
   }
+
+  for (int r = 0; r < h_ + 2 * kVPad; ++r) {
+    for (int c = 0; c < w_ + 2 * kHPad; ++c) {
+      const int32_t ref_value = ref_16_dst[r * stride + c];
+      const int32_t tst_value = tst_16_dst[r * stride + c];
+
+      EXPECT_EQ(tst_value, ref_value)
+          << "Error in 16 bit buffer "
+          << "Error at row: " << (r - kVPad) << ", col: " << (c - kHPad);
+    }
+  }
 }
 
 typedef tuple<int, int> BlockDimension;
@@ -242,8 +269,8 @@ class ConvolveScaleTestBase : public ::testing::Test {
 
  protected:
   void SetParams(const BaseParams &params, int bd) {
-    width_ = std::tr1::get<0>(params.dims);
-    height_ = std::tr1::get<1>(params.dims);
+    width_ = ::testing::get<0>(params.dims);
+    height_ = ::testing::get<1>(params.dims);
     ntaps_x_ = params.ntaps_x;
     ntaps_y_ = params.ntaps_y;
     bd_ = bd;
@@ -251,19 +278,54 @@ class ConvolveScaleTestBase : public ::testing::Test {
 
     filter_x_.set(ntaps_x_, false);
     filter_y_.set(ntaps_y_, true);
-    convolve_params_ = get_conv_params_no_round(0, avg_ != false, 0, NULL, 0);
+    convolve_params_ =
+        get_conv_params_no_round(0, avg_ != false, 0, NULL, 0, 1, bd);
 
     delete image_;
     image_ = new TestImage<SrcPixel>(width_, height_, bd_);
   }
 
+  void SetConvParamOffset(int i, int j, int is_compound, int do_average,
+                          int use_jnt_comp_avg) {
+    if (i == -1 && j == -1) {
+      convolve_params_.use_jnt_comp_avg = use_jnt_comp_avg;
+      convolve_params_.is_compound = is_compound;
+      convolve_params_.do_average = do_average;
+    } else {
+      convolve_params_.use_jnt_comp_avg = use_jnt_comp_avg;
+      convolve_params_.fwd_offset = quant_dist_lookup_table[i][j][0];
+      convolve_params_.bck_offset = quant_dist_lookup_table[i][j][1];
+      convolve_params_.is_compound = is_compound;
+      convolve_params_.do_average = do_average;
+    }
+  }
+
   void Run() {
     ACMRandom rnd(ACMRandom::DeterministicSeed());
     for (int i = 0; i < kTestIters; ++i) {
+      int is_compound = 0;
+      SetConvParamOffset(-1, -1, is_compound, 0, 0);
       Prep(&rnd);
       RunOne(true);
       RunOne(false);
       image_->Check();
+
+      is_compound = 1;
+      for (int do_average = 0; do_average < 2; do_average++) {
+        for (int use_jnt_comp_avg = 0; use_jnt_comp_avg < 2;
+             use_jnt_comp_avg++) {
+          for (int j = 0; j < 2; ++j) {
+            for (int k = 0; k < 4; ++k) {
+              SetConvParamOffset(j, k, is_compound, do_average,
+                                 use_jnt_comp_avg);
+              Prep(&rnd);
+              RunOne(true);
+              RunOne(false);
+              image_->Check();
+            }
+          }
+        }
+      }
     }
   }
 
@@ -327,7 +389,7 @@ class ConvolveScaleTestBase : public ::testing::Test {
 typedef tuple<int, int> BlockDimension;
 
 typedef void (*LowbdConvolveFunc)(const uint8_t *src, int src_stride,
-                                  int32_t *dst, int dst_stride, int w, int h,
+                                  uint8_t *dst, int dst_stride, int w, int h,
                                   InterpFilterParams *filter_params_x,
                                   InterpFilterParams *filter_params_y,
                                   const int subpel_x_qn, const int x_step_qn,
@@ -359,10 +421,10 @@ class LowBDConvolveScaleTest
 
   void RunOne(bool ref) {
     const uint8_t *src = image_->GetSrcData(ref, false);
-    CONV_BUF_TYPE *dst = image_->GetDstData(ref, false);
+    uint8_t *dst = image_->GetDstData(ref, false);
+    convolve_params_.dst = image_->GetDst16Data(ref, false);
     const int src_stride = image_->src_stride();
     const int dst_stride = image_->dst_stride();
-
     if (ref) {
       av1_convolve_2d_scale_c(src, src_stride, dst, dst_stride, width_, height_,
                               &filter_x_.params_, &filter_y_.params_, subpel_x_,
@@ -387,7 +449,7 @@ const BlockDimension kBlockDim[] = {
   make_tuple(64, 128), make_tuple(128, 64), make_tuple(128, 128),
 };
 
-const NTaps kNTaps[] = { EIGHT_TAP, TEN_TAP, TWELVE_TAP };
+const NTaps kNTaps[] = { EIGHT_TAP };
 
 TEST_P(LowBDConvolveScaleTest, Check) { Run(); }
 TEST_P(LowBDConvolveScaleTest, DISABLED_Speed) { SpeedTest(); }
@@ -399,9 +461,8 @@ INSTANTIATE_TEST_CASE_P(
                        ::testing::ValuesIn(kNTaps), ::testing::ValuesIn(kNTaps),
                        ::testing::Bool()));
 
-#if CONFIG_HIGHBITDEPTH
 typedef void (*HighbdConvolveFunc)(const uint16_t *src, int src_stride,
-                                   int32_t *dst, int dst_stride, int w, int h,
+                                   uint16_t *dst, int dst_stride, int w, int h,
                                    InterpFilterParams *filter_params_x,
                                    InterpFilterParams *filter_params_y,
                                    const int subpel_x_qn, const int x_step_qn,
@@ -433,7 +494,8 @@ class HighBDConvolveScaleTest
 
   void RunOne(bool ref) {
     const uint16_t *src = image_->GetSrcData(ref, false);
-    CONV_BUF_TYPE *dst = image_->GetDstData(ref, false);
+    uint16_t *dst = image_->GetDstData(ref, false);
+    convolve_params_.dst = image_->GetDst16Data(ref, false);
     const int src_stride = image_->src_stride();
     const int dst_stride = image_->dst_stride();
 
@@ -464,6 +526,4 @@ INSTANTIATE_TEST_CASE_P(
                        ::testing::ValuesIn(kBlockDim),
                        ::testing::ValuesIn(kNTaps), ::testing::ValuesIn(kNTaps),
                        ::testing::Bool(), ::testing::ValuesIn(kBDs)));
-
-#endif  // CONFIG_HIGHBITDEPTH
 }  // namespace

third_party/aom/test/av1_convolve_test.cc

@@ -1,514 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <algorithm>
-#include <vector>
-
-#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
-
-#include "./av1_rtcd.h"
-#include "./aom_dsp_rtcd.h"
-#include "aom_dsp/aom_dsp_common.h"
-#include "aom_ports/mem.h"
-#include "av1/common/filter.h"
-#include "av1/common/convolve.h"
-#include "test/acm_random.h"
-#include "test/util.h"
-
-using libaom_test::ACMRandom;
-
-namespace {
-using std::tr1::tuple;
-static void filter_block1d_horiz_c(const uint8_t *src_ptr, int src_stride,
-                                   const int16_t *filter, int tap,
-                                   uint8_t *dst_ptr, int dst_stride, int w,
-                                   int h) {
-  src_ptr -= tap / 2 - 1;
-  for (int r = 0; r < h; ++r) {
-    for (int c = 0; c < w; ++c) {
-      int sum = 0;
-      for (int i = 0; i < tap; ++i) {
-        sum += src_ptr[c + i] * filter[i];
-      }
-      dst_ptr[c] = clip_pixel(ROUND_POWER_OF_TWO(sum, FILTER_BITS));
-    }
-    src_ptr += src_stride;
-    dst_ptr += dst_stride;
-  }
-}
-
-static void filter_block1d_vert_c(const uint8_t *src_ptr, int src_stride,
-                                  const int16_t *filter, int tap,
-                                  uint8_t *dst_ptr, int dst_stride, int w,
-                                  int h) {
-  src_ptr -= (tap / 2 - 1) * src_stride;
-  for (int r = 0; r < h; ++r) {
-    for (int c = 0; c < w; ++c) {
-      int sum = 0;
-      for (int i = 0; i < tap; ++i) {
-        sum += src_ptr[c + i * src_stride] * filter[i];
-      }
-      dst_ptr[c] = clip_pixel(ROUND_POWER_OF_TWO(sum, FILTER_BITS));
-    }
-    src_ptr += src_stride;
-    dst_ptr += dst_stride;
-  }
-}
-
-static int match(const uint8_t *out, int out_stride, const uint8_t *ref_out,
-                 int ref_out_stride, int w, int h) {
-  for (int r = 0; r < h; ++r) {
-    for (int c = 0; c < w; ++c) {
-      if (out[r * out_stride + c] != ref_out[r * ref_out_stride + c]) return 0;
-    }
-  }
-  return 1;
-}
-
-typedef void (*ConvolveFunc)(const uint8_t *src, int src_stride, uint8_t *dst,
-                             int dst_stride, int w, int h,
-                             const InterpFilterParams filter_params,
-                             const int subpel_q4, int step_q4,
-                             ConvolveParams *conv_params);
-
-struct ConvolveFunctions {
-  ConvolveFunctions(ConvolveFunc hf, ConvolveFunc vf) : hf_(hf), vf_(vf) {}
-  ConvolveFunc hf_;
-  ConvolveFunc vf_;
-};
-
-typedef tuple<ConvolveFunctions *, InterpFilter /*filter_x*/,
-              InterpFilter /*filter_y*/>
-    ConvolveParam;
-
-class Av1ConvolveTest : public ::testing::TestWithParam<ConvolveParam> {
- public:
-  virtual void SetUp() {
-    rnd_(ACMRandom::DeterministicSeed());
-    cfs_ = GET_PARAM(0);
-    interp_filter_ls_[0] = GET_PARAM(2);
-    interp_filter_ls_[2] = interp_filter_ls_[0];
-    interp_filter_ls_[1] = GET_PARAM(1);
-    interp_filter_ls_[3] = interp_filter_ls_[1];
-  }
-  virtual void TearDown() {
-    while (buf_ls_.size() > 0) {
-      uint8_t *buf = buf_ls_.back();
-      aom_free(buf);
-      buf_ls_.pop_back();
-    }
-  }
-  virtual uint8_t *add_input(int w, int h, int *stride) {
-    uint8_t *buf =
-        reinterpret_cast<uint8_t *>(aom_memalign(kDataAlignment, kBufferSize));
-    buf_ls_.push_back(buf);
-    *stride = w + MAX_FILTER_TAP - 1;
-    int offset = MAX_FILTER_TAP / 2 - 1;
-    for (int r = 0; r < h + MAX_FILTER_TAP - 1; ++r) {
-      for (int c = 0; c < w + MAX_FILTER_TAP - 1; ++c) {
-        buf[r * (*stride) + c] = rnd_.Rand8();
-      }
-    }
-    return buf + offset * (*stride) + offset;
-  }
-  virtual uint8_t *add_output(int w, int /*h*/, int *stride) {
-    uint8_t *buf =
-        reinterpret_cast<uint8_t *>(aom_memalign(kDataAlignment, kBufferSize));
-    buf_ls_.push_back(buf);
-    *stride = w;
-    return buf;
-  }
-  virtual void random_init_buf(uint8_t *buf, int w, int h, int stride) {
-    for (int r = 0; r < h; ++r) {
-      for (int c = 0; c < w; ++c) {
-        buf[r * stride + c] = rnd_.Rand8();
-      }
-    }
-  }
-
- protected:
-  static const int kDataAlignment = 16;
-  static const int kOuterBlockSize = MAX_SB_SIZE + MAX_FILTER_TAP - 1;
-  static const int kBufferSize = kOuterBlockSize * kOuterBlockSize;
-  std::vector<uint8_t *> buf_ls_;
-  InterpFilter interp_filter_ls_[4];
-  ConvolveFunctions *cfs_;
-  ACMRandom rnd_;
-};
-
-int bsize_ls[] = { 1, 2, 4, 8, 16, 32, 64, 3, 7, 15, 31, 63 };
-int bsize_num = NELEMENTS(bsize_ls);
-
-TEST_P(Av1ConvolveTest, av1_convolve_vert) {
-  const int y_step_q4 = 16;
-  ConvolveParams conv_params = get_conv_params(0, 0, 0);
-
-  int in_stride, out_stride, ref_out_stride, avg_out_stride, ref_avg_out_stride;
-  uint8_t *in = add_input(MAX_SB_SIZE, MAX_SB_SIZE, &in_stride);
-  uint8_t *out = add_output(MAX_SB_SIZE, MAX_SB_SIZE, &out_stride);
-  uint8_t *ref_out = add_output(MAX_SB_SIZE, MAX_SB_SIZE, &ref_out_stride);
-  uint8_t *avg_out = add_output(MAX_SB_SIZE, MAX_SB_SIZE, &avg_out_stride);
-  uint8_t *ref_avg_out =
-      add_output(MAX_SB_SIZE, MAX_SB_SIZE, &ref_avg_out_stride);
-  for (int hb_idx = 0; hb_idx < bsize_num; ++hb_idx) {
-    for (int vb_idx = 0; vb_idx < bsize_num; ++vb_idx) {
-      int w = bsize_ls[hb_idx];
-      int h = bsize_ls[vb_idx];
-      for (int subpel_y_q4 = 0; subpel_y_q4 < SUBPEL_SHIFTS; ++subpel_y_q4) {
-        InterpFilter filter_y = interp_filter_ls_[0];
-        InterpFilterParams param_vert = av1_get_interp_filter_params(filter_y);
-        const int16_t *filter_vert =
-            av1_get_interp_filter_subpel_kernel(param_vert, subpel_y_q4);
-
-        filter_block1d_vert_c(in, in_stride, filter_vert, param_vert.taps,
-                              ref_out, ref_out_stride, w, h);
-
-        conv_params.ref = 0;
-        conv_params.do_average = 0;
-        cfs_->vf_(in, in_stride, out, out_stride, w, h, param_vert, subpel_y_q4,
-                  y_step_q4, &conv_params);
-        EXPECT_EQ(match(out, out_stride, ref_out, ref_out_stride, w, h), 1)
-            << " hb_idx " << hb_idx << " vb_idx " << vb_idx << " filter_y "
-            << filter_y << " subpel_y_q4 " << subpel_y_q4;
-
-        random_init_buf(avg_out, w, h, avg_out_stride);
-        for (int r = 0; r < h; ++r) {
-          for (int c = 0; c < w; ++c) {
-            ref_avg_out[r * ref_avg_out_stride + c] = ROUND_POWER_OF_TWO(
-                avg_out[r * avg_out_stride + c] + out[r * out_stride + c], 1);
-          }
-        }
-        conv_params.ref = 1;
-        conv_params.do_average = 1;
-        cfs_->vf_(in, in_stride, avg_out, avg_out_stride, w, h, param_vert,
-                  subpel_y_q4, y_step_q4, &conv_params);
-        EXPECT_EQ(match(avg_out, avg_out_stride, ref_avg_out,
-                        ref_avg_out_stride, w, h),
-                  1)
-            << " hb_idx " << hb_idx << " vb_idx " << vb_idx << " filter_y "
-            << filter_y << " subpel_y_q4 " << subpel_y_q4;
-      }
-    }
-  }
-};
-
-TEST_P(Av1ConvolveTest, av1_convolve_horiz) {
-  const int x_step_q4 = 16;
-  ConvolveParams conv_params = get_conv_params(0, 0, 0);
-
-  int in_stride, out_stride, ref_out_stride, avg_out_stride, ref_avg_out_stride;
-  uint8_t *in = add_input(MAX_SB_SIZE, MAX_SB_SIZE, &in_stride);
-  uint8_t *out = add_output(MAX_SB_SIZE, MAX_SB_SIZE, &out_stride);
-  uint8_t *ref_out = add_output(MAX_SB_SIZE, MAX_SB_SIZE, &ref_out_stride);
-  uint8_t *avg_out = add_output(MAX_SB_SIZE, MAX_SB_SIZE, &avg_out_stride);
-  uint8_t *ref_avg_out =
-      add_output(MAX_SB_SIZE, MAX_SB_SIZE, &ref_avg_out_stride);
-  for (int hb_idx = 0; hb_idx < bsize_num; ++hb_idx) {
-    for (int vb_idx = 0; vb_idx < bsize_num; ++vb_idx) {
-      int w = bsize_ls[hb_idx];
-      int h = bsize_ls[vb_idx];
-      for (int subpel_x_q4 = 0; subpel_x_q4 < SUBPEL_SHIFTS; ++subpel_x_q4) {
-        InterpFilter filter_x = interp_filter_ls_[1];
-        InterpFilterParams param_horiz = av1_get_interp_filter_params(filter_x);
-        const int16_t *filter_horiz =
-            av1_get_interp_filter_subpel_kernel(param_horiz, subpel_x_q4);
-
-        filter_block1d_horiz_c(in, in_stride, filter_horiz, param_horiz.taps,
-                               ref_out, ref_out_stride, w, h);
-
-        conv_params.ref = 0;
-        conv_params.do_average = 0;
-        cfs_->hf_(in, in_stride, out, out_stride, w, h, param_horiz,
-                  subpel_x_q4, x_step_q4, &conv_params);
-        EXPECT_EQ(match(out, out_stride, ref_out, ref_out_stride, w, h), 1)
-            << " hb_idx " << hb_idx << " vb_idx " << vb_idx << " filter_x "
-            << filter_x << " subpel_x_q4 " << subpel_x_q4;
-
-        random_init_buf(avg_out, w, h, avg_out_stride);
-        for (int r = 0; r < h; ++r) {
-          for (int c = 0; c < w; ++c) {
-            ref_avg_out[r * ref_avg_out_stride + c] = ROUND_POWER_OF_TWO(
-                avg_out[r * avg_out_stride + c] + out[r * out_stride + c], 1);
-          }
-        }
-        conv_params.ref = 1;
-        conv_params.do_average = 1;
-        cfs_->hf_(in, in_stride, avg_out, avg_out_stride, w, h, param_horiz,
-                  subpel_x_q4, x_step_q4, &conv_params);
-        EXPECT_EQ(match(avg_out, avg_out_stride, ref_avg_out,
-                        ref_avg_out_stride, w, h),
-                  1)
-            << "hb_idx " << hb_idx << "vb_idx" << vb_idx << " filter_x "
-            << filter_x << "subpel_x_q4 " << subpel_x_q4;
-      }
-    }
-  }
-};
-
-ConvolveFunctions convolve_functions_c(av1_convolve_horiz_c,
-                                       av1_convolve_vert_c);
-
-InterpFilter filter_ls[] = { EIGHTTAP_REGULAR, EIGHTTAP_SMOOTH,
-                             MULTITAP_SHARP };
-
-INSTANTIATE_TEST_CASE_P(
-    C, Av1ConvolveTest,
-    ::testing::Combine(::testing::Values(&convolve_functions_c),
-                       ::testing::ValuesIn(filter_ls),
-                       ::testing::ValuesIn(filter_ls)));
-
-#if CONFIG_HIGHBITDEPTH
-#ifndef __clang_analyzer__
-TEST(AV1ConvolveTest, av1_highbd_convolve) {
-  ACMRandom rnd(ACMRandom::DeterministicSeed());
-  InterpFilters interp_filters = av1_broadcast_interp_filter(EIGHTTAP_REGULAR);
-  InterpFilterParams filter_params =
-      av1_get_interp_filter_params(EIGHTTAP_REGULAR);
-  int filter_size = filter_params.taps;
-  int filter_center = filter_size / 2 - 1;
-  uint16_t src[12 * 12];
-  int src_stride = filter_size;
-  uint16_t dst[1] = { 0 };
-  int dst_stride = 1;
-  int x_step_q4 = 16;
-  int y_step_q4 = 16;
-  int avg = 0;
-  int bd = 10;
-  int w = 1;
-  int h = 1;
-
-  int subpel_x_q4;
-  int subpel_y_q4;
-
-  for (int i = 0; i < filter_size * filter_size; i++) {
-    src[i] = rnd.Rand16() % (1 << bd);
-  }
-
-  for (subpel_x_q4 = 0; subpel_x_q4 < SUBPEL_SHIFTS; subpel_x_q4++) {
-    for (subpel_y_q4 = 0; subpel_y_q4 < SUBPEL_SHIFTS; subpel_y_q4++) {
-      av1_highbd_convolve(
-          CONVERT_TO_BYTEPTR(src + src_stride * filter_center + filter_center),
-          src_stride, CONVERT_TO_BYTEPTR(dst), dst_stride, w, h, interp_filters,
-          subpel_x_q4, x_step_q4, subpel_y_q4, y_step_q4, avg, bd);
-
-      const int16_t *x_filter =
-          av1_get_interp_filter_subpel_kernel(filter_params, subpel_x_q4);
-      const int16_t *y_filter =
-          av1_get_interp_filter_subpel_kernel(filter_params, subpel_y_q4);
-
-      int temp[12];
-      int dst_ref = 0;
-      for (int r = 0; r < filter_size; r++) {
-        temp[r] = 0;
-        for (int c = 0; c < filter_size; c++) {
-          temp[r] += x_filter[c] * src[r * filter_size + c];
-        }
-        temp[r] =
-            clip_pixel_highbd(ROUND_POWER_OF_TWO(temp[r], FILTER_BITS), bd);
-        dst_ref += temp[r] * y_filter[r];
-      }
-      dst_ref = clip_pixel_highbd(ROUND_POWER_OF_TWO(dst_ref, FILTER_BITS), bd);
-      EXPECT_EQ(dst[0], dst_ref);
-    }
-  }
-}
-#endif
-
-TEST(AV1ConvolveTest, av1_highbd_convolve_avg) {
-  ACMRandom rnd(ACMRandom::DeterministicSeed());
-  InterpFilters interp_filters = av1_broadcast_interp_filter(EIGHTTAP_REGULAR);
-  InterpFilterParams filter_params =
-      av1_get_interp_filter_params(EIGHTTAP_REGULAR);
-  int filter_size = filter_params.taps;
-  int filter_center = filter_size / 2 - 1;
-  uint16_t src0[12 * 12];
-  uint16_t src1[12 * 12];
-  int src_stride = filter_size;
-  uint16_t dst0[1] = { 0 };
-  uint16_t dst1[1] = { 0 };
-  uint16_t dst[1] = { 0 };
-  int dst_stride = 1;
-  int x_step_q4 = 16;
-  int y_step_q4 = 16;
-  int avg = 0;
-  int bd = 10;
-
-  int w = 1;
-  int h = 1;
-
-  int subpel_x_q4;
-  int subpel_y_q4;
-
-  for (int i = 0; i < filter_size * filter_size; i++) {
-    src0[i] = rnd.Rand16() % (1 << bd);
-    src1[i] = rnd.Rand16() % (1 << bd);
-  }
-
-  for (subpel_x_q4 = 0; subpel_x_q4 < SUBPEL_SHIFTS; subpel_x_q4++) {
-    for (subpel_y_q4 = 0; subpel_y_q4 < SUBPEL_SHIFTS; subpel_y_q4++) {
-      int offset = filter_size * filter_center + filter_center;
-
-      avg = 0;
-      av1_highbd_convolve(CONVERT_TO_BYTEPTR(src0 + offset), src_stride,
-                          CONVERT_TO_BYTEPTR(dst0), dst_stride, w, h,
-                          interp_filters, subpel_x_q4, x_step_q4, subpel_y_q4,
-                          y_step_q4, avg, bd);
-      avg = 0;
-      av1_highbd_convolve(CONVERT_TO_BYTEPTR(src1 + offset), src_stride,
-                          CONVERT_TO_BYTEPTR(dst1), dst_stride, w, h,
-                          interp_filters, subpel_x_q4, x_step_q4, subpel_y_q4,
-                          y_step_q4, avg, bd);
-
-      avg = 0;
-      av1_highbd_convolve(CONVERT_TO_BYTEPTR(src0 + offset), src_stride,
-                          CONVERT_TO_BYTEPTR(dst), dst_stride, w, h,
-                          interp_filters, subpel_x_q4, x_step_q4, subpel_y_q4,
-                          y_step_q4, avg, bd);
-      avg = 1;
-      av1_highbd_convolve(CONVERT_TO_BYTEPTR(src1 + offset), src_stride,
-                          CONVERT_TO_BYTEPTR(dst), dst_stride, w, h,
-                          interp_filters, subpel_x_q4, x_step_q4, subpel_y_q4,
-                          y_step_q4, avg, bd);
-
-      EXPECT_EQ(dst[0], ROUND_POWER_OF_TWO(dst0[0] + dst1[0], 1));
-    }
-  }
-}
-#endif  // CONFIG_HIGHBITDEPTH
-
-#define CONVOLVE_SPEED_TEST 0
-#if CONVOLVE_SPEED_TEST
-#define highbd_convolve_speed(func, block_size, frame_size)                  \
-  TEST(AV1ConvolveTest, func##_speed_##block_size##_##frame_size) {          \
-    ACMRandom rnd(ACMRandom::DeterministicSeed());                           \
-    InterpFilter interp_filter = EIGHTTAP;                                   \
-    InterpFilterParams filter_params =                                       \
-        av1_get_interp_filter_params(interp_filter);                         \
-    int filter_size = filter_params.tap;                                     \
-    int filter_center = filter_size / 2 - 1;                                 \
-    DECLARE_ALIGNED(16, uint16_t,                                            \
-                    src[(frame_size + 7) * (frame_size + 7)]) = { 0 };       \
-    int src_stride = frame_size + 7;                                         \
-    DECLARE_ALIGNED(16, uint16_t, dst[frame_size * frame_size]) = { 0 };     \
-    int dst_stride = frame_size;                                             \
-    int x_step_q4 = 16;                                                      \
-    int y_step_q4 = 16;                                                      \
-    int subpel_x_q4 = 8;                                                     \
-    int subpel_y_q4 = 6;                                                     \
-    int bd = 10;                                                             \
-                                                                             \
-    int w = block_size;                                                      \
-    int h = block_size;                                                      \
-                                                                             \
-    const int16_t *filter_x =                                                \
-        av1_get_interp_filter_kernel(filter_params, subpel_x_q4);            \
-    const int16_t *filter_y =                                                \
-        av1_get_interp_filter_kernel(filter_params, subpel_y_q4);            \
-                                                                             \
-    for (int i = 0; i < src_stride * src_stride; i++) {                      \
-      src[i] = rnd.Rand16() % (1 << bd);                                     \
-    }                                                                        \
-                                                                             \
-    int offset = filter_center * src_stride + filter_center;                 \
-    int row_offset = 0;                                                      \
-    int col_offset = 0;                                                      \
-    for (int i = 0; i < 100000; i++) {                                       \
-      int src_total_offset = offset + col_offset * src_stride + row_offset;  \
-      int dst_total_offset = col_offset * dst_stride + row_offset;           \
-      func(CONVERT_TO_BYTEPTR(src + src_total_offset), src_stride,           \
-           CONVERT_TO_BYTEPTR(dst + dst_total_offset), dst_stride, filter_x, \
-           x_step_q4, filter_y, y_step_q4, w, h, bd);                        \
-      if (offset + w + w < frame_size) {                                     \
-        row_offset += w;                                                     \
-      } else {                                                               \
-        row_offset = 0;                                                      \
-        col_offset += h;                                                     \
-      }                                                                      \
-      if (col_offset + h >= frame_size) {                                    \
-        col_offset = 0;                                                      \
-      }                                                                      \
-    }                                                                        \
-  }
-
-#define lowbd_convolve_speed(func, block_size, frame_size)                  \
-  TEST(AV1ConvolveTest, func##_speed_l_##block_size##_##frame_size) {       \
-    ACMRandom rnd(ACMRandom::DeterministicSeed());                          \
-    InterpFilter interp_filter = EIGHTTAP;                                  \
-    InterpFilterParams filter_params =                                      \
-        av1_get_interp_filter_params(interp_filter);                        \
-    int filter_size = filter_params.tap;                                    \
-    int filter_center = filter_size / 2 - 1;                                \
-    DECLARE_ALIGNED(16, uint8_t, src[(frame_size + 7) * (frame_size + 7)]); \
-    int src_stride = frame_size + 7;                                        \
-    DECLARE_ALIGNED(16, uint8_t, dst[frame_size * frame_size]);             \
-    int dst_stride = frame_size;                                            \
-    int x_step_q4 = 16;                                                     \
-    int y_step_q4 = 16;                                                     \
-    int subpel_x_q4 = 8;                                                    \
-    int subpel_y_q4 = 6;                                                    \
-    int bd = 8;                                                             \
-                                                                            \
-    int w = block_size;                                                     \
-    int h = block_size;                                                     \
-                                                                            \
-    const int16_t *filter_x =                                               \
-        av1_get_interp_filter_kernel(filter_params, subpel_x_q4);           \
-    const int16_t *filter_y =                                               \
-        av1_get_interp_filter_kernel(filter_params, subpel_y_q4);           \
-                                                                            \
-    for (int i = 0; i < src_stride * src_stride; i++) {                     \
-      src[i] = rnd.Rand16() % (1 << bd);                                    \
-    }                                                                       \
-                                                                            \
-    int offset = filter_center * src_stride + filter_center;                \
-    int row_offset = 0;                                                     \
-    int col_offset = 0;                                                     \
-    for (int i = 0; i < 100000; i++) {                                      \
-      func(src + offset, src_stride, dst, dst_stride, filter_x, x_step_q4,  \
-           filter_y, y_step_q4, w, h);                                      \
-      if (offset + w + w < frame_size) {                                    \
-        row_offset += w;                                                    \
-      } else {                                                              \
-        row_offset = 0;                                                     \
-        col_offset += h;                                                    \
-      }                                                                     \
-      if (col_offset + h >= frame_size) {                                   \
-        col_offset = 0;                                                     \
-      }                                                                     \
-    }                                                                       \
-  }
-
-// This experiment shows that when frame size is 64x64
-// aom_highbd_convolve8_sse2 and aom_convolve8_sse2's speed are similar.
-// However when frame size becomes 1024x1024
-// aom_highbd_convolve8_sse2 is around 50% slower than aom_convolve8_sse2
-// we think the bottleneck is from memory IO
-highbd_convolve_speed(aom_highbd_convolve8_sse2, 8, 64);
-highbd_convolve_speed(aom_highbd_convolve8_sse2, 16, 64);
-highbd_convolve_speed(aom_highbd_convolve8_sse2, 32, 64);
-highbd_convolve_speed(aom_highbd_convolve8_sse2, 64, 64);
-
-lowbd_convolve_speed(aom_convolve8_sse2, 8, 64);
-lowbd_convolve_speed(aom_convolve8_sse2, 16, 64);
-lowbd_convolve_speed(aom_convolve8_sse2, 32, 64);
-lowbd_convolve_speed(aom_convolve8_sse2, 64, 64);
-
-highbd_convolve_speed(aom_highbd_convolve8_sse2, 8, 1024);
-highbd_convolve_speed(aom_highbd_convolve8_sse2, 16, 1024);
-highbd_convolve_speed(aom_highbd_convolve8_sse2, 32, 1024);
-highbd_convolve_speed(aom_highbd_convolve8_sse2, 64, 1024);
-
-lowbd_convolve_speed(aom_convolve8_sse2, 8, 1024);
-lowbd_convolve_speed(aom_convolve8_sse2, 16, 1024);
-lowbd_convolve_speed(aom_convolve8_sse2, 32, 1024);
-lowbd_convolve_speed(aom_convolve8_sse2, 64, 1024);
-#endif  // CONVOLVE_SPEED_TEST
-}  // namespace

third_party/aom/test/av1_dct_test.cc

@@ -1,112 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
-
-#include <math.h>
-#include <stdlib.h>
-#include <new>
-
-#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
-#include "test/acm_random.h"
-#include "test/util.h"
-#include "./aom_config.h"
-#include "aom_ports/msvc.h"
-
-#undef CONFIG_COEFFICIENT_RANGE_CHECKING
-#define CONFIG_COEFFICIENT_RANGE_CHECKING 1
-#define AV1_DCT_GTEST
-#include "av1/encoder/dct.c"
-#if CONFIG_DAALA_DCT4 || CONFIG_DAALA_DCT8 || CONFIG_DAALA_DCT16 || \
-    CONFIG_DAALA_DCT32
-#include "av1/common/daala_tx.c"
-#endif
-
-using libaom_test::ACMRandom;
-
-namespace {
-void reference_dct_1d(const double *in, double *out, int size) {
-  const double kInvSqrt2 = 0.707106781186547524400844362104;
-  for (int k = 0; k < size; ++k) {
-    out[k] = 0;
-    for (int n = 0; n < size; ++n) {
-      out[k] += in[n] * cos(PI * (2 * n + 1) * k / (2 * size));
-    }
-    if (k == 0) out[k] = out[k] * kInvSqrt2;
-  }
-}
-
-typedef void (*FdctFuncRef)(const double *in, double *out, int size);
-typedef void (*IdctFuncRef)(const double *in, double *out, int size);
-typedef void (*FdctFunc)(const tran_low_t *in, tran_low_t *out);
-typedef void (*IdctFunc)(const tran_low_t *in, tran_low_t *out);
-
-class TransTestBase {
- public:
-  virtual ~TransTestBase() {}
-
- protected:
-  void RunFwdAccuracyCheck() {
-    tran_low_t *input = new tran_low_t[txfm_size_];
-    tran_low_t *output = new tran_low_t[txfm_size_];
-    double *ref_input = new double[txfm_size_];
-    double *ref_output = new double[txfm_size_];
-
-    ACMRandom rnd(ACMRandom::DeterministicSeed());
-    const int count_test_block = 5000;
-    for (int ti = 0; ti < count_test_block; ++ti) {
-      for (int ni = 0; ni < txfm_size_; ++ni) {
-        input[ni] = rnd.Rand8() - rnd.Rand8();
-        ref_input[ni] = static_cast<double>(input[ni]);
-      }
-
-      fwd_txfm_(input, output);
-      fwd_txfm_ref_(ref_input, ref_output, txfm_size_);
-
-      for (int ni = 0; ni < txfm_size_; ++ni) {
-        EXPECT_LE(
-            abs(output[ni] - static_cast<tran_low_t>(round(ref_output[ni]))),
-            max_error_);
-      }
-    }
-
-    delete[] input;
-    delete[] output;
-    delete[] ref_input;
-    delete[] ref_output;
-  }
-
-  double max_error_;
-  int txfm_size_;
-  FdctFunc fwd_txfm_;
-  FdctFuncRef fwd_txfm_ref_;
-};
-
-typedef std::tr1::tuple<FdctFunc, FdctFuncRef, int, int> FdctParam;
-class AV1FwdTxfm : public TransTestBase,
-                   public ::testing::TestWithParam<FdctParam> {
- public:
-  virtual void SetUp() {
-    fwd_txfm_ = GET_PARAM(0);
-    fwd_txfm_ref_ = GET_PARAM(1);
-    txfm_size_ = GET_PARAM(2);
-    max_error_ = GET_PARAM(3);
-  }
-  virtual void TearDown() {}
-};
-
-TEST_P(AV1FwdTxfm, RunFwdAccuracyCheck) { RunFwdAccuracyCheck(); }
-
-INSTANTIATE_TEST_CASE_P(
-    C, AV1FwdTxfm,
-    ::testing::Values(FdctParam(&fdct4, &reference_dct_1d, 4, 1),
-                      FdctParam(&fdct8, &reference_dct_1d, 8, 1),
-                      FdctParam(&fdct16, &reference_dct_1d, 16, 2),
-                      FdctParam(&fdct32, &reference_dct_1d, 32, 3)));
-}  // namespace

third_party/aom/test/av1_ext_tile_test.cc

@@ -46,6 +46,7 @@ class AV1ExtTileTest
     cfg.allow_lowbitdepth = 1;
 
     decoder_ = codec_->CreateDecoder(cfg, 0);
+    decoder_->Control(AV1_SET_TILE_MODE, 1);
     decoder_->Control(AV1_SET_DECODE_TILE_ROW, -1);
     decoder_->Control(AV1_SET_DECODE_TILE_COL, -1);
 
@@ -86,13 +87,8 @@ class AV1ExtTileTest
       encoder->Control(AV1E_SET_TILE_ROWS, kTileSize);
       // TODO(yunqingwang): test single_tile_decoding = 0.
       encoder->Control(AV1E_SET_SINGLE_TILE_DECODING, 1);
-#if CONFIG_EXT_PARTITION
       // Always use 64x64 max partition.
       encoder->Control(AV1E_SET_SUPERBLOCK_SIZE, AOM_SUPERBLOCK_SIZE_64X64);
-#endif
-#if CONFIG_LOOPFILTERING_ACROSS_TILES
-      encoder->Control(AV1E_SET_TILE_LOOPFILTER, 0);
-#endif
     }
 
     if (video->frame() == 1) {
@@ -174,6 +170,23 @@ class AV1ExtTileTest
     }
   }
 
+  void TestRoundTrip() {
+    ::libaom_test::I420VideoSource video(
+        "hantro_collage_w352h288.yuv", kImgWidth, kImgHeight, 30, 1, 0, kLimit);
+    cfg_.rc_target_bitrate = 500;
+    cfg_.g_error_resilient = AOM_ERROR_RESILIENT_DEFAULT;
+    cfg_.large_scale_tile = 1;
+    cfg_.g_lag_in_frames = 0;
+    cfg_.g_threads = 1;
+
+    // Tile encoding
+    init_flags_ = AOM_CODEC_USE_PSNR;
+    ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
+
+    // Compare to check if two vectors are equal.
+    ASSERT_EQ(md5_, tile_md5_);
+  }
+
   ::libaom_test::TestMode encoding_mode_;
   int set_cpu_used_;
   ::libaom_test::Decoder *decoder_;
@@ -182,25 +195,19 @@ class AV1ExtTileTest
   std::vector<std::string> tile_md5_;
 };
 
-TEST_P(AV1ExtTileTest, DecoderResultTest) {
-  ::libaom_test::I420VideoSource video("hantro_collage_w352h288.yuv", kImgWidth,
-                                       kImgHeight, 30, 1, 0, kLimit);
-  cfg_.rc_target_bitrate = 500;
-  cfg_.g_error_resilient = AOM_ERROR_RESILIENT_DEFAULT;
-  cfg_.large_scale_tile = 1;
-  cfg_.g_lag_in_frames = 0;
-  cfg_.g_threads = 1;
-
-  // Tile encoding
-  init_flags_ = AOM_CODEC_USE_PSNR;
-  ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
-
-  // Compare to check if two vectors are equal.
-  ASSERT_EQ(md5_, tile_md5_);
-}
+TEST_P(AV1ExtTileTest, DISABLED_DecoderResultTest) { TestRoundTrip(); }
 
 AV1_INSTANTIATE_TEST_CASE(
     // Now only test 2-pass mode.
     AV1ExtTileTest, ::testing::Values(::libaom_test::kTwoPassGood),
-    ::testing::Range(0, 4));
+    ::testing::Range(1, 4));
+
+class AV1ExtTileTestLarge : public AV1ExtTileTest {};
+
+TEST_P(AV1ExtTileTestLarge, DISABLED_DecoderResultTest) { TestRoundTrip(); }
+
+AV1_INSTANTIATE_TEST_CASE(
+    // Now only test 2-pass mode.
+    AV1ExtTileTestLarge, ::testing::Values(::libaom_test::kTwoPassGood),
+    ::testing::Range(0, 1));
 }  // namespace

third_party/aom/test/av1_fht16x16_test.cc

@@ -1,276 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
-
-#include "./av1_rtcd.h"
-#include "./aom_dsp_rtcd.h"
-
-#include "test/acm_random.h"
-#include "test/clear_system_state.h"
-#include "test/register_state_check.h"
-#include "test/transform_test_base.h"
-#include "test/util.h"
-#include "aom_ports/mem.h"
-
-using libaom_test::ACMRandom;
-
-namespace {
-typedef void (*IhtFunc)(const tran_low_t *in, uint8_t *out, int stride,
-                        const TxfmParam *txfm_param);
-using std::tr1::tuple;
-using libaom_test::FhtFunc;
-typedef tuple<FhtFunc, IhtFunc, TX_TYPE, aom_bit_depth_t, int> Ht16x16Param;
-
-void fht16x16_ref(const int16_t *in, tran_low_t *out, int stride,
-                  TxfmParam *txfm_param) {
-  av1_fht16x16_c(in, out, stride, txfm_param);
-}
-
-void iht16x16_ref(const tran_low_t *in, uint8_t *dest, int stride,
-                  const TxfmParam *txfm_param) {
-  av1_iht16x16_256_add_c(in, dest, stride, txfm_param);
-}
-
-#if CONFIG_HIGHBITDEPTH
-typedef void (*IHbdHtFunc)(const tran_low_t *in, uint8_t *out, int stride,
-                           TX_TYPE tx_type, int bd);
-typedef void (*HbdHtFunc)(const int16_t *input, int32_t *output, int stride,
-                          TX_TYPE tx_type, int bd);
-
-// Target optimized function, tx_type, bit depth
-typedef tuple<HbdHtFunc, TX_TYPE, int> HighbdHt16x16Param;
-
-void highbd_fht16x16_ref(const int16_t *in, int32_t *out, int stride,
-                         TX_TYPE tx_type, int bd) {
-  av1_fwd_txfm2d_16x16_c(in, out, stride, tx_type, bd);
-}
-#endif  // CONFIG_HIGHBITDEPTH
-
-class AV1Trans16x16HT : public libaom_test::TransformTestBase,
-                        public ::testing::TestWithParam<Ht16x16Param> {
- public:
-  virtual ~AV1Trans16x16HT() {}
-
-  virtual void SetUp() {
-    fwd_txfm_ = GET_PARAM(0);
-    inv_txfm_ = GET_PARAM(1);
-    pitch_ = 16;
-    height_ = 16;
-    fwd_txfm_ref = fht16x16_ref;
-    inv_txfm_ref = iht16x16_ref;
-    bit_depth_ = GET_PARAM(3);
-    mask_ = (1 << bit_depth_) - 1;
-    num_coeffs_ = GET_PARAM(4);
-    txfm_param_.tx_type = GET_PARAM(2);
-  }
-  virtual void TearDown() { libaom_test::ClearSystemState(); }
-
- protected:
-  void RunFwdTxfm(const int16_t *in, tran_low_t *out, int stride) {
-    fwd_txfm_(in, out, stride, &txfm_param_);
-  }
-
-  void RunInvTxfm(const tran_low_t *out, uint8_t *dst, int stride) {
-    inv_txfm_(out, dst, stride, &txfm_param_);
-  }
-
-  FhtFunc fwd_txfm_;
-  IhtFunc inv_txfm_;
-};
-
-TEST_P(AV1Trans16x16HT, MemCheck) { RunMemCheck(); }
-TEST_P(AV1Trans16x16HT, AccuracyCheck) { RunAccuracyCheck(1, 0.001); }
-TEST_P(AV1Trans16x16HT, InvAccuracyCheck) { RunInvAccuracyCheck(1); }
-TEST_P(AV1Trans16x16HT, CoeffCheck) { RunCoeffCheck(); }
-TEST_P(AV1Trans16x16HT, InvCoeffCheck) { RunInvCoeffCheck(); }
-
-#if CONFIG_HIGHBITDEPTH
-class AV1HighbdTrans16x16HT
-    : public ::testing::TestWithParam<HighbdHt16x16Param> {
- public:
-  virtual ~AV1HighbdTrans16x16HT() {}
-
-  virtual void SetUp() {
-    fwd_txfm_ = GET_PARAM(0);
-    fwd_txfm_ref_ = highbd_fht16x16_ref;
-    tx_type_ = GET_PARAM(1);
-    bit_depth_ = GET_PARAM(2);
-    mask_ = (1 << bit_depth_) - 1;
-    num_coeffs_ = 256;
-
-    input_ = reinterpret_cast<int16_t *>(
-        aom_memalign(32, sizeof(int16_t) * num_coeffs_));
-    output_ = reinterpret_cast<int32_t *>(
-        aom_memalign(32, sizeof(int32_t) * num_coeffs_));
-    output_ref_ = reinterpret_cast<int32_t *>(
-        aom_memalign(32, sizeof(int32_t) * num_coeffs_));
-  }
-
-  virtual void TearDown() {
-    aom_free(input_);
-    aom_free(output_);
-    aom_free(output_ref_);
-    libaom_test::ClearSystemState();
-  }
-
- protected:
-  void RunBitexactCheck();
-
- private:
-  HbdHtFunc fwd_txfm_;
-  HbdHtFunc fwd_txfm_ref_;
-  TX_TYPE tx_type_;
-  int bit_depth_;
-  int mask_;
-  int num_coeffs_;
-  int16_t *input_;
-  int32_t *output_;
-  int32_t *output_ref_;
-};
-
-void AV1HighbdTrans16x16HT::RunBitexactCheck() {
-  ACMRandom rnd(ACMRandom::DeterministicSeed());
-  int i, j;
-  const int stride = 16;
-  const int num_tests = 1000;
-
-  for (i = 0; i < num_tests; ++i) {
-    for (j = 0; j < num_coeffs_; ++j) {
-      input_[j] = (rnd.Rand16() & mask_) - (rnd.Rand16() & mask_);
-    }
-
-    fwd_txfm_ref_(input_, output_ref_, stride, tx_type_, bit_depth_);
-    ASM_REGISTER_STATE_CHECK(
-        fwd_txfm_(input_, output_, stride, tx_type_, bit_depth_));
-
-    for (j = 0; j < num_coeffs_; ++j) {
-      EXPECT_EQ(output_ref_[j], output_[j])
-          << "Not bit-exact result at index: " << j << " at test block: " << i;
-    }
-  }
-}
-
-TEST_P(AV1HighbdTrans16x16HT, HighbdCoeffCheck) { RunBitexactCheck(); }
-#endif  // CONFIG_HIGHBITDEPTH
-
-using std::tr1::make_tuple;
-
-#if HAVE_SSE2 && !CONFIG_DAALA_DCT16
-const Ht16x16Param kArrayHt16x16Param_sse2[] = {
-  make_tuple(&av1_fht16x16_sse2, &av1_iht16x16_256_add_sse2, DCT_DCT,
-             AOM_BITS_8, 256),
-  make_tuple(&av1_fht16x16_sse2, &av1_iht16x16_256_add_sse2, ADST_DCT,
-             AOM_BITS_8, 256),
-  make_tuple(&av1_fht16x16_sse2, &av1_iht16x16_256_add_sse2, DCT_ADST,
-             AOM_BITS_8, 256),
-  make_tuple(&av1_fht16x16_sse2, &av1_iht16x16_256_add_sse2, ADST_ADST,
-             AOM_BITS_8, 256),
-#if CONFIG_EXT_TX
-  make_tuple(&av1_fht16x16_sse2, &av1_iht16x16_256_add_sse2, FLIPADST_DCT,
-             AOM_BITS_8, 256),
-  make_tuple(&av1_fht16x16_sse2, &av1_iht16x16_256_add_sse2, DCT_FLIPADST,
-             AOM_BITS_8, 256),
-  make_tuple(&av1_fht16x16_sse2, &av1_iht16x16_256_add_sse2, FLIPADST_FLIPADST,
-             AOM_BITS_8, 256),
-  make_tuple(&av1_fht16x16_sse2, &av1_iht16x16_256_add_sse2, ADST_FLIPADST,
-             AOM_BITS_8, 256),
-  make_tuple(&av1_fht16x16_sse2, &av1_iht16x16_256_add_sse2, FLIPADST_ADST,
-             AOM_BITS_8, 256),
-  make_tuple(&av1_fht16x16_sse2, &av1_iht16x16_256_add_sse2, IDTX, AOM_BITS_8,
-             256),
-  make_tuple(&av1_fht16x16_sse2, &av1_iht16x16_256_add_sse2, V_DCT, AOM_BITS_8,
-             256),
-  make_tuple(&av1_fht16x16_sse2, &av1_iht16x16_256_add_sse2, H_DCT, AOM_BITS_8,
-             256),
-  make_tuple(&av1_fht16x16_sse2, &av1_iht16x16_256_add_sse2, V_ADST, AOM_BITS_8,
-             256),
-  make_tuple(&av1_fht16x16_sse2, &av1_iht16x16_256_add_sse2, H_ADST, AOM_BITS_8,
-             256),
-  make_tuple(&av1_fht16x16_sse2, &av1_iht16x16_256_add_sse2, V_FLIPADST,
-             AOM_BITS_8, 256),
-  make_tuple(&av1_fht16x16_sse2, &av1_iht16x16_256_add_sse2, H_FLIPADST,
-             AOM_BITS_8, 256)
-#endif  // CONFIG_EXT_TX
-};
-INSTANTIATE_TEST_CASE_P(SSE2, AV1Trans16x16HT,
-                        ::testing::ValuesIn(kArrayHt16x16Param_sse2));
-#endif  // HAVE_SSE2
-
-#if HAVE_AVX2 && !CONFIG_DAALA_DCT16
-const Ht16x16Param kArrayHt16x16Param_avx2[] = {
-  make_tuple(&av1_fht16x16_avx2, &av1_iht16x16_256_add_avx2, DCT_DCT,
-             AOM_BITS_8, 256),
-  make_tuple(&av1_fht16x16_avx2, &av1_iht16x16_256_add_avx2, ADST_DCT,
-             AOM_BITS_8, 256),
-  make_tuple(&av1_fht16x16_avx2, &av1_iht16x16_256_add_avx2, DCT_ADST,
-             AOM_BITS_8, 256),
-  make_tuple(&av1_fht16x16_avx2, &av1_iht16x16_256_add_avx2, ADST_ADST,
-             AOM_BITS_8, 256),
-#if CONFIG_EXT_TX
-  make_tuple(&av1_fht16x16_avx2, &av1_iht16x16_256_add_avx2, FLIPADST_DCT,
-             AOM_BITS_8, 256),
-  make_tuple(&av1_fht16x16_avx2, &av1_iht16x16_256_add_avx2, DCT_FLIPADST,
-             AOM_BITS_8, 256),
-  make_tuple(&av1_fht16x16_avx2, &av1_iht16x16_256_add_avx2, FLIPADST_FLIPADST,
-             AOM_BITS_8, 256),
-  make_tuple(&av1_fht16x16_avx2, &av1_iht16x16_256_add_avx2, ADST_FLIPADST,
-             AOM_BITS_8, 256),
-  make_tuple(&av1_fht16x16_avx2, &av1_iht16x16_256_add_avx2, FLIPADST_ADST,
-             AOM_BITS_8, 256),
-  make_tuple(&av1_fht16x16_avx2, &av1_iht16x16_256_add_avx2, IDTX, AOM_BITS_8,
-             256),
-  make_tuple(&av1_fht16x16_avx2, &av1_iht16x16_256_add_avx2, V_DCT, AOM_BITS_8,
-             256),
-  make_tuple(&av1_fht16x16_avx2, &av1_iht16x16_256_add_avx2, H_DCT, AOM_BITS_8,
-             256),
-  make_tuple(&av1_fht16x16_avx2, &av1_iht16x16_256_add_avx2, V_ADST, AOM_BITS_8,
-             256),
-  make_tuple(&av1_fht16x16_avx2, &av1_iht16x16_256_add_avx2, H_ADST, AOM_BITS_8,
-             256),
-  make_tuple(&av1_fht16x16_avx2, &av1_iht16x16_256_add_avx2, V_FLIPADST,
-             AOM_BITS_8, 256),
-  make_tuple(&av1_fht16x16_avx2, &av1_iht16x16_256_add_avx2, H_FLIPADST,
-             AOM_BITS_8, 256)
-#endif  // CONFIG_EXT_TX
-};
-INSTANTIATE_TEST_CASE_P(AVX2, AV1Trans16x16HT,
-                        ::testing::ValuesIn(kArrayHt16x16Param_avx2));
-#endif  // HAVE_AVX2
-
-#if HAVE_SSE4_1 && CONFIG_HIGHBITDEPTH && !CONFIG_DAALA_DCT16
-const HighbdHt16x16Param kArrayHBDHt16x16Param_sse4_1[] = {
-  make_tuple(&av1_fwd_txfm2d_16x16_sse4_1, DCT_DCT, 10),
-  make_tuple(&av1_fwd_txfm2d_16x16_sse4_1, DCT_DCT, 12),
-  make_tuple(&av1_fwd_txfm2d_16x16_sse4_1, ADST_DCT, 10),
-  make_tuple(&av1_fwd_txfm2d_16x16_sse4_1, ADST_DCT, 12),
-  make_tuple(&av1_fwd_txfm2d_16x16_sse4_1, DCT_ADST, 10),
-  make_tuple(&av1_fwd_txfm2d_16x16_sse4_1, DCT_ADST, 12),
-  make_tuple(&av1_fwd_txfm2d_16x16_sse4_1, ADST_ADST, 10),
-  make_tuple(&av1_fwd_txfm2d_16x16_sse4_1, ADST_ADST, 12),
-#if CONFIG_EXT_TX
-  make_tuple(&av1_fwd_txfm2d_16x16_sse4_1, FLIPADST_DCT, 10),
-  make_tuple(&av1_fwd_txfm2d_16x16_sse4_1, FLIPADST_DCT, 12),
-  make_tuple(&av1_fwd_txfm2d_16x16_sse4_1, DCT_FLIPADST, 10),
-  make_tuple(&av1_fwd_txfm2d_16x16_sse4_1, DCT_FLIPADST, 12),
-  make_tuple(&av1_fwd_txfm2d_16x16_sse4_1, FLIPADST_FLIPADST, 10),
-  make_tuple(&av1_fwd_txfm2d_16x16_sse4_1, FLIPADST_FLIPADST, 12),
-  make_tuple(&av1_fwd_txfm2d_16x16_sse4_1, ADST_FLIPADST, 10),
-  make_tuple(&av1_fwd_txfm2d_16x16_sse4_1, ADST_FLIPADST, 12),
-  make_tuple(&av1_fwd_txfm2d_16x16_sse4_1, FLIPADST_ADST, 10),
-  make_tuple(&av1_fwd_txfm2d_16x16_sse4_1, FLIPADST_ADST, 12),
-#endif  // CONFIG_EXT_TX
-};
-INSTANTIATE_TEST_CASE_P(SSE4_1, AV1HighbdTrans16x16HT,
-                        ::testing::ValuesIn(kArrayHBDHt16x16Param_sse4_1));
-#endif  // HAVE_SSE4_1 && CONFIG_HIGHBITDEPTH && !CONFIG_DAALA_DCT16
-
-}  // namespace

third_party/aom/test/av1_fht16x32_test.cc

@@ -1,157 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
-
-#include "./aom_dsp_rtcd.h"
-#include "./av1_rtcd.h"
-
-#include "aom_ports/mem.h"
-#include "test/acm_random.h"
-#include "test/clear_system_state.h"
-#include "test/register_state_check.h"
-#include "test/transform_test_base.h"
-#include "test/util.h"
-
-using libaom_test::ACMRandom;
-
-namespace {
-typedef void (*IhtFunc)(const tran_low_t *in, uint8_t *out, int stride,
-                        const TxfmParam *txfm_param);
-using std::tr1::tuple;
-using libaom_test::FhtFunc;
-typedef tuple<FhtFunc, IhtFunc, TX_TYPE, aom_bit_depth_t, int> Ht16x32Param;
-
-void fht16x32_ref(const int16_t *in, tran_low_t *out, int stride,
-                  TxfmParam *txfm_param) {
-  av1_fht16x32_c(in, out, stride, txfm_param);
-}
-
-void iht16x32_ref(const tran_low_t *in, uint8_t *out, int stride,
-                  const TxfmParam *txfm_param) {
-  av1_iht16x32_512_add_c(in, out, stride, txfm_param);
-}
-
-class AV1Trans16x32HT : public libaom_test::TransformTestBase,
-                        public ::testing::TestWithParam<Ht16x32Param> {
- public:
-  virtual ~AV1Trans16x32HT() {}
-
-  virtual void SetUp() {
-    fwd_txfm_ = GET_PARAM(0);
-    inv_txfm_ = GET_PARAM(1);
-    pitch_ = 16;
-    height_ = 32;
-    fwd_txfm_ref = fht16x32_ref;
-    inv_txfm_ref = iht16x32_ref;
-    bit_depth_ = GET_PARAM(3);
-    mask_ = (1 << bit_depth_) - 1;
-    num_coeffs_ = GET_PARAM(4);
-    txfm_param_.tx_type = GET_PARAM(2);
-  }
-  virtual void TearDown() { libaom_test::ClearSystemState(); }
-
- protected:
-  void RunFwdTxfm(const int16_t *in, tran_low_t *out, int stride) {
-    fwd_txfm_(in, out, stride, &txfm_param_);
-  }
-
-  void RunInvTxfm(const tran_low_t *out, uint8_t *dst, int stride) {
-    inv_txfm_(out, dst, stride, &txfm_param_);
-  }
-
-  FhtFunc fwd_txfm_;
-  IhtFunc inv_txfm_;
-};
-
-TEST_P(AV1Trans16x32HT, AccuracyCheck) { RunAccuracyCheck(4, 0.2); }
-TEST_P(AV1Trans16x32HT, CoeffCheck) { RunCoeffCheck(); }
-TEST_P(AV1Trans16x32HT, MemCheck) { RunMemCheck(); }
-TEST_P(AV1Trans16x32HT, InvCoeffCheck) { RunInvCoeffCheck(); }
-TEST_P(AV1Trans16x32HT, InvAccuracyCheck) { RunInvAccuracyCheck(4); }
-
-using std::tr1::make_tuple;
-const Ht16x32Param kArrayHt16x32Param_c[] = {
-  make_tuple(&av1_fht16x32_c, &av1_iht16x32_512_add_c, DCT_DCT, AOM_BITS_8,
-             512),
-  make_tuple(&av1_fht16x32_c, &av1_iht16x32_512_add_c, ADST_DCT, AOM_BITS_8,
-             512),
-  make_tuple(&av1_fht16x32_c, &av1_iht16x32_512_add_c, DCT_ADST, AOM_BITS_8,
-             512),
-  make_tuple(&av1_fht16x32_c, &av1_iht16x32_512_add_c, ADST_ADST, AOM_BITS_8,
-             512),
-#if CONFIG_EXT_TX
-  make_tuple(&av1_fht16x32_c, &av1_iht16x32_512_add_c, FLIPADST_DCT, AOM_BITS_8,
-             512),
-  make_tuple(&av1_fht16x32_c, &av1_iht16x32_512_add_c, DCT_FLIPADST, AOM_BITS_8,
-             512),
-  make_tuple(&av1_fht16x32_c, &av1_iht16x32_512_add_c, FLIPADST_FLIPADST,
-             AOM_BITS_8, 512),
-  make_tuple(&av1_fht16x32_c, &av1_iht16x32_512_add_c, ADST_FLIPADST,
-             AOM_BITS_8, 512),
-  make_tuple(&av1_fht16x32_c, &av1_iht16x32_512_add_c, FLIPADST_ADST,
-             AOM_BITS_8, 512),
-  make_tuple(&av1_fht16x32_c, &av1_iht16x32_512_add_c, IDTX, AOM_BITS_8, 512),
-  make_tuple(&av1_fht16x32_c, &av1_iht16x32_512_add_c, V_DCT, AOM_BITS_8, 512),
-  make_tuple(&av1_fht16x32_c, &av1_iht16x32_512_add_c, H_DCT, AOM_BITS_8, 512),
-  make_tuple(&av1_fht16x32_c, &av1_iht16x32_512_add_c, V_ADST, AOM_BITS_8, 512),
-  make_tuple(&av1_fht16x32_c, &av1_iht16x32_512_add_c, H_ADST, AOM_BITS_8, 512),
-  make_tuple(&av1_fht16x32_c, &av1_iht16x32_512_add_c, V_FLIPADST, AOM_BITS_8,
-             512),
-  make_tuple(&av1_fht16x32_c, &av1_iht16x32_512_add_c, H_FLIPADST, AOM_BITS_8,
-             512)
-#endif  // CONFIG_EXT_TX
-};
-INSTANTIATE_TEST_CASE_P(C, AV1Trans16x32HT,
-                        ::testing::ValuesIn(kArrayHt16x32Param_c));
-
-#if HAVE_SSE2
-const Ht16x32Param kArrayHt16x32Param_sse2[] = {
-  make_tuple(&av1_fht16x32_sse2, &av1_iht16x32_512_add_sse2, DCT_DCT,
-             AOM_BITS_8, 512),
-  make_tuple(&av1_fht16x32_sse2, &av1_iht16x32_512_add_sse2, ADST_DCT,
-             AOM_BITS_8, 512),
-  make_tuple(&av1_fht16x32_sse2, &av1_iht16x32_512_add_sse2, DCT_ADST,
-             AOM_BITS_8, 512),
-  make_tuple(&av1_fht16x32_sse2, &av1_iht16x32_512_add_sse2, ADST_ADST,
-             AOM_BITS_8, 512),
-#if CONFIG_EXT_TX
-  make_tuple(&av1_fht16x32_sse2, &av1_iht16x32_512_add_sse2, FLIPADST_DCT,
-             AOM_BITS_8, 512),
-  make_tuple(&av1_fht16x32_sse2, &av1_iht16x32_512_add_sse2, DCT_FLIPADST,
-             AOM_BITS_8, 512),
-  make_tuple(&av1_fht16x32_sse2, &av1_iht16x32_512_add_sse2, FLIPADST_FLIPADST,
-             AOM_BITS_8, 512),
-  make_tuple(&av1_fht16x32_sse2, &av1_iht16x32_512_add_sse2, ADST_FLIPADST,
-             AOM_BITS_8, 512),
-  make_tuple(&av1_fht16x32_sse2, &av1_iht16x32_512_add_sse2, FLIPADST_ADST,
-             AOM_BITS_8, 512),
-  make_tuple(&av1_fht16x32_sse2, &av1_iht16x32_512_add_sse2, IDTX, AOM_BITS_8,
-             512),
-  make_tuple(&av1_fht16x32_sse2, &av1_iht16x32_512_add_sse2, V_DCT, AOM_BITS_8,
-             512),
-  make_tuple(&av1_fht16x32_sse2, &av1_iht16x32_512_add_sse2, H_DCT, AOM_BITS_8,
-             512),
-  make_tuple(&av1_fht16x32_sse2, &av1_iht16x32_512_add_sse2, V_ADST, AOM_BITS_8,
-             512),
-  make_tuple(&av1_fht16x32_sse2, &av1_iht16x32_512_add_sse2, H_ADST, AOM_BITS_8,
-             512),
-  make_tuple(&av1_fht16x32_sse2, &av1_iht16x32_512_add_sse2, V_FLIPADST,
-             AOM_BITS_8, 512),
-  make_tuple(&av1_fht16x32_sse2, &av1_iht16x32_512_add_sse2, H_FLIPADST,
-             AOM_BITS_8, 512)
-#endif  // CONFIG_EXT_TX
-};
-INSTANTIATE_TEST_CASE_P(SSE2, AV1Trans16x32HT,
-                        ::testing::ValuesIn(kArrayHt16x32Param_sse2));
-#endif  // HAVE_SSE2
-
-}  // namespace

third_party/aom/test/av1_fht16x8_test.cc

@@ -1,155 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
-
-#include "./aom_dsp_rtcd.h"
-#include "./av1_rtcd.h"
-
-#include "aom_ports/mem.h"
-#include "test/acm_random.h"
-#include "test/clear_system_state.h"
-#include "test/register_state_check.h"
-#include "test/transform_test_base.h"
-#include "test/util.h"
-
-using libaom_test::ACMRandom;
-
-namespace {
-typedef void (*IhtFunc)(const tran_low_t *in, uint8_t *out, int stride,
-                        const TxfmParam *txfm_param);
-using std::tr1::tuple;
-using libaom_test::FhtFunc;
-typedef tuple<FhtFunc, IhtFunc, TX_TYPE, aom_bit_depth_t, int> Ht16x8Param;
-
-void fht16x8_ref(const int16_t *in, tran_low_t *out, int stride,
-                 TxfmParam *txfm_param) {
-  av1_fht16x8_c(in, out, stride, txfm_param);
-}
-
-void iht16x8_ref(const tran_low_t *in, uint8_t *out, int stride,
-                 const TxfmParam *txfm_param) {
-  av1_iht16x8_128_add_c(in, out, stride, txfm_param);
-}
-
-class AV1Trans16x8HT : public libaom_test::TransformTestBase,
-                       public ::testing::TestWithParam<Ht16x8Param> {
- public:
-  virtual ~AV1Trans16x8HT() {}
-
-  virtual void SetUp() {
-    fwd_txfm_ = GET_PARAM(0);
-    inv_txfm_ = GET_PARAM(1);
-    pitch_ = 16;
-    height_ = 8;
-    inv_txfm_ref = iht16x8_ref;
-    fwd_txfm_ref = fht16x8_ref;
-    bit_depth_ = GET_PARAM(3);
-    mask_ = (1 << bit_depth_) - 1;
-    num_coeffs_ = GET_PARAM(4);
-    txfm_param_.tx_type = GET_PARAM(2);
-  }
-  virtual void TearDown() { libaom_test::ClearSystemState(); }
-
- protected:
-  void RunFwdTxfm(const int16_t *in, tran_low_t *out, int stride) {
-    fwd_txfm_(in, out, stride, &txfm_param_);
-  }
-
-  void RunInvTxfm(const tran_low_t *out, uint8_t *dst, int stride) {
-    inv_txfm_(out, dst, stride, &txfm_param_);
-  }
-
-  FhtFunc fwd_txfm_;
-  IhtFunc inv_txfm_;
-};
-
-TEST_P(AV1Trans16x8HT, AccuracyCheck) { RunAccuracyCheck(1, 0.001); }
-TEST_P(AV1Trans16x8HT, CoeffCheck) { RunCoeffCheck(); }
-TEST_P(AV1Trans16x8HT, MemCheck) { RunMemCheck(); }
-TEST_P(AV1Trans16x8HT, InvCoeffCheck) { RunInvCoeffCheck(); }
-TEST_P(AV1Trans16x8HT, InvAccuracyCheck) { RunInvAccuracyCheck(1); }
-
-using std::tr1::make_tuple;
-
-const Ht16x8Param kArrayHt16x8Param_c[] = {
-  make_tuple(&av1_fht16x8_c, &av1_iht16x8_128_add_c, DCT_DCT, AOM_BITS_8, 128),
-  make_tuple(&av1_fht16x8_c, &av1_iht16x8_128_add_c, ADST_DCT, AOM_BITS_8, 128),
-  make_tuple(&av1_fht16x8_c, &av1_iht16x8_128_add_c, DCT_ADST, AOM_BITS_8, 128),
-  make_tuple(&av1_fht16x8_c, &av1_iht16x8_128_add_c, ADST_ADST, AOM_BITS_8,
-             128),
-#if CONFIG_EXT_TX
-  make_tuple(&av1_fht16x8_c, &av1_iht16x8_128_add_c, FLIPADST_DCT, AOM_BITS_8,
-             128),
-  make_tuple(&av1_fht16x8_c, &av1_iht16x8_128_add_c, DCT_FLIPADST, AOM_BITS_8,
-             128),
-  make_tuple(&av1_fht16x8_c, &av1_iht16x8_128_add_c, FLIPADST_FLIPADST,
-             AOM_BITS_8, 128),
-  make_tuple(&av1_fht16x8_c, &av1_iht16x8_128_add_c, ADST_FLIPADST, AOM_BITS_8,
-             128),
-  make_tuple(&av1_fht16x8_c, &av1_iht16x8_128_add_c, FLIPADST_ADST, AOM_BITS_8,
-             128),
-  make_tuple(&av1_fht16x8_c, &av1_iht16x8_128_add_c, IDTX, AOM_BITS_8, 128),
-  make_tuple(&av1_fht16x8_c, &av1_iht16x8_128_add_c, V_DCT, AOM_BITS_8, 128),
-  make_tuple(&av1_fht16x8_c, &av1_iht16x8_128_add_c, H_DCT, AOM_BITS_8, 128),
-  make_tuple(&av1_fht16x8_c, &av1_iht16x8_128_add_c, V_ADST, AOM_BITS_8, 128),
-  make_tuple(&av1_fht16x8_c, &av1_iht16x8_128_add_c, H_ADST, AOM_BITS_8, 128),
-  make_tuple(&av1_fht16x8_c, &av1_iht16x8_128_add_c, V_FLIPADST, AOM_BITS_8,
-             128),
-  make_tuple(&av1_fht16x8_c, &av1_iht16x8_128_add_c, H_FLIPADST, AOM_BITS_8,
-             128)
-#endif  // CONFIG_EXT_TX
-};
-INSTANTIATE_TEST_CASE_P(C, AV1Trans16x8HT,
-                        ::testing::ValuesIn(kArrayHt16x8Param_c));
-
-#if HAVE_SSE2
-const Ht16x8Param kArrayHt16x8Param_sse2[] = {
-  make_tuple(&av1_fht16x8_sse2, &av1_iht16x8_128_add_sse2, DCT_DCT, AOM_BITS_8,
-             128),
-  make_tuple(&av1_fht16x8_sse2, &av1_iht16x8_128_add_sse2, ADST_DCT, AOM_BITS_8,
-             128),
-  make_tuple(&av1_fht16x8_sse2, &av1_iht16x8_128_add_sse2, DCT_ADST, AOM_BITS_8,
-             128),
-  make_tuple(&av1_fht16x8_sse2, &av1_iht16x8_128_add_sse2, ADST_ADST,
-             AOM_BITS_8, 128),
-#if CONFIG_EXT_TX
-  make_tuple(&av1_fht16x8_sse2, &av1_iht16x8_128_add_sse2, FLIPADST_DCT,
-             AOM_BITS_8, 128),
-  make_tuple(&av1_fht16x8_sse2, &av1_iht16x8_128_add_sse2, DCT_FLIPADST,
-             AOM_BITS_8, 128),
-  make_tuple(&av1_fht16x8_sse2, &av1_iht16x8_128_add_sse2, FLIPADST_FLIPADST,
-             AOM_BITS_8, 128),
-  make_tuple(&av1_fht16x8_sse2, &av1_iht16x8_128_add_sse2, ADST_FLIPADST,
-             AOM_BITS_8, 128),
-  make_tuple(&av1_fht16x8_sse2, &av1_iht16x8_128_add_sse2, FLIPADST_ADST,
-             AOM_BITS_8, 128),
-  make_tuple(&av1_fht16x8_sse2, &av1_iht16x8_128_add_sse2, IDTX, AOM_BITS_8,
-             128),
-  make_tuple(&av1_fht16x8_sse2, &av1_iht16x8_128_add_sse2, V_DCT, AOM_BITS_8,
-             128),
-  make_tuple(&av1_fht16x8_sse2, &av1_iht16x8_128_add_sse2, H_DCT, AOM_BITS_8,
-             128),
-  make_tuple(&av1_fht16x8_sse2, &av1_iht16x8_128_add_sse2, V_ADST, AOM_BITS_8,
-             128),
-  make_tuple(&av1_fht16x8_sse2, &av1_iht16x8_128_add_sse2, H_ADST, AOM_BITS_8,
-             128),
-  make_tuple(&av1_fht16x8_sse2, &av1_iht16x8_128_add_sse2, V_FLIPADST,
-             AOM_BITS_8, 128),
-  make_tuple(&av1_fht16x8_sse2, &av1_iht16x8_128_add_sse2, H_FLIPADST,
-             AOM_BITS_8, 128)
-#endif  // CONFIG_EXT_TX
-};
-INSTANTIATE_TEST_CASE_P(SSE2, AV1Trans16x8HT,
-                        ::testing::ValuesIn(kArrayHt16x8Param_sse2));
-#endif  // HAVE_SSE2
-
-}  // namespace

third_party/aom/test/av1_fht32x16_test.cc

@@ -1,157 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
-
-#include "./aom_dsp_rtcd.h"
-#include "./av1_rtcd.h"
-
-#include "aom_ports/mem.h"
-#include "test/acm_random.h"
-#include "test/clear_system_state.h"
-#include "test/register_state_check.h"
-#include "test/transform_test_base.h"
-#include "test/util.h"
-
-using libaom_test::ACMRandom;
-
-namespace {
-typedef void (*IhtFunc)(const tran_low_t *in, uint8_t *out, int stride,
-                        const TxfmParam *txfm_param);
-using std::tr1::tuple;
-using libaom_test::FhtFunc;
-typedef tuple<FhtFunc, IhtFunc, TX_TYPE, aom_bit_depth_t, int> Ht32x16Param;
-
-void fht32x16_ref(const int16_t *in, tran_low_t *out, int stride,
-                  TxfmParam *txfm_param) {
-  av1_fht32x16_c(in, out, stride, txfm_param);
-}
-
-void iht32x16_ref(const tran_low_t *in, uint8_t *out, int stride,
-                  const TxfmParam *txfm_param) {
-  av1_iht32x16_512_add_c(in, out, stride, txfm_param);
-}
-
-class AV1Trans32x16HT : public libaom_test::TransformTestBase,
-                        public ::testing::TestWithParam<Ht32x16Param> {
- public:
-  virtual ~AV1Trans32x16HT() {}
-
-  virtual void SetUp() {
-    fwd_txfm_ = GET_PARAM(0);
-    inv_txfm_ = GET_PARAM(1);
-    pitch_ = 32;
-    height_ = 16;
-    fwd_txfm_ref = fht32x16_ref;
-    inv_txfm_ref = iht32x16_ref;
-    bit_depth_ = GET_PARAM(3);
-    mask_ = (1 << bit_depth_) - 1;
-    num_coeffs_ = GET_PARAM(4);
-    txfm_param_.tx_type = GET_PARAM(2);
-  }
-  virtual void TearDown() { libaom_test::ClearSystemState(); }
-
- protected:
-  void RunFwdTxfm(const int16_t *in, tran_low_t *out, int stride) {
-    fwd_txfm_(in, out, stride, &txfm_param_);
-  }
-
-  void RunInvTxfm(const tran_low_t *out, uint8_t *dst, int stride) {
-    inv_txfm_(out, dst, stride, &txfm_param_);
-  }
-
-  FhtFunc fwd_txfm_;
-  IhtFunc inv_txfm_;
-};
-
-TEST_P(AV1Trans32x16HT, MemCheck) { RunMemCheck(); }
-TEST_P(AV1Trans32x16HT, AccuracyCheck) { RunAccuracyCheck(4, 0.2); }
-TEST_P(AV1Trans32x16HT, CoeffCheck) { RunCoeffCheck(); }
-TEST_P(AV1Trans32x16HT, InvCoeffCheck) { RunInvCoeffCheck(); }
-TEST_P(AV1Trans32x16HT, InvAccuracyCheck) { RunInvAccuracyCheck(4); }
-
-using std::tr1::make_tuple;
-const Ht32x16Param kArrayHt32x16Param_c[] = {
-  make_tuple(&av1_fht32x16_c, &av1_iht32x16_512_add_c, DCT_DCT, AOM_BITS_8,
-             512),
-  make_tuple(&av1_fht32x16_c, &av1_iht32x16_512_add_c, ADST_DCT, AOM_BITS_8,
-             512),
-  make_tuple(&av1_fht32x16_c, &av1_iht32x16_512_add_c, DCT_ADST, AOM_BITS_8,
-             512),
-  make_tuple(&av1_fht32x16_c, &av1_iht32x16_512_add_c, ADST_ADST, AOM_BITS_8,
-             512),
-#if CONFIG_EXT_TX
-  make_tuple(&av1_fht32x16_c, &av1_iht32x16_512_add_c, FLIPADST_DCT, AOM_BITS_8,
-             512),
-  make_tuple(&av1_fht32x16_c, &av1_iht32x16_512_add_c, DCT_FLIPADST, AOM_BITS_8,
-             512),
-  make_tuple(&av1_fht32x16_c, &av1_iht32x16_512_add_c, FLIPADST_FLIPADST,
-             AOM_BITS_8, 512),
-  make_tuple(&av1_fht32x16_c, &av1_iht32x16_512_add_c, ADST_FLIPADST,
-             AOM_BITS_8, 512),
-  make_tuple(&av1_fht32x16_c, &av1_iht32x16_512_add_c, FLIPADST_ADST,
-             AOM_BITS_8, 512),
-  make_tuple(&av1_fht32x16_c, &av1_iht32x16_512_add_c, IDTX, AOM_BITS_8, 512),
-  make_tuple(&av1_fht32x16_c, &av1_iht32x16_512_add_c, V_DCT, AOM_BITS_8, 512),
-  make_tuple(&av1_fht32x16_c, &av1_iht32x16_512_add_c, H_DCT, AOM_BITS_8, 512),
-  make_tuple(&av1_fht32x16_c, &av1_iht32x16_512_add_c, V_ADST, AOM_BITS_8, 512),
-  make_tuple(&av1_fht32x16_c, &av1_iht32x16_512_add_c, H_ADST, AOM_BITS_8, 512),
-  make_tuple(&av1_fht32x16_c, &av1_iht32x16_512_add_c, V_FLIPADST, AOM_BITS_8,
-             512),
-  make_tuple(&av1_fht32x16_c, &av1_iht32x16_512_add_c, H_FLIPADST, AOM_BITS_8,
-             512)
-#endif  // CONFIG_EXT_TX
-};
-INSTANTIATE_TEST_CASE_P(C, AV1Trans32x16HT,
-                        ::testing::ValuesIn(kArrayHt32x16Param_c));
-
-#if HAVE_SSE2
-const Ht32x16Param kArrayHt32x16Param_sse2[] = {
-  make_tuple(&av1_fht32x16_sse2, &av1_iht32x16_512_add_sse2, DCT_DCT,
-             AOM_BITS_8, 512),
-  make_tuple(&av1_fht32x16_sse2, &av1_iht32x16_512_add_sse2, ADST_DCT,
-             AOM_BITS_8, 512),
-  make_tuple(&av1_fht32x16_sse2, &av1_iht32x16_512_add_sse2, DCT_ADST,
-             AOM_BITS_8, 512),
-  make_tuple(&av1_fht32x16_sse2, &av1_iht32x16_512_add_sse2, ADST_ADST,
-             AOM_BITS_8, 512),
-#if CONFIG_EXT_TX
-  make_tuple(&av1_fht32x16_sse2, &av1_iht32x16_512_add_sse2, FLIPADST_DCT,
-             AOM_BITS_8, 512),
-  make_tuple(&av1_fht32x16_sse2, &av1_iht32x16_512_add_sse2, DCT_FLIPADST,
-             AOM_BITS_8, 512),
-  make_tuple(&av1_fht32x16_sse2, &av1_iht32x16_512_add_sse2, FLIPADST_FLIPADST,
-             AOM_BITS_8, 512),
-  make_tuple(&av1_fht32x16_sse2, &av1_iht32x16_512_add_sse2, ADST_FLIPADST,
-             AOM_BITS_8, 512),
-  make_tuple(&av1_fht32x16_sse2, &av1_iht32x16_512_add_sse2, FLIPADST_ADST,
-             AOM_BITS_8, 512),
-  make_tuple(&av1_fht32x16_sse2, &av1_iht32x16_512_add_sse2, IDTX, AOM_BITS_8,
-             512),
-  make_tuple(&av1_fht32x16_sse2, &av1_iht32x16_512_add_sse2, V_DCT, AOM_BITS_8,
-             512),
-  make_tuple(&av1_fht32x16_sse2, &av1_iht32x16_512_add_sse2, H_DCT, AOM_BITS_8,
-             512),
-  make_tuple(&av1_fht32x16_sse2, &av1_iht32x16_512_add_sse2, V_ADST, AOM_BITS_8,
-             512),
-  make_tuple(&av1_fht32x16_sse2, &av1_iht32x16_512_add_sse2, H_ADST, AOM_BITS_8,
-             512),
-  make_tuple(&av1_fht32x16_sse2, &av1_iht32x16_512_add_sse2, V_FLIPADST,
-             AOM_BITS_8, 512),
-  make_tuple(&av1_fht32x16_sse2, &av1_iht32x16_512_add_sse2, H_FLIPADST,
-             AOM_BITS_8, 512)
-#endif  // CONFIG_EXT_TX
-};
-INSTANTIATE_TEST_CASE_P(SSE2, AV1Trans32x16HT,
-                        ::testing::ValuesIn(kArrayHt32x16Param_sse2));
-#endif  // HAVE_SSE2
-
-}  // namespace

third_party/aom/test/av1_fht32x32_test.cc

@@ -1,227 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
-
-#include "./av1_rtcd.h"
-#include "./aom_dsp_rtcd.h"
-
-#include "test/acm_random.h"
-#include "test/clear_system_state.h"
-#include "test/register_state_check.h"
-#include "test/transform_test_base.h"
-#include "test/util.h"
-#include "aom_ports/mem.h"
-
-using libaom_test::ACMRandom;
-
-namespace {
-typedef void (*IhtFunc)(const tran_low_t *in, uint8_t *out, int stride,
-                        const TxfmParam *txfm_param);
-using std::tr1::tuple;
-using libaom_test::FhtFunc;
-typedef tuple<FhtFunc, IhtFunc, TX_TYPE, aom_bit_depth_t, int> Ht32x32Param;
-
-void fht32x32_ref(const int16_t *in, tran_low_t *out, int stride,
-                  TxfmParam *txfm_param) {
-  av1_fht32x32_c(in, out, stride, txfm_param);
-}
-
-#if CONFIG_HIGHBITDEPTH
-typedef void (*IHbdHtFunc)(const tran_low_t *in, uint8_t *out, int stride,
-                           TX_TYPE tx_type, int bd);
-typedef void (*HbdHtFunc)(const int16_t *input, int32_t *output, int stride,
-                          TX_TYPE tx_type, int bd);
-
-// Target optimized function, tx_type, bit depth
-typedef tuple<HbdHtFunc, TX_TYPE, int> HighbdHt32x32Param;
-
-void highbd_fht32x32_ref(const int16_t *in, int32_t *out, int stride,
-                         TX_TYPE tx_type, int bd) {
-  av1_fwd_txfm2d_32x32_c(in, out, stride, tx_type, bd);
-}
-#endif  // CONFIG_HIGHBITDEPTH
-
-#if (HAVE_SSE2 || HAVE_AVX2) && !CONFIG_DAALA_DCT32
-void dummy_inv_txfm(const tran_low_t *in, uint8_t *out, int stride,
-                    const TxfmParam *txfm_param) {
-  (void)in;
-  (void)out;
-  (void)stride;
-  (void)txfm_param;
-}
-#endif
-
-class AV1Trans32x32HT : public libaom_test::TransformTestBase,
-                        public ::testing::TestWithParam<Ht32x32Param> {
- public:
-  virtual ~AV1Trans32x32HT() {}
-
-  virtual void SetUp() {
-    fwd_txfm_ = GET_PARAM(0);
-    inv_txfm_ = GET_PARAM(1);
-    pitch_ = 32;
-    height_ = 32;
-    fwd_txfm_ref = fht32x32_ref;
-    bit_depth_ = GET_PARAM(3);
-    mask_ = (1 << bit_depth_) - 1;
-    num_coeffs_ = GET_PARAM(4);
-    txfm_param_.tx_type = GET_PARAM(2);
-  }
-  virtual void TearDown() { libaom_test::ClearSystemState(); }
-
- protected:
-  void RunFwdTxfm(const int16_t *in, tran_low_t *out, int stride) {
-    fwd_txfm_(in, out, stride, &txfm_param_);
-  }
-
-  void RunInvTxfm(const tran_low_t *out, uint8_t *dst, int stride) {
-    inv_txfm_(out, dst, stride, &txfm_param_);
-  }
-
-  FhtFunc fwd_txfm_;
-  IhtFunc inv_txfm_;
-};
-
-TEST_P(AV1Trans32x32HT, CoeffCheck) { RunCoeffCheck(); }
-TEST_P(AV1Trans32x32HT, MemCheck) { RunMemCheck(); }
-
-#if CONFIG_HIGHBITDEPTH
-class AV1HighbdTrans32x32HT
-    : public ::testing::TestWithParam<HighbdHt32x32Param> {
- public:
-  virtual ~AV1HighbdTrans32x32HT() {}
-
-  virtual void SetUp() {
-    fwd_txfm_ = GET_PARAM(0);
-    fwd_txfm_ref_ = highbd_fht32x32_ref;
-    tx_type_ = GET_PARAM(1);
-    bit_depth_ = GET_PARAM(2);
-    mask_ = (1 << bit_depth_) - 1;
-    num_coeffs_ = 1024;
-
-    input_ = reinterpret_cast<int16_t *>(
-        aom_memalign(32, sizeof(int16_t) * num_coeffs_));
-    output_ = reinterpret_cast<int32_t *>(
-        aom_memalign(32, sizeof(int32_t) * num_coeffs_));
-    output_ref_ = reinterpret_cast<int32_t *>(
-        aom_memalign(32, sizeof(int32_t) * num_coeffs_));
-  }
-
-  virtual void TearDown() {
-    aom_free(input_);
-    aom_free(output_);
-    aom_free(output_ref_);
-    libaom_test::ClearSystemState();
-  }
-
- protected:
-  void RunBitexactCheck();
-
- private:
-  HbdHtFunc fwd_txfm_;
-  HbdHtFunc fwd_txfm_ref_;
-  TX_TYPE tx_type_;
-  int bit_depth_;
-  int mask_;
-  int num_coeffs_;
-  int16_t *input_;
-  int32_t *output_;
-  int32_t *output_ref_;
-};
-
-void AV1HighbdTrans32x32HT::RunBitexactCheck() {
-  ACMRandom rnd(ACMRandom::DeterministicSeed());
-  int i, j;
-  const int stride = 32;
-  const int num_tests = 1000;
-
-  for (i = 0; i < num_tests; ++i) {
-    for (j = 0; j < num_coeffs_; ++j) {
-      input_[j] = (rnd.Rand16() & mask_) - (rnd.Rand16() & mask_);
-    }
-
-    fwd_txfm_ref_(input_, output_ref_, stride, tx_type_, bit_depth_);
-    ASM_REGISTER_STATE_CHECK(
-        fwd_txfm_(input_, output_, stride, tx_type_, bit_depth_));
-
-    for (j = 0; j < num_coeffs_; ++j) {
-      EXPECT_EQ(output_ref_[j], output_[j])
-          << "Not bit-exact result at index: " << j << " at test block: " << i;
-    }
-  }
-}
-
-TEST_P(AV1HighbdTrans32x32HT, HighbdCoeffCheck) { RunBitexactCheck(); }
-#endif  // CONFIG_HIGHBITDEPTH
-
-using std::tr1::make_tuple;
-
-#if HAVE_SSE2 && !CONFIG_DAALA_DCT32
-const Ht32x32Param kArrayHt32x32Param_sse2[] = {
-  make_tuple(&av1_fht32x32_sse2, &dummy_inv_txfm, DCT_DCT, AOM_BITS_8, 1024),
-  make_tuple(&av1_fht32x32_sse2, &dummy_inv_txfm, ADST_DCT, AOM_BITS_8, 1024),
-  make_tuple(&av1_fht32x32_sse2, &dummy_inv_txfm, DCT_ADST, AOM_BITS_8, 1024),
-  make_tuple(&av1_fht32x32_sse2, &dummy_inv_txfm, ADST_ADST, AOM_BITS_8, 1024),
-#if CONFIG_EXT_TX
-  make_tuple(&av1_fht32x32_sse2, &dummy_inv_txfm, FLIPADST_DCT, AOM_BITS_8,
-             1024),
-  make_tuple(&av1_fht32x32_sse2, &dummy_inv_txfm, DCT_FLIPADST, AOM_BITS_8,
-             1024),
-  make_tuple(&av1_fht32x32_sse2, &dummy_inv_txfm, FLIPADST_FLIPADST, AOM_BITS_8,
-             1024),
-  make_tuple(&av1_fht32x32_sse2, &dummy_inv_txfm, ADST_FLIPADST, AOM_BITS_8,
-             1024),
-  make_tuple(&av1_fht32x32_sse2, &dummy_inv_txfm, FLIPADST_ADST, AOM_BITS_8,
-             1024),
-  make_tuple(&av1_fht32x32_sse2, &dummy_inv_txfm, IDTX, AOM_BITS_8, 1024),
-  make_tuple(&av1_fht32x32_sse2, &dummy_inv_txfm, V_DCT, AOM_BITS_8, 1024),
-  make_tuple(&av1_fht32x32_sse2, &dummy_inv_txfm, H_DCT, AOM_BITS_8, 1024),
-  make_tuple(&av1_fht32x32_sse2, &dummy_inv_txfm, V_ADST, AOM_BITS_8, 1024),
-  make_tuple(&av1_fht32x32_sse2, &dummy_inv_txfm, H_ADST, AOM_BITS_8, 1024),
-  make_tuple(&av1_fht32x32_sse2, &dummy_inv_txfm, V_FLIPADST, AOM_BITS_8, 1024),
-  make_tuple(&av1_fht32x32_sse2, &dummy_inv_txfm, H_FLIPADST, AOM_BITS_8, 1024)
-#endif  // CONFIG_EXT_TX
-};
-INSTANTIATE_TEST_CASE_P(SSE2, AV1Trans32x32HT,
-                        ::testing::ValuesIn(kArrayHt32x32Param_sse2));
-#endif  // HAVE_SSE2 && !CONFIG_DAALA_DCT32
-
-#if HAVE_AVX2 && !CONFIG_DAALA_DCT32
-const Ht32x32Param kArrayHt32x32Param_avx2[] = {
-  make_tuple(&av1_fht32x32_avx2, &dummy_inv_txfm, DCT_DCT, AOM_BITS_8, 1024),
-  make_tuple(&av1_fht32x32_avx2, &dummy_inv_txfm, ADST_DCT, AOM_BITS_8, 1024),
-  make_tuple(&av1_fht32x32_avx2, &dummy_inv_txfm, DCT_ADST, AOM_BITS_8, 1024),
-  make_tuple(&av1_fht32x32_avx2, &dummy_inv_txfm, ADST_ADST, AOM_BITS_8, 1024),
-#if CONFIG_EXT_TX
-  make_tuple(&av1_fht32x32_avx2, &dummy_inv_txfm, FLIPADST_DCT, AOM_BITS_8,
-             1024),
-  make_tuple(&av1_fht32x32_avx2, &dummy_inv_txfm, DCT_FLIPADST, AOM_BITS_8,
-             1024),
-  make_tuple(&av1_fht32x32_avx2, &dummy_inv_txfm, FLIPADST_FLIPADST, AOM_BITS_8,
-             1024),
-  make_tuple(&av1_fht32x32_avx2, &dummy_inv_txfm, ADST_FLIPADST, AOM_BITS_8,
-             1024),
-  make_tuple(&av1_fht32x32_avx2, &dummy_inv_txfm, FLIPADST_ADST, AOM_BITS_8,
-             1024),
-  make_tuple(&av1_fht32x32_avx2, &dummy_inv_txfm, IDTX, AOM_BITS_8, 1024),
-  make_tuple(&av1_fht32x32_avx2, &dummy_inv_txfm, V_DCT, AOM_BITS_8, 1024),
-  make_tuple(&av1_fht32x32_avx2, &dummy_inv_txfm, H_DCT, AOM_BITS_8, 1024),
-  make_tuple(&av1_fht32x32_avx2, &dummy_inv_txfm, V_ADST, AOM_BITS_8, 1024),
-  make_tuple(&av1_fht32x32_avx2, &dummy_inv_txfm, H_ADST, AOM_BITS_8, 1024),
-  make_tuple(&av1_fht32x32_avx2, &dummy_inv_txfm, V_FLIPADST, AOM_BITS_8, 1024),
-  make_tuple(&av1_fht32x32_avx2, &dummy_inv_txfm, H_FLIPADST, AOM_BITS_8, 1024)
-#endif  // CONFIG_EXT_TX
-};
-INSTANTIATE_TEST_CASE_P(AVX2, AV1Trans32x32HT,
-                        ::testing::ValuesIn(kArrayHt32x32Param_avx2));
-#endif  // HAVE_AVX2 && !CONFIG_DAALA_DCT32
-}  // namespace

third_party/aom/test/av1_fht4x4_test.cc

@@ -1,235 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
-
-#include "./av1_rtcd.h"
-#include "./aom_dsp_rtcd.h"
-
-#include "test/acm_random.h"
-#include "test/clear_system_state.h"
-#include "test/register_state_check.h"
-#include "test/transform_test_base.h"
-#include "test/util.h"
-#include "aom_ports/mem.h"
-
-using libaom_test::ACMRandom;
-
-namespace {
-typedef void (*IhtFunc)(const tran_low_t *in, uint8_t *out, int stride,
-                        const TxfmParam *txfm_param);
-using std::tr1::tuple;
-using libaom_test::FhtFunc;
-typedef tuple<FhtFunc, IhtFunc, TX_TYPE, aom_bit_depth_t, int> Ht4x4Param;
-
-void fht4x4_ref(const int16_t *in, tran_low_t *out, int stride,
-                TxfmParam *txfm_param) {
-  av1_fht4x4_c(in, out, stride, txfm_param);
-}
-
-void iht4x4_ref(const tran_low_t *in, uint8_t *out, int stride,
-                const TxfmParam *txfm_param) {
-  av1_iht4x4_16_add_c(in, out, stride, txfm_param);
-}
-
-#if CONFIG_HIGHBITDEPTH
-typedef void (*IhighbdHtFunc)(const tran_low_t *in, uint8_t *out, int stride,
-                              TX_TYPE tx_type, int bd);
-typedef void (*HBDFhtFunc)(const int16_t *input, int32_t *output, int stride,
-                           TX_TYPE tx_type, int bd);
-
-// HighbdHt4x4Param argument list:
-// <Target optimized function, tx_type, bit depth>
-typedef tuple<HBDFhtFunc, TX_TYPE, int> HighbdHt4x4Param;
-
-void highbe_fht4x4_ref(const int16_t *in, int32_t *out, int stride,
-                       TX_TYPE tx_type, int bd) {
-  av1_fwd_txfm2d_4x4_c(in, out, stride, tx_type, bd);
-}
-#endif  // CONFIG_HIGHBITDEPTH
-
-class AV1Trans4x4HT : public libaom_test::TransformTestBase,
-                      public ::testing::TestWithParam<Ht4x4Param> {
- public:
-  virtual ~AV1Trans4x4HT() {}
-
-  virtual void SetUp() {
-    fwd_txfm_ = GET_PARAM(0);
-    inv_txfm_ = GET_PARAM(1);
-    pitch_ = 4;
-    height_ = 4;
-    fwd_txfm_ref = fht4x4_ref;
-    inv_txfm_ref = iht4x4_ref;
-    bit_depth_ = GET_PARAM(3);
-    mask_ = (1 << bit_depth_) - 1;
-    num_coeffs_ = GET_PARAM(4);
-    txfm_param_.tx_type = GET_PARAM(2);
-  }
-  virtual void TearDown() { libaom_test::ClearSystemState(); }
-
- protected:
-  void RunFwdTxfm(const int16_t *in, tran_low_t *out, int stride) {
-    fwd_txfm_(in, out, stride, &txfm_param_);
-  }
-
-  void RunInvTxfm(const tran_low_t *out, uint8_t *dst, int stride) {
-    inv_txfm_(out, dst, stride, &txfm_param_);
-  }
-
-  FhtFunc fwd_txfm_;
-  IhtFunc inv_txfm_;
-};
-
-TEST_P(AV1Trans4x4HT, MemCheck) { RunMemCheck(); }
-TEST_P(AV1Trans4x4HT, CoeffCheck) { RunCoeffCheck(); }
-// Note:
-//  TODO(luoyi): Add tx_type, 9-15 for inverse transform.
-//  Need cleanup since same tests may be done in fdct4x4_test.cc
-// TEST_P(AV1Trans4x4HT, AccuracyCheck) { RunAccuracyCheck(0); }
-// TEST_P(AV1Trans4x4HT, InvAccuracyCheck) { RunInvAccuracyCheck(0); }
-// TEST_P(AV1Trans4x4HT, InvCoeffCheck) { RunInvCoeffCheck(); }
-
-#if CONFIG_HIGHBITDEPTH
-class AV1HighbdTrans4x4HT : public ::testing::TestWithParam<HighbdHt4x4Param> {
- public:
-  virtual ~AV1HighbdTrans4x4HT() {}
-
-  virtual void SetUp() {
-    fwd_txfm_ = GET_PARAM(0);
-    fwd_txfm_ref_ = highbe_fht4x4_ref;
-    tx_type_ = GET_PARAM(1);
-    bit_depth_ = GET_PARAM(2);
-    mask_ = (1 << bit_depth_) - 1;
-    num_coeffs_ = 16;
-
-    input_ = reinterpret_cast<int16_t *>(
-        aom_memalign(16, sizeof(int16_t) * num_coeffs_));
-    output_ = reinterpret_cast<int32_t *>(
-        aom_memalign(16, sizeof(int32_t) * num_coeffs_));
-    output_ref_ = reinterpret_cast<int32_t *>(
-        aom_memalign(16, sizeof(int32_t) * num_coeffs_));
-  }
-
-  virtual void TearDown() {
-    aom_free(input_);
-    aom_free(output_);
-    aom_free(output_ref_);
-    libaom_test::ClearSystemState();
-  }
-
- protected:
-  void RunBitexactCheck();
-
- private:
-  HBDFhtFunc fwd_txfm_;
-  HBDFhtFunc fwd_txfm_ref_;
-  TX_TYPE tx_type_;
-  int bit_depth_;
-  int mask_;
-  int num_coeffs_;
-  int16_t *input_;
-  int32_t *output_;
-  int32_t *output_ref_;
-};
-
-void AV1HighbdTrans4x4HT::RunBitexactCheck() {
-  ACMRandom rnd(ACMRandom::DeterministicSeed());
-  int i, j;
-  const int stride = 4;
-  const int num_tests = 1000;
-  const int num_coeffs = 16;
-
-  for (i = 0; i < num_tests; ++i) {
-    for (j = 0; j < num_coeffs; ++j) {
-      input_[j] = (rnd.Rand16() & mask_) - (rnd.Rand16() & mask_);
-    }
-
-    fwd_txfm_ref_(input_, output_ref_, stride, tx_type_, bit_depth_);
-    fwd_txfm_(input_, output_, stride, tx_type_, bit_depth_);
-
-    for (j = 0; j < num_coeffs; ++j) {
-      EXPECT_EQ(output_[j], output_ref_[j])
-          << "Not bit-exact result at index: " << j << " at test block: " << i;
-    }
-  }
-}
-
-TEST_P(AV1HighbdTrans4x4HT, HighbdCoeffCheck) { RunBitexactCheck(); }
-#endif  // CONFIG_HIGHBITDEPTH
-
-using std::tr1::make_tuple;
-
-#if HAVE_SSE2 && !CONFIG_DAALA_DCT4
-const Ht4x4Param kArrayHt4x4Param_sse2[] = {
-  make_tuple(&av1_fht4x4_sse2, &av1_iht4x4_16_add_sse2, DCT_DCT, AOM_BITS_8,
-             16),
-  make_tuple(&av1_fht4x4_sse2, &av1_iht4x4_16_add_sse2, ADST_DCT, AOM_BITS_8,
-             16),
-  make_tuple(&av1_fht4x4_sse2, &av1_iht4x4_16_add_sse2, DCT_ADST, AOM_BITS_8,
-             16),
-  make_tuple(&av1_fht4x4_sse2, &av1_iht4x4_16_add_sse2, ADST_ADST, AOM_BITS_8,
-             16),
-#if CONFIG_EXT_TX
-  make_tuple(&av1_fht4x4_sse2, &av1_iht4x4_16_add_sse2, FLIPADST_DCT,
-             AOM_BITS_8, 16),
-  make_tuple(&av1_fht4x4_sse2, &av1_iht4x4_16_add_sse2, DCT_FLIPADST,
-             AOM_BITS_8, 16),
-  make_tuple(&av1_fht4x4_sse2, &av1_iht4x4_16_add_sse2, FLIPADST_FLIPADST,
-             AOM_BITS_8, 16),
-  make_tuple(&av1_fht4x4_sse2, &av1_iht4x4_16_add_sse2, ADST_FLIPADST,
-             AOM_BITS_8, 16),
-  make_tuple(&av1_fht4x4_sse2, &av1_iht4x4_16_add_sse2, FLIPADST_ADST,
-             AOM_BITS_8, 16),
-  make_tuple(&av1_fht4x4_sse2, &av1_iht4x4_16_add_sse2, IDTX, AOM_BITS_8, 16),
-  make_tuple(&av1_fht4x4_sse2, &av1_iht4x4_16_add_sse2, V_DCT, AOM_BITS_8, 16),
-  make_tuple(&av1_fht4x4_sse2, &av1_iht4x4_16_add_sse2, H_DCT, AOM_BITS_8, 16),
-  make_tuple(&av1_fht4x4_sse2, &av1_iht4x4_16_add_sse2, V_ADST, AOM_BITS_8, 16),
-  make_tuple(&av1_fht4x4_sse2, &av1_iht4x4_16_add_sse2, H_ADST, AOM_BITS_8, 16),
-  make_tuple(&av1_fht4x4_sse2, &av1_iht4x4_16_add_sse2, V_FLIPADST, AOM_BITS_8,
-             16),
-  make_tuple(&av1_fht4x4_sse2, &av1_iht4x4_16_add_sse2, H_FLIPADST, AOM_BITS_8,
-             16)
-#endif  // CONFIG_EXT_TX
-};
-INSTANTIATE_TEST_CASE_P(SSE2, AV1Trans4x4HT,
-                        ::testing::ValuesIn(kArrayHt4x4Param_sse2));
-#endif  // HAVE_SSE2
-
-#if HAVE_SSE4_1 && CONFIG_HIGHBITDEPTH && !CONFIG_DAALA_DCT4
-const HighbdHt4x4Param kArrayHighbdHt4x4Param[] = {
-  make_tuple(&av1_fwd_txfm2d_4x4_sse4_1, DCT_DCT, 10),
-  make_tuple(&av1_fwd_txfm2d_4x4_sse4_1, DCT_DCT, 12),
-  make_tuple(&av1_fwd_txfm2d_4x4_sse4_1, ADST_DCT, 10),
-  make_tuple(&av1_fwd_txfm2d_4x4_sse4_1, ADST_DCT, 12),
-  make_tuple(&av1_fwd_txfm2d_4x4_sse4_1, DCT_ADST, 10),
-  make_tuple(&av1_fwd_txfm2d_4x4_sse4_1, DCT_ADST, 12),
-  make_tuple(&av1_fwd_txfm2d_4x4_sse4_1, ADST_ADST, 10),
-  make_tuple(&av1_fwd_txfm2d_4x4_sse4_1, ADST_ADST, 12),
-#if CONFIG_EXT_TX
-  make_tuple(&av1_fwd_txfm2d_4x4_sse4_1, FLIPADST_DCT, 10),
-  make_tuple(&av1_fwd_txfm2d_4x4_sse4_1, FLIPADST_DCT, 12),
-  make_tuple(&av1_fwd_txfm2d_4x4_sse4_1, DCT_FLIPADST, 10),
-  make_tuple(&av1_fwd_txfm2d_4x4_sse4_1, DCT_FLIPADST, 12),
-  make_tuple(&av1_fwd_txfm2d_4x4_sse4_1, FLIPADST_FLIPADST, 10),
-  make_tuple(&av1_fwd_txfm2d_4x4_sse4_1, FLIPADST_FLIPADST, 12),
-  make_tuple(&av1_fwd_txfm2d_4x4_sse4_1, ADST_FLIPADST, 10),
-  make_tuple(&av1_fwd_txfm2d_4x4_sse4_1, ADST_FLIPADST, 12),
-  make_tuple(&av1_fwd_txfm2d_4x4_sse4_1, FLIPADST_ADST, 10),
-  make_tuple(&av1_fwd_txfm2d_4x4_sse4_1, FLIPADST_ADST, 12),
-#endif  // CONFIG_EXT_TX
-};
-
-INSTANTIATE_TEST_CASE_P(SSE4_1, AV1HighbdTrans4x4HT,
-                        ::testing::ValuesIn(kArrayHighbdHt4x4Param));
-
-#endif  // HAVE_SSE4_1 && CONFIG_HIGHBITDEPTH && !CONFIG_DAALA_DCT4
-
-}  // namespace

third_party/aom/test/av1_fht4x8_test.cc

@@ -1,145 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
-
-#include "./aom_dsp_rtcd.h"
-#include "./av1_rtcd.h"
-
-#include "aom_ports/mem.h"
-#include "test/acm_random.h"
-#include "test/clear_system_state.h"
-#include "test/register_state_check.h"
-#include "test/transform_test_base.h"
-#include "test/util.h"
-
-using libaom_test::ACMRandom;
-
-namespace {
-typedef void (*IhtFunc)(const tran_low_t *in, uint8_t *out, int stride,
-                        const TxfmParam *txfm_param);
-using std::tr1::tuple;
-using libaom_test::FhtFunc;
-typedef tuple<FhtFunc, IhtFunc, TX_TYPE, aom_bit_depth_t, int> Ht4x8Param;
-
-void fht4x8_ref(const int16_t *in, tran_low_t *out, int stride,
-                TxfmParam *txfm_param) {
-  av1_fht4x8_c(in, out, stride, txfm_param);
-}
-
-void iht4x8_ref(const tran_low_t *in, uint8_t *out, int stride,
-                const TxfmParam *txfm_param) {
-  av1_iht4x8_32_add_c(in, out, stride, txfm_param);
-}
-
-class AV1Trans4x8HT : public libaom_test::TransformTestBase,
-                      public ::testing::TestWithParam<Ht4x8Param> {
- public:
-  virtual ~AV1Trans4x8HT() {}
-
-  virtual void SetUp() {
-    fwd_txfm_ = GET_PARAM(0);
-    inv_txfm_ = GET_PARAM(1);
-    pitch_ = 4;
-    height_ = 8;
-    fwd_txfm_ref = fht4x8_ref;
-    inv_txfm_ref = iht4x8_ref;
-    bit_depth_ = GET_PARAM(3);
-    mask_ = (1 << bit_depth_) - 1;
-    num_coeffs_ = GET_PARAM(4);
-    txfm_param_.tx_type = GET_PARAM(2);
-  }
-  virtual void TearDown() { libaom_test::ClearSystemState(); }
-
- protected:
-  void RunFwdTxfm(const int16_t *in, tran_low_t *out, int stride) {
-    fwd_txfm_(in, out, stride, &txfm_param_);
-  }
-
-  void RunInvTxfm(const tran_low_t *out, uint8_t *dst, int stride) {
-    inv_txfm_(out, dst, stride, &txfm_param_);
-  }
-
-  FhtFunc fwd_txfm_;
-  IhtFunc inv_txfm_;
-};
-
-TEST_P(AV1Trans4x8HT, AccuracyCheck) { RunAccuracyCheck(0, 0.00001); }
-TEST_P(AV1Trans4x8HT, CoeffCheck) { RunCoeffCheck(); }
-TEST_P(AV1Trans4x8HT, MemCheck) { RunMemCheck(); }
-TEST_P(AV1Trans4x8HT, InvCoeffCheck) { RunInvCoeffCheck(); }
-TEST_P(AV1Trans4x8HT, InvAccuracyCheck) { RunInvAccuracyCheck(0); }
-
-using std::tr1::make_tuple;
-
-const Ht4x8Param kArrayHt4x8Param_c[] = {
-  make_tuple(&av1_fht4x8_c, &av1_iht4x8_32_add_c, DCT_DCT, AOM_BITS_8, 32),
-  make_tuple(&av1_fht4x8_c, &av1_iht4x8_32_add_c, ADST_DCT, AOM_BITS_8, 32),
-  make_tuple(&av1_fht4x8_c, &av1_iht4x8_32_add_c, DCT_ADST, AOM_BITS_8, 32),
-  make_tuple(&av1_fht4x8_c, &av1_iht4x8_32_add_c, ADST_ADST, AOM_BITS_8, 32),
-#if CONFIG_EXT_TX
-  make_tuple(&av1_fht4x8_c, &av1_iht4x8_32_add_c, FLIPADST_DCT, AOM_BITS_8, 32),
-  make_tuple(&av1_fht4x8_c, &av1_iht4x8_32_add_c, DCT_FLIPADST, AOM_BITS_8, 32),
-  make_tuple(&av1_fht4x8_c, &av1_iht4x8_32_add_c, FLIPADST_FLIPADST, AOM_BITS_8,
-             32),
-  make_tuple(&av1_fht4x8_c, &av1_iht4x8_32_add_c, ADST_FLIPADST, AOM_BITS_8,
-             32),
-  make_tuple(&av1_fht4x8_c, &av1_iht4x8_32_add_c, FLIPADST_ADST, AOM_BITS_8,
-             32),
-  make_tuple(&av1_fht4x8_c, &av1_iht4x8_32_add_c, IDTX, AOM_BITS_8, 32),
-  make_tuple(&av1_fht4x8_c, &av1_iht4x8_32_add_c, V_DCT, AOM_BITS_8, 32),
-  make_tuple(&av1_fht4x8_c, &av1_iht4x8_32_add_c, H_DCT, AOM_BITS_8, 32),
-  make_tuple(&av1_fht4x8_c, &av1_iht4x8_32_add_c, V_ADST, AOM_BITS_8, 32),
-  make_tuple(&av1_fht4x8_c, &av1_iht4x8_32_add_c, H_ADST, AOM_BITS_8, 32),
-  make_tuple(&av1_fht4x8_c, &av1_iht4x8_32_add_c, V_FLIPADST, AOM_BITS_8, 32),
-  make_tuple(&av1_fht4x8_c, &av1_iht4x8_32_add_c, H_FLIPADST, AOM_BITS_8, 32)
-#endif  // CONFIG_EXT_TX
-};
-INSTANTIATE_TEST_CASE_P(C, AV1Trans4x8HT,
-                        ::testing::ValuesIn(kArrayHt4x8Param_c));
-
-#if HAVE_SSE2
-const Ht4x8Param kArrayHt4x8Param_sse2[] = {
-  make_tuple(&av1_fht4x8_sse2, &av1_iht4x8_32_add_sse2, DCT_DCT, AOM_BITS_8,
-             32),
-  make_tuple(&av1_fht4x8_sse2, &av1_iht4x8_32_add_sse2, ADST_DCT, AOM_BITS_8,
-             32),
-  make_tuple(&av1_fht4x8_sse2, &av1_iht4x8_32_add_sse2, DCT_ADST, AOM_BITS_8,
-             32),
-  make_tuple(&av1_fht4x8_sse2, &av1_iht4x8_32_add_sse2, ADST_ADST, AOM_BITS_8,
-             32),
-#if CONFIG_EXT_TX
-  make_tuple(&av1_fht4x8_sse2, &av1_iht4x8_32_add_sse2, FLIPADST_DCT,
-             AOM_BITS_8, 32),
-  make_tuple(&av1_fht4x8_sse2, &av1_iht4x8_32_add_sse2, DCT_FLIPADST,
-             AOM_BITS_8, 32),
-  make_tuple(&av1_fht4x8_sse2, &av1_iht4x8_32_add_sse2, FLIPADST_FLIPADST,
-             AOM_BITS_8, 32),
-  make_tuple(&av1_fht4x8_sse2, &av1_iht4x8_32_add_sse2, ADST_FLIPADST,
-             AOM_BITS_8, 32),
-  make_tuple(&av1_fht4x8_sse2, &av1_iht4x8_32_add_sse2, FLIPADST_ADST,
-             AOM_BITS_8, 32),
-  make_tuple(&av1_fht4x8_sse2, &av1_iht4x8_32_add_sse2, IDTX, AOM_BITS_8, 32),
-  make_tuple(&av1_fht4x8_sse2, &av1_iht4x8_32_add_sse2, V_DCT, AOM_BITS_8, 32),
-  make_tuple(&av1_fht4x8_sse2, &av1_iht4x8_32_add_sse2, H_DCT, AOM_BITS_8, 32),
-  make_tuple(&av1_fht4x8_sse2, &av1_iht4x8_32_add_sse2, V_ADST, AOM_BITS_8, 32),
-  make_tuple(&av1_fht4x8_sse2, &av1_iht4x8_32_add_sse2, H_ADST, AOM_BITS_8, 32),
-  make_tuple(&av1_fht4x8_sse2, &av1_iht4x8_32_add_sse2, V_FLIPADST, AOM_BITS_8,
-             32),
-  make_tuple(&av1_fht4x8_sse2, &av1_iht4x8_32_add_sse2, H_FLIPADST, AOM_BITS_8,
-             32)
-#endif  // CONFIG_EXT_TX
-};
-INSTANTIATE_TEST_CASE_P(SSE2, AV1Trans4x8HT,
-                        ::testing::ValuesIn(kArrayHt4x8Param_sse2));
-#endif  // HAVE_SSE2
-
-}  // namespace

third_party/aom/test/av1_fht64x64_test.cc

@@ -1,124 +0,0 @@
-/*
- * Copyright (c) 2017, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
-
-#include "./av1_rtcd.h"
-
-#include "aom_ports/mem.h"
-#include "test/acm_random.h"
-#include "test/clear_system_state.h"
-#include "test/register_state_check.h"
-#include "test/transform_test_base.h"
-#include "test/util.h"
-
-#if CONFIG_TX64X64
-
-using libaom_test::ACMRandom;
-
-namespace {
-typedef void (*IhtFunc)(const tran_low_t *in, uint8_t *out, int stride,
-                        const TxfmParam *txfm_param);
-using std::tr1::tuple;
-using libaom_test::FhtFunc;
-typedef tuple<FhtFunc, IhtFunc, TX_TYPE, aom_bit_depth_t, int> Ht64x64Param;
-
-void fht64x64_ref(const int16_t *in, tran_low_t *out, int stride,
-                  TxfmParam *txfm_param) {
-  av1_fht64x64_c(in, out, stride, txfm_param);
-}
-
-void iht64x64_ref(const tran_low_t *in, uint8_t *dest, int stride,
-                  const TxfmParam *txfm_param) {
-  av1_iht64x64_4096_add_c(in, dest, stride, txfm_param);
-}
-
-class AV1Trans64x64HT : public libaom_test::TransformTestBase,
-                        public ::testing::TestWithParam<Ht64x64Param> {
- public:
-  virtual ~AV1Trans64x64HT() {}
-
-  virtual void SetUp() {
-    fwd_txfm_ = GET_PARAM(0);
-    inv_txfm_ = GET_PARAM(1);
-    pitch_ = 64;
-    height_ = 64;
-    fwd_txfm_ref = fht64x64_ref;
-    inv_txfm_ref = iht64x64_ref;
-    bit_depth_ = GET_PARAM(3);
-    mask_ = (1 << bit_depth_) - 1;
-    num_coeffs_ = GET_PARAM(4);
-    txfm_param_.tx_type = GET_PARAM(2);
-  }
-  virtual void TearDown() { libaom_test::ClearSystemState(); }
-
- protected:
-  void RunFwdTxfm(const int16_t *in, tran_low_t *out, int stride) {
-    fwd_txfm_(in, out, stride, &txfm_param_);
-  }
-
-  void RunInvTxfm(const tran_low_t *out, uint8_t *dst, int stride) {
-    inv_txfm_(out, dst, stride, &txfm_param_);
-  }
-
-  FhtFunc fwd_txfm_;
-  IhtFunc inv_txfm_;
-};
-
-TEST_P(AV1Trans64x64HT, AccuracyCheck) { RunAccuracyCheck(4, 0.2); }
-TEST_P(AV1Trans64x64HT, CoeffCheck) { RunCoeffCheck(); }
-TEST_P(AV1Trans64x64HT, MemCheck) { RunMemCheck(); }
-TEST_P(AV1Trans64x64HT, InvCoeffCheck) { RunInvCoeffCheck(); }
-TEST_P(AV1Trans64x64HT, InvAccuracyCheck) { RunInvAccuracyCheck(4); }
-
-using std::tr1::make_tuple;
-
-const Ht64x64Param kArrayHt64x64Param_c[] = {
-  make_tuple(&av1_fht64x64_c, &av1_iht64x64_4096_add_c, DCT_DCT, AOM_BITS_8,
-             4096),
-  make_tuple(&av1_fht64x64_c, &av1_iht64x64_4096_add_c, ADST_DCT, AOM_BITS_8,
-             4096),
-  make_tuple(&av1_fht64x64_c, &av1_iht64x64_4096_add_c, DCT_ADST, AOM_BITS_8,
-             4096),
-  make_tuple(&av1_fht64x64_c, &av1_iht64x64_4096_add_c, ADST_ADST, AOM_BITS_8,
-             4096),
-#if CONFIG_EXT_TX
-  make_tuple(&av1_fht64x64_c, &av1_iht64x64_4096_add_c, FLIPADST_DCT,
-             AOM_BITS_8, 4096),
-  make_tuple(&av1_fht64x64_c, &av1_iht64x64_4096_add_c, DCT_FLIPADST,
-             AOM_BITS_8, 4096),
-  make_tuple(&av1_fht64x64_c, &av1_iht64x64_4096_add_c, FLIPADST_FLIPADST,
-             AOM_BITS_8, 4096),
-  make_tuple(&av1_fht64x64_c, &av1_iht64x64_4096_add_c, ADST_FLIPADST,
-             AOM_BITS_8, 4096),
-  make_tuple(&av1_fht64x64_c, &av1_iht64x64_4096_add_c, FLIPADST_ADST,
-             AOM_BITS_8, 4096),
-  make_tuple(&av1_fht64x64_c, &av1_iht64x64_4096_add_c, IDTX, AOM_BITS_8, 4096),
-  make_tuple(&av1_fht64x64_c, &av1_iht64x64_4096_add_c, V_DCT, AOM_BITS_8,
-             4096),
-  make_tuple(&av1_fht64x64_c, &av1_iht64x64_4096_add_c, H_DCT, AOM_BITS_8,
-             4096),
-  make_tuple(&av1_fht64x64_c, &av1_iht64x64_4096_add_c, V_ADST, AOM_BITS_8,
-             4096),
-  make_tuple(&av1_fht64x64_c, &av1_iht64x64_4096_add_c, H_ADST, AOM_BITS_8,
-             4096),
-  make_tuple(&av1_fht64x64_c, &av1_iht64x64_4096_add_c, V_FLIPADST, AOM_BITS_8,
-             4096),
-  make_tuple(&av1_fht64x64_c, &av1_iht64x64_4096_add_c, H_FLIPADST, AOM_BITS_8,
-             4096)
-#endif  // CONFIG_EXT_TX
-};
-INSTANTIATE_TEST_CASE_P(C, AV1Trans64x64HT,
-                        ::testing::ValuesIn(kArrayHt64x64Param_c));
-
-}  // namespace
-
-#endif  // CONFIG_TX64X64

third_party/aom/test/av1_fht8x16_test.cc

@@ -1,154 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
-
-#include "./aom_dsp_rtcd.h"
-#include "./av1_rtcd.h"
-
-#include "aom_ports/mem.h"
-#include "test/acm_random.h"
-#include "test/clear_system_state.h"
-#include "test/register_state_check.h"
-#include "test/transform_test_base.h"
-#include "test/util.h"
-
-using libaom_test::ACMRandom;
-
-namespace {
-typedef void (*IhtFunc)(const tran_low_t *in, uint8_t *out, int stride,
-                        const TxfmParam *txfm_param);
-using std::tr1::tuple;
-using libaom_test::FhtFunc;
-typedef tuple<FhtFunc, IhtFunc, TX_TYPE, aom_bit_depth_t, int> Ht8x16Param;
-
-void fht8x16_ref(const int16_t *in, tran_low_t *out, int stride,
-                 TxfmParam *txfm_param) {
-  av1_fht8x16_c(in, out, stride, txfm_param);
-}
-
-void iht8x16_ref(const tran_low_t *in, uint8_t *out, int stride,
-                 const TxfmParam *txfm_param) {
-  av1_iht8x16_128_add_c(in, out, stride, txfm_param);
-}
-
-class AV1Trans8x16HT : public libaom_test::TransformTestBase,
-                       public ::testing::TestWithParam<Ht8x16Param> {
- public:
-  virtual ~AV1Trans8x16HT() {}
-
-  virtual void SetUp() {
-    fwd_txfm_ = GET_PARAM(0);
-    inv_txfm_ = GET_PARAM(1);
-    pitch_ = 8;
-    height_ = 16;
-    inv_txfm_ref = iht8x16_ref;
-    fwd_txfm_ref = fht8x16_ref;
-    bit_depth_ = GET_PARAM(3);
-    mask_ = (1 << bit_depth_) - 1;
-    num_coeffs_ = GET_PARAM(4);
-    txfm_param_.tx_type = GET_PARAM(2);
-  }
-  virtual void TearDown() { libaom_test::ClearSystemState(); }
-
- protected:
-  void RunFwdTxfm(const int16_t *in, tran_low_t *out, int stride) {
-    fwd_txfm_(in, out, stride, &txfm_param_);
-  }
-
-  void RunInvTxfm(const tran_low_t *out, uint8_t *dst, int stride) {
-    inv_txfm_(out, dst, stride, &txfm_param_);
-  }
-
-  FhtFunc fwd_txfm_;
-  IhtFunc inv_txfm_;
-};
-
-TEST_P(AV1Trans8x16HT, AccuracyCheck) { RunAccuracyCheck(1, 0.001); }
-TEST_P(AV1Trans8x16HT, MemCheck) { RunMemCheck(); }
-TEST_P(AV1Trans8x16HT, CoeffCheck) { RunCoeffCheck(); }
-TEST_P(AV1Trans8x16HT, InvCoeffCheck) { RunInvCoeffCheck(); }
-TEST_P(AV1Trans8x16HT, InvAccuracyCheck) { RunInvAccuracyCheck(1); }
-
-using std::tr1::make_tuple;
-
-const Ht8x16Param kArrayHt8x16Param_c[] = {
-  make_tuple(&av1_fht8x16_c, &av1_iht8x16_128_add_c, DCT_DCT, AOM_BITS_8, 128),
-  make_tuple(&av1_fht8x16_c, &av1_iht8x16_128_add_c, ADST_DCT, AOM_BITS_8, 128),
-  make_tuple(&av1_fht8x16_c, &av1_iht8x16_128_add_c, DCT_ADST, AOM_BITS_8, 128),
-  make_tuple(&av1_fht8x16_c, &av1_iht8x16_128_add_c, ADST_ADST, AOM_BITS_8,
-             128),
-#if CONFIG_EXT_TX
-  make_tuple(&av1_fht8x16_c, &av1_iht8x16_128_add_c, FLIPADST_DCT, AOM_BITS_8,
-             128),
-  make_tuple(&av1_fht8x16_c, &av1_iht8x16_128_add_c, DCT_FLIPADST, AOM_BITS_8,
-             128),
-  make_tuple(&av1_fht8x16_c, &av1_iht8x16_128_add_c, FLIPADST_FLIPADST,
-             AOM_BITS_8, 128),
-  make_tuple(&av1_fht8x16_c, &av1_iht8x16_128_add_c, ADST_FLIPADST, AOM_BITS_8,
-             128),
-  make_tuple(&av1_fht8x16_c, &av1_iht8x16_128_add_c, FLIPADST_ADST, AOM_BITS_8,
-             128),
-  make_tuple(&av1_fht8x16_c, &av1_iht8x16_128_add_c, IDTX, AOM_BITS_8, 128),
-  make_tuple(&av1_fht8x16_c, &av1_iht8x16_128_add_c, V_DCT, AOM_BITS_8, 128),
-  make_tuple(&av1_fht8x16_c, &av1_iht8x16_128_add_c, H_DCT, AOM_BITS_8, 128),
-  make_tuple(&av1_fht8x16_c, &av1_iht8x16_128_add_c, V_ADST, AOM_BITS_8, 128),
-  make_tuple(&av1_fht8x16_c, &av1_iht8x16_128_add_c, H_ADST, AOM_BITS_8, 128),
-  make_tuple(&av1_fht8x16_c, &av1_iht8x16_128_add_c, V_FLIPADST, AOM_BITS_8,
-             128),
-  make_tuple(&av1_fht8x16_c, &av1_iht8x16_128_add_c, H_FLIPADST, AOM_BITS_8,
-             128)
-#endif  // CONFIG_EXT_TX
-};
-INSTANTIATE_TEST_CASE_P(C, AV1Trans8x16HT,
-                        ::testing::ValuesIn(kArrayHt8x16Param_c));
-
-#if HAVE_SSE2
-const Ht8x16Param kArrayHt8x16Param_sse2[] = {
-  make_tuple(&av1_fht8x16_sse2, &av1_iht8x16_128_add_sse2, DCT_DCT, AOM_BITS_8,
-             128),
-  make_tuple(&av1_fht8x16_sse2, &av1_iht8x16_128_add_sse2, ADST_DCT, AOM_BITS_8,
-             128),
-  make_tuple(&av1_fht8x16_sse2, &av1_iht8x16_128_add_sse2, DCT_ADST, AOM_BITS_8,
-             128),
-  make_tuple(&av1_fht8x16_sse2, &av1_iht8x16_128_add_sse2, ADST_ADST,
-             AOM_BITS_8, 128),
-#if CONFIG_EXT_TX
-  make_tuple(&av1_fht8x16_sse2, &av1_iht8x16_128_add_sse2, FLIPADST_DCT,
-             AOM_BITS_8, 128),
-  make_tuple(&av1_fht8x16_sse2, &av1_iht8x16_128_add_sse2, DCT_FLIPADST,
-             AOM_BITS_8, 128),
-  make_tuple(&av1_fht8x16_sse2, &av1_iht8x16_128_add_sse2, FLIPADST_FLIPADST,
-             AOM_BITS_8, 128),
-  make_tuple(&av1_fht8x16_sse2, &av1_iht8x16_128_add_sse2, ADST_FLIPADST,
-             AOM_BITS_8, 128),
-  make_tuple(&av1_fht8x16_sse2, &av1_iht8x16_128_add_sse2, FLIPADST_ADST,
-             AOM_BITS_8, 128),
-  make_tuple(&av1_fht8x16_sse2, &av1_iht8x16_128_add_sse2, IDTX, AOM_BITS_8,
-             128),
-  make_tuple(&av1_fht8x16_sse2, &av1_iht8x16_128_add_sse2, V_DCT, AOM_BITS_8,
-             128),
-  make_tuple(&av1_fht8x16_sse2, &av1_iht8x16_128_add_sse2, H_DCT, AOM_BITS_8,
-             128),
-  make_tuple(&av1_fht8x16_sse2, &av1_iht8x16_128_add_sse2, V_ADST, AOM_BITS_8,
-             128),
-  make_tuple(&av1_fht8x16_sse2, &av1_iht8x16_128_add_sse2, H_ADST, AOM_BITS_8,
-             128),
-  make_tuple(&av1_fht8x16_sse2, &av1_iht8x16_128_add_sse2, V_FLIPADST,
-             AOM_BITS_8, 128),
-  make_tuple(&av1_fht8x16_sse2, &av1_iht8x16_128_add_sse2, H_FLIPADST,
-             AOM_BITS_8, 128)
-#endif  // CONFIG_EXT_TX
-};
-INSTANTIATE_TEST_CASE_P(SSE2, AV1Trans8x16HT,
-                        ::testing::ValuesIn(kArrayHt8x16Param_sse2));
-#endif  // HAVE_SSE2
-
-}  // namespace

third_party/aom/test/av1_fht8x4_test.cc

@@ -1,144 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
-
-#include "./aom_dsp_rtcd.h"
-#include "./av1_rtcd.h"
-
-#include "aom_ports/mem.h"
-#include "test/acm_random.h"
-#include "test/clear_system_state.h"
-#include "test/register_state_check.h"
-#include "test/transform_test_base.h"
-#include "test/util.h"
-
-using libaom_test::ACMRandom;
-
-namespace {
-typedef void (*IhtFunc)(const tran_low_t *in, uint8_t *out, int stride,
-                        const TxfmParam *txfm_param);
-using std::tr1::tuple;
-using libaom_test::FhtFunc;
-typedef tuple<FhtFunc, IhtFunc, TX_TYPE, aom_bit_depth_t, int> Ht8x4Param;
-
-void fht8x4_ref(const int16_t *in, tran_low_t *out, int stride,
-                TxfmParam *txfm_param) {
-  av1_fht8x4_c(in, out, stride, txfm_param);
-}
-
-void iht8x4_ref(const tran_low_t *in, uint8_t *out, int stride,
-                const TxfmParam *txfm_param) {
-  av1_iht8x4_32_add_c(in, out, stride, txfm_param);
-}
-
-class AV1Trans8x4HT : public libaom_test::TransformTestBase,
-                      public ::testing::TestWithParam<Ht8x4Param> {
- public:
-  virtual ~AV1Trans8x4HT() {}
-
-  virtual void SetUp() {
-    fwd_txfm_ = GET_PARAM(0);
-    inv_txfm_ = GET_PARAM(1);
-    pitch_ = 8;
-    height_ = 4;
-    fwd_txfm_ref = fht8x4_ref;
-    inv_txfm_ref = iht8x4_ref;
-    bit_depth_ = GET_PARAM(3);
-    mask_ = (1 << bit_depth_) - 1;
-    num_coeffs_ = GET_PARAM(4);
-    txfm_param_.tx_type = GET_PARAM(2);
-  }
-  virtual void TearDown() { libaom_test::ClearSystemState(); }
-
- protected:
-  void RunFwdTxfm(const int16_t *in, tran_low_t *out, int stride) {
-    fwd_txfm_(in, out, stride, &txfm_param_);
-  }
-
-  void RunInvTxfm(const tran_low_t *out, uint8_t *dst, int stride) {
-    inv_txfm_(out, dst, stride, &txfm_param_);
-  }
-
-  FhtFunc fwd_txfm_;
-  IhtFunc inv_txfm_;
-};
-
-TEST_P(AV1Trans8x4HT, AccuracyCheck) { RunAccuracyCheck(0, 0.00001); }
-TEST_P(AV1Trans8x4HT, CoeffCheck) { RunCoeffCheck(); }
-TEST_P(AV1Trans8x4HT, MemCheck) { RunMemCheck(); }
-TEST_P(AV1Trans8x4HT, InvCoeffCheck) { RunInvCoeffCheck(); }
-TEST_P(AV1Trans8x4HT, InvAccuracyCheck) { RunInvAccuracyCheck(0); }
-
-using std::tr1::make_tuple;
-
-const Ht8x4Param kArrayHt8x4Param_c[] = {
-  make_tuple(&av1_fht8x4_c, &av1_iht8x4_32_add_c, DCT_DCT, AOM_BITS_8, 32),
-  make_tuple(&av1_fht8x4_c, &av1_iht8x4_32_add_c, ADST_DCT, AOM_BITS_8, 32),
-  make_tuple(&av1_fht8x4_c, &av1_iht8x4_32_add_c, DCT_ADST, AOM_BITS_8, 32),
-  make_tuple(&av1_fht8x4_c, &av1_iht8x4_32_add_c, ADST_ADST, AOM_BITS_8, 32),
-#if CONFIG_EXT_TX
-  make_tuple(&av1_fht8x4_c, &av1_iht8x4_32_add_c, FLIPADST_DCT, AOM_BITS_8, 32),
-  make_tuple(&av1_fht8x4_c, &av1_iht8x4_32_add_c, DCT_FLIPADST, AOM_BITS_8, 32),
-  make_tuple(&av1_fht8x4_c, &av1_iht8x4_32_add_c, FLIPADST_FLIPADST, AOM_BITS_8,
-             32),
-  make_tuple(&av1_fht8x4_c, &av1_iht8x4_32_add_c, ADST_FLIPADST, AOM_BITS_8,
-             32),
-  make_tuple(&av1_fht8x4_c, &av1_iht8x4_32_add_c, FLIPADST_ADST, AOM_BITS_8,
-             32),
-  make_tuple(&av1_fht8x4_c, &av1_iht8x4_32_add_c, IDTX, AOM_BITS_8, 32),
-  make_tuple(&av1_fht8x4_c, &av1_iht8x4_32_add_c, V_DCT, AOM_BITS_8, 32),
-  make_tuple(&av1_fht8x4_c, &av1_iht8x4_32_add_c, H_DCT, AOM_BITS_8, 32),
-  make_tuple(&av1_fht8x4_c, &av1_iht8x4_32_add_c, V_ADST, AOM_BITS_8, 32),
-  make_tuple(&av1_fht8x4_c, &av1_iht8x4_32_add_c, H_ADST, AOM_BITS_8, 32),
-  make_tuple(&av1_fht8x4_c, &av1_iht8x4_32_add_c, V_FLIPADST, AOM_BITS_8, 32),
-  make_tuple(&av1_fht8x4_c, &av1_iht8x4_32_add_c, H_FLIPADST, AOM_BITS_8, 32)
-#endif  // CONFIG_EXT_TX
-};
-INSTANTIATE_TEST_CASE_P(C, AV1Trans8x4HT,
-                        ::testing::ValuesIn(kArrayHt8x4Param_c));
-
-#if HAVE_SSE2
-const Ht8x4Param kArrayHt8x4Param_sse2[] = {
-  make_tuple(&av1_fht8x4_sse2, &av1_iht8x4_32_add_sse2, DCT_DCT, AOM_BITS_8,
-             32),
-  make_tuple(&av1_fht8x4_sse2, &av1_iht8x4_32_add_sse2, ADST_DCT, AOM_BITS_8,
-             32),
-  make_tuple(&av1_fht8x4_sse2, &av1_iht8x4_32_add_sse2, DCT_ADST, AOM_BITS_8,
-             32),
-  make_tuple(&av1_fht8x4_sse2, &av1_iht8x4_32_add_sse2, ADST_ADST, AOM_BITS_8,
-             32),
-#if CONFIG_EXT_TX
-  make_tuple(&av1_fht8x4_sse2, &av1_iht8x4_32_add_sse2, FLIPADST_DCT,
-             AOM_BITS_8, 32),
-  make_tuple(&av1_fht8x4_sse2, &av1_iht8x4_32_add_sse2, DCT_FLIPADST,
-             AOM_BITS_8, 32),
-  make_tuple(&av1_fht8x4_sse2, &av1_iht8x4_32_add_sse2, FLIPADST_FLIPADST,
-             AOM_BITS_8, 32),
-  make_tuple(&av1_fht8x4_sse2, &av1_iht8x4_32_add_sse2, ADST_FLIPADST,
-             AOM_BITS_8, 32),
-  make_tuple(&av1_fht8x4_sse2, &av1_iht8x4_32_add_sse2, FLIPADST_ADST,
-             AOM_BITS_8, 32),
-  make_tuple(&av1_fht8x4_sse2, &av1_iht8x4_32_add_sse2, IDTX, AOM_BITS_8, 32),
-  make_tuple(&av1_fht8x4_sse2, &av1_iht8x4_32_add_sse2, V_DCT, AOM_BITS_8, 32),
-  make_tuple(&av1_fht8x4_sse2, &av1_iht8x4_32_add_sse2, H_DCT, AOM_BITS_8, 32),
-  make_tuple(&av1_fht8x4_sse2, &av1_iht8x4_32_add_sse2, V_ADST, AOM_BITS_8, 32),
-  make_tuple(&av1_fht8x4_sse2, &av1_iht8x4_32_add_sse2, H_ADST, AOM_BITS_8, 32),
-  make_tuple(&av1_fht8x4_sse2, &av1_iht8x4_32_add_sse2, V_FLIPADST, AOM_BITS_8,
-             32),
-  make_tuple(&av1_fht8x4_sse2, &av1_iht8x4_32_add_sse2, H_FLIPADST, AOM_BITS_8,
-             32)
-#endif  // CONFIG_EXT_TX
-};
-INSTANTIATE_TEST_CASE_P(SSE2, AV1Trans8x4HT,
-                        ::testing::ValuesIn(kArrayHt8x4Param_sse2));
-#endif  // HAVE_SSE2
-
-}  // namespace

third_party/aom/test/av1_fht8x8_test.cc

@@ -1,233 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
-
-#include "./av1_rtcd.h"
-#include "./aom_dsp_rtcd.h"
-
-#include "test/acm_random.h"
-#include "test/clear_system_state.h"
-#include "test/register_state_check.h"
-#include "test/transform_test_base.h"
-#include "test/util.h"
-#include "aom_ports/mem.h"
-
-using libaom_test::ACMRandom;
-
-namespace {
-typedef void (*IhtFunc)(const tran_low_t *in, uint8_t *out, int stride,
-                        const TxfmParam *txfm_param);
-
-using libaom_test::FhtFunc;
-using std::tr1::tuple;
-typedef tuple<FhtFunc, IhtFunc, TX_TYPE, aom_bit_depth_t, int> Ht8x8Param;
-
-void fht8x8_ref(const int16_t *in, tran_low_t *out, int stride,
-                TxfmParam *txfm_param) {
-  av1_fht8x8_c(in, out, stride, txfm_param);
-}
-
-void iht8x8_ref(const tran_low_t *in, uint8_t *out, int stride,
-                const TxfmParam *txfm_param) {
-  av1_iht8x8_64_add_c(in, out, stride, txfm_param);
-}
-
-#if CONFIG_HIGHBITDEPTH
-typedef void (*IHbdHtFunc)(const tran_low_t *in, uint8_t *out, int stride,
-                           TX_TYPE tx_type, int bd);
-typedef void (*HbdHtFunc)(const int16_t *input, int32_t *output, int stride,
-                          TX_TYPE tx_type, int bd);
-// Target optimized function, tx_type, bit depth
-typedef tuple<HbdHtFunc, TX_TYPE, int> HighbdHt8x8Param;
-
-void highbd_fht8x8_ref(const int16_t *in, int32_t *out, int stride,
-                       TX_TYPE tx_type, int bd) {
-  av1_fwd_txfm2d_8x8_c(in, out, stride, tx_type, bd);
-}
-#endif  // CONFIG_HIGHBITDEPTH
-
-class AV1Trans8x8HT : public libaom_test::TransformTestBase,
-                      public ::testing::TestWithParam<Ht8x8Param> {
- public:
-  virtual ~AV1Trans8x8HT() {}
-
-  virtual void SetUp() {
-    fwd_txfm_ = GET_PARAM(0);
-    inv_txfm_ = GET_PARAM(1);
-    pitch_ = 8;
-    height_ = 8;
-    fwd_txfm_ref = fht8x8_ref;
-    inv_txfm_ref = iht8x8_ref;
-    bit_depth_ = GET_PARAM(3);
-    mask_ = (1 << bit_depth_) - 1;
-    num_coeffs_ = GET_PARAM(4);
-    txfm_param_.tx_type = GET_PARAM(2);
-  }
-  virtual void TearDown() { libaom_test::ClearSystemState(); }
-
- protected:
-  void RunFwdTxfm(const int16_t *in, tran_low_t *out, int stride) {
-    fwd_txfm_(in, out, stride, &txfm_param_);
-  }
-
-  void RunInvTxfm(const tran_low_t *out, uint8_t *dst, int stride) {
-    inv_txfm_(out, dst, stride, &txfm_param_);
-  }
-
-  FhtFunc fwd_txfm_;
-  IhtFunc inv_txfm_;
-};
-
-TEST_P(AV1Trans8x8HT, MemCheck) { RunMemCheck(); }
-TEST_P(AV1Trans8x8HT, CoeffCheck) { RunCoeffCheck(); }
-// Note:
-//  TODO(luoyi): Add tx_type, 9-15 for inverse transform.
-//  Need cleanup since same tests may be done in fdct8x8_test.cc
-// TEST_P(AV1Trans8x8HT, AccuracyCheck) { RunAccuracyCheck(0); }
-// TEST_P(AV1Trans8x8HT, InvAccuracyCheck) { RunInvAccuracyCheck(0); }
-// TEST_P(AV1Trans8x8HT, InvCoeffCheck) { RunInvCoeffCheck(); }
-
-#if CONFIG_HIGHBITDEPTH
-class AV1HighbdTrans8x8HT : public ::testing::TestWithParam<HighbdHt8x8Param> {
- public:
-  virtual ~AV1HighbdTrans8x8HT() {}
-
-  virtual void SetUp() {
-    fwd_txfm_ = GET_PARAM(0);
-    fwd_txfm_ref_ = highbd_fht8x8_ref;
-    tx_type_ = GET_PARAM(1);
-    bit_depth_ = GET_PARAM(2);
-    mask_ = (1 << bit_depth_) - 1;
-    num_coeffs_ = 64;
-
-    input_ = reinterpret_cast<int16_t *>(
-        aom_memalign(16, sizeof(int16_t) * num_coeffs_));
-    output_ = reinterpret_cast<int32_t *>(
-        aom_memalign(16, sizeof(int32_t) * num_coeffs_));
-    output_ref_ = reinterpret_cast<int32_t *>(
-        aom_memalign(16, sizeof(int32_t) * num_coeffs_));
-  }
-
-  virtual void TearDown() {
-    aom_free(input_);
-    aom_free(output_);
-    aom_free(output_ref_);
-    libaom_test::ClearSystemState();
-  }
-
- protected:
-  void RunBitexactCheck();
-
- private:
-  HbdHtFunc fwd_txfm_;
-  HbdHtFunc fwd_txfm_ref_;
-  TX_TYPE tx_type_;
-  int bit_depth_;
-  int mask_;
-  int num_coeffs_;
-  int16_t *input_;
-  int32_t *output_;
-  int32_t *output_ref_;
-};
-
-void AV1HighbdTrans8x8HT::RunBitexactCheck() {
-  ACMRandom rnd(ACMRandom::DeterministicSeed());
-  int i, j;
-  const int stride = 8;
-  const int num_tests = 1000;
-  const int num_coeffs = 64;
-
-  for (i = 0; i < num_tests; ++i) {
-    for (j = 0; j < num_coeffs; ++j) {
-      input_[j] = (rnd.Rand16() & mask_) - (rnd.Rand16() & mask_);
-    }
-
-    fwd_txfm_ref_(input_, output_ref_, stride, tx_type_, bit_depth_);
-    ASM_REGISTER_STATE_CHECK(
-        fwd_txfm_(input_, output_, stride, tx_type_, bit_depth_));
-
-    for (j = 0; j < num_coeffs; ++j) {
-      EXPECT_EQ(output_ref_[j], output_[j])
-          << "Not bit-exact result at index: " << j << " at test block: " << i;
-    }
-  }
-}
-
-TEST_P(AV1HighbdTrans8x8HT, HighbdCoeffCheck) { RunBitexactCheck(); }
-#endif  // CONFIG_HIGHBITDEPTH
-
-using std::tr1::make_tuple;
-
-#if HAVE_SSE2 && !CONFIG_DAALA_DCT8
-const Ht8x8Param kArrayHt8x8Param_sse2[] = {
-  make_tuple(&av1_fht8x8_sse2, &av1_iht8x8_64_add_sse2, DCT_DCT, AOM_BITS_8,
-             64),
-  make_tuple(&av1_fht8x8_sse2, &av1_iht8x8_64_add_sse2, ADST_DCT, AOM_BITS_8,
-             64),
-  make_tuple(&av1_fht8x8_sse2, &av1_iht8x8_64_add_sse2, DCT_ADST, AOM_BITS_8,
-             64),
-  make_tuple(&av1_fht8x8_sse2, &av1_iht8x8_64_add_sse2, ADST_ADST, AOM_BITS_8,
-             64),
-#if CONFIG_EXT_TX
-  make_tuple(&av1_fht8x8_sse2, &av1_iht8x8_64_add_sse2, FLIPADST_DCT,
-             AOM_BITS_8, 64),
-  make_tuple(&av1_fht8x8_sse2, &av1_iht8x8_64_add_sse2, DCT_FLIPADST,
-             AOM_BITS_8, 64),
-  make_tuple(&av1_fht8x8_sse2, &av1_iht8x8_64_add_sse2, FLIPADST_FLIPADST,
-             AOM_BITS_8, 64),
-  make_tuple(&av1_fht8x8_sse2, &av1_iht8x8_64_add_sse2, ADST_FLIPADST,
-             AOM_BITS_8, 64),
-  make_tuple(&av1_fht8x8_sse2, &av1_iht8x8_64_add_sse2, FLIPADST_ADST,
-             AOM_BITS_8, 64),
-  make_tuple(&av1_fht8x8_sse2, &av1_iht8x8_64_add_sse2, IDTX, AOM_BITS_8, 64),
-  make_tuple(&av1_fht8x8_sse2, &av1_iht8x8_64_add_sse2, V_DCT, AOM_BITS_8, 64),
-  make_tuple(&av1_fht8x8_sse2, &av1_iht8x8_64_add_sse2, H_DCT, AOM_BITS_8, 64),
-  make_tuple(&av1_fht8x8_sse2, &av1_iht8x8_64_add_sse2, V_ADST, AOM_BITS_8, 64),
-  make_tuple(&av1_fht8x8_sse2, &av1_iht8x8_64_add_sse2, H_ADST, AOM_BITS_8, 64),
-  make_tuple(&av1_fht8x8_sse2, &av1_iht8x8_64_add_sse2, V_FLIPADST, AOM_BITS_8,
-             64),
-  make_tuple(&av1_fht8x8_sse2, &av1_iht8x8_64_add_sse2, H_FLIPADST, AOM_BITS_8,
-             64)
-#endif  // CONFIG_EXT_TX
-};
-INSTANTIATE_TEST_CASE_P(SSE2, AV1Trans8x8HT,
-                        ::testing::ValuesIn(kArrayHt8x8Param_sse2));
-#endif  // HAVE_SSE2
-
-#if HAVE_SSE4_1 && CONFIG_HIGHBITDEPTH && !CONFIG_DAALA_DCT8
-const HighbdHt8x8Param kArrayHBDHt8x8Param_sse4_1[] = {
-  make_tuple(&av1_fwd_txfm2d_8x8_sse4_1, DCT_DCT, 10),
-  make_tuple(&av1_fwd_txfm2d_8x8_sse4_1, DCT_DCT, 12),
-  make_tuple(&av1_fwd_txfm2d_8x8_sse4_1, ADST_DCT, 10),
-  make_tuple(&av1_fwd_txfm2d_8x8_sse4_1, ADST_DCT, 12),
-  make_tuple(&av1_fwd_txfm2d_8x8_sse4_1, DCT_ADST, 10),
-  make_tuple(&av1_fwd_txfm2d_8x8_sse4_1, DCT_ADST, 12),
-  make_tuple(&av1_fwd_txfm2d_8x8_sse4_1, ADST_ADST, 10),
-  make_tuple(&av1_fwd_txfm2d_8x8_sse4_1, ADST_ADST, 12),
-#if CONFIG_EXT_TX
-  make_tuple(&av1_fwd_txfm2d_8x8_sse4_1, FLIPADST_DCT, 10),
-  make_tuple(&av1_fwd_txfm2d_8x8_sse4_1, FLIPADST_DCT, 12),
-  make_tuple(&av1_fwd_txfm2d_8x8_sse4_1, DCT_FLIPADST, 10),
-  make_tuple(&av1_fwd_txfm2d_8x8_sse4_1, DCT_FLIPADST, 12),
-  make_tuple(&av1_fwd_txfm2d_8x8_sse4_1, FLIPADST_FLIPADST, 10),
-  make_tuple(&av1_fwd_txfm2d_8x8_sse4_1, FLIPADST_FLIPADST, 12),
-  make_tuple(&av1_fwd_txfm2d_8x8_sse4_1, ADST_FLIPADST, 10),
-  make_tuple(&av1_fwd_txfm2d_8x8_sse4_1, ADST_FLIPADST, 12),
-  make_tuple(&av1_fwd_txfm2d_8x8_sse4_1, FLIPADST_ADST, 10),
-  make_tuple(&av1_fwd_txfm2d_8x8_sse4_1, FLIPADST_ADST, 12),
-#endif  // CONFIG_EXT_TX
-};
-INSTANTIATE_TEST_CASE_P(SSE4_1, AV1HighbdTrans8x8HT,
-                        ::testing::ValuesIn(kArrayHBDHt8x8Param_sse4_1));
-#endif  // HAVE_SSE4_1 && CONFIG_HIGHBITDEPTH && !CONFIG_DAALA_DCT8
-
-}  // namespace

third_party/aom/test/av1_fwd_txfm1d_test.cc

@@ -9,36 +9,37 @@
  * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
  */
 
-#include "av1/common/av1_fwd_txfm1d.h"
+#include "av1/encoder/av1_fwd_txfm1d.h"
 #include "test/av1_txfm_test.h"
 
 using libaom_test::ACMRandom;
+using libaom_test::TYPE_ADST;
+using libaom_test::TYPE_DCT;
+using libaom_test::TYPE_IDTX;
+using libaom_test::TYPE_TXFM;
 using libaom_test::input_base;
 using libaom_test::reference_hybrid_1d;
-using libaom_test::TYPE_TXFM;
-using libaom_test::TYPE_DCT;
-using libaom_test::TYPE_ADST;
 
 namespace {
-const int txfm_type_num = 2;
-const TYPE_TXFM txfm_type_ls[2] = { TYPE_DCT, TYPE_ADST };
+const int txfm_type_num = 3;
+const TYPE_TXFM txfm_type_ls[txfm_type_num] = { TYPE_DCT, TYPE_ADST,
+                                                TYPE_IDTX };
 
 const int txfm_size_num = 5;
-const int txfm_size_ls[5] = { 4, 8, 16, 32, 64 };
 
-const TxfmFunc fwd_txfm_func_ls[2][5] = {
-#if CONFIG_TX64X64
-  { av1_fdct4_new, av1_fdct8_new, av1_fdct16_new, av1_fdct32_new,
-    av1_fdct64_new },
-#else
-  { av1_fdct4_new, av1_fdct8_new, av1_fdct16_new, av1_fdct32_new, NULL },
-#endif
-  { av1_fadst4_new, av1_fadst8_new, av1_fadst16_new, av1_fadst32_new, NULL }
+const int txfm_size_ls[] = { 4, 8, 16, 32, 64 };
+
+const TxfmFunc fwd_txfm_func_ls[][txfm_type_num] = {
+  { av1_fdct4_new, av1_fadst4_new, av1_fidentity4_c },
+  { av1_fdct8_new, av1_fadst8_new, av1_fidentity8_c },
+  { av1_fdct16_new, av1_fadst16_new, av1_fidentity16_c },
+  { av1_fdct32_new, NULL, av1_fidentity32_c },
+  { av1_fdct64_new, NULL, NULL },
 };
 
 // the maximum stage number of fwd/inv 1d dct/adst txfm is 12
-const int8_t cos_bit[12] = { 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14 };
-const int8_t range_bit[12] = { 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32 };
+const int8_t cos_bit = 14;
+const int8_t range_bit[12] = { 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20 };
 
 TEST(av1_fwd_txfm1d, round_shift) {
   EXPECT_EQ(round_shift(7, 1), 4);
@@ -51,10 +52,10 @@ TEST(av1_fwd_txfm1d, round_shift) {
   EXPECT_EQ(round_shift(-8, 2), -2);
 }
 
-TEST(av1_fwd_txfm1d, cospi_arr_data) {
+TEST(av1_fwd_txfm1d, av1_cospi_arr_data) {
   for (int i = 0; i < 7; i++) {
     for (int j = 0; j < 64; j++) {
-      EXPECT_EQ(cospi_arr_data[i][j],
+      EXPECT_EQ(av1_cospi_arr_data[i][j],
                 (int32_t)round(cos(M_PI * j / 128) * (1 << (cos_bit_min + i))));
     }
   }
@@ -71,7 +72,7 @@ TEST(av1_fwd_txfm1d, accuracy) {
 
     for (int ti = 0; ti < txfm_type_num; ++ti) {
       TYPE_TXFM txfm_type = txfm_type_ls[ti];
-      TxfmFunc fwd_txfm_func = fwd_txfm_func_ls[ti][si];
+      TxfmFunc fwd_txfm_func = fwd_txfm_func_ls[si][ti];
       int max_error = 7;
 
       const int count_test_block = 5000;
@@ -86,9 +87,10 @@ TEST(av1_fwd_txfm1d, accuracy) {
           reference_hybrid_1d(ref_input, ref_output, txfm_size, txfm_type);
 
           for (int ni = 0; ni < txfm_size; ++ni) {
-            EXPECT_LE(
+            ASSERT_LE(
                 abs(output[ni] - static_cast<int32_t>(round(ref_output[ni]))),
-                max_error);
+                max_error)
+                << "tx size = " << txfm_size << ", tx type = " << txfm_type;
           }
         }
       }

third_party/aom/test/av1_fwd_txfm2d_test.cc

@@ -12,24 +12,26 @@
 #include <math.h>
 #include <stdio.h>
 #include <stdlib.h>
+#include <vector>
+
+#include "config/av1_rtcd.h"
 
 #include "test/acm_random.h"
 #include "test/util.h"
 #include "test/av1_txfm_test.h"
 #include "av1/common/av1_txfm.h"
-#include "./av1_rtcd.h"
 
 using libaom_test::ACMRandom;
-using libaom_test::input_base;
+using libaom_test::TYPE_TXFM;
 using libaom_test::bd;
 using libaom_test::compute_avg_abs_error;
-using libaom_test::Fwd_Txfm2d_Func;
-using libaom_test::TYPE_TXFM;
+using libaom_test::input_base;
+
+using std::vector;
 
 namespace {
-#if CONFIG_HIGHBITDEPTH
 // tx_type_, tx_size_, max_error_, max_avg_error_
-typedef std::tr1::tuple<TX_TYPE, TX_SIZE, double, double> AV1FwdTxfm2dParam;
+typedef ::testing::tuple<TX_TYPE, TX_SIZE, double, double> AV1FwdTxfm2dParam;
 
 class AV1FwdTxfm2d : public ::testing::TestWithParam<AV1FwdTxfm2dParam> {
  public:
@@ -39,22 +41,16 @@ class AV1FwdTxfm2d : public ::testing::TestWithParam<AV1FwdTxfm2dParam> {
     max_error_ = GET_PARAM(2);
     max_avg_error_ = GET_PARAM(3);
     count_ = 500;
-    TXFM_2D_FLIP_CFG fwd_txfm_flip_cfg =
-        av1_get_fwd_txfm_cfg(tx_type_, tx_size_);
-    // TODO(sarahparker) this test will need to be updated when these
-    // functions are extended to support rectangular transforms
-    int amplify_bit = fwd_txfm_flip_cfg.row_cfg->shift[0] +
-                      fwd_txfm_flip_cfg.row_cfg->shift[1] +
-                      fwd_txfm_flip_cfg.row_cfg->shift[2];
+    TXFM_2D_FLIP_CFG fwd_txfm_flip_cfg;
+    av1_get_fwd_txfm_cfg(tx_type_, tx_size_, &fwd_txfm_flip_cfg);
+    amplify_factor_ = libaom_test::get_amplification_factor(tx_type_, tx_size_);
+    tx_width_ = tx_size_wide[fwd_txfm_flip_cfg.tx_size];
+    tx_height_ = tx_size_high[fwd_txfm_flip_cfg.tx_size];
     ud_flip_ = fwd_txfm_flip_cfg.ud_flip;
     lr_flip_ = fwd_txfm_flip_cfg.lr_flip;
-    amplify_factor_ =
-        amplify_bit >= 0 ? (1 << amplify_bit) : (1.0 / (1 << -amplify_bit));
 
     fwd_txfm_ = libaom_test::fwd_txfm_func_ls[tx_size_];
-    txfm1d_size_ = libaom_test::get_txfm1d_size(tx_size_);
-    txfm2d_size_ = txfm1d_size_ * txfm1d_size_;
-    get_txfm1d_type(tx_type_, &type0_, &type1_);
+    txfm2d_size_ = tx_width_ * tx_height_;
     input_ = reinterpret_cast<int16_t *>(
         aom_memalign(16, sizeof(input_[0]) * txfm2d_size_));
     output_ = reinterpret_cast<int32_t *>(
@@ -76,33 +72,40 @@ class AV1FwdTxfm2d : public ::testing::TestWithParam<AV1FwdTxfm2dParam> {
         ref_output_[ni] = 0;
       }
 
-      fwd_txfm_(input_, output_, txfm1d_size_, tx_type_, bd);
+      fwd_txfm_(input_, output_, tx_width_, tx_type_, bd);
 
-      if (lr_flip_ && ud_flip_)
-        libaom_test::fliplrud(ref_input_, txfm1d_size_, txfm1d_size_);
-      else if (lr_flip_)
-        libaom_test::fliplr(ref_input_, txfm1d_size_, txfm1d_size_);
-      else if (ud_flip_)
-        libaom_test::flipud(ref_input_, txfm1d_size_, txfm1d_size_);
-
-      reference_hybrid_2d(ref_input_, ref_output_, txfm1d_size_, type0_,
-                          type1_);
-
-      for (int ni = 0; ni < txfm2d_size_; ++ni) {
-        ref_output_[ni] = round(ref_output_[ni] * amplify_factor_);
-        EXPECT_GE(max_error_,
-                  fabs(output_[ni] - ref_output_[ni]) / amplify_factor_);
+      if (lr_flip_ && ud_flip_) {
+        libaom_test::fliplrud(ref_input_, tx_width_, tx_height_, tx_width_);
+      } else if (lr_flip_) {
+        libaom_test::fliplr(ref_input_, tx_width_, tx_height_, tx_width_);
+      } else if (ud_flip_) {
+        libaom_test::flipud(ref_input_, tx_width_, tx_height_, tx_width_);
       }
+
+      libaom_test::reference_hybrid_2d(ref_input_, ref_output_, tx_type_,
+                                       tx_size_);
+
+      double actual_max_error = 0;
+      for (int ni = 0; ni < txfm2d_size_; ++ni) {
+        ref_output_[ni] = round(ref_output_[ni]);
+        const double this_error =
+            fabs(output_[ni] - ref_output_[ni]) / amplify_factor_;
+        actual_max_error = AOMMAX(actual_max_error, this_error);
+      }
+      EXPECT_GE(max_error_, actual_max_error)
+          << "tx_size = " << tx_size_ << ", tx_type = " << tx_type_;
+      if (actual_max_error > max_error_) {  // exit early.
+        break;
+      }
+
       avg_abs_error += compute_avg_abs_error<int32_t, double>(
           output_, ref_output_, txfm2d_size_);
     }
 
     avg_abs_error /= amplify_factor_;
     avg_abs_error /= count_;
-    // max_abs_avg_error comes from upper bound of avg_abs_error
-    // printf("type0: %d type1: %d txfm_size: %d accuracy_avg_abs_error:
-    // %f\n", type0_, type1_, txfm1d_size_, avg_abs_error);
-    EXPECT_GE(max_avg_error_, avg_abs_error);
+    EXPECT_GE(max_avg_error_, avg_abs_error)
+        << "tx_size = " << tx_size_ << ", tx_type = " << tx_type_;
   }
 
   virtual void TearDown() {
@@ -119,11 +122,10 @@ class AV1FwdTxfm2d : public ::testing::TestWithParam<AV1FwdTxfm2dParam> {
   double amplify_factor_;
   TX_TYPE tx_type_;
   TX_SIZE tx_size_;
-  int txfm1d_size_;
+  int tx_width_;
+  int tx_height_;
   int txfm2d_size_;
-  Fwd_Txfm2d_Func fwd_txfm_;
-  TYPE_TXFM type0_;
-  TYPE_TXFM type1_;
+  FwdTxfm2dFunc fwd_txfm_;
   int16_t *input_;
   int32_t *output_;
   double *ref_input_;
@@ -132,76 +134,209 @@ class AV1FwdTxfm2d : public ::testing::TestWithParam<AV1FwdTxfm2dParam> {
   int lr_flip_;  // flip left to right
 };
 
-TEST_P(AV1FwdTxfm2d, RunFwdAccuracyCheck) { RunFwdAccuracyCheck(); }
-const AV1FwdTxfm2dParam av1_fwd_txfm2d_param_c[] = {
-#if CONFIG_EXT_TX
-  AV1FwdTxfm2dParam(FLIPADST_DCT, TX_4X4, 2, 0.2),
-  AV1FwdTxfm2dParam(DCT_FLIPADST, TX_4X4, 2, 0.2),
-  AV1FwdTxfm2dParam(FLIPADST_FLIPADST, TX_4X4, 2, 0.2),
-  AV1FwdTxfm2dParam(ADST_FLIPADST, TX_4X4, 2, 0.2),
-  AV1FwdTxfm2dParam(FLIPADST_ADST, TX_4X4, 2, 0.2),
-  AV1FwdTxfm2dParam(FLIPADST_DCT, TX_8X8, 5, 0.6),
-  AV1FwdTxfm2dParam(DCT_FLIPADST, TX_8X8, 5, 0.6),
-  AV1FwdTxfm2dParam(FLIPADST_FLIPADST, TX_8X8, 5, 0.6),
-  AV1FwdTxfm2dParam(ADST_FLIPADST, TX_8X8, 5, 0.6),
-  AV1FwdTxfm2dParam(FLIPADST_ADST, TX_8X8, 5, 0.6),
-  AV1FwdTxfm2dParam(FLIPADST_DCT, TX_16X16, 11, 1.5),
-  AV1FwdTxfm2dParam(DCT_FLIPADST, TX_16X16, 11, 1.5),
-  AV1FwdTxfm2dParam(FLIPADST_FLIPADST, TX_16X16, 11, 1.5),
-  AV1FwdTxfm2dParam(ADST_FLIPADST, TX_16X16, 11, 1.5),
-  AV1FwdTxfm2dParam(FLIPADST_ADST, TX_16X16, 11, 1.5),
-  AV1FwdTxfm2dParam(FLIPADST_DCT, TX_32X32, 70, 7),
-  AV1FwdTxfm2dParam(DCT_FLIPADST, TX_32X32, 70, 7),
-  AV1FwdTxfm2dParam(FLIPADST_FLIPADST, TX_32X32, 70, 7),
-  AV1FwdTxfm2dParam(ADST_FLIPADST, TX_32X32, 70, 7),
-  AV1FwdTxfm2dParam(FLIPADST_ADST, TX_32X32, 70, 7),
-#endif
-  AV1FwdTxfm2dParam(DCT_DCT, TX_4X4, 2, 0.2),
-  AV1FwdTxfm2dParam(ADST_DCT, TX_4X4, 2, 0.2),
-  AV1FwdTxfm2dParam(DCT_ADST, TX_4X4, 2, 0.2),
-  AV1FwdTxfm2dParam(ADST_ADST, TX_4X4, 2, 0.2),
-  AV1FwdTxfm2dParam(DCT_DCT, TX_8X8, 5, 0.6),
-  AV1FwdTxfm2dParam(ADST_DCT, TX_8X8, 5, 0.6),
-  AV1FwdTxfm2dParam(DCT_ADST, TX_8X8, 5, 0.6),
-  AV1FwdTxfm2dParam(ADST_ADST, TX_8X8, 5, 0.6),
-  AV1FwdTxfm2dParam(DCT_DCT, TX_16X16, 11, 1.5),
-  AV1FwdTxfm2dParam(ADST_DCT, TX_16X16, 11, 1.5),
-  AV1FwdTxfm2dParam(DCT_ADST, TX_16X16, 11, 1.5),
-  AV1FwdTxfm2dParam(ADST_ADST, TX_16X16, 11, 1.5),
-  AV1FwdTxfm2dParam(DCT_DCT, TX_32X32, 70, 7),
-  AV1FwdTxfm2dParam(ADST_DCT, TX_32X32, 70, 7),
-  AV1FwdTxfm2dParam(DCT_ADST, TX_32X32, 70, 7),
-  AV1FwdTxfm2dParam(ADST_ADST, TX_32X32, 70, 7)
+static double avg_error_ls[TX_SIZES_ALL] = {
+  0.5,   // 4x4 transform
+  0.5,   // 8x8 transform
+  1.2,   // 16x16 transform
+  6.1,   // 32x32 transform
+  3.4,   // 64x64 transform
+  0.57,  // 4x8 transform
+  0.68,  // 8x4 transform
+  0.92,  // 8x16 transform
+  1.1,   // 16x8 transform
+  4.1,   // 16x32 transform
+  6,     // 32x16 transform
+  3.5,   // 32x64 transform
+  5.7,   // 64x32 transform
+  0.6,   // 4x16 transform
+  0.9,   // 16x4 transform
+  1.2,   // 8x32 transform
+  1.7,   // 32x8 transform
+  2.0,   // 16x64 transform
+  4.7,   // 64x16 transform
 };
 
+static double max_error_ls[TX_SIZES_ALL] = {
+  3,    // 4x4 transform
+  5,    // 8x8 transform
+  11,   // 16x16 transform
+  70,   // 32x32 transform
+  64,   // 64x64 transform
+  3.9,  // 4x8 transform
+  4.3,  // 8x4 transform
+  12,   // 8x16 transform
+  12,   // 16x8 transform
+  32,   // 16x32 transform
+  46,   // 32x16 transform
+  136,  // 32x64 transform
+  136,  // 64x32 transform
+  5,    // 4x16 transform
+  6,    // 16x4 transform
+  21,   // 8x32 transform
+  13,   // 32x8 transform
+  30,   // 16x64 transform
+  36,   // 64x16 transform
+};
+
+vector<AV1FwdTxfm2dParam> GetTxfm2dParamList() {
+  vector<AV1FwdTxfm2dParam> param_list;
+  for (int s = 0; s < TX_SIZES; ++s) {
+    const double max_error = max_error_ls[s];
+    const double avg_error = avg_error_ls[s];
+    for (int t = 0; t < TX_TYPES; ++t) {
+      const TX_TYPE tx_type = static_cast<TX_TYPE>(t);
+      const TX_SIZE tx_size = static_cast<TX_SIZE>(s);
+      if (libaom_test::IsTxSizeTypeValid(tx_size, tx_type)) {
+        param_list.push_back(
+            AV1FwdTxfm2dParam(tx_type, tx_size, max_error, avg_error));
+      }
+    }
+  }
+  return param_list;
+}
+
 INSTANTIATE_TEST_CASE_P(C, AV1FwdTxfm2d,
-                        ::testing::ValuesIn(av1_fwd_txfm2d_param_c));
+                        ::testing::ValuesIn(GetTxfm2dParamList()));
+
+TEST_P(AV1FwdTxfm2d, RunFwdAccuracyCheck) { RunFwdAccuracyCheck(); }
 
 TEST(AV1FwdTxfm2d, CfgTest) {
   for (int bd_idx = 0; bd_idx < BD_NUM; ++bd_idx) {
     int bd = libaom_test::bd_arr[bd_idx];
     int8_t low_range = libaom_test::low_range_arr[bd_idx];
     int8_t high_range = libaom_test::high_range_arr[bd_idx];
-    // TODO(angiebird): include rect txfm in this test
-    for (int tx_size = 0; tx_size < TX_SIZES; ++tx_size) {
+    for (int tx_size = 0; tx_size < TX_SIZES_ALL; ++tx_size) {
       for (int tx_type = 0; tx_type < TX_TYPES; ++tx_type) {
-        TXFM_2D_FLIP_CFG cfg = av1_get_fwd_txfm_cfg(
-            static_cast<TX_TYPE>(tx_type), static_cast<TX_SIZE>(tx_size));
+        if (libaom_test::IsTxSizeTypeValid(static_cast<TX_SIZE>(tx_size),
+                                           static_cast<TX_TYPE>(tx_type)) ==
+            false) {
+          continue;
+        }
+        TXFM_2D_FLIP_CFG cfg;
+        av1_get_fwd_txfm_cfg(static_cast<TX_TYPE>(tx_type),
+                             static_cast<TX_SIZE>(tx_size), &cfg);
         int8_t stage_range_col[MAX_TXFM_STAGE_NUM];
         int8_t stage_range_row[MAX_TXFM_STAGE_NUM];
         av1_gen_fwd_stage_range(stage_range_col, stage_range_row, &cfg, bd);
-        const TXFM_1D_CFG *col_cfg = cfg.col_cfg;
-        const TXFM_1D_CFG *row_cfg = cfg.row_cfg;
-        libaom_test::txfm_stage_range_check(stage_range_col, col_cfg->stage_num,
-                                            col_cfg->cos_bit, low_range,
+        libaom_test::txfm_stage_range_check(stage_range_col, cfg.stage_num_col,
+                                            cfg.cos_bit_col, low_range,
                                             high_range);
-        libaom_test::txfm_stage_range_check(stage_range_row, row_cfg->stage_num,
-                                            row_cfg->cos_bit, low_range,
+        libaom_test::txfm_stage_range_check(stage_range_row, cfg.stage_num_row,
+                                            cfg.cos_bit_row, low_range,
                                             high_range);
       }
     }
   }
 }
 
-#endif  // CONFIG_HIGHBITDEPTH
+typedef void (*lowbd_fwd_txfm_func)(const int16_t *src_diff, tran_low_t *coeff,
+                                    int diff_stride, TxfmParam *txfm_param);
+
+void AV1FwdTxfm2dMatchTest(TX_SIZE tx_size, lowbd_fwd_txfm_func target_func) {
+  const int bd = 8;
+  TxfmParam param;
+  memset(&param, 0, sizeof(param));
+  const int rows = tx_size_high[tx_size];
+  const int cols = tx_size_wide[tx_size];
+  // printf("%d x %d\n", cols, rows);
+  for (int tx_type = 0; tx_type < TX_TYPES; ++tx_type) {
+    if (libaom_test::IsTxSizeTypeValid(
+            tx_size, static_cast<TX_TYPE>(tx_type)) == false) {
+      continue;
+    }
+
+    FwdTxfm2dFunc ref_func = libaom_test::fwd_txfm_func_ls[tx_size];
+    if (ref_func != NULL) {
+      DECLARE_ALIGNED(16, int16_t, input[64 * 64]) = { 0 };
+      DECLARE_ALIGNED(16, int32_t, output[64 * 64]);
+      DECLARE_ALIGNED(16, int32_t, ref_output[64 * 64]);
+      int input_stride = 64;
+      ACMRandom rnd(ACMRandom::DeterministicSeed());
+      for (int cnt = 0; cnt < 500; ++cnt) {
+        if (cnt == 0) {
+          for (int r = 0; r < rows; ++r) {
+            for (int c = 0; c < cols; ++c) {
+              input[r * input_stride + c] = (1 << bd) - 1;
+            }
+          }
+        } else {
+          for (int r = 0; r < rows; ++r) {
+            for (int c = 0; c < cols; ++c) {
+              input[r * input_stride + c] = rnd.Rand16() % (1 << bd);
+            }
+          }
+        }
+        param.tx_type = (TX_TYPE)tx_type;
+        param.tx_size = (TX_SIZE)tx_size;
+        param.tx_set_type = EXT_TX_SET_ALL16;
+        param.bd = bd;
+        ref_func(input, ref_output, input_stride, (TX_TYPE)tx_type, bd);
+        target_func(input, output, input_stride, &param);
+        const int check_rows = AOMMIN(32, rows);
+        const int check_cols = AOMMIN(32, rows * cols / check_rows);
+        for (int r = 0; r < check_rows; ++r) {
+          for (int c = 0; c < check_cols; ++c) {
+            ASSERT_EQ(ref_output[r * check_cols + c],
+                      output[r * check_cols + c])
+                << "[" << r << "," << c << "] cnt:" << cnt
+                << " tx_size: " << tx_size << " tx_type: " << tx_type;
+          }
+        }
+      }
+    }
+  }
+}
+
+typedef ::testing::tuple<TX_SIZE, lowbd_fwd_txfm_func> LbdFwdTxfm2dParam;
+
+class AV1FwdTxfm2dTest : public ::testing::TestWithParam<LbdFwdTxfm2dParam> {};
+
+TEST_P(AV1FwdTxfm2dTest, match) {
+  AV1FwdTxfm2dMatchTest(GET_PARAM(0), GET_PARAM(1));
+}
+
+using ::testing::Combine;
+using ::testing::Values;
+using ::testing::ValuesIn;
+
+#if HAVE_SSE2
+static TX_SIZE fwd_txfm_for_sse2[] = {
+  TX_4X4,
+  TX_8X8,
+  TX_16X16,
+  TX_32X32,
+  // TX_64X64,
+  TX_4X8,
+  TX_8X4,
+  TX_8X16,
+  TX_16X8,
+  TX_16X32,
+  TX_32X16,
+  // TX_32X64,
+  // TX_64X32,
+  TX_4X16,
+  TX_16X4,
+  TX_8X32,
+  TX_32X8,
+  TX_16X64,
+  TX_64X16,
+};
+
+INSTANTIATE_TEST_CASE_P(SSE2, AV1FwdTxfm2dTest,
+                        Combine(ValuesIn(fwd_txfm_for_sse2),
+                                Values(av1_lowbd_fwd_txfm_sse2)));
+#endif  // HAVE_SSE2
+
+#if HAVE_SSE4_1
+static TX_SIZE fwd_txfm_for_sse41[] = {
+  TX_4X4,
+  TX_64X64,
+  TX_32X64,
+  TX_64X32,
+};
+
+INSTANTIATE_TEST_CASE_P(SSE4_1, AV1FwdTxfm2dTest,
+                        Combine(ValuesIn(fwd_txfm_for_sse41),
+                                Values(av1_lowbd_fwd_txfm_sse4_1)));
+#endif  // HAVE_SSE4_1
 }  // namespace

third_party/aom/test/av1_highbd_iht_test.cc

@@ -11,7 +11,8 @@
 
 #include "third_party/googletest/src/googletest/include/gtest/gtest.h"
 
-#include "./av1_rtcd.h"
+#include "config/av1_rtcd.h"
+
 #include "test/acm_random.h"
 #include "test/clear_system_state.h"
 #include "test/register_state_check.h"
@@ -22,7 +23,7 @@
 
 namespace {
 
-using std::tr1::tuple;
+using ::testing::tuple;
 using libaom_test::ACMRandom;
 
 typedef void (*HbdHtFunc)(const int16_t *input, int32_t *output, int stride,
@@ -88,6 +89,8 @@ class AV1HighbdInvHTNxN : public ::testing::TestWithParam<IHbdHtParam> {
       return 16;
     } else if (1024 == num_coeffs_) {
       return 32;
+    } else if (4096 == num_coeffs_) {
+      return 64;
     } else {
       return 0;
     }
@@ -133,28 +136,24 @@ void AV1HighbdInvHTNxN::RunBitexactCheck() {
 
 TEST_P(AV1HighbdInvHTNxN, InvTransResultCheck) { RunBitexactCheck(); }
 
-using std::tr1::make_tuple;
+using ::testing::make_tuple;
 
-#if HAVE_SSE4_1 && CONFIG_HIGHBITDEPTH && \
-    !(CONFIG_DAALA_DCT4 && CONFIG_DAALA_DCT8 && CONFIG_DAALA_DCT16)
-#if !CONFIG_DAALA_DCT4
+#if HAVE_SSE4_1
 #define PARAM_LIST_4X4                                   \
   &av1_fwd_txfm2d_4x4_c, &av1_inv_txfm2d_add_4x4_sse4_1, \
       &av1_inv_txfm2d_add_4x4_c, 16
-#endif
-#if !CONFIG_DAALA_DCT8
 #define PARAM_LIST_8X8                                   \
   &av1_fwd_txfm2d_8x8_c, &av1_inv_txfm2d_add_8x8_sse4_1, \
       &av1_inv_txfm2d_add_8x8_c, 64
-#endif
-#if !CONFIG_DAALA_DCT16
 #define PARAM_LIST_16X16                                     \
   &av1_fwd_txfm2d_16x16_c, &av1_inv_txfm2d_add_16x16_sse4_1, \
       &av1_inv_txfm2d_add_16x16_c, 256
-#endif
+#define PARAM_LIST_64X64                                     \
+  &av1_fwd_txfm2d_64x64_c, &av1_inv_txfm2d_add_64x64_sse4_1, \
+      &av1_inv_txfm2d_add_64x64_c, 4096
+
 const IHbdHtParam kArrayIhtParam[] = {
-// 16x16
-#if !CONFIG_DAALA_DCT16
+  // 16x16
   make_tuple(PARAM_LIST_16X16, DCT_DCT, 10),
   make_tuple(PARAM_LIST_16X16, DCT_DCT, 12),
   make_tuple(PARAM_LIST_16X16, ADST_DCT, 10),
@@ -163,7 +162,6 @@ const IHbdHtParam kArrayIhtParam[] = {
   make_tuple(PARAM_LIST_16X16, DCT_ADST, 12),
   make_tuple(PARAM_LIST_16X16, ADST_ADST, 10),
   make_tuple(PARAM_LIST_16X16, ADST_ADST, 12),
-#if CONFIG_EXT_TX
   make_tuple(PARAM_LIST_16X16, FLIPADST_DCT, 10),
   make_tuple(PARAM_LIST_16X16, FLIPADST_DCT, 12),
   make_tuple(PARAM_LIST_16X16, DCT_FLIPADST, 10),
@@ -174,10 +172,7 @@ const IHbdHtParam kArrayIhtParam[] = {
   make_tuple(PARAM_LIST_16X16, ADST_FLIPADST, 12),
   make_tuple(PARAM_LIST_16X16, FLIPADST_ADST, 10),
   make_tuple(PARAM_LIST_16X16, FLIPADST_ADST, 12),
-#endif
-#endif
-// 8x8
-#if !CONFIG_DAALA_DCT8
+  // 8x8
   make_tuple(PARAM_LIST_8X8, DCT_DCT, 10),
   make_tuple(PARAM_LIST_8X8, DCT_DCT, 12),
   make_tuple(PARAM_LIST_8X8, ADST_DCT, 10),
@@ -186,7 +181,6 @@ const IHbdHtParam kArrayIhtParam[] = {
   make_tuple(PARAM_LIST_8X8, DCT_ADST, 12),
   make_tuple(PARAM_LIST_8X8, ADST_ADST, 10),
   make_tuple(PARAM_LIST_8X8, ADST_ADST, 12),
-#if CONFIG_EXT_TX
   make_tuple(PARAM_LIST_8X8, FLIPADST_DCT, 10),
   make_tuple(PARAM_LIST_8X8, FLIPADST_DCT, 12),
   make_tuple(PARAM_LIST_8X8, DCT_FLIPADST, 10),
@@ -197,10 +191,7 @@ const IHbdHtParam kArrayIhtParam[] = {
   make_tuple(PARAM_LIST_8X8, ADST_FLIPADST, 12),
   make_tuple(PARAM_LIST_8X8, FLIPADST_ADST, 10),
   make_tuple(PARAM_LIST_8X8, FLIPADST_ADST, 12),
-#endif
-#endif
-// 4x4
-#if !CONFIG_DAALA_DCT4
+  // 4x4
   make_tuple(PARAM_LIST_4X4, DCT_DCT, 10),
   make_tuple(PARAM_LIST_4X4, DCT_DCT, 12),
   make_tuple(PARAM_LIST_4X4, ADST_DCT, 10),
@@ -209,7 +200,6 @@ const IHbdHtParam kArrayIhtParam[] = {
   make_tuple(PARAM_LIST_4X4, DCT_ADST, 12),
   make_tuple(PARAM_LIST_4X4, ADST_ADST, 10),
   make_tuple(PARAM_LIST_4X4, ADST_ADST, 12),
-#if CONFIG_EXT_TX
   make_tuple(PARAM_LIST_4X4, FLIPADST_DCT, 10),
   make_tuple(PARAM_LIST_4X4, FLIPADST_DCT, 12),
   make_tuple(PARAM_LIST_4X4, DCT_FLIPADST, 10),
@@ -220,16 +210,15 @@ const IHbdHtParam kArrayIhtParam[] = {
   make_tuple(PARAM_LIST_4X4, ADST_FLIPADST, 12),
   make_tuple(PARAM_LIST_4X4, FLIPADST_ADST, 10),
   make_tuple(PARAM_LIST_4X4, FLIPADST_ADST, 12),
-#endif
-#endif
+  make_tuple(PARAM_LIST_64X64, DCT_DCT, 10),
+  make_tuple(PARAM_LIST_64X64, DCT_DCT, 12),
 };
 
 INSTANTIATE_TEST_CASE_P(SSE4_1, AV1HighbdInvHTNxN,
                         ::testing::ValuesIn(kArrayIhtParam));
-#endif  // HAVE_SSE4_1 && CONFIG_HIGHBITDEPTH &&
-        //  !(CONFIG_DAALA_DCT4 && CONFIG_DAALA_DCT8 && CONFIG_DAALA_DCT16)
+#endif  // HAVE_SSE4_1
 
-#if HAVE_AVX2 && CONFIG_HIGHBITDEPTH && !CONFIG_DAALA_DCT32
+#if HAVE_AVX2
 #define PARAM_LIST_32X32                                   \
   &av1_fwd_txfm2d_32x32_c, &av1_inv_txfm2d_add_32x32_avx2, \
       &av1_inv_txfm2d_add_32x32_c, 1024
@@ -243,5 +232,5 @@ const IHbdHtParam kArrayIhtParam32x32[] = {
 INSTANTIATE_TEST_CASE_P(AVX2, AV1HighbdInvHTNxN,
                         ::testing::ValuesIn(kArrayIhtParam32x32));
 
-#endif  // HAVE_AVX2 && CONFIG_HIGHBITDEPTH
+#endif  // HAVE_AVX2
 }  // namespace

third_party/aom/test/av1_horz_only_frame_superres_test.cc

@@ -0,0 +1,362 @@
+/*
+ * Copyright (c) 2018, Alliance for Open Media. All rights reserved
+ *
+ * This source code is subject to the terms of the BSD 2 Clause License and
+ * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
+ * was not distributed with this source code in the LICENSE file, you can
+ * obtain it at www.aomedia.org/license/software. If the Alliance for Open
+ * Media Patent License 1.0 was not distributed with this source code in the
+ * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
+ */
+
+#include <vector>
+
+#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
+
+#include "config/av1_rtcd.h"
+
+#include "aom_ports/aom_timer.h"
+#include "av1/common/convolve.h"
+#include "av1/common/resize.h"
+#include "test/acm_random.h"
+#include "test/clear_system_state.h"
+#include "test/register_state_check.h"
+#include "test/util.h"
+
+namespace {
+const int kTestIters = 10;
+const int kPerfIters = 1000;
+
+const int kVPad = 32;
+const int kHPad = 32;
+
+using ::testing::make_tuple;
+using ::testing::tuple;
+using libaom_test::ACMRandom;
+
+template <typename Pixel>
+class TestImage {
+ public:
+  TestImage(int w_src, int h, int superres_denom, int x0, int bd)
+      : w_src_(w_src), h_(h), superres_denom_(superres_denom), x0_(x0),
+        bd_(bd) {
+    assert(bd < 16);
+    assert(bd <= 8 * static_cast<int>(sizeof(Pixel)));
+    assert(9 <= superres_denom && superres_denom <= 16);
+    assert(SCALE_NUMERATOR == 8);
+    assert(0 <= x0_ && x0_ <= RS_SCALE_SUBPEL_MASK);
+
+    w_dst_ = w_src_;
+    av1_calculate_unscaled_superres_size(&w_dst_, NULL, superres_denom);
+
+    src_stride_ = ALIGN_POWER_OF_TWO(w_src_ + 2 * kHPad, 4);
+    dst_stride_ = ALIGN_POWER_OF_TWO(w_dst_ + 2 * kHPad, 4);
+
+    // Allocate image data
+    src_data_.resize(2 * src_block_size());
+    dst_data_.resize(2 * dst_block_size());
+  }
+
+  void Initialize(ACMRandom *rnd);
+  void Check() const;
+
+  int src_stride() const { return src_stride_; }
+  int dst_stride() const { return dst_stride_; }
+
+  int src_block_size() const { return (h_ + 2 * kVPad) * src_stride(); }
+  int dst_block_size() const { return (h_ + 2 * kVPad) * dst_stride(); }
+
+  int src_width() const { return w_src_; }
+  int dst_width() const { return w_dst_; }
+  int height() const { return h_; }
+  int x0() const { return x0_; }
+
+  const Pixel *GetSrcData(bool ref, bool borders) const {
+    const Pixel *block = &src_data_[ref ? 0 : src_block_size()];
+    return borders ? block : block + kHPad + src_stride_ * kVPad;
+  }
+
+  Pixel *GetDstData(bool ref, bool borders) {
+    Pixel *block = &dst_data_[ref ? 0 : dst_block_size()];
+    return borders ? block : block + kHPad + dst_stride_ * kVPad;
+  }
+
+ private:
+  int w_src_, w_dst_, h_, superres_denom_, x0_, bd_;
+  int src_stride_, dst_stride_;
+
+  std::vector<Pixel> src_data_;
+  std::vector<Pixel> dst_data_;
+};
+
+template <typename Pixel>
+void FillEdge(ACMRandom *rnd, int num_pixels, int bd, bool trash, Pixel *data) {
+  if (!trash) {
+    memset(data, 0, sizeof(*data) * num_pixels);
+    return;
+  }
+  const Pixel mask = (1 << bd) - 1;
+  for (int i = 0; i < num_pixels; ++i) data[i] = rnd->Rand16() & mask;
+}
+
+template <typename Pixel>
+void PrepBuffers(ACMRandom *rnd, int w, int h, int stride, int bd,
+                 bool trash_edges, Pixel *data) {
+  assert(rnd);
+  const Pixel mask = (1 << bd) - 1;
+
+  // Fill in the first buffer with random data
+  // Top border
+  FillEdge(rnd, stride * kVPad, bd, trash_edges, data);
+  for (int r = 0; r < h; ++r) {
+    Pixel *row_data = data + (kVPad + r) * stride;
+    // Left border, contents, right border
+    FillEdge(rnd, kHPad, bd, trash_edges, row_data);
+    for (int c = 0; c < w; ++c) row_data[kHPad + c] = rnd->Rand16() & mask;
+    FillEdge(rnd, kHPad, bd, trash_edges, row_data + kHPad + w);
+  }
+  // Bottom border
+  FillEdge(rnd, stride * kVPad, bd, trash_edges, data + stride * (kVPad + h));
+
+  const int bpp = sizeof(*data);
+  const int block_elts = stride * (h + 2 * kVPad);
+  const int block_size = bpp * block_elts;
+
+  // Now copy that to the second buffer
+  memcpy(data + block_elts, data, block_size);
+}
+
+template <typename Pixel>
+void TestImage<Pixel>::Initialize(ACMRandom *rnd) {
+  PrepBuffers(rnd, w_src_, h_, src_stride_, bd_, false, &src_data_[0]);
+  PrepBuffers(rnd, w_dst_, h_, dst_stride_, bd_, true, &dst_data_[0]);
+}
+
+template <typename Pixel>
+void TestImage<Pixel>::Check() const {
+  const int num_pixels = dst_block_size();
+  const Pixel *ref_dst = &dst_data_[0];
+  const Pixel *tst_dst = &dst_data_[num_pixels];
+
+  // If memcmp returns 0, there's nothing to do.
+  if (0 == memcmp(ref_dst, tst_dst, sizeof(*ref_dst) * num_pixels)) return;
+
+  // Otherwise, iterate through the buffer looking for differences, *ignoring
+  // the edges*
+  const int stride = dst_stride_;
+  for (int r = kVPad; r < h_ + kVPad; ++r) {
+    for (int c = kVPad; c < w_dst_ + kHPad; ++c) {
+      const int32_t ref_value = ref_dst[r * stride + c];
+      const int32_t tst_value = tst_dst[r * stride + c];
+
+      EXPECT_EQ(tst_value, ref_value)
+          << "Error at row: " << (r - kVPad) << ", col: " << (c - kHPad)
+          << ", superres_denom: " << superres_denom_ << ", height: " << h_
+          << ", src_width: " << w_src_ << ", dst_width: " << w_dst_
+          << ", x0: " << x0_;
+    }
+  }
+}
+
+template <typename Pixel>
+class ConvolveHorizRSTestBase : public ::testing::Test {
+ public:
+  ConvolveHorizRSTestBase() : image_(NULL) {}
+  virtual ~ConvolveHorizRSTestBase() {}
+  virtual void TearDown() { libaom_test::ClearSystemState(); }
+
+  // Implemented by subclasses (SetUp depends on the parameters passed
+  // in and RunOne depends on the function to be tested. These can't
+  // be templated for low/high bit depths because they have different
+  // numbers of parameters)
+  virtual void SetUp() = 0;
+  virtual void RunOne(bool ref) = 0;
+
+ protected:
+  void SetBitDepth(int bd) { bd_ = bd; }
+
+  void CorrectnessTest() {
+    ACMRandom rnd(ACMRandom::DeterministicSeed());
+    for (int i = 0; i < kTestIters; ++i) {
+      for (int superres_denom = 9; superres_denom <= 16; superres_denom++) {
+        // Get a random height between 512 and 767
+        int height = rnd.Rand8() + 512;
+
+        // Get a random src width between 128 and 383
+        int width_src = rnd.Rand8() + 128;
+
+        // x0 is normally calculated by get_upscale_convolve_x0 in
+        // av1/common/resize.c. However, this test should work for
+        // any value of x0 between 0 and RS_SCALE_SUBPEL_MASK
+        // (inclusive), so we choose one at random.
+        int x0 = rnd.Rand16() % (RS_SCALE_SUBPEL_MASK + 1);
+
+        image_ =
+            new TestImage<Pixel>(width_src, height, superres_denom, x0, bd_);
+
+        Prep(&rnd);
+        RunOne(true);
+        RunOne(false);
+        image_->Check();
+
+        delete image_;
+      }
+    }
+  }
+
+  void SpeedTest() {
+    // Pick some specific parameters to test
+    int height = 767;
+    int width_src = 129;
+    int superres_denom = 13;
+    int x0 = RS_SCALE_SUBPEL_MASK >> 1;
+
+    image_ = new TestImage<Pixel>(width_src, height, superres_denom, x0, bd_);
+
+    ACMRandom rnd(ACMRandom::DeterministicSeed());
+    Prep(&rnd);
+
+    aom_usec_timer ref_timer;
+    aom_usec_timer_start(&ref_timer);
+    for (int i = 0; i < kPerfIters; ++i) RunOne(true);
+    aom_usec_timer_mark(&ref_timer);
+    const int64_t ref_time = aom_usec_timer_elapsed(&ref_timer);
+
+    aom_usec_timer tst_timer;
+    aom_usec_timer_start(&tst_timer);
+    for (int i = 0; i < kPerfIters; ++i) RunOne(false);
+    aom_usec_timer_mark(&tst_timer);
+    const int64_t tst_time = aom_usec_timer_elapsed(&tst_timer);
+
+    std::cout << "[          ] C time = " << ref_time / 1000
+              << " ms, SIMD time = " << tst_time / 1000 << " ms\n";
+
+    EXPECT_GT(ref_time, tst_time)
+        << "Error: ConvolveHorizRSTest (Speed Test), SIMD slower than C.\n"
+        << "C time: " << ref_time << " us\n"
+        << "SIMD time: " << tst_time << " us\n";
+  }
+
+  void Prep(ACMRandom *rnd) {
+    assert(rnd);
+    image_->Initialize(rnd);
+  }
+
+  int bd_;
+  TestImage<Pixel> *image_;
+};
+
+typedef void (*LowBDConvolveHorizRsFunc)(const uint8_t *src, int src_stride,
+                                         uint8_t *dst, int dst_stride, int w,
+                                         int h, const int16_t *x_filters,
+                                         const int x0_qn, const int x_step_qn);
+
+// Test parameter list:
+//  <tst_fun_>
+typedef tuple<LowBDConvolveHorizRsFunc> LowBDParams;
+
+class LowBDConvolveHorizRSTest
+    : public ConvolveHorizRSTestBase<uint8_t>,
+      public ::testing::WithParamInterface<LowBDParams> {
+ public:
+  virtual ~LowBDConvolveHorizRSTest() {}
+
+  void SetUp() {
+    tst_fun_ = GET_PARAM(0);
+    const int bd = 8;
+    SetBitDepth(bd);
+  }
+
+  void RunOne(bool ref) {
+    const uint8_t *src = image_->GetSrcData(ref, false);
+    uint8_t *dst = image_->GetDstData(ref, false);
+    const int src_stride = image_->src_stride();
+    const int dst_stride = image_->dst_stride();
+    const int width_src = image_->src_width();
+    const int width_dst = image_->dst_width();
+    const int height = image_->height();
+    const int x0_qn = image_->x0();
+
+    const int32_t x_step_qn =
+        av1_get_upscale_convolve_step(width_src, width_dst);
+
+    if (ref) {
+      av1_convolve_horiz_rs_c(src, src_stride, dst, dst_stride, width_dst,
+                              height, &av1_resize_filter_normative[0][0], x0_qn,
+                              x_step_qn);
+    } else {
+      tst_fun_(src, src_stride, dst, dst_stride, width_dst, height,
+               &av1_resize_filter_normative[0][0], x0_qn, x_step_qn);
+    }
+  }
+
+ private:
+  LowBDConvolveHorizRsFunc tst_fun_;
+};
+
+TEST_P(LowBDConvolveHorizRSTest, Correctness) { CorrectnessTest(); }
+TEST_P(LowBDConvolveHorizRSTest, DISABLED_Speed) { SpeedTest(); }
+
+INSTANTIATE_TEST_CASE_P(SSE4_1, LowBDConvolveHorizRSTest,
+                        ::testing::Values(av1_convolve_horiz_rs_sse4_1));
+
+typedef void (*HighBDConvolveHorizRsFunc)(const uint16_t *src, int src_stride,
+                                          uint16_t *dst, int dst_stride, int w,
+                                          int h, const int16_t *x_filters,
+                                          const int x0_qn, const int x_step_qn,
+                                          int bd);
+
+// Test parameter list:
+//  <tst_fun_, bd_>
+typedef tuple<HighBDConvolveHorizRsFunc, int> HighBDParams;
+
+class HighBDConvolveHorizRSTest
+    : public ConvolveHorizRSTestBase<uint16_t>,
+      public ::testing::WithParamInterface<HighBDParams> {
+ public:
+  virtual ~HighBDConvolveHorizRSTest() {}
+
+  void SetUp() {
+    tst_fun_ = GET_PARAM(0);
+    const int bd = GET_PARAM(1);
+    SetBitDepth(bd);
+  }
+
+  void RunOne(bool ref) {
+    const uint16_t *src = image_->GetSrcData(ref, false);
+    uint16_t *dst = image_->GetDstData(ref, false);
+    const int src_stride = image_->src_stride();
+    const int dst_stride = image_->dst_stride();
+    const int width_src = image_->src_width();
+    const int width_dst = image_->dst_width();
+    const int height = image_->height();
+    const int x0_qn = image_->x0();
+
+    const int32_t x_step_qn =
+        av1_get_upscale_convolve_step(width_src, width_dst);
+
+    if (ref) {
+      av1_highbd_convolve_horiz_rs_c(
+          src, src_stride, dst, dst_stride, width_dst, height,
+          &av1_resize_filter_normative[0][0], x0_qn, x_step_qn, bd_);
+    } else {
+      tst_fun_(src, src_stride, dst, dst_stride, width_dst, height,
+               &av1_resize_filter_normative[0][0], x0_qn, x_step_qn, bd_);
+    }
+  }
+
+ private:
+  HighBDConvolveHorizRsFunc tst_fun_;
+};
+
+const int kBDs[] = { 8, 10, 12 };
+
+TEST_P(HighBDConvolveHorizRSTest, Correctness) { CorrectnessTest(); }
+TEST_P(HighBDConvolveHorizRSTest, DISABLED_Speed) { SpeedTest(); }
+
+INSTANTIATE_TEST_CASE_P(
+    SSE4_1, HighBDConvolveHorizRSTest,
+    ::testing::Combine(::testing::Values(av1_highbd_convolve_horiz_rs_sse4_1),
+                       ::testing::ValuesIn(kBDs)));
+
+}  // namespace

third_party/aom/test/av1_inv_txfm1d_test.cc

@@ -13,39 +13,35 @@
 
 #include "test/av1_txfm_test.h"
 #include "test/util.h"
-#include "av1/common/av1_fwd_txfm1d.h"
 #include "av1/common/av1_inv_txfm1d.h"
+#include "av1/encoder/av1_fwd_txfm1d.h"
 
 using libaom_test::ACMRandom;
 using libaom_test::input_base;
 
 namespace {
 const int txfm_type_num = 2;
-const int txfm_size_ls[5] = { 4, 8, 16, 32, 64 };
+const int txfm_size_ls[] = { 4, 8, 16, 32, 64 };
 
-const TxfmFunc fwd_txfm_func_ls[][2] = {
+const TxfmFunc fwd_txfm_func_ls[][txfm_type_num] = {
   { av1_fdct4_new, av1_fadst4_new },
   { av1_fdct8_new, av1_fadst8_new },
   { av1_fdct16_new, av1_fadst16_new },
-  { av1_fdct32_new, av1_fadst32_new },
-#if CONFIG_TX64X64
+  { av1_fdct32_new, NULL },
   { av1_fdct64_new, NULL },
-#endif
 };
 
-const TxfmFunc inv_txfm_func_ls[][2] = {
+const TxfmFunc inv_txfm_func_ls[][txfm_type_num] = {
   { av1_idct4_new, av1_iadst4_new },
   { av1_idct8_new, av1_iadst8_new },
   { av1_idct16_new, av1_iadst16_new },
-  { av1_idct32_new, av1_iadst32_new },
-#if CONFIG_TX64X64
+  { av1_idct32_new, NULL },
   { av1_idct64_new, NULL },
-#endif
 };
 
 // the maximum stage number of fwd/inv 1d dct/adst txfm is 12
-const int8_t cos_bit[12] = { 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13 };
-const int8_t range_bit[12] = { 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32 };
+const int8_t cos_bit = 13;
+const int8_t range_bit[12] = { 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20 };
 
 void reference_idct_1d_int(const int32_t *in, int32_t *out, int size) {
   double input[64];
@@ -54,8 +50,11 @@ void reference_idct_1d_int(const int32_t *in, int32_t *out, int size) {
   double output[64];
   libaom_test::reference_idct_1d(input, output, size);
 
-  for (int i = 0; i < size; ++i)
+  for (int i = 0; i < size; ++i) {
+    ASSERT_GE(output[i], INT32_MIN);
+    ASSERT_LE(output[i], INT32_MAX);
     out[i] = static_cast<int32_t>(round(output[i]));
+  }
 }
 
 void random_matrix(int32_t *dst, int len, ACMRandom *rnd) {
@@ -73,24 +72,32 @@ void random_matrix(int32_t *dst, int len, ACMRandom *rnd) {
 TEST(av1_inv_txfm1d, InvAccuracyCheck) {
   ACMRandom rnd(ACMRandom::DeterministicSeed());
   const int count_test_block = 20000;
-  const int max_error[] = { 6, 10, 19, 28 };
+  const int max_error[] = { 6, 10, 19, 31, 40 };
+  ASSERT_EQ(NELEMENTS(max_error), TX_SIZES);
+  ASSERT_EQ(NELEMENTS(inv_txfm_func_ls), TX_SIZES);
   for (int k = 0; k < count_test_block; ++k) {
     // choose a random transform to test
-    const int txfm_type = rnd.Rand8() % NELEMENTS(inv_txfm_func_ls);
-    const int txfm_size = txfm_size_ls[txfm_type];
-    const TxfmFunc txfm_func = inv_txfm_func_ls[txfm_type][0];
+    const TX_SIZE tx_size = static_cast<TX_SIZE>(rnd.Rand8() % TX_SIZES);
+    const int tx_size_pix = txfm_size_ls[tx_size];
+    const TxfmFunc inv_txfm_func = inv_txfm_func_ls[tx_size][0];
 
     int32_t input[64];
-    random_matrix(input, txfm_size, &rnd);
+    random_matrix(input, tx_size_pix, &rnd);
+
+    // 64x64 transform assumes last 32 values are zero.
+    memset(input + 32, 0, 32 * sizeof(input[0]));
 
     int32_t ref_output[64];
-    reference_idct_1d_int(input, ref_output, txfm_size);
+    reference_idct_1d_int(input, ref_output, tx_size_pix);
 
     int32_t output[64];
-    txfm_func(input, output, cos_bit, range_bit);
+    inv_txfm_func(input, output, cos_bit, range_bit);
 
-    for (int i = 0; i < txfm_size; ++i) {
-      EXPECT_LE(abs(output[i] - ref_output[i]), max_error[txfm_type]);
+    for (int i = 0; i < tx_size_pix; ++i) {
+      EXPECT_LE(abs(output[i] - ref_output[i]), max_error[tx_size])
+          << "tx_size = " << tx_size << ", i = " << i
+          << ", output[i] = " << output[i]
+          << ", ref_output[i] = " << ref_output[i];
     }
   }
 }

third_party/aom/test/av1_inv_txfm2d_test.cc

@@ -12,26 +12,35 @@
 #include <math.h>
 #include <stdio.h>
 #include <stdlib.h>
+#include <vector>
 
-#include "./av1_rtcd.h"
-#include "test/acm_random.h"
-#include "test/util.h"
-#include "test/av1_txfm_test.h"
+#include "config/av1_rtcd.h"
+
+#include "aom_ports/aom_timer.h"
 #include "av1/common/av1_inv_txfm1d_cfg.h"
+#include "av1/common/scan.h"
+#include "test/acm_random.h"
+#include "test/av1_txfm_test.h"
+#include "test/util.h"
 
 using libaom_test::ACMRandom;
-using libaom_test::input_base;
+using libaom_test::InvTxfm2dFunc;
+using libaom_test::LbdInvTxfm2dFunc;
 using libaom_test::bd;
 using libaom_test::compute_avg_abs_error;
-using libaom_test::Fwd_Txfm2d_Func;
-using libaom_test::Inv_Txfm2d_Func;
+using libaom_test::input_base;
+
+using ::testing::Combine;
+using ::testing::Range;
+using ::testing::Values;
+
+using std::vector;
 
 namespace {
 
-#if CONFIG_HIGHBITDEPTH
 // AV1InvTxfm2dParam argument list:
 // tx_type_, tx_size_, max_error_, max_avg_error_
-typedef std::tr1::tuple<TX_TYPE, TX_SIZE, int, double> AV1InvTxfm2dParam;
+typedef ::testing::tuple<TX_TYPE, TX_SIZE, int, double> AV1InvTxfm2dParam;
 
 class AV1InvTxfm2d : public ::testing::TestWithParam<AV1InvTxfm2dParam> {
  public:
@@ -46,171 +55,313 @@ class AV1InvTxfm2d : public ::testing::TestWithParam<AV1InvTxfm2dParam> {
     int tx_w = tx_size_wide[tx_size_];
     int tx_h = tx_size_high[tx_size_];
     int txfm2d_size = tx_w * tx_h;
-    const Fwd_Txfm2d_Func fwd_txfm_func =
-        libaom_test::fwd_txfm_func_ls[tx_size_];
-    const Inv_Txfm2d_Func inv_txfm_func =
-        libaom_test::inv_txfm_func_ls[tx_size_];
+    const FwdTxfm2dFunc fwd_txfm_func = libaom_test::fwd_txfm_func_ls[tx_size_];
+    const InvTxfm2dFunc inv_txfm_func = libaom_test::inv_txfm_func_ls[tx_size_];
     double avg_abs_error = 0;
     ACMRandom rnd(ACMRandom::DeterministicSeed());
 
     const int count = 500;
 
     for (int ci = 0; ci < count; ci++) {
-      int16_t expected[64 * 64] = { 0 };
-      ASSERT_LT(txfm2d_size, NELEMENTS(expected));
+      DECLARE_ALIGNED(16, int16_t, input[64 * 64]) = { 0 };
+      ASSERT_LE(txfm2d_size, NELEMENTS(input));
 
       for (int ni = 0; ni < txfm2d_size; ++ni) {
         if (ci == 0) {
           int extreme_input = input_base - 1;
-          expected[ni] = extreme_input;  // extreme case
+          input[ni] = extreme_input;  // extreme case
         } else {
-          expected[ni] = rnd.Rand16() % input_base;
+          input[ni] = rnd.Rand16() % input_base;
         }
       }
 
-      int32_t coeffs[64 * 64] = { 0 };
-      ASSERT_LT(txfm2d_size, NELEMENTS(coeffs));
-      fwd_txfm_func(expected, coeffs, tx_w, tx_type_, bd);
+      DECLARE_ALIGNED(16, uint16_t, expected[64 * 64]) = { 0 };
+      ASSERT_LE(txfm2d_size, NELEMENTS(expected));
+      if (TxfmUsesApproximation()) {
+        // Compare reference forward HT + inverse HT vs forward HT + inverse HT.
+        double ref_input[64 * 64];
+        ASSERT_LE(txfm2d_size, NELEMENTS(ref_input));
+        for (int ni = 0; ni < txfm2d_size; ++ni) {
+          ref_input[ni] = input[ni];
+        }
+        double ref_coeffs[64 * 64] = { 0 };
+        ASSERT_LE(txfm2d_size, NELEMENTS(ref_coeffs));
+        ASSERT_EQ(tx_type_, DCT_DCT);
+        libaom_test::reference_hybrid_2d(ref_input, ref_coeffs, tx_type_,
+                                         tx_size_);
+        DECLARE_ALIGNED(16, int32_t, ref_coeffs_int[64 * 64]) = { 0 };
+        ASSERT_LE(txfm2d_size, NELEMENTS(ref_coeffs_int));
+        for (int ni = 0; ni < txfm2d_size; ++ni) {
+          ref_coeffs_int[ni] = (int32_t)round(ref_coeffs[ni]);
+        }
+        inv_txfm_func(ref_coeffs_int, expected, tx_w, tx_type_, bd);
+      } else {
+        // Compare original input vs forward HT + inverse HT.
+        for (int ni = 0; ni < txfm2d_size; ++ni) {
+          expected[ni] = input[ni];
+        }
+      }
 
-      uint16_t actual[64 * 64] = { 0 };
-      ASSERT_LT(txfm2d_size, NELEMENTS(actual));
+      DECLARE_ALIGNED(16, int32_t, coeffs[64 * 64]) = { 0 };
+      ASSERT_LE(txfm2d_size, NELEMENTS(coeffs));
+      fwd_txfm_func(input, coeffs, tx_w, tx_type_, bd);
+
+      DECLARE_ALIGNED(16, uint16_t, actual[64 * 64]) = { 0 };
+      ASSERT_LE(txfm2d_size, NELEMENTS(actual));
       inv_txfm_func(coeffs, actual, tx_w, tx_type_, bd);
 
+      double actual_max_error = 0;
       for (int ni = 0; ni < txfm2d_size; ++ni) {
-        EXPECT_GE(max_error_, abs(expected[ni] - actual[ni]));
+        const double this_error = abs(expected[ni] - actual[ni]);
+        actual_max_error = AOMMAX(actual_max_error, this_error);
       }
-      avg_abs_error += compute_avg_abs_error<int16_t, uint16_t>(
+      EXPECT_GE(max_error_, actual_max_error)
+          << " tx_w: " << tx_w << " tx_h " << tx_h << " tx_type: " << tx_type_;
+      if (actual_max_error > max_error_) {  // exit early.
+        break;
+      }
+      avg_abs_error += compute_avg_abs_error<uint16_t, uint16_t>(
           expected, actual, txfm2d_size);
     }
 
     avg_abs_error /= count;
-    // max_abs_avg_error comes from upper bound of
-    // printf("txfm1d_size: %d accuracy_avg_abs_error: %f\n",
-    // txfm1d_size_, avg_abs_error);
     EXPECT_GE(max_avg_error_, avg_abs_error)
         << " tx_w: " << tx_w << " tx_h " << tx_h << " tx_type: " << tx_type_;
   }
 
  private:
+  bool TxfmUsesApproximation() {
+    if (tx_size_wide[tx_size_] == 64 || tx_size_high[tx_size_] == 64) {
+      return true;
+    }
+    return false;
+  }
+
   int max_error_;
   double max_avg_error_;
   TX_TYPE tx_type_;
   TX_SIZE tx_size_;
 };
 
-TEST_P(AV1InvTxfm2d, RunRoundtripCheck) { RunRoundtripCheck(); }
-
-const AV1InvTxfm2dParam av1_inv_txfm2d_param[] = {
-#if CONFIG_EXT_TX
-#if CONFIG_RECT_TX
-  AV1InvTxfm2dParam(DCT_DCT, TX_4X8, 2, 0.007),
-  AV1InvTxfm2dParam(ADST_DCT, TX_4X8, 2, 0.012),
-  AV1InvTxfm2dParam(DCT_ADST, TX_4X8, 2, 0.012),
-  AV1InvTxfm2dParam(ADST_ADST, TX_4X8, 2, 0.012),
-  AV1InvTxfm2dParam(FLIPADST_DCT, TX_4X8, 2, 0.012),
-  AV1InvTxfm2dParam(DCT_FLIPADST, TX_4X8, 2, 0.012),
-  AV1InvTxfm2dParam(FLIPADST_FLIPADST, TX_4X8, 2, 0.012),
-  AV1InvTxfm2dParam(ADST_FLIPADST, TX_4X8, 2, 0.012),
-  AV1InvTxfm2dParam(FLIPADST_ADST, TX_4X8, 2, 0.012),
-
-  AV1InvTxfm2dParam(DCT_DCT, TX_8X4, 2, 0.007),
-  AV1InvTxfm2dParam(ADST_DCT, TX_8X4, 2, 0.012),
-  AV1InvTxfm2dParam(DCT_ADST, TX_8X4, 2, 0.012),
-  AV1InvTxfm2dParam(ADST_ADST, TX_8X4, 2, 0.012),
-  AV1InvTxfm2dParam(FLIPADST_DCT, TX_8X4, 2, 0.007),
-  AV1InvTxfm2dParam(DCT_FLIPADST, TX_8X4, 2, 0.012),
-  AV1InvTxfm2dParam(FLIPADST_FLIPADST, TX_8X4, 2, 0.012),
-  AV1InvTxfm2dParam(ADST_FLIPADST, TX_8X4, 2, 0.012),
-  AV1InvTxfm2dParam(FLIPADST_ADST, TX_8X4, 2, 0.012),
-
-  AV1InvTxfm2dParam(DCT_DCT, TX_8X16, 2, 0.025),
-  AV1InvTxfm2dParam(ADST_DCT, TX_8X16, 2, 0.020),
-  AV1InvTxfm2dParam(DCT_ADST, TX_8X16, 2, 0.027),
-  AV1InvTxfm2dParam(ADST_ADST, TX_8X16, 2, 0.023),
-  AV1InvTxfm2dParam(FLIPADST_DCT, TX_8X16, 2, 0.020),
-  AV1InvTxfm2dParam(DCT_FLIPADST, TX_8X16, 2, 0.027),
-  AV1InvTxfm2dParam(FLIPADST_FLIPADST, TX_8X16, 2, 0.032),
-  AV1InvTxfm2dParam(ADST_FLIPADST, TX_8X16, 2, 0.023),
-  AV1InvTxfm2dParam(FLIPADST_ADST, TX_8X16, 2, 0.023),
-
-  AV1InvTxfm2dParam(DCT_DCT, TX_16X8, 2, 0.007),
-  AV1InvTxfm2dParam(ADST_DCT, TX_16X8, 2, 0.012),
-  AV1InvTxfm2dParam(DCT_ADST, TX_16X8, 2, 0.024),
-  AV1InvTxfm2dParam(ADST_ADST, TX_16X8, 2, 0.033),
-  AV1InvTxfm2dParam(FLIPADST_DCT, TX_16X8, 2, 0.015),
-  AV1InvTxfm2dParam(DCT_FLIPADST, TX_16X8, 2, 0.032),
-  AV1InvTxfm2dParam(FLIPADST_FLIPADST, TX_16X8, 2, 0.032),
-  AV1InvTxfm2dParam(ADST_FLIPADST, TX_16X8, 2, 0.033),
-  AV1InvTxfm2dParam(FLIPADST_ADST, TX_16X8, 2, 0.032),
-#endif
-  AV1InvTxfm2dParam(FLIPADST_DCT, TX_4X4, 2, 0.002),
-  AV1InvTxfm2dParam(DCT_FLIPADST, TX_4X4, 2, 0.002),
-  AV1InvTxfm2dParam(FLIPADST_FLIPADST, TX_4X4, 2, 0.002),
-  AV1InvTxfm2dParam(ADST_FLIPADST, TX_4X4, 2, 0.002),
-  AV1InvTxfm2dParam(FLIPADST_ADST, TX_4X4, 2, 0.002),
-  AV1InvTxfm2dParam(FLIPADST_DCT, TX_8X8, 2, 0.02),
-  AV1InvTxfm2dParam(DCT_FLIPADST, TX_8X8, 2, 0.02),
-  AV1InvTxfm2dParam(FLIPADST_FLIPADST, TX_8X8, 2, 0.02),
-  AV1InvTxfm2dParam(ADST_FLIPADST, TX_8X8, 2, 0.02),
-  AV1InvTxfm2dParam(FLIPADST_ADST, TX_8X8, 2, 0.02),
-  AV1InvTxfm2dParam(FLIPADST_DCT, TX_16X16, 2, 0.04),
-  AV1InvTxfm2dParam(DCT_FLIPADST, TX_16X16, 2, 0.04),
-  AV1InvTxfm2dParam(FLIPADST_FLIPADST, TX_16X16, 11, 0.04),
-  AV1InvTxfm2dParam(ADST_FLIPADST, TX_16X16, 2, 0.04),
-  AV1InvTxfm2dParam(FLIPADST_ADST, TX_16X16, 2, 0.04),
-  AV1InvTxfm2dParam(FLIPADST_DCT, TX_32X32, 4, 0.4),
-  AV1InvTxfm2dParam(DCT_FLIPADST, TX_32X32, 4, 0.4),
-  AV1InvTxfm2dParam(FLIPADST_FLIPADST, TX_32X32, 4, 0.4),
-  AV1InvTxfm2dParam(ADST_FLIPADST, TX_32X32, 4, 0.4),
-  AV1InvTxfm2dParam(FLIPADST_ADST, TX_32X32, 4, 0.4),
-#endif
-  AV1InvTxfm2dParam(DCT_DCT, TX_4X4, 2, 0.002),
-  AV1InvTxfm2dParam(ADST_DCT, TX_4X4, 2, 0.002),
-  AV1InvTxfm2dParam(DCT_ADST, TX_4X4, 2, 0.002),
-  AV1InvTxfm2dParam(ADST_ADST, TX_4X4, 2, 0.002),
-  AV1InvTxfm2dParam(DCT_DCT, TX_8X8, 2, 0.02),
-  AV1InvTxfm2dParam(ADST_DCT, TX_8X8, 2, 0.02),
-  AV1InvTxfm2dParam(DCT_ADST, TX_8X8, 2, 0.02),
-  AV1InvTxfm2dParam(ADST_ADST, TX_8X8, 2, 0.02),
-  AV1InvTxfm2dParam(DCT_DCT, TX_16X16, 2, 0.04),
-  AV1InvTxfm2dParam(ADST_DCT, TX_16X16, 2, 0.04),
-  AV1InvTxfm2dParam(DCT_ADST, TX_16X16, 2, 0.04),
-  AV1InvTxfm2dParam(ADST_ADST, TX_16X16, 2, 0.04),
-  AV1InvTxfm2dParam(DCT_DCT, TX_32X32, 4, 0.4),
-  AV1InvTxfm2dParam(ADST_DCT, TX_32X32, 4, 0.4),
-  AV1InvTxfm2dParam(DCT_ADST, TX_32X32, 4, 0.4),
-  AV1InvTxfm2dParam(ADST_ADST, TX_32X32, 4, 0.4)
+static int max_error_ls[TX_SIZES_ALL] = {
+  2,  // 4x4 transform
+  2,  // 8x8 transform
+  2,  // 16x16 transform
+  4,  // 32x32 transform
+  3,  // 64x64 transform
+  2,  // 4x8 transform
+  2,  // 8x4 transform
+  2,  // 8x16 transform
+  2,  // 16x8 transform
+  3,  // 16x32 transform
+  3,  // 32x16 transform
+  5,  // 32x64 transform
+  5,  // 64x32 transform
+  2,  // 4x16 transform
+  2,  // 16x4 transform
+  2,  // 8x32 transform
+  2,  // 32x8 transform
+  3,  // 16x64 transform
+  3,  // 64x16 transform
 };
 
+static double avg_error_ls[TX_SIZES_ALL] = {
+  0.002,  // 4x4 transform
+  0.05,   // 8x8 transform
+  0.07,   // 16x16 transform
+  0.4,    // 32x32 transform
+  0.3,    // 64x64 transform
+  0.02,   // 4x8 transform
+  0.02,   // 8x4 transform
+  0.04,   // 8x16 transform
+  0.07,   // 16x8 transform
+  0.4,    // 16x32 transform
+  0.5,    // 32x16 transform
+  0.38,   // 32x64 transform
+  0.39,   // 64x32 transform
+  0.2,    // 4x16 transform
+  0.2,    // 16x4 transform
+  0.2,    // 8x32 transform
+  0.2,    // 32x8 transform
+  0.38,   // 16x64 transform
+  0.38,   // 64x16 transform
+};
+
+vector<AV1InvTxfm2dParam> GetInvTxfm2dParamList() {
+  vector<AV1InvTxfm2dParam> param_list;
+  for (int s = 0; s < TX_SIZES; ++s) {
+    const int max_error = max_error_ls[s];
+    const double avg_error = avg_error_ls[s];
+    for (int t = 0; t < TX_TYPES; ++t) {
+      const TX_TYPE tx_type = static_cast<TX_TYPE>(t);
+      const TX_SIZE tx_size = static_cast<TX_SIZE>(s);
+      if (libaom_test::IsTxSizeTypeValid(tx_size, tx_type)) {
+        param_list.push_back(
+            AV1InvTxfm2dParam(tx_type, tx_size, max_error, avg_error));
+      }
+    }
+  }
+  return param_list;
+}
+
 INSTANTIATE_TEST_CASE_P(C, AV1InvTxfm2d,
-                        ::testing::ValuesIn(av1_inv_txfm2d_param));
+                        ::testing::ValuesIn(GetInvTxfm2dParamList()));
+
+TEST_P(AV1InvTxfm2d, RunRoundtripCheck) { RunRoundtripCheck(); }
 
 TEST(AV1InvTxfm2d, CfgTest) {
   for (int bd_idx = 0; bd_idx < BD_NUM; ++bd_idx) {
     int bd = libaom_test::bd_arr[bd_idx];
     int8_t low_range = libaom_test::low_range_arr[bd_idx];
     int8_t high_range = libaom_test::high_range_arr[bd_idx];
-    // TODO(angiebird): include rect txfm in this test
-    for (int tx_size = 0; tx_size < TX_SIZES; ++tx_size) {
+    for (int tx_size = 0; tx_size < TX_SIZES_ALL; ++tx_size) {
       for (int tx_type = 0; tx_type < TX_TYPES; ++tx_type) {
-        TXFM_2D_FLIP_CFG cfg = av1_get_inv_txfm_cfg(
-            static_cast<TX_TYPE>(tx_type), static_cast<TX_SIZE>(tx_size));
+        if (libaom_test::IsTxSizeTypeValid(static_cast<TX_SIZE>(tx_size),
+                                           static_cast<TX_TYPE>(tx_type)) ==
+            false) {
+          continue;
+        }
+        TXFM_2D_FLIP_CFG cfg;
+        av1_get_inv_txfm_cfg(static_cast<TX_TYPE>(tx_type),
+                             static_cast<TX_SIZE>(tx_size), &cfg);
         int8_t stage_range_col[MAX_TXFM_STAGE_NUM];
         int8_t stage_range_row[MAX_TXFM_STAGE_NUM];
         av1_gen_inv_stage_range(stage_range_col, stage_range_row, &cfg,
-                                fwd_shift_sum[tx_size], bd);
-        const TXFM_1D_CFG *col_cfg = cfg.col_cfg;
-        const TXFM_1D_CFG *row_cfg = cfg.row_cfg;
-        libaom_test::txfm_stage_range_check(stage_range_col, col_cfg->stage_num,
-                                            col_cfg->cos_bit, low_range,
+                                (TX_SIZE)tx_size, bd);
+        libaom_test::txfm_stage_range_check(stage_range_col, cfg.stage_num_col,
+                                            cfg.cos_bit_col, low_range,
                                             high_range);
-        libaom_test::txfm_stage_range_check(stage_range_row, row_cfg->stage_num,
-                                            row_cfg->cos_bit, low_range,
+        libaom_test::txfm_stage_range_check(stage_range_row, cfg.stage_num_row,
+                                            cfg.cos_bit_row, low_range,
                                             high_range);
       }
     }
   }
 }
-#endif  // CONFIG_HIGHBITDEPTH
+
+typedef ::testing::tuple<const LbdInvTxfm2dFunc> AV1LbdInvTxfm2dParam;
+class AV1LbdInvTxfm2d : public ::testing::TestWithParam<AV1LbdInvTxfm2dParam> {
+ public:
+  virtual void SetUp() { target_func_ = GET_PARAM(0); }
+  void RunAV1InvTxfm2dTest(TX_TYPE tx_type, TX_SIZE tx_size, int run_times);
+
+ private:
+  LbdInvTxfm2dFunc target_func_;
+};
+
+void AV1LbdInvTxfm2d::RunAV1InvTxfm2dTest(TX_TYPE tx_type, TX_SIZE tx_size,
+                                          int run_times) {
+  FwdTxfm2dFunc fwd_func_ = libaom_test::fwd_txfm_func_ls[tx_size];
+  InvTxfm2dFunc ref_func_ = libaom_test::inv_txfm_func_ls[tx_size];
+  if (fwd_func_ == NULL || ref_func_ == NULL || target_func_ == NULL) {
+    return;
+  }
+  const int bd = 8;
+  const int BLK_WIDTH = 64;
+  const int BLK_SIZE = BLK_WIDTH * BLK_WIDTH;
+  DECLARE_ALIGNED(16, int16_t, input[BLK_SIZE]) = { 0 };
+  DECLARE_ALIGNED(32, int32_t, inv_input[BLK_SIZE]) = { 0 };
+  DECLARE_ALIGNED(16, uint8_t, output[BLK_SIZE]) = { 0 };
+  DECLARE_ALIGNED(16, uint16_t, ref_output[BLK_SIZE]) = { 0 };
+  int stride = BLK_WIDTH;
+  int rows = tx_size_high[tx_size];
+  int cols = tx_size_wide[tx_size];
+  const int rows_nonezero = AOMMIN(32, rows);
+  const int cols_nonezero = AOMMIN(32, cols);
+  run_times /= (rows * cols);
+  run_times = AOMMAX(1, run_times);
+  const SCAN_ORDER *scan_order = get_default_scan(tx_size, tx_type);
+  const int16_t *scan = scan_order->scan;
+  const int16_t eobmax = rows_nonezero * cols_nonezero;
+  ACMRandom rnd(ACMRandom::DeterministicSeed());
+  int randTimes = run_times == 1 ? (eobmax + 500) : 1;
+  for (int cnt = 0; cnt < randTimes; ++cnt) {
+    const int16_t max_in = (1 << (bd)) - 1;
+    for (int r = 0; r < BLK_WIDTH; ++r) {
+      for (int c = 0; c < BLK_WIDTH; ++c) {
+        input[r * cols + c] = (cnt == 0) ? max_in : rnd.Rand8Extremes();
+        output[r * stride + c] = (cnt == 0) ? 128 : rnd.Rand8();
+        ref_output[r * stride + c] = output[r * stride + c];
+      }
+    }
+    fwd_func_(input, inv_input, stride, tx_type, bd);
+
+    // produce eob input by setting high freq coeffs to zero
+    const int eob = AOMMIN(cnt + 1, eobmax);
+    for (int i = eob; i < eobmax; i++) {
+      inv_input[scan[i]] = 0;
+    }
+
+    aom_usec_timer timer;
+    aom_usec_timer_start(&timer);
+    for (int i = 0; i < run_times; ++i) {
+      ref_func_(inv_input, ref_output, stride, tx_type, bd);
+    }
+    aom_usec_timer_mark(&timer);
+    const double time1 = static_cast<double>(aom_usec_timer_elapsed(&timer));
+    aom_usec_timer_start(&timer);
+    for (int i = 0; i < run_times; ++i) {
+      target_func_(inv_input, output, stride, tx_type, tx_size, eob);
+    }
+    aom_usec_timer_mark(&timer);
+    const double time2 = static_cast<double>(aom_usec_timer_elapsed(&timer));
+    if (run_times > 10) {
+      printf("txfm[%d] %3dx%-3d:%7.2f/%7.2fns", tx_type, cols, rows, time1,
+             time2);
+      printf("(%3.2f)\n", time1 / time2);
+    }
+    for (int r = 0; r < rows; ++r) {
+      for (int c = 0; c < cols; ++c) {
+        uint8_t ref_value = static_cast<uint8_t>(ref_output[r * stride + c]);
+        ASSERT_EQ(ref_value, output[r * stride + c])
+            << "[" << r << "," << c << "] " << cnt
+            << " tx_size: " << static_cast<int>(tx_size)
+            << " tx_type: " << tx_type << " eob " << eob;
+      }
+    }
+  }
+}
+
+TEST_P(AV1LbdInvTxfm2d, match) {
+  for (int j = 0; j < (int)(TX_SIZES_ALL); ++j) {
+    for (int i = 0; i < (int)TX_TYPES; ++i) {
+      if (libaom_test::IsTxSizeTypeValid(static_cast<TX_SIZE>(j),
+                                         static_cast<TX_TYPE>(i))) {
+        RunAV1InvTxfm2dTest(static_cast<TX_TYPE>(i), static_cast<TX_SIZE>(j),
+                            1);
+      }
+    }
+  }
+}
+
+TEST_P(AV1LbdInvTxfm2d, DISABLED_Speed) {
+  for (int j = 0; j < (int)(TX_SIZES_ALL); ++j) {
+    for (int i = 0; i < (int)TX_TYPES; ++i) {
+      if (libaom_test::IsTxSizeTypeValid(static_cast<TX_SIZE>(j),
+                                         static_cast<TX_TYPE>(i))) {
+        RunAV1InvTxfm2dTest(static_cast<TX_TYPE>(i), static_cast<TX_SIZE>(j),
+                            10000000);
+      }
+    }
+  }
+}
+
+#if HAVE_SSSE3
+#if defined(_MSC_VER) || defined(__SSSE3__)
+#include "av1/common/x86/av1_inv_txfm_ssse3.h"
+INSTANTIATE_TEST_CASE_P(SSSE3, AV1LbdInvTxfm2d,
+                        ::testing::Values(av1_lowbd_inv_txfm2d_add_ssse3));
+#endif  // _MSC_VER || __SSSE3__
+#endif  // HAVE_SSSE3
+
+#if HAVE_AVX2
+extern "C" void av1_lowbd_inv_txfm2d_add_avx2(const int32_t *input,
+                                              uint8_t *output, int stride,
+                                              TX_TYPE tx_type, TX_SIZE tx_size,
+                                              int eob);
+
+INSTANTIATE_TEST_CASE_P(AVX2, AV1LbdInvTxfm2d,
+                        ::testing::Values(av1_lowbd_inv_txfm2d_add_avx2));
+#endif  // HAVE_AVX2
 
 }  // namespace

third_party/aom/test/av1_inv_txfm_test.cc

@@ -1,271 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
-
-#include <math.h>
-#include <stdlib.h>
-#include <string.h>
-
-#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
-
-#include "./av1_rtcd.h"
-#include "./aom_dsp_rtcd.h"
-#include "test/acm_random.h"
-#include "test/av1_txfm_test.h"
-#include "test/clear_system_state.h"
-#include "test/register_state_check.h"
-#include "test/util.h"
-#include "av1/common/blockd.h"
-#include "av1/common/scan.h"
-#include "aom/aom_integer.h"
-#include "aom_dsp/inv_txfm.h"
-
-using libaom_test::ACMRandom;
-
-namespace {
-
-typedef void (*IdctFunc)(const tran_low_t *in, tran_low_t *out);
-
-class TransTestBase {
- public:
-  virtual ~TransTestBase() {}
-
- protected:
-  void RunInvAccuracyCheck() {
-    tran_low_t input[64];
-    tran_low_t output[64];
-    double ref_input[64];
-    double ref_output[64];
-
-    ACMRandom rnd(ACMRandom::DeterministicSeed());
-    const int count_test_block = 5000;
-    for (int ti = 0; ti < count_test_block; ++ti) {
-      for (int ni = 0; ni < txfm_size_; ++ni) {
-        input[ni] = rnd.Rand8() - rnd.Rand8();
-        ref_input[ni] = static_cast<double>(input[ni]);
-      }
-
-      inv_txfm_(input, output);
-      libaom_test::reference_idct_1d(ref_input, ref_output, txfm_size_);
-
-      for (int ni = 0; ni < txfm_size_; ++ni) {
-        EXPECT_LE(
-            abs(output[ni] - static_cast<tran_low_t>(round(ref_output[ni]))),
-            max_error_);
-      }
-    }
-  }
-
-  double max_error_;
-  int txfm_size_;
-  IdctFunc inv_txfm_;
-};
-
-typedef std::tr1::tuple<IdctFunc, int, int> IdctParam;
-class AV1InvTxfm : public TransTestBase,
-                   public ::testing::TestWithParam<IdctParam> {
- public:
-  virtual void SetUp() {
-    inv_txfm_ = GET_PARAM(0);
-    txfm_size_ = GET_PARAM(1);
-    max_error_ = GET_PARAM(2);
-  }
-  virtual void TearDown() {}
-};
-
-TEST_P(AV1InvTxfm, RunInvAccuracyCheck) { RunInvAccuracyCheck(); }
-
-INSTANTIATE_TEST_CASE_P(C, AV1InvTxfm,
-                        ::testing::Values(IdctParam(&aom_idct4_c, 4, 1),
-                                          IdctParam(&aom_idct8_c, 8, 2),
-                                          IdctParam(&aom_idct16_c, 16, 4),
-                                          IdctParam(&aom_idct32_c, 32, 6)));
-
-#if CONFIG_AV1_ENCODER
-typedef void (*FwdTxfmFunc)(const int16_t *in, tran_low_t *out, int stride);
-typedef void (*InvTxfmFunc)(const tran_low_t *in, uint8_t *out, int stride);
-typedef std::tr1::tuple<FwdTxfmFunc, InvTxfmFunc, InvTxfmFunc, TX_SIZE, int>
-    PartialInvTxfmParam;
-#if !CONFIG_ADAPT_SCAN
-const int kMaxNumCoeffs = 1024;
-#endif
-class AV1PartialIDctTest
-    : public ::testing::TestWithParam<PartialInvTxfmParam> {
- public:
-  virtual ~AV1PartialIDctTest() {}
-  virtual void SetUp() {
-    ftxfm_ = GET_PARAM(0);
-    full_itxfm_ = GET_PARAM(1);
-    partial_itxfm_ = GET_PARAM(2);
-    tx_size_ = GET_PARAM(3);
-    last_nonzero_ = GET_PARAM(4);
-  }
-
-  virtual void TearDown() { libaom_test::ClearSystemState(); }
-
- protected:
-  int last_nonzero_;
-  TX_SIZE tx_size_;
-  FwdTxfmFunc ftxfm_;
-  InvTxfmFunc full_itxfm_;
-  InvTxfmFunc partial_itxfm_;
-};
-
-#if !CONFIG_ADAPT_SCAN
-static MB_MODE_INFO get_mbmi() {
-  MB_MODE_INFO mbmi;
-  mbmi.ref_frame[0] = LAST_FRAME;
-  assert(is_inter_block(&mbmi));
-  return mbmi;
-}
-
-TEST_P(AV1PartialIDctTest, RunQuantCheck) {
-  int size;
-  switch (tx_size_) {
-    case TX_4X4: size = 4; break;
-    case TX_8X8: size = 8; break;
-    case TX_16X16: size = 16; break;
-    case TX_32X32: size = 32; break;
-    default: FAIL() << "Wrong Size!"; break;
-  }
-  DECLARE_ALIGNED(16, tran_low_t, test_coef_block1[kMaxNumCoeffs]);
-  DECLARE_ALIGNED(16, tran_low_t, test_coef_block2[kMaxNumCoeffs]);
-  DECLARE_ALIGNED(16, uint8_t, dst1[kMaxNumCoeffs]);
-  DECLARE_ALIGNED(16, uint8_t, dst2[kMaxNumCoeffs]);
-
-  const int count_test_block = 1000;
-  const int block_size = size * size;
-
-  DECLARE_ALIGNED(16, int16_t, input_extreme_block[kMaxNumCoeffs]);
-  DECLARE_ALIGNED(16, tran_low_t, output_ref_block[kMaxNumCoeffs]);
-
-  int max_error = 0;
-  for (int m = 0; m < count_test_block; ++m) {
-    // clear out destination buffer
-    memset(dst1, 0, sizeof(*dst1) * block_size);
-    memset(dst2, 0, sizeof(*dst2) * block_size);
-    memset(test_coef_block1, 0, sizeof(*test_coef_block1) * block_size);
-    memset(test_coef_block2, 0, sizeof(*test_coef_block2) * block_size);
-
-    ACMRandom rnd(ACMRandom::DeterministicSeed());
-
-    for (int n = 0; n < count_test_block; ++n) {
-      // Initialize a test block with input range [-255, 255].
-      if (n == 0) {
-        for (int j = 0; j < block_size; ++j) input_extreme_block[j] = 255;
-      } else if (n == 1) {
-        for (int j = 0; j < block_size; ++j) input_extreme_block[j] = -255;
-      } else {
-        for (int j = 0; j < block_size; ++j) {
-          input_extreme_block[j] = rnd.Rand8() % 2 ? 255 : -255;
-        }
-      }
-
-      ftxfm_(input_extreme_block, output_ref_block, size);
-
-      // quantization with maximum allowed step sizes
-      test_coef_block1[0] = (output_ref_block[0] / 1336) * 1336;
-      MB_MODE_INFO mbmi = get_mbmi();
-      for (int j = 1; j < last_nonzero_; ++j)
-        test_coef_block1[get_scan((const AV1_COMMON *)NULL, tx_size_, DCT_DCT,
-                                  &mbmi)
-                             ->scan[j]] = (output_ref_block[j] / 1828) * 1828;
-    }
-
-    ASM_REGISTER_STATE_CHECK(full_itxfm_(test_coef_block1, dst1, size));
-    ASM_REGISTER_STATE_CHECK(partial_itxfm_(test_coef_block1, dst2, size));
-
-    for (int j = 0; j < block_size; ++j) {
-      const int diff = dst1[j] - dst2[j];
-      const int error = diff * diff;
-      if (max_error < error) max_error = error;
-    }
-  }
-
-  EXPECT_EQ(0, max_error)
-      << "Error: partial inverse transform produces different results";
-}
-
-TEST_P(AV1PartialIDctTest, ResultsMatch) {
-  ACMRandom rnd(ACMRandom::DeterministicSeed());
-  int size;
-  switch (tx_size_) {
-    case TX_4X4: size = 4; break;
-    case TX_8X8: size = 8; break;
-    case TX_16X16: size = 16; break;
-    case TX_32X32: size = 32; break;
-    default: FAIL() << "Wrong Size!"; break;
-  }
-  DECLARE_ALIGNED(16, tran_low_t, test_coef_block1[kMaxNumCoeffs]);
-  DECLARE_ALIGNED(16, tran_low_t, test_coef_block2[kMaxNumCoeffs]);
-  DECLARE_ALIGNED(16, uint8_t, dst1[kMaxNumCoeffs]);
-  DECLARE_ALIGNED(16, uint8_t, dst2[kMaxNumCoeffs]);
-  const int count_test_block = 1000;
-  const int max_coeff = 32766 / 4;
-  const int block_size = size * size;
-  int max_error = 0;
-  for (int i = 0; i < count_test_block; ++i) {
-    // clear out destination buffer
-    memset(dst1, 0, sizeof(*dst1) * block_size);
-    memset(dst2, 0, sizeof(*dst2) * block_size);
-    memset(test_coef_block1, 0, sizeof(*test_coef_block1) * block_size);
-    memset(test_coef_block2, 0, sizeof(*test_coef_block2) * block_size);
-    int max_energy_leftover = max_coeff * max_coeff;
-    for (int j = 0; j < last_nonzero_; ++j) {
-      int16_t coef = static_cast<int16_t>(sqrt(1.0 * max_energy_leftover) *
-                                          (rnd.Rand16() - 32768) / 65536);
-      max_energy_leftover -= coef * coef;
-      if (max_energy_leftover < 0) {
-        max_energy_leftover = 0;
-        coef = 0;
-      }
-      MB_MODE_INFO mbmi = get_mbmi();
-      test_coef_block1[get_scan((const AV1_COMMON *)NULL, tx_size_, DCT_DCT,
-                                &mbmi)
-                           ->scan[j]] = coef;
-    }
-
-    memcpy(test_coef_block2, test_coef_block1,
-           sizeof(*test_coef_block2) * block_size);
-
-    ASM_REGISTER_STATE_CHECK(full_itxfm_(test_coef_block1, dst1, size));
-    ASM_REGISTER_STATE_CHECK(partial_itxfm_(test_coef_block2, dst2, size));
-
-    for (int j = 0; j < block_size; ++j) {
-      const int diff = dst1[j] - dst2[j];
-      const int error = diff * diff;
-      if (max_error < error) max_error = error;
-    }
-  }
-
-  EXPECT_EQ(0, max_error)
-      << "Error: partial inverse transform produces different results";
-}
-#endif
-using std::tr1::make_tuple;
-
-INSTANTIATE_TEST_CASE_P(
-    C, AV1PartialIDctTest,
-    ::testing::Values(make_tuple(&aom_fdct32x32_c, &aom_idct32x32_1024_add_c,
-                                 &aom_idct32x32_34_add_c, TX_32X32, 34),
-                      make_tuple(&aom_fdct32x32_c, &aom_idct32x32_1024_add_c,
-                                 &aom_idct32x32_1_add_c, TX_32X32, 1),
-                      make_tuple(&aom_fdct16x16_c, &aom_idct16x16_256_add_c,
-                                 &aom_idct16x16_10_add_c, TX_16X16, 10),
-                      make_tuple(&aom_fdct16x16_c, &aom_idct16x16_256_add_c,
-                                 &aom_idct16x16_1_add_c, TX_16X16, 1),
-                      make_tuple(&aom_fdct8x8_c, &aom_idct8x8_64_add_c,
-                                 &aom_idct8x8_12_add_c, TX_8X8, 12),
-                      make_tuple(&aom_fdct8x8_c, &aom_idct8x8_64_add_c,
-                                 &aom_idct8x8_1_add_c, TX_8X8, 1),
-                      make_tuple(&aom_fdct4x4_c, &aom_idct4x4_16_add_c,
-                                 &aom_idct4x4_1_add_c, TX_4X4, 1)));
-#endif  // CONFIG_AV1_ENCODER
-}  // namespace

third_party/aom/test/av1_quantize_test.cc

@@ -12,8 +12,9 @@
 
 #include "third_party/googletest/src/googletest/include/gtest/gtest.h"
 
-#include "./aom_config.h"
-#include "./av1_rtcd.h"
+#include "config/aom_config.h"
+#include "config/av1_rtcd.h"
+
 #include "test/acm_random.h"
 #include "test/clear_system_state.h"
 #include "test/register_state_check.h"
@@ -22,8 +23,8 @@
 namespace {
 
 typedef void (*QuantizeFpFunc)(
-    const tran_low_t *coeff_ptr, intptr_t count, int skip_block,
-    const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr,
+    const tran_low_t *coeff_ptr, intptr_t count, const int16_t *zbin_ptr,
+    const int16_t *round_ptr, const int16_t *quant_ptr,
     const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr,
     tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr,
     const int16_t *scan, const int16_t *iscan, int log_scale);
@@ -50,20 +51,19 @@ class AV1QuantizeTest : public ::testing::TestWithParam<QuantizeFuncParams> {
   void RunQuantizeTest() {
     ACMRandom rnd(ACMRandom::DeterministicSeed());
     DECLARE_ALIGNED(16, tran_low_t, coeff_ptr[maxSize]);
-    DECLARE_ALIGNED(16, int16_t, zbin_ptr[2]);
-    DECLARE_ALIGNED(16, int16_t, round_ptr[2]);
-    DECLARE_ALIGNED(16, int16_t, quant_ptr[2]);
-    DECLARE_ALIGNED(16, int16_t, quant_shift_ptr[2]);
+    DECLARE_ALIGNED(16, int16_t, zbin_ptr[8]);
+    DECLARE_ALIGNED(16, int16_t, round_ptr[8]);
+    DECLARE_ALIGNED(16, int16_t, quant_ptr[8]);
+    DECLARE_ALIGNED(16, int16_t, quant_shift_ptr[8]);
     DECLARE_ALIGNED(16, tran_low_t, qcoeff_ptr[maxSize]);
     DECLARE_ALIGNED(16, tran_low_t, dqcoeff_ptr[maxSize]);
     DECLARE_ALIGNED(16, tran_low_t, ref_qcoeff_ptr[maxSize]);
     DECLARE_ALIGNED(16, tran_low_t, ref_dqcoeff_ptr[maxSize]);
-    DECLARE_ALIGNED(16, int16_t, dequant_ptr[2]);
+    DECLARE_ALIGNED(16, int16_t, dequant_ptr[8]);
     uint16_t eob;
     uint16_t ref_eob;
     int err_count_total = 0;
     int first_failure = -1;
-    int skip_block = 0;
     int count = params_.coeffCount;
     const TX_SIZE txSize = getTxSize(count);
     int log_scale = (txSize == TX_32X32);
@@ -86,20 +86,26 @@ class AV1QuantizeTest : public ::testing::TestWithParam<QuantizeFuncParams> {
         quant_ptr[j] = (1 << 16) / dequant_ptr[j];
         round_ptr[j] = (abs(rnd(roundFactorRange)) * dequant_ptr[j]) >> 7;
       }
-
-      quanFuncRef(coeff_ptr, count, skip_block, zbin_ptr, round_ptr, quant_ptr,
+      for (int j = 2; j < 8; ++j) {
+        zbin_ptr[j] = zbin_ptr[1];
+        quant_shift_ptr[j] = quant_shift_ptr[1];
+        dequant_ptr[j] = dequant_ptr[1];
+        quant_ptr[j] = quant_ptr[1];
+        round_ptr[j] = round_ptr[1];
+      }
+      quanFuncRef(coeff_ptr, count, zbin_ptr, round_ptr, quant_ptr,
                   quant_shift_ptr, ref_qcoeff_ptr, ref_dqcoeff_ptr, dequant_ptr,
                   &ref_eob, scanOrder.scan, scanOrder.iscan, log_scale);
 
       ASM_REGISTER_STATE_CHECK(
-          quanFunc(coeff_ptr, count, skip_block, zbin_ptr, round_ptr, quant_ptr,
+          quanFunc(coeff_ptr, count, zbin_ptr, round_ptr, quant_ptr,
                    quant_shift_ptr, qcoeff_ptr, dqcoeff_ptr, dequant_ptr, &eob,
                    scanOrder.scan, scanOrder.iscan, log_scale));
 
       for (int j = 0; j < count; ++j) {
         err_count += (ref_qcoeff_ptr[j] != qcoeff_ptr[j]) |
                      (ref_dqcoeff_ptr[j] != dqcoeff_ptr[j]);
-        EXPECT_EQ(ref_qcoeff_ptr[j], qcoeff_ptr[j])
+        ASSERT_EQ(ref_qcoeff_ptr[j], qcoeff_ptr[j])
             << "qcoeff error: i = " << i << " j = " << j << "\n";
         EXPECT_EQ(ref_dqcoeff_ptr[j], dqcoeff_ptr[j])
             << "dqcoeff error: i = " << i << " j = " << j << "\n";
@@ -120,18 +126,17 @@ class AV1QuantizeTest : public ::testing::TestWithParam<QuantizeFuncParams> {
   void RunEobTest() {
     ACMRandom rnd(ACMRandom::DeterministicSeed());
     DECLARE_ALIGNED(16, tran_low_t, coeff_ptr[maxSize]);
-    DECLARE_ALIGNED(16, int16_t, zbin_ptr[2]);
-    DECLARE_ALIGNED(16, int16_t, round_ptr[2]);
-    DECLARE_ALIGNED(16, int16_t, quant_ptr[2]);
-    DECLARE_ALIGNED(16, int16_t, quant_shift_ptr[2]);
+    DECLARE_ALIGNED(16, int16_t, zbin_ptr[8]);
+    DECLARE_ALIGNED(16, int16_t, round_ptr[8]);
+    DECLARE_ALIGNED(16, int16_t, quant_ptr[8]);
+    DECLARE_ALIGNED(16, int16_t, quant_shift_ptr[8]);
     DECLARE_ALIGNED(16, tran_low_t, qcoeff_ptr[maxSize]);
     DECLARE_ALIGNED(16, tran_low_t, dqcoeff_ptr[maxSize]);
     DECLARE_ALIGNED(16, tran_low_t, ref_qcoeff_ptr[maxSize]);
     DECLARE_ALIGNED(16, tran_low_t, ref_dqcoeff_ptr[maxSize]);
-    DECLARE_ALIGNED(16, int16_t, dequant_ptr[2]);
+    DECLARE_ALIGNED(16, int16_t, dequant_ptr[8]);
     uint16_t eob;
     uint16_t ref_eob;
-    int skip_block = 0;
     int count = params_.coeffCount;
     const TX_SIZE txSize = getTxSize(count);
     int log_scale = (txSize == TX_32X32);
@@ -157,13 +162,20 @@ class AV1QuantizeTest : public ::testing::TestWithParam<QuantizeFuncParams> {
         quant_ptr[j] = (1 << 16) / dequant_ptr[j];
         round_ptr[j] = (abs(rnd(roundFactorRange)) * dequant_ptr[j]) >> 7;
       }
+      for (int j = 2; j < 8; ++j) {
+        zbin_ptr[j] = zbin_ptr[1];
+        quant_shift_ptr[j] = quant_shift_ptr[1];
+        dequant_ptr[j] = dequant_ptr[1];
+        quant_ptr[j] = quant_ptr[1];
+        round_ptr[j] = round_ptr[1];
+      }
 
-      quanFuncRef(coeff_ptr, count, skip_block, zbin_ptr, round_ptr, quant_ptr,
+      quanFuncRef(coeff_ptr, count, zbin_ptr, round_ptr, quant_ptr,
                   quant_shift_ptr, ref_qcoeff_ptr, ref_dqcoeff_ptr, dequant_ptr,
                   &ref_eob, scanOrder.scan, scanOrder.iscan, log_scale);
 
       ASM_REGISTER_STATE_CHECK(
-          quanFunc(coeff_ptr, count, skip_block, zbin_ptr, round_ptr, quant_ptr,
+          quanFunc(coeff_ptr, count, zbin_ptr, round_ptr, quant_ptr,
                    quant_shift_ptr, qcoeff_ptr, dqcoeff_ptr, dequant_ptr, &eob,
                    scanOrder.scan, scanOrder.iscan, log_scale));
       EXPECT_EQ(ref_eob, eob) << "eob error: "
@@ -196,7 +208,7 @@ TEST_P(AV1QuantizeTest, EobVerify) { RunEobTest(); }
 
 #if HAVE_SSE4_1
 const QuantizeFuncParams qfps[4] = {
-  QuantizeFuncParams(av1_highbd_quantize_fp_sse4_1, &av1_highbd_quantize_fp_c,
+  QuantizeFuncParams(&av1_highbd_quantize_fp_sse4_1, &av1_highbd_quantize_fp_c,
                      16),
   QuantizeFuncParams(&av1_highbd_quantize_fp_sse4_1, &av1_highbd_quantize_fp_c,
                      64),
@@ -208,4 +220,20 @@ const QuantizeFuncParams qfps[4] = {
 
 INSTANTIATE_TEST_CASE_P(SSE4_1, AV1QuantizeTest, ::testing::ValuesIn(qfps));
 #endif  // HAVE_SSE4_1
+
+#if HAVE_AVX2
+const QuantizeFuncParams qfps_avx2[4] = {
+  QuantizeFuncParams(&av1_highbd_quantize_fp_avx2, &av1_highbd_quantize_fp_c,
+                     16),
+  QuantizeFuncParams(&av1_highbd_quantize_fp_avx2, &av1_highbd_quantize_fp_c,
+                     64),
+  QuantizeFuncParams(&av1_highbd_quantize_fp_avx2, &av1_highbd_quantize_fp_c,
+                     256),
+  QuantizeFuncParams(&av1_highbd_quantize_fp_avx2, &av1_highbd_quantize_fp_c,
+                     1024),
+};
+
+INSTANTIATE_TEST_CASE_P(AVX2, AV1QuantizeTest, ::testing::ValuesIn(qfps_avx2));
+#endif  // HAVE_AVX2
+
 }  // namespace

third_party/aom/test/av1_round_shift_array_test.cc

@@ -0,0 +1,127 @@
+/*
+ * Copyright (c) 2018, Alliance for Open Media. All rights reserved
+ *
+ * This source code is subject to the terms of the BSD 2 Clause License and
+ * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
+ * was not distributed with this source code in the LICENSE file, you can
+ * obtain it at www.aomedia.org/license/software. If the Alliance for Open
+ * Media Patent License 1.0 was not distributed with this source code in the
+ * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
+ */
+
+#include <math.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+#include "config/aom_dsp_rtcd.h"
+
+#include "aom_mem/aom_mem.h"
+#include "aom_ports/aom_timer.h"
+#include "aom_ports/mem.h"
+#include "test/acm_random.h"
+#include "test/clear_system_state.h"
+#include "test/util.h"
+#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
+
+namespace AV1CompRoundShift {
+
+typedef void (*comp_round_shift_array_func)(int32_t *arr, int size, int bit);
+
+const int kValidBitCheck[] = {
+  -4, -3, -2, -1, 0, 1, 2, 3, 4,
+};
+
+typedef ::testing::tuple<comp_round_shift_array_func, BLOCK_SIZE, int>
+    CompRoundShiftParam;
+
+class AV1CompRoundShiftTest
+    : public ::testing::TestWithParam<CompRoundShiftParam> {
+ public:
+  ~AV1CompRoundShiftTest();
+
+  void SetUp() { rnd_.Reset(libaom_test::ACMRandom::DeterministicSeed()); }
+  void TearDown() { libaom_test::ClearSystemState(); }
+
+ protected:
+  void RunCheckOutput(comp_round_shift_array_func test_impl, BLOCK_SIZE bsize,
+                      int bit);
+  void RunSpeedTest(comp_round_shift_array_func test_impl, BLOCK_SIZE bsize,
+                    int bit);
+
+  libaom_test::ACMRandom rnd_;
+};
+
+AV1CompRoundShiftTest::~AV1CompRoundShiftTest() { ; }
+
+void AV1CompRoundShiftTest::RunCheckOutput(
+    comp_round_shift_array_func test_impl, BLOCK_SIZE bsize, int bit) {
+  const int w = block_size_wide[bsize];
+  const int h = block_size_high[bsize];
+  const int blk_wd = 64;
+  DECLARE_ALIGNED(32, int32_t, pred_[blk_wd]);
+  DECLARE_ALIGNED(32, int32_t, ref_buffer_[blk_wd]);
+  for (int i = 0; i < (blk_wd); ++i) {
+    ref_buffer_[i] = pred_[i] = rnd_.Rand31() / 16;
+  }
+  av1_round_shift_array_c(ref_buffer_, w, bit);
+  test_impl(pred_, w, bit);
+  for (int x = 0; x < w; ++x) {
+    ASSERT_EQ(ref_buffer_[x], pred_[x]) << w << "x" << h << "mismatch @"
+                                        << "(" << x << ")";
+  }
+}
+
+void AV1CompRoundShiftTest::RunSpeedTest(comp_round_shift_array_func test_impl,
+                                         BLOCK_SIZE bsize, int bit) {
+  const int w = block_size_wide[bsize];
+  const int h = block_size_high[bsize];
+  const int blk_wd = 64;
+  DECLARE_ALIGNED(32, int32_t, ref_buffer_[blk_wd]);
+  for (int i = 0; i < (blk_wd); ++i) {
+    ref_buffer_[i] = rnd_.Rand31();
+  }
+
+  const int num_loops = 1000000000 / (w + h);
+  comp_round_shift_array_func funcs[2] = { av1_round_shift_array_c, test_impl };
+  double elapsed_time[2] = { 0 };
+  for (int i = 0; i < 2; ++i) {
+    aom_usec_timer timer;
+    aom_usec_timer_start(&timer);
+    comp_round_shift_array_func func = funcs[i];
+    for (int j = 0; j < num_loops; ++j) {
+      func(ref_buffer_, w, bit);
+    }
+    aom_usec_timer_mark(&timer);
+    double time = static_cast<double>(aom_usec_timer_elapsed(&timer));
+    elapsed_time[i] = 1000.0 * time / num_loops;
+  }
+  printf("av1_round_shift_array %3dx%-3d: bit : %d %7.2f/%7.2fns", w, h, bit,
+         elapsed_time[0], elapsed_time[1]);
+  printf("(%3.2f)\n", elapsed_time[0] / elapsed_time[1]);
+}
+
+TEST_P(AV1CompRoundShiftTest, CheckOutput) {
+  RunCheckOutput(GET_PARAM(0), GET_PARAM(1), GET_PARAM(2));
+}
+
+TEST_P(AV1CompRoundShiftTest, DISABLED_Speed) {
+  RunSpeedTest(GET_PARAM(0), GET_PARAM(1), GET_PARAM(2));
+}
+
+#if HAVE_SSE4_1
+INSTANTIATE_TEST_CASE_P(
+    SSE4_1, AV1CompRoundShiftTest,
+    ::testing::Combine(::testing::Values(&av1_round_shift_array_sse4_1),
+                       ::testing::ValuesIn(txsize_to_bsize),
+                       ::testing::ValuesIn(kValidBitCheck)));
+#endif
+
+#if HAVE_NEON
+INSTANTIATE_TEST_CASE_P(
+    NEON, AV1CompRoundShiftTest,
+    ::testing::Combine(::testing::Values(&av1_round_shift_array_neon),
+                       ::testing::ValuesIn(txsize_to_bsize),
+                       ::testing::ValuesIn(kValidBitCheck)));
+#endif
+
+};  // namespace AV1CompRoundShift

third_party/aom/test/av1_txfm_test.cc

@@ -34,7 +34,6 @@ void get_txfm1d_type(TX_TYPE txfm2d_type, TYPE_TXFM *type0, TYPE_TXFM *type1) {
       *type0 = TYPE_ADST;
       *type1 = TYPE_ADST;
       break;
-#if CONFIG_EXT_TX
     case FLIPADST_DCT:
       *type0 = TYPE_ADST;
       *type1 = TYPE_DCT;
@@ -55,7 +54,34 @@ void get_txfm1d_type(TX_TYPE txfm2d_type, TYPE_TXFM *type0, TYPE_TXFM *type1) {
       *type0 = TYPE_ADST;
       *type1 = TYPE_ADST;
       break;
-#endif  // CONFIG_EXT_TX
+    case IDTX:
+      *type0 = TYPE_IDTX;
+      *type1 = TYPE_IDTX;
+      break;
+    case H_DCT:
+      *type0 = TYPE_IDTX;
+      *type1 = TYPE_DCT;
+      break;
+    case V_DCT:
+      *type0 = TYPE_DCT;
+      *type1 = TYPE_IDTX;
+      break;
+    case H_ADST:
+      *type0 = TYPE_IDTX;
+      *type1 = TYPE_ADST;
+      break;
+    case V_ADST:
+      *type0 = TYPE_ADST;
+      *type1 = TYPE_IDTX;
+      break;
+    case H_FLIPADST:
+      *type0 = TYPE_IDTX;
+      *type1 = TYPE_ADST;
+      break;
+    case V_FLIPADST:
+      *type0 = TYPE_ADST;
+      *type1 = TYPE_IDTX;
+      break;
     default:
       *type0 = TYPE_DCT;
       *type1 = TYPE_DCT;
@@ -64,6 +90,7 @@ void get_txfm1d_type(TX_TYPE txfm2d_type, TYPE_TXFM *type0, TYPE_TXFM *type1) {
   }
 }
 
+double Sqrt2 = pow(2, 0.5);
 double invSqrt2 = 1 / pow(2, 0.5);
 
 double dct_matrix(double n, double k, int size) {
@@ -92,7 +119,63 @@ void reference_idct_1d(const double *in, double *out, int size) {
   }
 }
 
+// TODO(any): Copied from the old 'fadst4' (same as the new 'av1_fadst4_new'
+// function). Should be replaced by a proper reference function that takes
+// 'double' input & output.
+static void fadst4_new(const tran_low_t *input, tran_low_t *output) {
+  tran_high_t x0, x1, x2, x3;
+  tran_high_t s0, s1, s2, s3, s4, s5, s6, s7;
+
+  x0 = input[0];
+  x1 = input[1];
+  x2 = input[2];
+  x3 = input[3];
+
+  if (!(x0 | x1 | x2 | x3)) {
+    output[0] = output[1] = output[2] = output[3] = 0;
+    return;
+  }
+
+  s0 = sinpi_1_9 * x0;
+  s1 = sinpi_4_9 * x0;
+  s2 = sinpi_2_9 * x1;
+  s3 = sinpi_1_9 * x1;
+  s4 = sinpi_3_9 * x2;
+  s5 = sinpi_4_9 * x3;
+  s6 = sinpi_2_9 * x3;
+  s7 = x0 + x1 - x3;
+
+  x0 = s0 + s2 + s5;
+  x1 = sinpi_3_9 * s7;
+  x2 = s1 - s3 + s6;
+  x3 = s4;
+
+  s0 = x0 + x3;
+  s1 = x1;
+  s2 = x2 - x3;
+  s3 = x2 - x0 + x3;
+
+  // 1-D transform scaling factor is sqrt(2).
+  output[0] = (tran_low_t)fdct_round_shift(s0);
+  output[1] = (tran_low_t)fdct_round_shift(s1);
+  output[2] = (tran_low_t)fdct_round_shift(s2);
+  output[3] = (tran_low_t)fdct_round_shift(s3);
+}
+
 void reference_adst_1d(const double *in, double *out, int size) {
+  if (size == 4) {  // Special case.
+    tran_low_t int_input[4];
+    for (int i = 0; i < 4; ++i) {
+      int_input[i] = static_cast<tran_low_t>(round(in[i]));
+    }
+    tran_low_t int_output[4];
+    fadst4_new(int_input, int_output);
+    for (int i = 0; i < 4; ++i) {
+      out[i] = int_output[i];
+    }
+    return;
+  }
+
   for (int k = 0; k < size; ++k) {
     out[k] = 0;
     for (int n = 0; n < size; ++n) {
@@ -101,96 +184,188 @@ void reference_adst_1d(const double *in, double *out, int size) {
   }
 }
 
+void reference_idtx_1d(const double *in, double *out, int size) {
+  double scale = 0;
+  if (size == 4)
+    scale = Sqrt2;
+  else if (size == 8)
+    scale = 2;
+  else if (size == 16)
+    scale = 2 * Sqrt2;
+  else if (size == 32)
+    scale = 4;
+  else if (size == 64)
+    scale = 4 * Sqrt2;
+  for (int k = 0; k < size; ++k) {
+    out[k] = in[k] * scale;
+  }
+}
+
 void reference_hybrid_1d(double *in, double *out, int size, int type) {
   if (type == TYPE_DCT)
     reference_dct_1d(in, out, size);
-  else
+  else if (type == TYPE_ADST)
     reference_adst_1d(in, out, size);
+  else
+    reference_idtx_1d(in, out, size);
 }
 
-void reference_hybrid_2d(double *in, double *out, int size, int type0,
-                         int type1) {
-  double *tempOut = new double[size * size];
+double get_amplification_factor(TX_TYPE tx_type, TX_SIZE tx_size) {
+  TXFM_2D_FLIP_CFG fwd_txfm_flip_cfg;
+  av1_get_fwd_txfm_cfg(tx_type, tx_size, &fwd_txfm_flip_cfg);
+  const int tx_width = tx_size_wide[fwd_txfm_flip_cfg.tx_size];
+  const int tx_height = tx_size_high[fwd_txfm_flip_cfg.tx_size];
+  const int8_t *shift = fwd_txfm_flip_cfg.shift;
+  const int amplify_bit = shift[0] + shift[1] + shift[2];
+  double amplify_factor =
+      amplify_bit >= 0 ? (1 << amplify_bit) : (1.0 / (1 << -amplify_bit));
 
-  for (int r = 0; r < size; r++) {
-    // out ->tempOut
-    for (int c = 0; c < size; c++) {
-      tempOut[r * size + c] = in[c * size + r];
+  // For rectangular transforms, we need to multiply by an extra factor.
+  const int rect_type = get_rect_tx_log_ratio(tx_width, tx_height);
+  if (abs(rect_type) == 1) {
+    amplify_factor *= pow(2, 0.5);
+  }
+  return amplify_factor;
+}
+
+void reference_hybrid_2d(double *in, double *out, TX_TYPE tx_type,
+                         TX_SIZE tx_size) {
+  // Get transform type and size of each dimension.
+  TYPE_TXFM type0;
+  TYPE_TXFM type1;
+  get_txfm1d_type(tx_type, &type0, &type1);
+  const int tx_width = tx_size_wide[tx_size];
+  const int tx_height = tx_size_high[tx_size];
+
+  double *const temp_in = new double[AOMMAX(tx_width, tx_height)];
+  double *const temp_out = new double[AOMMAX(tx_width, tx_height)];
+  double *const out_interm = new double[tx_width * tx_height];
+  const int stride = tx_width;
+
+  // Transform columns.
+  for (int c = 0; c < tx_width; ++c) {
+    for (int r = 0; r < tx_height; ++r) {
+      temp_in[r] = in[r * stride + c];
+    }
+    reference_hybrid_1d(temp_in, temp_out, tx_height, type0);
+    for (int r = 0; r < tx_height; ++r) {
+      out_interm[r * stride + c] = temp_out[r];
     }
   }
 
-  // dct each row: in -> out
-  for (int r = 0; r < size; r++) {
-    reference_hybrid_1d(tempOut + r * size, out + r * size, size, type0);
+  // Transform rows.
+  for (int r = 0; r < tx_height; ++r) {
+    reference_hybrid_1d(out_interm + r * stride, out + r * stride, tx_width,
+                        type1);
   }
 
-  for (int r = 0; r < size; r++) {
-    // out ->tempOut
-    for (int c = 0; c < size; c++) {
-      tempOut[r * size + c] = out[c * size + r];
+  delete[] temp_in;
+  delete[] temp_out;
+  delete[] out_interm;
+
+  // These transforms use an approximate 2D DCT transform, by only keeping the
+  // top-left quarter of the coefficients, and repacking them in the first
+  // quarter indices.
+  // TODO(urvang): Refactor this code.
+  if (tx_width == 64 && tx_height == 64) {  // tx_size == TX_64X64
+    // Zero out top-right 32x32 area.
+    for (int row = 0; row < 32; ++row) {
+      memset(out + row * 64 + 32, 0, 32 * sizeof(*out));
+    }
+    // Zero out the bottom 64x32 area.
+    memset(out + 32 * 64, 0, 32 * 64 * sizeof(*out));
+    // Re-pack non-zero coeffs in the first 32x32 indices.
+    for (int row = 1; row < 32; ++row) {
+      memcpy(out + row * 32, out + row * 64, 32 * sizeof(*out));
+    }
+  } else if (tx_width == 32 && tx_height == 64) {  // tx_size == TX_32X64
+    // Zero out the bottom 32x32 area.
+    memset(out + 32 * 32, 0, 32 * 32 * sizeof(*out));
+    // Note: no repacking needed here.
+  } else if (tx_width == 64 && tx_height == 32) {  // tx_size == TX_64X32
+    // Zero out right 32x32 area.
+    for (int row = 0; row < 32; ++row) {
+      memset(out + row * 64 + 32, 0, 32 * sizeof(*out));
+    }
+    // Re-pack non-zero coeffs in the first 32x32 indices.
+    for (int row = 1; row < 32; ++row) {
+      memcpy(out + row * 32, out + row * 64, 32 * sizeof(*out));
+    }
+  } else if (tx_width == 16 && tx_height == 64) {  // tx_size == TX_16X64
+    // Zero out the bottom 16x32 area.
+    memset(out + 16 * 32, 0, 16 * 32 * sizeof(*out));
+    // Note: no repacking needed here.
+  } else if (tx_width == 64 && tx_height == 16) {  // tx_size == TX_64X16
+    // Zero out right 32x16 area.
+    for (int row = 0; row < 16; ++row) {
+      memset(out + row * 64 + 32, 0, 32 * sizeof(*out));
+    }
+    // Re-pack non-zero coeffs in the first 32x16 indices.
+    for (int row = 1; row < 16; ++row) {
+      memcpy(out + row * 32, out + row * 64, 32 * sizeof(*out));
     }
   }
 
-  for (int r = 0; r < size; r++) {
-    reference_hybrid_1d(tempOut + r * size, out + r * size, size, type1);
-  }
-  delete[] tempOut;
-}
-
-template <typename Type>
-void fliplr(Type *dest, int stride, int length) {
-  int i, j;
-  for (i = 0; i < length; ++i) {
-    for (j = 0; j < length / 2; ++j) {
-      const Type tmp = dest[i * stride + j];
-      dest[i * stride + j] = dest[i * stride + length - 1 - j];
-      dest[i * stride + length - 1 - j] = tmp;
+  // Apply appropriate scale.
+  const double amplify_factor = get_amplification_factor(tx_type, tx_size);
+  for (int c = 0; c < tx_width; ++c) {
+    for (int r = 0; r < tx_height; ++r) {
+      out[r * stride + c] *= amplify_factor;
     }
   }
 }
 
 template <typename Type>
-void flipud(Type *dest, int stride, int length) {
-  int i, j;
-  for (j = 0; j < length; ++j) {
-    for (i = 0; i < length / 2; ++i) {
-      const Type tmp = dest[i * stride + j];
-      dest[i * stride + j] = dest[(length - 1 - i) * stride + j];
-      dest[(length - 1 - i) * stride + j] = tmp;
+void fliplr(Type *dest, int width, int height, int stride) {
+  for (int r = 0; r < height; ++r) {
+    for (int c = 0; c < width / 2; ++c) {
+      const Type tmp = dest[r * stride + c];
+      dest[r * stride + c] = dest[r * stride + width - 1 - c];
+      dest[r * stride + width - 1 - c] = tmp;
     }
   }
 }
 
 template <typename Type>
-void fliplrud(Type *dest, int stride, int length) {
-  int i, j;
-  for (i = 0; i < length / 2; ++i) {
-    for (j = 0; j < length; ++j) {
-      const Type tmp = dest[i * stride + j];
-      dest[i * stride + j] = dest[(length - 1 - i) * stride + length - 1 - j];
-      dest[(length - 1 - i) * stride + length - 1 - j] = tmp;
+void flipud(Type *dest, int width, int height, int stride) {
+  for (int c = 0; c < width; ++c) {
+    for (int r = 0; r < height / 2; ++r) {
+      const Type tmp = dest[r * stride + c];
+      dest[r * stride + c] = dest[(height - 1 - r) * stride + c];
+      dest[(height - 1 - r) * stride + c] = tmp;
     }
   }
 }
 
-template void fliplr<double>(double *dest, int stride, int length);
-template void flipud<double>(double *dest, int stride, int length);
-template void fliplrud<double>(double *dest, int stride, int length);
+template <typename Type>
+void fliplrud(Type *dest, int width, int height, int stride) {
+  for (int r = 0; r < height / 2; ++r) {
+    for (int c = 0; c < width; ++c) {
+      const Type tmp = dest[r * stride + c];
+      dest[r * stride + c] = dest[(height - 1 - r) * stride + width - 1 - c];
+      dest[(height - 1 - r) * stride + width - 1 - c] = tmp;
+    }
+  }
+}
+
+template void fliplr<double>(double *dest, int width, int height, int stride);
+template void flipud<double>(double *dest, int width, int height, int stride);
+template void fliplrud<double>(double *dest, int width, int height, int stride);
 
 int bd_arr[BD_NUM] = { 8, 10, 12 };
-int8_t low_range_arr[BD_NUM] = { 16, 32, 32 };
+
+int8_t low_range_arr[BD_NUM] = { 18, 32, 32 };
 int8_t high_range_arr[BD_NUM] = { 32, 32, 32 };
 
 void txfm_stage_range_check(const int8_t *stage_range, int stage_num,
-                            const int8_t *cos_bit, int low_range,
-                            int high_range) {
+                            int8_t cos_bit, int low_range, int high_range) {
   for (int i = 0; i < stage_num; ++i) {
     EXPECT_LE(stage_range[i], low_range);
+    ASSERT_LE(stage_range[i] + cos_bit, high_range) << "stage = " << i;
   }
   for (int i = 0; i < stage_num - 1; ++i) {
     // make sure there is no overflow while doing half_btf()
-    EXPECT_LE(stage_range[i] + cos_bit[i], high_range);
-    EXPECT_LE(stage_range[i + 1] + cos_bit[i], high_range);
+    ASSERT_LE(stage_range[i + 1] + cos_bit, high_range) << "stage = " << i;
   }
 }
 }  // namespace libaom_test

third_party/aom/test/av1_txfm_test.h

@@ -19,17 +19,20 @@
 #endif
 #include <math.h>
 
+#include "config/av1_rtcd.h"
+
 #include "third_party/googletest/src/googletest/include/gtest/gtest.h"
 
 #include "test/acm_random.h"
-#include "av1/common/enums.h"
 #include "av1/common/av1_txfm.h"
-#include "./av1_rtcd.h"
+#include "av1/common/blockd.h"
+#include "av1/common/enums.h"
 
 namespace libaom_test {
 typedef enum {
   TYPE_DCT = 0,
   TYPE_ADST,
+  TYPE_IDTX,
   TYPE_IDCT,
   TYPE_IADST,
   TYPE_LAST
@@ -46,8 +49,10 @@ void reference_adst_1d(const double *in, double *out, int size);
 
 void reference_hybrid_1d(double *in, double *out, int size, int type);
 
-void reference_hybrid_2d(double *in, double *out, int size, int type0,
-                         int type1);
+double get_amplification_factor(TX_TYPE tx_type, TX_SIZE tx_size);
+
+void reference_hybrid_2d(double *in, double *out, TX_TYPE tx_type,
+                         TX_SIZE tx_size);
 template <typename Type1, typename Type2>
 static double compute_avg_abs_error(const Type1 *a, const Type2 *b,
                                     const int size) {
@@ -60,81 +65,62 @@ static double compute_avg_abs_error(const Type1 *a, const Type2 *b,
 }
 
 template <typename Type>
-void fliplr(Type *dest, int stride, int length);
+void fliplr(Type *dest, int width, int height, int stride);
 
 template <typename Type>
-void flipud(Type *dest, int stride, int length);
+void flipud(Type *dest, int width, int height, int stride);
 
 template <typename Type>
-void fliplrud(Type *dest, int stride, int length);
+void fliplrud(Type *dest, int width, int height, int stride);
 
-typedef void (*TxfmFunc)(const int32_t *in, int32_t *out, const int8_t *cos_bit,
+typedef void (*TxfmFunc)(const int32_t *in, int32_t *out, const int8_t cos_bit,
                          const int8_t *range_bit);
 
-typedef void (*Fwd_Txfm2d_Func)(const int16_t *, int32_t *, int, TX_TYPE, int);
-typedef void (*Inv_Txfm2d_Func)(const int32_t *, uint16_t *, int, TX_TYPE, int);
+typedef void (*InvTxfm2dFunc)(const int32_t *, uint16_t *, int, TX_TYPE, int);
+typedef void (*LbdInvTxfm2dFunc)(const int32_t *, uint8_t *, int, TX_TYPE,
+                                 TX_SIZE, int);
 
 static const int bd = 10;
 static const int input_base = (1 << bd);
 
-#if CONFIG_HIGHBITDEPTH
+static INLINE bool IsTxSizeTypeValid(TX_SIZE tx_size, TX_TYPE tx_type) {
+  const TX_SIZE tx_size_sqr_up = txsize_sqr_up_map[tx_size];
+  TxSetType tx_set_type;
+  if (tx_size_sqr_up > TX_32X32) {
+    tx_set_type = EXT_TX_SET_DCTONLY;
+  } else if (tx_size_sqr_up == TX_32X32) {
+    tx_set_type = EXT_TX_SET_DCT_IDTX;
+  } else {
+    tx_set_type = EXT_TX_SET_ALL16;
+  }
+  return av1_ext_tx_used[tx_set_type][tx_type] != 0;
+}
+
 #if CONFIG_AV1_ENCODER
 
-static const Fwd_Txfm2d_Func fwd_txfm_func_ls[TX_SIZES_ALL] = {
-#if CONFIG_CHROMA_2X2
-  NULL,
-#endif
-  av1_fwd_txfm2d_4x4_c,
-  av1_fwd_txfm2d_8x8_c,
-  av1_fwd_txfm2d_16x16_c,
-  av1_fwd_txfm2d_32x32_c,
-#if CONFIG_TX64X64
-  av1_fwd_txfm2d_64x64_c,
-#endif  // CONFIG_TX64X64
-  av1_fwd_txfm2d_4x8_c,
-  av1_fwd_txfm2d_8x4_c,
-  av1_fwd_txfm2d_8x16_c,
-  av1_fwd_txfm2d_16x8_c,
-  av1_fwd_txfm2d_16x32_c,
-  av1_fwd_txfm2d_32x16_c,
-#if CONFIG_TX64X64
-  av1_fwd_txfm2d_32x64_c,
-  av1_fwd_txfm2d_64x32_c,
-#endif  // CONFIG_TX64X64
-  NULL,
-  NULL,
-  NULL,
-  NULL,
+static const FwdTxfm2dFunc fwd_txfm_func_ls[TX_SIZES_ALL] = {
+  av1_fwd_txfm2d_4x4_c,   av1_fwd_txfm2d_8x8_c,   av1_fwd_txfm2d_16x16_c,
+  av1_fwd_txfm2d_32x32_c, av1_fwd_txfm2d_64x64_c, av1_fwd_txfm2d_4x8_c,
+  av1_fwd_txfm2d_8x4_c,   av1_fwd_txfm2d_8x16_c,  av1_fwd_txfm2d_16x8_c,
+  av1_fwd_txfm2d_16x32_c, av1_fwd_txfm2d_32x16_c, av1_fwd_txfm2d_32x64_c,
+  av1_fwd_txfm2d_64x32_c, av1_fwd_txfm2d_4x16_c,  av1_fwd_txfm2d_16x4_c,
+  av1_fwd_txfm2d_8x32_c,  av1_fwd_txfm2d_32x8_c,  av1_fwd_txfm2d_16x64_c,
+  av1_fwd_txfm2d_64x16_c,
 };
 #endif
 
-static const Inv_Txfm2d_Func inv_txfm_func_ls[TX_SIZES_ALL] = {
-#if CONFIG_CHROMA_2X2
-  NULL,
-#endif
-  av1_inv_txfm2d_add_4x4_c,
-  av1_inv_txfm2d_add_8x8_c,
-  av1_inv_txfm2d_add_16x16_c,
-  av1_inv_txfm2d_add_32x32_c,
-#if CONFIG_TX64X64
-  av1_inv_txfm2d_add_64x64_c,
-#endif  // CONFIG_TX64X64
-  av1_inv_txfm2d_add_4x8_c,
-  av1_inv_txfm2d_add_8x4_c,
-  av1_inv_txfm2d_add_8x16_c,
-  av1_inv_txfm2d_add_16x8_c,
-  av1_inv_txfm2d_add_16x32_c,
-  av1_inv_txfm2d_add_32x16_c,
-#if CONFIG_TX64X64
-  av1_inv_txfm2d_add_32x64_c,
-  av1_inv_txfm2d_add_64x32_c,
-#endif  // CONFIG_TX64X64
-  NULL,
-  NULL,
-  NULL,
-  NULL,
+static const InvTxfm2dFunc inv_txfm_func_ls[TX_SIZES_ALL] = {
+  av1_inv_txfm2d_add_4x4_c,   av1_inv_txfm2d_add_8x8_c,
+  av1_inv_txfm2d_add_16x16_c, av1_inv_txfm2d_add_32x32_c,
+  av1_inv_txfm2d_add_64x64_c, av1_inv_txfm2d_add_4x8_c,
+  av1_inv_txfm2d_add_8x4_c,   av1_inv_txfm2d_add_8x16_c,
+  av1_inv_txfm2d_add_16x8_c,  av1_inv_txfm2d_add_16x32_c,
+  av1_inv_txfm2d_add_32x16_c, av1_inv_txfm2d_add_32x64_c,
+  av1_inv_txfm2d_add_64x32_c, av1_inv_txfm2d_add_4x16_c,
+  av1_inv_txfm2d_add_16x4_c,  av1_inv_txfm2d_add_8x32_c,
+  av1_inv_txfm2d_add_32x8_c,  av1_inv_txfm2d_add_16x64_c,
+  av1_inv_txfm2d_add_64x16_c,
 };
-#endif  // CONFIG_HIGHBITDEPTH
 
 #define BD_NUM 3
 
@@ -143,7 +129,7 @@ extern int8_t low_range_arr[];
 extern int8_t high_range_arr[];
 
 void txfm_stage_range_check(const int8_t *stage_range, int stage_num,
-                            const int8_t *cos_bit, int low_range,
+                            const int8_t cos_bit, int low_range,
                             int high_range);
 }  // namespace libaom_test
 #endif  // AV1_TXFM_TEST_H_

third_party/aom/test/av1_wedge_utils_test.cc

@@ -11,10 +11,9 @@
 
 #include "third_party/googletest/src/googletest/include/gtest/gtest.h"
 
-#include "./aom_config.h"
-
-#include "./aom_dsp_rtcd.h"
-#include "./av1_rtcd.h"
+#include "config/aom_config.h"
+#include "config/aom_dsp_rtcd.h"
+#include "config/av1_rtcd.h"
 
 #include "aom_dsp/aom_dsp_common.h"
 
@@ -100,7 +99,7 @@ TEST_F(WedgeUtilsSSEFuncTest, ResidualBlendingEquiv) {
       p1[j] = clamp(s[j] + rng_(33) - 16, 0, UINT8_MAX);
     }
 
-    aom_blend_a64_mask(p, w, p0, w, p1, w, m, w, h, w, 0, 0);
+    aom_blend_a64_mask(p, w, p0, w, p1, w, m, w, w, h, 0, 0);
 
     aom_subtract_block(h, w, r0, w, s, w, p0, w);
     aom_subtract_block(h, w, r1, w, s, w, p1, w);

third_party/aom/test/avg_test.cc

@@ -1,299 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
-
-#include <limits.h>
-#include <stdio.h>
-#include <string.h>
-
-#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
-
-#include "./aom_config.h"
-#include "./aom_dsp_rtcd.h"
-
-#include "test/acm_random.h"
-#include "test/clear_system_state.h"
-#include "test/register_state_check.h"
-#include "test/util.h"
-#include "aom_mem/aom_mem.h"
-
-using libaom_test::ACMRandom;
-
-namespace {
-class AverageTestBase : public ::testing::Test {
- public:
-  AverageTestBase(int width, int height) : width_(width), height_(height) {}
-
-  static void SetUpTestCase() {
-    source_data_ = reinterpret_cast<uint8_t *>(
-        aom_memalign(kDataAlignment, kDataBlockSize));
-  }
-
-  static void TearDownTestCase() {
-    aom_free(source_data_);
-    source_data_ = NULL;
-  }
-
-  virtual void TearDown() { libaom_test::ClearSystemState(); }
-
- protected:
-  // Handle blocks up to 4 blocks 64x64 with stride up to 128
-  static const int kDataAlignment = 16;
-  static const int kDataBlockSize = 64 * 128;
-
-  virtual void SetUp() {
-    source_stride_ = (width_ + 31) & ~31;
-    rnd_.Reset(ACMRandom::DeterministicSeed());
-  }
-
-  void FillConstant(uint8_t fill_constant) {
-    for (int i = 0; i < width_ * height_; ++i) {
-      source_data_[i] = fill_constant;
-    }
-  }
-
-  void FillRandom() {
-    for (int i = 0; i < width_ * height_; ++i) {
-      source_data_[i] = rnd_.Rand8();
-    }
-  }
-
-  int width_, height_;
-  static uint8_t *source_data_;
-  int source_stride_;
-
-  ACMRandom rnd_;
-};
-
-typedef void (*IntProRowFunc)(int16_t hbuf[16], uint8_t const *ref,
-                              const int ref_stride, const int height);
-
-typedef std::tr1::tuple<int, IntProRowFunc, IntProRowFunc> IntProRowParam;
-
-class IntProRowTest : public AverageTestBase,
-                      public ::testing::WithParamInterface<IntProRowParam> {
- public:
-  IntProRowTest()
-      : AverageTestBase(16, GET_PARAM(0)), hbuf_asm_(NULL), hbuf_c_(NULL) {
-    asm_func_ = GET_PARAM(1);
-    c_func_ = GET_PARAM(2);
-  }
-
- protected:
-  virtual void SetUp() {
-    hbuf_asm_ = reinterpret_cast<int16_t *>(
-        aom_memalign(kDataAlignment, sizeof(*hbuf_asm_) * 16));
-    hbuf_c_ = reinterpret_cast<int16_t *>(
-        aom_memalign(kDataAlignment, sizeof(*hbuf_c_) * 16));
-  }
-
-  virtual void TearDown() {
-    aom_free(hbuf_c_);
-    hbuf_c_ = NULL;
-    aom_free(hbuf_asm_);
-    hbuf_asm_ = NULL;
-  }
-
-  void RunComparison() {
-    ASM_REGISTER_STATE_CHECK(c_func_(hbuf_c_, source_data_, 0, height_));
-    ASM_REGISTER_STATE_CHECK(asm_func_(hbuf_asm_, source_data_, 0, height_));
-    EXPECT_EQ(0, memcmp(hbuf_c_, hbuf_asm_, sizeof(*hbuf_c_) * 16))
-        << "Output mismatch";
-  }
-
- private:
-  IntProRowFunc asm_func_;
-  IntProRowFunc c_func_;
-  int16_t *hbuf_asm_;
-  int16_t *hbuf_c_;
-};
-
-typedef int16_t (*IntProColFunc)(uint8_t const *ref, const int width);
-
-typedef std::tr1::tuple<int, IntProColFunc, IntProColFunc> IntProColParam;
-
-class IntProColTest : public AverageTestBase,
-                      public ::testing::WithParamInterface<IntProColParam> {
- public:
-  IntProColTest() : AverageTestBase(GET_PARAM(0), 1), sum_asm_(0), sum_c_(0) {
-    asm_func_ = GET_PARAM(1);
-    c_func_ = GET_PARAM(2);
-  }
-
- protected:
-  void RunComparison() {
-    ASM_REGISTER_STATE_CHECK(sum_c_ = c_func_(source_data_, width_));
-    ASM_REGISTER_STATE_CHECK(sum_asm_ = asm_func_(source_data_, width_));
-    EXPECT_EQ(sum_c_, sum_asm_) << "Output mismatch";
-  }
-
- private:
-  IntProColFunc asm_func_;
-  IntProColFunc c_func_;
-  int16_t sum_asm_;
-  int16_t sum_c_;
-};
-
-typedef int (*SatdFunc)(const int16_t *coeffs, int length);
-typedef std::tr1::tuple<int, SatdFunc> SatdTestParam;
-
-class SatdTest : public ::testing::Test,
-                 public ::testing::WithParamInterface<SatdTestParam> {
- protected:
-  virtual void SetUp() {
-    satd_size_ = GET_PARAM(0);
-    satd_func_ = GET_PARAM(1);
-    rnd_.Reset(ACMRandom::DeterministicSeed());
-    src_ = reinterpret_cast<int16_t *>(
-        aom_memalign(16, sizeof(*src_) * satd_size_));
-    ASSERT_TRUE(src_ != NULL);
-  }
-
-  virtual void TearDown() {
-    libaom_test::ClearSystemState();
-    aom_free(src_);
-  }
-
-  void FillConstant(const int16_t val) {
-    for (int i = 0; i < satd_size_; ++i) src_[i] = val;
-  }
-
-  void FillRandom() {
-    for (int i = 0; i < satd_size_; ++i) src_[i] = rnd_.Rand16();
-  }
-
-  void Check(int expected) {
-    int total;
-    ASM_REGISTER_STATE_CHECK(total = satd_func_(src_, satd_size_));
-    EXPECT_EQ(expected, total);
-  }
-
-  int satd_size_;
-
- private:
-  int16_t *src_;
-  SatdFunc satd_func_;
-  ACMRandom rnd_;
-};
-
-uint8_t *AverageTestBase::source_data_ = NULL;
-
-TEST_P(IntProRowTest, MinValue) {
-  FillConstant(0);
-  RunComparison();
-}
-
-TEST_P(IntProRowTest, MaxValue) {
-  FillConstant(255);
-  RunComparison();
-}
-
-TEST_P(IntProRowTest, Random) {
-  FillRandom();
-  RunComparison();
-}
-
-TEST_P(IntProColTest, MinValue) {
-  FillConstant(0);
-  RunComparison();
-}
-
-TEST_P(IntProColTest, MaxValue) {
-  FillConstant(255);
-  RunComparison();
-}
-
-TEST_P(IntProColTest, Random) {
-  FillRandom();
-  RunComparison();
-}
-
-TEST_P(SatdTest, MinValue) {
-  const int kMin = -32640;
-  const int expected = -kMin * satd_size_;
-  FillConstant(kMin);
-  Check(expected);
-}
-
-TEST_P(SatdTest, MaxValue) {
-  const int kMax = 32640;
-  const int expected = kMax * satd_size_;
-  FillConstant(kMax);
-  Check(expected);
-}
-
-TEST_P(SatdTest, Random) {
-  int expected;
-  switch (satd_size_) {
-    case 16: expected = 205298; break;
-    case 64: expected = 1113950; break;
-    case 256: expected = 4268415; break;
-    case 1024: expected = 16954082; break;
-    default:
-      FAIL() << "Invalid satd size (" << satd_size_
-             << ") valid: 16/64/256/1024";
-  }
-  FillRandom();
-  Check(expected);
-}
-
-using std::tr1::make_tuple;
-
-INSTANTIATE_TEST_CASE_P(C, SatdTest,
-                        ::testing::Values(make_tuple(16, &aom_satd_c),
-                                          make_tuple(64, &aom_satd_c),
-                                          make_tuple(256, &aom_satd_c),
-                                          make_tuple(1024, &aom_satd_c)));
-
-#if HAVE_SSE2
-INSTANTIATE_TEST_CASE_P(
-    SSE2, IntProRowTest,
-    ::testing::Values(make_tuple(16, &aom_int_pro_row_sse2, &aom_int_pro_row_c),
-                      make_tuple(32, &aom_int_pro_row_sse2, &aom_int_pro_row_c),
-                      make_tuple(64, &aom_int_pro_row_sse2,
-                                 &aom_int_pro_row_c)));
-
-INSTANTIATE_TEST_CASE_P(
-    SSE2, IntProColTest,
-    ::testing::Values(make_tuple(16, &aom_int_pro_col_sse2, &aom_int_pro_col_c),
-                      make_tuple(32, &aom_int_pro_col_sse2, &aom_int_pro_col_c),
-                      make_tuple(64, &aom_int_pro_col_sse2,
-                                 &aom_int_pro_col_c)));
-
-INSTANTIATE_TEST_CASE_P(SSE2, SatdTest,
-                        ::testing::Values(make_tuple(16, &aom_satd_sse2),
-                                          make_tuple(64, &aom_satd_sse2),
-                                          make_tuple(256, &aom_satd_sse2),
-                                          make_tuple(1024, &aom_satd_sse2)));
-#endif
-
-#if HAVE_NEON
-INSTANTIATE_TEST_CASE_P(
-    NEON, IntProRowTest,
-    ::testing::Values(make_tuple(16, &aom_int_pro_row_neon, &aom_int_pro_row_c),
-                      make_tuple(32, &aom_int_pro_row_neon, &aom_int_pro_row_c),
-                      make_tuple(64, &aom_int_pro_row_neon,
-                                 &aom_int_pro_row_c)));
-
-INSTANTIATE_TEST_CASE_P(
-    NEON, IntProColTest,
-    ::testing::Values(make_tuple(16, &aom_int_pro_col_neon, &aom_int_pro_col_c),
-                      make_tuple(32, &aom_int_pro_col_neon, &aom_int_pro_col_c),
-                      make_tuple(64, &aom_int_pro_col_neon,
-                                 &aom_int_pro_col_c)));
-
-INSTANTIATE_TEST_CASE_P(NEON, SatdTest,
-                        ::testing::Values(make_tuple(16, &aom_satd_neon),
-                                          make_tuple(64, &aom_satd_neon),
-                                          make_tuple(256, &aom_satd_neon),
-                                          make_tuple(1024, &aom_satd_neon)));
-#endif
-
-}  // namespace

third_party/aom/test/best_encode.sh

@@ -0,0 +1,103 @@
+#!/bin/bash
+#
+# Copyright (c) 2016, Alliance for Open Media. All rights reserved
+#
+# This source code is subject to the terms of the BSD 2 Clause License and
+# the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
+# was not distributed with this source code in the LICENSE file, you can
+# obtain it at www.aomedia.org/license/software. If the Alliance for Open
+# Media Patent License 1.0 was not distributed with this source code in the
+# PATENTS file, you can obtain it at www.aomedia.org/license/patent.
+#
+# Author: jimbankoski@google.com (Jim Bankoski)
+
+if [[ $# -ne 2 ]]; then
+  echo "Encodes a file using best known settings (slow!)"
+  echo "  Usage:    be [FILE] [BITRATE]"
+  echo "  Example:  be akiyo_cif.y4m 200"
+  exit
+fi
+
+f=$1  # file is first parameter
+b=$2  # bitrate is second parameter
+
+if [[ -e $f.fpf ]]; then
+  # First-pass file found, do second pass only
+  aomenc \
+    $f \
+    -o $f-$b.av1.webm \
+    -p 2 \
+    --pass=2 \
+    --fpf=$f.fpf \
+    --best \
+    --cpu-used=0 \
+    --target-bitrate=$b \
+    --auto-alt-ref=1 \
+    -v \
+    --minsection-pct=0 \
+    --maxsection-pct=800 \
+    --lag-in-frames=25 \
+    --kf-min-dist=0 \
+    --kf-max-dist=99999 \
+    --static-thresh=0 \
+    --min-q=0 \
+    --max-q=63 \
+    --drop-frame=0 \
+    --bias-pct=50 \
+    --minsection-pct=0 \
+    --maxsection-pct=800 \
+    --psnr \
+    --arnr-maxframes=7 \
+    --arnr-strength=3 \
+    --arnr-type=3
+else
+  # No first-pass file found, do 2-pass encode
+  aomenc \
+    $f \
+    -o $f-$b.av1.webm \
+    -p 2 \
+    --pass=1 \
+    --fpf=$f.fpf \
+    --best \
+    --cpu-used=0 \
+    --target-bitrate=$b \
+    --auto-alt-ref=1 \
+    -v \
+    --minsection-pct=0 \
+    --maxsection-pct=800 \
+    --lag-in-frames=25 \
+    --kf-min-dist=0 \
+    --kf-max-dist=99999 \
+    --static-thresh=0 \
+    --min-q=0 \
+    --max-q=63 \
+    --drop-frame=0
+
+  aomenc \
+    $f \
+    -o $f-$b.av1.webm \
+    -p 2 \
+    --pass=2 \
+    --fpf=$f.fpf \
+    --best \
+    --cpu-used=0 \
+    --target-bitrate=$b \
+    --auto-alt-ref=1 \
+    -v \
+    --minsection-pct=0 \
+    --maxsection-pct=800 \
+    --lag-in-frames=25 \
+    --kf-min-dist=0 \
+    --kf-max-dist=99999 \
+    --static-thresh=0 \
+    --min-q=0 \
+    --max-q=63 \
+    --drop-frame=0 \
+    --bias-pct=50 \
+    --minsection-pct=0 \
+    --maxsection-pct=800 \
+    --psnr \
+    --arnr-maxframes=7 \
+    --arnr-strength=3 \
+    --arnr-type=3
+fi

third_party/aom/test/binary_codes_test.cc

@@ -7,7 +7,7 @@
  * obtain it at www.aomedia.org/license/software. If the Alliance for Open
  * Media Patent License 1.0 was not distributed with this source code in the
  * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
+ */
 
 #include <math.h>
 #include <stdlib.h>
@@ -15,7 +15,8 @@
 
 #include "third_party/googletest/src/googletest/include/gtest/gtest.h"
 
-#include "./aom_config.h"
+#include "config/aom_config.h"
+
 #include "test/acm_random.h"
 #include "aom/aom_integer.h"
 #include "aom_dsp/bitreader.h"
@@ -29,57 +30,6 @@ using libaom_test::ACMRandom;
 
 namespace {
 
-// Test for Bilevel code with reference
-TEST(AV1, TestPrimitiveRefbilivel) {
-  ACMRandom rnd(ACMRandom::DeterministicSeed());
-  const int kBufferSize = 65536;
-  aom_writer bw;
-  uint8_t bw_buffer[kBufferSize];
-  const uint16_t kRanges = 8;
-  const uint16_t kNearRanges = 8;
-  const uint16_t kReferences = 8;
-  const uint16_t kValues = 16;
-  const uint16_t range_vals[kRanges] = { 1, 13, 64, 120, 230, 420, 1100, 8000 };
-  uint16_t enc_values[kRanges][kNearRanges][kReferences][kValues][4];
-  aom_start_encode(&bw, bw_buffer);
-  for (int n = 0; n < kRanges; ++n) {
-    const uint16_t range = range_vals[n];
-    for (int p = 0; p < kNearRanges; ++p) {
-      const uint16_t near_range = 1 + rnd(range);
-      for (int r = 0; r < kReferences; ++r) {
-        const uint16_t ref = rnd(range);
-        for (int v = 0; v < kValues; ++v) {
-          const uint16_t value = rnd(range);
-          enc_values[n][p][r][v][0] = range;
-          enc_values[n][p][r][v][1] = near_range;
-          enc_values[n][p][r][v][2] = ref;
-          enc_values[n][p][r][v][3] = value;
-          aom_write_primitive_refbilevel(&bw, range, near_range, ref, value);
-        }
-      }
-    }
-  }
-  aom_stop_encode(&bw);
-  aom_reader br;
-  aom_reader_init(&br, bw_buffer, bw.pos, NULL, NULL);
-  GTEST_ASSERT_GE(aom_reader_tell(&br), 0u);
-  GTEST_ASSERT_LE(aom_reader_tell(&br), 1u);
-  for (int n = 0; n < kRanges; ++n) {
-    for (int p = 0; p < kNearRanges; ++p) {
-      for (int r = 0; r < kReferences; ++r) {
-        for (int v = 0; v < kValues; ++v) {
-          const uint16_t range = enc_values[n][p][r][v][0];
-          const uint16_t near_range = enc_values[n][p][r][v][1];
-          const uint16_t ref = enc_values[n][p][r][v][2];
-          const uint16_t value = aom_read_primitive_refbilevel(
-              &br, range, near_range, ref, ACCT_STR);
-          GTEST_ASSERT_EQ(value, enc_values[n][p][r][v][3]);
-        }
-      }
-    }
-  }
-}
-
 // Test for Finite subexponential code with reference
 TEST(AV1, TestPrimitiveRefsubexpfin) {
   ACMRandom rnd(ACMRandom::DeterministicSeed());
@@ -111,7 +61,7 @@ TEST(AV1, TestPrimitiveRefsubexpfin) {
   }
   aom_stop_encode(&bw);
   aom_reader br;
-  aom_reader_init(&br, bw_buffer, bw.pos, NULL, NULL);
+  aom_reader_init(&br, bw_buffer, bw.pos);
   GTEST_ASSERT_GE(aom_reader_tell(&br), 0u);
   GTEST_ASSERT_LE(aom_reader_tell(&br), 1u);
   for (int n = 0; n < kRanges; ++n) {

third_party/aom/test/blend_a64_mask_1d_test.cc

@@ -17,11 +17,11 @@
 #include "test/register_state_check.h"
 #include "test/function_equivalence_test.h"
 
-#include "./aom_config.h"
-#include "./aom_dsp_rtcd.h"
-#include "aom/aom_integer.h"
+#include "config/aom_config.h"
+#include "config/aom_dsp_rtcd.h"
+#include "config/av1_rtcd.h"
 
-#include "./av1_rtcd.h"
+#include "aom/aom_integer.h"
 
 #include "av1/common/enums.h"
 
@@ -46,8 +46,8 @@ class BlendA64Mask1DTest : public FunctionEquivalenceTest<F> {
   virtual void Execute(const T *p_src0, const T *p_src1) = 0;
 
   void Common() {
-    w_ = 1 << this->rng_(MAX_SB_SIZE_LOG2 + 1);
-    h_ = 1 << this->rng_(MAX_SB_SIZE_LOG2 + 1);
+    w_ = 2 << this->rng_(MAX_SB_SIZE_LOG2);
+    h_ = 2 << this->rng_(MAX_SB_SIZE_LOG2);
 
     dst_offset_ = this->rng_(33);
     dst_stride_ = this->rng_(kMaxWidth + 1 - w_) + w_;
@@ -116,7 +116,7 @@ class BlendA64Mask1DTest : public FunctionEquivalenceTest<F> {
 
 typedef void (*F8B)(uint8_t *dst, uint32_t dst_stride, const uint8_t *src0,
                     uint32_t src0_stride, const uint8_t *src1,
-                    uint32_t src1_stride, const uint8_t *mask, int h, int w);
+                    uint32_t src1_stride, const uint8_t *mask, int w, int h);
 typedef libaom_test::FuncParam<F8B> TestFuncs;
 
 class BlendA64Mask1DTest8B : public BlendA64Mask1DTest<F8B, uint8_t> {
@@ -124,10 +124,10 @@ class BlendA64Mask1DTest8B : public BlendA64Mask1DTest<F8B, uint8_t> {
   void Execute(const uint8_t *p_src0, const uint8_t *p_src1) {
     params_.ref_func(dst_ref_ + dst_offset_, dst_stride_, p_src0 + src0_offset_,
                      src0_stride_, p_src1 + src1_offset_, src1_stride_, mask_,
-                     h_, w_);
+                     w_, h_);
     ASM_REGISTER_STATE_CHECK(params_.tst_func(
         dst_tst_ + dst_offset_, dst_stride_, p_src0 + src0_offset_,
-        src0_stride_, p_src1 + src1_offset_, src1_stride_, mask_, h_, w_));
+        src0_stride_, p_src1 + src1_offset_, src1_stride_, mask_, w_, h_));
   }
 };
 
@@ -167,7 +167,7 @@ TEST_P(BlendA64Mask1DTest8B, ExtremeValues) {
 static void blend_a64_hmask_ref(uint8_t *dst, uint32_t dst_stride,
                                 const uint8_t *src0, uint32_t src0_stride,
                                 const uint8_t *src1, uint32_t src1_stride,
-                                const uint8_t *mask, int h, int w) {
+                                const uint8_t *mask, int w, int h) {
   uint8_t mask2d[BlendA64Mask1DTest8B::kMaxMaskSize]
                 [BlendA64Mask1DTest8B::kMaxMaskSize];
 
@@ -175,14 +175,14 @@ static void blend_a64_hmask_ref(uint8_t *dst, uint32_t dst_stride,
     for (int col = 0; col < w; ++col) mask2d[row][col] = mask[col];
 
   aom_blend_a64_mask_c(dst, dst_stride, src0, src0_stride, src1, src1_stride,
-                       &mask2d[0][0], BlendA64Mask1DTest8B::kMaxMaskSize, h, w,
+                       &mask2d[0][0], BlendA64Mask1DTest8B::kMaxMaskSize, w, h,
                        0, 0);
 }
 
 static void blend_a64_vmask_ref(uint8_t *dst, uint32_t dst_stride,
                                 const uint8_t *src0, uint32_t src0_stride,
                                 const uint8_t *src1, uint32_t src1_stride,
-                                const uint8_t *mask, int h, int w) {
+                                const uint8_t *mask, int w, int h) {
   uint8_t mask2d[BlendA64Mask1DTest8B::kMaxMaskSize]
                 [BlendA64Mask1DTest8B::kMaxMaskSize];
 
@@ -190,7 +190,7 @@ static void blend_a64_vmask_ref(uint8_t *dst, uint32_t dst_stride,
     for (int col = 0; col < w; ++col) mask2d[row][col] = mask[row];
 
   aom_blend_a64_mask_c(dst, dst_stride, src0, src0_stride, src1, src1_stride,
-                       &mask2d[0][0], BlendA64Mask1DTest8B::kMaxMaskSize, h, w,
+                       &mask2d[0][0], BlendA64Mask1DTest8B::kMaxMaskSize, w, h,
                        0, 0);
 }
 
@@ -207,14 +207,21 @@ INSTANTIATE_TEST_CASE_P(
         TestFuncs(blend_a64_vmask_ref, aom_blend_a64_vmask_sse4_1)));
 #endif  // HAVE_SSE4_1
 
-#if CONFIG_HIGHBITDEPTH
+#if HAVE_NEON
+INSTANTIATE_TEST_CASE_P(NEON, BlendA64Mask1DTest8B,
+                        ::testing::Values(TestFuncs(blend_a64_hmask_ref,
+                                                    aom_blend_a64_hmask_neon),
+                                          TestFuncs(blend_a64_vmask_ref,
+                                                    aom_blend_a64_vmask_neon)));
+#endif  // HAVE_NEON
+
 //////////////////////////////////////////////////////////////////////////////
 // High bit-depth version
 //////////////////////////////////////////////////////////////////////////////
 
 typedef void (*FHBD)(uint8_t *dst, uint32_t dst_stride, const uint8_t *src0,
                      uint32_t src0_stride, const uint8_t *src1,
-                     uint32_t src1_stride, const uint8_t *mask, int h, int w,
+                     uint32_t src1_stride, const uint8_t *mask, int w, int h,
                      int bd);
 typedef libaom_test::FuncParam<FHBD> TestFuncsHBD;
 
@@ -224,11 +231,11 @@ class BlendA64Mask1DTestHBD : public BlendA64Mask1DTest<FHBD, uint16_t> {
     params_.ref_func(CONVERT_TO_BYTEPTR(dst_ref_ + dst_offset_), dst_stride_,
                      CONVERT_TO_BYTEPTR(p_src0 + src0_offset_), src0_stride_,
                      CONVERT_TO_BYTEPTR(p_src1 + src1_offset_), src1_stride_,
-                     mask_, h_, w_, bit_depth_);
+                     mask_, w_, h_, bit_depth_);
     ASM_REGISTER_STATE_CHECK(params_.tst_func(
         CONVERT_TO_BYTEPTR(dst_tst_ + dst_offset_), dst_stride_,
         CONVERT_TO_BYTEPTR(p_src0 + src0_offset_), src0_stride_,
-        CONVERT_TO_BYTEPTR(p_src1 + src1_offset_), src1_stride_, mask_, h_, w_,
+        CONVERT_TO_BYTEPTR(p_src1 + src1_offset_), src1_stride_, mask_, w_, h_,
         bit_depth_));
   }
 
@@ -287,7 +294,7 @@ TEST_P(BlendA64Mask1DTestHBD, ExtremeValues) {
 static void highbd_blend_a64_hmask_ref(
     uint8_t *dst, uint32_t dst_stride, const uint8_t *src0,
     uint32_t src0_stride, const uint8_t *src1, uint32_t src1_stride,
-    const uint8_t *mask, int h, int w, int bd) {
+    const uint8_t *mask, int w, int h, int bd) {
   uint8_t mask2d[BlendA64Mask1DTestHBD::kMaxMaskSize]
                 [BlendA64Mask1DTestHBD::kMaxMaskSize];
 
@@ -296,13 +303,13 @@ static void highbd_blend_a64_hmask_ref(
 
   aom_highbd_blend_a64_mask_c(
       dst, dst_stride, src0, src0_stride, src1, src1_stride, &mask2d[0][0],
-      BlendA64Mask1DTestHBD::kMaxMaskSize, h, w, 0, 0, bd);
+      BlendA64Mask1DTestHBD::kMaxMaskSize, w, h, 0, 0, bd);
 }
 
 static void highbd_blend_a64_vmask_ref(
     uint8_t *dst, uint32_t dst_stride, const uint8_t *src0,
     uint32_t src0_stride, const uint8_t *src1, uint32_t src1_stride,
-    const uint8_t *mask, int h, int w, int bd) {
+    const uint8_t *mask, int w, int h, int bd) {
   uint8_t mask2d[BlendA64Mask1DTestHBD::kMaxMaskSize]
                 [BlendA64Mask1DTestHBD::kMaxMaskSize];
 
@@ -311,7 +318,7 @@ static void highbd_blend_a64_vmask_ref(
 
   aom_highbd_blend_a64_mask_c(
       dst, dst_stride, src0, src0_stride, src1, src1_stride, &mask2d[0][0],
-      BlendA64Mask1DTestHBD::kMaxMaskSize, h, w, 0, 0, bd);
+      BlendA64Mask1DTestHBD::kMaxMaskSize, w, h, 0, 0, bd);
 }
 
 INSTANTIATE_TEST_CASE_P(
@@ -329,6 +336,4 @@ INSTANTIATE_TEST_CASE_P(
                       TestFuncsHBD(highbd_blend_a64_vmask_ref,
                                    aom_highbd_blend_a64_vmask_sse4_1)));
 #endif  // HAVE_SSE4_1
-
-#endif  // CONFIG_HIGHBITDEPTH
 }  // namespace

third_party/aom/test/blend_a64_mask_test.cc

@@ -17,11 +17,11 @@
 #include "test/register_state_check.h"
 #include "test/function_equivalence_test.h"
 
-#include "./aom_config.h"
-#include "./aom_dsp_rtcd.h"
-#include "aom/aom_integer.h"
+#include "config/aom_config.h"
+#include "config/aom_dsp_rtcd.h"
+#include "config/av1_rtcd.h"
 
-#include "./av1_rtcd.h"
+#include "aom/aom_integer.h"
 
 #include "av1/common/enums.h"
 
@@ -31,8 +31,8 @@ using libaom_test::FunctionEquivalenceTest;
 
 namespace {
 
-template <typename F, typename T>
-class BlendA64MaskTest : public FunctionEquivalenceTest<F> {
+template <typename BlendA64Func, typename SrcPixel, typename DstPixel>
+class BlendA64MaskTest : public FunctionEquivalenceTest<BlendA64Func> {
  protected:
   static const int kIterations = 10000;
   static const int kMaxWidth = MAX_SB_SIZE * 5;  // * 5 to cover longer strides
@@ -43,14 +43,44 @@ class BlendA64MaskTest : public FunctionEquivalenceTest<F> {
 
   virtual ~BlendA64MaskTest() {}
 
-  virtual void Execute(const T *p_src0, const T *p_src1) = 0;
+  virtual void Execute(const SrcPixel *p_src0, const SrcPixel *p_src1) = 0;
 
-  void Common() {
-    w_ = 1 << this->rng_(MAX_SB_SIZE_LOG2 + 1);
-    h_ = 1 << this->rng_(MAX_SB_SIZE_LOG2 + 1);
+  template <typename Pixel>
+  void GetSources(Pixel **src0, Pixel **src1, Pixel * /*dst*/) {
+    switch (this->rng_(3)) {
+      case 0:  // Separate sources
+        *src0 = src0_;
+        *src1 = src1_;
+        break;
+      case 1:  // src0 == dst
+        *src0 = dst_tst_;
+        src0_stride_ = dst_stride_;
+        src0_offset_ = dst_offset_;
+        *src1 = src1_;
+        break;
+      case 2:  // src1 == dst
+        *src0 = src0_;
+        *src1 = dst_tst_;
+        src1_stride_ = dst_stride_;
+        src1_offset_ = dst_offset_;
+        break;
+      default: FAIL();
+    }
+  }
 
-    subx_ = this->rng_(2);
-    suby_ = this->rng_(2);
+  void GetSources(uint16_t **src0, uint16_t **src1, uint8_t * /*dst*/) {
+    *src0 = src0_;
+    *src1 = src1_;
+  }
+
+  uint8_t Rand1() { return this->rng_.Rand8() & 1; }
+
+  void RunTest() {
+    w_ = 4 << this->rng_(MAX_SB_SIZE_LOG2 - 1);
+    h_ = 4 << this->rng_(MAX_SB_SIZE_LOG2 - 1);
+
+    subx_ = Rand1();
+    suby_ = Rand1();
 
     dst_offset_ = this->rng_(33);
     dst_stride_ = this->rng_(kMaxWidth + 1 - w_) + w_;
@@ -64,49 +94,35 @@ class BlendA64MaskTest : public FunctionEquivalenceTest<F> {
     mask_stride_ =
         this->rng_(kMaxWidth + 1 - w_ * (subx_ ? 2 : 1)) + w_ * (subx_ ? 2 : 1);
 
-    T *p_src0;
-    T *p_src1;
+    SrcPixel *p_src0;
+    SrcPixel *p_src1;
 
-    switch (this->rng_(3)) {
-      case 0:  // Separate sources
-        p_src0 = src0_;
-        p_src1 = src1_;
-        break;
-      case 1:  // src0 == dst
-        p_src0 = dst_tst_;
-        src0_stride_ = dst_stride_;
-        src0_offset_ = dst_offset_;
-        p_src1 = src1_;
-        break;
-      case 2:  // src1 == dst
-        p_src0 = src0_;
-        p_src1 = dst_tst_;
-        src1_stride_ = dst_stride_;
-        src1_offset_ = dst_offset_;
-        break;
-      default: FAIL();
-    }
+    p_src0 = src0_;
+    p_src1 = src1_;
+
+    GetSources(&p_src0, &p_src1, &dst_ref_[0]);
 
     Execute(p_src0, p_src1);
 
     for (int r = 0; r < h_; ++r) {
       for (int c = 0; c < w_; ++c) {
         ASSERT_EQ(dst_ref_[dst_offset_ + r * dst_stride_ + c],
-                  dst_tst_[dst_offset_ + r * dst_stride_ + c]);
+                  dst_tst_[dst_offset_ + r * dst_stride_ + c])
+            << w_ << "x" << h_ << " r: " << r << " c: " << c;
       }
     }
   }
 
-  T dst_ref_[kBufSize];
-  T dst_tst_[kBufSize];
+  DstPixel dst_ref_[kBufSize];
+  DstPixel dst_tst_[kBufSize];
   uint32_t dst_stride_;
   uint32_t dst_offset_;
 
-  T src0_[kBufSize];
+  SrcPixel src0_[kBufSize];
   uint32_t src0_stride_;
   uint32_t src0_offset_;
 
-  T src1_[kBufSize];
+  SrcPixel src1_[kBufSize];
   uint32_t src1_stride_;
   uint32_t src1_offset_;
 
@@ -127,19 +143,19 @@ class BlendA64MaskTest : public FunctionEquivalenceTest<F> {
 typedef void (*F8B)(uint8_t *dst, uint32_t dst_stride, const uint8_t *src0,
                     uint32_t src0_stride, const uint8_t *src1,
                     uint32_t src1_stride, const uint8_t *mask,
-                    uint32_t mask_stride, int h, int w, int suby, int subx);
+                    uint32_t mask_stride, int w, int h, int subx, int suby);
 typedef libaom_test::FuncParam<F8B> TestFuncs;
 
-class BlendA64MaskTest8B : public BlendA64MaskTest<F8B, uint8_t> {
+class BlendA64MaskTest8B : public BlendA64MaskTest<F8B, uint8_t, uint8_t> {
  protected:
   void Execute(const uint8_t *p_src0, const uint8_t *p_src1) {
     params_.ref_func(dst_ref_ + dst_offset_, dst_stride_, p_src0 + src0_offset_,
                      src0_stride_, p_src1 + src1_offset_, src1_stride_, mask_,
-                     kMaxMaskWidth, h_, w_, suby_, subx_);
+                     kMaxMaskWidth, w_, h_, subx_, suby_);
     ASM_REGISTER_STATE_CHECK(params_.tst_func(
         dst_tst_ + dst_offset_, dst_stride_, p_src0 + src0_offset_,
         src0_stride_, p_src1 + src1_offset_, src1_stride_, mask_, kMaxMaskWidth,
-        h_, w_, suby_, subx_));
+        w_, h_, subx_, suby_));
   }
 };
 
@@ -156,7 +172,7 @@ TEST_P(BlendA64MaskTest8B, RandomValues) {
     for (int i = 0; i < kMaxMaskSize; ++i)
       mask_[i] = rng_(AOM_BLEND_A64_MAX_ALPHA + 1);
 
-    Common();
+    RunTest();
   }
 }
 
@@ -172,7 +188,7 @@ TEST_P(BlendA64MaskTest8B, ExtremeValues) {
     for (int i = 0; i < kMaxMaskSize; ++i)
       mask_[i] = rng_(2) + AOM_BLEND_A64_MAX_ALPHA - 1;
 
-    Common();
+    RunTest();
   }
 }
 
@@ -182,7 +198,85 @@ INSTANTIATE_TEST_CASE_P(SSE4_1, BlendA64MaskTest8B,
                             aom_blend_a64_mask_c, aom_blend_a64_mask_sse4_1)));
 #endif  // HAVE_SSE4_1
 
-#if CONFIG_HIGHBITDEPTH
+//////////////////////////////////////////////////////////////////////////////
+// 8 bit _d16 version
+//////////////////////////////////////////////////////////////////////////////
+
+typedef void (*F8B_D16)(uint8_t *dst, uint32_t dst_stride, const uint16_t *src0,
+                        uint32_t src0_stride, const uint16_t *src1,
+                        uint32_t src1_stride, const uint8_t *mask,
+                        uint32_t mask_stride, int w, int h, int subx, int suby,
+                        ConvolveParams *conv_params);
+typedef libaom_test::FuncParam<F8B_D16> TestFuncs_d16;
+
+class BlendA64MaskTest8B_d16
+    : public BlendA64MaskTest<F8B_D16, uint16_t, uint8_t> {
+ protected:
+  // max number of bits used by the source
+  static const int kSrcMaxBitsMask = 0x3fff;
+
+  void Execute(const uint16_t *p_src0, const uint16_t *p_src1) {
+    ConvolveParams conv_params;
+    conv_params.round_0 = ROUND0_BITS;
+    conv_params.round_1 = COMPOUND_ROUND1_BITS;
+    params_.ref_func(dst_ref_ + dst_offset_, dst_stride_, p_src0 + src0_offset_,
+                     src0_stride_, p_src1 + src1_offset_, src1_stride_, mask_,
+                     kMaxMaskWidth, w_, h_, subx_, suby_, &conv_params);
+    ASM_REGISTER_STATE_CHECK(params_.tst_func(
+        dst_tst_ + dst_offset_, dst_stride_, p_src0 + src0_offset_,
+        src0_stride_, p_src1 + src1_offset_, src1_stride_, mask_, kMaxMaskWidth,
+        w_, h_, subx_, suby_, &conv_params));
+  }
+};
+
+TEST_P(BlendA64MaskTest8B_d16, RandomValues) {
+  for (int iter = 0; iter < kIterations && !HasFatalFailure(); ++iter) {
+    for (int i = 0; i < kBufSize; ++i) {
+      dst_ref_[i] = rng_.Rand8();
+      dst_tst_[i] = rng_.Rand8();
+
+      src0_[i] = rng_.Rand16() & kSrcMaxBitsMask;
+      src1_[i] = rng_.Rand16() & kSrcMaxBitsMask;
+    }
+
+    for (int i = 0; i < kMaxMaskSize; ++i)
+      mask_[i] = rng_(AOM_BLEND_A64_MAX_ALPHA + 1);
+
+    RunTest();
+  }
+}
+
+TEST_P(BlendA64MaskTest8B_d16, ExtremeValues) {
+  for (int iter = 0; iter < kIterations && !HasFatalFailure(); ++iter) {
+    for (int i = 0; i < kBufSize; ++i) {
+      dst_ref_[i] = 255;
+      dst_tst_[i] = 255;
+
+      src0_[i] = kSrcMaxBitsMask;
+      src1_[i] = kSrcMaxBitsMask;
+    }
+
+    for (int i = 0; i < kMaxMaskSize; ++i)
+      mask_[i] = AOM_BLEND_A64_MAX_ALPHA - 1;
+
+    RunTest();
+  }
+}
+
+#if HAVE_SSE4_1
+INSTANTIATE_TEST_CASE_P(
+    SSE4_1, BlendA64MaskTest8B_d16,
+    ::testing::Values(TestFuncs_d16(aom_lowbd_blend_a64_d16_mask_c,
+                                    aom_lowbd_blend_a64_d16_mask_sse4_1)));
+#endif  // HAVE_SSE4_1
+
+#if HAVE_NEON
+INSTANTIATE_TEST_CASE_P(
+    NEON, BlendA64MaskTest8B_d16,
+    ::testing::Values(TestFuncs_d16(aom_lowbd_blend_a64_d16_mask_c,
+                                    aom_lowbd_blend_a64_d16_mask_neon)));
+#endif  // HAVE_NEON
+
 //////////////////////////////////////////////////////////////////////////////
 // High bit-depth version
 //////////////////////////////////////////////////////////////////////////////
@@ -190,22 +284,22 @@ INSTANTIATE_TEST_CASE_P(SSE4_1, BlendA64MaskTest8B,
 typedef void (*FHBD)(uint8_t *dst, uint32_t dst_stride, const uint8_t *src0,
                      uint32_t src0_stride, const uint8_t *src1,
                      uint32_t src1_stride, const uint8_t *mask,
-                     uint32_t mask_stride, int h, int w, int suby, int subx,
+                     uint32_t mask_stride, int w, int h, int subx, int suby,
                      int bd);
 typedef libaom_test::FuncParam<FHBD> TestFuncsHBD;
 
-class BlendA64MaskTestHBD : public BlendA64MaskTest<FHBD, uint16_t> {
+class BlendA64MaskTestHBD : public BlendA64MaskTest<FHBD, uint16_t, uint16_t> {
  protected:
   void Execute(const uint16_t *p_src0, const uint16_t *p_src1) {
     params_.ref_func(CONVERT_TO_BYTEPTR(dst_ref_ + dst_offset_), dst_stride_,
                      CONVERT_TO_BYTEPTR(p_src0 + src0_offset_), src0_stride_,
                      CONVERT_TO_BYTEPTR(p_src1 + src1_offset_), src1_stride_,
-                     mask_, kMaxMaskWidth, h_, w_, suby_, subx_, bit_depth_);
+                     mask_, kMaxMaskWidth, w_, h_, subx_, suby_, bit_depth_);
     ASM_REGISTER_STATE_CHECK(params_.tst_func(
         CONVERT_TO_BYTEPTR(dst_tst_ + dst_offset_), dst_stride_,
         CONVERT_TO_BYTEPTR(p_src0 + src0_offset_), src0_stride_,
         CONVERT_TO_BYTEPTR(p_src1 + src1_offset_), src1_stride_, mask_,
-        kMaxMaskWidth, h_, w_, suby_, subx_, bit_depth_));
+        kMaxMaskWidth, w_, h_, subx_, suby_, bit_depth_));
   }
 
   int bit_depth_;
@@ -231,7 +325,7 @@ TEST_P(BlendA64MaskTestHBD, RandomValues) {
     for (int i = 0; i < kMaxMaskSize; ++i)
       mask_[i] = rng_(AOM_BLEND_A64_MAX_ALPHA + 1);
 
-    Common();
+    RunTest();
   }
 }
 
@@ -256,7 +350,7 @@ TEST_P(BlendA64MaskTestHBD, ExtremeValues) {
     for (int i = 0; i < kMaxMaskSize; ++i)
       mask_[i] = rng_(2) + AOM_BLEND_A64_MAX_ALPHA - 1;
 
-    Common();
+    RunTest();
   }
 }
 
@@ -266,5 +360,104 @@ INSTANTIATE_TEST_CASE_P(
     ::testing::Values(TestFuncsHBD(aom_highbd_blend_a64_mask_c,
                                    aom_highbd_blend_a64_mask_sse4_1)));
 #endif  // HAVE_SSE4_1
-#endif  // CONFIG_HIGHBITDEPTH
+
+//////////////////////////////////////////////////////////////////////////////
+// HBD _d16 version
+//////////////////////////////////////////////////////////////////////////////
+
+typedef void (*FHBD_D16)(uint8_t *dst, uint32_t dst_stride,
+                         const CONV_BUF_TYPE *src0, uint32_t src0_stride,
+                         const CONV_BUF_TYPE *src1, uint32_t src1_stride,
+                         const uint8_t *mask, uint32_t mask_stride, int w,
+                         int h, int subx, int suby, ConvolveParams *conv_params,
+                         const int bd);
+typedef libaom_test::FuncParam<FHBD_D16> TestFuncsHBD_d16;
+
+class BlendA64MaskTestHBD_d16
+    : public BlendA64MaskTest<FHBD_D16, uint16_t, uint16_t> {
+ protected:
+  // max number of bits used by the source
+  static const int kSrcMaxBitsMask = (1 << 14) - 1;
+  static const int kSrcMaxBitsMaskHBD = (1 << 16) - 1;
+
+  void Execute(const uint16_t *p_src0, const uint16_t *p_src1) {
+    ConvolveParams conv_params;
+    conv_params.round_0 = (bit_depth_ == 12) ? ROUND0_BITS + 2 : ROUND0_BITS;
+    conv_params.round_1 = COMPOUND_ROUND1_BITS;
+
+    params_.ref_func(CONVERT_TO_BYTEPTR(dst_ref_ + dst_offset_), dst_stride_,
+                     p_src0 + src0_offset_, src0_stride_, p_src1 + src1_offset_,
+                     src1_stride_, mask_, kMaxMaskWidth, w_, h_, subx_, suby_,
+                     &conv_params, bit_depth_);
+    if (params_.tst_func) {
+      ASM_REGISTER_STATE_CHECK(params_.tst_func(
+          CONVERT_TO_BYTEPTR(dst_tst_ + dst_offset_), dst_stride_,
+          p_src0 + src0_offset_, src0_stride_, p_src1 + src1_offset_,
+          src1_stride_, mask_, kMaxMaskWidth, w_, h_, subx_, suby_,
+          &conv_params, bit_depth_));
+    }
+  }
+
+  int bit_depth_;
+  int src_max_bits_mask_;
+};
+
+TEST_P(BlendA64MaskTestHBD_d16, RandomValues) {
+  if (params_.tst_func == NULL) return;
+  for (int iter = 0; iter < kIterations && !HasFatalFailure(); ++iter) {
+    switch (rng_(3)) {
+      case 0: bit_depth_ = 8; break;
+      case 1: bit_depth_ = 10; break;
+      default: bit_depth_ = 12; break;
+    }
+    src_max_bits_mask_ =
+        (bit_depth_ == 8) ? kSrcMaxBitsMask : kSrcMaxBitsMaskHBD;
+
+    for (int i = 0; i < kBufSize; ++i) {
+      dst_ref_[i] = rng_.Rand8();
+      dst_tst_[i] = rng_.Rand8();
+
+      src0_[i] = rng_.Rand16() & src_max_bits_mask_;
+      src1_[i] = rng_.Rand16() & src_max_bits_mask_;
+    }
+
+    for (int i = 0; i < kMaxMaskSize; ++i)
+      mask_[i] = rng_(AOM_BLEND_A64_MAX_ALPHA + 1);
+
+    RunTest();
+  }
+}
+
+TEST_P(BlendA64MaskTestHBD_d16, SaturatedValues) {
+  for (bit_depth_ = 8; bit_depth_ <= 12; bit_depth_ += 2) {
+    src_max_bits_mask_ =
+        (bit_depth_ == 8) ? kSrcMaxBitsMask : kSrcMaxBitsMaskHBD;
+
+    for (int i = 0; i < kBufSize; ++i) {
+      dst_ref_[i] = 0;
+      dst_tst_[i] = (1 << bit_depth_) - 1;
+
+      src0_[i] = src_max_bits_mask_;
+      src1_[i] = src_max_bits_mask_;
+    }
+
+    for (int i = 0; i < kMaxMaskSize; ++i) mask_[i] = AOM_BLEND_A64_MAX_ALPHA;
+
+    RunTest();
+  }
+}
+
+INSTANTIATE_TEST_CASE_P(
+    C, BlendA64MaskTestHBD_d16,
+    ::testing::Values(TestFuncsHBD_d16(aom_highbd_blend_a64_d16_mask_c, NULL)));
+
+// TODO(slavarnway): Enable the following in the avx2 commit. (56501)
+#if 0
+#if HAVE_AVX2
+INSTANTIATE_TEST_CASE_P(
+    SSE4_1, BlendA64MaskTestHBD,
+    ::testing::Values(TestFuncsHBD(aom_highbd_blend_a64_mask_c,
+                                   aom_highbd_blend_a64_mask_avx2)));
+#endif  // HAVE_AVX2
+#endif
 }  // namespace

third_party/aom/test/block_error_test.cc

@@ -1,136 +0,0 @@
-/*
- * Copyright (c) 2017, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
-
-#include "./aom_config.h"
-#include "./av1_rtcd.h"
-
-#include "test/acm_random.h"
-#include "test/clear_system_state.h"
-#include "test/register_state_check.h"
-#include "test/util.h"
-
-namespace {
-using libaom_test::ACMRandom;
-
-typedef int64_t (*BlockErrorFunc)(const tran_low_t *coeff,
-                                  const tran_low_t *dqcoeff, intptr_t size,
-                                  int64_t *ssz);
-#if CONFIG_HIGHBITDEPTH
-typedef int64_t (*HbdBlockErrorFunc)(const tran_low_t *coeff,
-                                     const tran_low_t *dqcoeff, intptr_t size,
-                                     int64_t *ssz, int bd);
-#endif
-
-typedef std::tr1::tuple<BlockErrorFunc, BlockErrorFunc, TX_SIZE,
-                        aom_bit_depth_t>
-    BlockErrorParam;
-
-const int kTestNum = 10000;
-
-class BlockErrorTest : public ::testing::TestWithParam<BlockErrorParam> {
- public:
-  BlockErrorTest()
-      : blk_err_ref_(GET_PARAM(0)), blk_err_(GET_PARAM(1)),
-        tx_size_(GET_PARAM(2)), bd_(GET_PARAM(3)) {}
-
-  virtual ~BlockErrorTest() {}
-
-  virtual void SetUp() {
-    const intptr_t block_size = getCoeffNum();
-    coeff_ = reinterpret_cast<tran_low_t *>(
-        aom_memalign(16, 2 * block_size * sizeof(tran_low_t)));
-  }
-
-  virtual void TearDown() {
-    aom_free(coeff_);
-    coeff_ = NULL;
-    libaom_test::ClearSystemState();
-  }
-
-  void BlockErrorRun(int testNum) {
-    int i;
-    int64_t error_ref, error;
-    int64_t sse_ref, sse;
-    const intptr_t block_size = getCoeffNum();
-    tran_low_t *dqcoeff = coeff_ + block_size;
-    for (i = 0; i < testNum; ++i) {
-      FillRandomData();
-
-      error_ref = blk_err_ref_(coeff_, dqcoeff, block_size, &sse_ref);
-      ASM_REGISTER_STATE_CHECK(error =
-                                   blk_err_(coeff_, dqcoeff, block_size, &sse));
-
-      EXPECT_EQ(error_ref, error) << "Error doesn't match on test: " << i;
-      EXPECT_EQ(sse_ref, sse) << "SSE doesn't match on test: " << i;
-    }
-  }
-
-  intptr_t getCoeffNum() { return tx_size_2d[tx_size_]; }
-
-  void FillRandomData() {
-    const intptr_t block_size = getCoeffNum();
-    tran_low_t *dqcoeff = coeff_ + block_size;
-    intptr_t i;
-    int16_t margin = 512;
-    for (i = 0; i < block_size; ++i) {
-      coeff_[i] = GetRandomNumWithRange(INT16_MIN + margin, INT16_MAX - margin);
-      dqcoeff[i] = coeff_[i] + GetRandomDeltaWithRange(margin);
-    }
-  }
-
-  void FillConstantData() {
-    const intptr_t block_size = getCoeffNum();
-    tran_low_t *dqcoeff = coeff_ + block_size;
-    intptr_t i;
-    for (i = 0; i < block_size; ++i) {
-      coeff_[i] = 5;
-      dqcoeff[i] = 7;
-    }
-  }
-
-  tran_low_t GetRandomNumWithRange(int16_t min, int16_t max) {
-    return clamp((int16_t)rnd_.Rand16(), min, max);
-  }
-
-  tran_low_t GetRandomDeltaWithRange(int16_t delta) {
-    tran_low_t value = (int16_t)rnd_.Rand16();
-    value %= delta;
-    return value;
-  }
-
-  BlockErrorFunc blk_err_ref_;
-  BlockErrorFunc blk_err_;
-  TX_SIZE tx_size_;
-  aom_bit_depth_t bd_;
-  ACMRandom rnd_;
-  tran_low_t *coeff_;
-};
-
-TEST_P(BlockErrorTest, BitExact) { BlockErrorRun(kTestNum); }
-
-using std::tr1::make_tuple;
-
-#if !CONFIG_HIGHBITDEPTH && HAVE_SSE2
-const BlockErrorParam kBlkErrParamArraySse2[] = { make_tuple(
-    &av1_block_error_c, &av1_block_error_sse2, TX_32X32, AOM_BITS_8) };
-INSTANTIATE_TEST_CASE_P(SSE2, BlockErrorTest,
-                        ::testing::ValuesIn(kBlkErrParamArraySse2));
-#endif
-
-#if HAVE_AVX2
-const BlockErrorParam kBlkErrParamArrayAvx2[] = { make_tuple(
-    &av1_block_error_c, &av1_block_error_avx2, TX_32X32, AOM_BITS_8) };
-INSTANTIATE_TEST_CASE_P(AVX2, BlockErrorTest,
-                        ::testing::ValuesIn(kBlkErrParamArrayAvx2));
-#endif
-}  // namespace

third_party/aom/test/boolcoder_test.cc

@@ -7,7 +7,7 @@
  * obtain it at www.aomedia.org/license/software. If the Alliance for Open
  * Media Patent License 1.0 was not distributed with this source code in the
  * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
+ */
 
 #include <math.h>
 #include <stdlib.h>
@@ -69,7 +69,7 @@ TEST(AV1, TestBitIO) {
         aom_stop_encode(&bw);
 
         aom_reader br;
-        aom_reader_init(&br, bw_buffer, bw.pos, NULL, NULL);
+        aom_reader_init(&br, bw_buffer, bw.pos);
         bit_rnd.Reset(random_seed);
         for (int i = 0; i < kBitsToTest; ++i) {
           if (bit_method == 2) {
@@ -86,7 +86,7 @@ TEST(AV1, TestBitIO) {
   }
 }
 
-#define FRAC_DIFF_TOTAL_ERROR 0.16
+#define FRAC_DIFF_TOTAL_ERROR 0.18
 
 TEST(AV1, TestTell) {
   const int kBufferSize = 10000;
@@ -102,7 +102,7 @@ TEST(AV1, TestTell) {
     }
     aom_stop_encode(&bw);
     aom_reader br;
-    aom_reader_init(&br, bw_buffer, bw.pos, NULL, NULL);
+    aom_reader_init(&br, bw_buffer, bw.pos);
     uint32_t last_tell = aom_reader_tell(&br);
     uint32_t last_tell_frac = aom_reader_tell_frac(&br);
     double frac_diff_total = 0;

third_party/aom/test/borders_test.cc

@@ -7,7 +7,7 @@
  * obtain it at www.aomedia.org/license/software. If the Alliance for Open
  * Media Patent License 1.0 was not distributed with this source code in the
  * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
+ */
 
 #include <climits>
 #include <vector>
@@ -19,12 +19,12 @@
 
 namespace {
 
-class BordersTest
+class BordersTestLarge
     : public ::libaom_test::CodecTestWithParam<libaom_test::TestMode>,
       public ::libaom_test::EncoderTest {
  protected:
-  BordersTest() : EncoderTest(GET_PARAM(0)) {}
-  virtual ~BordersTest() {}
+  BordersTestLarge() : EncoderTest(GET_PARAM(0)) {}
+  virtual ~BordersTestLarge() {}
 
   virtual void SetUp() {
     InitializeConfig();
@@ -47,7 +47,7 @@ class BordersTest
   }
 };
 
-TEST_P(BordersTest, TestEncodeHighBitrate) {
+TEST_P(BordersTestLarge, TestEncodeHighBitrate) {
   // Validate that this non multiple of 64 wide clip encodes and decodes
   // without a mismatch when passing in a very low max q.  This pushes
   // the encoder to producing lots of big partitions which will likely
@@ -63,7 +63,7 @@ TEST_P(BordersTest, TestEncodeHighBitrate) {
 
   ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
 }
-TEST_P(BordersTest, TestLowBitrate) {
+TEST_P(BordersTestLarge, TestLowBitrate) {
   // Validate that this clip encodes and decodes without a mismatch
   // when passing in a very high min q.  This pushes the encoder to producing
   // lots of small partitions which might will test the other condition.
@@ -80,6 +80,6 @@ TEST_P(BordersTest, TestLowBitrate) {
   ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
 }
 
-AV1_INSTANTIATE_TEST_CASE(BordersTest,
+AV1_INSTANTIATE_TEST_CASE(BordersTestLarge,
                           ::testing::Values(::libaom_test::kTwoPassGood));
 }  // namespace

third_party/aom/test/cdef_test.cc

@@ -7,15 +7,16 @@
  * obtain it at www.aomedia.org/license/software. If the Alliance for Open
  * Media Patent License 1.0 was not distributed with this source code in the
  * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
+ */
 
 #include <cstdlib>
 #include <string>
 
 #include "third_party/googletest/src/googletest/include/gtest/gtest.h"
 
-#include "./aom_config.h"
-#include "./av1_rtcd.h"
+#include "config/aom_config.h"
+#include "config/av1_rtcd.h"
+
 #include "aom_ports/aom_timer.h"
 #include "av1/common/cdef_block.h"
 #include "test/acm_random.h"
@@ -27,7 +28,8 @@ using libaom_test::ACMRandom;
 
 namespace {
 
-typedef std::tr1::tuple<cdef_filter_block_func, cdef_filter_block_func, int>
+typedef ::testing::tuple<cdef_filter_block_func, cdef_filter_block_func,
+                         BLOCK_SIZE, int, int>
     cdef_dir_param_t;
 
 class CDEFBlockTest : public ::testing::TestWithParam<cdef_dir_param_t> {
@@ -37,12 +39,16 @@ class CDEFBlockTest : public ::testing::TestWithParam<cdef_dir_param_t> {
     cdef = GET_PARAM(0);
     ref_cdef = GET_PARAM(1);
     bsize = GET_PARAM(2);
+    boundary = GET_PARAM(3);
+    depth = GET_PARAM(4);
   }
 
   virtual void TearDown() { libaom_test::ClearSystemState(); }
 
  protected:
   int bsize;
+  int boundary;
+  int depth;
   cdef_filter_block_func cdef;
   cdef_filter_block_func ref_cdef;
 };
@@ -50,7 +56,7 @@ class CDEFBlockTest : public ::testing::TestWithParam<cdef_dir_param_t> {
 typedef CDEFBlockTest CDEFSpeedTest;
 
 void test_cdef(int bsize, int iterations, cdef_filter_block_func cdef,
-               cdef_filter_block_func ref_cdef) {
+               cdef_filter_block_func ref_cdef, int boundary, int depth) {
   const int size = 8;
   const int ysize = size + 2 * CDEF_VBORDER;
   ACMRandom rnd(ACMRandom::DeterministicSeed());
@@ -61,80 +67,73 @@ void test_cdef(int bsize, int iterations, cdef_filter_block_func cdef,
   memset(d, 0, sizeof(d));
 
   int error = 0, pristrength = 0, secstrength, dir;
-  int boundary, pridamping, secdamping, depth, bits, level, count,
+  int pridamping, secdamping, bits, level, count,
       errdepth = 0, errpristrength = 0, errsecstrength = 0, errboundary = 0,
       errpridamping = 0, errsecdamping = 0;
   unsigned int pos = 0;
 
-  for (boundary = 0; boundary < 16; boundary++) {
-    for (depth = 8; depth <= 12; depth += 2) {
-      const unsigned int max_pos = size * size >> (depth == 8);
-      for (pridamping = 3 + depth - 8;
-           pridamping < 7 - 3 * !!boundary + depth - 8; pridamping++) {
-        for (secdamping = 3 + depth - 8;
-             secdamping < 7 - 3 * !!boundary + depth - 8; secdamping++) {
-          for (count = 0; count < iterations; count++) {
-            for (level = 0; level < (1 << depth) && !error;
-                 level += (2 + 6 * !!boundary) << (depth - 8)) {
-              for (bits = 1; bits <= depth && !error;
-                   bits += 1 + 3 * !!boundary) {
-                for (unsigned int i = 0; i < sizeof(s) / sizeof(*s); i++)
-                  s[i] = clamp((rnd.Rand16() & ((1 << bits) - 1)) + level, 0,
-                               (1 << depth) - 1);
-                if (boundary) {
-                  if (boundary & 1) {  // Left
-                    for (int i = 0; i < ysize; i++)
-                      for (int j = 0; j < CDEF_HBORDER; j++)
-                        s[i * CDEF_BSTRIDE + j] = CDEF_VERY_LARGE;
-                  }
-                  if (boundary & 2) {  // Right
-                    for (int i = 0; i < ysize; i++)
-                      for (int j = CDEF_HBORDER + size; j < CDEF_BSTRIDE; j++)
-                        s[i * CDEF_BSTRIDE + j] = CDEF_VERY_LARGE;
-                  }
-                  if (boundary & 4) {  // Above
-                    for (int i = 0; i < CDEF_VBORDER; i++)
-                      for (int j = 0; j < CDEF_BSTRIDE; j++)
-                        s[i * CDEF_BSTRIDE + j] = CDEF_VERY_LARGE;
-                  }
-                  if (boundary & 8) {  // Below
-                    for (int i = CDEF_VBORDER + size; i < ysize; i++)
-                      for (int j = 0; j < CDEF_BSTRIDE; j++)
-                        s[i * CDEF_BSTRIDE + j] = CDEF_VERY_LARGE;
-                  }
-                }
-                for (dir = 0; dir < 8; dir++) {
-                  for (pristrength = 0;
-                       pristrength <= 19 << (depth - 8) && !error;
-                       pristrength += (1 + 4 * !!boundary) << (depth - 8)) {
-                    if (pristrength == 16) pristrength = 19;
-                    for (secstrength = 0;
-                         secstrength <= 4 << (depth - 8) && !error;
-                         secstrength += 1 << (depth - 8)) {
-                      if (secstrength == 3 << (depth - 8)) continue;
-                      ref_cdef(depth == 8 ? (uint8_t *)ref_d : 0, ref_d, size,
-                               s + CDEF_HBORDER + CDEF_VBORDER * CDEF_BSTRIDE,
-                               pristrength, secstrength, dir, pridamping,
-                               secdamping, bsize, (1 << depth) - 1);
-                      // If cdef and ref_cdef are the same, we're just testing
-                      // speed
-                      if (cdef != ref_cdef)
-                        ASM_REGISTER_STATE_CHECK(
-                            cdef(depth == 8 ? (uint8_t *)d : 0, d, size,
-                                 s + CDEF_HBORDER + CDEF_VBORDER * CDEF_BSTRIDE,
-                                 pristrength, secstrength, dir, pridamping,
-                                 secdamping, bsize, (1 << depth) - 1));
-                      if (ref_cdef != cdef) {
-                        for (pos = 0; pos < max_pos && !error; pos++) {
-                          error = ref_d[pos] != d[pos];
-                          errdepth = depth;
-                          errpristrength = pristrength;
-                          errsecstrength = secstrength;
-                          errboundary = boundary;
-                          errpridamping = pridamping;
-                          errsecdamping = secdamping;
-                        }
-                      }
+  const unsigned int max_pos = size * size >> static_cast<int>(depth == 8);
+  for (pridamping = 3 + depth - 8; pridamping < 7 - 3 * !!boundary + depth - 8;
+       pridamping++) {
+    for (secdamping = 3 + depth - 8;
+         secdamping < 7 - 3 * !!boundary + depth - 8; secdamping++) {
+      for (count = 0; count < iterations; count++) {
+        for (level = 0; level < (1 << depth) && !error;
+             level += (2 + 6 * !!boundary) << (depth - 8)) {
+          for (bits = 1; bits <= depth && !error; bits += 1 + 3 * !!boundary) {
+            for (unsigned int i = 0; i < sizeof(s) / sizeof(*s); i++)
+              s[i] = clamp((rnd.Rand16() & ((1 << bits) - 1)) + level, 0,
+                           (1 << depth) - 1);
+            if (boundary) {
+              if (boundary & 1) {  // Left
+                for (int i = 0; i < ysize; i++)
+                  for (int j = 0; j < CDEF_HBORDER; j++)
+                    s[i * CDEF_BSTRIDE + j] = CDEF_VERY_LARGE;
+              }
+              if (boundary & 2) {  // Right
+                for (int i = 0; i < ysize; i++)
+                  for (int j = CDEF_HBORDER + size; j < CDEF_BSTRIDE; j++)
+                    s[i * CDEF_BSTRIDE + j] = CDEF_VERY_LARGE;
+              }
+              if (boundary & 4) {  // Above
+                for (int i = 0; i < CDEF_VBORDER; i++)
+                  for (int j = 0; j < CDEF_BSTRIDE; j++)
+                    s[i * CDEF_BSTRIDE + j] = CDEF_VERY_LARGE;
+              }
+              if (boundary & 8) {  // Below
+                for (int i = CDEF_VBORDER + size; i < ysize; i++)
+                  for (int j = 0; j < CDEF_BSTRIDE; j++)
+                    s[i * CDEF_BSTRIDE + j] = CDEF_VERY_LARGE;
+              }
+            }
+            for (dir = 0; dir < 8; dir++) {
+              for (pristrength = 0; pristrength <= 19 << (depth - 8) && !error;
+                   pristrength += (1 + 4 * !!boundary) << (depth - 8)) {
+                if (pristrength == 16) pristrength = 19;
+                for (secstrength = 0; secstrength <= 4 << (depth - 8) && !error;
+                     secstrength += 1 << (depth - 8)) {
+                  if (secstrength == 3 << (depth - 8)) continue;
+                  ref_cdef(depth == 8 ? (uint8_t *)ref_d : 0, ref_d, size,
+                           s + CDEF_HBORDER + CDEF_VBORDER * CDEF_BSTRIDE,
+                           pristrength, secstrength, dir, pridamping,
+                           secdamping, bsize, (1 << depth) - 1, depth - 8);
+                  // If cdef and ref_cdef are the same, we're just testing
+                  // speed
+                  if (cdef != ref_cdef)
+                    ASM_REGISTER_STATE_CHECK(
+                        cdef(depth == 8 ? (uint8_t *)d : 0, d, size,
+                             s + CDEF_HBORDER + CDEF_VBORDER * CDEF_BSTRIDE,
+                             pristrength, secstrength, dir, pridamping,
+                             secdamping, bsize, (1 << depth) - 1, depth - 8));
+                  if (ref_cdef != cdef) {
+                    for (pos = 0; pos < max_pos && !error; pos++) {
+                      error = ref_d[pos] != d[pos];
+                      errdepth = depth;
+                      errpristrength = pristrength;
+                      errsecstrength = secstrength;
+                      errboundary = boundary;
+                      errpridamping = pridamping;
+                      errsecdamping = secdamping;
                     }
                   }
                 }
@@ -145,6 +144,7 @@ void test_cdef(int bsize, int iterations, cdef_filter_block_func cdef,
       }
     }
   }
+
   pos--;
   EXPECT_EQ(0, error) << "Error: CDEFBlockTest, SIMD and C mismatch."
                       << std::endl
@@ -162,25 +162,20 @@ void test_cdef(int bsize, int iterations, cdef_filter_block_func cdef,
 }
 
 void test_cdef_speed(int bsize, int iterations, cdef_filter_block_func cdef,
-                     cdef_filter_block_func ref_cdef) {
+                     cdef_filter_block_func ref_cdef, int boundary, int depth) {
   aom_usec_timer ref_timer;
   aom_usec_timer timer;
 
   aom_usec_timer_start(&ref_timer);
-  test_cdef(bsize, iterations, ref_cdef, ref_cdef);
+  test_cdef(bsize, iterations, ref_cdef, ref_cdef, boundary, depth);
   aom_usec_timer_mark(&ref_timer);
   int ref_elapsed_time = (int)aom_usec_timer_elapsed(&ref_timer);
 
   aom_usec_timer_start(&timer);
-  test_cdef(bsize, iterations, cdef, cdef);
+  test_cdef(bsize, iterations, cdef, cdef, boundary, depth);
   aom_usec_timer_mark(&timer);
   int elapsed_time = (int)aom_usec_timer_elapsed(&timer);
 
-#if 0
-  std::cout << "[          ] C time = " << ref_elapsed_time / 1000
-            << " ms, SIMD time = " << elapsed_time / 1000 << " ms" << std::endl;
-#endif
-
   EXPECT_GT(ref_elapsed_time, elapsed_time)
       << "Error: CDEFSpeedTest, SIMD slower than C." << std::endl
       << "C time: " << ref_elapsed_time << " us" << std::endl
@@ -190,7 +185,7 @@ void test_cdef_speed(int bsize, int iterations, cdef_filter_block_func cdef,
 typedef int (*find_dir_t)(const uint16_t *img, int stride, int32_t *var,
                           int coeff_shift);
 
-typedef std::tr1::tuple<find_dir_t, find_dir_t> find_dir_param_t;
+typedef ::testing::tuple<find_dir_t, find_dir_t> find_dir_param_t;
 
 class CDEFFindDirTest : public ::testing::TestWithParam<find_dir_param_t> {
  public:
@@ -268,11 +263,6 @@ void test_finddir_speed(int (*finddir)(const uint16_t *img, int stride,
   aom_usec_timer_mark(&timer);
   int elapsed_time = (int)aom_usec_timer_elapsed(&timer);
 
-#if 0
-  std::cout << "[          ] C time = " << ref_elapsed_time / 1000
-            << " ms, SIMD time = " << elapsed_time / 1000 << " ms" << std::endl;
-#endif
-
   EXPECT_GT(ref_elapsed_time, elapsed_time)
       << "Error: CDEFFindDirSpeedTest, SIMD slower than C." << std::endl
       << "C time: " << ref_elapsed_time << " us" << std::endl
@@ -280,11 +270,11 @@ void test_finddir_speed(int (*finddir)(const uint16_t *img, int stride,
 }
 
 TEST_P(CDEFBlockTest, TestSIMDNoMismatch) {
-  test_cdef(bsize, 1, cdef, ref_cdef);
+  test_cdef(bsize, 1, cdef, ref_cdef, boundary, depth);
 }
 
 TEST_P(CDEFSpeedTest, DISABLED_TestSpeed) {
-  test_cdef_speed(bsize, 4, cdef, ref_cdef);
+  test_cdef_speed(bsize, 4, cdef, ref_cdef, boundary, depth);
 }
 
 TEST_P(CDEFFindDirTest, TestSIMDNoMismatch) {
@@ -295,7 +285,7 @@ TEST_P(CDEFFindDirSpeedTest, DISABLED_TestSpeed) {
   test_finddir_speed(finddir, ref_finddir);
 }
 
-using std::tr1::make_tuple;
+using ::testing::make_tuple;
 
 // VS compiling for 32 bit targets does not support vector types in
 // structs as arguments, which makes the v256 type of the intrinsics
@@ -304,9 +294,11 @@ using std::tr1::make_tuple;
 #if HAVE_SSE2
 INSTANTIATE_TEST_CASE_P(
     SSE2, CDEFBlockTest,
-    ::testing::Values(
-        make_tuple(&cdef_filter_block_sse2, &cdef_filter_block_c, BLOCK_4X4),
-        make_tuple(&cdef_filter_block_sse2, &cdef_filter_block_c, BLOCK_8X8)));
+    ::testing::Combine(::testing::Values(&cdef_filter_block_sse2),
+                       ::testing::Values(&cdef_filter_block_c),
+                       ::testing::Values(BLOCK_4X4, BLOCK_4X8, BLOCK_8X4,
+                                         BLOCK_8X8),
+                       ::testing::Range(0, 16), ::testing::Range(8, 13, 2)));
 INSTANTIATE_TEST_CASE_P(SSE2, CDEFFindDirTest,
                         ::testing::Values(make_tuple(&cdef_find_dir_sse2,
                                                      &cdef_find_dir_c)));
@@ -314,9 +306,11 @@ INSTANTIATE_TEST_CASE_P(SSE2, CDEFFindDirTest,
 #if HAVE_SSSE3
 INSTANTIATE_TEST_CASE_P(
     SSSE3, CDEFBlockTest,
-    ::testing::Values(
-        make_tuple(&cdef_filter_block_ssse3, &cdef_filter_block_c, BLOCK_4X4),
-        make_tuple(&cdef_filter_block_ssse3, &cdef_filter_block_c, BLOCK_8X8)));
+    ::testing::Combine(::testing::Values(&cdef_filter_block_ssse3),
+                       ::testing::Values(&cdef_filter_block_c),
+                       ::testing::Values(BLOCK_4X4, BLOCK_4X8, BLOCK_8X4,
+                                         BLOCK_8X8),
+                       ::testing::Range(0, 16), ::testing::Range(8, 13, 2)));
 INSTANTIATE_TEST_CASE_P(SSSE3, CDEFFindDirTest,
                         ::testing::Values(make_tuple(&cdef_find_dir_ssse3,
                                                      &cdef_find_dir_c)));
@@ -325,10 +319,11 @@ INSTANTIATE_TEST_CASE_P(SSSE3, CDEFFindDirTest,
 #if HAVE_SSE4_1
 INSTANTIATE_TEST_CASE_P(
     SSE4_1, CDEFBlockTest,
-    ::testing::Values(make_tuple(&cdef_filter_block_sse4_1,
-                                 &cdef_filter_block_c, BLOCK_4X4),
-                      make_tuple(&cdef_filter_block_sse4_1,
-                                 &cdef_filter_block_c, BLOCK_8X8)));
+    ::testing::Combine(::testing::Values(&cdef_filter_block_sse4_1),
+                       ::testing::Values(&cdef_filter_block_c),
+                       ::testing::Values(BLOCK_4X4, BLOCK_4X8, BLOCK_8X4,
+                                         BLOCK_8X8),
+                       ::testing::Range(0, 16), ::testing::Range(8, 13, 2)));
 INSTANTIATE_TEST_CASE_P(SSE4_1, CDEFFindDirTest,
                         ::testing::Values(make_tuple(&cdef_find_dir_sse4_1,
                                                      &cdef_find_dir_c)));
@@ -337,9 +332,11 @@ INSTANTIATE_TEST_CASE_P(SSE4_1, CDEFFindDirTest,
 #if HAVE_AVX2
 INSTANTIATE_TEST_CASE_P(
     AVX2, CDEFBlockTest,
-    ::testing::Values(
-        make_tuple(&cdef_filter_block_avx2, &cdef_filter_block_c, BLOCK_4X4),
-        make_tuple(&cdef_filter_block_avx2, &cdef_filter_block_c, BLOCK_8X8)));
+    ::testing::Combine(::testing::Values(&cdef_filter_block_avx2),
+                       ::testing::Values(&cdef_filter_block_c),
+                       ::testing::Values(BLOCK_4X4, BLOCK_4X8, BLOCK_8X4,
+                                         BLOCK_8X8),
+                       ::testing::Range(0, 16), ::testing::Range(8, 13, 2)));
 INSTANTIATE_TEST_CASE_P(AVX2, CDEFFindDirTest,
                         ::testing::Values(make_tuple(&cdef_find_dir_avx2,
                                                      &cdef_find_dir_c)));
@@ -348,9 +345,11 @@ INSTANTIATE_TEST_CASE_P(AVX2, CDEFFindDirTest,
 #if HAVE_NEON
 INSTANTIATE_TEST_CASE_P(
     NEON, CDEFBlockTest,
-    ::testing::Values(
-        make_tuple(&cdef_filter_block_neon, &cdef_filter_block_c, BLOCK_4X4),
-        make_tuple(&cdef_filter_block_neon, &cdef_filter_block_c, BLOCK_8X8)));
+    ::testing::Combine(::testing::Values(&cdef_filter_block_neon),
+                       ::testing::Values(&cdef_filter_block_c),
+                       ::testing::Values(BLOCK_4X4, BLOCK_4X8, BLOCK_8X4,
+                                         BLOCK_8X8),
+                       ::testing::Range(0, 16), ::testing::Range(8, 13, 2)));
 INSTANTIATE_TEST_CASE_P(NEON, CDEFFindDirTest,
                         ::testing::Values(make_tuple(&cdef_find_dir_neon,
                                                      &cdef_find_dir_c)));
@@ -360,9 +359,11 @@ INSTANTIATE_TEST_CASE_P(NEON, CDEFFindDirTest,
 #if HAVE_SSE2
 INSTANTIATE_TEST_CASE_P(
     SSE2, CDEFSpeedTest,
-    ::testing::Values(
-        make_tuple(&cdef_filter_block_sse2, &cdef_filter_block_c, BLOCK_4X4),
-        make_tuple(&cdef_filter_block_sse2, &cdef_filter_block_c, BLOCK_8X8)));
+    ::testing::Combine(::testing::Values(&cdef_filter_block_sse2),
+                       ::testing::Values(&cdef_filter_block_c),
+                       ::testing::Values(BLOCK_4X4, BLOCK_4X8, BLOCK_8X4,
+                                         BLOCK_8X8),
+                       ::testing::Range(0, 16), ::testing::Range(8, 13, 2)));
 INSTANTIATE_TEST_CASE_P(SSE2, CDEFFindDirSpeedTest,
                         ::testing::Values(make_tuple(&cdef_find_dir_sse2,
                                                      &cdef_find_dir_c)));
@@ -371,9 +372,11 @@ INSTANTIATE_TEST_CASE_P(SSE2, CDEFFindDirSpeedTest,
 #if HAVE_SSSE3
 INSTANTIATE_TEST_CASE_P(
     SSSE3, CDEFSpeedTest,
-    ::testing::Values(
-        make_tuple(&cdef_filter_block_ssse3, &cdef_filter_block_c, BLOCK_4X4),
-        make_tuple(&cdef_filter_block_ssse3, &cdef_filter_block_c, BLOCK_8X8)));
+    ::testing::Combine(::testing::Values(&cdef_filter_block_ssse3),
+                       ::testing::Values(&cdef_filter_block_c),
+                       ::testing::Values(BLOCK_4X4, BLOCK_4X8, BLOCK_8X4,
+                                         BLOCK_8X8),
+                       ::testing::Range(0, 16), ::testing::Range(8, 13, 2)));
 INSTANTIATE_TEST_CASE_P(SSSE3, CDEFFindDirSpeedTest,
                         ::testing::Values(make_tuple(&cdef_find_dir_ssse3,
                                                      &cdef_find_dir_c)));
@@ -382,10 +385,11 @@ INSTANTIATE_TEST_CASE_P(SSSE3, CDEFFindDirSpeedTest,
 #if HAVE_SSE4_1
 INSTANTIATE_TEST_CASE_P(
     SSE4_1, CDEFSpeedTest,
-    ::testing::Values(make_tuple(&cdef_filter_block_sse4_1,
-                                 &cdef_filter_block_c, BLOCK_4X4),
-                      make_tuple(&cdef_filter_block_sse4_1,
-                                 &cdef_filter_block_c, BLOCK_8X8)));
+    ::testing::Combine(::testing::Values(&cdef_filter_block_sse4_1),
+                       ::testing::Values(&cdef_filter_block_c),
+                       ::testing::Values(BLOCK_4X4, BLOCK_4X8, BLOCK_8X4,
+                                         BLOCK_8X8),
+                       ::testing::Range(0, 16), ::testing::Range(8, 13, 2)));
 INSTANTIATE_TEST_CASE_P(SSE4_1, CDEFFindDirSpeedTest,
                         ::testing::Values(make_tuple(&cdef_find_dir_sse4_1,
                                                      &cdef_find_dir_c)));
@@ -394,9 +398,11 @@ INSTANTIATE_TEST_CASE_P(SSE4_1, CDEFFindDirSpeedTest,
 #if HAVE_AVX2
 INSTANTIATE_TEST_CASE_P(
     AVX2, CDEFSpeedTest,
-    ::testing::Values(
-        make_tuple(&cdef_filter_block_avx2, &cdef_filter_block_c, BLOCK_4X4),
-        make_tuple(&cdef_filter_block_avx2, &cdef_filter_block_c, BLOCK_8X8)));
+    ::testing::Combine(::testing::Values(&cdef_filter_block_avx2),
+                       ::testing::Values(&cdef_filter_block_c),
+                       ::testing::Values(BLOCK_4X4, BLOCK_4X8, BLOCK_8X4,
+                                         BLOCK_8X8),
+                       ::testing::Range(0, 16), ::testing::Range(8, 13, 2)));
 INSTANTIATE_TEST_CASE_P(AVX2, CDEFFindDirSpeedTest,
                         ::testing::Values(make_tuple(&cdef_find_dir_avx2,
                                                      &cdef_find_dir_c)));
@@ -405,9 +411,11 @@ INSTANTIATE_TEST_CASE_P(AVX2, CDEFFindDirSpeedTest,
 #if HAVE_NEON
 INSTANTIATE_TEST_CASE_P(
     NEON, CDEFSpeedTest,
-    ::testing::Values(
-        make_tuple(&cdef_filter_block_neon, &cdef_filter_block_c, BLOCK_4X4),
-        make_tuple(&cdef_filter_block_neon, &cdef_filter_block_c, BLOCK_8X8)));
+    ::testing::Combine(::testing::Values(&cdef_filter_block_neon),
+                       ::testing::Values(&cdef_filter_block_c),
+                       ::testing::Values(BLOCK_4X4, BLOCK_4X8, BLOCK_8X4,
+                                         BLOCK_8X8),
+                       ::testing::Range(0, 16), ::testing::Range(8, 13, 2)));
 INSTANTIATE_TEST_CASE_P(NEON, CDEFFindDirSpeedTest,
                         ::testing::Values(make_tuple(&cdef_find_dir_neon,
                                                      &cdef_find_dir_c)));

third_party/aom/test/cfl_test.cc

@@ -0,0 +1,567 @@
+/*
+ * Copyright (c) 2017, Alliance for Open Media. All rights reserved
+ *
+ * This source code is subject to the terms of the BSD 2 Clause License and
+ * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
+ * was not distributed with this source code in the LICENSE file, you can
+ * obtain it at www.aomedia.org/license/software. If the Alliance for Open
+ * Media Patent License 1.0 was not distributed with this source code in the
+ * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
+ */
+
+#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
+
+#include "config/av1_rtcd.h"
+
+#include "aom_ports/aom_timer.h"
+#include "test/util.h"
+#include "test/acm_random.h"
+
+using ::testing::make_tuple;
+
+using libaom_test::ACMRandom;
+
+#define NUM_ITERATIONS (100)
+#define NUM_ITERATIONS_SPEED (INT16_MAX)
+
+#define ALL_CFL_TX_SIZES(function)                                     \
+  make_tuple(TX_4X4, &function), make_tuple(TX_4X8, &function),        \
+      make_tuple(TX_4X16, &function), make_tuple(TX_8X4, &function),   \
+      make_tuple(TX_8X8, &function), make_tuple(TX_8X16, &function),   \
+      make_tuple(TX_8X32, &function), make_tuple(TX_16X4, &function),  \
+      make_tuple(TX_16X8, &function), make_tuple(TX_16X16, &function), \
+      make_tuple(TX_16X32, &function), make_tuple(TX_32X8, &function), \
+      make_tuple(TX_32X16, &function), make_tuple(TX_32X32, &function)
+
+#define ALL_CFL_TX_SIZES_SUBSAMPLE(fun420, fun422, fun444) \
+  make_tuple(TX_4X4, &fun420, &fun422, &fun444),           \
+      make_tuple(TX_4X8, &fun420, &fun422, &fun444),       \
+      make_tuple(TX_4X16, &fun420, &fun422, &fun444),      \
+      make_tuple(TX_8X4, &fun420, &fun422, &fun444),       \
+      make_tuple(TX_8X8, &fun420, &fun422, &fun444),       \
+      make_tuple(TX_8X16, &fun420, &fun422, &fun444),      \
+      make_tuple(TX_8X32, &fun420, &fun422, &fun444),      \
+      make_tuple(TX_16X4, &fun420, &fun422, &fun444),      \
+      make_tuple(TX_16X8, &fun420, &fun422, &fun444),      \
+      make_tuple(TX_16X16, &fun420, &fun422, &fun444),     \
+      make_tuple(TX_16X32, &fun420, &fun422, &fun444),     \
+      make_tuple(TX_32X8, &fun420, &fun422, &fun444),      \
+      make_tuple(TX_32X16, &fun420, &fun422, &fun444),     \
+      make_tuple(TX_32X32, &fun420, &fun422, &fun444)
+
+namespace {
+
+template <typename A>
+static void assert_eq(const A *a, const A *b, int width, int height) {
+  for (int j = 0; j < height; j++) {
+    for (int i = 0; i < width; i++) {
+      ASSERT_EQ(a[j * CFL_BUF_LINE + i], b[j * CFL_BUF_LINE + i]);
+    }
+  }
+}
+
+static void assertFaster(int ref_elapsed_time, int elapsed_time) {
+  EXPECT_GT(ref_elapsed_time, elapsed_time)
+      << "Error: CFLSubtractSpeedTest, SIMD slower than C." << std::endl
+      << "C time: " << ref_elapsed_time << " us" << std::endl
+      << "SIMD time: " << elapsed_time << " us" << std::endl;
+}
+
+static void printSpeed(int ref_elapsed_time, int elapsed_time, int width,
+                       int height) {
+  std::cout.precision(2);
+  std::cout << "[          ] " << width << "x" << height
+            << ": C time = " << ref_elapsed_time
+            << " us, SIMD time = " << elapsed_time << " us"
+            << " (~" << ref_elapsed_time / (double)elapsed_time << "x) "
+            << std::endl;
+}
+
+class CFLTest {
+ public:
+  virtual ~CFLTest() {}
+  void init(TX_SIZE tx) {
+    tx_size = tx;
+    width = tx_size_wide[tx_size];
+    height = tx_size_high[tx_size];
+    rnd(ACMRandom::DeterministicSeed());
+  }
+
+ protected:
+  TX_SIZE tx_size;
+  int width;
+  int height;
+  ACMRandom rnd;
+};
+
+template <typename I>
+class CFLTestWithData : public CFLTest {
+ public:
+  virtual ~CFLTestWithData() {}
+
+ protected:
+  I data[CFL_BUF_SQUARE];
+  I data_ref[CFL_BUF_SQUARE];
+  void randData(I (ACMRandom::*random)()) {
+    for (int j = 0; j < this->height; j++) {
+      for (int i = 0; i < this->width; i++) {
+        const I d = (this->rnd.*random)();
+        data[j * CFL_BUF_LINE + i] = d;
+        data_ref[j * CFL_BUF_LINE + i] = d;
+      }
+    }
+  }
+};
+
+template <typename I>
+class CFLTestWithAlignedData : public CFLTest {
+ public:
+  CFLTestWithAlignedData() {
+    chroma_pels_ref =
+        reinterpret_cast<I *>(aom_memalign(32, sizeof(I) * CFL_BUF_SQUARE));
+    chroma_pels =
+        reinterpret_cast<I *>(aom_memalign(32, sizeof(I) * CFL_BUF_SQUARE));
+    sub_luma_pels_ref = reinterpret_cast<int16_t *>(
+        aom_memalign(32, sizeof(int16_t) * CFL_BUF_SQUARE));
+    sub_luma_pels = reinterpret_cast<int16_t *>(
+        aom_memalign(32, sizeof(int16_t) * CFL_BUF_SQUARE));
+    memset(chroma_pels_ref, 0, sizeof(I) * CFL_BUF_SQUARE);
+    memset(chroma_pels, 0, sizeof(I) * CFL_BUF_SQUARE);
+    memset(sub_luma_pels_ref, 0, sizeof(int16_t) * CFL_BUF_SQUARE);
+    memset(sub_luma_pels, 0, sizeof(int16_t) * CFL_BUF_SQUARE);
+  }
+  ~CFLTestWithAlignedData() {
+    aom_free(chroma_pels_ref);
+    aom_free(sub_luma_pels_ref);
+    aom_free(chroma_pels);
+    aom_free(sub_luma_pels);
+  }
+
+ protected:
+  I *chroma_pels_ref;
+  I *chroma_pels;
+  int16_t *sub_luma_pels_ref;
+  int16_t *sub_luma_pels;
+  int alpha_q3;
+  I dc;
+  void randData(int bd) {
+    alpha_q3 = this->rnd(33) - 16;
+    dc = this->rnd(1 << bd);
+    for (int j = 0; j < this->height; j++) {
+      for (int i = 0; i < this->width; i++) {
+        chroma_pels[j * CFL_BUF_LINE + i] = dc;
+        chroma_pels_ref[j * CFL_BUF_LINE + i] = dc;
+        sub_luma_pels_ref[j * CFL_BUF_LINE + i] =
+            sub_luma_pels[j * CFL_BUF_LINE + i] = this->rnd(1 << (bd + 3));
+      }
+    }
+  }
+};
+
+typedef cfl_subtract_average_fn (*sub_avg_fn)(TX_SIZE tx_size);
+typedef ::testing::tuple<TX_SIZE, sub_avg_fn> sub_avg_param;
+class CFLSubAvgTest : public ::testing::TestWithParam<sub_avg_param>,
+                      public CFLTestWithData<int16_t> {
+ public:
+  virtual void SetUp() {
+    CFLTest::init(::testing::get<0>(this->GetParam()));
+    sub_avg = ::testing::get<1>(this->GetParam())(tx_size);
+    sub_avg_ref = get_subtract_average_fn_c(tx_size);
+  }
+  virtual ~CFLSubAvgTest() {}
+
+ protected:
+  cfl_subtract_average_fn sub_avg;
+  cfl_subtract_average_fn sub_avg_ref;
+};
+
+TEST_P(CFLSubAvgTest, SubAvgTest) {
+  for (int it = 0; it < NUM_ITERATIONS; it++) {
+    randData(&ACMRandom::Rand15Signed);
+    sub_avg((uint16_t *)data, data);
+    sub_avg_ref((uint16_t *)data_ref, data_ref);
+    assert_eq<int16_t>(data, data_ref, width, height);
+  }
+}
+
+TEST_P(CFLSubAvgTest, DISABLED_SubAvgSpeedTest) {
+  aom_usec_timer ref_timer;
+  aom_usec_timer timer;
+  randData(&ACMRandom::Rand15Signed);
+  aom_usec_timer_start(&ref_timer);
+  for (int k = 0; k < NUM_ITERATIONS_SPEED; k++) {
+    sub_avg_ref((uint16_t *)data_ref, data_ref);
+  }
+  aom_usec_timer_mark(&ref_timer);
+  int ref_elapsed_time = (int)aom_usec_timer_elapsed(&ref_timer);
+  aom_usec_timer_start(&timer);
+  for (int k = 0; k < NUM_ITERATIONS_SPEED; k++) {
+    sub_avg((uint16_t *)data, data);
+  }
+  aom_usec_timer_mark(&timer);
+  int elapsed_time = (int)aom_usec_timer_elapsed(&timer);
+  printSpeed(ref_elapsed_time, elapsed_time, width, height);
+  assertFaster(ref_elapsed_time, elapsed_time);
+}
+
+template <typename S, typename T, typename I>
+class CFLSubsampleTest : public ::testing::TestWithParam<S>,
+                         public CFLTestWithData<I> {
+ public:
+  virtual void SetUp() {
+    CFLTest::init(::testing::get<0>(this->GetParam()));
+    fun_420 = ::testing::get<1>(this->GetParam())(this->tx_size);
+    fun_422 = ::testing::get<2>(this->GetParam())(this->tx_size);
+    fun_444 = ::testing::get<3>(this->GetParam())(this->tx_size);
+  }
+
+ protected:
+  T fun_420;
+  T fun_422;
+  T fun_444;
+  T fun_420_ref;
+  T fun_422_ref;
+  T fun_444_ref;
+
+  void subsampleTest(T fun, T fun_ref, int sub_width, int sub_height,
+                     I (ACMRandom::*random)()) {
+    uint16_t sub_luma_pels[CFL_BUF_SQUARE];
+    uint16_t sub_luma_pels_ref[CFL_BUF_SQUARE];
+
+    for (int it = 0; it < NUM_ITERATIONS; it++) {
+      CFLTestWithData<I>::randData(random);
+      fun(this->data, CFL_BUF_LINE, sub_luma_pels);
+      fun_ref(this->data_ref, CFL_BUF_LINE, sub_luma_pels_ref);
+      assert_eq<uint16_t>(sub_luma_pels, sub_luma_pels_ref, sub_width,
+                          sub_height);
+    }
+  }
+
+  void subsampleSpeedTest(T fun, T fun_ref, I (ACMRandom::*random)()) {
+    uint16_t sub_luma_pels[CFL_BUF_SQUARE];
+    uint16_t sub_luma_pels_ref[CFL_BUF_SQUARE];
+    aom_usec_timer ref_timer;
+    aom_usec_timer timer;
+
+    CFLTestWithData<I>::randData(random);
+    aom_usec_timer_start(&ref_timer);
+    for (int k = 0; k < NUM_ITERATIONS_SPEED; k++) {
+      fun_ref(this->data_ref, CFL_BUF_LINE, sub_luma_pels);
+    }
+    aom_usec_timer_mark(&ref_timer);
+    int ref_elapsed_time = (int)aom_usec_timer_elapsed(&ref_timer);
+    aom_usec_timer_start(&timer);
+    for (int k = 0; k < NUM_ITERATIONS_SPEED; k++) {
+      fun(this->data, CFL_BUF_LINE, sub_luma_pels_ref);
+    }
+    aom_usec_timer_mark(&timer);
+    int elapsed_time = (int)aom_usec_timer_elapsed(&timer);
+    printSpeed(ref_elapsed_time, elapsed_time, this->width, this->height);
+    assertFaster(ref_elapsed_time, elapsed_time);
+  }
+};
+
+typedef cfl_subsample_lbd_fn (*get_subsample_lbd_fn)(TX_SIZE tx_size);
+typedef ::testing::tuple<TX_SIZE, get_subsample_lbd_fn, get_subsample_lbd_fn,
+                         get_subsample_lbd_fn>
+    subsample_lbd_param;
+class CFLSubsampleLBDTest
+    : public CFLSubsampleTest<subsample_lbd_param, cfl_subsample_lbd_fn,
+                              uint8_t> {
+ public:
+  virtual ~CFLSubsampleLBDTest() {}
+  virtual void SetUp() {
+    CFLSubsampleTest::SetUp();
+    fun_420_ref = cfl_get_luma_subsampling_420_lbd_c(tx_size);
+    fun_422_ref = cfl_get_luma_subsampling_422_lbd_c(tx_size);
+    fun_444_ref = cfl_get_luma_subsampling_444_lbd_c(tx_size);
+  }
+};
+
+TEST_P(CFLSubsampleLBDTest, SubsampleLBD420Test) {
+  subsampleTest(fun_420, fun_420_ref, width >> 1, height >> 1,
+                &ACMRandom::Rand8);
+}
+
+TEST_P(CFLSubsampleLBDTest, DISABLED_SubsampleLBD420SpeedTest) {
+  subsampleSpeedTest(fun_420, fun_420_ref, &ACMRandom::Rand8);
+}
+
+TEST_P(CFLSubsampleLBDTest, SubsampleLBD422Test) {
+  subsampleTest(fun_422, fun_422_ref, width >> 1, height, &ACMRandom::Rand8);
+}
+
+TEST_P(CFLSubsampleLBDTest, DISABLED_SubsampleLBD422SpeedTest) {
+  subsampleSpeedTest(fun_422, fun_422_ref, &ACMRandom::Rand8);
+}
+
+TEST_P(CFLSubsampleLBDTest, SubsampleLBD444Test) {
+  subsampleTest(fun_444, fun_444_ref, width, height, &ACMRandom::Rand8);
+}
+
+TEST_P(CFLSubsampleLBDTest, DISABLED_SubsampleLBD444SpeedTest) {
+  subsampleSpeedTest(fun_444, fun_444_ref, &ACMRandom::Rand8);
+}
+
+typedef cfl_subsample_hbd_fn (*get_subsample_hbd_fn)(TX_SIZE tx_size);
+typedef ::testing::tuple<TX_SIZE, get_subsample_hbd_fn, get_subsample_hbd_fn,
+                         get_subsample_hbd_fn>
+    subsample_hbd_param;
+class CFLSubsampleHBDTest
+    : public CFLSubsampleTest<subsample_hbd_param, cfl_subsample_hbd_fn,
+                              uint16_t> {
+ public:
+  virtual ~CFLSubsampleHBDTest() {}
+  virtual void SetUp() {
+    CFLSubsampleTest::SetUp();
+    fun_420_ref = cfl_get_luma_subsampling_420_hbd_c(tx_size);
+    fun_422_ref = cfl_get_luma_subsampling_422_hbd_c(tx_size);
+    fun_444_ref = cfl_get_luma_subsampling_444_hbd_c(tx_size);
+  }
+};
+
+TEST_P(CFLSubsampleHBDTest, SubsampleHBD420Test) {
+  subsampleTest(fun_420, fun_420_ref, width >> 1, height >> 1,
+                &ACMRandom::Rand12);
+}
+
+TEST_P(CFLSubsampleHBDTest, DISABLED_SubsampleHBD420SpeedTest) {
+  subsampleSpeedTest(fun_420, fun_420_ref, &ACMRandom::Rand12);
+}
+
+TEST_P(CFLSubsampleHBDTest, SubsampleHBD422Test) {
+  subsampleTest(fun_422, fun_422_ref, width >> 1, height, &ACMRandom::Rand12);
+}
+
+TEST_P(CFLSubsampleHBDTest, DISABLED_SubsampleHBD422SpeedTest) {
+  subsampleSpeedTest(fun_422, fun_422_ref, &ACMRandom::Rand12);
+}
+
+TEST_P(CFLSubsampleHBDTest, SubsampleHBD444Test) {
+  subsampleTest(fun_444, fun_444_ref, width, height, &ACMRandom::Rand12);
+}
+
+TEST_P(CFLSubsampleHBDTest, DISABLED_SubsampleHBD444SpeedTest) {
+  subsampleSpeedTest(fun_444, fun_444_ref, &ACMRandom::Rand12);
+}
+
+typedef cfl_predict_lbd_fn (*get_predict_fn)(TX_SIZE tx_size);
+typedef ::testing::tuple<TX_SIZE, get_predict_fn> predict_param;
+class CFLPredictTest : public ::testing::TestWithParam<predict_param>,
+                       public CFLTestWithAlignedData<uint8_t> {
+ public:
+  virtual void SetUp() {
+    CFLTest::init(::testing::get<0>(this->GetParam()));
+    predict = ::testing::get<1>(this->GetParam())(tx_size);
+    predict_ref = get_predict_lbd_fn_c(tx_size);
+  }
+  virtual ~CFLPredictTest() {}
+
+ protected:
+  cfl_predict_lbd_fn predict;
+  cfl_predict_lbd_fn predict_ref;
+};
+
+TEST_P(CFLPredictTest, PredictTest) {
+  for (int it = 0; it < NUM_ITERATIONS; it++) {
+    randData(8);
+    predict(sub_luma_pels, chroma_pels, CFL_BUF_LINE, alpha_q3);
+    predict_ref(sub_luma_pels_ref, chroma_pels_ref, CFL_BUF_LINE, alpha_q3);
+    assert_eq<uint8_t>(chroma_pels, chroma_pels_ref, width, height);
+  }
+}
+TEST_P(CFLPredictTest, DISABLED_PredictSpeedTest) {
+  aom_usec_timer ref_timer;
+  aom_usec_timer timer;
+  randData(8);
+  aom_usec_timer_start(&ref_timer);
+  for (int k = 0; k < NUM_ITERATIONS_SPEED; k++) {
+    predict_ref(sub_luma_pels_ref, chroma_pels_ref, CFL_BUF_LINE, alpha_q3);
+  }
+  aom_usec_timer_mark(&ref_timer);
+  int ref_elapsed_time = (int)aom_usec_timer_elapsed(&ref_timer);
+
+  aom_usec_timer_start(&timer);
+  for (int k = 0; k < NUM_ITERATIONS_SPEED; k++) {
+    predict(sub_luma_pels, chroma_pels, CFL_BUF_LINE, alpha_q3);
+  }
+  aom_usec_timer_mark(&timer);
+  int elapsed_time = (int)aom_usec_timer_elapsed(&timer);
+  printSpeed(ref_elapsed_time, elapsed_time, width, height);
+  assertFaster(ref_elapsed_time, elapsed_time);
+}
+
+typedef cfl_predict_hbd_fn (*get_predict_fn_hbd)(TX_SIZE tx_size);
+typedef ::testing::tuple<TX_SIZE, get_predict_fn_hbd> predict_param_hbd;
+class CFLPredictHBDTest : public ::testing::TestWithParam<predict_param_hbd>,
+                          public CFLTestWithAlignedData<uint16_t> {
+ public:
+  virtual void SetUp() {
+    CFLTest::init(::testing::get<0>(this->GetParam()));
+    predict = ::testing::get<1>(this->GetParam())(tx_size);
+    predict_ref = get_predict_hbd_fn_c(tx_size);
+  }
+  virtual ~CFLPredictHBDTest() {}
+
+ protected:
+  cfl_predict_hbd_fn predict;
+  cfl_predict_hbd_fn predict_ref;
+};
+
+TEST_P(CFLPredictHBDTest, PredictHBDTest) {
+  int bd = 12;
+  for (int it = 0; it < NUM_ITERATIONS; it++) {
+    randData(bd);
+    predict(sub_luma_pels, chroma_pels, CFL_BUF_LINE, alpha_q3, bd);
+    predict_ref(sub_luma_pels_ref, chroma_pels_ref, CFL_BUF_LINE, alpha_q3, bd);
+    assert_eq<uint16_t>(chroma_pels, chroma_pels_ref, width, height);
+  }
+}
+TEST_P(CFLPredictHBDTest, DISABLED_PredictHBDSpeedTest) {
+  aom_usec_timer ref_timer;
+  aom_usec_timer timer;
+  const int bd = 12;
+  randData(bd);
+  aom_usec_timer_start(&ref_timer);
+  for (int k = 0; k < NUM_ITERATIONS_SPEED; k++) {
+    predict_ref(sub_luma_pels_ref, chroma_pels_ref, CFL_BUF_LINE, alpha_q3, bd);
+  }
+  aom_usec_timer_mark(&ref_timer);
+  int ref_elapsed_time = (int)aom_usec_timer_elapsed(&ref_timer);
+
+  aom_usec_timer_start(&timer);
+  for (int k = 0; k < NUM_ITERATIONS_SPEED; k++) {
+    predict(sub_luma_pels, chroma_pels, CFL_BUF_LINE, alpha_q3, bd);
+  }
+  aom_usec_timer_mark(&timer);
+  int elapsed_time = (int)aom_usec_timer_elapsed(&timer);
+  printSpeed(ref_elapsed_time, elapsed_time, width, height);
+  assertFaster(ref_elapsed_time, elapsed_time);
+}
+
+#if HAVE_SSE2
+const sub_avg_param sub_avg_sizes_sse2[] = { ALL_CFL_TX_SIZES(
+    get_subtract_average_fn_sse2) };
+
+INSTANTIATE_TEST_CASE_P(SSE2, CFLSubAvgTest,
+                        ::testing::ValuesIn(sub_avg_sizes_sse2));
+
+#endif
+
+#if HAVE_SSSE3
+const subsample_lbd_param subsample_lbd_sizes_ssse3[] = {
+  ALL_CFL_TX_SIZES_SUBSAMPLE(cfl_get_luma_subsampling_420_lbd_ssse3,
+                             cfl_get_luma_subsampling_422_lbd_ssse3,
+                             cfl_get_luma_subsampling_444_lbd_ssse3)
+};
+
+const subsample_hbd_param subsample_hbd_sizes_ssse3[] = {
+  ALL_CFL_TX_SIZES_SUBSAMPLE(cfl_get_luma_subsampling_420_hbd_ssse3,
+                             cfl_get_luma_subsampling_422_hbd_ssse3,
+                             cfl_get_luma_subsampling_444_hbd_ssse3)
+};
+
+const predict_param predict_sizes_ssse3[] = { ALL_CFL_TX_SIZES(
+    get_predict_lbd_fn_ssse3) };
+
+const predict_param_hbd predict_sizes_hbd_ssse3[] = { ALL_CFL_TX_SIZES(
+    get_predict_hbd_fn_ssse3) };
+
+INSTANTIATE_TEST_CASE_P(SSSE3, CFLSubsampleLBDTest,
+                        ::testing::ValuesIn(subsample_lbd_sizes_ssse3));
+
+INSTANTIATE_TEST_CASE_P(SSSE3, CFLSubsampleHBDTest,
+                        ::testing::ValuesIn(subsample_hbd_sizes_ssse3));
+
+INSTANTIATE_TEST_CASE_P(SSSE3, CFLPredictTest,
+                        ::testing::ValuesIn(predict_sizes_ssse3));
+
+INSTANTIATE_TEST_CASE_P(SSSE3, CFLPredictHBDTest,
+                        ::testing::ValuesIn(predict_sizes_hbd_ssse3));
+#endif
+
+#if HAVE_AVX2
+const sub_avg_param sub_avg_sizes_avx2[] = { ALL_CFL_TX_SIZES(
+    get_subtract_average_fn_avx2) };
+
+const subsample_lbd_param subsample_lbd_sizes_avx2[] = {
+  ALL_CFL_TX_SIZES_SUBSAMPLE(cfl_get_luma_subsampling_420_lbd_avx2,
+                             cfl_get_luma_subsampling_422_lbd_avx2,
+                             cfl_get_luma_subsampling_444_lbd_avx2)
+};
+
+const subsample_hbd_param subsample_hbd_sizes_avx2[] = {
+  ALL_CFL_TX_SIZES_SUBSAMPLE(cfl_get_luma_subsampling_420_hbd_avx2,
+                             cfl_get_luma_subsampling_422_hbd_avx2,
+                             cfl_get_luma_subsampling_444_hbd_avx2)
+};
+
+const predict_param predict_sizes_avx2[] = { ALL_CFL_TX_SIZES(
+    get_predict_lbd_fn_avx2) };
+
+const predict_param_hbd predict_sizes_hbd_avx2[] = { ALL_CFL_TX_SIZES(
+    get_predict_hbd_fn_avx2) };
+
+INSTANTIATE_TEST_CASE_P(AVX2, CFLSubAvgTest,
+                        ::testing::ValuesIn(sub_avg_sizes_avx2));
+
+INSTANTIATE_TEST_CASE_P(AVX2, CFLSubsampleLBDTest,
+                        ::testing::ValuesIn(subsample_lbd_sizes_avx2));
+
+INSTANTIATE_TEST_CASE_P(AVX2, CFLSubsampleHBDTest,
+                        ::testing::ValuesIn(subsample_hbd_sizes_avx2));
+
+INSTANTIATE_TEST_CASE_P(AVX2, CFLPredictTest,
+                        ::testing::ValuesIn(predict_sizes_avx2));
+
+INSTANTIATE_TEST_CASE_P(AVX2, CFLPredictHBDTest,
+                        ::testing::ValuesIn(predict_sizes_hbd_avx2));
+#endif
+
+#if HAVE_NEON
+
+const sub_avg_param sub_avg_sizes_neon[] = { ALL_CFL_TX_SIZES(
+    get_subtract_average_fn_neon) };
+
+const subsample_lbd_param subsample_lbd_sizes_neon[] = {
+  ALL_CFL_TX_SIZES_SUBSAMPLE(cfl_get_luma_subsampling_420_lbd_neon,
+                             cfl_get_luma_subsampling_422_lbd_neon,
+                             cfl_get_luma_subsampling_444_lbd_neon)
+};
+
+const subsample_hbd_param subsample_hbd_sizes_neon[] = {
+  ALL_CFL_TX_SIZES_SUBSAMPLE(cfl_get_luma_subsampling_420_hbd_neon,
+                             cfl_get_luma_subsampling_422_hbd_neon,
+                             cfl_get_luma_subsampling_444_hbd_neon)
+};
+
+const predict_param predict_sizes_neon[] = { ALL_CFL_TX_SIZES(
+    get_predict_lbd_fn_neon) };
+
+const predict_param_hbd predict_sizes_hbd_neon[] = { ALL_CFL_TX_SIZES(
+    get_predict_hbd_fn_neon) };
+
+INSTANTIATE_TEST_CASE_P(NEON, CFLSubAvgTest,
+                        ::testing::ValuesIn(sub_avg_sizes_neon));
+
+INSTANTIATE_TEST_CASE_P(NEON, CFLSubsampleLBDTest,
+                        ::testing::ValuesIn(subsample_lbd_sizes_neon));
+
+INSTANTIATE_TEST_CASE_P(NEON, CFLSubsampleHBDTest,
+                        ::testing::ValuesIn(subsample_hbd_sizes_neon));
+
+INSTANTIATE_TEST_CASE_P(NEON, CFLPredictTest,
+                        ::testing::ValuesIn(predict_sizes_neon));
+
+INSTANTIATE_TEST_CASE_P(NEON, CFLPredictHBDTest,
+                        ::testing::ValuesIn(predict_sizes_hbd_neon));
+#endif
+
+#if HAVE_VSX
+const sub_avg_param sub_avg_sizes_vsx[] = { ALL_CFL_TX_SIZES(
+    get_subtract_average_fn_vsx) };
+
+INSTANTIATE_TEST_CASE_P(VSX, CFLSubAvgTest,
+                        ::testing::ValuesIn(sub_avg_sizes_vsx));
+#endif
+}  // namespace

third_party/aom/test/clear_system_state.h

@@ -11,7 +11,8 @@
 #ifndef TEST_CLEAR_SYSTEM_STATE_H_
 #define TEST_CLEAR_SYSTEM_STATE_H_
 
-#include "./aom_config.h"
+#include "config/aom_config.h"
+
 #if ARCH_X86 || ARCH_X86_64
 #include "aom_ports/x86.h"
 #endif

third_party/aom/test/clpf_test.cc

@@ -1,437 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
-
-#include <cstdlib>
-#include <string>
-
-#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
-
-#include "./aom_config.h"
-#include "./av1_rtcd.h"
-#include "aom_ports/aom_timer.h"
-#include "av1/common/cdef_block.h"
-#include "test/acm_random.h"
-#include "test/clear_system_state.h"
-#include "test/register_state_check.h"
-#include "test/util.h"
-
-using libaom_test::ACMRandom;
-
-namespace {
-
-typedef void (*clpf_block_t)(uint8_t *dst, const uint16_t *src, int dstride,
-                             int sstride, int sizex, int sizey,
-                             unsigned int strength, unsigned int bitdepth);
-
-typedef std::tr1::tuple<clpf_block_t, clpf_block_t, int, int>
-    clpf_block_param_t;
-
-class CDEFClpfBlockTest : public ::testing::TestWithParam<clpf_block_param_t> {
- public:
-  virtual ~CDEFClpfBlockTest() {}
-  virtual void SetUp() {
-    clpf = GET_PARAM(0);
-    ref_clpf = GET_PARAM(1);
-    sizex = GET_PARAM(2);
-    sizey = GET_PARAM(3);
-  }
-
-  virtual void TearDown() { libaom_test::ClearSystemState(); }
-
- protected:
-  int sizex;
-  int sizey;
-  clpf_block_t clpf;
-  clpf_block_t ref_clpf;
-};
-
-typedef CDEFClpfBlockTest CDEFClpfSpeedTest;
-
-#if CONFIG_HIGHBITDEPTH
-typedef void (*clpf_block_hbd_t)(uint16_t *dst, const uint16_t *src,
-                                 int dstride, int sstride, int sizex, int sizey,
-                                 unsigned int strength, unsigned int bitdepth);
-
-typedef std::tr1::tuple<clpf_block_hbd_t, clpf_block_hbd_t, int, int>
-    clpf_block_hbd_param_t;
-
-class CDEFClpfBlockHbdTest
-    : public ::testing::TestWithParam<clpf_block_hbd_param_t> {
- public:
-  virtual ~CDEFClpfBlockHbdTest() {}
-  virtual void SetUp() {
-    clpf = GET_PARAM(0);
-    ref_clpf = GET_PARAM(1);
-    sizex = GET_PARAM(2);
-    sizey = GET_PARAM(3);
-  }
-
-  virtual void TearDown() { libaom_test::ClearSystemState(); }
-
- protected:
-  int sizex;
-  int sizey;
-  clpf_block_hbd_t clpf;
-  clpf_block_hbd_t ref_clpf;
-};
-
-typedef CDEFClpfBlockHbdTest ClpfHbdSpeedTest;
-#endif
-
-template <typename pixel>
-void test_clpf(int w, int h, unsigned int depth, unsigned int iterations,
-               void (*clpf)(pixel *dst, const uint16_t *src, int dstride,
-                            int sstride, int sizex, int sizey,
-                            unsigned int strength, unsigned int bitdepth),
-               void (*ref_clpf)(pixel *dst, const uint16_t *src, int dstride,
-                                int sstride, int sizex, int sizey,
-                                unsigned int strength, unsigned int bitdepth)) {
-  const int size = 24;
-  ACMRandom rnd(ACMRandom::DeterministicSeed());
-  DECLARE_ALIGNED(16, uint16_t, s[size * size]);
-  DECLARE_ALIGNED(16, pixel, d[size * size]);
-  DECLARE_ALIGNED(16, pixel, ref_d[size * size]);
-  memset(ref_d, 0, size * size * sizeof(*ref_d));
-  memset(d, 0, size * size * sizeof(*d));
-
-  int error = 0, pos = 0, xpos = 8, ypos = 8;
-  unsigned int strength = 0, bits, level, count, damp = 0, boundary = 0;
-
-  assert(size >= w + 16 && size >= h + 16);
-  assert(depth >= 8);
-
-  // Test every combination of:
-  // * Input with up to <depth> bits of noise
-  // * Noise level around every value from 0 to (1<<depth)-1
-  // * All strengths
-  // * All dampings
-  // * Boundaries
-  // If clpf and ref_clpf are the same, we're just testing speed
-  for (boundary = 0; boundary < 16; boundary++) {
-    for (count = 0; count < iterations; count++) {
-      for (level = 0; level < (1U << depth) && !error;
-           level += (1 + 4 * !!boundary) << (depth - 8)) {
-        for (bits = 1; bits <= depth && !error; bits++) {
-          for (damp = 4 + depth - 8; damp < depth - 1 && !error; damp++) {
-            for (int i = 0; i < size * size; i++)
-              s[i] = clamp((rnd.Rand16() & ((1 << bits) - 1)) + level, 0,
-                           (1 << depth) - 1);
-            if (boundary) {
-              if (boundary & 1) {  // Left
-                for (int i = 0; i < size; i++)
-                  for (int j = 0; j < xpos; j++)
-                    s[i * size + j] = CDEF_VERY_LARGE;
-              }
-              if (boundary & 2) {  // Right
-                for (int i = 0; i < size; i++)
-                  for (int j = xpos + w; j < size; j++)
-                    s[i * size + j] = CDEF_VERY_LARGE;
-              }
-              if (boundary & 4) {  // Above
-                for (int i = 0; i < ypos; i++)
-                  for (int j = 0; j < size; j++)
-                    s[i * size + j] = CDEF_VERY_LARGE;
-              }
-              if (boundary & 8) {  // Below
-                for (int i = ypos + h; i < size; i++)
-                  for (int j = 0; j < size; j++)
-                    s[i * size + j] = CDEF_VERY_LARGE;
-              }
-            }
-            for (strength = depth - 8; strength < depth - 5 && !error;
-                 strength += !error) {
-              ref_clpf(ref_d + ypos * size + xpos, s + ypos * size + xpos, size,
-                       size, w, h, 1 << strength, damp);
-              if (clpf != ref_clpf)
-                ASM_REGISTER_STATE_CHECK(clpf(d + ypos * size + xpos,
-                                              s + ypos * size + xpos, size,
-                                              size, w, h, 1 << strength, damp));
-              if (ref_clpf != clpf) {
-                for (pos = 0; pos < size * size && !error; pos++) {
-                  error = ref_d[pos] != d[pos];
-                }
-              }
-            }
-          }
-        }
-      }
-    }
-  }
-
-  pos--;
-  EXPECT_EQ(0, error)
-      << "Error: CDEFClpfBlockTest, SIMD and C mismatch." << std::endl
-      << "First error at " << pos % size << "," << pos / size << " ("
-      << (int16_t)ref_d[pos] << " != " << (int16_t)d[pos] << ") " << std::endl
-      << "strength: " << (1 << strength) << std::endl
-      << "damping: " << damp << std::endl
-      << "depth: " << depth << std::endl
-      << "boundary: " << boundary << std::endl
-      << "w: " << w << std::endl
-      << "h: " << h << std::endl
-      << "A=" << (pos > 2 * size ? (int16_t)s[pos - 2 * size] : -1) << std::endl
-      << "B=" << (pos > size ? (int16_t)s[pos - size] : -1) << std::endl
-      << "C=" << (pos % size - 2 >= 0 ? (int16_t)s[pos - 2] : -1) << std::endl
-      << "D=" << (pos % size - 1 >= 0 ? (int16_t)s[pos - 1] : -1) << std::endl
-      << "X=" << (int16_t)s[pos] << std::endl
-      << "E=" << (pos % size + 1 < size ? (int16_t)s[pos + 1] : -1) << std::endl
-      << "F=" << (pos % size + 2 < size ? (int16_t)s[pos + 2] : -1) << std::endl
-      << "G=" << (pos + size < size * size ? (int16_t)s[pos + size] : -1)
-      << std::endl
-      << "H="
-      << (pos + 2 * size < size * size ? (int16_t)s[pos + 2 * size] : -1)
-      << std::endl;
-}
-
-template <typename pixel>
-void test_clpf_speed(int w, int h, unsigned int depth, unsigned int iterations,
-                     void (*clpf)(pixel *dst, const uint16_t *src, int dstride,
-                                  int sstride, int sizex, int sizey,
-                                  unsigned int strength, unsigned int bitdepth),
-                     void (*ref_clpf)(pixel *dst, const uint16_t *src,
-                                      int dstride, int sstride, int sizex,
-                                      int sizey, unsigned int strength,
-                                      unsigned int bitdepth)) {
-  aom_usec_timer ref_timer;
-  aom_usec_timer timer;
-
-  aom_usec_timer_start(&ref_timer);
-  test_clpf(w, h, depth, iterations, ref_clpf, ref_clpf);
-  aom_usec_timer_mark(&ref_timer);
-  int ref_elapsed_time = (int)aom_usec_timer_elapsed(&ref_timer);
-
-  aom_usec_timer_start(&timer);
-  test_clpf(w, h, depth, iterations, clpf, clpf);
-  aom_usec_timer_mark(&timer);
-  int elapsed_time = (int)aom_usec_timer_elapsed(&timer);
-
-#if 0
-  std::cout << "[          ] C time = " << ref_elapsed_time / 1000
-            << " ms, SIMD time = " << elapsed_time / 1000 << " ms" << std::endl;
-#endif
-
-  EXPECT_GT(ref_elapsed_time, elapsed_time)
-      << "Error: CDEFClpfSpeedTest, SIMD slower than C." << std::endl
-      << "C time: " << ref_elapsed_time << " us" << std::endl
-      << "SIMD time: " << elapsed_time << " us" << std::endl;
-}
-
-TEST_P(CDEFClpfBlockTest, TestSIMDNoMismatch) {
-  test_clpf(sizex, sizey, 8, 1, clpf, ref_clpf);
-}
-
-TEST_P(CDEFClpfSpeedTest, DISABLED_TestSpeed) {
-  test_clpf_speed(sizex, sizey, 8, 16, clpf, ref_clpf);
-}
-
-#if CONFIG_HIGHBITDEPTH
-TEST_P(CDEFClpfBlockHbdTest, TestSIMDNoMismatch) {
-  test_clpf(sizex, sizey, 12, 1, clpf, ref_clpf);
-}
-
-TEST_P(ClpfHbdSpeedTest, DISABLED_TestSpeed) {
-  test_clpf_speed(sizex, sizey, 12, 4, clpf, ref_clpf);
-}
-#endif
-
-using std::tr1::make_tuple;
-
-// VS compiling for 32 bit targets does not support vector types in
-// structs as arguments, which makes the v256 type of the intrinsics
-// hard to support, so optimizations for this target are disabled.
-#if defined(_WIN64) || !defined(_MSC_VER) || defined(__clang__)
-// Test all supported architectures and block sizes
-#if HAVE_SSE2
-INSTANTIATE_TEST_CASE_P(
-    SSE2, CDEFClpfBlockTest,
-    ::testing::Values(
-        make_tuple(&aom_clpf_block_sse2, &aom_clpf_block_c, 8, 8),
-        make_tuple(&aom_clpf_block_sse2, &aom_clpf_block_c, 8, 4),
-        make_tuple(&aom_clpf_block_sse2, &aom_clpf_block_c, 4, 8),
-        make_tuple(&aom_clpf_block_sse2, &aom_clpf_block_c, 4, 4),
-        make_tuple(&aom_clpf_hblock_sse2, &aom_clpf_hblock_c, 8, 8),
-        make_tuple(&aom_clpf_hblock_sse2, &aom_clpf_hblock_c, 8, 4),
-        make_tuple(&aom_clpf_hblock_sse2, &aom_clpf_hblock_c, 4, 8),
-        make_tuple(&aom_clpf_hblock_sse2, &aom_clpf_hblock_c, 4, 4)));
-#endif
-
-#if HAVE_SSSE3
-INSTANTIATE_TEST_CASE_P(
-    SSSE3, CDEFClpfBlockTest,
-    ::testing::Values(
-        make_tuple(&aom_clpf_block_ssse3, &aom_clpf_block_c, 8, 8),
-        make_tuple(&aom_clpf_block_ssse3, &aom_clpf_block_c, 8, 4),
-        make_tuple(&aom_clpf_block_ssse3, &aom_clpf_block_c, 4, 8),
-        make_tuple(&aom_clpf_block_ssse3, &aom_clpf_block_c, 4, 4),
-        make_tuple(&aom_clpf_hblock_ssse3, &aom_clpf_hblock_c, 8, 8),
-        make_tuple(&aom_clpf_hblock_ssse3, &aom_clpf_hblock_c, 8, 4),
-        make_tuple(&aom_clpf_hblock_ssse3, &aom_clpf_hblock_c, 4, 8),
-        make_tuple(&aom_clpf_hblock_ssse3, &aom_clpf_hblock_c, 4, 4)));
-#endif
-
-#if HAVE_SSE4_1
-INSTANTIATE_TEST_CASE_P(
-    SSE4_1, CDEFClpfBlockTest,
-    ::testing::Values(
-        make_tuple(&aom_clpf_block_sse4_1, &aom_clpf_block_c, 8, 8),
-        make_tuple(&aom_clpf_block_sse4_1, &aom_clpf_block_c, 8, 4),
-        make_tuple(&aom_clpf_block_sse4_1, &aom_clpf_block_c, 4, 8),
-        make_tuple(&aom_clpf_block_sse4_1, &aom_clpf_block_c, 4, 4),
-        make_tuple(&aom_clpf_hblock_sse4_1, &aom_clpf_hblock_c, 8, 8),
-        make_tuple(&aom_clpf_hblock_sse4_1, &aom_clpf_hblock_c, 8, 4),
-        make_tuple(&aom_clpf_hblock_sse4_1, &aom_clpf_hblock_c, 4, 8),
-        make_tuple(&aom_clpf_hblock_sse4_1, &aom_clpf_hblock_c, 4, 4)));
-#endif
-
-#if HAVE_NEON
-INSTANTIATE_TEST_CASE_P(
-    NEON, CDEFClpfBlockTest,
-    ::testing::Values(
-        make_tuple(&aom_clpf_block_neon, &aom_clpf_block_c, 8, 8),
-        make_tuple(&aom_clpf_block_neon, &aom_clpf_block_c, 8, 4),
-        make_tuple(&aom_clpf_block_neon, &aom_clpf_block_c, 4, 8),
-        make_tuple(&aom_clpf_block_neon, &aom_clpf_block_c, 4, 4),
-        make_tuple(&aom_clpf_hblock_neon, &aom_clpf_hblock_c, 8, 8),
-        make_tuple(&aom_clpf_hblock_neon, &aom_clpf_hblock_c, 8, 4),
-        make_tuple(&aom_clpf_hblock_neon, &aom_clpf_hblock_c, 4, 8),
-        make_tuple(&aom_clpf_hblock_neon, &aom_clpf_hblock_c, 4, 4)));
-#endif
-
-#if CONFIG_HIGHBITDEPTH
-#if HAVE_SSE2
-INSTANTIATE_TEST_CASE_P(
-    SSE2, CDEFClpfBlockHbdTest,
-    ::testing::Values(
-        make_tuple(&aom_clpf_block_hbd_sse2, &aom_clpf_block_hbd_c, 8, 8),
-        make_tuple(&aom_clpf_block_hbd_sse2, &aom_clpf_block_hbd_c, 8, 4),
-        make_tuple(&aom_clpf_block_hbd_sse2, &aom_clpf_block_hbd_c, 4, 8),
-        make_tuple(&aom_clpf_block_hbd_sse2, &aom_clpf_block_hbd_c, 4, 4),
-        make_tuple(&aom_clpf_hblock_hbd_sse2, &aom_clpf_hblock_hbd_c, 8, 8),
-        make_tuple(&aom_clpf_hblock_hbd_sse2, &aom_clpf_hblock_hbd_c, 8, 4),
-        make_tuple(&aom_clpf_hblock_hbd_sse2, &aom_clpf_hblock_hbd_c, 4, 8),
-        make_tuple(&aom_clpf_hblock_hbd_sse2, &aom_clpf_hblock_hbd_c, 4, 4)));
-#endif
-
-#if HAVE_SSSE3
-INSTANTIATE_TEST_CASE_P(
-    SSSE3, CDEFClpfBlockHbdTest,
-    ::testing::Values(
-        make_tuple(&aom_clpf_block_hbd_ssse3, &aom_clpf_block_hbd_c, 8, 8),
-        make_tuple(&aom_clpf_block_hbd_ssse3, &aom_clpf_block_hbd_c, 8, 4),
-        make_tuple(&aom_clpf_block_hbd_ssse3, &aom_clpf_block_hbd_c, 4, 8),
-        make_tuple(&aom_clpf_block_hbd_ssse3, &aom_clpf_block_hbd_c, 4, 4),
-        make_tuple(&aom_clpf_hblock_hbd_ssse3, &aom_clpf_hblock_hbd_c, 8, 8),
-        make_tuple(&aom_clpf_hblock_hbd_ssse3, &aom_clpf_hblock_hbd_c, 8, 4),
-        make_tuple(&aom_clpf_hblock_hbd_ssse3, &aom_clpf_hblock_hbd_c, 4, 8),
-        make_tuple(&aom_clpf_hblock_hbd_ssse3, &aom_clpf_hblock_hbd_c, 4, 4)));
-#endif
-
-#if HAVE_SSE4_1
-INSTANTIATE_TEST_CASE_P(
-    SSE4_1, CDEFClpfBlockHbdTest,
-    ::testing::Values(
-        make_tuple(&aom_clpf_block_hbd_sse4_1, &aom_clpf_block_hbd_c, 8, 8),
-        make_tuple(&aom_clpf_block_hbd_sse4_1, &aom_clpf_block_hbd_c, 8, 4),
-        make_tuple(&aom_clpf_block_hbd_sse4_1, &aom_clpf_block_hbd_c, 4, 8),
-        make_tuple(&aom_clpf_block_hbd_sse4_1, &aom_clpf_block_hbd_c, 4, 4),
-        make_tuple(&aom_clpf_hblock_hbd_sse4_1, &aom_clpf_hblock_hbd_c, 8, 8),
-        make_tuple(&aom_clpf_hblock_hbd_sse4_1, &aom_clpf_hblock_hbd_c, 8, 4),
-        make_tuple(&aom_clpf_hblock_hbd_sse4_1, &aom_clpf_hblock_hbd_c, 4, 8),
-        make_tuple(&aom_clpf_hblock_hbd_sse4_1, &aom_clpf_hblock_hbd_c, 4, 4)));
-#endif
-
-#if HAVE_NEON
-INSTANTIATE_TEST_CASE_P(
-    NEON, CDEFClpfBlockHbdTest,
-    ::testing::Values(
-        make_tuple(&aom_clpf_block_hbd_neon, &aom_clpf_block_hbd_c, 8, 8),
-        make_tuple(&aom_clpf_block_hbd_neon, &aom_clpf_block_hbd_c, 8, 4),
-        make_tuple(&aom_clpf_block_hbd_neon, &aom_clpf_block_hbd_c, 4, 8),
-        make_tuple(&aom_clpf_block_hbd_neon, &aom_clpf_block_hbd_c, 4, 4),
-        make_tuple(&aom_clpf_hblock_hbd_neon, &aom_clpf_hblock_hbd_c, 8, 8),
-        make_tuple(&aom_clpf_hblock_hbd_neon, &aom_clpf_hblock_hbd_c, 8, 4),
-        make_tuple(&aom_clpf_hblock_hbd_neon, &aom_clpf_hblock_hbd_c, 4, 8),
-        make_tuple(&aom_clpf_hblock_hbd_neon, &aom_clpf_hblock_hbd_c, 4, 4)));
-#endif
-#endif  // CONFIG_HIGHBITDEPTH
-
-// Test speed for all supported architectures
-#if HAVE_SSE2
-INSTANTIATE_TEST_CASE_P(
-    SSE2, CDEFClpfSpeedTest,
-    ::testing::Values(make_tuple(&aom_clpf_block_sse2, &aom_clpf_block_c, 8, 8),
-                      make_tuple(&aom_clpf_hblock_sse2, &aom_clpf_hblock_c, 8,
-                                 8)));
-#endif
-
-#if HAVE_SSSE3
-INSTANTIATE_TEST_CASE_P(SSSE3, CDEFClpfSpeedTest,
-                        ::testing::Values(make_tuple(&aom_clpf_block_ssse3,
-                                                     &aom_clpf_block_c, 8, 8),
-                                          make_tuple(&aom_clpf_hblock_ssse3,
-                                                     &aom_clpf_hblock_c, 8,
-                                                     8)));
-#endif
-
-#if HAVE_SSE4_1
-INSTANTIATE_TEST_CASE_P(SSE4_1, CDEFClpfSpeedTest,
-                        ::testing::Values(make_tuple(&aom_clpf_block_sse4_1,
-                                                     &aom_clpf_block_c, 8, 8),
-                                          make_tuple(&aom_clpf_hblock_sse4_1,
-                                                     &aom_clpf_hblock_c, 8,
-                                                     8)));
-
-#endif
-
-#if HAVE_NEON
-INSTANTIATE_TEST_CASE_P(
-    NEON, CDEFClpfSpeedTest,
-    ::testing::Values(make_tuple(&aom_clpf_block_neon, &aom_clpf_block_c, 8, 8),
-                      make_tuple(&aom_clpf_hblock_neon, &aom_clpf_hblock_c, 8,
-                                 8)));
-#endif
-
-#if CONFIG_HIGHBITDEPTH
-#if HAVE_SSE2
-INSTANTIATE_TEST_CASE_P(
-    SSE2, ClpfHbdSpeedTest,
-    ::testing::Values(
-        make_tuple(&aom_clpf_block_hbd_sse2, &aom_clpf_block_hbd_c, 8, 8),
-        make_tuple(&aom_clpf_hblock_hbd_sse2, &aom_clpf_hblock_hbd_c, 8, 8)));
-#endif
-
-#if HAVE_SSSE3
-INSTANTIATE_TEST_CASE_P(
-    SSSE3, ClpfHbdSpeedTest,
-    ::testing::Values(
-        make_tuple(&aom_clpf_block_hbd_ssse3, &aom_clpf_block_hbd_c, 8, 8),
-        make_tuple(&aom_clpf_hblock_hbd_ssse3, &aom_clpf_hblock_hbd_c, 8, 8)));
-#endif
-
-#if HAVE_SSE4_1
-INSTANTIATE_TEST_CASE_P(
-    SSE4_1, ClpfHbdSpeedTest,
-    ::testing::Values(
-        make_tuple(&aom_clpf_block_hbd_sse4_1, &aom_clpf_block_hbd_c, 8, 8),
-        make_tuple(&aom_clpf_hblock_hbd_sse4_1, &aom_clpf_hblock_hbd_c, 8, 8)));
-#endif
-
-#if HAVE_NEON
-INSTANTIATE_TEST_CASE_P(
-    NEON, ClpfHbdSpeedTest,
-    ::testing::Values(
-        make_tuple(&aom_clpf_block_hbd_neon, &aom_clpf_block_hbd_c, 8, 8),
-        make_tuple(&aom_clpf_hblock_hbd_neon, &aom_clpf_hblock_hbd_c, 8, 8)));
-#endif
-#endif  // CONFIG_HIGHBITDEPTH
-#endif  // defined(_WIN64) || !defined(_MSC_VER)
-
-}  // namespace

third_party/aom/test/codec_factory.h

@@ -11,7 +11,8 @@
 #ifndef TEST_CODEC_FACTORY_H_
 #define TEST_CODEC_FACTORY_H_
 
-#include "./aom_config.h"
+#include "config/aom_config.h"
+
 #include "aom/aom_decoder.h"
 #include "aom/aom_encoder.h"
 #if CONFIG_AV1_ENCODER
@@ -39,7 +40,6 @@ class CodecFactory {
                                  const aom_codec_flags_t flags) const = 0;
 
   virtual Encoder *CreateEncoder(aom_codec_enc_cfg_t cfg,
-                                 unsigned long deadline,
                                  const unsigned long init_flags,
                                  TwopassStatsStore *stats) const = 0;
 
@@ -54,22 +54,26 @@ class CodecFactory {
 template <class T1>
 class CodecTestWithParam
     : public ::testing::TestWithParam<
-          std::tr1::tuple<const libaom_test::CodecFactory *, T1> > {};
+          ::testing::tuple<const libaom_test::CodecFactory *, T1> > {};
 
 template <class T1, class T2>
 class CodecTestWith2Params
     : public ::testing::TestWithParam<
-          std::tr1::tuple<const libaom_test::CodecFactory *, T1, T2> > {};
+          ::testing::tuple<const libaom_test::CodecFactory *, T1, T2> > {};
 
 template <class T1, class T2, class T3>
 class CodecTestWith3Params
     : public ::testing::TestWithParam<
-          std::tr1::tuple<const libaom_test::CodecFactory *, T1, T2, T3> > {};
+          ::testing::tuple<const libaom_test::CodecFactory *, T1, T2, T3> > {};
+
+template <class T1, class T2, class T3, class T4>
+class CodecTestWith4Params
+    : public ::testing::TestWithParam< ::testing::tuple<
+          const libaom_test::CodecFactory *, T1, T2, T3, T4> > {};
 
 /*
  * AV1 Codec Definitions
  */
-#if CONFIG_AV1
 class AV1Decoder : public Decoder {
  public:
   explicit AV1Decoder(aom_codec_dec_cfg_t cfg) : Decoder(cfg) {}
@@ -89,9 +93,9 @@ class AV1Decoder : public Decoder {
 
 class AV1Encoder : public Encoder {
  public:
-  AV1Encoder(aom_codec_enc_cfg_t cfg, unsigned long deadline,
-             const unsigned long init_flags, TwopassStatsStore *stats)
-      : Encoder(cfg, deadline, init_flags, stats) {}
+  AV1Encoder(aom_codec_enc_cfg_t cfg, const uint32_t init_flags,
+             TwopassStatsStore *stats)
+      : Encoder(cfg, init_flags, stats) {}
 
  protected:
   virtual aom_codec_iface_t *CodecInterface() const {
@@ -123,14 +127,12 @@ class AV1CodecFactory : public CodecFactory {
   }
 
   virtual Encoder *CreateEncoder(aom_codec_enc_cfg_t cfg,
-                                 unsigned long deadline,
                                  const unsigned long init_flags,
                                  TwopassStatsStore *stats) const {
 #if CONFIG_AV1_ENCODER
-    return new AV1Encoder(cfg, deadline, init_flags, stats);
+    return new AV1Encoder(cfg, init_flags, stats);
 #else
     (void)cfg;
-    (void)deadline;
     (void)init_flags;
     (void)stats;
     return NULL;
@@ -158,9 +160,6 @@ const libaom_test::AV1CodecFactory kAV1;
           ::testing::Values(static_cast<const libaom_test::CodecFactory *>( \
               &libaom_test::kAV1)),                                         \
           __VA_ARGS__))
-#else
-#define AV1_INSTANTIATE_TEST_CASE(test, ...)
-#endif  // CONFIG_AV1
 
 }  // namespace libaom_test
 #endif  // TEST_CODEC_FACTORY_H_

third_party/aom/test/coding_path_sync.cc

@@ -13,7 +13,7 @@
 #include "third_party/googletest/src/googletest/include/gtest/gtest.h"
 #include "test/acm_random.h"
 
-#include "./aom_config.h"
+#include "config/aom_config.h"
 
 #include "aom_ports/mem.h"  // ROUND_POWER_OF_TWO
 #include "aom/aomcx.h"
@@ -32,11 +32,17 @@ class CompressedSource {
     aom_codec_enc_cfg_t cfg;
     aom_codec_enc_config_default(algo, &cfg, 0);
 
-    const int max_q = cfg.rc_max_quantizer;
+    // force the quantizer, to reduce the sensitivity on encoding choices.
+    // e.g, we don't want this test to break when the rate control is modified.
+    {
+      const int max_q = cfg.rc_max_quantizer;
+      const int min_q = cfg.rc_min_quantizer;
+      const int q = rnd_.PseudoUniform(max_q - min_q + 1) + min_q;
 
-    cfg.rc_end_usage = AOM_CQ;
-    cfg.rc_max_quantizer = max_q;
-    cfg.rc_min_quantizer = max_q;
+      cfg.rc_end_usage = AOM_Q;
+      cfg.rc_max_quantizer = q;
+      cfg.rc_min_quantizer = q;
+    }
 
     // choose the picture size
     {
@@ -44,9 +50,26 @@ class CompressedSource {
       height_ = rnd_.PseudoUniform(kHeight - 8) + 8;
     }
 
+    // choose the chroma subsampling
+    {
+      const aom_img_fmt_t fmts[] = {
+        AOM_IMG_FMT_I420,
+        AOM_IMG_FMT_I422,
+        AOM_IMG_FMT_I444,
+      };
+
+      format_ = fmts[rnd_.PseudoUniform(NELEMENTS(fmts))];
+    }
+
     cfg.g_w = width_;
     cfg.g_h = height_;
     cfg.g_lag_in_frames = 0;
+    if (format_ == AOM_IMG_FMT_I420)
+      cfg.g_profile = 0;
+    else if (format_ == AOM_IMG_FMT_I444)
+      cfg.g_profile = 1;
+    else if (format_ == AOM_IMG_FMT_I422)
+      cfg.g_profile = 2;
 
     aom_codec_enc_init(&enc_, algo, &cfg, 0);
   }
@@ -54,7 +77,7 @@ class CompressedSource {
   ~CompressedSource() { aom_codec_destroy(&enc_); }
 
   const aom_codec_cx_pkt_t *ReadFrame() {
-    uint8_t buf[kWidth * kHeight * 3 / 2] = { 0 };
+    uint8_t buf[kWidth * kHeight * 3] = { 0 };
 
     // render regular pattern
     const int period = rnd_.Rand8() % 32 + 1;
@@ -67,8 +90,8 @@ class CompressedSource {
       buf[i] = (i + phase) % period < period / 2 ? val_a : val_b;
 
     aom_image_t img;
-    aom_img_wrap(&img, AOM_IMG_FMT_I420, width_, height_, 0, buf);
-    aom_codec_encode(&enc_, &img, frame_count_++, 1, 0, 0);
+    aom_img_wrap(&img, format_, width_, height_, 0, buf);
+    aom_codec_encode(&enc_, &img, frame_count_++, 1, 0);
 
     aom_codec_iter_t iter = NULL;
 
@@ -86,6 +109,7 @@ class CompressedSource {
   static const int kHeight = 128;
 
   ACMRandom rnd_;
+  aom_img_fmt_t format_;
   aom_codec_ctx_t enc_;
   int frame_count_;
   int width_, height_;
@@ -128,7 +152,7 @@ class Decoder {
 
   std::vector<int16_t> decode(const aom_codec_cx_pkt_t *pkt) {
     aom_codec_decode(&dec_, static_cast<uint8_t *>(pkt->data.frame.buf),
-                     static_cast<unsigned int>(pkt->data.frame.sz), NULL, 0);
+                     pkt->data.frame.sz, NULL);
 
     aom_codec_iter_t iter = NULL;
     return Serialize(aom_codec_get_frame(&dec_, &iter));
@@ -140,18 +164,41 @@ class Decoder {
 
 // Try to reveal a mismatch between LBD and HBD coding paths.
 TEST(CodingPathSync, SearchForHbdLbdMismatch) {
-  const int count_tests = 100;
+  const int count_tests = 10;
   for (int i = 0; i < count_tests; ++i) {
     Decoder dec_hbd(0);
     Decoder dec_lbd(1);
 
     CompressedSource enc(i);
-    const aom_codec_cx_pkt_t *frame = enc.ReadFrame();
 
-    std::vector<int16_t> lbd_yuv = dec_lbd.decode(frame);
-    std::vector<int16_t> hbd_yuv = dec_hbd.decode(frame);
+    for (int k = 0; k < 3; ++k) {
+      const aom_codec_cx_pkt_t *frame = enc.ReadFrame();
 
-    ASSERT_EQ(lbd_yuv, hbd_yuv);
+      std::vector<int16_t> lbd_yuv = dec_lbd.decode(frame);
+      std::vector<int16_t> hbd_yuv = dec_hbd.decode(frame);
+
+      ASSERT_EQ(lbd_yuv, hbd_yuv);
+    }
+  }
+}
+
+TEST(CodingPathSyncLarge, SearchForHbdLbdMismatchLarge) {
+  const int count_tests = 100;
+  const int seed = 1234;
+  for (int i = 0; i < count_tests; ++i) {
+    Decoder dec_hbd(0);
+    Decoder dec_lbd(1);
+
+    CompressedSource enc(seed + i);
+
+    for (int k = 0; k < 5; ++k) {
+      const aom_codec_cx_pkt_t *frame = enc.ReadFrame();
+
+      std::vector<int16_t> lbd_yuv = dec_lbd.decode(frame);
+      std::vector<int16_t> hbd_yuv = dec_hbd.decode(frame);
+
+      ASSERT_EQ(lbd_yuv, hbd_yuv);
+    }
   }
 }
 

third_party/aom/test/comp_avg_pred_test.cc

@@ -0,0 +1,72 @@
+/*
+ * Copyright (c) 2018, Alliance for Open Media. All rights reserved
+ *
+ * This source code is subject to the terms of the BSD 2 Clause License and
+ * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
+ * was not distributed with this source code in the LICENSE file, you can
+ * obtain it at www.aomedia.org/license/software. If the Alliance for Open
+ * Media Patent License 1.0 was not distributed with this source code in the
+ * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
+ */
+
+#include "test/comp_avg_pred_test.h"
+
+using ::testing::make_tuple;
+using ::testing::tuple;
+using libaom_test::ACMRandom;
+using libaom_test::AV1JNTCOMPAVG::AV1HighBDJNTCOMPAVGTest;
+using libaom_test::AV1JNTCOMPAVG::AV1HighBDJNTCOMPAVGUPSAMPLEDTest;
+using libaom_test::AV1JNTCOMPAVG::AV1JNTCOMPAVGTest;
+using libaom_test::AV1JNTCOMPAVG::AV1JNTCOMPAVGUPSAMPLEDTest;
+
+namespace {
+
+TEST_P(AV1JNTCOMPAVGTest, DISABLED_Speed) { RunSpeedTest(GET_PARAM(0)); }
+
+TEST_P(AV1JNTCOMPAVGTest, CheckOutput) { RunCheckOutput(GET_PARAM(0)); }
+
+#if HAVE_SSSE3
+INSTANTIATE_TEST_CASE_P(
+    SSSE3, AV1JNTCOMPAVGTest,
+    libaom_test::AV1JNTCOMPAVG::BuildParams(aom_jnt_comp_avg_pred_ssse3));
+#endif
+
+TEST_P(AV1JNTCOMPAVGUPSAMPLEDTest, DISABLED_Speed) {
+  RunSpeedTest(GET_PARAM(0));
+}
+
+TEST_P(AV1JNTCOMPAVGUPSAMPLEDTest, CheckOutput) {
+  RunCheckOutput(GET_PARAM(0));
+}
+
+#if HAVE_SSSE3
+INSTANTIATE_TEST_CASE_P(SSSE3, AV1JNTCOMPAVGUPSAMPLEDTest,
+                        libaom_test::AV1JNTCOMPAVG::BuildParams(
+                            aom_jnt_comp_avg_upsampled_pred_ssse3));
+#endif
+
+TEST_P(AV1HighBDJNTCOMPAVGTest, DISABLED_Speed) { RunSpeedTest(GET_PARAM(1)); }
+
+TEST_P(AV1HighBDJNTCOMPAVGTest, CheckOutput) { RunCheckOutput(GET_PARAM(1)); }
+
+#if HAVE_SSE2
+INSTANTIATE_TEST_CASE_P(
+    SSE2, AV1HighBDJNTCOMPAVGTest,
+    libaom_test::AV1JNTCOMPAVG::BuildParams(aom_highbd_jnt_comp_avg_pred_sse2));
+#endif
+
+TEST_P(AV1HighBDJNTCOMPAVGUPSAMPLEDTest, DISABLED_Speed) {
+  RunSpeedTest(GET_PARAM(1));
+}
+
+TEST_P(AV1HighBDJNTCOMPAVGUPSAMPLEDTest, CheckOutput) {
+  RunCheckOutput(GET_PARAM(1));
+}
+
+#if HAVE_SSE2
+INSTANTIATE_TEST_CASE_P(SSE2, AV1HighBDJNTCOMPAVGUPSAMPLEDTest,
+                        libaom_test::AV1JNTCOMPAVG::BuildParams(
+                            aom_highbd_jnt_comp_avg_upsampled_pred_sse2));
+#endif
+
+}  // namespace

third_party/aom/test/comp_avg_pred_test.h

@@ -0,0 +1,546 @@
+/*
+ * Copyright (c) 2018, Alliance for Open Media. All rights reserved
+ *
+ * This source code is subject to the terms of the BSD 2 Clause License and
+ * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
+ * was not distributed with this source code in the LICENSE file, you can
+ * obtain it at www.aomedia.org/license/software. If the Alliance for Open
+ * Media Patent License 1.0 was not distributed with this source code in the
+ * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
+ */
+
+#ifndef TEST_COMP_AVG_PRED_TEST_H_
+#define TEST_COMP_AVG_PRED_TEST_H_
+
+#include "config/aom_dsp_rtcd.h"
+
+#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
+#include "test/acm_random.h"
+#include "test/util.h"
+#include "test/clear_system_state.h"
+#include "test/register_state_check.h"
+#include "av1/common/common_data.h"
+#include "aom_ports/aom_timer.h"
+
+namespace libaom_test {
+const int kMaxSize = 128 + 32;  // padding
+
+namespace AV1JNTCOMPAVG {
+
+typedef void (*jntcompavg_func)(uint8_t *comp_pred, const uint8_t *pred,
+                                int width, int height, const uint8_t *ref,
+                                int ref_stride,
+                                const JNT_COMP_PARAMS *jcp_param);
+
+typedef void (*jntcompavgupsampled_func)(
+    MACROBLOCKD *xd, const struct AV1Common *const cm, int mi_row, int mi_col,
+    const MV *const mv, uint8_t *comp_pred, const uint8_t *pred, int width,
+    int height, int subpel_x_q3, int subpel_y_q3, const uint8_t *ref,
+    int ref_stride, const JNT_COMP_PARAMS *jcp_param);
+
+typedef void (*highbdjntcompavg_func)(uint16_t *comp_pred, const uint8_t *pred8,
+                                      int width, int height,
+                                      const uint8_t *ref8, int ref_stride,
+                                      const JNT_COMP_PARAMS *jcp_param);
+
+typedef void (*highbdjntcompavgupsampled_func)(
+    MACROBLOCKD *xd, const struct AV1Common *const cm, int mi_row, int mi_col,
+    const MV *const mv, uint16_t *comp_pred, const uint8_t *pred8, int width,
+    int height, int subpel_x_q3, int subpel_y_q3, const uint8_t *ref8,
+    int ref_stride, int bd, const JNT_COMP_PARAMS *jcp_param);
+
+typedef ::testing::tuple<jntcompavg_func, BLOCK_SIZE> JNTCOMPAVGParam;
+
+typedef ::testing::tuple<jntcompavgupsampled_func, BLOCK_SIZE>
+    JNTCOMPAVGUPSAMPLEDParam;
+
+typedef ::testing::tuple<int, highbdjntcompavg_func, BLOCK_SIZE>
+    HighbdJNTCOMPAVGParam;
+
+typedef ::testing::tuple<int, highbdjntcompavgupsampled_func, BLOCK_SIZE>
+    HighbdJNTCOMPAVGUPSAMPLEDParam;
+
+::testing::internal::ParamGenerator<JNTCOMPAVGParam> BuildParams(
+    jntcompavg_func filter) {
+  return ::testing::Combine(::testing::Values(filter),
+                            ::testing::Range(BLOCK_4X4, BLOCK_SIZES_ALL));
+}
+
+::testing::internal::ParamGenerator<JNTCOMPAVGUPSAMPLEDParam> BuildParams(
+    jntcompavgupsampled_func filter) {
+  return ::testing::Combine(::testing::Values(filter),
+                            ::testing::Range(BLOCK_4X4, BLOCK_SIZES_ALL));
+}
+
+::testing::internal::ParamGenerator<HighbdJNTCOMPAVGParam> BuildParams(
+    highbdjntcompavg_func filter) {
+  return ::testing::Combine(::testing::Range(8, 13, 2),
+                            ::testing::Values(filter),
+                            ::testing::Range(BLOCK_4X4, BLOCK_SIZES_ALL));
+}
+
+::testing::internal::ParamGenerator<HighbdJNTCOMPAVGUPSAMPLEDParam> BuildParams(
+    highbdjntcompavgupsampled_func filter) {
+  return ::testing::Combine(::testing::Range(8, 13, 2),
+                            ::testing::Values(filter),
+                            ::testing::Range(BLOCK_4X4, BLOCK_SIZES_ALL));
+}
+
+class AV1JNTCOMPAVGTest : public ::testing::TestWithParam<JNTCOMPAVGParam> {
+ public:
+  ~AV1JNTCOMPAVGTest() {}
+  void SetUp() { rnd_.Reset(ACMRandom::DeterministicSeed()); }
+  void TearDown() { libaom_test::ClearSystemState(); }
+
+ protected:
+  void RunCheckOutput(jntcompavg_func test_impl) {
+    const int w = kMaxSize, h = kMaxSize;
+    const int block_idx = GET_PARAM(1);
+
+    uint8_t pred8[kMaxSize * kMaxSize];
+    uint8_t ref8[kMaxSize * kMaxSize];
+    uint8_t output[kMaxSize * kMaxSize];
+    uint8_t output2[kMaxSize * kMaxSize];
+
+    for (int i = 0; i < h; ++i)
+      for (int j = 0; j < w; ++j) {
+        pred8[i * w + j] = rnd_.Rand8();
+        ref8[i * w + j] = rnd_.Rand8();
+      }
+    const int in_w = block_size_wide[block_idx];
+    const int in_h = block_size_high[block_idx];
+
+    JNT_COMP_PARAMS jnt_comp_params;
+    jnt_comp_params.use_jnt_comp_avg = 1;
+
+    for (int ii = 0; ii < 2; ii++) {
+      for (int jj = 0; jj < 4; jj++) {
+        jnt_comp_params.fwd_offset = quant_dist_lookup_table[ii][jj][0];
+        jnt_comp_params.bck_offset = quant_dist_lookup_table[ii][jj][1];
+
+        const int offset_r = 3 + rnd_.PseudoUniform(h - in_h - 7);
+        const int offset_c = 3 + rnd_.PseudoUniform(w - in_w - 7);
+        aom_jnt_comp_avg_pred_c(output, pred8 + offset_r * w + offset_c, in_w,
+                                in_h, ref8 + offset_r * w + offset_c, in_w,
+                                &jnt_comp_params);
+        test_impl(output2, pred8 + offset_r * w + offset_c, in_w, in_h,
+                  ref8 + offset_r * w + offset_c, in_w, &jnt_comp_params);
+
+        for (int i = 0; i < in_h; ++i) {
+          for (int j = 0; j < in_w; ++j) {
+            int idx = i * in_w + j;
+            ASSERT_EQ(output[idx], output2[idx])
+                << "Mismatch at unit tests for AV1JNTCOMPAVGTest\n"
+                << in_w << "x" << in_h << " Pixel mismatch at index " << idx
+                << " = (" << i << ", " << j << ")";
+          }
+        }
+      }
+    }
+  }
+  void RunSpeedTest(jntcompavg_func test_impl) {
+    const int w = kMaxSize, h = kMaxSize;
+    const int block_idx = GET_PARAM(1);
+
+    uint8_t pred8[kMaxSize * kMaxSize];
+    uint8_t ref8[kMaxSize * kMaxSize];
+    uint8_t output[kMaxSize * kMaxSize];
+    uint8_t output2[kMaxSize * kMaxSize];
+
+    for (int i = 0; i < h; ++i)
+      for (int j = 0; j < w; ++j) {
+        pred8[i * w + j] = rnd_.Rand8();
+        ref8[i * w + j] = rnd_.Rand8();
+      }
+    const int in_w = block_size_wide[block_idx];
+    const int in_h = block_size_high[block_idx];
+
+    JNT_COMP_PARAMS jnt_comp_params;
+    jnt_comp_params.use_jnt_comp_avg = 1;
+
+    jnt_comp_params.fwd_offset = quant_dist_lookup_table[0][0][0];
+    jnt_comp_params.bck_offset = quant_dist_lookup_table[0][0][1];
+
+    const int num_loops = 1000000000 / (in_w + in_h);
+    aom_usec_timer timer;
+    aom_usec_timer_start(&timer);
+
+    for (int i = 0; i < num_loops; ++i)
+      aom_jnt_comp_avg_pred_c(output, pred8, in_w, in_h, ref8, in_w,
+                              &jnt_comp_params);
+
+    aom_usec_timer_mark(&timer);
+    const int elapsed_time = static_cast<int>(aom_usec_timer_elapsed(&timer));
+    printf("jntcompavg c_code %3dx%-3d: %7.2f us\n", in_w, in_h,
+           1000.0 * elapsed_time / num_loops);
+
+    aom_usec_timer timer1;
+    aom_usec_timer_start(&timer1);
+
+    for (int i = 0; i < num_loops; ++i)
+      test_impl(output2, pred8, in_w, in_h, ref8, in_w, &jnt_comp_params);
+
+    aom_usec_timer_mark(&timer1);
+    const int elapsed_time1 = static_cast<int>(aom_usec_timer_elapsed(&timer1));
+    printf("jntcompavg test_code %3dx%-3d: %7.2f us\n", in_w, in_h,
+           1000.0 * elapsed_time1 / num_loops);
+  }
+
+  libaom_test::ACMRandom rnd_;
+};  // class AV1JNTCOMPAVGTest
+
+class AV1JNTCOMPAVGUPSAMPLEDTest
+    : public ::testing::TestWithParam<JNTCOMPAVGUPSAMPLEDParam> {
+ public:
+  ~AV1JNTCOMPAVGUPSAMPLEDTest() {}
+  void SetUp() { rnd_.Reset(ACMRandom::DeterministicSeed()); }
+  void TearDown() { libaom_test::ClearSystemState(); }
+
+ protected:
+  void RunCheckOutput(jntcompavgupsampled_func test_impl) {
+    const int w = kMaxSize, h = kMaxSize;
+    const int block_idx = GET_PARAM(1);
+
+    uint8_t pred8[kMaxSize * kMaxSize];
+    uint8_t ref8[kMaxSize * kMaxSize];
+    DECLARE_ALIGNED(16, uint8_t, output[MAX_SB_SQUARE]);
+    DECLARE_ALIGNED(16, uint8_t, output2[MAX_SB_SQUARE]);
+
+    for (int i = 0; i < h; ++i)
+      for (int j = 0; j < w; ++j) {
+        pred8[i * w + j] = rnd_.Rand8();
+        ref8[i * w + j] = rnd_.Rand8();
+      }
+    const int in_w = block_size_wide[block_idx];
+    const int in_h = block_size_high[block_idx];
+
+    JNT_COMP_PARAMS jnt_comp_params;
+    jnt_comp_params.use_jnt_comp_avg = 1;
+    int sub_x_q3, sub_y_q3;
+    for (sub_x_q3 = 0; sub_x_q3 < 8; ++sub_x_q3) {
+      for (sub_y_q3 = 0; sub_y_q3 < 8; ++sub_y_q3) {
+        for (int ii = 0; ii < 2; ii++) {
+          for (int jj = 0; jj < 4; jj++) {
+            jnt_comp_params.fwd_offset = quant_dist_lookup_table[ii][jj][0];
+            jnt_comp_params.bck_offset = quant_dist_lookup_table[ii][jj][1];
+
+            const int offset_r = 3 + rnd_.PseudoUniform(h - in_h - 7);
+            const int offset_c = 3 + rnd_.PseudoUniform(w - in_w - 7);
+
+            aom_jnt_comp_avg_upsampled_pred_c(
+                NULL, NULL, 0, 0, NULL, output, pred8 + offset_r * w + offset_c,
+                in_w, in_h, sub_x_q3, sub_y_q3, ref8 + offset_r * w + offset_c,
+                in_w, &jnt_comp_params);
+            test_impl(NULL, NULL, 0, 0, NULL, output2,
+                      pred8 + offset_r * w + offset_c, in_w, in_h, sub_x_q3,
+                      sub_y_q3, ref8 + offset_r * w + offset_c, in_w,
+                      &jnt_comp_params);
+
+            for (int i = 0; i < in_h; ++i) {
+              for (int j = 0; j < in_w; ++j) {
+                int idx = i * in_w + j;
+                ASSERT_EQ(output[idx], output2[idx])
+                    << "Mismatch at unit tests for AV1JNTCOMPAVGUPSAMPLEDTest\n"
+                    << in_w << "x" << in_h << " Pixel mismatch at index " << idx
+                    << " = (" << i << ", " << j << "), sub pixel offset = ("
+                    << sub_y_q3 << ", " << sub_x_q3 << ")";
+              }
+            }
+          }
+        }
+      }
+    }
+  }
+  void RunSpeedTest(jntcompavgupsampled_func test_impl) {
+    const int w = kMaxSize, h = kMaxSize;
+    const int block_idx = GET_PARAM(1);
+
+    uint8_t pred8[kMaxSize * kMaxSize];
+    uint8_t ref8[kMaxSize * kMaxSize];
+    DECLARE_ALIGNED(16, uint8_t, output[MAX_SB_SQUARE]);
+    DECLARE_ALIGNED(16, uint8_t, output2[MAX_SB_SQUARE]);
+
+    for (int i = 0; i < h; ++i)
+      for (int j = 0; j < w; ++j) {
+        pred8[i * w + j] = rnd_.Rand8();
+        ref8[i * w + j] = rnd_.Rand8();
+      }
+    const int in_w = block_size_wide[block_idx];
+    const int in_h = block_size_high[block_idx];
+
+    JNT_COMP_PARAMS jnt_comp_params;
+    jnt_comp_params.use_jnt_comp_avg = 1;
+
+    jnt_comp_params.fwd_offset = quant_dist_lookup_table[0][0][0];
+    jnt_comp_params.bck_offset = quant_dist_lookup_table[0][0][1];
+
+    int sub_x_q3 = 0;
+    int sub_y_q3 = 0;
+
+    const int num_loops = 1000000000 / (in_w + in_h);
+    aom_usec_timer timer;
+    aom_usec_timer_start(&timer);
+
+    for (int i = 0; i < num_loops; ++i)
+      aom_jnt_comp_avg_upsampled_pred_c(NULL, NULL, 0, 0, NULL, output, pred8,
+                                        in_w, in_h, sub_x_q3, sub_y_q3, ref8,
+                                        in_w, &jnt_comp_params);
+
+    aom_usec_timer_mark(&timer);
+    const int elapsed_time = static_cast<int>(aom_usec_timer_elapsed(&timer));
+    printf("jntcompavgupsampled c_code %3dx%-3d: %7.2f us\n", in_w, in_h,
+           1000.0 * elapsed_time / num_loops);
+
+    aom_usec_timer timer1;
+    aom_usec_timer_start(&timer1);
+
+    for (int i = 0; i < num_loops; ++i)
+      test_impl(NULL, NULL, 0, 0, NULL, output2, pred8, in_w, in_h, sub_x_q3,
+                sub_y_q3, ref8, in_w, &jnt_comp_params);
+
+    aom_usec_timer_mark(&timer1);
+    const int elapsed_time1 = static_cast<int>(aom_usec_timer_elapsed(&timer1));
+    printf("jntcompavgupsampled test_code %3dx%-3d: %7.2f us\n", in_w, in_h,
+           1000.0 * elapsed_time1 / num_loops);
+  }
+
+  libaom_test::ACMRandom rnd_;
+};  // class AV1JNTCOMPAVGUPSAMPLEDTest
+
+class AV1HighBDJNTCOMPAVGTest
+    : public ::testing::TestWithParam<HighbdJNTCOMPAVGParam> {
+ public:
+  ~AV1HighBDJNTCOMPAVGTest() {}
+  void SetUp() { rnd_.Reset(ACMRandom::DeterministicSeed()); }
+
+  void TearDown() { libaom_test::ClearSystemState(); }
+
+ protected:
+  void RunCheckOutput(highbdjntcompavg_func test_impl) {
+    const int w = kMaxSize, h = kMaxSize;
+    const int block_idx = GET_PARAM(2);
+    const int bd = GET_PARAM(0);
+    uint16_t pred8[kMaxSize * kMaxSize];
+    uint16_t ref8[kMaxSize * kMaxSize];
+    uint16_t output[kMaxSize * kMaxSize];
+    uint16_t output2[kMaxSize * kMaxSize];
+
+    for (int i = 0; i < h; ++i)
+      for (int j = 0; j < w; ++j) {
+        pred8[i * w + j] = rnd_.Rand16() & ((1 << bd) - 1);
+        ref8[i * w + j] = rnd_.Rand16() & ((1 << bd) - 1);
+      }
+    const int in_w = block_size_wide[block_idx];
+    const int in_h = block_size_high[block_idx];
+
+    JNT_COMP_PARAMS jnt_comp_params;
+    jnt_comp_params.use_jnt_comp_avg = 1;
+
+    for (int ii = 0; ii < 2; ii++) {
+      for (int jj = 0; jj < 4; jj++) {
+        jnt_comp_params.fwd_offset = quant_dist_lookup_table[ii][jj][0];
+        jnt_comp_params.bck_offset = quant_dist_lookup_table[ii][jj][1];
+
+        const int offset_r = 3 + rnd_.PseudoUniform(h - in_h - 7);
+        const int offset_c = 3 + rnd_.PseudoUniform(w - in_w - 7);
+        aom_highbd_jnt_comp_avg_pred_c(
+            output, CONVERT_TO_BYTEPTR(pred8) + offset_r * w + offset_c, in_w,
+            in_h, CONVERT_TO_BYTEPTR(ref8) + offset_r * w + offset_c, in_w,
+            &jnt_comp_params);
+        test_impl(output2, CONVERT_TO_BYTEPTR(pred8) + offset_r * w + offset_c,
+                  in_w, in_h,
+                  CONVERT_TO_BYTEPTR(ref8) + offset_r * w + offset_c, in_w,
+                  &jnt_comp_params);
+
+        for (int i = 0; i < in_h; ++i) {
+          for (int j = 0; j < in_w; ++j) {
+            int idx = i * in_w + j;
+            ASSERT_EQ(output[idx], output2[idx])
+                << "Mismatch at unit tests for AV1HighBDJNTCOMPAVGTest\n"
+                << in_w << "x" << in_h << " Pixel mismatch at index " << idx
+                << " = (" << i << ", " << j << ")";
+          }
+        }
+      }
+    }
+  }
+  void RunSpeedTest(highbdjntcompavg_func test_impl) {
+    const int w = kMaxSize, h = kMaxSize;
+    const int block_idx = GET_PARAM(2);
+    const int bd = GET_PARAM(0);
+    uint16_t pred8[kMaxSize * kMaxSize];
+    uint16_t ref8[kMaxSize * kMaxSize];
+    uint16_t output[kMaxSize * kMaxSize];
+    uint16_t output2[kMaxSize * kMaxSize];
+
+    for (int i = 0; i < h; ++i)
+      for (int j = 0; j < w; ++j) {
+        pred8[i * w + j] = rnd_.Rand16() & ((1 << bd) - 1);
+        ref8[i * w + j] = rnd_.Rand16() & ((1 << bd) - 1);
+      }
+    const int in_w = block_size_wide[block_idx];
+    const int in_h = block_size_high[block_idx];
+
+    JNT_COMP_PARAMS jnt_comp_params;
+    jnt_comp_params.use_jnt_comp_avg = 1;
+
+    jnt_comp_params.fwd_offset = quant_dist_lookup_table[0][0][0];
+    jnt_comp_params.bck_offset = quant_dist_lookup_table[0][0][1];
+
+    const int num_loops = 1000000000 / (in_w + in_h);
+    aom_usec_timer timer;
+    aom_usec_timer_start(&timer);
+
+    for (int i = 0; i < num_loops; ++i)
+      aom_highbd_jnt_comp_avg_pred_c(output, CONVERT_TO_BYTEPTR(pred8), in_w,
+                                     in_h, CONVERT_TO_BYTEPTR(ref8), in_w,
+                                     &jnt_comp_params);
+
+    aom_usec_timer_mark(&timer);
+    const int elapsed_time = static_cast<int>(aom_usec_timer_elapsed(&timer));
+    printf("highbdjntcompavg c_code %3dx%-3d: %7.2f us\n", in_w, in_h,
+           1000.0 * elapsed_time / num_loops);
+
+    aom_usec_timer timer1;
+    aom_usec_timer_start(&timer1);
+
+    for (int i = 0; i < num_loops; ++i)
+      test_impl(output2, CONVERT_TO_BYTEPTR(pred8), in_w, in_h,
+                CONVERT_TO_BYTEPTR(ref8), in_w, &jnt_comp_params);
+
+    aom_usec_timer_mark(&timer1);
+    const int elapsed_time1 = static_cast<int>(aom_usec_timer_elapsed(&timer1));
+    printf("highbdjntcompavg test_code %3dx%-3d: %7.2f us\n", in_w, in_h,
+           1000.0 * elapsed_time1 / num_loops);
+  }
+
+  libaom_test::ACMRandom rnd_;
+};  // class AV1HighBDJNTCOMPAVGTest
+
+class AV1HighBDJNTCOMPAVGUPSAMPLEDTest
+    : public ::testing::TestWithParam<HighbdJNTCOMPAVGUPSAMPLEDParam> {
+ public:
+  ~AV1HighBDJNTCOMPAVGUPSAMPLEDTest() {}
+  void SetUp() { rnd_.Reset(ACMRandom::DeterministicSeed()); }
+  void TearDown() { libaom_test::ClearSystemState(); }
+
+ protected:
+  void RunCheckOutput(highbdjntcompavgupsampled_func test_impl) {
+    const int w = kMaxSize, h = kMaxSize;
+    const int block_idx = GET_PARAM(2);
+    const int bd = GET_PARAM(0);
+    uint16_t pred8[kMaxSize * kMaxSize];
+    uint16_t ref8[kMaxSize * kMaxSize];
+    uint16_t output[kMaxSize * kMaxSize];
+    uint16_t output2[kMaxSize * kMaxSize];
+
+    for (int i = 0; i < h; ++i)
+      for (int j = 0; j < w; ++j) {
+        pred8[i * w + j] = rnd_.Rand16() & ((1 << bd) - 1);
+        ref8[i * w + j] = rnd_.Rand16() & ((1 << bd) - 1);
+      }
+    const int in_w = block_size_wide[block_idx];
+    const int in_h = block_size_high[block_idx];
+
+    JNT_COMP_PARAMS jnt_comp_params;
+    jnt_comp_params.use_jnt_comp_avg = 1;
+    int sub_x_q3, sub_y_q3;
+
+    for (sub_x_q3 = 0; sub_x_q3 < 8; ++sub_x_q3) {
+      for (sub_y_q3 = 0; sub_y_q3 < 8; ++sub_y_q3) {
+        for (int ii = 0; ii < 2; ii++) {
+          for (int jj = 0; jj < 4; jj++) {
+            jnt_comp_params.fwd_offset = quant_dist_lookup_table[ii][jj][0];
+            jnt_comp_params.bck_offset = quant_dist_lookup_table[ii][jj][1];
+
+            const int offset_r = 3 + rnd_.PseudoUniform(h - in_h - 7);
+            const int offset_c = 3 + rnd_.PseudoUniform(w - in_w - 7);
+
+            aom_highbd_jnt_comp_avg_upsampled_pred_c(
+                NULL, NULL, 0, 0, NULL, output,
+                CONVERT_TO_BYTEPTR(pred8) + offset_r * w + offset_c, in_w, in_h,
+                sub_x_q3, sub_y_q3,
+                CONVERT_TO_BYTEPTR(ref8) + offset_r * w + offset_c, in_w, bd,
+                &jnt_comp_params);
+            test_impl(NULL, NULL, 0, 0, NULL, output2,
+                      CONVERT_TO_BYTEPTR(pred8) + offset_r * w + offset_c, in_w,
+                      in_h, sub_x_q3, sub_y_q3,
+                      CONVERT_TO_BYTEPTR(ref8) + offset_r * w + offset_c, in_w,
+                      bd, &jnt_comp_params);
+
+            for (int i = 0; i < in_h; ++i) {
+              for (int j = 0; j < in_w; ++j) {
+                int idx = i * in_w + j;
+                ASSERT_EQ(output[idx], output2[idx])
+                    << "Mismatch at unit tests for "
+                       "AV1HighBDJNTCOMPAVGUPSAMPLEDTest\n"
+                    << in_w << "x" << in_h << " Pixel mismatch at index " << idx
+                    << " = (" << i << ", " << j << "), sub pixel offset = ("
+                    << sub_y_q3 << ", " << sub_x_q3 << ")";
+              }
+            }
+          }
+        }
+      }
+    }
+  }
+  void RunSpeedTest(highbdjntcompavgupsampled_func test_impl) {
+    const int w = kMaxSize, h = kMaxSize;
+    const int block_idx = GET_PARAM(2);
+    const int bd = GET_PARAM(0);
+    uint16_t pred8[kMaxSize * kMaxSize];
+    uint16_t ref8[kMaxSize * kMaxSize];
+    uint16_t output[kMaxSize * kMaxSize];
+    uint16_t output2[kMaxSize * kMaxSize];
+
+    for (int i = 0; i < h; ++i)
+      for (int j = 0; j < w; ++j) {
+        pred8[i * w + j] = rnd_.Rand16() & ((1 << bd) - 1);
+        ref8[i * w + j] = rnd_.Rand16() & ((1 << bd) - 1);
+      }
+    const int in_w = block_size_wide[block_idx];
+    const int in_h = block_size_high[block_idx];
+
+    JNT_COMP_PARAMS jnt_comp_params;
+    jnt_comp_params.use_jnt_comp_avg = 1;
+
+    jnt_comp_params.fwd_offset = quant_dist_lookup_table[0][0][0];
+    jnt_comp_params.bck_offset = quant_dist_lookup_table[0][0][1];
+    int sub_x_q3 = 0;
+    int sub_y_q3 = 0;
+    const int num_loops = 1000000000 / (in_w + in_h);
+    aom_usec_timer timer;
+    aom_usec_timer_start(&timer);
+
+    for (int i = 0; i < num_loops; ++i)
+      aom_highbd_jnt_comp_avg_upsampled_pred_c(
+          NULL, NULL, 0, 0, NULL, output, CONVERT_TO_BYTEPTR(pred8), in_w, in_h,
+          sub_x_q3, sub_y_q3, CONVERT_TO_BYTEPTR(ref8), in_w, bd,
+          &jnt_comp_params);
+
+    aom_usec_timer_mark(&timer);
+    const int elapsed_time = static_cast<int>(aom_usec_timer_elapsed(&timer));
+    printf("highbdjntcompavgupsampled c_code %3dx%-3d: %7.2f us\n", in_w, in_h,
+           1000.0 * elapsed_time / num_loops);
+
+    aom_usec_timer timer1;
+    aom_usec_timer_start(&timer1);
+
+    for (int i = 0; i < num_loops; ++i)
+      test_impl(NULL, NULL, 0, 0, NULL, output2, CONVERT_TO_BYTEPTR(pred8),
+                in_w, in_h, sub_x_q3, sub_y_q3, CONVERT_TO_BYTEPTR(ref8), in_w,
+                bd, &jnt_comp_params);
+
+    aom_usec_timer_mark(&timer1);
+    const int elapsed_time1 = static_cast<int>(aom_usec_timer_elapsed(&timer1));
+    printf("highbdjntcompavgupsampled test_code %3dx%-3d: %7.2f us\n", in_w,
+           in_h, 1000.0 * elapsed_time1 / num_loops);
+  }
+
+  libaom_test::ACMRandom rnd_;
+};  // class AV1HighBDJNTCOMPAVGUPSAMPLEDTest
+
+}  // namespace AV1JNTCOMPAVG
+}  // namespace libaom_test
+
+#endif  // TEST_COMP_AVG_PRED_TEST_H_

third_party/aom/test/comp_mask_variance_test.cc

@@ -0,0 +1,273 @@
+/*
+ * Copyright (c) 2018, Alliance for Open Media. All rights reserved
+ *
+ * This source code is subject to the terms of the BSD 2 Clause License and
+ * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
+ * was not distributed with this source code in the LICENSE file, you can
+ * obtain it at www.aomedia.org/license/software. If the Alliance for Open
+ * Media Patent License 1.0 was not distributed with this source code in the
+ * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
+ */
+
+#include <cstdlib>
+#include <new>
+
+#include "config/aom_config.h"
+#include "config/aom_dsp_rtcd.h"
+
+#include "aom/aom_codec.h"
+#include "aom/aom_integer.h"
+#include "aom_dsp/variance.h"
+#include "aom_mem/aom_mem.h"
+#include "aom_ports/aom_timer.h"
+#include "aom_ports/mem.h"
+#include "av1/common/reconinter.h"
+#include "test/acm_random.h"
+#include "test/clear_system_state.h"
+#include "test/register_state_check.h"
+#include "test/util.h"
+#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
+
+namespace AV1CompMaskVariance {
+typedef void (*comp_mask_pred_func)(uint8_t *comp_pred, const uint8_t *pred,
+                                    int width, int height, const uint8_t *ref,
+                                    int ref_stride, const uint8_t *mask,
+                                    int mask_stride, int invert_mask);
+#if HAVE_SSSE3 || HAVE_AV2
+const BLOCK_SIZE kValidBlockSize[] = {
+  BLOCK_8X8,   BLOCK_8X16, BLOCK_8X32,  BLOCK_16X8,  BLOCK_16X16,
+  BLOCK_16X32, BLOCK_32X8, BLOCK_32X16, BLOCK_32X32,
+};
+#endif
+typedef ::testing::tuple<comp_mask_pred_func, BLOCK_SIZE> CompMaskPredParam;
+
+class AV1CompMaskVarianceTest
+    : public ::testing::TestWithParam<CompMaskPredParam> {
+ public:
+  ~AV1CompMaskVarianceTest();
+  void SetUp();
+
+  void TearDown();
+
+ protected:
+  void RunCheckOutput(comp_mask_pred_func test_impl, BLOCK_SIZE bsize, int inv);
+  void RunSpeedTest(comp_mask_pred_func test_impl, BLOCK_SIZE bsize);
+  bool CheckResult(int width, int height) {
+    for (int y = 0; y < height; ++y) {
+      for (int x = 0; x < width; ++x) {
+        const int idx = y * width + x;
+        if (comp_pred1_[idx] != comp_pred2_[idx]) {
+          printf("%dx%d mismatch @%d(%d,%d) ", width, height, idx, y, x);
+          printf("%d != %d ", comp_pred1_[idx], comp_pred2_[idx]);
+          return false;
+        }
+      }
+    }
+    return true;
+  }
+
+  libaom_test::ACMRandom rnd_;
+  uint8_t *comp_pred1_;
+  uint8_t *comp_pred2_;
+  uint8_t *pred_;
+  uint8_t *ref_buffer_;
+  uint8_t *ref_;
+};
+
+AV1CompMaskVarianceTest::~AV1CompMaskVarianceTest() { ; }
+
+void AV1CompMaskVarianceTest::SetUp() {
+  rnd_.Reset(libaom_test::ACMRandom::DeterministicSeed());
+  av1_init_wedge_masks();
+  comp_pred1_ = (uint8_t *)aom_memalign(16, MAX_SB_SQUARE);
+  comp_pred2_ = (uint8_t *)aom_memalign(16, MAX_SB_SQUARE);
+  pred_ = (uint8_t *)aom_memalign(16, MAX_SB_SQUARE);
+  ref_buffer_ = (uint8_t *)aom_memalign(16, MAX_SB_SQUARE + (8 * MAX_SB_SIZE));
+  ref_ = ref_buffer_ + (8 * MAX_SB_SIZE);
+  for (int i = 0; i < MAX_SB_SQUARE; ++i) {
+    pred_[i] = rnd_.Rand8();
+  }
+  for (int i = 0; i < MAX_SB_SQUARE + (8 * MAX_SB_SIZE); ++i) {
+    ref_buffer_[i] = rnd_.Rand8();
+  }
+}
+
+void AV1CompMaskVarianceTest::TearDown() {
+  aom_free(comp_pred1_);
+  aom_free(comp_pred2_);
+  aom_free(pred_);
+  aom_free(ref_buffer_);
+  libaom_test::ClearSystemState();
+}
+
+void AV1CompMaskVarianceTest::RunCheckOutput(comp_mask_pred_func test_impl,
+                                             BLOCK_SIZE bsize, int inv) {
+  const int w = block_size_wide[bsize];
+  const int h = block_size_high[bsize];
+
+  int wedge_types = (1 << get_wedge_bits_lookup(bsize));
+  for (int wedge_index = 0; wedge_index < wedge_types; ++wedge_index) {
+    const uint8_t *mask = av1_get_contiguous_soft_mask(wedge_index, 1, bsize);
+
+    aom_comp_mask_pred_c(comp_pred1_, pred_, w, h, ref_, MAX_SB_SIZE, mask, w,
+                         inv);
+    test_impl(comp_pred2_, pred_, w, h, ref_, MAX_SB_SIZE, mask, w, inv);
+
+    ASSERT_EQ(CheckResult(w, h), true)
+        << " wedge " << wedge_index << " inv " << inv;
+  }
+}
+
+void AV1CompMaskVarianceTest::RunSpeedTest(comp_mask_pred_func test_impl,
+                                           BLOCK_SIZE bsize) {
+  const int w = block_size_wide[bsize];
+  const int h = block_size_high[bsize];
+
+  int wedge_types = (1 << get_wedge_bits_lookup(bsize));
+  int wedge_index = wedge_types / 2;
+  const uint8_t *mask = av1_get_contiguous_soft_mask(wedge_index, 1, bsize);
+  const int num_loops = 1000000000 / (w + h);
+
+  comp_mask_pred_func funcs[2] = { aom_comp_mask_pred_c, test_impl };
+  double elapsed_time[2] = { 0 };
+  for (int i = 0; i < 2; ++i) {
+    aom_usec_timer timer;
+    aom_usec_timer_start(&timer);
+    comp_mask_pred_func func = funcs[i];
+    for (int j = 0; j < num_loops; ++j) {
+      func(comp_pred1_, pred_, w, h, ref_, MAX_SB_SIZE, mask, w, 0);
+    }
+    aom_usec_timer_mark(&timer);
+    double time = static_cast<double>(aom_usec_timer_elapsed(&timer));
+    elapsed_time[i] = 1000.0 * time / num_loops;
+  }
+  printf("compMask %3dx%-3d: %7.2f/%7.2fns", w, h, elapsed_time[0],
+         elapsed_time[1]);
+  printf("(%3.2f)\n", elapsed_time[0] / elapsed_time[1]);
+}
+
+TEST_P(AV1CompMaskVarianceTest, CheckOutput) {
+  // inv = 0, 1
+  RunCheckOutput(GET_PARAM(0), GET_PARAM(1), 0);
+  RunCheckOutput(GET_PARAM(0), GET_PARAM(1), 1);
+}
+
+TEST_P(AV1CompMaskVarianceTest, DISABLED_Speed) {
+  RunSpeedTest(GET_PARAM(0), GET_PARAM(1));
+}
+
+#if HAVE_SSSE3
+INSTANTIATE_TEST_CASE_P(
+    SSSE3, AV1CompMaskVarianceTest,
+    ::testing::Combine(::testing::Values(&aom_comp_mask_pred_ssse3),
+                       ::testing::ValuesIn(kValidBlockSize)));
+#endif
+
+#if HAVE_AVX2
+INSTANTIATE_TEST_CASE_P(
+    AVX2, AV1CompMaskVarianceTest,
+    ::testing::Combine(::testing::Values(&aom_comp_mask_pred_avx2),
+                       ::testing::ValuesIn(kValidBlockSize)));
+#endif
+
+#ifndef aom_comp_mask_pred
+// can't run this test if aom_comp_mask_pred is defined to aom_comp_mask_pred_c
+class AV1CompMaskUpVarianceTest : public AV1CompMaskVarianceTest {
+ public:
+  ~AV1CompMaskUpVarianceTest();
+
+ protected:
+  void RunCheckOutput(comp_mask_pred_func test_impl, BLOCK_SIZE bsize, int inv);
+  void RunSpeedTest(comp_mask_pred_func test_impl, BLOCK_SIZE bsize,
+                    int havSub);
+};
+
+AV1CompMaskUpVarianceTest::~AV1CompMaskUpVarianceTest() { ; }
+
+void AV1CompMaskUpVarianceTest::RunCheckOutput(comp_mask_pred_func test_impl,
+                                               BLOCK_SIZE bsize, int inv) {
+  const int w = block_size_wide[bsize];
+  const int h = block_size_high[bsize];
+  int wedge_types = (1 << get_wedge_bits_lookup(bsize));
+
+  // loop through subx and suby
+  for (int sub = 0; sub < 8 * 8; ++sub) {
+    int subx = sub & 0x7;
+    int suby = (sub >> 3);
+    for (int wedge_index = 0; wedge_index < wedge_types; ++wedge_index) {
+      const uint8_t *mask = av1_get_contiguous_soft_mask(wedge_index, 1, bsize);
+
+      aom_comp_mask_pred = aom_comp_mask_pred_c;  // ref
+      aom_comp_mask_upsampled_pred(NULL, NULL, 0, 0, NULL, comp_pred1_, pred_,
+                                   w, h, subx, suby, ref_, MAX_SB_SIZE, mask, w,
+                                   inv);
+
+      aom_comp_mask_pred = test_impl;  // test
+      aom_comp_mask_upsampled_pred(NULL, NULL, 0, 0, NULL, comp_pred2_, pred_,
+                                   w, h, subx, suby, ref_, MAX_SB_SIZE, mask, w,
+                                   inv);
+      ASSERT_EQ(CheckResult(w, h), true)
+          << " wedge " << wedge_index << " inv " << inv << "sub (" << subx
+          << "," << suby << ")";
+    }
+  }
+}
+
+void AV1CompMaskUpVarianceTest::RunSpeedTest(comp_mask_pred_func test_impl,
+                                             BLOCK_SIZE bsize, int havSub) {
+  const int w = block_size_wide[bsize];
+  const int h = block_size_high[bsize];
+  const int subx = havSub ? 3 : 0;
+  const int suby = havSub ? 4 : 0;
+
+  int wedge_types = (1 << get_wedge_bits_lookup(bsize));
+  int wedge_index = wedge_types / 2;
+  const uint8_t *mask = av1_get_contiguous_soft_mask(wedge_index, 1, bsize);
+
+  const int num_loops = 1000000000 / (w + h);
+  comp_mask_pred_func funcs[2] = { &aom_comp_mask_pred_c, test_impl };
+  double elapsed_time[2] = { 0 };
+  for (int i = 0; i < 2; ++i) {
+    aom_usec_timer timer;
+    aom_usec_timer_start(&timer);
+    aom_comp_mask_pred = funcs[i];
+    for (int j = 0; j < num_loops; ++j) {
+      aom_comp_mask_upsampled_pred(NULL, NULL, 0, 0, NULL, comp_pred1_, pred_,
+                                   w, h, subx, suby, ref_, MAX_SB_SIZE, mask, w,
+                                   0);
+    }
+    aom_usec_timer_mark(&timer);
+    double time = static_cast<double>(aom_usec_timer_elapsed(&timer));
+    elapsed_time[i] = 1000.0 * time / num_loops;
+  }
+  printf("CompMaskUp[%d] %3dx%-3d:%7.2f/%7.2fns", havSub, w, h, elapsed_time[0],
+         elapsed_time[1]);
+  printf("(%3.2f)\n", elapsed_time[0] / elapsed_time[1]);
+}
+
+TEST_P(AV1CompMaskUpVarianceTest, CheckOutput) {
+  // inv mask = 0, 1
+  RunCheckOutput(GET_PARAM(0), GET_PARAM(1), 0);
+  RunCheckOutput(GET_PARAM(0), GET_PARAM(1), 1);
+}
+
+TEST_P(AV1CompMaskUpVarianceTest, DISABLED_Speed) {
+  RunSpeedTest(GET_PARAM(0), GET_PARAM(1), 1);
+}
+
+#if HAVE_SSSE3
+INSTANTIATE_TEST_CASE_P(
+    SSSE3, AV1CompMaskUpVarianceTest,
+    ::testing::Combine(::testing::Values(&aom_comp_mask_pred_ssse3),
+                       ::testing::ValuesIn(kValidBlockSize)));
+#endif
+
+#if HAVE_AVX2
+INSTANTIATE_TEST_CASE_P(
+    AVX2, AV1CompMaskUpVarianceTest,
+    ::testing::Combine(::testing::Values(&aom_comp_mask_pred_avx2),
+                       ::testing::ValuesIn(kValidBlockSize)));
+#endif
+
+#endif  // ifndef aom_comp_mask_pred
+}  // namespace AV1CompMaskVariance

third_party/aom/test/convolve_round_test.cc

@@ -11,7 +11,8 @@
 
 #include <assert.h>
 
-#include "./av1_rtcd.h"
+#include "config/av1_rtcd.h"
+
 #include "aom/aom_integer.h"
 #include "aom_ports/aom_timer.h"
 #include "test/acm_random.h"
@@ -51,7 +52,7 @@ void highbd_convolve_rounding_12(CONVOLVE_ROUNDING_PARAM) {
 
 typedef enum { LOWBITDEPTH_TEST, HIGHBITDEPTH_TEST } DataPathType;
 
-using std::tr1::tuple;
+using ::testing::tuple;
 
 typedef tuple<ConvolveRoundFunc, ConvolveRoundFunc, DataPathType>
     ConvolveRoundParam;
@@ -92,11 +93,9 @@ class ConvolveRoundTest : public ::testing::TestWithParam<ConvolveRoundParam> {
     if (data_path_ == LOWBITDEPTH_TEST) {
       dst = reinterpret_cast<uint8_t *>(dst_);
       dst_ref = reinterpret_cast<uint8_t *>(dst_ref_);
-#if CONFIG_HIGHBITDEPTH
     } else if (data_path_ == HIGHBITDEPTH_TEST) {
       dst = CONVERT_TO_BYTEPTR(dst_);
       dst_ref = CONVERT_TO_BYTEPTR(dst_ref_);
-#endif
     } else {
       assert(0);
     }
@@ -163,10 +162,8 @@ class ConvolveRoundTest : public ::testing::TestWithParam<ConvolveRoundParam> {
 
 TEST_P(ConvolveRoundTest, BitExactCheck) { ConvolveRoundingRun(); }
 
-using std::tr1::make_tuple;
-
+using ::testing::make_tuple;
 #if HAVE_AVX2
-#if CONFIG_HIGHBITDEPTH
 const ConvolveRoundParam kConvRndParamArray[] = {
   make_tuple(&av1_convolve_rounding_c, &av1_convolve_rounding_avx2,
              LOWBITDEPTH_TEST),
@@ -180,11 +177,6 @@ const ConvolveRoundParam kConvRndParamArray[] = {
              &highbd_convolve_rounding_12<av1_highbd_convolve_rounding_avx2>,
              HIGHBITDEPTH_TEST)
 };
-#else
-const ConvolveRoundParam kConvRndParamArray[] = { make_tuple(
-    &av1_convolve_rounding_c, &av1_convolve_rounding_avx2, LOWBITDEPTH_TEST) };
-#endif
-
 INSTANTIATE_TEST_CASE_P(AVX2, ConvolveRoundTest,
                         ::testing::ValuesIn(kConvRndParamArray));
 #endif  // HAVE_AVX2

third_party/aom/test/corner_match_test.cc

@@ -8,11 +8,11 @@
  * Media Patent License 1.0 was not distributed with this source code in the
  * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
  */
+#include "config/av1_rtcd.h"
 
 #include "third_party/googletest/src/googletest/include/gtest/gtest.h"
 #include "test/acm_random.h"
 #include "test/util.h"
-#include "./av1_rtcd.h"
 #include "test/clear_system_state.h"
 #include "test/register_state_check.h"
 
@@ -24,8 +24,8 @@ namespace AV1CornerMatch {
 
 using libaom_test::ACMRandom;
 
-using std::tr1::tuple;
-using std::tr1::make_tuple;
+using ::testing::make_tuple;
+using ::testing::tuple;
 typedef tuple<int> CornerMatchParam;
 
 class AV1CornerMatchTest : public ::testing::TestWithParam<CornerMatchParam> {

third_party/aom/test/cpu_speed_test.cc

@@ -7,7 +7,7 @@
  * obtain it at www.aomedia.org/license/software. If the Alliance for Open
  * Media Patent License 1.0 was not distributed with this source code in the
  * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
+ */
 
 #include "third_party/googletest/src/googletest/include/gtest/gtest.h"
 #include "test/codec_factory.h"

third_party/aom/test/datarate_test.cc

@@ -7,9 +7,10 @@
  * obtain it at www.aomedia.org/license/software. If the Alliance for Open
  * Media Patent License 1.0 was not distributed with this source code in the
  * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
+ */
+
+#include "config/aom_config.h"
 
-#include "./aom_config.h"
 #include "third_party/googletest/src/googletest/include/gtest/gtest.h"
 #include "test/codec_factory.h"
 #include "test/encode_test_driver.h"
@@ -215,6 +216,7 @@ TEST_P(DatarateTestLarge, ChangingDropFrameThresh) {
   cfg_.rc_end_usage = AOM_CBR;
   cfg_.rc_target_bitrate = 200;
   cfg_.g_lag_in_frames = 0;
+  cfg_.g_error_resilient = 1;
   // TODO(marpan): Investigate datarate target failures with a smaller keyframe
   // interval (128).
   cfg_.kf_max_dist = 9999;

third_party/aom/test/dct16x16_test.cc

@@ -1,888 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
-
-#include <math.h>
-#include <stdlib.h>
-#include <string.h>
-
-#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
-
-#include "./av1_rtcd.h"
-#include "./aom_dsp_rtcd.h"
-#include "test/acm_random.h"
-#include "test/clear_system_state.h"
-#include "test/register_state_check.h"
-#include "test/util.h"
-#include "av1/common/entropy.h"
-#include "av1/common/scan.h"
-#include "aom/aom_codec.h"
-#include "aom/aom_integer.h"
-#include "aom_ports/mem.h"
-#include "aom_ports/msvc.h"  // for round()
-
-using libaom_test::ACMRandom;
-
-namespace {
-
-const int kNumCoeffs = 256;
-const double C1 = 0.995184726672197;
-const double C2 = 0.98078528040323;
-const double C3 = 0.956940335732209;
-const double C4 = 0.923879532511287;
-const double C5 = 0.881921264348355;
-const double C6 = 0.831469612302545;
-const double C7 = 0.773010453362737;
-const double C8 = 0.707106781186548;
-const double C9 = 0.634393284163646;
-const double C10 = 0.555570233019602;
-const double C11 = 0.471396736825998;
-const double C12 = 0.38268343236509;
-const double C13 = 0.290284677254462;
-const double C14 = 0.195090322016128;
-const double C15 = 0.098017140329561;
-
-void butterfly_16x16_dct_1d(double input[16], double output[16]) {
-  double step[16];
-  double intermediate[16];
-  double temp1, temp2;
-
-  // step 1
-  step[0] = input[0] + input[15];
-  step[1] = input[1] + input[14];
-  step[2] = input[2] + input[13];
-  step[3] = input[3] + input[12];
-  step[4] = input[4] + input[11];
-  step[5] = input[5] + input[10];
-  step[6] = input[6] + input[9];
-  step[7] = input[7] + input[8];
-  step[8] = input[7] - input[8];
-  step[9] = input[6] - input[9];
-  step[10] = input[5] - input[10];
-  step[11] = input[4] - input[11];
-  step[12] = input[3] - input[12];
-  step[13] = input[2] - input[13];
-  step[14] = input[1] - input[14];
-  step[15] = input[0] - input[15];
-
-  // step 2
-  output[0] = step[0] + step[7];
-  output[1] = step[1] + step[6];
-  output[2] = step[2] + step[5];
-  output[3] = step[3] + step[4];
-  output[4] = step[3] - step[4];
-  output[5] = step[2] - step[5];
-  output[6] = step[1] - step[6];
-  output[7] = step[0] - step[7];
-
-  temp1 = step[8] * C7;
-  temp2 = step[15] * C9;
-  output[8] = temp1 + temp2;
-
-  temp1 = step[9] * C11;
-  temp2 = step[14] * C5;
-  output[9] = temp1 - temp2;
-
-  temp1 = step[10] * C3;
-  temp2 = step[13] * C13;
-  output[10] = temp1 + temp2;
-
-  temp1 = step[11] * C15;
-  temp2 = step[12] * C1;
-  output[11] = temp1 - temp2;
-
-  temp1 = step[11] * C1;
-  temp2 = step[12] * C15;
-  output[12] = temp2 + temp1;
-
-  temp1 = step[10] * C13;
-  temp2 = step[13] * C3;
-  output[13] = temp2 - temp1;
-
-  temp1 = step[9] * C5;
-  temp2 = step[14] * C11;
-  output[14] = temp2 + temp1;
-
-  temp1 = step[8] * C9;
-  temp2 = step[15] * C7;
-  output[15] = temp2 - temp1;
-
-  // step 3
-  step[0] = output[0] + output[3];
-  step[1] = output[1] + output[2];
-  step[2] = output[1] - output[2];
-  step[3] = output[0] - output[3];
-
-  temp1 = output[4] * C14;
-  temp2 = output[7] * C2;
-  step[4] = temp1 + temp2;
-
-  temp1 = output[5] * C10;
-  temp2 = output[6] * C6;
-  step[5] = temp1 + temp2;
-
-  temp1 = output[5] * C6;
-  temp2 = output[6] * C10;
-  step[6] = temp2 - temp1;
-
-  temp1 = output[4] * C2;
-  temp2 = output[7] * C14;
-  step[7] = temp2 - temp1;
-
-  step[8] = output[8] + output[11];
-  step[9] = output[9] + output[10];
-  step[10] = output[9] - output[10];
-  step[11] = output[8] - output[11];
-
-  step[12] = output[12] + output[15];
-  step[13] = output[13] + output[14];
-  step[14] = output[13] - output[14];
-  step[15] = output[12] - output[15];
-
-  // step 4
-  output[0] = (step[0] + step[1]);
-  output[8] = (step[0] - step[1]);
-
-  temp1 = step[2] * C12;
-  temp2 = step[3] * C4;
-  temp1 = temp1 + temp2;
-  output[4] = 2 * (temp1 * C8);
-
-  temp1 = step[2] * C4;
-  temp2 = step[3] * C12;
-  temp1 = temp2 - temp1;
-  output[12] = 2 * (temp1 * C8);
-
-  output[2] = 2 * ((step[4] + step[5]) * C8);
-  output[14] = 2 * ((step[7] - step[6]) * C8);
-
-  temp1 = step[4] - step[5];
-  temp2 = step[6] + step[7];
-  output[6] = (temp1 + temp2);
-  output[10] = (temp1 - temp2);
-
-  intermediate[8] = step[8] + step[14];
-  intermediate[9] = step[9] + step[15];
-
-  temp1 = intermediate[8] * C12;
-  temp2 = intermediate[9] * C4;
-  temp1 = temp1 - temp2;
-  output[3] = 2 * (temp1 * C8);
-
-  temp1 = intermediate[8] * C4;
-  temp2 = intermediate[9] * C12;
-  temp1 = temp2 + temp1;
-  output[13] = 2 * (temp1 * C8);
-
-  output[9] = 2 * ((step[10] + step[11]) * C8);
-
-  intermediate[11] = step[10] - step[11];
-  intermediate[12] = step[12] + step[13];
-  intermediate[13] = step[12] - step[13];
-  intermediate[14] = step[8] - step[14];
-  intermediate[15] = step[9] - step[15];
-
-  output[15] = (intermediate[11] + intermediate[12]);
-  output[1] = -(intermediate[11] - intermediate[12]);
-
-  output[7] = 2 * (intermediate[13] * C8);
-
-  temp1 = intermediate[14] * C12;
-  temp2 = intermediate[15] * C4;
-  temp1 = temp1 - temp2;
-  output[11] = -2 * (temp1 * C8);
-
-  temp1 = intermediate[14] * C4;
-  temp2 = intermediate[15] * C12;
-  temp1 = temp2 + temp1;
-  output[5] = 2 * (temp1 * C8);
-}
-
-void reference_16x16_dct_2d(int16_t input[256], double output[256]) {
-  // First transform columns
-  for (int i = 0; i < 16; ++i) {
-    double temp_in[16], temp_out[16];
-    for (int j = 0; j < 16; ++j) temp_in[j] = input[j * 16 + i];
-    butterfly_16x16_dct_1d(temp_in, temp_out);
-    for (int j = 0; j < 16; ++j) output[j * 16 + i] = temp_out[j];
-  }
-  // Then transform rows
-  for (int i = 0; i < 16; ++i) {
-    double temp_in[16], temp_out[16];
-    for (int j = 0; j < 16; ++j) temp_in[j] = output[j + i * 16];
-    butterfly_16x16_dct_1d(temp_in, temp_out);
-    // Scale by some magic number
-    for (int j = 0; j < 16; ++j) output[j + i * 16] = temp_out[j] / 2;
-  }
-}
-
-typedef void (*FdctFunc)(const int16_t *in, tran_low_t *out, int stride);
-typedef void (*IdctFunc)(const tran_low_t *in, uint8_t *out, int stride);
-typedef void (*FhtFunc)(const int16_t *in, tran_low_t *out, int stride,
-                        TxfmParam *txfm_param);
-typedef void (*IhtFunc)(const tran_low_t *in, uint8_t *out, int stride,
-                        const TxfmParam *txfm_param);
-
-typedef std::tr1::tuple<FdctFunc, IdctFunc, TX_TYPE, aom_bit_depth_t>
-    Dct16x16Param;
-typedef std::tr1::tuple<FhtFunc, IhtFunc, TX_TYPE, aom_bit_depth_t>
-    Ht16x16Param;
-typedef std::tr1::tuple<IdctFunc, IdctFunc, TX_TYPE, aom_bit_depth_t>
-    Idct16x16Param;
-
-void fdct16x16_ref(const int16_t *in, tran_low_t *out, int stride,
-                   TxfmParam * /*txfm_param*/) {
-  aom_fdct16x16_c(in, out, stride);
-}
-
-void idct16x16_ref(const tran_low_t *in, uint8_t *dest, int stride,
-                   const TxfmParam * /*txfm_param*/) {
-  aom_idct16x16_256_add_c(in, dest, stride);
-}
-
-void fht16x16_ref(const int16_t *in, tran_low_t *out, int stride,
-                  TxfmParam *txfm_param) {
-  av1_fht16x16_c(in, out, stride, txfm_param);
-}
-
-void iht16x16_ref(const tran_low_t *in, uint8_t *dest, int stride,
-                  const TxfmParam *txfm_param) {
-  av1_iht16x16_256_add_c(in, dest, stride, txfm_param);
-}
-
-#if CONFIG_HIGHBITDEPTH
-void fht16x16_10(const int16_t *in, tran_low_t *out, int stride,
-                 TxfmParam *txfm_param) {
-  av1_fwd_txfm2d_16x16_c(in, out, stride, txfm_param->tx_type, 10);
-}
-
-void fht16x16_12(const int16_t *in, tran_low_t *out, int stride,
-                 TxfmParam *txfm_param) {
-  av1_fwd_txfm2d_16x16_c(in, out, stride, txfm_param->tx_type, 12);
-}
-
-void iht16x16_10(const tran_low_t *in, uint8_t *out, int stride,
-                 const TxfmParam *txfm_param) {
-  av1_inv_txfm2d_add_16x16_c(in, CONVERT_TO_SHORTPTR(out), stride,
-                             txfm_param->tx_type, 10);
-}
-
-void iht16x16_12(const tran_low_t *in, uint8_t *out, int stride,
-                 const TxfmParam *txfm_param) {
-  av1_inv_txfm2d_add_16x16_c(in, CONVERT_TO_SHORTPTR(out), stride,
-                             txfm_param->tx_type, 12);
-}
-#endif  // CONFIG_HIGHBITDEPTH
-
-class Trans16x16TestBase {
- public:
-  virtual ~Trans16x16TestBase() {}
-
- protected:
-  virtual void RunFwdTxfm(int16_t *in, tran_low_t *out, int stride) = 0;
-
-  virtual void RunInvTxfm(tran_low_t *out, uint8_t *dst, int stride) = 0;
-
-  void RunAccuracyCheck() {
-    ACMRandom rnd(ACMRandom::DeterministicSeed());
-    uint32_t max_error = 0;
-    int64_t total_error = 0;
-    const int count_test_block = 10000;
-    for (int i = 0; i < count_test_block; ++i) {
-      DECLARE_ALIGNED(16, int16_t, test_input_block[kNumCoeffs]);
-      DECLARE_ALIGNED(16, tran_low_t, test_temp_block[kNumCoeffs]);
-      DECLARE_ALIGNED(16, uint8_t, dst[kNumCoeffs]);
-      DECLARE_ALIGNED(16, uint8_t, src[kNumCoeffs]);
-#if CONFIG_HIGHBITDEPTH
-      DECLARE_ALIGNED(16, uint16_t, dst16[kNumCoeffs]);
-      DECLARE_ALIGNED(16, uint16_t, src16[kNumCoeffs]);
-#endif
-
-      // Initialize a test block with input range [-mask_, mask_].
-      for (int j = 0; j < kNumCoeffs; ++j) {
-        if (bit_depth_ == AOM_BITS_8) {
-          src[j] = rnd.Rand8();
-          dst[j] = rnd.Rand8();
-          test_input_block[j] = src[j] - dst[j];
-#if CONFIG_HIGHBITDEPTH
-        } else {
-          src16[j] = rnd.Rand16() & mask_;
-          dst16[j] = rnd.Rand16() & mask_;
-          test_input_block[j] = src16[j] - dst16[j];
-#endif
-        }
-      }
-
-      ASM_REGISTER_STATE_CHECK(
-          RunFwdTxfm(test_input_block, test_temp_block, pitch_));
-      if (bit_depth_ == AOM_BITS_8) {
-        ASM_REGISTER_STATE_CHECK(RunInvTxfm(test_temp_block, dst, pitch_));
-#if CONFIG_HIGHBITDEPTH
-      } else {
-        ASM_REGISTER_STATE_CHECK(
-            RunInvTxfm(test_temp_block, CONVERT_TO_BYTEPTR(dst16), pitch_));
-#endif
-      }
-
-      for (int j = 0; j < kNumCoeffs; ++j) {
-#if CONFIG_HIGHBITDEPTH
-        const int32_t diff =
-            bit_depth_ == AOM_BITS_8 ? dst[j] - src[j] : dst16[j] - src16[j];
-#else
-        const int32_t diff = dst[j] - src[j];
-#endif
-        const uint32_t error = diff * diff;
-        if (max_error < error) max_error = error;
-        total_error += error;
-      }
-    }
-
-    EXPECT_GE(1u << 2 * (bit_depth_ - 8), max_error)
-        << "Error: 16x16 FHT/IHT has an individual round trip error > 1";
-
-    EXPECT_GE(count_test_block << 2 * (bit_depth_ - 8), total_error)
-        << "Error: 16x16 FHT/IHT has average round trip error > 1 per block";
-  }
-
-  void RunCoeffCheck() {
-    ACMRandom rnd(ACMRandom::DeterministicSeed());
-    const int count_test_block = 1000;
-    DECLARE_ALIGNED(16, int16_t, input_block[kNumCoeffs]);
-    DECLARE_ALIGNED(16, tran_low_t, output_ref_block[kNumCoeffs]);
-    DECLARE_ALIGNED(16, tran_low_t, output_block[kNumCoeffs]);
-
-    for (int i = 0; i < count_test_block; ++i) {
-      // Initialize a test block with input range [-mask_, mask_].
-      for (int j = 0; j < kNumCoeffs; ++j)
-        input_block[j] = (rnd.Rand16() & mask_) - (rnd.Rand16() & mask_);
-
-      fwd_txfm_ref(input_block, output_ref_block, pitch_, &txfm_param_);
-      ASM_REGISTER_STATE_CHECK(RunFwdTxfm(input_block, output_block, pitch_));
-
-      // The minimum quant value is 4.
-      for (int j = 0; j < kNumCoeffs; ++j)
-        EXPECT_EQ(output_block[j], output_ref_block[j]);
-    }
-  }
-
-  void RunMemCheck() {
-    ACMRandom rnd(ACMRandom::DeterministicSeed());
-    const int count_test_block = 1000;
-    DECLARE_ALIGNED(16, int16_t, input_extreme_block[kNumCoeffs]);
-    DECLARE_ALIGNED(16, tran_low_t, output_ref_block[kNumCoeffs]);
-    DECLARE_ALIGNED(16, tran_low_t, output_block[kNumCoeffs]);
-
-    for (int i = 0; i < count_test_block; ++i) {
-      // Initialize a test block with input range [-mask_, mask_].
-      for (int j = 0; j < kNumCoeffs; ++j) {
-        input_extreme_block[j] = rnd.Rand8() % 2 ? mask_ : -mask_;
-      }
-      if (i == 0) {
-        for (int j = 0; j < kNumCoeffs; ++j) input_extreme_block[j] = mask_;
-      } else if (i == 1) {
-        for (int j = 0; j < kNumCoeffs; ++j) input_extreme_block[j] = -mask_;
-      }
-
-      fwd_txfm_ref(input_extreme_block, output_ref_block, pitch_, &txfm_param_);
-      ASM_REGISTER_STATE_CHECK(
-          RunFwdTxfm(input_extreme_block, output_block, pitch_));
-
-      // The minimum quant value is 4.
-      for (int j = 0; j < kNumCoeffs; ++j) {
-        EXPECT_EQ(output_block[j], output_ref_block[j]);
-        EXPECT_GE(4 * DCT_MAX_VALUE << (bit_depth_ - 8), abs(output_block[j]))
-            << "Error: 16x16 FDCT has coefficient larger than 4*DCT_MAX_VALUE";
-      }
-    }
-  }
-
-  void RunQuantCheck(int dc_thred, int ac_thred) {
-    ACMRandom rnd(ACMRandom::DeterministicSeed());
-    const int count_test_block = 100000;
-    DECLARE_ALIGNED(16, int16_t, input_extreme_block[kNumCoeffs]);
-    DECLARE_ALIGNED(16, tran_low_t, output_ref_block[kNumCoeffs]);
-
-    DECLARE_ALIGNED(16, uint8_t, dst[kNumCoeffs]);
-    DECLARE_ALIGNED(16, uint8_t, ref[kNumCoeffs]);
-#if CONFIG_HIGHBITDEPTH
-    DECLARE_ALIGNED(16, uint16_t, dst16[kNumCoeffs]);
-    DECLARE_ALIGNED(16, uint16_t, ref16[kNumCoeffs]);
-#endif
-
-    for (int i = 0; i < count_test_block; ++i) {
-      // Initialize a test block with input range [-mask_, mask_].
-      for (int j = 0; j < kNumCoeffs; ++j) {
-        input_extreme_block[j] = rnd.Rand8() % 2 ? mask_ : -mask_;
-      }
-      if (i == 0)
-        for (int j = 0; j < kNumCoeffs; ++j) input_extreme_block[j] = mask_;
-      if (i == 1)
-        for (int j = 0; j < kNumCoeffs; ++j) input_extreme_block[j] = -mask_;
-
-      fwd_txfm_ref(input_extreme_block, output_ref_block, pitch_, &txfm_param_);
-
-      // clear reconstructed pixel buffers
-      memset(dst, 0, kNumCoeffs * sizeof(uint8_t));
-      memset(ref, 0, kNumCoeffs * sizeof(uint8_t));
-#if CONFIG_HIGHBITDEPTH
-      memset(dst16, 0, kNumCoeffs * sizeof(uint16_t));
-      memset(ref16, 0, kNumCoeffs * sizeof(uint16_t));
-#endif
-
-      // quantization with maximum allowed step sizes
-      output_ref_block[0] = (output_ref_block[0] / dc_thred) * dc_thred;
-      for (int j = 1; j < kNumCoeffs; ++j)
-        output_ref_block[j] = (output_ref_block[j] / ac_thred) * ac_thred;
-      if (bit_depth_ == AOM_BITS_8) {
-        inv_txfm_ref(output_ref_block, ref, pitch_, &txfm_param_);
-        ASM_REGISTER_STATE_CHECK(RunInvTxfm(output_ref_block, dst, pitch_));
-#if CONFIG_HIGHBITDEPTH
-      } else {
-        inv_txfm_ref(output_ref_block, CONVERT_TO_BYTEPTR(ref16), pitch_,
-                     &txfm_param_);
-        ASM_REGISTER_STATE_CHECK(
-            RunInvTxfm(output_ref_block, CONVERT_TO_BYTEPTR(dst16), pitch_));
-#endif
-      }
-      if (bit_depth_ == AOM_BITS_8) {
-        for (int j = 0; j < kNumCoeffs; ++j) EXPECT_EQ(ref[j], dst[j]);
-#if CONFIG_HIGHBITDEPTH
-      } else {
-        for (int j = 0; j < kNumCoeffs; ++j) EXPECT_EQ(ref16[j], dst16[j]);
-#endif
-      }
-    }
-  }
-
-  void RunInvAccuracyCheck() {
-    ACMRandom rnd(ACMRandom::DeterministicSeed());
-    const int count_test_block = 1000;
-    DECLARE_ALIGNED(16, int16_t, in[kNumCoeffs]);
-    DECLARE_ALIGNED(16, tran_low_t, coeff[kNumCoeffs]);
-    DECLARE_ALIGNED(16, uint8_t, dst[kNumCoeffs]);
-    DECLARE_ALIGNED(16, uint8_t, src[kNumCoeffs]);
-#if CONFIG_HIGHBITDEPTH
-    DECLARE_ALIGNED(16, uint16_t, dst16[kNumCoeffs]);
-    DECLARE_ALIGNED(16, uint16_t, src16[kNumCoeffs]);
-#endif  // CONFIG_HIGHBITDEPTH
-
-    for (int i = 0; i < count_test_block; ++i) {
-      double out_r[kNumCoeffs];
-
-      // Initialize a test block with input range [-255, 255].
-      for (int j = 0; j < kNumCoeffs; ++j) {
-        if (bit_depth_ == AOM_BITS_8) {
-          src[j] = rnd.Rand8();
-          dst[j] = rnd.Rand8();
-          in[j] = src[j] - dst[j];
-#if CONFIG_HIGHBITDEPTH
-        } else {
-          src16[j] = rnd.Rand16() & mask_;
-          dst16[j] = rnd.Rand16() & mask_;
-          in[j] = src16[j] - dst16[j];
-#endif  // CONFIG_HIGHBITDEPTH
-        }
-      }
-
-      reference_16x16_dct_2d(in, out_r);
-      for (int j = 0; j < kNumCoeffs; ++j)
-        coeff[j] = static_cast<tran_low_t>(round(out_r[j]));
-
-      if (bit_depth_ == AOM_BITS_8) {
-        ASM_REGISTER_STATE_CHECK(RunInvTxfm(coeff, dst, 16));
-#if CONFIG_HIGHBITDEPTH
-      } else {
-        ASM_REGISTER_STATE_CHECK(
-            RunInvTxfm(coeff, CONVERT_TO_BYTEPTR(dst16), 16));
-#endif  // CONFIG_HIGHBITDEPTH
-      }
-
-      for (int j = 0; j < kNumCoeffs; ++j) {
-#if CONFIG_HIGHBITDEPTH
-        const int diff =
-            bit_depth_ == AOM_BITS_8 ? dst[j] - src[j] : dst16[j] - src16[j];
-#else
-        const int diff = dst[j] - src[j];
-#endif  // CONFIG_HIGHBITDEPTH
-        const uint32_t error = diff * diff;
-        EXPECT_GE(1u, error)
-            << "Error: 16x16 IDCT has error " << error << " at index " << j;
-      }
-    }
-  }
-
-  void CompareInvReference(IdctFunc ref_txfm, int thresh) {
-    ACMRandom rnd(ACMRandom::DeterministicSeed());
-    const int count_test_block = 10000;
-    const int eob = 10;
-    const int16_t *scan = av1_default_scan_orders[TX_16X16].scan;
-    DECLARE_ALIGNED(16, tran_low_t, coeff[kNumCoeffs]);
-    DECLARE_ALIGNED(16, uint8_t, dst[kNumCoeffs]);
-    DECLARE_ALIGNED(16, uint8_t, ref[kNumCoeffs]);
-#if CONFIG_HIGHBITDEPTH
-    DECLARE_ALIGNED(16, uint16_t, dst16[kNumCoeffs]);
-    DECLARE_ALIGNED(16, uint16_t, ref16[kNumCoeffs]);
-#endif  // CONFIG_HIGHBITDEPTH
-
-    for (int i = 0; i < count_test_block; ++i) {
-      for (int j = 0; j < kNumCoeffs; ++j) {
-        if (j < eob) {
-          // Random values less than the threshold, either positive or negative
-          coeff[scan[j]] = rnd(thresh) * (1 - 2 * (i % 2));
-        } else {
-          coeff[scan[j]] = 0;
-        }
-        if (bit_depth_ == AOM_BITS_8) {
-          dst[j] = 0;
-          ref[j] = 0;
-#if CONFIG_HIGHBITDEPTH
-        } else {
-          dst16[j] = 0;
-          ref16[j] = 0;
-#endif  // CONFIG_HIGHBITDEPTH
-        }
-      }
-      if (bit_depth_ == AOM_BITS_8) {
-        ref_txfm(coeff, ref, pitch_);
-        ASM_REGISTER_STATE_CHECK(RunInvTxfm(coeff, dst, pitch_));
-      } else {
-#if CONFIG_HIGHBITDEPTH
-        ref_txfm(coeff, CONVERT_TO_BYTEPTR(ref16), pitch_);
-        ASM_REGISTER_STATE_CHECK(
-            RunInvTxfm(coeff, CONVERT_TO_BYTEPTR(dst16), pitch_));
-#endif  // CONFIG_HIGHBITDEPTH
-      }
-
-      for (int j = 0; j < kNumCoeffs; ++j) {
-#if CONFIG_HIGHBITDEPTH
-        const int diff =
-            bit_depth_ == AOM_BITS_8 ? dst[j] - ref[j] : dst16[j] - ref16[j];
-#else
-        const int diff = dst[j] - ref[j];
-#endif  // CONFIG_HIGHBITDEPTH
-        const uint32_t error = diff * diff;
-        EXPECT_EQ(0u, error) << "Error: 16x16 IDCT Comparison has error "
-                             << error << " at index " << j;
-      }
-    }
-  }
-
-  int pitch_;
-  aom_bit_depth_t bit_depth_;
-  int mask_;
-  FhtFunc fwd_txfm_ref;
-  IhtFunc inv_txfm_ref;
-  TxfmParam txfm_param_;
-};
-
-class Trans16x16DCT : public Trans16x16TestBase,
-                      public ::testing::TestWithParam<Dct16x16Param> {
- public:
-  virtual ~Trans16x16DCT() {}
-
-  virtual void SetUp() {
-    fwd_txfm_ = GET_PARAM(0);
-    inv_txfm_ = GET_PARAM(1);
-    bit_depth_ = GET_PARAM(3);
-    pitch_ = 16;
-    fwd_txfm_ref = fdct16x16_ref;
-    inv_txfm_ref = idct16x16_ref;
-    mask_ = (1 << bit_depth_) - 1;
-    inv_txfm_ref = idct16x16_ref;
-    txfm_param_.tx_type = GET_PARAM(2);
-  }
-  virtual void TearDown() { libaom_test::ClearSystemState(); }
-
- protected:
-  void RunFwdTxfm(int16_t *in, tran_low_t *out, int stride) {
-    fwd_txfm_(in, out, stride);
-  }
-  void RunInvTxfm(tran_low_t *out, uint8_t *dst, int stride) {
-    inv_txfm_(out, dst, stride);
-  }
-
-  FdctFunc fwd_txfm_;
-  IdctFunc inv_txfm_;
-};
-
-TEST_P(Trans16x16DCT, AccuracyCheck) { RunAccuracyCheck(); }
-
-TEST_P(Trans16x16DCT, CoeffCheck) { RunCoeffCheck(); }
-
-TEST_P(Trans16x16DCT, MemCheck) { RunMemCheck(); }
-
-TEST_P(Trans16x16DCT, QuantCheck) {
-  // Use maximally allowed quantization step sizes for DC and AC
-  // coefficients respectively.
-  RunQuantCheck(1336, 1828);
-}
-
-TEST_P(Trans16x16DCT, InvAccuracyCheck) { RunInvAccuracyCheck(); }
-
-class Trans16x16HT : public Trans16x16TestBase,
-                     public ::testing::TestWithParam<Ht16x16Param> {
- public:
-  virtual ~Trans16x16HT() {}
-
-  virtual void SetUp() {
-    fwd_txfm_ = GET_PARAM(0);
-    inv_txfm_ = GET_PARAM(1);
-    bit_depth_ = GET_PARAM(3);
-    pitch_ = 16;
-    mask_ = (1 << bit_depth_) - 1;
-    txfm_param_.tx_type = GET_PARAM(2);
-#if CONFIG_HIGHBITDEPTH
-    switch (bit_depth_) {
-      case AOM_BITS_10:
-        fwd_txfm_ref = fht16x16_10;
-        inv_txfm_ref = iht16x16_10;
-        break;
-      case AOM_BITS_12:
-        fwd_txfm_ref = fht16x16_12;
-        inv_txfm_ref = iht16x16_12;
-        break;
-      default:
-        fwd_txfm_ref = fht16x16_ref;
-        inv_txfm_ref = iht16x16_ref;
-        break;
-    }
-#else
-    fwd_txfm_ref = fht16x16_ref;
-    inv_txfm_ref = iht16x16_ref;
-#endif
-  }
-  virtual void TearDown() { libaom_test::ClearSystemState(); }
-
- protected:
-  void RunFwdTxfm(int16_t *in, tran_low_t *out, int stride) {
-    fwd_txfm_(in, out, stride, &txfm_param_);
-  }
-  void RunInvTxfm(tran_low_t *out, uint8_t *dst, int stride) {
-    inv_txfm_(out, dst, stride, &txfm_param_);
-  }
-
-  FhtFunc fwd_txfm_;
-  IhtFunc inv_txfm_;
-};
-
-TEST_P(Trans16x16HT, AccuracyCheck) { RunAccuracyCheck(); }
-
-TEST_P(Trans16x16HT, CoeffCheck) { RunCoeffCheck(); }
-
-TEST_P(Trans16x16HT, MemCheck) { RunMemCheck(); }
-
-TEST_P(Trans16x16HT, QuantCheck) {
-  // The encoder skips any non-DC intra prediction modes,
-  // when the quantization step size goes beyond 988.
-  RunQuantCheck(429, 729);
-}
-
-class InvTrans16x16DCT : public Trans16x16TestBase,
-                         public ::testing::TestWithParam<Idct16x16Param> {
- public:
-  virtual ~InvTrans16x16DCT() {}
-
-  virtual void SetUp() {
-    ref_txfm_ = GET_PARAM(0);
-    inv_txfm_ = GET_PARAM(1);
-    thresh_ = GET_PARAM(2);
-    bit_depth_ = GET_PARAM(3);
-    pitch_ = 16;
-    mask_ = (1 << bit_depth_) - 1;
-  }
-  virtual void TearDown() { libaom_test::ClearSystemState(); }
-
- protected:
-  void RunFwdTxfm(int16_t * /*in*/, tran_low_t * /*out*/, int /*stride*/) {}
-  void RunInvTxfm(tran_low_t *out, uint8_t *dst, int stride) {
-    inv_txfm_(out, dst, stride);
-  }
-
-  IdctFunc ref_txfm_;
-  IdctFunc inv_txfm_;
-  int thresh_;
-};
-
-TEST_P(InvTrans16x16DCT, CompareReference) {
-  CompareInvReference(ref_txfm_, thresh_);
-}
-
-class PartialTrans16x16Test : public ::testing::TestWithParam<
-                                  std::tr1::tuple<FdctFunc, aom_bit_depth_t> > {
- public:
-  virtual ~PartialTrans16x16Test() {}
-  virtual void SetUp() {
-    fwd_txfm_ = GET_PARAM(0);
-    bit_depth_ = GET_PARAM(1);
-  }
-
-  virtual void TearDown() { libaom_test::ClearSystemState(); }
-
- protected:
-  aom_bit_depth_t bit_depth_;
-  FdctFunc fwd_txfm_;
-};
-
-TEST_P(PartialTrans16x16Test, Extremes) {
-#if CONFIG_HIGHBITDEPTH
-  const int16_t maxval =
-      static_cast<int16_t>(clip_pixel_highbd(1 << 30, bit_depth_));
-#else
-  const int16_t maxval = 255;
-#endif
-  const int minval = -maxval;
-  DECLARE_ALIGNED(16, int16_t, input[kNumCoeffs]);
-  DECLARE_ALIGNED(16, tran_low_t, output[kNumCoeffs]);
-
-  for (int i = 0; i < kNumCoeffs; ++i) input[i] = maxval;
-  output[0] = 0;
-  ASM_REGISTER_STATE_CHECK(fwd_txfm_(input, output, 16));
-  EXPECT_EQ((maxval * kNumCoeffs) >> 1, output[0]);
-
-  for (int i = 0; i < kNumCoeffs; ++i) input[i] = minval;
-  output[0] = 0;
-  ASM_REGISTER_STATE_CHECK(fwd_txfm_(input, output, 16));
-  EXPECT_EQ((minval * kNumCoeffs) >> 1, output[0]);
-}
-
-TEST_P(PartialTrans16x16Test, Random) {
-#if CONFIG_HIGHBITDEPTH
-  const int16_t maxval =
-      static_cast<int16_t>(clip_pixel_highbd(1 << 30, bit_depth_));
-#else
-  const int16_t maxval = 255;
-#endif
-  DECLARE_ALIGNED(16, int16_t, input[kNumCoeffs]);
-  DECLARE_ALIGNED(16, tran_low_t, output[kNumCoeffs]);
-  ACMRandom rnd(ACMRandom::DeterministicSeed());
-
-  int sum = 0;
-  for (int i = 0; i < kNumCoeffs; ++i) {
-    const int val = (i & 1) ? -rnd(maxval + 1) : rnd(maxval + 1);
-    input[i] = val;
-    sum += val;
-  }
-  output[0] = 0;
-  ASM_REGISTER_STATE_CHECK(fwd_txfm_(input, output, 16));
-  EXPECT_EQ(sum >> 1, output[0]);
-}
-
-using std::tr1::make_tuple;
-
-#if CONFIG_HIGHBITDEPTH
-INSTANTIATE_TEST_CASE_P(C, Trans16x16DCT,
-                        ::testing::Values(make_tuple(&aom_fdct16x16_c,
-                                                     &aom_idct16x16_256_add_c,
-                                                     DCT_DCT, AOM_BITS_8)));
-#else
-INSTANTIATE_TEST_CASE_P(C, Trans16x16DCT,
-                        ::testing::Values(make_tuple(&aom_fdct16x16_c,
-                                                     &aom_idct16x16_256_add_c,
-                                                     DCT_DCT, AOM_BITS_8)));
-#endif  // CONFIG_HIGHBITDEPTH
-
-#if CONFIG_HIGHBITDEPTH
-INSTANTIATE_TEST_CASE_P(
-    C, Trans16x16HT,
-    ::testing::Values(
-        make_tuple(&fht16x16_10, &iht16x16_10, DCT_DCT, AOM_BITS_10),
-        make_tuple(&fht16x16_10, &iht16x16_10, ADST_DCT, AOM_BITS_10),
-        make_tuple(&fht16x16_10, &iht16x16_10, DCT_ADST, AOM_BITS_10),
-        make_tuple(&fht16x16_10, &iht16x16_10, ADST_ADST, AOM_BITS_10),
-        make_tuple(&fht16x16_12, &iht16x16_12, DCT_DCT, AOM_BITS_12),
-        make_tuple(&fht16x16_12, &iht16x16_12, ADST_DCT, AOM_BITS_12),
-        make_tuple(&fht16x16_12, &iht16x16_12, DCT_ADST, AOM_BITS_12),
-        make_tuple(&fht16x16_12, &iht16x16_12, ADST_ADST, AOM_BITS_12),
-        make_tuple(&av1_fht16x16_c, &av1_iht16x16_256_add_c, DCT_DCT,
-                   AOM_BITS_8),
-        make_tuple(&av1_fht16x16_c, &av1_iht16x16_256_add_c, ADST_DCT,
-                   AOM_BITS_8),
-        make_tuple(&av1_fht16x16_c, &av1_iht16x16_256_add_c, DCT_ADST,
-                   AOM_BITS_8),
-        make_tuple(&av1_fht16x16_c, &av1_iht16x16_256_add_c, ADST_ADST,
-                   AOM_BITS_8)));
-#else
-INSTANTIATE_TEST_CASE_P(
-    C, Trans16x16HT,
-    ::testing::Values(make_tuple(&av1_fht16x16_c, &av1_iht16x16_256_add_c,
-                                 DCT_DCT, AOM_BITS_8),
-                      make_tuple(&av1_fht16x16_c, &av1_iht16x16_256_add_c,
-                                 ADST_DCT, AOM_BITS_8),
-                      make_tuple(&av1_fht16x16_c, &av1_iht16x16_256_add_c,
-                                 DCT_ADST, AOM_BITS_8),
-                      make_tuple(&av1_fht16x16_c, &av1_iht16x16_256_add_c,
-                                 ADST_ADST, AOM_BITS_8)));
-#endif  // CONFIG_HIGHBITDEPTH
-
-#if HAVE_NEON_ASM && !CONFIG_HIGHBITDEPTH
-INSTANTIATE_TEST_CASE_P(
-    NEON, Trans16x16DCT,
-    ::testing::Values(make_tuple(&aom_fdct16x16_c, &aom_idct16x16_256_add_neon,
-                                 DCT_DCT, AOM_BITS_8)));
-#endif
-
-#if HAVE_SSE2 && !CONFIG_HIGHBITDEPTH
-INSTANTIATE_TEST_CASE_P(SSE2, Trans16x16DCT,
-                        ::testing::Values(make_tuple(
-                            &aom_fdct16x16_sse2, &aom_idct16x16_256_add_sse2,
-                            DCT_DCT, AOM_BITS_8)));
-#if !CONFIG_DAALA_DCT16
-INSTANTIATE_TEST_CASE_P(
-    SSE2, Trans16x16HT,
-    ::testing::Values(make_tuple(&av1_fht16x16_sse2, &av1_iht16x16_256_add_sse2,
-                                 DCT_DCT, AOM_BITS_8),
-                      make_tuple(&av1_fht16x16_sse2, &av1_iht16x16_256_add_sse2,
-                                 ADST_DCT, AOM_BITS_8),
-                      make_tuple(&av1_fht16x16_sse2, &av1_iht16x16_256_add_sse2,
-                                 DCT_ADST, AOM_BITS_8),
-                      make_tuple(&av1_fht16x16_sse2, &av1_iht16x16_256_add_sse2,
-                                 ADST_ADST, AOM_BITS_8)));
-#endif  // CONFIG_DAALA_DCT16
-#endif  // HAVE_SSE2 && !CONFIG_HIGHBITDEPTH
-
-#if HAVE_SSE2 && CONFIG_HIGHBITDEPTH
-INSTANTIATE_TEST_CASE_P(SSE2, Trans16x16DCT,
-                        ::testing::Values(make_tuple(&aom_fdct16x16_sse2,
-                                                     &aom_idct16x16_256_add_c,
-                                                     DCT_DCT, AOM_BITS_8)));
-#if !CONFIG_DAALA_DCT16
-INSTANTIATE_TEST_CASE_P(
-    SSE2, Trans16x16HT,
-    ::testing::Values(make_tuple(&av1_fht16x16_sse2, &av1_iht16x16_256_add_c,
-                                 DCT_DCT, AOM_BITS_8),
-                      make_tuple(&av1_fht16x16_sse2, &av1_iht16x16_256_add_c,
-                                 ADST_DCT, AOM_BITS_8),
-                      make_tuple(&av1_fht16x16_sse2, &av1_iht16x16_256_add_c,
-                                 DCT_ADST, AOM_BITS_8),
-                      make_tuple(&av1_fht16x16_sse2, &av1_iht16x16_256_add_c,
-                                 ADST_ADST, AOM_BITS_8)));
-#endif
-#endif  // HAVE_SSE2 && CONFIG_HIGHBITDEPTH
-
-#if HAVE_MSA && !CONFIG_HIGHBITDEPTH
-INSTANTIATE_TEST_CASE_P(MSA, Trans16x16DCT,
-                        ::testing::Values(make_tuple(&aom_fdct16x16_msa,
-                                                     &aom_idct16x16_256_add_msa,
-                                                     DCT_DCT, AOM_BITS_8)));
-#if !CONFIG_EXT_TX && !CONFIG_DAALA_DCT16
-// TODO(yaowu): re-enable this after msa versions are updated to match C.
-INSTANTIATE_TEST_CASE_P(
-    DISABLED_MSA, Trans16x16HT,
-    ::testing::Values(make_tuple(&av1_fht16x16_msa, &av1_iht16x16_256_add_msa,
-                                 DCT_DCT, AOM_BITS_8),
-                      make_tuple(&av1_fht16x16_msa, &av1_iht16x16_256_add_msa,
-                                 ADST_DCT, AOM_BITS_8),
-                      make_tuple(&av1_fht16x16_msa, &av1_iht16x16_256_add_msa,
-                                 DCT_ADST, AOM_BITS_8),
-                      make_tuple(&av1_fht16x16_msa, &av1_iht16x16_256_add_msa,
-                                 ADST_ADST, AOM_BITS_8)));
-#endif  // !CONFIG_EXT_TX && !CONFIG_DAALA_DCT16
-#endif  // HAVE_MSA && !CONFIG_HIGHBITDEPTH
-}  // namespace

third_party/aom/test/dct32x32_test.cc

@@ -1,425 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
-
-#include <math.h>
-#include <stdlib.h>
-#include <string.h>
-
-#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
-
-#include "./av1_rtcd.h"
-#include "./aom_config.h"
-#include "./aom_dsp_rtcd.h"
-#include "test/acm_random.h"
-#include "test/clear_system_state.h"
-#include "test/register_state_check.h"
-#include "test/util.h"
-#include "av1/common/entropy.h"
-#include "aom/aom_codec.h"
-#include "aom/aom_integer.h"
-#include "aom_ports/mem.h"
-#include "aom_ports/msvc.h"  // for round()
-
-using libaom_test::ACMRandom;
-
-namespace {
-
-const int kNumCoeffs = 1024;
-const double kPi = 3.141592653589793238462643383279502884;
-void reference_32x32_dct_1d(const double in[32], double out[32]) {
-  const double kInvSqrt2 = 0.707106781186547524400844362104;
-  for (int k = 0; k < 32; k++) {
-    out[k] = 0.0;
-    for (int n = 0; n < 32; n++)
-      out[k] += in[n] * cos(kPi * (2 * n + 1) * k / 64.0);
-    if (k == 0) out[k] = out[k] * kInvSqrt2;
-  }
-}
-
-void reference_32x32_dct_2d(const int16_t input[kNumCoeffs],
-                            double output[kNumCoeffs]) {
-  // First transform columns
-  for (int i = 0; i < 32; ++i) {
-    double temp_in[32], temp_out[32];
-    for (int j = 0; j < 32; ++j) temp_in[j] = input[j * 32 + i];
-    reference_32x32_dct_1d(temp_in, temp_out);
-    for (int j = 0; j < 32; ++j) output[j * 32 + i] = temp_out[j];
-  }
-  // Then transform rows
-  for (int i = 0; i < 32; ++i) {
-    double temp_in[32], temp_out[32];
-    for (int j = 0; j < 32; ++j) temp_in[j] = output[j + i * 32];
-    reference_32x32_dct_1d(temp_in, temp_out);
-    // Scale by some magic number
-    for (int j = 0; j < 32; ++j) output[j + i * 32] = temp_out[j] / 4;
-  }
-}
-
-typedef void (*FwdTxfmFunc)(const int16_t *in, tran_low_t *out, int stride);
-typedef void (*InvTxfmFunc)(const tran_low_t *in, uint8_t *out, int stride);
-
-typedef std::tr1::tuple<FwdTxfmFunc, InvTxfmFunc, int, aom_bit_depth_t>
-    Trans32x32Param;
-
-class Trans32x32Test : public ::testing::TestWithParam<Trans32x32Param> {
- public:
-  virtual ~Trans32x32Test() {}
-  virtual void SetUp() {
-    fwd_txfm_ = GET_PARAM(0);
-    inv_txfm_ = GET_PARAM(1);
-    version_ = GET_PARAM(2);  // 0: high precision forward transform
-                              // 1: low precision version for rd loop
-    bit_depth_ = GET_PARAM(3);
-    mask_ = (1 << bit_depth_) - 1;
-  }
-
-  virtual void TearDown() { libaom_test::ClearSystemState(); }
-
- protected:
-  int version_;
-  aom_bit_depth_t bit_depth_;
-  int mask_;
-  FwdTxfmFunc fwd_txfm_;
-  InvTxfmFunc inv_txfm_;
-};
-
-TEST_P(Trans32x32Test, AccuracyCheck) {
-  ACMRandom rnd(ACMRandom::DeterministicSeed());
-  uint32_t max_error = 0;
-  int64_t total_error = 0;
-  const int count_test_block = 10000;
-  DECLARE_ALIGNED(16, int16_t, test_input_block[kNumCoeffs]);
-  DECLARE_ALIGNED(16, tran_low_t, test_temp_block[kNumCoeffs]);
-  DECLARE_ALIGNED(16, uint8_t, dst[kNumCoeffs]);
-  DECLARE_ALIGNED(16, uint8_t, src[kNumCoeffs]);
-#if CONFIG_HIGHBITDEPTH
-  DECLARE_ALIGNED(16, uint16_t, dst16[kNumCoeffs]);
-  DECLARE_ALIGNED(16, uint16_t, src16[kNumCoeffs]);
-#endif
-
-  for (int i = 0; i < count_test_block; ++i) {
-    // Initialize a test block with input range [-mask_, mask_].
-    for (int j = 0; j < kNumCoeffs; ++j) {
-      if (bit_depth_ == AOM_BITS_8) {
-        src[j] = rnd.Rand8();
-        dst[j] = rnd.Rand8();
-        test_input_block[j] = src[j] - dst[j];
-#if CONFIG_HIGHBITDEPTH
-      } else {
-        src16[j] = rnd.Rand16() & mask_;
-        dst16[j] = rnd.Rand16() & mask_;
-        test_input_block[j] = src16[j] - dst16[j];
-#endif
-      }
-    }
-
-    ASM_REGISTER_STATE_CHECK(fwd_txfm_(test_input_block, test_temp_block, 32));
-    if (bit_depth_ == AOM_BITS_8) {
-      ASM_REGISTER_STATE_CHECK(inv_txfm_(test_temp_block, dst, 32));
-#if CONFIG_HIGHBITDEPTH
-    } else {
-      ASM_REGISTER_STATE_CHECK(
-          inv_txfm_(test_temp_block, CONVERT_TO_BYTEPTR(dst16), 32));
-#endif
-    }
-
-    for (int j = 0; j < kNumCoeffs; ++j) {
-#if CONFIG_HIGHBITDEPTH
-      const int32_t diff =
-          bit_depth_ == AOM_BITS_8 ? dst[j] - src[j] : dst16[j] - src16[j];
-#else
-      const int32_t diff = dst[j] - src[j];
-#endif
-      const uint32_t error = diff * diff;
-      if (max_error < error) max_error = error;
-      total_error += error;
-    }
-  }
-
-  if (version_ == 1) {
-    max_error /= 2;
-    total_error /= 45;
-  }
-
-  EXPECT_GE(1u << 2 * (bit_depth_ - 8), max_error)
-      << "Error: 32x32 FDCT/IDCT has an individual round-trip error > 1";
-
-  EXPECT_GE(count_test_block << 2 * (bit_depth_ - 8), total_error)
-      << "Error: 32x32 FDCT/IDCT has average round-trip error > 1 per block";
-}
-
-TEST_P(Trans32x32Test, CoeffCheck) {
-  ACMRandom rnd(ACMRandom::DeterministicSeed());
-  const int count_test_block = 1000;
-
-  DECLARE_ALIGNED(16, int16_t, input_block[kNumCoeffs]);
-  DECLARE_ALIGNED(16, tran_low_t, output_ref_block[kNumCoeffs]);
-  DECLARE_ALIGNED(16, tran_low_t, output_block[kNumCoeffs]);
-
-  for (int i = 0; i < count_test_block; ++i) {
-    for (int j = 0; j < kNumCoeffs; ++j)
-      input_block[j] = (rnd.Rand16() & mask_) - (rnd.Rand16() & mask_);
-
-    const int stride = 32;
-    aom_fdct32x32_c(input_block, output_ref_block, stride);
-    ASM_REGISTER_STATE_CHECK(fwd_txfm_(input_block, output_block, stride));
-
-    if (version_ == 0) {
-      for (int j = 0; j < kNumCoeffs; ++j)
-        EXPECT_EQ(output_block[j], output_ref_block[j])
-            << "Error: 32x32 FDCT versions have mismatched coefficients";
-    } else {
-      for (int j = 0; j < kNumCoeffs; ++j)
-        EXPECT_GE(6, abs(output_block[j] - output_ref_block[j]))
-            << "Error: 32x32 FDCT rd has mismatched coefficients";
-    }
-  }
-}
-
-TEST_P(Trans32x32Test, MemCheck) {
-  ACMRandom rnd(ACMRandom::DeterministicSeed());
-  const int count_test_block = 2000;
-
-  DECLARE_ALIGNED(16, int16_t, input_extreme_block[kNumCoeffs]);
-  DECLARE_ALIGNED(16, tran_low_t, output_ref_block[kNumCoeffs]);
-  DECLARE_ALIGNED(16, tran_low_t, output_block[kNumCoeffs]);
-
-  for (int i = 0; i < count_test_block; ++i) {
-    // Initialize a test block with input range [-mask_, mask_].
-    for (int j = 0; j < kNumCoeffs; ++j) {
-      input_extreme_block[j] = rnd.Rand8() & 1 ? mask_ : -mask_;
-    }
-    if (i == 0) {
-      for (int j = 0; j < kNumCoeffs; ++j) input_extreme_block[j] = mask_;
-    } else if (i == 1) {
-      for (int j = 0; j < kNumCoeffs; ++j) input_extreme_block[j] = -mask_;
-    }
-
-    const int stride = 32;
-    aom_fdct32x32_c(input_extreme_block, output_ref_block, stride);
-    ASM_REGISTER_STATE_CHECK(
-        fwd_txfm_(input_extreme_block, output_block, stride));
-
-    // The minimum quant value is 4.
-    for (int j = 0; j < kNumCoeffs; ++j) {
-      if (version_ == 0) {
-        EXPECT_EQ(output_block[j], output_ref_block[j])
-            << "Error: 32x32 FDCT versions have mismatched coefficients";
-      } else {
-        EXPECT_GE(6, abs(output_block[j] - output_ref_block[j]))
-            << "Error: 32x32 FDCT rd has mismatched coefficients";
-      }
-      EXPECT_GE(4 * DCT_MAX_VALUE << (bit_depth_ - 8), abs(output_ref_block[j]))
-          << "Error: 32x32 FDCT C has coefficient larger than 4*DCT_MAX_VALUE";
-      EXPECT_GE(4 * DCT_MAX_VALUE << (bit_depth_ - 8), abs(output_block[j]))
-          << "Error: 32x32 FDCT has coefficient larger than "
-          << "4*DCT_MAX_VALUE";
-    }
-  }
-}
-
-TEST_P(Trans32x32Test, InverseAccuracy) {
-  ACMRandom rnd(ACMRandom::DeterministicSeed());
-  const int count_test_block = 1000;
-  DECLARE_ALIGNED(16, int16_t, in[kNumCoeffs]);
-  DECLARE_ALIGNED(16, tran_low_t, coeff[kNumCoeffs]);
-  DECLARE_ALIGNED(16, uint8_t, dst[kNumCoeffs]);
-  DECLARE_ALIGNED(16, uint8_t, src[kNumCoeffs]);
-#if CONFIG_HIGHBITDEPTH
-  DECLARE_ALIGNED(16, uint16_t, dst16[kNumCoeffs]);
-  DECLARE_ALIGNED(16, uint16_t, src16[kNumCoeffs]);
-#endif
-
-  for (int i = 0; i < count_test_block; ++i) {
-    double out_r[kNumCoeffs];
-
-    // Initialize a test block with input range [-255, 255]
-    for (int j = 0; j < kNumCoeffs; ++j) {
-      if (bit_depth_ == AOM_BITS_8) {
-        src[j] = rnd.Rand8();
-        dst[j] = rnd.Rand8();
-        in[j] = src[j] - dst[j];
-#if CONFIG_HIGHBITDEPTH
-      } else {
-        src16[j] = rnd.Rand16() & mask_;
-        dst16[j] = rnd.Rand16() & mask_;
-        in[j] = src16[j] - dst16[j];
-#endif
-      }
-    }
-
-    reference_32x32_dct_2d(in, out_r);
-    for (int j = 0; j < kNumCoeffs; ++j)
-      coeff[j] = static_cast<tran_low_t>(round(out_r[j]));
-    if (bit_depth_ == AOM_BITS_8) {
-      ASM_REGISTER_STATE_CHECK(inv_txfm_(coeff, dst, 32));
-#if CONFIG_HIGHBITDEPTH
-    } else {
-      ASM_REGISTER_STATE_CHECK(inv_txfm_(coeff, CONVERT_TO_BYTEPTR(dst16), 32));
-#endif
-    }
-    for (int j = 0; j < kNumCoeffs; ++j) {
-#if CONFIG_HIGHBITDEPTH
-      const int diff =
-          bit_depth_ == AOM_BITS_8 ? dst[j] - src[j] : dst16[j] - src16[j];
-#else
-      const int diff = dst[j] - src[j];
-#endif
-      const int error = diff * diff;
-      EXPECT_GE(1, error) << "Error: 32x32 IDCT has error " << error
-                          << " at index " << j;
-    }
-  }
-}
-
-class PartialTrans32x32Test
-    : public ::testing::TestWithParam<
-          std::tr1::tuple<FwdTxfmFunc, aom_bit_depth_t> > {
- public:
-  virtual ~PartialTrans32x32Test() {}
-  virtual void SetUp() {
-    fwd_txfm_ = GET_PARAM(0);
-    bit_depth_ = GET_PARAM(1);
-  }
-
-  virtual void TearDown() { libaom_test::ClearSystemState(); }
-
- protected:
-  aom_bit_depth_t bit_depth_;
-  FwdTxfmFunc fwd_txfm_;
-};
-
-TEST_P(PartialTrans32x32Test, Extremes) {
-#if CONFIG_HIGHBITDEPTH
-  const int16_t maxval =
-      static_cast<int16_t>(clip_pixel_highbd(1 << 30, bit_depth_));
-#else
-  const int16_t maxval = 255;
-#endif
-  const int minval = -maxval;
-  DECLARE_ALIGNED(16, int16_t, input[kNumCoeffs]);
-  DECLARE_ALIGNED(16, tran_low_t, output[kNumCoeffs]);
-
-  for (int i = 0; i < kNumCoeffs; ++i) input[i] = maxval;
-  output[0] = 0;
-  ASM_REGISTER_STATE_CHECK(fwd_txfm_(input, output, 32));
-  EXPECT_EQ((maxval * kNumCoeffs) >> 3, output[0]);
-
-  for (int i = 0; i < kNumCoeffs; ++i) input[i] = minval;
-  output[0] = 0;
-  ASM_REGISTER_STATE_CHECK(fwd_txfm_(input, output, 32));
-  EXPECT_EQ((minval * kNumCoeffs) >> 3, output[0]);
-}
-
-TEST_P(PartialTrans32x32Test, Random) {
-#if CONFIG_HIGHBITDEPTH
-  const int16_t maxval =
-      static_cast<int16_t>(clip_pixel_highbd(1 << 30, bit_depth_));
-#else
-  const int16_t maxval = 255;
-#endif
-  DECLARE_ALIGNED(16, int16_t, input[kNumCoeffs]);
-  DECLARE_ALIGNED(16, tran_low_t, output[kNumCoeffs]);
-  ACMRandom rnd(ACMRandom::DeterministicSeed());
-
-  int sum = 0;
-  for (int i = 0; i < kNumCoeffs; ++i) {
-    const int val = (i & 1) ? -rnd(maxval + 1) : rnd(maxval + 1);
-    input[i] = val;
-    sum += val;
-  }
-  output[0] = 0;
-  ASM_REGISTER_STATE_CHECK(fwd_txfm_(input, output, 32));
-  EXPECT_EQ(sum >> 3, output[0]);
-}
-
-using std::tr1::make_tuple;
-
-#if CONFIG_HIGHBITDEPTH
-INSTANTIATE_TEST_CASE_P(
-    C, Trans32x32Test,
-    ::testing::Values(make_tuple(&aom_fdct32x32_c, &aom_idct32x32_1024_add_c, 0,
-                                 AOM_BITS_8),
-                      make_tuple(&aom_fdct32x32_rd_c, &aom_idct32x32_1024_add_c,
-                                 1, AOM_BITS_8)));
-#else
-INSTANTIATE_TEST_CASE_P(
-    C, Trans32x32Test,
-    ::testing::Values(make_tuple(&aom_fdct32x32_c, &aom_idct32x32_1024_add_c, 0,
-                                 AOM_BITS_8),
-                      make_tuple(&aom_fdct32x32_rd_c, &aom_idct32x32_1024_add_c,
-                                 1, AOM_BITS_8)));
-#endif  // CONFIG_HIGHBITDEPTH
-
-#if HAVE_NEON && !CONFIG_HIGHBITDEPTH
-INSTANTIATE_TEST_CASE_P(
-    NEON, Trans32x32Test,
-    ::testing::Values(make_tuple(&aom_fdct32x32_c, &aom_idct32x32_1024_add_neon,
-                                 DCT_DCT, AOM_BITS_8),
-                      make_tuple(&aom_fdct32x32_rd_c,
-                                 &aom_idct32x32_1024_add_neon, ADST_DCT,
-                                 AOM_BITS_8)));
-#endif  // HAVE_NEON && !CONFIG_HIGHBITDEPTH
-
-#if HAVE_SSE2 && !CONFIG_HIGHBITDEPTH
-INSTANTIATE_TEST_CASE_P(
-    SSE2, Trans32x32Test,
-    ::testing::Values(make_tuple(&aom_fdct32x32_sse2,
-                                 &aom_idct32x32_1024_add_sse2, DCT_DCT,
-                                 AOM_BITS_8),
-                      make_tuple(&aom_fdct32x32_rd_sse2,
-                                 &aom_idct32x32_1024_add_sse2, ADST_DCT,
-                                 AOM_BITS_8)));
-#endif  // HAVE_SSE2 && !CONFIG_HIGHBITDEPTH
-
-#if HAVE_SSE2 && CONFIG_HIGHBITDEPTH
-INSTANTIATE_TEST_CASE_P(SSE2, Trans32x32Test,
-                        ::testing::Values(make_tuple(&aom_fdct32x32_sse2,
-                                                     &aom_idct32x32_1024_add_c,
-                                                     DCT_DCT, AOM_BITS_8),
-                                          make_tuple(&aom_fdct32x32_rd_sse2,
-                                                     &aom_idct32x32_1024_add_c,
-                                                     ADST_DCT, AOM_BITS_8)));
-#endif  // HAVE_SSE2 && CONFIG_HIGHBITDEPTH
-
-#if HAVE_AVX2 && !CONFIG_HIGHBITDEPTH
-INSTANTIATE_TEST_CASE_P(
-    AVX2, Trans32x32Test,
-    ::testing::Values(make_tuple(&aom_fdct32x32_avx2,
-                                 &aom_idct32x32_1024_add_sse2, DCT_DCT,
-                                 AOM_BITS_8),
-                      make_tuple(&aom_fdct32x32_rd_avx2,
-                                 &aom_idct32x32_1024_add_sse2, ADST_DCT,
-                                 AOM_BITS_8)));
-#endif  // HAVE_AVX2 && !CONFIG_HIGHBITDEPTH
-
-#if HAVE_AVX2 && CONFIG_HIGHBITDEPTH
-INSTANTIATE_TEST_CASE_P(
-    AVX2, Trans32x32Test,
-    ::testing::Values(make_tuple(&aom_fdct32x32_avx2,
-                                 &aom_idct32x32_1024_add_sse2, DCT_DCT,
-                                 AOM_BITS_8),
-                      make_tuple(&aom_fdct32x32_rd_avx2,
-                                 &aom_idct32x32_1024_add_sse2, ADST_DCT,
-                                 AOM_BITS_8)));
-#endif  // HAVE_AVX2 && CONFIG_HIGHBITDEPTH
-
-#if HAVE_MSA && !CONFIG_HIGHBITDEPTH
-INSTANTIATE_TEST_CASE_P(
-    MSA, Trans32x32Test,
-    ::testing::Values(make_tuple(&aom_fdct32x32_msa,
-                                 &aom_idct32x32_1024_add_msa, DCT_DCT,
-                                 AOM_BITS_8),
-                      make_tuple(&aom_fdct32x32_rd_msa,
-                                 &aom_idct32x32_1024_add_msa, ADST_DCT,
-                                 AOM_BITS_8)));
-#endif  // HAVE_MSA && !CONFIG_HIGHBITDEPTH
-}  // namespace

third_party/aom/test/decode_api_test.cc

@@ -7,12 +7,12 @@
  * obtain it at www.aomedia.org/license/software. If the Alliance for Open
  * Media Patent License 1.0 was not distributed with this source code in the
  * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
+ */
 
 #include "third_party/googletest/src/googletest/include/gtest/gtest.h"
 
-#include "./aom_config.h"
-#include "test/ivf_video_source.h"
+#include "config/aom_config.h"
+
 #include "test/util.h"
 #include "aom/aomdx.h"
 #include "aom/aom_decoder.h"
@@ -30,12 +30,12 @@ TEST(DecodeAPI, InvalidParams) {
 
   EXPECT_EQ(AOM_CODEC_INVALID_PARAM, aom_codec_dec_init(NULL, NULL, NULL, 0));
   EXPECT_EQ(AOM_CODEC_INVALID_PARAM, aom_codec_dec_init(&dec, NULL, NULL, 0));
-  EXPECT_EQ(AOM_CODEC_INVALID_PARAM, aom_codec_decode(NULL, NULL, 0, NULL, 0));
-  EXPECT_EQ(AOM_CODEC_INVALID_PARAM, aom_codec_decode(NULL, buf, 0, NULL, 0));
+  EXPECT_EQ(AOM_CODEC_INVALID_PARAM, aom_codec_decode(NULL, NULL, 0, NULL));
+  EXPECT_EQ(AOM_CODEC_INVALID_PARAM, aom_codec_decode(NULL, buf, 0, NULL));
   EXPECT_EQ(AOM_CODEC_INVALID_PARAM,
-            aom_codec_decode(NULL, buf, NELEMENTS(buf), NULL, 0));
+            aom_codec_decode(NULL, buf, NELEMENTS(buf), NULL));
   EXPECT_EQ(AOM_CODEC_INVALID_PARAM,
-            aom_codec_decode(NULL, NULL, NELEMENTS(buf), NULL, 0));
+            aom_codec_decode(NULL, NULL, NELEMENTS(buf), NULL));
   EXPECT_EQ(AOM_CODEC_INVALID_PARAM, aom_codec_destroy(NULL));
   EXPECT_TRUE(aom_codec_error(NULL) != NULL);
 
@@ -44,14 +44,9 @@ TEST(DecodeAPI, InvalidParams) {
               aom_codec_dec_init(NULL, kCodecs[i], NULL, 0));
 
     EXPECT_EQ(AOM_CODEC_OK, aom_codec_dec_init(&dec, kCodecs[i], NULL, 0));
-#if !CONFIG_OBU
-    // Needs to be fixed
-    EXPECT_EQ(AOM_CODEC_UNSUP_BITSTREAM,
-              aom_codec_decode(&dec, buf, NELEMENTS(buf), NULL, 0));
-#endif
     EXPECT_EQ(AOM_CODEC_INVALID_PARAM,
-              aom_codec_decode(&dec, NULL, NELEMENTS(buf), NULL, 0));
-    EXPECT_EQ(AOM_CODEC_INVALID_PARAM, aom_codec_decode(&dec, buf, 0, NULL, 0));
+              aom_codec_decode(&dec, NULL, NELEMENTS(buf), NULL));
+    EXPECT_EQ(AOM_CODEC_INVALID_PARAM, aom_codec_decode(&dec, buf, 0, NULL));
 
     EXPECT_EQ(AOM_CODEC_OK, aom_codec_destroy(&dec));
   }

third_party/aom/test/decode_multithreaded_test.cc

@@ -0,0 +1,179 @@
+/*
+ * Copyright (c) 2018, Alliance for Open Media. All rights reserved
+ *
+ * This source code is subject to the terms of the BSD 2 Clause License and
+ * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
+ * was not distributed with this source code in the LICENSE file, you can
+ * obtain it at www.aomedia.org/license/software. If the Alliance for Open
+ * Media Patent License 1.0 was not distributed with this source code in the
+ * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
+ */
+
+#include <cstdio>
+#include <cstdlib>
+#include <string>
+
+#include "aom_mem/aom_mem.h"
+#include "test/codec_factory.h"
+#include "test/encode_test_driver.h"
+#include "test/i420_video_source.h"
+#include "test/md5_helper.h"
+#include "test/util.h"
+#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
+
+namespace {
+
+static const int kNumMultiThreadDecoders = 3;
+
+class AV1DecodeMultiThreadedTest
+    : public ::libaom_test::CodecTestWith4Params<int, int, int, int>,
+      public ::libaom_test::EncoderTest {
+ protected:
+  AV1DecodeMultiThreadedTest()
+      : EncoderTest(GET_PARAM(0)), md5_single_thread_(), md5_multi_thread_(),
+        n_tile_cols_(GET_PARAM(1)), n_tile_rows_(GET_PARAM(2)),
+        n_tile_groups_(GET_PARAM(3)), set_cpu_used_(GET_PARAM(4)) {
+    init_flags_ = AOM_CODEC_USE_PSNR;
+    aom_codec_dec_cfg_t cfg = aom_codec_dec_cfg_t();
+    cfg.w = 704;
+    cfg.h = 576;
+    cfg.threads = 1;
+    cfg.allow_lowbitdepth = 1;
+    single_thread_dec_ = codec_->CreateDecoder(cfg, 0);
+
+    // Test cfg.threads == powers of 2.
+    for (int i = 0; i < kNumMultiThreadDecoders; ++i) {
+      cfg.threads <<= 1;
+      multi_thread_dec_[i] = codec_->CreateDecoder(cfg, 0);
+    }
+
+    if (single_thread_dec_->IsAV1()) {
+      single_thread_dec_->Control(AV1_SET_DECODE_TILE_ROW, -1);
+      single_thread_dec_->Control(AV1_SET_DECODE_TILE_COL, -1);
+    }
+    for (int i = 0; i < kNumMultiThreadDecoders; ++i) {
+      if (multi_thread_dec_[i]->IsAV1()) {
+        multi_thread_dec_[i]->Control(AV1_SET_DECODE_TILE_ROW, -1);
+        multi_thread_dec_[i]->Control(AV1_SET_DECODE_TILE_COL, -1);
+      }
+    }
+  }
+
+  virtual ~AV1DecodeMultiThreadedTest() {
+    delete single_thread_dec_;
+    for (int i = 0; i < kNumMultiThreadDecoders; ++i)
+      delete multi_thread_dec_[i];
+  }
+
+  virtual void SetUp() {
+    InitializeConfig();
+    SetMode(libaom_test::kTwoPassGood);
+  }
+
+  virtual void PreEncodeFrameHook(libaom_test::VideoSource *video,
+                                  libaom_test::Encoder *encoder) {
+    if (video->frame() == 1) {
+      encoder->Control(AV1E_SET_TILE_COLUMNS, n_tile_cols_);
+      encoder->Control(AV1E_SET_TILE_ROWS, n_tile_rows_);
+      encoder->Control(AV1E_SET_NUM_TG, n_tile_groups_);
+      encoder->Control(AOME_SET_CPUUSED, set_cpu_used_);
+    }
+  }
+
+  void UpdateMD5(::libaom_test::Decoder *dec, const aom_codec_cx_pkt_t *pkt,
+                 ::libaom_test::MD5 *md5) {
+    const aom_codec_err_t res = dec->DecodeFrame(
+        reinterpret_cast<uint8_t *>(pkt->data.frame.buf), pkt->data.frame.sz);
+    if (res != AOM_CODEC_OK) {
+      abort_ = true;
+      ASSERT_EQ(AOM_CODEC_OK, res);
+    }
+    const aom_image_t *img = dec->GetDxData().Next();
+    md5->Add(img);
+  }
+
+  virtual void FramePktHook(const aom_codec_cx_pkt_t *pkt) {
+    UpdateMD5(single_thread_dec_, pkt, &md5_single_thread_);
+
+    for (int i = 0; i < kNumMultiThreadDecoders; ++i)
+      UpdateMD5(multi_thread_dec_[i], pkt, &md5_multi_thread_[i]);
+  }
+
+  void DoTest() {
+    const aom_rational timebase = { 33333333, 1000000000 };
+    cfg_.g_timebase = timebase;
+    cfg_.rc_target_bitrate = 500;
+    cfg_.g_lag_in_frames = 12;
+    cfg_.rc_end_usage = AOM_VBR;
+
+    libaom_test::I420VideoSource video("hantro_collage_w352h288.yuv", 704, 576,
+                                       timebase.den, timebase.num, 0, 5);
+    ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
+
+    const char *md5_single_thread_str = md5_single_thread_.Get();
+
+    for (int i = 0; i < kNumMultiThreadDecoders; ++i) {
+      const char *md5_multi_thread_str = md5_multi_thread_[i].Get();
+      ASSERT_STREQ(md5_single_thread_str, md5_multi_thread_str);
+    }
+  }
+
+  ::libaom_test::MD5 md5_single_thread_;
+  ::libaom_test::MD5 md5_multi_thread_[kNumMultiThreadDecoders];
+  ::libaom_test::Decoder *single_thread_dec_;
+  ::libaom_test::Decoder *multi_thread_dec_[kNumMultiThreadDecoders];
+
+ private:
+  int n_tile_cols_;
+  int n_tile_rows_;
+  int n_tile_groups_;
+  int set_cpu_used_;
+};
+
+// run an encode and do the decode both in single thread
+// and multi thread. Ensure that the MD5 of the output in both cases
+// is identical. If so, the test passes.
+TEST_P(AV1DecodeMultiThreadedTest, MD5Match) {
+  cfg_.large_scale_tile = 0;
+  single_thread_dec_->Control(AV1_SET_TILE_MODE, 0);
+  for (int i = 0; i < kNumMultiThreadDecoders; ++i)
+    multi_thread_dec_[i]->Control(AV1_SET_TILE_MODE, 0);
+  DoTest();
+}
+
+class AV1DecodeMultiThreadedTestLarge : public AV1DecodeMultiThreadedTest {};
+
+TEST_P(AV1DecodeMultiThreadedTestLarge, MD5Match) {
+  cfg_.large_scale_tile = 0;
+  single_thread_dec_->Control(AV1_SET_TILE_MODE, 0);
+  for (int i = 0; i < kNumMultiThreadDecoders; ++i)
+    multi_thread_dec_[i]->Control(AV1_SET_TILE_MODE, 0);
+  DoTest();
+}
+
+// TODO(ranjit): More tests have to be added using pre-generated MD5.
+AV1_INSTANTIATE_TEST_CASE(AV1DecodeMultiThreadedTest, ::testing::Values(1, 2),
+                          ::testing::Values(1, 2), ::testing::Values(1),
+                          ::testing::Values(3));
+AV1_INSTANTIATE_TEST_CASE(AV1DecodeMultiThreadedTestLarge,
+                          ::testing::Values(0, 1, 2, 6),
+                          ::testing::Values(0, 1, 2, 6),
+                          ::testing::Values(1, 4), ::testing::Values(0));
+
+class AV1DecodeMultiThreadedLSTestLarge
+    : public AV1DecodeMultiThreadedTestLarge {};
+
+TEST_P(AV1DecodeMultiThreadedLSTestLarge, DISABLED_MD5Match) {
+  cfg_.large_scale_tile = 1;
+  single_thread_dec_->Control(AV1_SET_TILE_MODE, 1);
+  for (int i = 0; i < kNumMultiThreadDecoders; ++i)
+    multi_thread_dec_[i]->Control(AV1_SET_TILE_MODE, 1);
+  DoTest();
+}
+
+AV1_INSTANTIATE_TEST_CASE(AV1DecodeMultiThreadedLSTestLarge,
+                          ::testing::Values(1, 2, 32),
+                          ::testing::Values(1, 2, 32), ::testing::Values(1),
+                          ::testing::Values(0, 3));
+
+}  // namespace

third_party/aom/test/decode_perf_test.cc

@@ -7,9 +7,14 @@
  * obtain it at www.aomedia.org/license/software. If the Alliance for Open
  * Media Patent License 1.0 was not distributed with this source code in the
  * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
+ */
 
 #include <string>
+
+#include "config/aom_version.h"
+
+#include "aom_ports/aom_timer.h"
+#include "common/ivfenc.h"
 #include "test/codec_factory.h"
 #include "test/decode_test_driver.h"
 #include "test/encode_test_driver.h"
@@ -18,25 +23,21 @@
 #include "test/md5_helper.h"
 #include "test/util.h"
 #include "test/webm_video_source.h"
-#include "aom_ports/aom_timer.h"
-#include "./ivfenc.h"
-#include "./aom_version.h"
 
-using std::tr1::make_tuple;
+using ::testing::make_tuple;
 
 namespace {
 
 #define VIDEO_NAME 0
 #define THREADS 1
 
-const int kMaxPsnr = 100;
 const double kUsecsInSec = 1000000.0;
 const char kNewEncodeOutputFile[] = "new_encode.ivf";
 
 /*
  DecodePerfTest takes a tuple of filename + number of threads to decode with
  */
-typedef std::tr1::tuple<const char *, unsigned> DecodePerfParam;
+typedef ::testing::tuple<const char *, unsigned> DecodePerfParam;
 
 // TODO(jimbankoski): Add actual test vectors here when available.
 // const DecodePerfParam kAV1DecodePerfVectors[] = {};
@@ -129,7 +130,8 @@ class AV1NewEncodeDecodePerfTest
   }
 
   virtual void BeginPassHook(unsigned int /*pass*/) {
-    const std::string data_path = getenv("LIBAOM_TEST_DATA_PATH");
+    const char *const env = getenv("LIBAOM_TEST_DATA_PATH");
+    const std::string data_path(env ? env : ".");
     const std::string path_to_source = data_path + "/" + kNewEncodeOutputFile;
     outfile_ = fopen(path_to_source.c_str(), "wb");
     ASSERT_TRUE(outfile_ != NULL);
@@ -157,7 +159,7 @@ class AV1NewEncodeDecodePerfTest
               pkt->data.frame.sz);
   }
 
-  virtual bool DoDecode() { return false; }
+  virtual bool DoDecode() const { return false; }
 
   void set_speed(unsigned int speed) { speed_ = speed; }
 

third_party/aom/test/decode_test_driver.cc

@@ -7,7 +7,7 @@
  * obtain it at www.aomedia.org/license/software. If the Alliance for Open
  * Media Patent License 1.0 was not distributed with this source code in the
  * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
+ */
 
 #include "third_party/googletest/src/googletest/include/gtest/gtest.h"
 
@@ -18,13 +18,12 @@
 
 namespace libaom_test {
 
-const char kVP8Name[] = "WebM Project VP8";
 const char kAV1Name[] = "AOMedia Project AV1 Decoder";
 
 aom_codec_err_t Decoder::PeekStream(const uint8_t *cxdata, size_t size,
                                     aom_codec_stream_info_t *stream_info) {
-  return aom_codec_peek_stream_info(
-      CodecInterface(), cxdata, static_cast<unsigned int>(size), stream_info);
+  return aom_codec_peek_stream_info(CodecInterface(), cxdata, size,
+                                    stream_info);
 }
 
 aom_codec_err_t Decoder::DecodeFrame(const uint8_t *cxdata, size_t size) {
@@ -36,39 +35,22 @@ aom_codec_err_t Decoder::DecodeFrame(const uint8_t *cxdata, size_t size,
   aom_codec_err_t res_dec;
   InitOnce();
   API_REGISTER_STATE_CHECK(
-      res_dec = aom_codec_decode(
-          &decoder_, cxdata, static_cast<unsigned int>(size), user_priv, 0));
+      res_dec = aom_codec_decode(&decoder_, cxdata, size, user_priv));
   return res_dec;
 }
 
-bool Decoder::IsVP8() const {
-  const char *codec_name = GetDecoderName();
-  return strncmp(kVP8Name, codec_name, sizeof(kVP8Name) - 1) == 0;
-}
-
 bool Decoder::IsAV1() const {
   const char *codec_name = GetDecoderName();
   return strncmp(kAV1Name, codec_name, sizeof(kAV1Name) - 1) == 0;
 }
 
-void DecoderTest::HandlePeekResult(Decoder *const decoder,
-                                   CompressedVideoSource *video,
+void DecoderTest::HandlePeekResult(Decoder *const /*decoder*/,
+                                   CompressedVideoSource * /*video*/,
                                    const aom_codec_err_t res_peek) {
-  const bool is_vp8 = decoder->IsVP8();
-  if (is_vp8) {
-    /* Vp8's implementation of PeekStream returns an error if the frame you
-     * pass it is not a keyframe, so we only expect AOM_CODEC_OK on the first
-     * frame, which must be a keyframe. */
-    if (video->frame_number() == 0) {
-      ASSERT_EQ(AOM_CODEC_OK, res_peek)
-          << "Peek return failed: " << aom_codec_err_to_string(res_peek);
-    }
-  } else {
-    /* The Av1 implementation of PeekStream returns an error only if the
-     * data passed to it isn't a valid Av1 chunk. */
-    ASSERT_EQ(AOM_CODEC_OK, res_peek)
-        << "Peek return failed: " << aom_codec_err_to_string(res_peek);
-  }
+  /* The Av1 implementation of PeekStream returns an error only if the
+   * data passed to it isn't a valid Av1 chunk. */
+  ASSERT_EQ(AOM_CODEC_OK, res_peek)
+      << "Peek return failed: " << aom_codec_err_to_string(res_peek);
 }
 
 void DecoderTest::RunLoop(CompressedVideoSource *video,
@@ -76,6 +58,7 @@ void DecoderTest::RunLoop(CompressedVideoSource *video,
   Decoder *const decoder = codec_->CreateDecoder(dec_cfg, flags_);
   ASSERT_TRUE(decoder != NULL);
   bool end_of_file = false;
+  bool peeked_stream = false;
 
   // Decode frames.
   for (video->Begin(); !::testing::Test::HasFailure() && !end_of_file;
@@ -83,15 +66,23 @@ void DecoderTest::RunLoop(CompressedVideoSource *video,
     PreDecodeFrameHook(*video, decoder);
 
     aom_codec_stream_info_t stream_info;
+    stream_info.is_annexb = 0;
+
     if (video->cxdata() != NULL) {
-      const aom_codec_err_t res_peek = decoder->PeekStream(
-          video->cxdata(), video->frame_size(), &stream_info);
-      HandlePeekResult(decoder, video, res_peek);
-      ASSERT_FALSE(::testing::Test::HasFailure());
+      if (!peeked_stream) {
+        // TODO(yaowu): PeekStream returns error for non-sequence_header_obu,
+        // therefore should only be tried once per sequence, this shall be fixed
+        // once PeekStream is updated to properly operate on other obus.
+        const aom_codec_err_t res_peek = decoder->PeekStream(
+            video->cxdata(), video->frame_size(), &stream_info);
+        HandlePeekResult(decoder, video, res_peek);
+        ASSERT_FALSE(::testing::Test::HasFailure());
+        peeked_stream = true;
+      }
 
       aom_codec_err_t res_dec =
           decoder->DecodeFrame(video->cxdata(), video->frame_size());
-      if (!HandleDecodeResult(res_dec, decoder)) break;
+      if (!HandleDecodeResult(res_dec, *video, decoder)) break;
     } else {
       // Signal end of the file to the decoder.
       const aom_codec_err_t res_dec = decoder->DecodeFrame(NULL, 0);

third_party/aom/test/decode_test_driver.h

@@ -13,7 +13,9 @@
 #define TEST_DECODE_TEST_DRIVER_H_
 #include <cstring>
 #include "third_party/googletest/src/googletest/include/gtest/gtest.h"
-#include "./aom_config.h"
+
+#include "config/aom_config.h"
+
 #include "aom/aom_decoder.h"
 
 namespace libaom_test {
@@ -93,8 +95,6 @@ class Decoder {
     return aom_codec_iface_name(CodecInterface());
   }
 
-  bool IsVP8() const;
-
   bool IsAV1() const;
 
   aom_codec_ctx_t *GetDecoder() { return &decoder_; }
@@ -134,6 +134,7 @@ class DecoderTest {
 
   // Hook to be called to handle decode result. Return true to continue.
   virtual bool HandleDecodeResult(const aom_codec_err_t res_dec,
+                                  const CompressedVideoSource & /*video*/,
                                   Decoder *decoder) {
     EXPECT_EQ(AOM_CODEC_OK, res_dec) << decoder->DecodeError();
     return AOM_CODEC_OK == res_dec;

third_party/aom/test/decode_to_md5.sh

@@ -16,7 +16,7 @@
 . $(dirname $0)/tools_common.sh
 
 # Environment check: Make sure input is available:
-#   $AOM_IVF_FILE and $AV1_IVF_FILE are required.
+#   $AV1_IVF_FILE is required.
 decode_to_md5_verify_environment() {
   if [ "$(av1_encode_available)" != "yes" ] && [ ! -e "${AV1_IVF_FILE}" ]; then
     return 1
@@ -27,7 +27,7 @@ decode_to_md5_verify_environment() {
 # interpreted as codec name and used solely to name the output file. $3 is the
 # expected md5 sum: It must match that of the final frame.
 decode_to_md5() {
-  local decoder="${LIBAOM_BIN_PATH}/decode_to_md5${AOM_TEST_EXE_SUFFIX}"
+  local decoder="$(aom_tool_path decode_to_md5)"
   local input_file="$1"
   local codec="$2"
   local expected_md5="$3"
@@ -45,14 +45,23 @@ decode_to_md5() {
 
   local md5_last_frame="$(tail -n1 "${output_file}" | awk '{print $1}')"
   local actual_md5="$(echo "${md5_last_frame}" | awk '{print $1}')"
-  [ "${actual_md5}" = "${expected_md5}" ] || return 1
+  if [ "${actual_md5}" = "${expected_md5}" ]; then
+    return 0
+  else
+    elog "MD5 mismatch:"
+    elog "Expected: ${expected_md5}"
+    elog "Actual: ${actual_md5}"
+    return 1
+  fi
 }
 
-decode_to_md5_av1() {
+DISABLED_decode_to_md5_av1() {
   # expected MD5 sum for the last frame.
-  local expected_md5="26d3ef1d60754a1f6acb603c3763efbe"
+  local expected_md5="567dd6d4b7a7170edddbf58bbcc3aff1"
   local file="${AV1_IVF_FILE}"
 
+  # TODO(urvang): Check in the encoded file (like libvpx does) to avoid
+  # encoding every time.
   if [ "$(av1_decode_available)" = "yes" ]; then
     if [ ! -e "${AV1_IVF_FILE}" ]; then
       file="${AOM_TEST_OUTPUT_DIR}/test_encode.ivf"
@@ -62,6 +71,7 @@ decode_to_md5_av1() {
   fi
 }
 
-decode_to_md5_tests="decode_to_md5_av1"
+# TODO(tomfinegan): Enable when the bitstream stabilizes.
+decode_to_md5_tests="DISABLED_decode_to_md5_av1"
 
 run_tests decode_to_md5_verify_environment "${decode_to_md5_tests}"

third_party/aom/test/decode_with_drops.sh

@@ -16,7 +16,7 @@
 . $(dirname $0)/tools_common.sh
 
 # Environment check: Make sure input is available:
-#   $AOM_IVF_FILE and $AV1_IVF_FILE are required.
+#   $AV1_IVF_FILE is required.
 decode_with_drops_verify_environment() {
   if [ "$(av1_encode_available)" != "yes" ] && [ ! -e "${AV1_IVF_FILE}" ]; then
     return 1
@@ -27,7 +27,7 @@ decode_with_drops_verify_environment() {
 # to name the output file. $3 is the drop mode, and is passed directly to
 # decode_with_drops.
 decode_with_drops() {
-  local decoder="${LIBAOM_BIN_PATH}/decode_with_drops${AOM_TEST_EXE_SUFFIX}"
+  local decoder="$(aom_tool_path decode_with_drops)"
   local input_file="$1"
   local codec="$2"
   local output_file="${AOM_TEST_OUTPUT_DIR}/decode_with_drops_${codec}"
@@ -47,21 +47,22 @@ decode_with_drops() {
 
 # Decodes $AV1_IVF_FILE while dropping frames, twice: once in sequence mode,
 # and once in pattern mode.
-decode_with_drops_av1() {
+DISABLED_decode_with_drops_av1() {
   if [ "$(av1_decode_available)" = "yes" ]; then
     local file="${AV1_IVF_FILE}"
     if [ ! -e "${AV1_IVF_FILE}" ]; then
       file="${AOM_TEST_OUTPUT_DIR}/test_encode.ivf"
       encode_yuv_raw_input_av1 "${file}" --ivf
     fi
-    # Drop frames 2 and 3.
-    decode_with_drops "${file}" "av1" "2-3"
+    # Drop frames 3 and 4.
+    decode_with_drops "${file}" "av1" "3-4"
 
     # Test pattern mode: Drop 3 of every 4 frames.
     decode_with_drops "${file}" "av1" "3/4"
   fi
 }
 
-decode_with_drops_tests="decode_with_drops_av1"
+# TODO(yaowu): Disable this test as trailing_bit check is expected to fail
+decode_with_drops_tests="DISABLED_decode_with_drops_av1"
 
 run_tests decode_with_drops_verify_environment "${decode_with_drops_tests}"

third_party/aom/test/dering_test.cc

@@ -1,383 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
-
-#include <cstdlib>
-#include <string>
-
-#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
-
-#include "./aom_config.h"
-#include "./av1_rtcd.h"
-#include "aom_ports/aom_timer.h"
-#include "av1/common/cdef_block.h"
-#include "test/acm_random.h"
-#include "test/clear_system_state.h"
-#include "test/register_state_check.h"
-#include "test/util.h"
-
-using libaom_test::ACMRandom;
-
-namespace {
-
-typedef std::tr1::tuple<cdef_direction_func, cdef_direction_func, int>
-    dering_dir_param_t;
-
-class CDEFDeringDirTest : public ::testing::TestWithParam<dering_dir_param_t> {
- public:
-  virtual ~CDEFDeringDirTest() {}
-  virtual void SetUp() {
-    dering = GET_PARAM(0);
-    ref_dering = GET_PARAM(1);
-    bsize = GET_PARAM(2);
-  }
-
-  virtual void TearDown() { libaom_test::ClearSystemState(); }
-
- protected:
-  int bsize;
-  cdef_direction_func dering;
-  cdef_direction_func ref_dering;
-};
-
-typedef CDEFDeringDirTest CDEFDeringSpeedTest;
-
-void test_dering(int bsize, int iterations, cdef_direction_func dering,
-                 cdef_direction_func ref_dering) {
-  const int size = 8;
-  const int ysize = size + 2 * CDEF_VBORDER;
-  ACMRandom rnd(ACMRandom::DeterministicSeed());
-  DECLARE_ALIGNED(16, uint16_t, s[ysize * CDEF_BSTRIDE]);
-  DECLARE_ALIGNED(16, static uint16_t, d[size * size]);
-  DECLARE_ALIGNED(16, static uint16_t, ref_d[size * size]);
-  memset(ref_d, 0, sizeof(ref_d));
-  memset(d, 0, sizeof(d));
-
-  int error = 0, threshold = 0, dir;
-  int boundary, damping, depth, bits, level, count,
-      errdepth = 0, errthreshold = 0, errboundary = 0, errdamping = 0;
-  unsigned int pos = 0;
-
-  for (boundary = 0; boundary < 16; boundary++) {
-    for (depth = 8; depth <= 12; depth += 2) {
-      for (damping = 5 + depth - 8; damping < 7 + depth - 8; damping++) {
-        for (count = 0; count < iterations; count++) {
-          for (level = 0; level < (1 << depth) && !error;
-               level += (1 + 4 * !!boundary) << (depth - 8)) {
-            for (bits = 1; bits <= depth && !error; bits++) {
-              for (unsigned int i = 0; i < sizeof(s) / sizeof(*s); i++)
-                s[i] = clamp((rnd.Rand16() & ((1 << bits) - 1)) + level, 0,
-                             (1 << depth) - 1);
-              if (boundary) {
-                if (boundary & 1) {  // Left
-                  for (int i = 0; i < ysize; i++)
-                    for (int j = 0; j < CDEF_HBORDER; j++)
-                      s[i * CDEF_BSTRIDE + j] = CDEF_VERY_LARGE;
-                }
-                if (boundary & 2) {  // Right
-                  for (int i = 0; i < ysize; i++)
-                    for (int j = CDEF_HBORDER + size; j < CDEF_BSTRIDE; j++)
-                      s[i * CDEF_BSTRIDE + j] = CDEF_VERY_LARGE;
-                }
-                if (boundary & 4) {  // Above
-                  for (int i = 0; i < CDEF_VBORDER; i++)
-                    for (int j = 0; j < CDEF_BSTRIDE; j++)
-                      s[i * CDEF_BSTRIDE + j] = CDEF_VERY_LARGE;
-                }
-                if (boundary & 8) {  // Below
-                  for (int i = CDEF_VBORDER + size; i < ysize; i++)
-                    for (int j = 0; j < CDEF_BSTRIDE; j++)
-                      s[i * CDEF_BSTRIDE + j] = CDEF_VERY_LARGE;
-                }
-              }
-              for (dir = 0; dir < 8; dir++) {
-                for (threshold = 0; threshold < 64 << (depth - 8) && !error;
-                     threshold += (1 + 4 * !!boundary) << (depth - 8)) {
-                  ref_dering(ref_d, size,
-                             s + CDEF_HBORDER + CDEF_VBORDER * CDEF_BSTRIDE,
-                             threshold, dir, damping);
-                  // If dering and ref_dering are the same, we're just testing
-                  // speed
-                  if (dering != ref_dering)
-                    ASM_REGISTER_STATE_CHECK(dering(
-                        d, size, s + CDEF_HBORDER + CDEF_VBORDER * CDEF_BSTRIDE,
-                        threshold, dir, damping));
-                  if (ref_dering != dering) {
-                    for (pos = 0; pos < sizeof(d) / sizeof(*d) && !error;
-                         pos++) {
-                      error = ref_d[pos] != d[pos];
-                      errdepth = depth;
-                      errthreshold = threshold;
-                      errboundary = boundary;
-                      errdamping = damping;
-                    }
-                  }
-                }
-              }
-            }
-          }
-        }
-      }
-    }
-  }
-
-  pos--;
-  EXPECT_EQ(0, error) << "Error: CDEFDeringDirTest, SIMD and C mismatch."
-                      << std::endl
-                      << "First error at " << pos % size << "," << pos / size
-                      << " (" << (int16_t)ref_d[pos] << " : " << (int16_t)d[pos]
-                      << ") " << std::endl
-                      << "threshold: " << errthreshold << std::endl
-                      << "damping: " << errdamping << std::endl
-                      << "depth: " << errdepth << std::endl
-                      << "size: " << bsize << std::endl
-                      << "boundary: " << errboundary << std::endl
-                      << std::endl;
-}
-
-void test_dering_speed(int bsize, int iterations, cdef_direction_func dering,
-                       cdef_direction_func ref_dering) {
-  aom_usec_timer ref_timer;
-  aom_usec_timer timer;
-
-  aom_usec_timer_start(&ref_timer);
-  test_dering(bsize, iterations, ref_dering, ref_dering);
-  aom_usec_timer_mark(&ref_timer);
-  int ref_elapsed_time = (int)aom_usec_timer_elapsed(&ref_timer);
-
-  aom_usec_timer_start(&timer);
-  test_dering(bsize, iterations, dering, dering);
-  aom_usec_timer_mark(&timer);
-  int elapsed_time = (int)aom_usec_timer_elapsed(&timer);
-
-#if 0
-  std::cout << "[          ] C time = " << ref_elapsed_time / 1000
-            << " ms, SIMD time = " << elapsed_time / 1000 << " ms" << std::endl;
-#endif
-
-  EXPECT_GT(ref_elapsed_time, elapsed_time)
-      << "Error: CDEFDeringSpeedTest, SIMD slower than C." << std::endl
-      << "C time: " << ref_elapsed_time << " us" << std::endl
-      << "SIMD time: " << elapsed_time << " us" << std::endl;
-}
-
-typedef int (*find_dir_t)(const uint16_t *img, int stride, int32_t *var,
-                          int coeff_shift);
-
-typedef std::tr1::tuple<find_dir_t, find_dir_t> find_dir_param_t;
-
-class CDEFDeringFindDirTest
-    : public ::testing::TestWithParam<find_dir_param_t> {
- public:
-  virtual ~CDEFDeringFindDirTest() {}
-  virtual void SetUp() {
-    finddir = GET_PARAM(0);
-    ref_finddir = GET_PARAM(1);
-  }
-
-  virtual void TearDown() { libaom_test::ClearSystemState(); }
-
- protected:
-  find_dir_t finddir;
-  find_dir_t ref_finddir;
-};
-
-typedef CDEFDeringFindDirTest CDEFDeringFindDirSpeedTest;
-
-void test_finddir(int (*finddir)(const uint16_t *img, int stride, int32_t *var,
-                                 int coeff_shift),
-                  int (*ref_finddir)(const uint16_t *img, int stride,
-                                     int32_t *var, int coeff_shift)) {
-  const int size = 8;
-  ACMRandom rnd(ACMRandom::DeterministicSeed());
-  DECLARE_ALIGNED(16, uint16_t, s[size * size]);
-
-  int error = 0;
-  int depth, bits, level, count, errdepth = 0;
-  int ref_res = 0, res = 0;
-  int32_t ref_var = 0, var = 0;
-
-  for (depth = 8; depth <= 12 && !error; depth += 2) {
-    for (count = 0; count < 512 && !error; count++) {
-      for (level = 0; level < (1 << depth) && !error;
-           level += 1 << (depth - 8)) {
-        for (bits = 1; bits <= depth && !error; bits++) {
-          for (unsigned int i = 0; i < sizeof(s) / sizeof(*s); i++)
-            s[i] = clamp((rnd.Rand16() & ((1 << bits) - 1)) + level, 0,
-                         (1 << depth) - 1);
-          for (int c = 0; c < 1 + 9 * (finddir == ref_finddir); c++)
-            ref_res = ref_finddir(s, size, &ref_var, depth - 8);
-          if (finddir != ref_finddir)
-            ASM_REGISTER_STATE_CHECK(res = finddir(s, size, &var, depth - 8));
-          if (ref_finddir != finddir) {
-            if (res != ref_res || var != ref_var) error = 1;
-            errdepth = depth;
-          }
-        }
-      }
-    }
-  }
-
-  EXPECT_EQ(0, error) << "Error: CDEFDeringFindDirTest, SIMD and C mismatch."
-                      << std::endl
-                      << "return: " << res << " : " << ref_res << std::endl
-                      << "var: " << var << " : " << ref_var << std::endl
-                      << "depth: " << errdepth << std::endl
-                      << std::endl;
-}
-
-void test_finddir_speed(int (*finddir)(const uint16_t *img, int stride,
-                                       int32_t *var, int coeff_shift),
-                        int (*ref_finddir)(const uint16_t *img, int stride,
-                                           int32_t *var, int coeff_shift)) {
-  aom_usec_timer ref_timer;
-  aom_usec_timer timer;
-
-  aom_usec_timer_start(&ref_timer);
-  test_finddir(ref_finddir, ref_finddir);
-  aom_usec_timer_mark(&ref_timer);
-  int ref_elapsed_time = (int)aom_usec_timer_elapsed(&ref_timer);
-
-  aom_usec_timer_start(&timer);
-  test_finddir(finddir, finddir);
-  aom_usec_timer_mark(&timer);
-  int elapsed_time = (int)aom_usec_timer_elapsed(&timer);
-
-#if 0
-  std::cout << "[          ] C time = " << ref_elapsed_time / 1000
-            << " ms, SIMD time = " << elapsed_time / 1000 << " ms" << std::endl;
-#endif
-
-  EXPECT_GT(ref_elapsed_time, elapsed_time)
-      << "Error: CDEFDeringFindDirSpeedTest, SIMD slower than C." << std::endl
-      << "C time: " << ref_elapsed_time << " us" << std::endl
-      << "SIMD time: " << elapsed_time << " us" << std::endl;
-}
-
-TEST_P(CDEFDeringDirTest, TestSIMDNoMismatch) {
-  test_dering(bsize, 1, dering, ref_dering);
-}
-
-TEST_P(CDEFDeringSpeedTest, DISABLED_TestSpeed) {
-  test_dering_speed(bsize, 4, dering, ref_dering);
-}
-
-TEST_P(CDEFDeringFindDirTest, TestSIMDNoMismatch) {
-  test_finddir(finddir, ref_finddir);
-}
-
-TEST_P(CDEFDeringFindDirSpeedTest, DISABLED_TestSpeed) {
-  test_finddir_speed(finddir, ref_finddir);
-}
-
-using std::tr1::make_tuple;
-
-// VS compiling for 32 bit targets does not support vector types in
-// structs as arguments, which makes the v256 type of the intrinsics
-// hard to support, so optimizations for this target are disabled.
-#if defined(_WIN64) || !defined(_MSC_VER) || defined(__clang__)
-#if HAVE_SSE2
-INSTANTIATE_TEST_CASE_P(SSE2, CDEFDeringDirTest,
-                        ::testing::Values(make_tuple(&cdef_direction_4x4_sse2,
-                                                     &cdef_direction_4x4_c, 4),
-                                          make_tuple(&cdef_direction_8x8_sse2,
-                                                     &cdef_direction_8x8_c,
-                                                     8)));
-INSTANTIATE_TEST_CASE_P(SSE2, CDEFDeringFindDirTest,
-                        ::testing::Values(make_tuple(&cdef_find_dir_sse2,
-                                                     &cdef_find_dir_c)));
-#endif
-#if HAVE_SSSE3
-INSTANTIATE_TEST_CASE_P(SSSE3, CDEFDeringDirTest,
-                        ::testing::Values(make_tuple(&cdef_direction_4x4_ssse3,
-                                                     &cdef_direction_4x4_c, 4),
-                                          make_tuple(&cdef_direction_8x8_ssse3,
-                                                     &cdef_direction_8x8_c,
-                                                     8)));
-INSTANTIATE_TEST_CASE_P(SSSE3, CDEFDeringFindDirTest,
-                        ::testing::Values(make_tuple(&cdef_find_dir_ssse3,
-                                                     &cdef_find_dir_c)));
-#endif
-
-#if HAVE_SSE4_1
-INSTANTIATE_TEST_CASE_P(SSE4_1, CDEFDeringDirTest,
-                        ::testing::Values(make_tuple(&cdef_direction_4x4_sse4_1,
-                                                     &cdef_direction_4x4_c, 4),
-                                          make_tuple(&cdef_direction_8x8_sse4_1,
-                                                     &cdef_direction_8x8_c,
-                                                     8)));
-INSTANTIATE_TEST_CASE_P(SSE4_1, CDEFDeringFindDirTest,
-                        ::testing::Values(make_tuple(&cdef_find_dir_sse4_1,
-                                                     &cdef_find_dir_c)));
-#endif
-
-#if HAVE_NEON
-INSTANTIATE_TEST_CASE_P(NEON, CDEFDeringDirTest,
-                        ::testing::Values(make_tuple(&cdef_direction_4x4_neon,
-                                                     &cdef_direction_4x4_c, 4),
-                                          make_tuple(&cdef_direction_8x8_neon,
-                                                     &cdef_direction_8x8_c,
-                                                     8)));
-INSTANTIATE_TEST_CASE_P(NEON, CDEFDeringFindDirTest,
-                        ::testing::Values(make_tuple(&cdef_find_dir_neon,
-                                                     &cdef_find_dir_c)));
-#endif
-
-// Test speed for all supported architectures
-#if HAVE_SSE2
-INSTANTIATE_TEST_CASE_P(SSE2, CDEFDeringSpeedTest,
-                        ::testing::Values(make_tuple(&cdef_direction_4x4_sse2,
-                                                     &cdef_direction_4x4_c, 4),
-                                          make_tuple(&cdef_direction_8x8_sse2,
-                                                     &cdef_direction_8x8_c,
-                                                     8)));
-INSTANTIATE_TEST_CASE_P(SSE2, CDEFDeringFindDirSpeedTest,
-                        ::testing::Values(make_tuple(&cdef_find_dir_sse2,
-                                                     &cdef_find_dir_c)));
-#endif
-
-#if HAVE_SSSE3
-INSTANTIATE_TEST_CASE_P(SSSE3, CDEFDeringSpeedTest,
-                        ::testing::Values(make_tuple(&cdef_direction_4x4_ssse3,
-                                                     &cdef_direction_4x4_c, 4),
-                                          make_tuple(&cdef_direction_8x8_ssse3,
-                                                     &cdef_direction_8x8_c,
-                                                     8)));
-INSTANTIATE_TEST_CASE_P(SSSE3, CDEFDeringFindDirSpeedTest,
-                        ::testing::Values(make_tuple(&cdef_find_dir_ssse3,
-                                                     &cdef_find_dir_c)));
-#endif
-
-#if HAVE_SSE4_1
-INSTANTIATE_TEST_CASE_P(SSE4_1, CDEFDeringSpeedTest,
-                        ::testing::Values(make_tuple(&cdef_direction_4x4_sse4_1,
-                                                     &cdef_direction_4x4_c, 4),
-                                          make_tuple(&cdef_direction_8x8_sse4_1,
-                                                     &cdef_direction_8x8_c,
-                                                     8)));
-INSTANTIATE_TEST_CASE_P(SSE4_1, CDEFDeringFindDirSpeedTest,
-                        ::testing::Values(make_tuple(&cdef_find_dir_sse4_1,
-                                                     &cdef_find_dir_c)));
-#endif
-
-#if HAVE_NEON
-INSTANTIATE_TEST_CASE_P(NEON, CDEFDeringSpeedTest,
-                        ::testing::Values(make_tuple(&cdef_direction_4x4_neon,
-                                                     &cdef_direction_4x4_c, 4),
-                                          make_tuple(&cdef_direction_8x8_neon,
-                                                     &cdef_direction_8x8_c,
-                                                     8)));
-INSTANTIATE_TEST_CASE_P(NEON, CDEFDeringFindDirSpeedTest,
-                        ::testing::Values(make_tuple(&cdef_find_dir_neon,
-                                                     &cdef_find_dir_c)));
-#endif
-
-#endif  // defined(_WIN64) || !defined(_MSC_VER)
-}  // namespace

third_party/aom/test/dr_prediction_test.cc

@@ -0,0 +1,359 @@
+/*
+ * Copyright (c) 2018, Alliance for Open Media. All rights reserved
+ *
+ * This source code is subject to the terms of the BSD 2 Clause License and
+ * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
+ * was not distributed with this source code in the LICENSE file, you can
+ * obtain it at www.aomedia.org/license/software. If the Alliance for Open
+ * Media Patent License 1.0 was not distributed with this source code in the
+ * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
+ */
+#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
+
+#include "config/aom_config.h"
+#include "config/aom_dsp_rtcd.h"
+
+#include "aom_mem/aom_mem.h"
+#include "aom_ports/aom_timer.h"
+#include "av1/common/blockd.h"
+#include "av1/common/pred_common.h"
+#include "av1/common/reconintra.h"
+#include "test/acm_random.h"
+#include "test/clear_system_state.h"
+#include "test/register_state_check.h"
+#include "test/util.h"
+
+namespace {
+
+const int kZ1Start = 0;
+const int kZ2Start = 90;
+const int kZ3Start = 180;
+
+const TX_SIZE kTxSize[] = { TX_4X4,   TX_8X8,   TX_16X16, TX_32X32, TX_64X64,
+                            TX_4X8,   TX_8X4,   TX_8X16,  TX_16X8,  TX_16X32,
+                            TX_32X16, TX_32X64, TX_64X32, TX_4X16,  TX_16X4,
+                            TX_8X32,  TX_32X8,  TX_16X64, TX_64X16 };
+
+const char *const kTxSizeStrings[] = {
+  "TX_4X4",   "TX_8X8",   "TX_16X16", "TX_32X32", "TX_64X64",
+  "TX_4X8",   "TX_8X4",   "TX_8X16",  "TX_16X8",  "TX_16X32",
+  "TX_32X16", "TX_32X64", "TX_64X32", "TX_4X16",  "TX_16X4",
+  "TX_8X32",  "TX_32X8",  "TX_16X64", "TX_64X16"
+};
+
+using libaom_test::ACMRandom;
+
+typedef void (*DrPred_Hbd)(uint16_t *dst, ptrdiff_t stride, int bw, int bh,
+                           const uint16_t *above, const uint16_t *left,
+                           int upsample_above, int upsample_left, int dx,
+                           int dy, int bd);
+
+typedef void (*DrPred)(uint8_t *dst, ptrdiff_t stride, int bw, int bh,
+                       const uint8_t *above, const uint8_t *left,
+                       int upsample_above, int upsample_left, int dx, int dy,
+                       int bd);
+
+typedef void (*Z1_Lbd)(uint8_t *dst, ptrdiff_t stride, int bw, int bh,
+                       const uint8_t *above, const uint8_t *left,
+                       int upsample_above, int dx, int dy);
+template <Z1_Lbd fn>
+void z1_wrapper(uint8_t *dst, ptrdiff_t stride, int bw, int bh,
+                const uint8_t *above, const uint8_t *left, int upsample_above,
+                int /*upsample_left*/, int dx, int dy, int /*bd*/) {
+  fn(dst, stride, bw, bh, above, left, upsample_above, dx, dy);
+}
+
+typedef void (*Z2_Lbd)(uint8_t *dst, ptrdiff_t stride, int bw, int bh,
+                       const uint8_t *above, const uint8_t *left,
+                       int upsample_above, int upsample_left, int dx, int dy);
+template <Z2_Lbd fn>
+void z2_wrapper(uint8_t *dst, ptrdiff_t stride, int bw, int bh,
+                const uint8_t *above, const uint8_t *left, int upsample_above,
+                int upsample_left, int dx, int dy, int /*bd*/) {
+  fn(dst, stride, bw, bh, above, left, upsample_above, upsample_left, dx, dy);
+}
+
+typedef void (*Z3_Lbd)(uint8_t *dst, ptrdiff_t stride, int bw, int bh,
+                       const uint8_t *above, const uint8_t *left,
+                       int upsample_left, int dx, int dy);
+template <Z3_Lbd fn>
+void z3_wrapper(uint8_t *dst, ptrdiff_t stride, int bw, int bh,
+                const uint8_t *above, const uint8_t *left,
+                int /*upsample_above*/, int upsample_left, int dx, int dy,
+                int /*bd*/) {
+  fn(dst, stride, bw, bh, above, left, upsample_left, dx, dy);
+}
+
+typedef void (*Z1_Hbd)(uint16_t *dst, ptrdiff_t stride, int bw, int bh,
+                       const uint16_t *above, const uint16_t *left,
+                       int upsample_above, int dx, int dy, int bd);
+template <Z1_Hbd fn>
+void z1_wrapper_hbd(uint16_t *dst, ptrdiff_t stride, int bw, int bh,
+                    const uint16_t *above, const uint16_t *left,
+                    int upsample_above, int /*upsample_left*/, int dx, int dy,
+                    int bd) {
+  fn(dst, stride, bw, bh, above, left, upsample_above, dx, dy, bd);
+}
+
+typedef void (*Z2_Hbd)(uint16_t *dst, ptrdiff_t stride, int bw, int bh,
+                       const uint16_t *above, const uint16_t *left,
+                       int upsample_above, int upsample_left, int dx, int dy,
+                       int bd);
+template <Z2_Hbd fn>
+void z2_wrapper_hbd(uint16_t *dst, ptrdiff_t stride, int bw, int bh,
+                    const uint16_t *above, const uint16_t *left,
+                    int upsample_above, int upsample_left, int dx, int dy,
+                    int bd) {
+  fn(dst, stride, bw, bh, above, left, upsample_above, upsample_left, dx, dy,
+     bd);
+}
+
+typedef void (*Z3_Hbd)(uint16_t *dst, ptrdiff_t stride, int bw, int bh,
+                       const uint16_t *above, const uint16_t *left,
+                       int upsample_left, int dx, int dy, int bd);
+template <Z3_Hbd fn>
+void z3_wrapper_hbd(uint16_t *dst, ptrdiff_t stride, int bw, int bh,
+                    const uint16_t *above, const uint16_t *left,
+                    int /*upsample_above*/, int upsample_left, int dx, int dy,
+                    int bd) {
+  fn(dst, stride, bw, bh, above, left, upsample_left, dx, dy, bd);
+}
+
+template <typename FuncType>
+struct DrPredFunc {
+  DrPredFunc(FuncType pred = NULL, FuncType tst = NULL, int bit_depth_value = 0,
+             int start_angle_value = 0)
+      : ref_fn(pred), tst_fn(tst), bit_depth(bit_depth_value),
+        start_angle(start_angle_value) {}
+
+  FuncType ref_fn;
+  FuncType tst_fn;
+  int bit_depth;
+  int start_angle;
+};
+
+template <typename Pixel, typename FuncType>
+class DrPredTest : public ::testing::TestWithParam<DrPredFunc<FuncType> > {
+ protected:
+  static const int kMaxNumTests = 100000;
+  static const int kIterations = 10;
+  static const int kDstStride = 64;
+  static const int kDstSize = kDstStride * kDstStride;
+  static const int kOffset = 16;
+  static const int kBufSize = ((2 * MAX_TX_SIZE) << 1) + 16;
+
+  DrPredTest()
+      : upsample_above_(0), upsample_left_(0), bw_(0), bh_(0), dx_(1), dy_(1),
+        bd_(8), txsize_(TX_4X4) {
+    params_ = this->GetParam();
+    start_angle_ = params_.start_angle;
+    stop_angle_ = start_angle_ + 90;
+
+    dst_ref_ = &dst_ref_data_[0];
+    dst_tst_ = &dst_tst_data_[0];
+    dst_stride_ = kDstStride;
+    above_ = &above_data_[kOffset];
+    left_ = &left_data_[kOffset];
+
+    for (int i = 0; i < kBufSize; ++i) {
+      above_data_[i] = rng_.Rand8();
+      left_data_[i] = rng_.Rand8();
+    }
+
+    for (int i = 0; i < kDstSize; ++i) {
+      dst_ref_[i] = 0;
+    }
+  }
+
+  virtual ~DrPredTest() {}
+
+  void Predict(bool speedtest, int tx) {
+    const int kNumTests = speedtest ? kMaxNumTests : 1;
+    aom_usec_timer timer;
+
+    aom_usec_timer_start(&timer);
+    for (int k = 0; k < kNumTests; ++k) {
+      params_.ref_fn(dst_ref_, dst_stride_, bw_, bh_, above_, left_,
+                     upsample_above_, upsample_left_, dx_, dy_, bd_);
+    }
+    aom_usec_timer_mark(&timer);
+    const int ref_time = static_cast<int>(aom_usec_timer_elapsed(&timer));
+
+    aom_usec_timer_start(&timer);
+    if (params_.tst_fn) {
+      for (int k = 0; k < kNumTests; ++k) {
+        ASM_REGISTER_STATE_CHECK(params_.tst_fn(dst_tst_, dst_stride_, bw_, bh_,
+                                                above_, left_, upsample_above_,
+                                                upsample_left_, dx_, dy_, bd_));
+      }
+    }
+    aom_usec_timer_mark(&timer);
+    const int tst_time = static_cast<int>(aom_usec_timer_elapsed(&timer));
+
+    OutputTimes(kNumTests, ref_time, tst_time, tx);
+  }
+
+  void RunTest(bool speedtest) {
+    for (int i = 0; i < kBufSize; ++i) {
+      above_data_[i] = left_data_[i] = (1 << bd_) - 1;
+    }
+
+    for (int tx = 0; tx < TX_SIZES_ALL; ++tx) {
+      if (params_.tst_fn == NULL) {
+        for (int i = 0; i < kDstSize; ++i) {
+          dst_tst_[i] = (1 << bd_) - 1;
+        }
+      } else {
+        for (int i = 0; i < kDstSize; ++i) {
+          dst_tst_[i] = 0;
+        }
+      }
+
+      bw_ = tx_size_wide[kTxSize[tx]];
+      bh_ = tx_size_high[kTxSize[tx]];
+
+      Predict(speedtest, tx);
+
+      for (int r = 0; r < bh_; ++r) {
+        for (int c = 0; c < bw_; ++c) {
+          ASSERT_EQ(dst_ref_[r * dst_stride_ + c],
+                    dst_tst_[r * dst_stride_ + c])
+              << bw_ << "x" << bh_ << " r: " << r << " c: " << c
+              << " dx: " << dx_ << " dy: " << dy_
+              << " upsample_above: " << upsample_above_
+              << " upsample_left: " << upsample_left_;
+        }
+      }
+    }
+  }
+
+  void OutputTimes(int num_tests, int ref_time, int tst_time, int tx) {
+    if (num_tests > 1) {
+      if (params_.tst_fn) {
+        const float x = static_cast<float>(ref_time) / tst_time;
+        printf("\t[%8s] :: ref time %6d, tst time %6d     %3.2f\n",
+               kTxSizeStrings[tx], ref_time, tst_time, x);
+      } else {
+        printf("\t[%8s] :: ref time %6d\n", kTxSizeStrings[tx], ref_time);
+      }
+    }
+  }
+
+  Pixel dst_ref_data_[kDstSize];
+  Pixel dst_tst_data_[kDstSize];
+
+  Pixel left_data_[kBufSize];
+  Pixel dummy_data_[kBufSize];
+  Pixel above_data_[kBufSize];
+
+  Pixel *dst_ref_;
+  Pixel *dst_tst_;
+  Pixel *above_;
+  Pixel *left_;
+  int dst_stride_;
+
+  int upsample_above_;
+  int upsample_left_;
+  int bw_;
+  int bh_;
+  int dx_;
+  int dy_;
+  int bd_;
+  TX_SIZE txsize_;
+
+  int start_angle_;
+  int stop_angle_;
+
+  ACMRandom rng_;
+
+  DrPredFunc<FuncType> params_;
+};
+
+class LowbdDrPredTest : public DrPredTest<uint8_t, DrPred> {};
+
+TEST_P(LowbdDrPredTest, SaturatedValues) {
+  for (int iter = 0; iter < kIterations && !HasFatalFailure(); ++iter) {
+    upsample_above_ = iter & 1;
+    for (int angle = start_angle_; angle < stop_angle_; ++angle) {
+      dx_ = av1_get_dx(angle);
+      dy_ = av1_get_dy(angle);
+      if (dx_ && dy_) RunTest(false);
+    }
+  }
+}
+
+TEST_P(LowbdDrPredTest, DISABLED_Speed) {
+  const int angles[] = { 3, 45, 87 };
+  for (upsample_above_ = 0; upsample_above_ < 2; ++upsample_above_) {
+    upsample_left_ = upsample_above_;
+    for (int i = 0; i < 3; ++i) {
+      dx_ = av1_get_dx(angles[i] + start_angle_);
+      dy_ = av1_get_dy(angles[i] + start_angle_);
+      printf("upsample_above: %d upsample_left: %d angle: %d ~~~~~~~~~~~~~~~\n",
+             upsample_above_, upsample_left_, angles[i] + start_angle_);
+      if (dx_ && dy_) RunTest(true);
+    }
+  }
+}
+
+using ::testing::make_tuple;
+
+INSTANTIATE_TEST_CASE_P(
+    C, LowbdDrPredTest,
+    ::testing::Values(DrPredFunc<DrPred>(&z1_wrapper<av1_dr_prediction_z1_c>,
+                                         NULL, AOM_BITS_8, kZ1Start),
+                      DrPredFunc<DrPred>(&z2_wrapper<av1_dr_prediction_z2_c>,
+                                         NULL, AOM_BITS_8, kZ2Start),
+                      DrPredFunc<DrPred>(&z3_wrapper<av1_dr_prediction_z3_c>,
+                                         NULL, AOM_BITS_8, kZ3Start)));
+
+class HighbdDrPredTest : public DrPredTest<uint16_t, DrPred_Hbd> {};
+
+TEST_P(HighbdDrPredTest, SaturatedValues) {
+  for (int iter = 0; iter < kIterations && !HasFatalFailure(); ++iter) {
+    upsample_above_ = iter & 1;
+    for (int angle = start_angle_; angle < stop_angle_; ++angle) {
+      dx_ = av1_get_dx(angle);
+      dy_ = av1_get_dy(angle);
+      if (dx_ && dy_) RunTest(false);
+    }
+  }
+}
+
+TEST_P(HighbdDrPredTest, DISABLED_Speed) {
+  const int angles[] = { 3, 45, 87 };
+  for (upsample_above_ = 0; upsample_above_ < 2; ++upsample_above_) {
+    upsample_left_ = upsample_above_;
+    for (int i = 0; i < 3; ++i) {
+      dx_ = av1_get_dx(angles[i] + start_angle_);
+      dy_ = av1_get_dy(angles[i] + start_angle_);
+      printf("upsample_above: %d upsample_left: %d angle: %d ~~~~~~~~~~~~~~~\n",
+             upsample_above_, upsample_left_, angles[i] + start_angle_);
+      if (dx_ && dy_) RunTest(true);
+    }
+  }
+}
+
+INSTANTIATE_TEST_CASE_P(
+    C, HighbdDrPredTest,
+    ::testing::Values(
+        DrPredFunc<DrPred_Hbd>(&z1_wrapper_hbd<av1_highbd_dr_prediction_z1_c>,
+                               NULL, AOM_BITS_8, kZ1Start),
+        DrPredFunc<DrPred_Hbd>(&z1_wrapper_hbd<av1_highbd_dr_prediction_z1_c>,
+                               NULL, AOM_BITS_10, kZ1Start),
+        DrPredFunc<DrPred_Hbd>(&z1_wrapper_hbd<av1_highbd_dr_prediction_z1_c>,
+                               NULL, AOM_BITS_12, kZ1Start),
+        DrPredFunc<DrPred_Hbd>(&z2_wrapper_hbd<av1_highbd_dr_prediction_z2_c>,
+                               NULL, AOM_BITS_8, kZ2Start),
+        DrPredFunc<DrPred_Hbd>(&z2_wrapper_hbd<av1_highbd_dr_prediction_z2_c>,
+                               NULL, AOM_BITS_10, kZ2Start),
+        DrPredFunc<DrPred_Hbd>(&z2_wrapper_hbd<av1_highbd_dr_prediction_z2_c>,
+                               NULL, AOM_BITS_12, kZ2Start),
+        DrPredFunc<DrPred_Hbd>(&z3_wrapper_hbd<av1_highbd_dr_prediction_z3_c>,
+                               NULL, AOM_BITS_8, kZ3Start),
+        DrPredFunc<DrPred_Hbd>(&z3_wrapper_hbd<av1_highbd_dr_prediction_z3_c>,
+                               NULL, AOM_BITS_10, kZ3Start),
+        DrPredFunc<DrPred_Hbd>(&z3_wrapper_hbd<av1_highbd_dr_prediction_z3_c>,
+                               NULL, AOM_BITS_12, kZ3Start)));
+
+}  // namespace

third_party/aom/test/dump_obu.sh

@@ -0,0 +1,70 @@
+#!/bin/sh
+## Copyright (c) 2018, Alliance for Open Media. All rights reserved
+##
+## This source code is subject to the terms of the BSD 2 Clause License and
+## the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
+## was not distributed with this source code in the LICENSE file, you can
+## obtain it at www.aomedia.org/license/software. If the Alliance for Open
+## Media Patent License 1.0 was not distributed with this source code in the
+## PATENTS file, you can obtain it at www.aomedia.org/license/patent.
+##
+## This file tests the libaom dump_obu tool. To add new tests to this
+## file, do the following:
+##   1. Write a shell function (this is your test).
+##   2. Add the function to dump_obu_tests (on a new line).
+##
+. $(dirname $0)/tools_common.sh
+
+readonly dump_obu_test_file="${AOM_TEST_OUTPUT_DIR}/av1_obu_test.ivf"
+
+dump_obu_verify_environment() {
+  if [ ! -e "${YUV_RAW_INPUT}" ]; then
+    elog "The file ${YUV_RAW_INPUT##*/} must exist in LIBAOM_TEST_DATA_PATH."
+    return 1
+  fi
+  if [ "$(dump_obu_available)" = "yes" ]; then
+    if [ -z "$(aom_tool_path dump_obu)" ]; then
+      elog "dump_obu not found in LIBAOM_BIN_PATH, its parent, or child tools/."
+    fi
+  fi
+}
+
+dump_obu_available() {
+  if [ "$(av1_decode_available)" = "yes" ] && \
+     [ "$(av1_encode_available)" = "yes" ]; then
+    echo yes
+  fi
+}
+
+aomenc_available() {
+  if [ -x "$(aom_tool_path aomenc)" ]; then
+    echo yes
+  fi
+}
+
+encode_test_file() {
+  if [ "$(aomenc_available)" = "yes" ]; then
+    local readonly encoder="$(aom_tool_path aomenc)"
+
+    eval "${encoder}" \
+      $(aomenc_encode_test_fast_params) \
+      $(yuv_raw_input) \
+      --ivf \
+      --output=${dump_obu_test_file} \
+      ${devnull}
+
+    if [ ! -e "${dump_obu_test_file}" ]; then
+      elog "dump_obu test input encode failed."
+      return 1
+    fi
+  fi
+}
+
+dump_obu() {
+  encode_test_file
+  eval $(aom_tool_path dump_obu) "${dump_obu_test_file}" ${devnull}
+}
+
+dump_obu_tests="dump_obu"
+
+run_tests dump_obu_verify_environment "${dump_obu_tests}"

third_party/aom/test/ec_test.cc

@@ -0,0 +1,159 @@
+/*
+ * Copyright (c) 2017, Alliance for Open Media. All rights reserved
+ *
+ * This source code is subject to the terms of the BSD 2 Clause License and
+ * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
+ * was not distributed with this source code in the LICENSE file, you can
+ * obtain it at www.aomedia.org/license/software. If the Alliance for Open
+ * Media Patent License 1.0 was not distributed with this source code in the
+ * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
+ */
+
+#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
+
+#include <cstdlib>
+
+#include "aom_dsp/entenc.h"
+#include "aom_dsp/entdec.h"
+
+TEST(EC_TEST, random_ec_test) {
+  od_ec_enc enc;
+  od_ec_dec dec;
+  int sz;
+  int i;
+  int ret;
+  unsigned int sym;
+  unsigned int seed;
+  unsigned char *ptr;
+  uint32_t ptr_sz;
+  char *seed_str;
+  ret = 0;
+  seed_str = getenv("EC_TEST_SEED");
+  if (seed_str) {
+    seed = atoi(seed_str);
+  } else {
+    seed = 0xdaa1a;
+  }
+  srand(seed);
+  od_ec_enc_init(&enc, 1);
+  /*Test compatibility between multiple different encode/decode routines.*/
+  for (i = 0; i < 409600; i++) {
+    unsigned *fz;
+    unsigned *fts;
+    unsigned *data;
+    unsigned *tell;
+    unsigned *enc_method;
+    int j;
+    sz = rand() / ((RAND_MAX >> (rand() % 9U)) + 1U);
+    fz = (unsigned *)malloc(sz * sizeof(*fz));
+    fts = (unsigned *)malloc(sz * sizeof(*fts));
+    data = (unsigned *)malloc(sz * sizeof(*data));
+    tell = (unsigned *)malloc((sz + 1) * sizeof(*tell));
+    enc_method = (unsigned *)malloc(sz * sizeof(*enc_method));
+    od_ec_enc_reset(&enc);
+    tell[0] = od_ec_enc_tell_frac(&enc);
+    for (j = 0; j < sz; j++) {
+      data[j] = rand() / ((RAND_MAX >> 1) + 1);
+
+      fts[j] = CDF_PROB_BITS;
+      fz[j] = (rand() % (CDF_PROB_TOP - 2)) >> (CDF_PROB_BITS - fts[j]);
+      fz[j] = OD_MAXI(fz[j], 1);
+      enc_method[j] = 3 + (rand() & 1);
+      switch (enc_method[j]) {
+        case 3: {
+          od_ec_encode_bool_q15(&enc, data[j],
+                                OD_ICDF(fz[j] << (CDF_PROB_BITS - fts[j])));
+          break;
+        }
+        case 4: {
+          uint16_t cdf[2];
+          cdf[0] = OD_ICDF(fz[j]);
+          cdf[1] = OD_ICDF(1U << fts[j]);
+          od_ec_encode_cdf_q15(&enc, data[j], cdf, 2);
+          break;
+        }
+      }
+
+      tell[j + 1] = od_ec_enc_tell_frac(&enc);
+    }
+    ptr = od_ec_enc_done(&enc, &ptr_sz);
+    EXPECT_GE(((od_ec_enc_tell(&enc) + 7U) >> 3), ptr_sz)
+        << "od_ec_enc_tell() lied: "
+           "there's "
+        << ptr_sz << " bytes instead of " << ((od_ec_enc_tell(&enc) + 7) >> 3)
+        << " (Random seed: " << seed << ")\n";
+    od_ec_dec_init(&dec, ptr, ptr_sz);
+    EXPECT_EQ(od_ec_dec_tell_frac(&dec), tell[0])
+        << "od_ec_dec_tell() mismatch between encoder and decoder "
+           "at symbol 0: "
+        << (unsigned)od_ec_dec_tell_frac(&dec) << " instead of " << tell[0]
+        << " (Random seed: " << seed << ").\n";
+    for (j = 0; j < sz; j++) {
+      int dec_method;
+      if (CDF_SHIFT == 0) {
+        dec_method = 3 + (rand() & 1);
+      } else {
+        dec_method = enc_method[j];
+      }
+      switch (dec_method) {
+        case 3: {
+          sym = od_ec_decode_bool_q15(
+              &dec, OD_ICDF(fz[j] << (CDF_PROB_BITS - fts[j])));
+          break;
+        }
+        case 4: {
+          uint16_t cdf[2];
+          cdf[0] = OD_ICDF(fz[j]);
+          cdf[1] = OD_ICDF(1U << fts[j]);
+          sym = od_ec_decode_cdf_q15(&dec, cdf, 2);
+          break;
+        }
+      }
+
+      EXPECT_EQ(sym, data[j])
+          << "Decoded " << sym << " instead of " << data[j]
+          << " with fz=" << fz[j] << " and ftb=" << fts[j] << "at position "
+          << j << " of " << sz << " (Random seed: " << seed << ").\n"
+          << "Encoding method: " << enc_method[j]
+          << " decoding method: " << dec_method << "\n";
+      EXPECT_EQ(od_ec_dec_tell_frac(&dec), tell[j + 1])
+          << "od_ec_dec_tell() mismatch between encoder and "
+             "decoder at symbol "
+          << j + 1 << ": " << (unsigned)od_ec_dec_tell_frac(&dec)
+          << " instead of " << tell[j + 1] << " (Random seed: " << seed
+          << ").\n";
+    }
+    free(enc_method);
+    free(tell);
+    free(data);
+    free(fts);
+    free(fz);
+  }
+  od_ec_enc_reset(&enc);
+  if (CDF_SHIFT == 0) {
+    od_ec_encode_bool_q15(&enc, 0, OD_ICDF(16384));
+    od_ec_encode_bool_q15(&enc, 0, OD_ICDF(16384));
+    od_ec_encode_bool_q15(&enc, 0, OD_ICDF(16384));
+    od_ec_encode_bool_q15(&enc, 0, OD_ICDF(16384));
+    od_ec_encode_bool_q15(&enc, 0, OD_ICDF(24576));
+    od_ec_enc_patch_initial_bits(&enc, 3, 2);
+    EXPECT_FALSE(enc.error) << "od_ec_enc_patch_initial_bits() failed.\n";
+    od_ec_enc_patch_initial_bits(&enc, 0, 5);
+    EXPECT_TRUE(enc.error)
+        << "od_ec_enc_patch_initial_bits() didn't fail when it should have.\n";
+    od_ec_enc_reset(&enc);
+    od_ec_encode_bool_q15(&enc, 0, OD_ICDF(16384));
+    od_ec_encode_bool_q15(&enc, 0, OD_ICDF(16384));
+    od_ec_encode_bool_q15(&enc, 1, OD_ICDF(32256));
+    od_ec_encode_bool_q15(&enc, 0, OD_ICDF(24576));
+    od_ec_enc_patch_initial_bits(&enc, 0, 2);
+    EXPECT_FALSE(enc.error) << "od_ec_enc_patch_initial_bits() failed.\n";
+    ptr = od_ec_enc_done(&enc, &ptr_sz);
+    EXPECT_EQ(ptr_sz, 2u);
+    EXPECT_EQ(ptr[0], 63)
+        << "Got " << ptr[0]
+        << " when expecting 63 for od_ec_enc_patch_initial_bits().\n";
+  }
+  od_ec_enc_clear(&enc);
+  EXPECT_EQ(ret, 0);
+}

third_party/aom/test/encode_api_test.cc

@@ -7,11 +7,12 @@
  * obtain it at www.aomedia.org/license/software. If the Alliance for Open
  * Media Patent License 1.0 was not distributed with this source code in the
  * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
+ */
 
 #include "third_party/googletest/src/googletest/include/gtest/gtest.h"
 
-#include "./aom_config.h"
+#include "config/aom_config.h"
+
 #include "test/util.h"
 #include "aom/aomcx.h"
 #include "aom/aom_encoder.h"
@@ -33,8 +34,8 @@ TEST(EncodeAPI, InvalidParams) {
 
   EXPECT_EQ(AOM_CODEC_INVALID_PARAM, aom_codec_enc_init(NULL, NULL, NULL, 0));
   EXPECT_EQ(AOM_CODEC_INVALID_PARAM, aom_codec_enc_init(&enc, NULL, NULL, 0));
-  EXPECT_EQ(AOM_CODEC_INVALID_PARAM, aom_codec_encode(NULL, NULL, 0, 0, 0, 0));
-  EXPECT_EQ(AOM_CODEC_INVALID_PARAM, aom_codec_encode(NULL, &img, 0, 0, 0, 0));
+  EXPECT_EQ(AOM_CODEC_INVALID_PARAM, aom_codec_encode(NULL, NULL, 0, 0, 0));
+  EXPECT_EQ(AOM_CODEC_INVALID_PARAM, aom_codec_encode(NULL, &img, 0, 0, 0));
   EXPECT_EQ(AOM_CODEC_INVALID_PARAM, aom_codec_destroy(NULL));
   EXPECT_EQ(AOM_CODEC_INVALID_PARAM,
             aom_codec_enc_config_default(NULL, NULL, 0));
@@ -53,7 +54,7 @@ TEST(EncodeAPI, InvalidParams) {
 
     EXPECT_EQ(AOM_CODEC_OK, aom_codec_enc_config_default(kCodecs[i], &cfg, 0));
     EXPECT_EQ(AOM_CODEC_OK, aom_codec_enc_init(&enc, kCodecs[i], &cfg, 0));
-    EXPECT_EQ(AOM_CODEC_OK, aom_codec_encode(&enc, NULL, 0, 0, 0, 0));
+    EXPECT_EQ(AOM_CODEC_OK, aom_codec_encode(&enc, NULL, 0, 0, 0));
 
     EXPECT_EQ(AOM_CODEC_OK, aom_codec_destroy(&enc));
   }

third_party/aom/test/encode_perf_test.cc

@@ -7,12 +7,14 @@
  * obtain it at www.aomedia.org/license/software. If the Alliance for Open
  * Media Patent License 1.0 was not distributed with this source code in the
  * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
+ */
 
 #include <string>
 #include "third_party/googletest/src/googletest/include/gtest/gtest.h"
-#include "./aom_config.h"
-#include "./aom_version.h"
+
+#include "config/aom_config.h"
+#include "config/aom_version.h"
+
 #include "test/codec_factory.h"
 #include "test/encode_test_driver.h"
 #include "test/i420_video_source.h"

third_party/aom/test/encode_test_driver.cc

@@ -7,13 +7,14 @@
  * obtain it at www.aomedia.org/license/software. If the Alliance for Open
  * Media Patent License 1.0 was not distributed with this source code in the
  * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
+ */
 
 #include <string>
 
 #include "third_party/googletest/src/googletest/include/gtest/gtest.h"
 
-#include "./aom_config.h"
+#include "config/aom_config.h"
+
 #include "aom_ports/mem.h"
 #include "test/codec_factory.h"
 #include "test/decode_test_driver.h"
@@ -34,21 +35,6 @@ void Encoder::InitEncoder(VideoSource *video) {
 
     res = aom_codec_enc_init(&encoder_, CodecInterface(), &cfg_, init_flags_);
     ASSERT_EQ(AOM_CODEC_OK, res) << EncoderError();
-
-#if CONFIG_AV1_ENCODER
-    if (CodecInterface() == &aom_codec_av1_cx_algo) {
-// Default to 1 tile column for AV1. With CONFIG_EXT_TILE, the
-// default is already the largest possible tile size
-#if !CONFIG_EXT_TILE
-      const int log2_tile_columns = 0;
-      res = aom_codec_control_(&encoder_, AV1E_SET_TILE_COLUMNS,
-                               log2_tile_columns);
-      ASSERT_EQ(AOM_CODEC_OK, res) << EncoderError();
-#endif  // !CONFIG_EXT_TILE
-    } else
-#endif
-    {
-    }
   }
 }
 
@@ -82,15 +68,14 @@ void Encoder::EncodeFrameInternal(const VideoSource &video,
   }
 
   // Encode the frame
-  API_REGISTER_STATE_CHECK(res = aom_codec_encode(&encoder_, img, video.pts(),
-                                                  video.duration(), frame_flags,
-                                                  deadline_));
+  API_REGISTER_STATE_CHECK(res =
+                               aom_codec_encode(&encoder_, img, video.pts(),
+                                                video.duration(), frame_flags));
   ASSERT_EQ(AOM_CODEC_OK, res) << EncoderError();
 }
 
 void Encoder::Flush() {
-  const aom_codec_err_t res =
-      aom_codec_encode(&encoder_, NULL, 0, 0, 0, deadline_);
+  const aom_codec_err_t res = aom_codec_encode(&encoder_, NULL, 0, 0, 0);
   if (!encoder_.priv)
     ASSERT_EQ(AOM_CODEC_ERROR, res) << EncoderError();
   else
@@ -105,11 +90,8 @@ void EncoderTest::InitializeConfig() {
 void EncoderTest::SetMode(TestMode mode) {
   switch (mode) {
     case kOnePassGood:
-    case kTwoPassGood: deadline_ = AOM_DL_GOOD_QUALITY; break;
-    case kRealTime:
-      deadline_ = AOM_DL_GOOD_QUALITY;
-      cfg_.g_lag_in_frames = 0;
-      break;
+    case kTwoPassGood: break;
+    case kRealTime: cfg_.g_lag_in_frames = 0; break;
     default: ASSERT_TRUE(false) << "Unexpected mode " << mode;
   }
   mode_ = mode;
@@ -149,14 +131,16 @@ static bool compare_img(const aom_image_t *img1, const aom_image_t *img2,
                         int *const mismatch_row, int *const mismatch_col,
                         int *const mismatch_plane, int *const mismatch_pix1,
                         int *const mismatch_pix2) {
-  if (img1->fmt != img2->fmt || img1->cs != img2->cs ||
-      img1->d_w != img2->d_w || img1->d_h != img2->d_h) {
+  if (img1->fmt != img2->fmt || img1->cp != img2->cp || img1->tc != img2->tc ||
+      img1->mc != img2->mc || img1->d_w != img2->d_w ||
+      img1->d_h != img2->d_h || img1->monochrome != img2->monochrome) {
     if (mismatch_row != NULL) *mismatch_row = -1;
     if (mismatch_col != NULL) *mismatch_col = -1;
     return false;
   }
 
-  for (int plane = 0; plane < 3; plane++) {
+  const int num_planes = img1->monochrome ? 1 : 3;
+  for (int plane = 0; plane < num_planes; plane++) {
     if (!compare_plane(img1->planes[plane], img1->stride[plane],
                        img2->planes[plane], img2->stride[plane],
                        aom_img_plane_width(img1, plane),
@@ -209,7 +193,7 @@ void EncoderTest::RunLoop(VideoSource *video) {
 
     BeginPassHook(pass);
     testing::internal::scoped_ptr<Encoder> encoder(
-        codec_->CreateEncoder(cfg_, deadline_, init_flags_, &stats_));
+        codec_->CreateEncoder(cfg_, init_flags_, &stats_));
     ASSERT_TRUE(encoder.get() != NULL);
 
     ASSERT_NO_FATAL_FAILURE(video->Begin());
@@ -228,10 +212,11 @@ void EncoderTest::RunLoop(VideoSource *video) {
       dec_init_flags |= AOM_CODEC_USE_INPUT_FRAGMENTS;
     testing::internal::scoped_ptr<Decoder> decoder(
         codec_->CreateDecoder(dec_cfg, dec_init_flags));
-#if CONFIG_AV1 && CONFIG_EXT_TILE
+#if CONFIG_AV1_DECODER
     if (decoder->IsAV1()) {
       // Set dec_cfg.tile_row = -1 and dec_cfg.tile_col = -1 so that the whole
       // frame is decoded.
+      decoder->Control(AV1_SET_TILE_MODE, cfg_.large_scale_tile);
       decoder->Control(AV1_SET_DECODE_TILE_ROW, -1);
       decoder->Control(AV1_SET_DECODE_TILE_COL, -1);
     }
@@ -256,8 +241,16 @@ void EncoderTest::RunLoop(VideoSource *video) {
           case AOM_CODEC_CX_FRAME_PKT:
             has_cxdata = true;
             if (decoder.get() != NULL && DoDecode()) {
-              aom_codec_err_t res_dec = decoder->DecodeFrame(
-                  (const uint8_t *)pkt->data.frame.buf, pkt->data.frame.sz);
+              aom_codec_err_t res_dec;
+              if (DoDecodeInvisible()) {
+                res_dec = decoder->DecodeFrame(
+                    (const uint8_t *)pkt->data.frame.buf, pkt->data.frame.sz);
+              } else {
+                res_dec = decoder->DecodeFrame(
+                    (const uint8_t *)pkt->data.frame.buf +
+                        (pkt->data.frame.sz - pkt->data.frame.vis_frame_size),
+                    pkt->data.frame.vis_frame_size);
+              }
 
               if (!HandleDecodeResult(res_dec, decoder.get())) break;
 

third_party/aom/test/encode_test_driver.h

@@ -16,7 +16,8 @@
 
 #include "third_party/googletest/src/googletest/include/gtest/gtest.h"
 
-#include "./aom_config.h"
+#include "config/aom_config.h"
+
 #if CONFIG_AV1_ENCODER
 #include "aom/aomcx.h"
 #endif
@@ -37,6 +38,9 @@ enum TestMode { kRealTime, kOnePassGood, kTwoPassGood };
 
 #define TWO_PASS_TEST_MODES ::testing::Values(::libaom_test::kTwoPassGood)
 
+#define NONREALTIME_TEST_MODES \
+  ::testing::Values(::libaom_test::kOnePassGood, ::libaom_test::kTwoPassGood)
+
 // Provides an object to handle the libaom get_cx_data() iteration pattern
 class CxDataIterator {
  public:
@@ -78,9 +82,9 @@ class TwopassStatsStore {
 // level of abstraction will be fleshed out as more tests are written.
 class Encoder {
  public:
-  Encoder(aom_codec_enc_cfg_t cfg, unsigned long deadline,
-          const unsigned long init_flags, TwopassStatsStore *stats)
-      : cfg_(cfg), deadline_(deadline), init_flags_(init_flags), stats_(stats) {
+  Encoder(aom_codec_enc_cfg_t cfg, const uint32_t init_flags,
+          TwopassStatsStore *stats)
+      : cfg_(cfg), init_flags_(init_flags), stats_(stats) {
     memset(&encoder_, 0, sizeof(encoder_));
   }
 
@@ -128,8 +132,6 @@ class Encoder {
     cfg_ = *cfg;
   }
 
-  void set_deadline(unsigned long deadline) { deadline_ = deadline; }
-
  protected:
   virtual aom_codec_iface_t *CodecInterface() const = 0;
 
@@ -147,7 +149,6 @@ class Encoder {
 
   aom_codec_ctx_t encoder_;
   aom_codec_enc_cfg_t cfg_;
-  unsigned long deadline_;
   unsigned long init_flags_;
   TwopassStatsStore *stats_;
 };
@@ -173,7 +174,7 @@ class EncoderTest {
   // Initialize the cfg_ member with the default configuration.
   void InitializeConfig();
 
-  // Map the TestMode enum to the deadline_ and passes_ variables.
+  // Map the TestMode enum to the passes_ variables.
   void SetMode(TestMode mode);
 
   // Set encoder flag.
@@ -206,9 +207,11 @@ class EncoderTest {
     return !(::testing::Test::HasFatalFailure() || abort_);
   }
 
-  const CodecFactory *codec_;
   // Hook to determine whether to decode frame after encoding
-  virtual bool DoDecode() const { return 1; }
+  virtual bool DoDecode() const { return true; }
+
+  // Hook to determine whether to decode invisible frames after encoding
+  virtual bool DoDecodeInvisible() const { return true; }
 
   // Hook to handle encode/decode mismatch
   virtual void MismatchHook(const aom_image_t *img1, const aom_image_t *img2);
@@ -230,10 +233,10 @@ class EncoderTest {
     return pkt;
   }
 
+  const CodecFactory *codec_;
   bool abort_;
   aom_codec_enc_cfg_t cfg_;
   unsigned int passes_;
-  unsigned long deadline_;
   TwopassStatsStore stats_;
   unsigned long init_flags_;
   unsigned long frame_flags_;

third_party/aom/test/encoder_parms_get_to_decoder.cc

@@ -1,164 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
-
-#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
-
-#include "test/codec_factory.h"
-#include "test/encode_test_driver.h"
-#include "test/util.h"
-#include "test/y4m_video_source.h"
-#include "av1/av1_dx_iface.c"
-
-namespace {
-
-const int kCpuUsed = 2;
-
-struct EncodePerfTestVideo {
-  const char *name;
-  uint32_t width;
-  uint32_t height;
-  uint32_t bitrate;
-  int frames;
-};
-
-const EncodePerfTestVideo kAV1EncodePerfTestVectors[] = {
-  { "niklas_1280_720_30.y4m", 1280, 720, 600, 10 },
-};
-
-struct EncodeParameters {
-  int32_t tile_rows;
-  int32_t tile_cols;
-  int32_t lossless;
-  int32_t error_resilient;
-  int32_t frame_parallel;
-  aom_color_range_t color_range;
-  aom_color_space_t cs;
-#if CONFIG_COLORSPACE_HEADERS
-  aom_transfer_function_t tf;
-  aom_chroma_sample_position_t csp;
-#endif
-  int render_size[2];
-  // TODO(JBB): quantizers / bitrate
-};
-
-const EncodeParameters kAV1EncodeParameterSet[] = {
-  { 0, 0, 0, 1, 0, AOM_CR_STUDIO_RANGE, AOM_CS_BT_601, { 0, 0 } },
-  { 0, 0, 0, 0, 0, AOM_CR_FULL_RANGE, AOM_CS_BT_709, { 0, 0 } },
-#if CONFIG_COLORSPACE_HEADERS
-  { 0, 0, 1, 0, 0, AOM_CR_FULL_RANGE, AOM_CS_BT_2020_NCL, { 0, 0 } },
-#else
-  { 0, 0, 1, 0, 0, AOM_CR_FULL_RANGE, AOM_CS_BT_2020, { 0, 0 } },
-#endif
-  { 0, 2, 0, 0, 1, AOM_CR_STUDIO_RANGE, AOM_CS_UNKNOWN, { 640, 480 } },
-  // TODO(JBB): Test profiles (requires more work).
-};
-
-class AvxEncoderParmsGetToDecoder
-    : public ::libaom_test::CodecTestWith2Params<EncodeParameters,
-                                                 EncodePerfTestVideo>,
-      public ::libaom_test::EncoderTest,
-{
- protected:
-  AvxEncoderParmsGetToDecoder()
-      : EncoderTest(GET_PARAM(0)), encode_parms(GET_PARAM(1)) {}
-
-  virtual ~AvxEncoderParmsGetToDecoder() {}
-
-  virtual void SetUp() {
-    InitializeConfig();
-    SetMode(::libaom_test::kTwoPassGood);
-    cfg_.g_lag_in_frames = 25;
-    cfg_.g_error_resilient = encode_parms.error_resilient;
-    dec_cfg_.threads = 4;
-    test_video_ = GET_PARAM(2);
-    cfg_.rc_target_bitrate = test_video_.bitrate;
-  }
-
-  virtual void PreEncodeFrameHook(::libaom_test::VideoSource *video,
-                                  ::libaom_test::Encoder *encoder) {
-    if (video->frame() == 1) {
-      encoder->Control(AV1E_SET_COLOR_SPACE, encode_parms.cs);
-#if CONFIG_COLORSPACE_HEADERS
-      encoder->Control(AV1E_SET_TRANSFER_FUNCTION, encode_parms.tf);
-      encoder->Control(AV1E_SET_CHROMA_SAMPLE_POSITION, encode_parms.csp);
-#endif
-      encoder->Control(AV1E_SET_COLOR_RANGE, encode_parms.color_range);
-      encoder->Control(AV1E_SET_LOSSLESS, encode_parms.lossless);
-      encoder->Control(AV1E_SET_FRAME_PARALLEL_DECODING,
-                       encode_parms.frame_parallel);
-      encoder->Control(AV1E_SET_TILE_ROWS, encode_parms.tile_rows);
-      encoder->Control(AV1E_SET_TILE_COLUMNS, encode_parms.tile_cols);
-      encoder->Control(AOME_SET_CPUUSED, kCpuUsed);
-      encoder->Control(AOME_SET_ENABLEAUTOALTREF, 1);
-      encoder->Control(AOME_SET_ARNR_MAXFRAMES, 7);
-      encoder->Control(AOME_SET_ARNR_STRENGTH, 5);
-      if (encode_parms.render_size[0] > 0 && encode_parms.render_size[1] > 0)
-        encoder->Control(AV1E_SET_RENDER_SIZE, encode_parms.render_size);
-    }
-  }
-
-  virtual bool HandleDecodeResult(const aom_codec_err_t res_dec,
-                                  libaom_test::Decoder *decoder) {
-    aom_codec_ctx_t *const av1_decoder = decoder->GetDecoder();
-    aom_codec_alg_priv_t *const priv =
-        reinterpret_cast<aom_codec_alg_priv_t *>(av1_decoder->priv);
-    FrameWorkerData *const worker_data =
-        reinterpret_cast<FrameWorkerData *>(priv->frame_workers[0].data1);
-    AV1_COMMON *const common = &worker_data->pbi->common;
-
-    if (encode_parms.lossless) {
-      EXPECT_EQ(0, common->base_qindex);
-      EXPECT_EQ(0, common->y_dc_delta_q);
-      EXPECT_EQ(0, common->uv_dc_delta_q);
-      EXPECT_EQ(0, common->uv_ac_delta_q);
-      EXPECT_EQ(ONLY_4X4, common->tx_mode);
-    }
-    EXPECT_EQ(encode_parms.error_resilient, common->error_resilient_mode);
-    if (encode_parms.error_resilient) {
-      EXPECT_EQ(0, common->use_prev_frame_mvs);
-    }
-    EXPECT_EQ(encode_parms.color_range, common->color_range);
-    EXPECT_EQ(encode_parms.cs, common->color_space);
-#if CONFIG_COLORSPACE_HEADERS
-    EXPECT_EQ(encode_parms.tf, common->transfer_function);
-    EXPECT_EQ(encode_parms.csp, common->chroma_sample_position);
-#endif
-    if (encode_parms.render_size[0] > 0 && encode_parms.render_size[1] > 0) {
-      EXPECT_EQ(encode_parms.render_size[0], common->render_width);
-      EXPECT_EQ(encode_parms.render_size[1], common->render_height);
-    }
-    EXPECT_EQ(encode_parms.tile_cols, common->log2_tile_cols);
-    EXPECT_EQ(encode_parms.tile_rows, common->log2_tile_rows);
-
-    EXPECT_EQ(AOM_CODEC_OK, res_dec) << decoder->DecodeError();
-    return AOM_CODEC_OK == res_dec;
-  }
-
-  EncodePerfTestVideo test_video_;
-
- private:
-  EncodeParameters encode_parms;
-};
-
-TEST_P(AvxEncoderParmsGetToDecoder, BitstreamParms) {
-  init_flags_ = AOM_CODEC_USE_PSNR;
-
-  testing::internal::scoped_ptr<libaom_test::VideoSource> video(
-      new libaom_test::Y4mVideoSource(test_video_.name, 0, test_video_.frames));
-  ASSERT_TRUE(video.get() != NULL);
-
-  ASSERT_NO_FATAL_FAILURE(RunLoop(video.get()));
-}
-
-AV1_INSTANTIATE_TEST_CASE(AvxEncoderParmsGetToDecoder,
-                          ::testing::ValuesIn(kAV1EncodeParameterSet),
-                          ::testing::ValuesIn(kAV1EncodePerfTestVectors));
-}  // namespace

third_party/aom/test/encodetxb_test.cc

@@ -0,0 +1,242 @@
+/*
+ * Copyright (c) 2017, Alliance for Open Media. All rights reserved
+ *
+ * This source code is subject to the terms of the BSD 2 Clause License and
+ * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
+ * was not distributed with this source code in the LICENSE file, you can
+ * obtain it at www.aomedia.org/license/software. If the Alliance for Open
+ * Media Patent License 1.0 was not distributed with this source code in the
+ * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
+ */
+
+#include <stdint.h>
+#include <stdio.h>
+#include <string.h>
+
+#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
+
+#include "config/aom_config.h"
+#include "config/av1_rtcd.h"
+
+#include "aom_ports/aom_timer.h"
+#include "aom_ports/mem.h"
+#include "av1/common/idct.h"
+#include "av1/common/onyxc_int.h"
+#include "av1/common/scan.h"
+#include "av1/common/txb_common.h"
+#include "test/acm_random.h"
+#include "test/clear_system_state.h"
+#include "test/register_state_check.h"
+#include "test/util.h"
+
+namespace {
+using libaom_test::ACMRandom;
+
+typedef void (*GetNzMapContextsFunc)(const uint8_t *const levels,
+                                     const int16_t *const scan,
+                                     const uint16_t eob, const TX_SIZE tx_size,
+                                     const TX_CLASS tx_class,
+                                     int8_t *const coeff_contexts);
+
+class EncodeTxbTest : public ::testing::TestWithParam<GetNzMapContextsFunc> {
+ public:
+  EncodeTxbTest() : get_nz_map_contexts_func_(GetParam()) {}
+
+  virtual ~EncodeTxbTest() {}
+
+  virtual void SetUp() {
+    coeff_contexts_ref_ = reinterpret_cast<int8_t *>(
+        aom_memalign(16, sizeof(*coeff_contexts_ref_) * MAX_TX_SQUARE));
+    ASSERT_TRUE(coeff_contexts_ref_ != NULL);
+    coeff_contexts_ = reinterpret_cast<int8_t *>(
+        aom_memalign(16, sizeof(*coeff_contexts_) * MAX_TX_SQUARE));
+    ASSERT_TRUE(coeff_contexts_ != NULL);
+  }
+
+  virtual void TearDown() {
+    aom_free(coeff_contexts_ref_);
+    aom_free(coeff_contexts_);
+    libaom_test::ClearSystemState();
+  }
+
+  void GetNzMapContextsRun() {
+    const int kNumTests = 10;
+    int result = 0;
+
+    for (int is_inter = 0; is_inter < 2; ++is_inter) {
+      for (int tx_type = DCT_DCT; tx_type < TX_TYPES; ++tx_type) {
+        const TX_CLASS tx_class = tx_type_to_class[tx_type];
+        for (int tx_size = TX_4X4; tx_size < TX_SIZES_ALL; ++tx_size) {
+          const int bwl = get_txb_bwl((TX_SIZE)tx_size);
+          const int width = get_txb_wide((TX_SIZE)tx_size);
+          const int height = get_txb_high((TX_SIZE)tx_size);
+          const int real_width = tx_size_wide[tx_size];
+          const int real_height = tx_size_high[tx_size];
+          const int16_t *const scan = av1_scan_orders[tx_size][tx_type].scan;
+
+          levels_ = set_levels(levels_buf_, width);
+          for (int i = 0; i < kNumTests && !result; ++i) {
+            for (int eob = 1; eob <= width * height && !result; ++eob) {
+              InitDataWithEob(scan, bwl, eob);
+
+              av1_get_nz_map_contexts_c(levels_, scan, eob, (TX_SIZE)tx_size,
+                                        tx_class, coeff_contexts_ref_);
+              get_nz_map_contexts_func_(levels_, scan, eob, (TX_SIZE)tx_size,
+                                        tx_class, coeff_contexts_);
+
+              result = Compare(scan, eob);
+
+              EXPECT_EQ(result, 0)
+                  << " tx_class " << tx_class << " width " << real_width
+                  << " height " << real_height << " eob " << eob;
+            }
+          }
+        }
+      }
+    }
+  }
+
+  void SpeedTestGetNzMapContextsRun() {
+    const int kNumTests = 2000000000;
+    aom_usec_timer timer;
+
+    printf("Note: Only test the largest possible eob case!\n");
+    for (int tx_size = TX_4X4; tx_size < TX_SIZES_ALL; ++tx_size) {
+      const int bwl = get_txb_bwl((TX_SIZE)tx_size);
+      const int width = get_txb_wide((TX_SIZE)tx_size);
+      const int height = get_txb_high((TX_SIZE)tx_size);
+      const int real_width = tx_size_wide[tx_size];
+      const int real_height = tx_size_high[tx_size];
+      const TX_TYPE tx_type = DCT_DCT;
+      const TX_CLASS tx_class = tx_type_to_class[tx_type];
+      const int16_t *const scan = av1_scan_orders[tx_size][tx_type].scan;
+      const int eob = width * height;
+      const int numTests = kNumTests / (width * height);
+
+      levels_ = set_levels(levels_buf_, width);
+      InitDataWithEob(scan, bwl, eob);
+
+      aom_usec_timer_start(&timer);
+      for (int i = 0; i < numTests; ++i) {
+        get_nz_map_contexts_func_(levels_, scan, eob, (TX_SIZE)tx_size,
+                                  tx_class, coeff_contexts_);
+      }
+      aom_usec_timer_mark(&timer);
+
+      const int elapsed_time = static_cast<int>(aom_usec_timer_elapsed(&timer));
+      printf("get_nz_map_contexts_%2dx%2d: %7.1f ms\n", real_width, real_height,
+             elapsed_time / 1000.0);
+    }
+  }
+
+ private:
+  void InitDataWithEob(const int16_t *const scan, const int bwl,
+                       const int eob) {
+    memset(levels_buf_, 0, sizeof(levels_buf_));
+    memset(coeff_contexts_, 0, sizeof(*coeff_contexts_) * MAX_TX_SQUARE);
+
+    for (int c = 0; c < eob; ++c) {
+      levels_[get_padded_idx(scan[c], bwl)] =
+          static_cast<uint8_t>(clamp(rnd_.Rand8(), 0, INT8_MAX));
+      coeff_contexts_[scan[c]] = rnd_.Rand16() >> 1;
+    }
+
+    memcpy(coeff_contexts_ref_, coeff_contexts_,
+           sizeof(*coeff_contexts_) * MAX_TX_SQUARE);
+  }
+
+  bool Compare(const int16_t *const scan, const int eob) const {
+    bool result = false;
+    if (memcmp(coeff_contexts_, coeff_contexts_ref_,
+               sizeof(*coeff_contexts_ref_) * MAX_TX_SQUARE)) {
+      for (int i = 0; i < eob; i++) {
+        const int pos = scan[i];
+        if (coeff_contexts_ref_[pos] != coeff_contexts_[pos]) {
+          printf("coeff_contexts_[%d] diff:%6d (ref),%6d (opt)\n", pos,
+                 coeff_contexts_ref_[pos], coeff_contexts_[pos]);
+          result = true;
+          break;
+        }
+      }
+    }
+    return result;
+  }
+
+  GetNzMapContextsFunc get_nz_map_contexts_func_;
+  ACMRandom rnd_;
+  uint8_t levels_buf_[TX_PAD_2D];
+  uint8_t *levels_;
+  int8_t *coeff_contexts_ref_;
+  int8_t *coeff_contexts_;
+};
+
+TEST_P(EncodeTxbTest, GetNzMapContexts) { GetNzMapContextsRun(); }
+
+TEST_P(EncodeTxbTest, DISABLED_SpeedTestGetNzMapContexts) {
+  SpeedTestGetNzMapContextsRun();
+}
+
+#if HAVE_SSE2
+INSTANTIATE_TEST_CASE_P(SSE2, EncodeTxbTest,
+                        ::testing::Values(av1_get_nz_map_contexts_sse2));
+#endif
+
+#if HAVE_SSE4_1
+class EncodeTxbInitLevelTest : public ::testing::TestWithParam<int> {
+ public:
+  virtual ~EncodeTxbInitLevelTest() {}
+  virtual void TearDown() { libaom_test::ClearSystemState(); }
+  void RunTest(int tx_size, int is_speed);
+};
+
+void EncodeTxbInitLevelTest::RunTest(int tx_size, int is_speed) {
+  const int width = get_txb_wide((TX_SIZE)tx_size);
+  const int height = get_txb_high((TX_SIZE)tx_size);
+  tran_low_t coeff[MAX_TX_SQUARE];
+
+  uint8_t levels_buf[2][TX_PAD_2D];
+  uint8_t *const levels0 = set_levels(levels_buf[0], width);
+  uint8_t *const levels1 = set_levels(levels_buf[1], width);
+
+  ACMRandom rnd(ACMRandom::DeterministicSeed());
+  for (int i = 0; i < width * height; i++) {
+    coeff[i] = rnd.Rand15Signed() + rnd.Rand15Signed();
+  }
+  for (int i = 0; i < TX_PAD_2D; i++) {
+    levels_buf[0][i] = rnd.Rand8();
+    levels_buf[1][i] = rnd.Rand8();
+  }
+  const int run_times = is_speed ? (width * height) * 10000 : 1;
+  aom_usec_timer timer;
+  aom_usec_timer_start(&timer);
+  for (int i = 0; i < run_times; ++i) {
+    av1_txb_init_levels_c(coeff, width, height, levels0);
+  }
+  const double t1 = get_time_mark(&timer);
+  aom_usec_timer_start(&timer);
+  for (int i = 0; i < run_times; ++i) {
+    av1_txb_init_levels_sse4_1(coeff, width, height, levels1);
+  }
+  const double t2 = get_time_mark(&timer);
+  if (is_speed) {
+    printf("init %3dx%-3d:%7.2f/%7.2fns", width, height, t1, t2);
+    printf("(%3.2f)\n", t1 / t2);
+  }
+  const int stride = width + TX_PAD_HOR;
+  for (int r = 0; r < height + TX_PAD_VER; ++r) {
+    for (int c = 0; c < stride; ++c) {
+      ASSERT_EQ(levels_buf[0][c + r * stride], levels_buf[1][c + r * stride])
+          << "[" << r << "," << c << "] " << run_times << width << "x"
+          << height;
+    }
+  }
+}
+
+TEST_P(EncodeTxbInitLevelTest, match) { RunTest(GetParam(), 0); }
+TEST_P(EncodeTxbInitLevelTest, DISABLED_Speed) { RunTest(GetParam(), 1); }
+
+INSTANTIATE_TEST_CASE_P(SSE4_1, EncodeTxbInitLevelTest,
+                        ::testing::Range(0, static_cast<int>(TX_SIZES_ALL), 1));
+#endif
+
+}  // namespace

third_party/aom/test/end_to_end_test.cc

@@ -7,7 +7,7 @@
  * obtain it at www.aomedia.org/license/software. If the Alliance for Open
  * Media Patent License 1.0 was not distributed with this source code in the
  * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
+ */
 
 #include "third_party/googletest/src/googletest/include/gtest/gtest.h"
 
@@ -30,7 +30,7 @@ const double kPsnrThreshold[][5] = {
 // AV1 HBD average PSNR is slightly lower than AV1.
 // We make two cases here to enable the testing and
 // guard picture quality.
-#if CONFIG_AV1_ENCODER && CONFIG_HIGHBITDEPTH
+#if CONFIG_AV1_ENCODER
   { 36.0, 37.0, 37.0, 37.0, 37.0 }, { 31.0, 36.0, 36.0, 36.0, 36.0 },
   { 31.0, 35.0, 35.0, 35.0, 35.0 }, { 31.0, 34.0, 34.0, 34.0, 34.0 },
   { 31.0, 33.0, 33.0, 33.0, 33.0 }, { 31.0, 32.0, 32.0, 32.0, 32.0 },
@@ -40,7 +40,7 @@ const double kPsnrThreshold[][5] = {
   { 34.0, 35.0, 35.0, 35.0, 35.0 }, { 33.0, 34.0, 34.0, 34.0, 34.0 },
   { 32.0, 33.0, 33.0, 33.0, 33.0 }, { 31.0, 32.0, 32.0, 32.0, 32.0 },
   { 30.0, 31.0, 31.0, 31.0, 31.0 }, { 29.0, 30.0, 30.0, 30.0, 30.0 },
-#endif  // CONFIG_HIGHBITDEPTH && CONFIG_AV1_ENCODER
+#endif  // CONFIG_AV1_ENCODER
 };
 
 typedef struct {
@@ -53,24 +53,20 @@ typedef struct {
 
 const TestVideoParam kTestVectors[] = {
   { "park_joy_90p_8_420.y4m", 8, AOM_IMG_FMT_I420, AOM_BITS_8, 0 },
-  { "park_joy_90p_8_422.y4m", 8, AOM_IMG_FMT_I422, AOM_BITS_8, 1 },
+  { "park_joy_90p_8_422.y4m", 8, AOM_IMG_FMT_I422, AOM_BITS_8, 2 },
   { "park_joy_90p_8_444.y4m", 8, AOM_IMG_FMT_I444, AOM_BITS_8, 1 },
-  { "park_joy_90p_8_440.yuv", 8, AOM_IMG_FMT_I440, AOM_BITS_8, 1 },
-#if CONFIG_HIGHBITDEPTH
-  { "park_joy_90p_10_420.y4m", 10, AOM_IMG_FMT_I42016, AOM_BITS_10, 2 },
-  { "park_joy_90p_10_422.y4m", 10, AOM_IMG_FMT_I42216, AOM_BITS_10, 3 },
-  { "park_joy_90p_10_444.y4m", 10, AOM_IMG_FMT_I44416, AOM_BITS_10, 3 },
-  { "park_joy_90p_10_440.yuv", 10, AOM_IMG_FMT_I44016, AOM_BITS_10, 3 },
+  { "park_joy_90p_10_420.y4m", 10, AOM_IMG_FMT_I42016, AOM_BITS_10, 0 },
+  { "park_joy_90p_10_422.y4m", 10, AOM_IMG_FMT_I42216, AOM_BITS_10, 2 },
+  { "park_joy_90p_10_444.y4m", 10, AOM_IMG_FMT_I44416, AOM_BITS_10, 1 },
   { "park_joy_90p_12_420.y4m", 12, AOM_IMG_FMT_I42016, AOM_BITS_12, 2 },
-  { "park_joy_90p_12_422.y4m", 12, AOM_IMG_FMT_I42216, AOM_BITS_12, 3 },
-  { "park_joy_90p_12_444.y4m", 12, AOM_IMG_FMT_I44416, AOM_BITS_12, 3 },
-  { "park_joy_90p_12_440.yuv", 12, AOM_IMG_FMT_I44016, AOM_BITS_12, 3 },
-#endif  // CONFIG_HIGHBITDEPTH
+  { "park_joy_90p_12_422.y4m", 12, AOM_IMG_FMT_I42216, AOM_BITS_12, 2 },
+  { "park_joy_90p_12_444.y4m", 12, AOM_IMG_FMT_I44416, AOM_BITS_12, 2 },
 };
 
 // Encoding modes tested
 const libaom_test::TestMode kEncodingModeVectors[] = {
-  ::libaom_test::kTwoPassGood, ::libaom_test::kOnePassGood,
+  ::libaom_test::kTwoPassGood,
+  ::libaom_test::kOnePassGood,
   ::libaom_test::kRealTime,
 };
 
@@ -150,6 +146,32 @@ class EndToEndTest
     return kPsnrThreshold[cpu_used_][encoding_mode_];
   }
 
+  void DoTest() {
+    cfg_.rc_target_bitrate = kBitrate;
+    cfg_.g_error_resilient = 0;
+    cfg_.g_profile = test_video_param_.profile;
+    cfg_.g_input_bit_depth = test_video_param_.input_bit_depth;
+    cfg_.g_bit_depth = test_video_param_.bit_depth;
+    init_flags_ = AOM_CODEC_USE_PSNR;
+    if (cfg_.g_bit_depth > 8) init_flags_ |= AOM_CODEC_USE_HIGHBITDEPTH;
+
+    testing::internal::scoped_ptr<libaom_test::VideoSource> video;
+    if (is_extension_y4m(test_video_param_.filename)) {
+      video.reset(new libaom_test::Y4mVideoSource(test_video_param_.filename, 0,
+                                                  kFrames));
+    } else {
+      video.reset(new libaom_test::YUVVideoSource(
+          test_video_param_.filename, test_video_param_.fmt, kWidth, kHeight,
+          kFramerate, 1, 0, kFrames));
+    }
+    ASSERT_TRUE(video.get() != NULL);
+
+    ASSERT_NO_FATAL_FAILURE(RunLoop(video.get()));
+    const double psnr = GetAveragePsnr();
+    EXPECT_GT(psnr, GetPsnrThreshold())
+        << "cpu used = " << cpu_used_ << ", encoding mode = " << encoding_mode_;
+  }
+
   TestVideoParam test_video_param_;
   int cpu_used_;
 
@@ -161,55 +183,9 @@ class EndToEndTest
 
 class EndToEndTestLarge : public EndToEndTest {};
 
-TEST_P(EndToEndTestLarge, EndtoEndPSNRTest) {
-  cfg_.rc_target_bitrate = kBitrate;
-  cfg_.g_error_resilient = 0;
-  cfg_.g_profile = test_video_param_.profile;
-  cfg_.g_input_bit_depth = test_video_param_.input_bit_depth;
-  cfg_.g_bit_depth = test_video_param_.bit_depth;
-  init_flags_ = AOM_CODEC_USE_PSNR;
-  if (cfg_.g_bit_depth > 8) init_flags_ |= AOM_CODEC_USE_HIGHBITDEPTH;
+TEST_P(EndToEndTestLarge, EndtoEndPSNRTest) { DoTest(); }
 
-  testing::internal::scoped_ptr<libaom_test::VideoSource> video;
-  if (is_extension_y4m(test_video_param_.filename)) {
-    video.reset(new libaom_test::Y4mVideoSource(test_video_param_.filename, 0,
-                                                kFrames));
-  } else {
-    video.reset(new libaom_test::YUVVideoSource(
-        test_video_param_.filename, test_video_param_.fmt, kWidth, kHeight,
-        kFramerate, 1, 0, kFrames));
-  }
-  ASSERT_TRUE(video.get() != NULL);
-
-  ASSERT_NO_FATAL_FAILURE(RunLoop(video.get()));
-  const double psnr = GetAveragePsnr();
-  EXPECT_GT(psnr, GetPsnrThreshold());
-}
-
-TEST_P(EndToEndTest, EndtoEndPSNRTest) {
-  cfg_.rc_target_bitrate = kBitrate;
-  cfg_.g_error_resilient = 0;
-  cfg_.g_profile = test_video_param_.profile;
-  cfg_.g_input_bit_depth = test_video_param_.input_bit_depth;
-  cfg_.g_bit_depth = test_video_param_.bit_depth;
-  init_flags_ = AOM_CODEC_USE_PSNR;
-  if (cfg_.g_bit_depth > 8) init_flags_ |= AOM_CODEC_USE_HIGHBITDEPTH;
-
-  testing::internal::scoped_ptr<libaom_test::VideoSource> video;
-  if (is_extension_y4m(test_video_param_.filename)) {
-    video.reset(new libaom_test::Y4mVideoSource(test_video_param_.filename, 0,
-                                                kFrames));
-  } else {
-    video.reset(new libaom_test::YUVVideoSource(
-        test_video_param_.filename, test_video_param_.fmt, kWidth, kHeight,
-        kFramerate, 1, 0, kFrames));
-  }
-  ASSERT_TRUE(video.get() != NULL);
-
-  ASSERT_NO_FATAL_FAILURE(RunLoop(video.get()));
-  const double psnr = GetAveragePsnr();
-  EXPECT_GT(psnr, GetPsnrThreshold());
-}
+TEST_P(EndToEndTest, EndtoEndPSNRTest) { DoTest(); }
 
 AV1_INSTANTIATE_TEST_CASE(EndToEndTestLarge,
                           ::testing::ValuesIn(kEncodingModeVectors),

third_party/aom/test/error_block_test.cc

@@ -7,7 +7,7 @@
  * obtain it at www.aomedia.org/license/software. If the Alliance for Open
  * Media Patent License 1.0 was not distributed with this source code in the
  * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
+ */
 
 #include <cmath>
 #include <cstdlib>
@@ -15,8 +15,9 @@
 
 #include "third_party/googletest/src/googletest/include/gtest/gtest.h"
 
-#include "./aom_config.h"
-#include "./av1_rtcd.h"
+#include "config/aom_config.h"
+#include "config/av1_rtcd.h"
+
 #include "test/acm_random.h"
 #include "test/clear_system_state.h"
 #include "test/register_state_check.h"
@@ -28,14 +29,13 @@
 using libaom_test::ACMRandom;
 
 namespace {
-#if CONFIG_HIGHBITDEPTH
 const int kNumIterations = 1000;
 
 typedef int64_t (*ErrorBlockFunc)(const tran_low_t *coeff,
                                   const tran_low_t *dqcoeff,
                                   intptr_t block_size, int64_t *ssz, int bps);
 
-typedef std::tr1::tuple<ErrorBlockFunc, ErrorBlockFunc, aom_bit_depth_t>
+typedef ::testing::tuple<ErrorBlockFunc, ErrorBlockFunc, aom_bit_depth_t>
     ErrorBlockParam;
 
 class ErrorBlockTest : public ::testing::TestWithParam<ErrorBlockParam> {
@@ -156,8 +156,8 @@ TEST_P(ErrorBlockTest, ExtremeValues) {
       << "First failed at test case " << first_failure;
 }
 
-#if HAVE_SSE2 || HAVE_AVX
-using std::tr1::make_tuple;
+#if (HAVE_SSE2 || HAVE_AVX)
+using ::testing::make_tuple;
 
 INSTANTIATE_TEST_CASE_P(
     SSE2, ErrorBlockTest,
@@ -168,6 +168,4 @@ INSTANTIATE_TEST_CASE_P(
                       make_tuple(&av1_highbd_block_error_sse2,
                                  &av1_highbd_block_error_c, AOM_BITS_8)));
 #endif  // HAVE_SSE2
-
-#endif  // CONFIG_HIGHBITDEPTH
 }  // namespace

third_party/aom/test/error_resilience_test.cc

@@ -7,7 +7,7 @@
  * obtain it at www.aomedia.org/license/software. If the Alliance for Open
  * Media Patent License 1.0 was not distributed with this source code in the
  * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
+ */
 
 #include "third_party/googletest/src/googletest/include/gtest/gtest.h"
 #include "test/codec_factory.h"
@@ -18,7 +18,13 @@
 namespace {
 
 const int kMaxErrorFrames = 12;
+const int kMaxInvisibleErrorFrames = 12;
 const int kMaxDroppableFrames = 12;
+const int kMaxErrorResilientFrames = 12;
+const int kMaxNoMFMVFrames = 12;
+const int kMaxPrimRefNoneFrames = 12;
+const int kMaxSFrames = 12;
+const int kCpuUsed = 1;
 
 class ErrorResilienceTestLarge
     : public ::libaom_test::CodecTestWithParam<libaom_test::TestMode>,
@@ -26,7 +32,7 @@ class ErrorResilienceTestLarge
  protected:
   ErrorResilienceTestLarge()
       : EncoderTest(GET_PARAM(0)), psnr_(0.0), nframes_(0), mismatch_psnr_(0.0),
-        mismatch_nframes_(0), encoding_mode_(GET_PARAM(1)) {
+        mismatch_nframes_(0), encoding_mode_(GET_PARAM(1)), allow_mismatch_(0) {
     Reset();
   }
 
@@ -34,8 +40,21 @@ class ErrorResilienceTestLarge
 
   void Reset() {
     error_nframes_ = 0;
+    invisible_error_nframes_ = 0;
     droppable_nframes_ = 0;
-    pattern_switch_ = 0;
+    error_resilient_nframes_ = 0;
+    nomfmv_nframes_ = 0;
+    prim_ref_none_nframes_ = 0;
+    s_nframes_ = 0;
+  }
+
+  void SetupEncoder(int bitrate, int lag) {
+    const aom_rational timebase = { 33333333, 1000000000 };
+    cfg_.g_timebase = timebase;
+    cfg_.rc_target_bitrate = bitrate;
+    cfg_.kf_mode = AOM_KF_DISABLED;
+    cfg_.g_lag_in_frames = lag;
+    init_flags_ = AOM_CODEC_USE_PSNR;
   }
 
   virtual void SetUp() {
@@ -46,6 +65,7 @@ class ErrorResilienceTestLarge
   virtual void BeginPassHook(unsigned int /*pass*/) {
     psnr_ = 0.0;
     nframes_ = 0;
+    decoded_nframes_ = 0;
     mismatch_psnr_ = 0.0;
     mismatch_nframes_ = 0;
   }
@@ -55,18 +75,71 @@ class ErrorResilienceTestLarge
     nframes_++;
   }
 
-  virtual void PreEncodeFrameHook(libaom_test::VideoSource *video) {
+  virtual void PreEncodeFrameHook(libaom_test::VideoSource *video,
+                                  libaom_test::Encoder *encoder) {
+    if (video->frame() == 0) encoder->Control(AOME_SET_CPUUSED, kCpuUsed);
     frame_flags_ &=
-        ~(AOM_EFLAG_NO_UPD_LAST | AOM_EFLAG_NO_UPD_GF | AOM_EFLAG_NO_UPD_ARF);
+        ~(AOM_EFLAG_NO_UPD_LAST | AOM_EFLAG_NO_UPD_GF | AOM_EFLAG_NO_UPD_ARF |
+          AOM_EFLAG_NO_REF_FRAME_MVS | AOM_EFLAG_ERROR_RESILIENT |
+          AOM_EFLAG_SET_S_FRAME | AOM_EFLAG_SET_PRIMARY_REF_NONE);
     if (droppable_nframes_ > 0 &&
         (cfg_.g_pass == AOM_RC_LAST_PASS || cfg_.g_pass == AOM_RC_ONE_PASS)) {
       for (unsigned int i = 0; i < droppable_nframes_; ++i) {
         if (droppable_frames_[i] == video->frame()) {
-          std::cout << "Encoding droppable frame: " << droppable_frames_[i]
-                    << "\n";
+          std::cout << "             Encoding droppable frame: "
+                    << droppable_frames_[i] << "\n";
           frame_flags_ |= (AOM_EFLAG_NO_UPD_LAST | AOM_EFLAG_NO_UPD_GF |
                            AOM_EFLAG_NO_UPD_ARF);
-          return;
+          break;
+        }
+      }
+    }
+
+    if (error_resilient_nframes_ > 0 &&
+        (cfg_.g_pass == AOM_RC_LAST_PASS || cfg_.g_pass == AOM_RC_ONE_PASS)) {
+      for (unsigned int i = 0; i < error_resilient_nframes_; ++i) {
+        if (error_resilient_frames_[i] == video->frame()) {
+          std::cout << "             Encoding error_resilient frame: "
+                    << error_resilient_frames_[i] << "\n";
+          frame_flags_ |= AOM_EFLAG_ERROR_RESILIENT;
+          break;
+        }
+      }
+    }
+
+    if (nomfmv_nframes_ > 0 &&
+        (cfg_.g_pass == AOM_RC_LAST_PASS || cfg_.g_pass == AOM_RC_ONE_PASS)) {
+      for (unsigned int i = 0; i < nomfmv_nframes_; ++i) {
+        if (nomfmv_frames_[i] == video->frame()) {
+          std::cout << "             Encoding no mfmv frame: "
+                    << nomfmv_frames_[i] << "\n";
+          frame_flags_ |= AOM_EFLAG_NO_REF_FRAME_MVS;
+          break;
+        }
+      }
+    }
+
+    if (prim_ref_none_nframes_ > 0 &&
+        (cfg_.g_pass == AOM_RC_LAST_PASS || cfg_.g_pass == AOM_RC_ONE_PASS)) {
+      for (unsigned int i = 0; i < prim_ref_none_nframes_; ++i) {
+        if (prim_ref_none_frames_[i] == video->frame()) {
+          std::cout << "             Encoding no PRIMARY_REF_NONE frame: "
+                    << prim_ref_none_frames_[i] << "\n";
+          frame_flags_ |= AOM_EFLAG_SET_PRIMARY_REF_NONE;
+          break;
+        }
+      }
+    }
+
+    encoder->Control(AV1E_SET_S_FRAME_MODE, 0);
+    if (s_nframes_ > 0 &&
+        (cfg_.g_pass == AOM_RC_LAST_PASS || cfg_.g_pass == AOM_RC_ONE_PASS)) {
+      for (unsigned int i = 0; i < s_nframes_; ++i) {
+        if (s_frames_[i] == video->frame()) {
+          std::cout << "             Encoding S frame: " << s_frames_[i]
+                    << "\n";
+          frame_flags_ |= AOM_EFLAG_SET_S_FRAME;
+          break;
         }
       }
     }
@@ -96,12 +169,37 @@ class ErrorResilienceTestLarge
     return 1;
   }
 
+  virtual bool DoDecodeInvisible() const {
+    if (invisible_error_nframes_ > 0 &&
+        (cfg_.g_pass == AOM_RC_LAST_PASS || cfg_.g_pass == AOM_RC_ONE_PASS)) {
+      for (unsigned int i = 0; i < invisible_error_nframes_; ++i) {
+        if (invisible_error_frames_[i] == nframes_ - 1) {
+          std::cout << "             Skipping decoding all invisible frames in "
+                       "frame pkt: "
+                    << invisible_error_frames_[i] << "\n";
+          return 0;
+        }
+      }
+    }
+    return 1;
+  }
+
   virtual void MismatchHook(const aom_image_t *img1, const aom_image_t *img2) {
-    double mismatch_psnr = compute_psnr(img1, img2);
-    mismatch_psnr_ += mismatch_psnr;
-    ++mismatch_nframes_;
-    // std::cout << "Mismatch frame psnr: " << mismatch_psnr << "\n";
-    ::libaom_test::EncoderTest::MismatchHook(img1, img2);
+    if (allow_mismatch_) {
+      double mismatch_psnr = compute_psnr(img1, img2);
+      mismatch_psnr_ += mismatch_psnr;
+      ++mismatch_nframes_;
+      // std::cout << "Mismatch frame psnr: " << mismatch_psnr << "\n";
+    } else {
+      ::libaom_test::EncoderTest::MismatchHook(img1, img2);
+    }
+  }
+
+  virtual void DecompressedFrameHook(const aom_image_t &img,
+                                     aom_codec_pts_t pts) {
+    (void)img;
+    (void)pts;
+    ++decoded_nframes_;
   }
 
   void SetErrorFrames(int num, unsigned int *list) {
@@ -114,6 +212,16 @@ class ErrorResilienceTestLarge
       error_frames_[i] = list[i];
   }
 
+  void SetInvisibleErrorFrames(int num, unsigned int *list) {
+    if (num > kMaxInvisibleErrorFrames)
+      num = kMaxInvisibleErrorFrames;
+    else if (num < 0)
+      num = 0;
+    invisible_error_nframes_ = num;
+    for (unsigned int i = 0; i < invisible_error_nframes_; ++i)
+      invisible_error_frames_[i] = list[i];
+  }
+
   void SetDroppableFrames(int num, unsigned int *list) {
     if (num > kMaxDroppableFrames)
       num = kMaxDroppableFrames;
@@ -124,42 +232,93 @@ class ErrorResilienceTestLarge
       droppable_frames_[i] = list[i];
   }
 
-  unsigned int GetMismatchFrames() { return mismatch_nframes_; }
+  void SetErrorResilientFrames(int num, unsigned int *list) {
+    if (num > kMaxErrorResilientFrames)
+      num = kMaxErrorResilientFrames;
+    else if (num < 0)
+      num = 0;
+    error_resilient_nframes_ = num;
+    for (unsigned int i = 0; i < error_resilient_nframes_; ++i)
+      error_resilient_frames_[i] = list[i];
+  }
 
-  void SetPatternSwitch(int frame_switch) { pattern_switch_ = frame_switch; }
+  void SetNoMFMVFrames(int num, unsigned int *list) {
+    if (num > kMaxNoMFMVFrames)
+      num = kMaxNoMFMVFrames;
+    else if (num < 0)
+      num = 0;
+    nomfmv_nframes_ = num;
+    for (unsigned int i = 0; i < nomfmv_nframes_; ++i)
+      nomfmv_frames_[i] = list[i];
+  }
+
+  void SetPrimaryRefNoneFrames(int num, unsigned int *list) {
+    if (num > kMaxPrimRefNoneFrames)
+      num = kMaxPrimRefNoneFrames;
+    else if (num < 0)
+      num = 0;
+    prim_ref_none_nframes_ = num;
+    for (unsigned int i = 0; i < prim_ref_none_nframes_; ++i)
+      prim_ref_none_frames_[i] = list[i];
+  }
+
+  void SetSFrames(int num, unsigned int *list) {
+    if (num > kMaxSFrames)
+      num = kMaxSFrames;
+    else if (num < 0)
+      num = 0;
+    s_nframes_ = num;
+    for (unsigned int i = 0; i < s_nframes_; ++i) s_frames_[i] = list[i];
+  }
+
+  unsigned int GetMismatchFrames() { return mismatch_nframes_; }
+  unsigned int GetEncodedFrames() { return nframes_; }
+  unsigned int GetDecodedFrames() { return decoded_nframes_; }
+
+  void SetAllowMismatch(int allow) { allow_mismatch_ = allow; }
 
  private:
   double psnr_;
   unsigned int nframes_;
+  unsigned int decoded_nframes_;
   unsigned int error_nframes_;
+  unsigned int invisible_error_nframes_;
   unsigned int droppable_nframes_;
-  unsigned int pattern_switch_;
+  unsigned int error_resilient_nframes_;
+  unsigned int nomfmv_nframes_;
+  unsigned int prim_ref_none_nframes_;
+  unsigned int s_nframes_;
   double mismatch_psnr_;
   unsigned int mismatch_nframes_;
   unsigned int error_frames_[kMaxErrorFrames];
+  unsigned int invisible_error_frames_[kMaxInvisibleErrorFrames];
   unsigned int droppable_frames_[kMaxDroppableFrames];
+  unsigned int error_resilient_frames_[kMaxErrorResilientFrames];
+  unsigned int nomfmv_frames_[kMaxNoMFMVFrames];
+  unsigned int prim_ref_none_frames_[kMaxPrimRefNoneFrames];
+  unsigned int s_frames_[kMaxSFrames];
   libaom_test::TestMode encoding_mode_;
+  int allow_mismatch_;
 };
 
 TEST_P(ErrorResilienceTestLarge, OnVersusOff) {
-  const aom_rational timebase = { 33333333, 1000000000 };
-  cfg_.g_timebase = timebase;
-  cfg_.rc_target_bitrate = 2000;
-  cfg_.g_lag_in_frames = 10;
-
-  init_flags_ = AOM_CODEC_USE_PSNR;
-
+  SetupEncoder(2000, 10);
   libaom_test::I420VideoSource video("hantro_collage_w352h288.yuv", 352, 288,
-                                     timebase.den, timebase.num, 0, 12);
+                                     cfg_.g_timebase.den, cfg_.g_timebase.num,
+                                     0, 12);
 
-  // Error resilient mode OFF.
+  // Global error resilient mode OFF.
   cfg_.g_error_resilient = 0;
   ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
   const double psnr_resilience_off = GetAveragePsnr();
   EXPECT_GT(psnr_resilience_off, 25.0);
 
-  // Error resilient mode ON.
-  cfg_.g_error_resilient = 1;
+  Reset();
+  // Error resilient mode ON for certain frames
+  unsigned int num_error_resilient_frames = 5;
+  unsigned int error_resilient_frame_list[] = { 3, 5, 6, 9, 11 };
+  SetErrorResilientFrames(num_error_resilient_frames,
+                          error_resilient_frame_list);
   ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
   const double psnr_resilience_on = GetAveragePsnr();
   EXPECT_GT(psnr_resilience_on, 25.0);
@@ -175,60 +334,105 @@ TEST_P(ErrorResilienceTestLarge, OnVersusOff) {
 // Check for successful decoding and no encoder/decoder mismatch
 // if we lose (i.e., drop before decoding) a set of droppable
 // frames (i.e., frames that don't update any reference buffers).
-// Check both isolated and consecutive loss.
 TEST_P(ErrorResilienceTestLarge, DropFramesWithoutRecovery) {
-  const aom_rational timebase = { 33333333, 1000000000 };
-  cfg_.g_timebase = timebase;
-  cfg_.rc_target_bitrate = 500;
-  // FIXME(debargha): Fix this to work for any lag.
-  // Currently this test only works for lag = 0
-  cfg_.g_lag_in_frames = 0;
-
-  init_flags_ = AOM_CODEC_USE_PSNR;
-
+  SetupEncoder(500, 10);
   libaom_test::I420VideoSource video("hantro_collage_w352h288.yuv", 352, 288,
-                                     timebase.den, timebase.num, 0, 20);
-
-  // Error resilient mode ON.
-  cfg_.g_error_resilient = 1;
-  cfg_.kf_mode = AOM_KF_DISABLED;
+                                     cfg_.g_timebase.den, cfg_.g_timebase.num,
+                                     0, 20);
 
   // Set an arbitrary set of error frames same as droppable frames.
-  // In addition to isolated loss/drop, add a long consecutive series
-  // (of size 9) of dropped frames.
-  unsigned int num_droppable_frames = 5;
-  unsigned int droppable_frame_list[] = { 5, 10, 13, 16, 19 };
+  unsigned int num_droppable_frames = 3;
+  unsigned int droppable_frame_list[] = { 5, 10, 13 };
   SetDroppableFrames(num_droppable_frames, droppable_frame_list);
   SetErrorFrames(num_droppable_frames, droppable_frame_list);
   ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
   // Test that no mismatches have been found
+  std::cout << "             Encoded frames: " << GetEncodedFrames() << "\n";
+  std::cout << "             Decoded frames: " << GetDecodedFrames() << "\n";
   std::cout << "             Mismatch frames: " << GetMismatchFrames() << "\n";
-  EXPECT_EQ(GetMismatchFrames(), (unsigned int)0);
-
-  // Reset previously set of error/droppable frames.
-  Reset();
-
-#if 0
-  // TODO(jkoleszar): This test is disabled for the time being as too
-  // sensitive. It's not clear how to set a reasonable threshold for
-  // this behavior.
-
-  // Now set an arbitrary set of error frames that are non-droppable
-  unsigned int num_error_frames = 3;
-  unsigned int error_frame_list[] = {3, 10, 20};
-  SetErrorFrames(num_error_frames, error_frame_list);
-  ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
-
-  // Test that dropping an arbitrary set of inter frames does not hurt too much
-  // Note the Average Mismatch PSNR is the average of the PSNR between
-  // decoded frame and encoder's version of the same frame for all frames
-  // with mismatch.
-  const double psnr_resilience_mismatch = GetAverageMismatchPsnr();
-  std::cout << "             Mismatch PSNR: "
-            << psnr_resilience_mismatch << "\n";
-  EXPECT_GT(psnr_resilience_mismatch, 20.0);
-#endif
+  EXPECT_EQ(GetEncodedFrames() - GetDecodedFrames(), num_droppable_frames);
 }
 
-AV1_INSTANTIATE_TEST_CASE(ErrorResilienceTestLarge, ONE_PASS_TEST_MODES);
+// Check for ParseAbility property of an error-resilient frame.
+// Encode a frame in error-resilient mode (E-frame), and disallow all
+// subsequent frames from using MFMV. If frames are dropped before the
+// E frame, all frames starting from the E frame should be parse-able.
+TEST_P(ErrorResilienceTestLarge, ParseAbilityTest) {
+  SetupEncoder(500, 10);
+
+  libaom_test::I420VideoSource video("hantro_collage_w352h288.yuv", 352, 288,
+                                     cfg_.g_timebase.den, cfg_.g_timebase.num,
+                                     0, 15);
+
+  SetAllowMismatch(1);
+
+  // Note that an E-frame cannot be forced on a frame that is a
+  // show_existing_frame, or a frame that comes directly after an invisible
+  // frame. Currently, this will cause an assertion failure.
+  // Set an arbitrary error resilient (E) frame
+  unsigned int num_error_resilient_frames = 1;
+  unsigned int error_resilient_frame_list[] = { 8 };
+  SetErrorResilientFrames(num_error_resilient_frames,
+                          error_resilient_frame_list);
+  // Ensure that any invisible frames before the E frame are dropped
+  SetInvisibleErrorFrames(num_error_resilient_frames,
+                          error_resilient_frame_list);
+  // Set all frames after the error resilient frame to not allow MFMV
+  unsigned int num_post_error_resilient_frames = 6;
+  unsigned int post_error_resilient_frame_list[] = { 9, 10, 11, 12, 13, 14 };
+  SetNoMFMVFrames(num_post_error_resilient_frames,
+                  post_error_resilient_frame_list);
+
+  // Set a few frames before the E frame that are lost (not decoded)
+  unsigned int num_error_frames = 5;
+  unsigned int error_frame_list[] = { 3, 4, 5, 6, 7 };
+  SetErrorFrames(num_error_frames, error_frame_list);
+
+  ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
+  std::cout << "             Encoded frames: " << GetEncodedFrames() << "\n";
+  std::cout << "             Decoded frames: " << GetDecodedFrames() << "\n";
+  std::cout << "             Mismatch frames: " << GetMismatchFrames() << "\n";
+  EXPECT_EQ(GetEncodedFrames() - GetDecodedFrames(), num_error_frames);
+  // All frames following the E-frame and the E-frame are expected to have
+  // mismatches, but still be parse-able.
+  EXPECT_LE(GetMismatchFrames(), num_post_error_resilient_frames + 1);
+}
+
+// Check for ParseAbility property of an S frame.
+// Encode an S-frame. If frames are dropped before the S-frame, all frames
+// starting from the S frame should be parse-able.
+TEST_P(ErrorResilienceTestLarge, SFrameTest) {
+  SetupEncoder(500, 10);
+
+  libaom_test::I420VideoSource video("hantro_collage_w352h288.yuv", 352, 288,
+                                     cfg_.g_timebase.den, cfg_.g_timebase.num,
+                                     0, 15);
+
+  SetAllowMismatch(1);
+
+  // Note that an S-frame cannot be forced on a frame that is a
+  // show_existing_frame. This issue still needs to be addressed.
+  // Set an arbitrary S-frame
+  unsigned int num_s_frames = 1;
+  unsigned int s_frame_list[] = { 6 };
+  SetSFrames(num_s_frames, s_frame_list);
+  // Ensure that any invisible frames before the S frame are dropped
+  SetInvisibleErrorFrames(num_s_frames, s_frame_list);
+
+  // Set a few frames before the S frame that are lost (not decoded)
+  unsigned int num_error_frames = 4;
+  unsigned int error_frame_list[] = { 2, 3, 4, 5 };
+  SetErrorFrames(num_error_frames, error_frame_list);
+
+  ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
+  std::cout << "             Encoded frames: " << GetEncodedFrames() << "\n";
+  std::cout << "             Decoded frames: " << GetDecodedFrames() << "\n";
+  std::cout << "             Mismatch frames: " << GetMismatchFrames() << "\n";
+  EXPECT_EQ(GetEncodedFrames() - GetDecodedFrames(), num_error_frames);
+  // All frames following the S-frame and the S-frame are expected to have
+  // mismatches, but still be parse-able.
+  EXPECT_LE(GetMismatchFrames(), GetEncodedFrames() - s_frame_list[0]);
+}
+
+AV1_INSTANTIATE_TEST_CASE(ErrorResilienceTestLarge, NONREALTIME_TEST_MODES);
 }  // namespace

third_party/aom/test/ethread_test.cc

@@ -7,7 +7,7 @@
  * obtain it at www.aomedia.org/license/software. If the Alliance for Open
  * Media Patent License 1.0 was not distributed with this source code in the
  * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
+ */
 
 #include <string>
 #include <vector>
@@ -16,7 +16,7 @@
 #include "test/encode_test_driver.h"
 #include "test/md5_helper.h"
 #include "test/util.h"
-#include "test/y4m_video_source.h"
+#include "test/yuv_video_source.h"
 
 namespace {
 class AVxEncoderThreadTest
@@ -32,12 +32,10 @@ class AVxEncoderThreadTest
     cfg.h = 720;
     cfg.allow_lowbitdepth = 1;
     decoder_ = codec_->CreateDecoder(cfg, 0);
-#if CONFIG_AV1
     if (decoder_->IsAV1()) {
       decoder_->Control(AV1_SET_DECODE_TILE_ROW, -1);
       decoder_->Control(AV1_SET_DECODE_TILE_COL, -1);
     }
-#endif
 
     size_enc_.clear();
     md5_dec_.clear();
@@ -71,9 +69,6 @@ class AVxEncoderThreadTest
                                   ::libaom_test::Encoder *encoder) {
     if (!encoder_initialized_) {
       SetTileSize(encoder);
-#if CONFIG_LOOPFILTERING_ACROSS_TILES
-      encoder->Control(AV1E_SET_TILE_LOOPFILTER, 0);
-#endif  // CONFIG_LOOPFILTERING_ACROSS_TILES
       encoder->Control(AOME_SET_CPUUSED, set_cpu_used_);
       if (encoding_mode_ != ::libaom_test::kRealTime) {
         encoder->Control(AOME_SET_ENABLEAUTOALTREF, 1);
@@ -118,7 +113,8 @@ class AVxEncoderThreadTest
   }
 
   void DoTest() {
-    ::libaom_test::Y4mVideoSource video("niklas_1280_720_30.y4m", 15, 18);
+    ::libaom_test::YUVVideoSource video(
+        "niklas_640_480_30.yuv", AOM_IMG_FMT_I420, 640, 480, 30, 1, 15, 18);
     cfg_.rc_target_bitrate = 1000;
 
     // Encode using single thread.
@@ -164,18 +160,16 @@ class AVxEncoderThreadTest
 };
 
 TEST_P(AVxEncoderThreadTest, EncoderResultTest) {
-#if CONFIG_AV1 && CONFIG_EXT_TILE
   cfg_.large_scale_tile = 0;
-#endif  // CONFIG_AV1 && CONFIG_EXT_TILE
+  decoder_->Control(AV1_SET_TILE_MODE, 0);
   DoTest();
 }
 
 class AVxEncoderThreadTestLarge : public AVxEncoderThreadTest {};
 
 TEST_P(AVxEncoderThreadTestLarge, EncoderResultTest) {
-#if CONFIG_AV1 && CONFIG_EXT_TILE
   cfg_.large_scale_tile = 0;
-#endif  // CONFIG_AV1 && CONFIG_EXT_TILE
+  decoder_->Control(AV1_SET_TILE_MODE, 0);
   DoTest();
 }
 
@@ -190,7 +184,6 @@ AV1_INSTANTIATE_TEST_CASE(AVxEncoderThreadTestLarge,
                                             ::libaom_test::kOnePassGood),
                           ::testing::Range(0, 2));
 
-#if CONFIG_AV1 && CONFIG_EXT_TILE
 class AVxEncoderThreadLSTest : public AVxEncoderThreadTest {
   virtual void SetTileSize(libaom_test::Encoder *encoder) {
     encoder->Control(AV1E_SET_TILE_COLUMNS, 1);
@@ -200,15 +193,17 @@ class AVxEncoderThreadLSTest : public AVxEncoderThreadTest {
   }
 };
 
-TEST_P(AVxEncoderThreadLSTest, EncoderResultTest) {
+TEST_P(AVxEncoderThreadLSTest, DISABLED_EncoderResultTest) {
   cfg_.large_scale_tile = 1;
+  decoder_->Control(AV1_SET_TILE_MODE, 1);
   DoTest();
 }
 
 class AVxEncoderThreadLSTestLarge : public AVxEncoderThreadLSTest {};
 
-TEST_P(AVxEncoderThreadLSTestLarge, EncoderResultTest) {
+TEST_P(AVxEncoderThreadLSTestLarge, DISABLED_EncoderResultTest) {
   cfg_.large_scale_tile = 1;
+  decoder_->Control(AV1_SET_TILE_MODE, 1);
   DoTest();
 }
 
@@ -220,5 +215,4 @@ AV1_INSTANTIATE_TEST_CASE(AVxEncoderThreadLSTestLarge,
                           ::testing::Values(::libaom_test::kTwoPassGood,
                                             ::libaom_test::kOnePassGood),
                           ::testing::Range(0, 2));
-#endif  // CONFIG_AV1 && CONFIG_EXT_TILE
 }  // namespace

third_party/aom/test/examples.sh

@@ -12,10 +12,10 @@
 ##
 . $(dirname $0)/tools_common.sh
 
-example_tests=$(ls $(dirname $0)/*.sh)
+example_tests=$(ls -r $(dirname $0)/*.sh)
 
 # List of script names to exclude.
-exclude_list="examples tools_common decode_to_md5"
+exclude_list="best_encode examples run_encodes tools_common"
 
 # Filter out the scripts in $exclude_list.
 for word in ${exclude_list}; do

third_party/aom/test/fdct4x4_test.cc

@@ -1,350 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
-
-#include <math.h>
-#include <stdlib.h>
-#include <string.h>
-
-#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
-
-#include "./av1_rtcd.h"
-#include "./aom_dsp_rtcd.h"
-#include "test/acm_random.h"
-#include "test/clear_system_state.h"
-#include "test/register_state_check.h"
-#include "test/transform_test_base.h"
-#include "test/util.h"
-#include "av1/common/entropy.h"
-#include "aom/aom_codec.h"
-#include "aom/aom_integer.h"
-#include "aom_ports/mem.h"
-
-using libaom_test::ACMRandom;
-
-namespace {
-typedef void (*FdctFunc)(const int16_t *in, tran_low_t *out, int stride);
-typedef void (*IdctFunc)(const tran_low_t *in, uint8_t *out, int stride);
-typedef void (*IhtFunc)(const tran_low_t *in, uint8_t *out, int stride,
-                        const TxfmParam *txfm_param);
-using libaom_test::FhtFunc;
-
-typedef std::tr1::tuple<FdctFunc, IdctFunc, TX_TYPE, aom_bit_depth_t, int>
-    Dct4x4Param;
-typedef std::tr1::tuple<FhtFunc, IhtFunc, TX_TYPE, aom_bit_depth_t, int>
-    Ht4x4Param;
-
-void fdct4x4_ref(const int16_t *in, tran_low_t *out, int stride,
-                 TxfmParam * /*txfm_param*/) {
-  aom_fdct4x4_c(in, out, stride);
-}
-
-void fht4x4_ref(const int16_t *in, tran_low_t *out, int stride,
-                TxfmParam *txfm_param) {
-  av1_fht4x4_c(in, out, stride, txfm_param);
-}
-
-void fwht4x4_ref(const int16_t *in, tran_low_t *out, int stride,
-                 TxfmParam * /*txfm_param*/) {
-  av1_fwht4x4_c(in, out, stride);
-}
-
-#if CONFIG_HIGHBITDEPTH
-void fht4x4_10(const int16_t *in, tran_low_t *out, int stride,
-               TxfmParam *txfm_param) {
-  av1_fwd_txfm2d_4x4_c(in, out, stride, txfm_param->tx_type, 10);
-}
-
-void fht4x4_12(const int16_t *in, tran_low_t *out, int stride,
-               TxfmParam *txfm_param) {
-  av1_fwd_txfm2d_4x4_c(in, out, stride, txfm_param->tx_type, 12);
-}
-
-void iht4x4_10(const tran_low_t *in, uint8_t *out, int stride,
-               const TxfmParam *txfm_param) {
-  av1_inv_txfm2d_add_4x4_c(in, CONVERT_TO_SHORTPTR(out), stride,
-                           txfm_param->tx_type, 10);
-}
-
-void iht4x4_12(const tran_low_t *in, uint8_t *out, int stride,
-               const TxfmParam *txfm_param) {
-  av1_inv_txfm2d_add_4x4_c(in, CONVERT_TO_SHORTPTR(out), stride,
-                           txfm_param->tx_type, 12);
-}
-
-void iwht4x4_10(const tran_low_t *in, uint8_t *out, int stride) {
-  aom_highbd_iwht4x4_16_add_c(in, out, stride, 10);
-}
-
-void iwht4x4_12(const tran_low_t *in, uint8_t *out, int stride) {
-  aom_highbd_iwht4x4_16_add_c(in, out, stride, 12);
-}
-#endif  // CONFIG_HIGHBITDEPTH
-
-class Trans4x4DCT : public libaom_test::TransformTestBase,
-                    public ::testing::TestWithParam<Dct4x4Param> {
- public:
-  virtual ~Trans4x4DCT() {}
-
-  virtual void SetUp() {
-    fwd_txfm_ = GET_PARAM(0);
-    inv_txfm_ = GET_PARAM(1);
-    pitch_ = 4;
-    height_ = 4;
-    fwd_txfm_ref = fdct4x4_ref;
-    bit_depth_ = GET_PARAM(3);
-    mask_ = (1 << bit_depth_) - 1;
-    num_coeffs_ = GET_PARAM(4);
-    txfm_param_.tx_type = GET_PARAM(2);
-  }
-  virtual void TearDown() { libaom_test::ClearSystemState(); }
-
- protected:
-  void RunFwdTxfm(const int16_t *in, tran_low_t *out, int stride) {
-    fwd_txfm_(in, out, stride);
-  }
-  void RunInvTxfm(const tran_low_t *out, uint8_t *dst, int stride) {
-    inv_txfm_(out, dst, stride);
-  }
-
-  FdctFunc fwd_txfm_;
-  IdctFunc inv_txfm_;
-};
-
-TEST_P(Trans4x4DCT, AccuracyCheck) { RunAccuracyCheck(0, 0.00001); }
-
-TEST_P(Trans4x4DCT, CoeffCheck) { RunCoeffCheck(); }
-
-TEST_P(Trans4x4DCT, MemCheck) { RunMemCheck(); }
-
-TEST_P(Trans4x4DCT, InvAccuracyCheck) { RunInvAccuracyCheck(1); }
-
-class Trans4x4HT : public libaom_test::TransformTestBase,
-                   public ::testing::TestWithParam<Ht4x4Param> {
- public:
-  virtual ~Trans4x4HT() {}
-
-  virtual void SetUp() {
-    fwd_txfm_ = GET_PARAM(0);
-    inv_txfm_ = GET_PARAM(1);
-    pitch_ = 4;
-    height_ = 4;
-    fwd_txfm_ref = fht4x4_ref;
-    bit_depth_ = GET_PARAM(3);
-    mask_ = (1 << bit_depth_) - 1;
-    num_coeffs_ = GET_PARAM(4);
-    txfm_param_.tx_type = GET_PARAM(2);
-#if CONFIG_HIGHBITDEPTH
-    switch (bit_depth_) {
-      case AOM_BITS_10: fwd_txfm_ref = fht4x4_10; break;
-      case AOM_BITS_12: fwd_txfm_ref = fht4x4_12; break;
-      default: fwd_txfm_ref = fht4x4_ref; break;
-    }
-#endif
-  }
-  virtual void TearDown() { libaom_test::ClearSystemState(); }
-
- protected:
-  void RunFwdTxfm(const int16_t *in, tran_low_t *out, int stride) {
-    fwd_txfm_(in, out, stride, &txfm_param_);
-  }
-
-  void RunInvTxfm(const tran_low_t *out, uint8_t *dst, int stride) {
-    inv_txfm_(out, dst, stride, &txfm_param_);
-  }
-
-  FhtFunc fwd_txfm_;
-  IhtFunc inv_txfm_;
-};
-
-TEST_P(Trans4x4HT, AccuracyCheck) { RunAccuracyCheck(1, 0.005); }
-
-TEST_P(Trans4x4HT, CoeffCheck) { RunCoeffCheck(); }
-
-TEST_P(Trans4x4HT, MemCheck) { RunMemCheck(); }
-
-TEST_P(Trans4x4HT, InvAccuracyCheck) { RunInvAccuracyCheck(1); }
-
-class Trans4x4WHT : public libaom_test::TransformTestBase,
-                    public ::testing::TestWithParam<Dct4x4Param> {
- public:
-  virtual ~Trans4x4WHT() {}
-
-  virtual void SetUp() {
-    fwd_txfm_ = GET_PARAM(0);
-    inv_txfm_ = GET_PARAM(1);
-    pitch_ = 4;
-    height_ = 4;
-    fwd_txfm_ref = fwht4x4_ref;
-    bit_depth_ = GET_PARAM(3);
-    mask_ = (1 << bit_depth_) - 1;
-    num_coeffs_ = GET_PARAM(4);
-  }
-  virtual void TearDown() { libaom_test::ClearSystemState(); }
-
- protected:
-  void RunFwdTxfm(const int16_t *in, tran_low_t *out, int stride) {
-    fwd_txfm_(in, out, stride);
-  }
-  void RunInvTxfm(const tran_low_t *out, uint8_t *dst, int stride) {
-    inv_txfm_(out, dst, stride);
-  }
-
-  FdctFunc fwd_txfm_;
-  IdctFunc inv_txfm_;
-};
-
-TEST_P(Trans4x4WHT, AccuracyCheck) { RunAccuracyCheck(0, 0.00001); }
-
-TEST_P(Trans4x4WHT, CoeffCheck) { RunCoeffCheck(); }
-
-TEST_P(Trans4x4WHT, MemCheck) { RunMemCheck(); }
-
-TEST_P(Trans4x4WHT, InvAccuracyCheck) { RunInvAccuracyCheck(0); }
-using std::tr1::make_tuple;
-
-INSTANTIATE_TEST_CASE_P(C, Trans4x4DCT,
-                        ::testing::Values(make_tuple(&aom_fdct4x4_c,
-                                                     &aom_idct4x4_16_add_c,
-                                                     DCT_DCT, AOM_BITS_8, 16)));
-
-#if CONFIG_HIGHBITDEPTH
-INSTANTIATE_TEST_CASE_P(
-    DISABLED_C, Trans4x4HT,
-    ::testing::Values(
-        make_tuple(&fht4x4_12, &iht4x4_12, DCT_DCT, AOM_BITS_12, 16),
-        make_tuple(&fht4x4_12, &iht4x4_12, ADST_DCT, AOM_BITS_12, 16),
-        make_tuple(&fht4x4_12, &iht4x4_12, DCT_ADST, AOM_BITS_12, 16),
-        make_tuple(&fht4x4_12, &iht4x4_12, ADST_ADST, AOM_BITS_12, 16)));
-
-INSTANTIATE_TEST_CASE_P(
-    C, Trans4x4HT,
-    ::testing::Values(
-        make_tuple(&fht4x4_10, &iht4x4_10, DCT_DCT, AOM_BITS_10, 16),
-        make_tuple(&fht4x4_10, &iht4x4_10, ADST_DCT, AOM_BITS_10, 16),
-        make_tuple(&fht4x4_10, &iht4x4_10, DCT_ADST, AOM_BITS_10, 16),
-        make_tuple(&fht4x4_10, &iht4x4_10, ADST_ADST, AOM_BITS_10, 16),
-        make_tuple(&av1_fht4x4_c, &av1_iht4x4_16_add_c, DCT_DCT, AOM_BITS_8,
-                   16),
-        make_tuple(&av1_fht4x4_c, &av1_iht4x4_16_add_c, ADST_DCT, AOM_BITS_8,
-                   16),
-        make_tuple(&av1_fht4x4_c, &av1_iht4x4_16_add_c, DCT_ADST, AOM_BITS_8,
-                   16),
-        make_tuple(&av1_fht4x4_c, &av1_iht4x4_16_add_c, ADST_ADST, AOM_BITS_8,
-                   16)));
-#else
-INSTANTIATE_TEST_CASE_P(
-    C, Trans4x4HT,
-    ::testing::Values(make_tuple(&av1_fht4x4_c, &av1_iht4x4_16_add_c, DCT_DCT,
-                                 AOM_BITS_8, 16),
-                      make_tuple(&av1_fht4x4_c, &av1_iht4x4_16_add_c, ADST_DCT,
-                                 AOM_BITS_8, 16),
-                      make_tuple(&av1_fht4x4_c, &av1_iht4x4_16_add_c, DCT_ADST,
-                                 AOM_BITS_8, 16),
-                      make_tuple(&av1_fht4x4_c, &av1_iht4x4_16_add_c, ADST_ADST,
-                                 AOM_BITS_8, 16)));
-#endif  // CONFIG_HIGHBITDEPTH
-
-#if CONFIG_HIGHBITDEPTH
-INSTANTIATE_TEST_CASE_P(
-    C, Trans4x4WHT,
-    ::testing::Values(make_tuple(&av1_highbd_fwht4x4_c, &iwht4x4_10, DCT_DCT,
-                                 AOM_BITS_10, 16),
-                      make_tuple(&av1_highbd_fwht4x4_c, &iwht4x4_12, DCT_DCT,
-                                 AOM_BITS_12, 16),
-                      make_tuple(&av1_fwht4x4_c, &aom_iwht4x4_16_add_c, DCT_DCT,
-                                 AOM_BITS_8, 16)));
-#else
-INSTANTIATE_TEST_CASE_P(C, Trans4x4WHT,
-                        ::testing::Values(make_tuple(&av1_fwht4x4_c,
-                                                     &aom_iwht4x4_16_add_c,
-                                                     DCT_DCT, AOM_BITS_8, 16)));
-#endif  // CONFIG_HIGHBITDEPTH
-
-#if HAVE_NEON_ASM && !CONFIG_HIGHBITDEPTH
-INSTANTIATE_TEST_CASE_P(NEON, Trans4x4DCT,
-                        ::testing::Values(make_tuple(&aom_fdct4x4_c,
-                                                     &aom_idct4x4_16_add_neon,
-                                                     DCT_DCT, AOM_BITS_8, 16)));
-#endif  // HAVE_NEON_ASM && !CONFIG_HIGHBITDEPTH
-
-#if HAVE_NEON && !CONFIG_HIGHBITDEPTH
-INSTANTIATE_TEST_CASE_P(
-    NEON, Trans4x4HT,
-    ::testing::Values(make_tuple(&av1_fht4x4_c, &av1_iht4x4_16_add_neon,
-                                 DCT_DCT, AOM_BITS_8, 16),
-                      make_tuple(&av1_fht4x4_c, &av1_iht4x4_16_add_neon,
-                                 ADST_DCT, AOM_BITS_8, 16),
-                      make_tuple(&av1_fht4x4_c, &av1_iht4x4_16_add_neon,
-                                 DCT_ADST, AOM_BITS_8, 16),
-                      make_tuple(&av1_fht4x4_c, &av1_iht4x4_16_add_neon,
-                                 ADST_ADST, AOM_BITS_8, 16)));
-#endif  // HAVE_NEON && !CONFIG_HIGHBITDEPTH
-
-#if HAVE_SSE2 && !CONFIG_DAALA_DCT4
-INSTANTIATE_TEST_CASE_P(
-    SSE2, Trans4x4WHT,
-    ::testing::Values(make_tuple(&av1_fwht4x4_c, &aom_iwht4x4_16_add_c, DCT_DCT,
-                                 AOM_BITS_8, 16),
-                      make_tuple(&av1_fwht4x4_c, &aom_iwht4x4_16_add_sse2,
-                                 DCT_DCT, AOM_BITS_8, 16)));
-#endif
-
-#if HAVE_SSE2 && !CONFIG_HIGHBITDEPTH
-INSTANTIATE_TEST_CASE_P(SSE2, Trans4x4DCT,
-                        ::testing::Values(make_tuple(&aom_fdct4x4_sse2,
-                                                     &aom_idct4x4_16_add_sse2,
-                                                     DCT_DCT, AOM_BITS_8, 16)));
-#if !CONFIG_DAALA_DCT4
-INSTANTIATE_TEST_CASE_P(
-    SSE2, Trans4x4HT,
-    ::testing::Values(make_tuple(&av1_fht4x4_sse2, &av1_iht4x4_16_add_sse2,
-                                 DCT_DCT, AOM_BITS_8, 16),
-                      make_tuple(&av1_fht4x4_sse2, &av1_iht4x4_16_add_sse2,
-                                 ADST_DCT, AOM_BITS_8, 16),
-                      make_tuple(&av1_fht4x4_sse2, &av1_iht4x4_16_add_sse2,
-                                 DCT_ADST, AOM_BITS_8, 16),
-                      make_tuple(&av1_fht4x4_sse2, &av1_iht4x4_16_add_sse2,
-                                 ADST_ADST, AOM_BITS_8, 16)));
-#endif  // !CONFIG_DAALA_DCT4
-#endif  // HAVE_SSE2 && !CONFIG_HIGHBITDEPTH
-
-#if HAVE_SSE2 && CONFIG_HIGHBITDEPTH && !CONFIG_DAALA_DCT4
-INSTANTIATE_TEST_CASE_P(
-    SSE2, Trans4x4HT,
-    ::testing::Values(make_tuple(&av1_fht4x4_sse2, &av1_iht4x4_16_add_c,
-                                 DCT_DCT, AOM_BITS_8, 16),
-                      make_tuple(&av1_fht4x4_sse2, &av1_iht4x4_16_add_c,
-                                 ADST_DCT, AOM_BITS_8, 16),
-                      make_tuple(&av1_fht4x4_sse2, &av1_iht4x4_16_add_c,
-                                 DCT_ADST, AOM_BITS_8, 16),
-                      make_tuple(&av1_fht4x4_sse2, &av1_iht4x4_16_add_c,
-                                 ADST_ADST, AOM_BITS_8, 16)));
-#endif  // HAVE_SSE2 && CONFIG_HIGHBITDEPTH && !CONFIG_DAALA_DCT4
-
-#if HAVE_MSA && !CONFIG_HIGHBITDEPTH
-INSTANTIATE_TEST_CASE_P(MSA, Trans4x4DCT,
-                        ::testing::Values(make_tuple(&aom_fdct4x4_msa,
-                                                     &aom_idct4x4_16_add_msa,
-                                                     DCT_DCT, AOM_BITS_8, 16)));
-#if !CONFIG_EXT_TX && !CONFIG_DAALA_DCT4
-INSTANTIATE_TEST_CASE_P(
-    MSA, Trans4x4HT,
-    ::testing::Values(make_tuple(&av1_fht4x4_msa, &av1_iht4x4_16_add_msa,
-                                 DCT_DCT, AOM_BITS_8, 16),
-                      make_tuple(&av1_fht4x4_msa, &av1_iht4x4_16_add_msa,
-                                 ADST_DCT, AOM_BITS_8, 16),
-                      make_tuple(&av1_fht4x4_msa, &av1_iht4x4_16_add_msa,
-                                 DCT_ADST, AOM_BITS_8, 16),
-                      make_tuple(&av1_fht4x4_msa, &av1_iht4x4_16_add_msa,
-                                 ADST_ADST, AOM_BITS_8, 16)));
-#endif  // !CONFIG_EXT_TX && && !CONFIG_DAALA_DCT4
-#endif  // HAVE_MSA && !CONFIG_HIGHBITDEPTH
-}  // namespace

third_party/aom/test/fdct8x8_test.cc

@@ -1,738 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <math.h>
-#include <stdlib.h>
-#include <string.h>
-
-#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
-
-#include "./av1_rtcd.h"
-#include "./aom_dsp_rtcd.h"
-#include "test/acm_random.h"
-#include "test/clear_system_state.h"
-#include "test/register_state_check.h"
-#include "test/util.h"
-#include "av1/common/entropy.h"
-#include "av1/common/scan.h"
-#include "aom/aom_codec.h"
-#include "aom/aom_integer.h"
-#include "aom_ports/mem.h"
-
-using libaom_test::ACMRandom;
-
-namespace {
-
-const int kNumCoeffs = 64;
-const double kPi = 3.141592653589793238462643383279502884;
-
-const int kSignBiasMaxDiff255 = 1500;
-const int kSignBiasMaxDiff15 = 10000;
-
-typedef void (*FdctFunc)(const int16_t *in, tran_low_t *out, int stride);
-typedef void (*IdctFunc)(const tran_low_t *in, uint8_t *out, int stride);
-typedef void (*FhtFunc)(const int16_t *in, tran_low_t *out, int stride,
-                        TxfmParam *txfm_param);
-typedef void (*IhtFunc)(const tran_low_t *in, uint8_t *out, int stride,
-                        const TxfmParam *txfm_param);
-
-typedef std::tr1::tuple<FdctFunc, IdctFunc, TX_TYPE, aom_bit_depth_t>
-    Dct8x8Param;
-typedef std::tr1::tuple<FhtFunc, IhtFunc, TX_TYPE, aom_bit_depth_t> Ht8x8Param;
-typedef std::tr1::tuple<IdctFunc, IdctFunc, int, aom_bit_depth_t> Idct8x8Param;
-
-void reference_8x8_dct_1d(const double in[8], double out[8]) {
-  const double kInvSqrt2 = 0.707106781186547524400844362104;
-  for (int k = 0; k < 8; k++) {
-    out[k] = 0.0;
-    for (int n = 0; n < 8; n++)
-      out[k] += in[n] * cos(kPi * (2 * n + 1) * k / 16.0);
-    if (k == 0) out[k] = out[k] * kInvSqrt2;
-  }
-}
-
-void reference_8x8_dct_2d(const int16_t input[kNumCoeffs],
-                          double output[kNumCoeffs]) {
-  // First transform columns
-  for (int i = 0; i < 8; ++i) {
-    double temp_in[8], temp_out[8];
-    for (int j = 0; j < 8; ++j) temp_in[j] = input[j * 8 + i];
-    reference_8x8_dct_1d(temp_in, temp_out);
-    for (int j = 0; j < 8; ++j) output[j * 8 + i] = temp_out[j];
-  }
-  // Then transform rows
-  for (int i = 0; i < 8; ++i) {
-    double temp_in[8], temp_out[8];
-    for (int j = 0; j < 8; ++j) temp_in[j] = output[j + i * 8];
-    reference_8x8_dct_1d(temp_in, temp_out);
-    // Scale by some magic number
-    for (int j = 0; j < 8; ++j) output[j + i * 8] = temp_out[j] * 2;
-  }
-}
-
-void fdct8x8_ref(const int16_t *in, tran_low_t *out, int stride,
-                 TxfmParam * /*txfm_param*/) {
-  aom_fdct8x8_c(in, out, stride);
-}
-
-void fht8x8_ref(const int16_t *in, tran_low_t *out, int stride,
-                TxfmParam *txfm_param) {
-  av1_fht8x8_c(in, out, stride, txfm_param);
-}
-
-#if CONFIG_HIGHBITDEPTH
-void fht8x8_10(const int16_t *in, tran_low_t *out, int stride,
-               TxfmParam *txfm_param) {
-  av1_fwd_txfm2d_8x8_c(in, out, stride, txfm_param->tx_type, 10);
-}
-
-void fht8x8_12(const int16_t *in, tran_low_t *out, int stride,
-               TxfmParam *txfm_param) {
-  av1_fwd_txfm2d_8x8_c(in, out, stride, txfm_param->tx_type, 12);
-}
-
-void iht8x8_10(const tran_low_t *in, uint8_t *out, int stride,
-               const TxfmParam *txfm_param) {
-  av1_inv_txfm2d_add_8x8_c(in, CONVERT_TO_SHORTPTR(out), stride,
-                           txfm_param->tx_type, 10);
-}
-
-void iht8x8_12(const tran_low_t *in, uint8_t *out, int stride,
-               const TxfmParam *txfm_param) {
-  av1_inv_txfm2d_add_8x8_c(in, CONVERT_TO_SHORTPTR(out), stride,
-                           txfm_param->tx_type, 12);
-}
-
-#endif  // CONFIG_HIGHBITDEPTH
-
-class FwdTrans8x8TestBase {
- public:
-  virtual ~FwdTrans8x8TestBase() {}
-
- protected:
-  virtual void RunFwdTxfm(int16_t *in, tran_low_t *out, int stride) = 0;
-  virtual void RunInvTxfm(tran_low_t *out, uint8_t *dst, int stride) = 0;
-
-  void RunSignBiasCheck() {
-    ACMRandom rnd(ACMRandom::DeterministicSeed());
-    DECLARE_ALIGNED(16, int16_t, test_input_block[64]);
-    DECLARE_ALIGNED(16, tran_low_t, test_output_block[64]);
-    int count_sign_block[64][2];
-    const int count_test_block = 100000;
-
-    memset(count_sign_block, 0, sizeof(count_sign_block));
-
-    for (int i = 0; i < count_test_block; ++i) {
-      // Initialize a test block with input range [-255, 255].
-      for (int j = 0; j < 64; ++j)
-        test_input_block[j] = ((rnd.Rand16() >> (16 - bit_depth_)) & mask_) -
-                              ((rnd.Rand16() >> (16 - bit_depth_)) & mask_);
-      ASM_REGISTER_STATE_CHECK(
-          RunFwdTxfm(test_input_block, test_output_block, pitch_));
-
-      for (int j = 0; j < 64; ++j) {
-        if (test_output_block[j] < 0)
-          ++count_sign_block[j][0];
-        else if (test_output_block[j] > 0)
-          ++count_sign_block[j][1];
-      }
-    }
-
-    for (int j = 0; j < 64; ++j) {
-      const int diff = abs(count_sign_block[j][0] - count_sign_block[j][1]);
-      const int max_diff = kSignBiasMaxDiff255;
-      EXPECT_LT(diff, max_diff << (bit_depth_ - 8))
-          << "Error: 8x8 FDCT/FHT has a sign bias > "
-          << 1. * max_diff / count_test_block * 100 << "%"
-          << " for input range [-255, 255] at index " << j
-          << " count0: " << count_sign_block[j][0]
-          << " count1: " << count_sign_block[j][1] << " diff: " << diff;
-    }
-
-    memset(count_sign_block, 0, sizeof(count_sign_block));
-
-    for (int i = 0; i < count_test_block; ++i) {
-      // Initialize a test block with input range [-mask_ / 16, mask_ / 16].
-      for (int j = 0; j < 64; ++j)
-        test_input_block[j] =
-            ((rnd.Rand16() & mask_) >> 4) - ((rnd.Rand16() & mask_) >> 4);
-      ASM_REGISTER_STATE_CHECK(
-          RunFwdTxfm(test_input_block, test_output_block, pitch_));
-
-      for (int j = 0; j < 64; ++j) {
-        if (test_output_block[j] < 0)
-          ++count_sign_block[j][0];
-        else if (test_output_block[j] > 0)
-          ++count_sign_block[j][1];
-      }
-    }
-
-    for (int j = 0; j < 64; ++j) {
-      const int diff = abs(count_sign_block[j][0] - count_sign_block[j][1]);
-      const int max_diff = kSignBiasMaxDiff15;
-      EXPECT_LT(diff, max_diff << (bit_depth_ - 8))
-          << "Error: 8x8 FDCT/FHT has a sign bias > "
-          << 1. * max_diff / count_test_block * 100 << "%"
-          << " for input range [-15, 15] at index " << j
-          << " count0: " << count_sign_block[j][0]
-          << " count1: " << count_sign_block[j][1] << " diff: " << diff;
-    }
-  }
-
-  void RunRoundTripErrorCheck() {
-    ACMRandom rnd(ACMRandom::DeterministicSeed());
-    int max_error = 0;
-    int total_error = 0;
-    const int count_test_block = 100000;
-    DECLARE_ALIGNED(16, int16_t, test_input_block[64]);
-    DECLARE_ALIGNED(16, tran_low_t, test_temp_block[64]);
-    DECLARE_ALIGNED(16, uint8_t, dst[64]);
-    DECLARE_ALIGNED(16, uint8_t, src[64]);
-#if CONFIG_HIGHBITDEPTH
-    DECLARE_ALIGNED(16, uint16_t, dst16[64]);
-    DECLARE_ALIGNED(16, uint16_t, src16[64]);
-#endif
-
-    for (int i = 0; i < count_test_block; ++i) {
-      // Initialize a test block with input range [-mask_, mask_].
-      for (int j = 0; j < 64; ++j) {
-        if (bit_depth_ == AOM_BITS_8) {
-          src[j] = rnd.Rand8();
-          dst[j] = rnd.Rand8();
-          test_input_block[j] = src[j] - dst[j];
-#if CONFIG_HIGHBITDEPTH
-        } else {
-          src16[j] = rnd.Rand16() & mask_;
-          dst16[j] = rnd.Rand16() & mask_;
-          test_input_block[j] = src16[j] - dst16[j];
-#endif
-        }
-      }
-
-      ASM_REGISTER_STATE_CHECK(
-          RunFwdTxfm(test_input_block, test_temp_block, pitch_));
-      for (int j = 0; j < 64; ++j) {
-        if (test_temp_block[j] > 0) {
-          test_temp_block[j] += 2;
-          test_temp_block[j] /= 4;
-          test_temp_block[j] *= 4;
-        } else {
-          test_temp_block[j] -= 2;
-          test_temp_block[j] /= 4;
-          test_temp_block[j] *= 4;
-        }
-      }
-      if (bit_depth_ == AOM_BITS_8) {
-        ASM_REGISTER_STATE_CHECK(RunInvTxfm(test_temp_block, dst, pitch_));
-#if CONFIG_HIGHBITDEPTH
-      } else {
-        ASM_REGISTER_STATE_CHECK(
-            RunInvTxfm(test_temp_block, CONVERT_TO_BYTEPTR(dst16), pitch_));
-#endif
-      }
-
-      for (int j = 0; j < 64; ++j) {
-#if CONFIG_HIGHBITDEPTH
-        const int diff =
-            bit_depth_ == AOM_BITS_8 ? dst[j] - src[j] : dst16[j] - src16[j];
-#else
-        const int diff = dst[j] - src[j];
-#endif
-        const int error = diff * diff;
-        if (max_error < error) max_error = error;
-        total_error += error;
-      }
-    }
-
-    EXPECT_GE(1 << 2 * (bit_depth_ - 8), max_error)
-        << "Error: 8x8 FDCT/IDCT or FHT/IHT has an individual"
-        << " roundtrip error > 1";
-
-    EXPECT_GE((count_test_block << 2 * (bit_depth_ - 8)) / 5, total_error)
-        << "Error: 8x8 FDCT/IDCT or FHT/IHT has average roundtrip "
-        << "error > 1/5 per block";
-  }
-
-  void RunExtremalCheck() {
-    ACMRandom rnd(ACMRandom::DeterministicSeed());
-    int max_error = 0;
-    int total_error = 0;
-    int total_coeff_error = 0;
-    const int count_test_block = 100000;
-    DECLARE_ALIGNED(16, int16_t, test_input_block[64]);
-    DECLARE_ALIGNED(16, tran_low_t, test_temp_block[64]);
-    DECLARE_ALIGNED(16, tran_low_t, ref_temp_block[64]);
-    DECLARE_ALIGNED(16, uint8_t, dst[64]);
-    DECLARE_ALIGNED(16, uint8_t, src[64]);
-#if CONFIG_HIGHBITDEPTH
-    DECLARE_ALIGNED(16, uint16_t, dst16[64]);
-    DECLARE_ALIGNED(16, uint16_t, src16[64]);
-#endif
-
-    for (int i = 0; i < count_test_block; ++i) {
-      // Initialize a test block with input range [-mask_, mask_].
-      for (int j = 0; j < 64; ++j) {
-        if (bit_depth_ == AOM_BITS_8) {
-          if (i == 0) {
-            src[j] = 255;
-            dst[j] = 0;
-          } else if (i == 1) {
-            src[j] = 0;
-            dst[j] = 255;
-          } else {
-            src[j] = rnd.Rand8() % 2 ? 255 : 0;
-            dst[j] = rnd.Rand8() % 2 ? 255 : 0;
-          }
-          test_input_block[j] = src[j] - dst[j];
-#if CONFIG_HIGHBITDEPTH
-        } else {
-          if (i == 0) {
-            src16[j] = mask_;
-            dst16[j] = 0;
-          } else if (i == 1) {
-            src16[j] = 0;
-            dst16[j] = mask_;
-          } else {
-            src16[j] = rnd.Rand8() % 2 ? mask_ : 0;
-            dst16[j] = rnd.Rand8() % 2 ? mask_ : 0;
-          }
-          test_input_block[j] = src16[j] - dst16[j];
-#endif
-        }
-      }
-
-      ASM_REGISTER_STATE_CHECK(
-          RunFwdTxfm(test_input_block, test_temp_block, pitch_));
-      ASM_REGISTER_STATE_CHECK(
-          fwd_txfm_ref(test_input_block, ref_temp_block, pitch_, &txfm_param_));
-      if (bit_depth_ == AOM_BITS_8) {
-        ASM_REGISTER_STATE_CHECK(RunInvTxfm(test_temp_block, dst, pitch_));
-#if CONFIG_HIGHBITDEPTH
-      } else {
-        ASM_REGISTER_STATE_CHECK(
-            RunInvTxfm(test_temp_block, CONVERT_TO_BYTEPTR(dst16), pitch_));
-#endif
-      }
-
-      for (int j = 0; j < 64; ++j) {
-#if CONFIG_HIGHBITDEPTH
-        const int diff =
-            bit_depth_ == AOM_BITS_8 ? dst[j] - src[j] : dst16[j] - src16[j];
-#else
-        const int diff = dst[j] - src[j];
-#endif
-        const int error = diff * diff;
-        if (max_error < error) max_error = error;
-        total_error += error;
-
-        const int coeff_diff = test_temp_block[j] - ref_temp_block[j];
-        total_coeff_error += abs(coeff_diff);
-      }
-
-      EXPECT_GE(1 << 2 * (bit_depth_ - 8), max_error)
-          << "Error: Extremal 8x8 FDCT/IDCT or FHT/IHT has"
-          << "an individual roundtrip error > 1";
-
-      EXPECT_GE((count_test_block << 2 * (bit_depth_ - 8)) / 5, total_error)
-          << "Error: Extremal 8x8 FDCT/IDCT or FHT/IHT has average"
-          << " roundtrip error > 1/5 per block";
-
-      EXPECT_EQ(0, total_coeff_error)
-          << "Error: Extremal 8x8 FDCT/FHT has"
-          << "overflow issues in the intermediate steps > 1";
-    }
-  }
-
-  void RunInvAccuracyCheck() {
-    ACMRandom rnd(ACMRandom::DeterministicSeed());
-    const int count_test_block = 1000;
-    DECLARE_ALIGNED(16, int16_t, in[kNumCoeffs]);
-    DECLARE_ALIGNED(16, tran_low_t, coeff[kNumCoeffs]);
-    DECLARE_ALIGNED(16, uint8_t, dst[kNumCoeffs]);
-    DECLARE_ALIGNED(16, uint8_t, src[kNumCoeffs]);
-#if CONFIG_HIGHBITDEPTH
-    DECLARE_ALIGNED(16, uint16_t, src16[kNumCoeffs]);
-    DECLARE_ALIGNED(16, uint16_t, dst16[kNumCoeffs]);
-#endif
-
-    for (int i = 0; i < count_test_block; ++i) {
-      double out_r[kNumCoeffs];
-
-      // Initialize a test block with input range [-255, 255].
-      for (int j = 0; j < kNumCoeffs; ++j) {
-        if (bit_depth_ == AOM_BITS_8) {
-          src[j] = rnd.Rand8() % 2 ? 255 : 0;
-          dst[j] = src[j] > 0 ? 0 : 255;
-          in[j] = src[j] - dst[j];
-#if CONFIG_HIGHBITDEPTH
-        } else {
-          src16[j] = rnd.Rand8() % 2 ? mask_ : 0;
-          dst16[j] = src16[j] > 0 ? 0 : mask_;
-          in[j] = src16[j] - dst16[j];
-#endif
-        }
-      }
-
-      reference_8x8_dct_2d(in, out_r);
-      for (int j = 0; j < kNumCoeffs; ++j)
-        coeff[j] = static_cast<tran_low_t>(round(out_r[j]));
-
-      if (bit_depth_ == AOM_BITS_8) {
-        ASM_REGISTER_STATE_CHECK(RunInvTxfm(coeff, dst, pitch_));
-#if CONFIG_HIGHBITDEPTH
-      } else {
-        ASM_REGISTER_STATE_CHECK(
-            RunInvTxfm(coeff, CONVERT_TO_BYTEPTR(dst16), pitch_));
-#endif
-      }
-
-      for (int j = 0; j < kNumCoeffs; ++j) {
-#if CONFIG_HIGHBITDEPTH
-        const int diff =
-            bit_depth_ == AOM_BITS_8 ? dst[j] - src[j] : dst16[j] - src16[j];
-#else
-        const int diff = dst[j] - src[j];
-#endif
-        const uint32_t error = diff * diff;
-        EXPECT_GE(1u << 2 * (bit_depth_ - 8), error)
-            << "Error: 8x8 IDCT has error " << error << " at index " << j;
-      }
-    }
-  }
-
-  void RunFwdAccuracyCheck() {
-    ACMRandom rnd(ACMRandom::DeterministicSeed());
-    const int count_test_block = 1000;
-    DECLARE_ALIGNED(16, int16_t, in[kNumCoeffs]);
-    DECLARE_ALIGNED(16, tran_low_t, coeff_r[kNumCoeffs]);
-    DECLARE_ALIGNED(16, tran_low_t, coeff[kNumCoeffs]);
-
-    for (int i = 0; i < count_test_block; ++i) {
-      double out_r[kNumCoeffs];
-
-      // Initialize a test block with input range [-mask_, mask_].
-      for (int j = 0; j < kNumCoeffs; ++j)
-        in[j] = rnd.Rand8() % 2 == 0 ? mask_ : -mask_;
-
-      RunFwdTxfm(in, coeff, pitch_);
-      reference_8x8_dct_2d(in, out_r);
-      for (int j = 0; j < kNumCoeffs; ++j)
-        coeff_r[j] = static_cast<tran_low_t>(round(out_r[j]));
-
-      for (int j = 0; j < kNumCoeffs; ++j) {
-        const int32_t diff = coeff[j] - coeff_r[j];
-        const uint32_t error = diff * diff;
-        EXPECT_GE(9u << 2 * (bit_depth_ - 8), error)
-            << "Error: 8x8 DCT has error " << error << " at index " << j;
-      }
-    }
-  }
-
-  void CompareInvReference(IdctFunc ref_txfm, int thresh) {
-    ACMRandom rnd(ACMRandom::DeterministicSeed());
-    const int count_test_block = 10000;
-    const int eob = 12;
-    DECLARE_ALIGNED(16, tran_low_t, coeff[kNumCoeffs]);
-    DECLARE_ALIGNED(16, uint8_t, dst[kNumCoeffs]);
-    DECLARE_ALIGNED(16, uint8_t, ref[kNumCoeffs]);
-#if CONFIG_HIGHBITDEPTH
-    DECLARE_ALIGNED(16, uint16_t, dst16[kNumCoeffs]);
-    DECLARE_ALIGNED(16, uint16_t, ref16[kNumCoeffs]);
-#endif
-    const int16_t *scan = av1_default_scan_orders[TX_8X8].scan;
-
-    for (int i = 0; i < count_test_block; ++i) {
-      for (int j = 0; j < kNumCoeffs; ++j) {
-        if (j < eob) {
-          // Random values less than the threshold, either positive or negative
-          coeff[scan[j]] = rnd(thresh) * (1 - 2 * (i % 2));
-        } else {
-          coeff[scan[j]] = 0;
-        }
-        if (bit_depth_ == AOM_BITS_8) {
-          dst[j] = 0;
-          ref[j] = 0;
-#if CONFIG_HIGHBITDEPTH
-        } else {
-          dst16[j] = 0;
-          ref16[j] = 0;
-#endif
-        }
-      }
-      if (bit_depth_ == AOM_BITS_8) {
-        ref_txfm(coeff, ref, pitch_);
-        ASM_REGISTER_STATE_CHECK(RunInvTxfm(coeff, dst, pitch_));
-#if CONFIG_HIGHBITDEPTH
-      } else {
-        ref_txfm(coeff, CONVERT_TO_BYTEPTR(ref16), pitch_);
-        ASM_REGISTER_STATE_CHECK(
-            RunInvTxfm(coeff, CONVERT_TO_BYTEPTR(dst16), pitch_));
-#endif
-      }
-
-      for (int j = 0; j < kNumCoeffs; ++j) {
-#if CONFIG_HIGHBITDEPTH
-        const int diff =
-            bit_depth_ == AOM_BITS_8 ? dst[j] - ref[j] : dst16[j] - ref16[j];
-#else
-        const int diff = dst[j] - ref[j];
-#endif
-        const uint32_t error = diff * diff;
-        EXPECT_EQ(0u, error)
-            << "Error: 8x8 IDCT has error " << error << " at index " << j;
-      }
-    }
-  }
-  int pitch_;
-  FhtFunc fwd_txfm_ref;
-  aom_bit_depth_t bit_depth_;
-  int mask_;
-  TxfmParam txfm_param_;
-};
-
-class FwdTrans8x8DCT : public FwdTrans8x8TestBase,
-                       public ::testing::TestWithParam<Dct8x8Param> {
- public:
-  virtual ~FwdTrans8x8DCT() {}
-
-  virtual void SetUp() {
-    fwd_txfm_ = GET_PARAM(0);
-    inv_txfm_ = GET_PARAM(1);
-    pitch_ = 8;
-    fwd_txfm_ref = fdct8x8_ref;
-    bit_depth_ = GET_PARAM(3);
-    mask_ = (1 << bit_depth_) - 1;
-    txfm_param_.tx_type = GET_PARAM(2);
-  }
-
-  virtual void TearDown() { libaom_test::ClearSystemState(); }
-
- protected:
-  void RunFwdTxfm(int16_t *in, tran_low_t *out, int stride) {
-    fwd_txfm_(in, out, stride);
-  }
-  void RunInvTxfm(tran_low_t *out, uint8_t *dst, int stride) {
-    inv_txfm_(out, dst, stride);
-  }
-
-  FdctFunc fwd_txfm_;
-  IdctFunc inv_txfm_;
-};
-
-TEST_P(FwdTrans8x8DCT, SignBiasCheck) { RunSignBiasCheck(); }
-
-TEST_P(FwdTrans8x8DCT, RoundTripErrorCheck) { RunRoundTripErrorCheck(); }
-
-TEST_P(FwdTrans8x8DCT, ExtremalCheck) { RunExtremalCheck(); }
-
-TEST_P(FwdTrans8x8DCT, FwdAccuracyCheck) { RunFwdAccuracyCheck(); }
-
-TEST_P(FwdTrans8x8DCT, InvAccuracyCheck) { RunInvAccuracyCheck(); }
-
-class FwdTrans8x8HT : public FwdTrans8x8TestBase,
-                      public ::testing::TestWithParam<Ht8x8Param> {
- public:
-  virtual ~FwdTrans8x8HT() {}
-
-  virtual void SetUp() {
-    fwd_txfm_ = GET_PARAM(0);
-    inv_txfm_ = GET_PARAM(1);
-    pitch_ = 8;
-    fwd_txfm_ref = fht8x8_ref;
-    bit_depth_ = GET_PARAM(3);
-    mask_ = (1 << bit_depth_) - 1;
-    txfm_param_.tx_type = GET_PARAM(2);
-#if CONFIG_HIGHBITDEPTH
-    switch (bit_depth_) {
-      case AOM_BITS_10: fwd_txfm_ref = fht8x8_10; break;
-      case AOM_BITS_12: fwd_txfm_ref = fht8x8_12; break;
-      default: fwd_txfm_ref = fht8x8_ref; break;
-    }
-#endif
-  }
-
-  virtual void TearDown() { libaom_test::ClearSystemState(); }
-
- protected:
-  void RunFwdTxfm(int16_t *in, tran_low_t *out, int stride) {
-    fwd_txfm_(in, out, stride, &txfm_param_);
-  }
-  void RunInvTxfm(tran_low_t *out, uint8_t *dst, int stride) {
-    inv_txfm_(out, dst, stride, &txfm_param_);
-  }
-
-  FhtFunc fwd_txfm_;
-  IhtFunc inv_txfm_;
-};
-
-TEST_P(FwdTrans8x8HT, SignBiasCheck) { RunSignBiasCheck(); }
-
-TEST_P(FwdTrans8x8HT, RoundTripErrorCheck) { RunRoundTripErrorCheck(); }
-
-TEST_P(FwdTrans8x8HT, ExtremalCheck) { RunExtremalCheck(); }
-
-class InvTrans8x8DCT : public FwdTrans8x8TestBase,
-                       public ::testing::TestWithParam<Idct8x8Param> {
- public:
-  virtual ~InvTrans8x8DCT() {}
-
-  virtual void SetUp() {
-    ref_txfm_ = GET_PARAM(0);
-    inv_txfm_ = GET_PARAM(1);
-    thresh_ = GET_PARAM(2);
-    pitch_ = 8;
-    bit_depth_ = GET_PARAM(3);
-    mask_ = (1 << bit_depth_) - 1;
-  }
-
-  virtual void TearDown() { libaom_test::ClearSystemState(); }
-
- protected:
-  void RunInvTxfm(tran_low_t *out, uint8_t *dst, int stride) {
-    inv_txfm_(out, dst, stride);
-  }
-  void RunFwdTxfm(int16_t * /*out*/, tran_low_t * /*dst*/, int /*stride*/) {}
-
-  IdctFunc ref_txfm_;
-  IdctFunc inv_txfm_;
-  int thresh_;
-};
-
-TEST_P(InvTrans8x8DCT, CompareReference) {
-  CompareInvReference(ref_txfm_, thresh_);
-}
-
-using std::tr1::make_tuple;
-
-#if CONFIG_HIGHBITDEPTH
-INSTANTIATE_TEST_CASE_P(C, FwdTrans8x8DCT,
-                        ::testing::Values(make_tuple(&aom_fdct8x8_c,
-                                                     &aom_idct8x8_64_add_c,
-                                                     DCT_DCT, AOM_BITS_8)));
-#else
-INSTANTIATE_TEST_CASE_P(C, FwdTrans8x8DCT,
-                        ::testing::Values(make_tuple(&aom_fdct8x8_c,
-                                                     &aom_idct8x8_64_add_c,
-                                                     DCT_DCT, AOM_BITS_8)));
-#endif  // CONFIG_HIGHBITDEPTH
-
-#if CONFIG_HIGHBITDEPTH
-INSTANTIATE_TEST_CASE_P(
-    C, FwdTrans8x8HT,
-    ::testing::Values(
-        make_tuple(&av1_fht8x8_c, &av1_iht8x8_64_add_c, DCT_DCT, AOM_BITS_8),
-        make_tuple(&fht8x8_10, &iht8x8_10, DCT_DCT, AOM_BITS_10),
-        make_tuple(&fht8x8_10, &iht8x8_10, ADST_DCT, AOM_BITS_10),
-        make_tuple(&fht8x8_10, &iht8x8_10, DCT_ADST, AOM_BITS_10),
-        make_tuple(&fht8x8_10, &iht8x8_10, ADST_ADST, AOM_BITS_10),
-        make_tuple(&fht8x8_12, &iht8x8_12, DCT_DCT, AOM_BITS_12),
-        make_tuple(&fht8x8_12, &iht8x8_12, ADST_DCT, AOM_BITS_12),
-        make_tuple(&fht8x8_12, &iht8x8_12, DCT_ADST, AOM_BITS_12),
-        make_tuple(&fht8x8_12, &iht8x8_12, ADST_ADST, AOM_BITS_12),
-        make_tuple(&av1_fht8x8_c, &av1_iht8x8_64_add_c, ADST_DCT, AOM_BITS_8),
-        make_tuple(&av1_fht8x8_c, &av1_iht8x8_64_add_c, DCT_ADST, AOM_BITS_8),
-        make_tuple(&av1_fht8x8_c, &av1_iht8x8_64_add_c, ADST_ADST,
-                   AOM_BITS_8)));
-#else
-INSTANTIATE_TEST_CASE_P(
-    C, FwdTrans8x8HT,
-    ::testing::Values(
-        make_tuple(&av1_fht8x8_c, &av1_iht8x8_64_add_c, DCT_DCT, AOM_BITS_8),
-        make_tuple(&av1_fht8x8_c, &av1_iht8x8_64_add_c, ADST_DCT, AOM_BITS_8),
-        make_tuple(&av1_fht8x8_c, &av1_iht8x8_64_add_c, DCT_ADST, AOM_BITS_8),
-        make_tuple(&av1_fht8x8_c, &av1_iht8x8_64_add_c, ADST_ADST,
-                   AOM_BITS_8)));
-#endif  // CONFIG_HIGHBITDEPTH
-
-#if HAVE_NEON_ASM && !CONFIG_HIGHBITDEPTH
-INSTANTIATE_TEST_CASE_P(NEON, FwdTrans8x8DCT,
-                        ::testing::Values(make_tuple(&aom_fdct8x8_neon,
-                                                     &aom_idct8x8_64_add_neon,
-                                                     DCT_DCT, AOM_BITS_8)));
-#endif  // HAVE_NEON_ASM && !CONFIG_HIGHBITDEPTH
-
-#if HAVE_NEON && !CONFIG_HIGHBITDEPTH
-INSTANTIATE_TEST_CASE_P(
-    NEON, FwdTrans8x8HT,
-    ::testing::Values(make_tuple(&av1_fht8x8_c, &av1_iht8x8_64_add_neon,
-                                 DCT_DCT, AOM_BITS_8),
-                      make_tuple(&av1_fht8x8_c, &av1_iht8x8_64_add_neon,
-                                 ADST_DCT, AOM_BITS_8),
-                      make_tuple(&av1_fht8x8_c, &av1_iht8x8_64_add_neon,
-                                 DCT_ADST, AOM_BITS_8),
-                      make_tuple(&av1_fht8x8_c, &av1_iht8x8_64_add_neon,
-                                 ADST_ADST, AOM_BITS_8)));
-#endif  // HAVE_NEON && !CONFIG_HIGHBITDEPTH
-
-#if HAVE_SSE2 && !CONFIG_HIGHBITDEPTH
-INSTANTIATE_TEST_CASE_P(SSE2, FwdTrans8x8DCT,
-                        ::testing::Values(make_tuple(&aom_fdct8x8_sse2,
-                                                     &aom_idct8x8_64_add_sse2,
-                                                     DCT_DCT, AOM_BITS_8)));
-#if !CONFIG_DAALA_DCT8
-INSTANTIATE_TEST_CASE_P(
-    SSE2, FwdTrans8x8HT,
-    ::testing::Values(make_tuple(&av1_fht8x8_sse2, &av1_iht8x8_64_add_sse2,
-                                 DCT_DCT, AOM_BITS_8),
-                      make_tuple(&av1_fht8x8_sse2, &av1_iht8x8_64_add_sse2,
-                                 ADST_DCT, AOM_BITS_8),
-                      make_tuple(&av1_fht8x8_sse2, &av1_iht8x8_64_add_sse2,
-                                 DCT_ADST, AOM_BITS_8),
-                      make_tuple(&av1_fht8x8_sse2, &av1_iht8x8_64_add_sse2,
-                                 ADST_ADST, AOM_BITS_8)));
-#endif  // !CONFIG_DAALA_DCT8
-#endif  // HAVE_SSE2 && !CONFIG_HIGHBITDEPTH
-
-#if HAVE_SSE2 && CONFIG_HIGHBITDEPTH
-INSTANTIATE_TEST_CASE_P(SSE2, FwdTrans8x8DCT,
-                        ::testing::Values(make_tuple(&aom_fdct8x8_sse2,
-                                                     &aom_idct8x8_64_add_c,
-                                                     DCT_DCT, AOM_BITS_8)));
-#if !CONFIG_DAALA_DCT8
-INSTANTIATE_TEST_CASE_P(
-    SSE2, FwdTrans8x8HT,
-    ::testing::Values(make_tuple(&av1_fht8x8_sse2, &av1_iht8x8_64_add_c,
-                                 DCT_DCT, AOM_BITS_8),
-                      make_tuple(&av1_fht8x8_sse2, &av1_iht8x8_64_add_c,
-                                 ADST_DCT, AOM_BITS_8),
-                      make_tuple(&av1_fht8x8_sse2, &av1_iht8x8_64_add_c,
-                                 DCT_ADST, AOM_BITS_8),
-                      make_tuple(&av1_fht8x8_sse2, &av1_iht8x8_64_add_c,
-                                 ADST_ADST, AOM_BITS_8)));
-#endif  // !CONFIG_DAALA_DCT8
-#endif  // HAVE_SSE2 && CONFIG_HIGHBITDEPTH
-
-#if HAVE_SSSE3 && ARCH_X86_64
-INSTANTIATE_TEST_CASE_P(SSSE3, FwdTrans8x8DCT,
-                        ::testing::Values(make_tuple(&aom_fdct8x8_ssse3,
-                                                     &aom_idct8x8_64_add_ssse3,
-                                                     DCT_DCT, AOM_BITS_8)));
-#endif
-
-#if HAVE_MSA && !CONFIG_HIGHBITDEPTH
-INSTANTIATE_TEST_CASE_P(MSA, FwdTrans8x8DCT,
-                        ::testing::Values(make_tuple(&aom_fdct8x8_msa,
-                                                     &aom_idct8x8_64_add_msa,
-                                                     DCT_DCT, AOM_BITS_8)));
-#if !CONFIG_EXT_TX && !CONFIG_DAALA_DCT8
-INSTANTIATE_TEST_CASE_P(
-    MSA, FwdTrans8x8HT,
-    ::testing::Values(make_tuple(&av1_fht8x8_msa, &av1_iht8x8_64_add_msa,
-                                 DCT_DCT, AOM_BITS_8),
-                      make_tuple(&av1_fht8x8_msa, &av1_iht8x8_64_add_msa,
-                                 ADST_DCT, AOM_BITS_8),
-                      make_tuple(&av1_fht8x8_msa, &av1_iht8x8_64_add_msa,
-                                 DCT_ADST, AOM_BITS_8),
-                      make_tuple(&av1_fht8x8_msa, &av1_iht8x8_64_add_msa,
-                                 ADST_ADST, AOM_BITS_8)));
-#endif  // !CONFIG_EXT_TX && !CONFIG_DAALA_DCT8
-#endif  // HAVE_MSA && !CONFIG_HIGHBITDEPTH
-}  // namespace

third_party/aom/test/fft_test.cc

@@ -0,0 +1,263 @@
+#include <math.h>
+
+#include <algorithm>
+#include <complex>
+#include <vector>
+
+#include "aom_dsp/fft_common.h"
+#include "aom_mem/aom_mem.h"
+#if ARCH_X86 || ARCH_X86_64
+#include "aom_ports/x86.h"
+#endif
+#include "av1/common/common.h"
+#include "config/aom_dsp_rtcd.h"
+#include "test/acm_random.h"
+#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
+
+namespace {
+
+typedef void (*tform_fun_t)(const float *input, float *temp, float *output);
+
+// Simple 1D FFT implementation
+template <typename InputType>
+void fft(const InputType *data, std::complex<float> *result, int n) {
+  if (n == 1) {
+    result[0] = data[0];
+    return;
+  }
+  std::vector<InputType> temp(n);
+  for (int k = 0; k < n / 2; ++k) {
+    temp[k] = data[2 * k];
+    temp[n / 2 + k] = data[2 * k + 1];
+  }
+  fft(&temp[0], result, n / 2);
+  fft(&temp[n / 2], result + n / 2, n / 2);
+  for (int k = 0; k < n / 2; ++k) {
+    std::complex<float> w = std::complex<float>((float)cos(2. * PI * k / n),
+                                                (float)-sin(2. * PI * k / n));
+    std::complex<float> a = result[k];
+    std::complex<float> b = result[n / 2 + k];
+    result[k] = a + w * b;
+    result[n / 2 + k] = a - w * b;
+  }
+}
+
+void transpose(std::vector<std::complex<float> > *data, int n) {
+  for (int y = 0; y < n; ++y) {
+    for (int x = y + 1; x < n; ++x) {
+      std::swap((*data)[y * n + x], (*data)[x * n + y]);
+    }
+  }
+}
+
+// Simple 2D FFT implementation
+template <class InputType>
+std::vector<std::complex<float> > fft2d(const InputType *input, int n) {
+  std::vector<std::complex<float> > rowfft(n * n);
+  std::vector<std::complex<float> > result(n * n);
+  for (int y = 0; y < n; ++y) {
+    fft(input + y * n, &rowfft[y * n], n);
+  }
+  transpose(&rowfft, n);
+  for (int y = 0; y < n; ++y) {
+    fft(&rowfft[y * n], &result[y * n], n);
+  }
+  transpose(&result, n);
+  return result;
+}
+
+struct FFTTestArg {
+  int n;
+  void (*fft)(const float *input, float *temp, float *output);
+  int flag;
+  FFTTestArg(int n_in, tform_fun_t fft_in, int flag_in)
+      : n(n_in), fft(fft_in), flag(flag_in) {}
+};
+
+std::ostream &operator<<(std::ostream &os, const FFTTestArg &test_arg) {
+  return os << "fft_arg { n:" << test_arg.n << " fft:" << test_arg.fft
+            << " flag:" << test_arg.flag << "}";
+}
+
+class FFT2DTest : public ::testing::TestWithParam<FFTTestArg> {
+ protected:
+  void SetUp() {
+    int n = GetParam().n;
+    input_ = (float *)aom_memalign(32, sizeof(*input_) * n * n);
+    temp_ = (float *)aom_memalign(32, sizeof(*temp_) * n * n);
+    output_ = (float *)aom_memalign(32, sizeof(*output_) * n * n * 2);
+    memset(input_, 0, sizeof(*input_) * n * n);
+    memset(temp_, 0, sizeof(*temp_) * n * n);
+    memset(output_, 0, sizeof(*output_) * n * n * 2);
+#if ARCH_X86 || ARCH_X86_64
+    disabled_ = GetParam().flag != 0 && !(x86_simd_caps() & GetParam().flag);
+#else
+    disabled_ = GetParam().flag != 0;
+#endif
+  }
+  void TearDown() {
+    aom_free(input_);
+    aom_free(temp_);
+    aom_free(output_);
+  }
+  int disabled_;
+  float *input_;
+  float *temp_;
+  float *output_;
+};
+
+TEST_P(FFT2DTest, Correct) {
+  if (disabled_) return;
+
+  int n = GetParam().n;
+  for (int i = 0; i < n * n; ++i) {
+    input_[i] = 1;
+    std::vector<std::complex<float> > expected = fft2d<float>(&input_[0], n);
+    GetParam().fft(&input_[0], &temp_[0], &output_[0]);
+    for (int y = 0; y < n; ++y) {
+      for (int x = 0; x < (n / 2) + 1; ++x) {
+        EXPECT_NEAR(expected[y * n + x].real(), output_[2 * (y * n + x)], 1e-5);
+        EXPECT_NEAR(expected[y * n + x].imag(), output_[2 * (y * n + x) + 1],
+                    1e-5);
+      }
+    }
+    input_[i] = 0;
+  }
+}
+
+TEST_P(FFT2DTest, Benchmark) {
+  if (disabled_) return;
+
+  int n = GetParam().n;
+  float sum = 0;
+  for (int i = 0; i < 1000 * (64 - n); ++i) {
+    input_[i % (n * n)] = 1;
+    GetParam().fft(&input_[0], &temp_[0], &output_[0]);
+    sum += output_[0];
+    input_[i % (n * n)] = 0;
+  }
+}
+
+INSTANTIATE_TEST_CASE_P(
+    FFT2DTestC, FFT2DTest,
+    ::testing::Values(FFTTestArg(2, aom_fft2x2_float_c, 0),
+                      FFTTestArg(4, aom_fft4x4_float_c, 0),
+                      FFTTestArg(8, aom_fft8x8_float_c, 0),
+                      FFTTestArg(16, aom_fft16x16_float_c, 0),
+                      FFTTestArg(32, aom_fft32x32_float_c, 0)));
+#if ARCH_X86 || ARCH_X86_64
+INSTANTIATE_TEST_CASE_P(
+    FFT2DTestSSE2, FFT2DTest,
+    ::testing::Values(FFTTestArg(4, aom_fft4x4_float_sse2, HAS_SSE2),
+                      FFTTestArg(8, aom_fft8x8_float_sse2, HAS_SSE2),
+                      FFTTestArg(16, aom_fft16x16_float_sse2, HAS_SSE2),
+                      FFTTestArg(32, aom_fft32x32_float_sse2, HAS_SSE2)));
+
+INSTANTIATE_TEST_CASE_P(
+    FFT2DTestAVX2, FFT2DTest,
+    ::testing::Values(FFTTestArg(8, aom_fft8x8_float_avx2, HAS_AVX2),
+                      FFTTestArg(16, aom_fft16x16_float_avx2, HAS_AVX2),
+                      FFTTestArg(32, aom_fft32x32_float_avx2, HAS_AVX2)));
+#endif
+
+struct IFFTTestArg {
+  int n;
+  tform_fun_t ifft;
+  int flag;
+  IFFTTestArg(int n_in, tform_fun_t ifft_in, int flag_in)
+      : n(n_in), ifft(ifft_in), flag(flag_in) {}
+};
+
+std::ostream &operator<<(std::ostream &os, const IFFTTestArg &test_arg) {
+  return os << "ifft_arg { n:" << test_arg.n << " fft:" << test_arg.ifft
+            << " flag:" << test_arg.flag << "}";
+}
+
+class IFFT2DTest : public ::testing::TestWithParam<IFFTTestArg> {
+ protected:
+  void SetUp() {
+    int n = GetParam().n;
+    input_ = (float *)aom_memalign(32, sizeof(*input_) * n * n * 2);
+    temp_ = (float *)aom_memalign(32, sizeof(*temp_) * n * n * 2);
+    output_ = (float *)aom_memalign(32, sizeof(*output_) * n * n);
+    memset(input_, 0, sizeof(*input_) * n * n * 2);
+    memset(temp_, 0, sizeof(*temp_) * n * n * 2);
+    memset(output_, 0, sizeof(*output_) * n * n);
+#if ARCH_X86 || ARCH_X86_64
+    disabled_ = GetParam().flag != 0 && !(x86_simd_caps() & GetParam().flag);
+#else
+    disabled_ = GetParam().flag != 0;
+#endif
+  }
+  void TearDown() {
+    aom_free(input_);
+    aom_free(temp_);
+    aom_free(output_);
+  }
+  int disabled_;
+  float *input_;
+  float *temp_;
+  float *output_;
+};
+
+TEST_P(IFFT2DTest, Correctness) {
+  if (disabled_) return;
+  int n = GetParam().n;
+  ASSERT_GE(n, 2);
+  std::vector<float> expected(n * n);
+  std::vector<float> actual(n * n);
+  // Do forward transform then invert to make sure we get back expected
+  for (int y = 0; y < n; ++y) {
+    for (int x = 0; x < n; ++x) {
+      expected[y * n + x] = 1;
+      std::vector<std::complex<float> > input_c = fft2d(&expected[0], n);
+      for (int i = 0; i < n * n; ++i) {
+        input_[2 * i + 0] = input_c[i].real();
+        input_[2 * i + 1] = input_c[i].imag();
+      }
+      GetParam().ifft(&input_[0], &temp_[0], &output_[0]);
+
+      for (int yy = 0; yy < n; ++yy) {
+        for (int xx = 0; xx < n; ++xx) {
+          EXPECT_NEAR(expected[yy * n + xx], output_[yy * n + xx] / (n * n),
+                      1e-5);
+        }
+      }
+      expected[y * n + x] = 0;
+    }
+  }
+};
+
+TEST_P(IFFT2DTest, Benchmark) {
+  if (disabled_) return;
+  int n = GetParam().n;
+  float sum = 0;
+  for (int i = 0; i < 1000 * (64 - n); ++i) {
+    input_[i % (n * n)] = 1;
+    GetParam().ifft(&input_[0], &temp_[0], &output_[0]);
+    sum += output_[0];
+    input_[i % (n * n)] = 0;
+  }
+}
+INSTANTIATE_TEST_CASE_P(
+    IFFT2DTestC, IFFT2DTest,
+    ::testing::Values(IFFTTestArg(2, aom_ifft2x2_float_c, 0),
+                      IFFTTestArg(4, aom_ifft4x4_float_c, 0),
+                      IFFTTestArg(8, aom_ifft8x8_float_c, 0),
+                      IFFTTestArg(16, aom_ifft16x16_float_c, 0),
+                      IFFTTestArg(32, aom_ifft32x32_float_c, 0)));
+#if ARCH_X86 || ARCH_X86_64
+INSTANTIATE_TEST_CASE_P(
+    IFFT2DTestSSE2, IFFT2DTest,
+    ::testing::Values(IFFTTestArg(4, aom_ifft4x4_float_sse2, HAS_SSE2),
+                      IFFTTestArg(8, aom_ifft8x8_float_sse2, HAS_SSE2),
+                      IFFTTestArg(16, aom_ifft16x16_float_sse2, HAS_SSE2),
+                      IFFTTestArg(32, aom_ifft32x32_float_sse2, HAS_SSE2)));
+
+INSTANTIATE_TEST_CASE_P(
+    IFFT2DTestAVX2, IFFT2DTest,
+    ::testing::Values(IFFTTestArg(8, aom_ifft8x8_float_avx2, HAS_AVX2),
+                      IFFTTestArg(16, aom_ifft16x16_float_avx2, HAS_AVX2),
+                      IFFTTestArg(32, aom_ifft32x32_float_avx2, HAS_AVX2)));
+#endif
+}  // namespace

third_party/aom/test/film_grain_table_test.cc

@@ -0,0 +1,239 @@
+#include <string>
+#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
+#include "aom_dsp/grain_table.h"
+#include "aom/internal/aom_codec_internal.h"
+#include "av1/encoder/grain_test_vectors.h"
+#include "test/video_source.h"
+
+void grain_equal(const aom_film_grain_t *expected,
+                 const aom_film_grain_t *actual) {
+  EXPECT_EQ(expected->apply_grain, actual->apply_grain);
+  EXPECT_EQ(expected->update_parameters, actual->update_parameters);
+  if (!expected->update_parameters) return;
+  EXPECT_EQ(expected->num_y_points, actual->num_y_points);
+  EXPECT_EQ(expected->num_cb_points, actual->num_cb_points);
+  EXPECT_EQ(expected->num_cr_points, actual->num_cr_points);
+  EXPECT_EQ(0, memcmp(expected->scaling_points_y, actual->scaling_points_y,
+                      expected->num_y_points *
+                          sizeof(expected->scaling_points_y[0])));
+  EXPECT_EQ(0, memcmp(expected->scaling_points_cb, actual->scaling_points_cb,
+                      expected->num_cb_points *
+                          sizeof(expected->scaling_points_cb[0])));
+  EXPECT_EQ(0, memcmp(expected->scaling_points_cr, actual->scaling_points_cr,
+                      expected->num_cr_points *
+                          sizeof(expected->scaling_points_cr[0])));
+  EXPECT_EQ(expected->scaling_shift, actual->scaling_shift);
+  EXPECT_EQ(expected->ar_coeff_lag, actual->ar_coeff_lag);
+  EXPECT_EQ(expected->ar_coeff_shift, actual->ar_coeff_shift);
+
+  const int num_pos_luma =
+      2 * expected->ar_coeff_lag * (expected->ar_coeff_lag + 1);
+  const int num_pos_chroma = num_pos_luma;
+  EXPECT_EQ(0, memcmp(expected->ar_coeffs_y, actual->ar_coeffs_y,
+                      sizeof(expected->ar_coeffs_y[0]) * num_pos_luma));
+  if (actual->num_cb_points || actual->chroma_scaling_from_luma) {
+    EXPECT_EQ(0, memcmp(expected->ar_coeffs_cb, actual->ar_coeffs_cb,
+                        sizeof(expected->ar_coeffs_cb[0]) * num_pos_chroma));
+  }
+  if (actual->num_cr_points || actual->chroma_scaling_from_luma) {
+    EXPECT_EQ(0, memcmp(expected->ar_coeffs_cr, actual->ar_coeffs_cr,
+                        sizeof(expected->ar_coeffs_cr[0]) * num_pos_chroma));
+  }
+  EXPECT_EQ(expected->overlap_flag, actual->overlap_flag);
+  EXPECT_EQ(expected->chroma_scaling_from_luma,
+            actual->chroma_scaling_from_luma);
+  EXPECT_EQ(expected->grain_scale_shift, actual->grain_scale_shift);
+  // EXPECT_EQ(expected->random_seed, actual->random_seed);
+
+  // clip_to_restricted and bit_depth aren't written
+  if (expected->num_cb_points) {
+    EXPECT_EQ(expected->cb_mult, actual->cb_mult);
+    EXPECT_EQ(expected->cb_luma_mult, actual->cb_luma_mult);
+    EXPECT_EQ(expected->cb_offset, actual->cb_offset);
+  }
+  if (expected->num_cr_points) {
+    EXPECT_EQ(expected->cr_mult, actual->cr_mult);
+    EXPECT_EQ(expected->cr_luma_mult, actual->cr_luma_mult);
+    EXPECT_EQ(expected->cr_offset, actual->cr_offset);
+  }
+}
+
+TEST(FilmGrainTableTest, AddAndLookupSingleSegment) {
+  aom_film_grain_table_t table;
+  memset(&table, 0, sizeof(table));
+
+  aom_film_grain_t grain;
+  EXPECT_FALSE(aom_film_grain_table_lookup(&table, 0, 1000, false, &grain));
+
+  aom_film_grain_table_append(&table, 1000, 2000, film_grain_test_vectors + 0);
+  EXPECT_FALSE(aom_film_grain_table_lookup(&table, 0, 1000, false, &grain));
+  EXPECT_FALSE(aom_film_grain_table_lookup(&table, 2000, 3000, false, &grain));
+
+  EXPECT_TRUE(aom_film_grain_table_lookup(&table, 1000, 2000, false, &grain));
+
+  grain.bit_depth = film_grain_test_vectors[0].bit_depth;
+  EXPECT_EQ(0, memcmp(&grain, film_grain_test_vectors + 0, sizeof(table)));
+
+  // Extend the existing segment
+  aom_film_grain_table_append(&table, 2000, 3000, film_grain_test_vectors + 0);
+  EXPECT_EQ(0, table.head->next);
+
+  // Lookup and remove and check that the entry is no longer there
+  EXPECT_TRUE(aom_film_grain_table_lookup(&table, 1000, 2000, true, &grain));
+  EXPECT_FALSE(aom_film_grain_table_lookup(&table, 1000, 2000, false, &grain));
+
+  EXPECT_TRUE(aom_film_grain_table_lookup(&table, 2000, 3000, true, &grain));
+  EXPECT_FALSE(aom_film_grain_table_lookup(&table, 2000, 3000, false, &grain));
+
+  EXPECT_EQ(0, table.head);
+  EXPECT_EQ(0, table.tail);
+  aom_film_grain_table_free(&table);
+}
+
+TEST(FilmGrainTableTest, SplitSingleSegment) {
+  aom_film_grain_table_t table;
+  aom_film_grain_t grain;
+  memset(&table, 0, sizeof(table));
+
+  aom_film_grain_table_append(&table, 0, 1000, film_grain_test_vectors + 0);
+
+  // Test lookup and remove that adjusts start time
+  EXPECT_TRUE(aom_film_grain_table_lookup(&table, 0, 100, true, &grain));
+  EXPECT_EQ(NULL, table.head->next);
+  EXPECT_EQ(100, table.head->start_time);
+
+  // Test lookup and remove that adjusts end time
+  EXPECT_TRUE(aom_film_grain_table_lookup(&table, 900, 1000, true, &grain));
+  EXPECT_EQ(NULL, table.head->next);
+  EXPECT_EQ(100, table.head->start_time);
+  EXPECT_EQ(900, table.head->end_time);
+
+  // Test lookup and remove that splits the first entry
+  EXPECT_TRUE(aom_film_grain_table_lookup(&table, 400, 600, true, &grain));
+  EXPECT_EQ(100, table.head->start_time);
+  EXPECT_EQ(400, table.head->end_time);
+
+  ASSERT_NE((void *)NULL, table.head->next);
+  EXPECT_EQ(table.tail, table.head->next);
+  EXPECT_EQ(600, table.head->next->start_time);
+  EXPECT_EQ(900, table.head->next->end_time);
+
+  aom_film_grain_table_free(&table);
+}
+
+TEST(FilmGrainTableTest, AddAndLookupMultipleSegments) {
+  aom_film_grain_table_t table;
+  memset(&table, 0, sizeof(table));
+
+  aom_film_grain_t grain;
+  const int kNumTestVectors =
+      sizeof(film_grain_test_vectors) / sizeof(film_grain_test_vectors[0]);
+  for (int i = 0; i < kNumTestVectors; ++i) {
+    aom_film_grain_table_append(&table, i * 1000, (i + 1) * 1000,
+                                film_grain_test_vectors + i);
+  }
+
+  for (int i = kNumTestVectors - 1; i >= 0; --i) {
+    EXPECT_TRUE(aom_film_grain_table_lookup(&table, i * 1000, (i + 1) * 1000,
+                                            true, &grain));
+    grain_equal(film_grain_test_vectors + i, &grain);
+    EXPECT_FALSE(aom_film_grain_table_lookup(&table, i * 1000, (i + 1) * 1000,
+                                             true, &grain));
+  }
+
+  // Verify that all the data has been removed
+  for (int i = 0; i < kNumTestVectors; ++i) {
+    EXPECT_FALSE(aom_film_grain_table_lookup(&table, i * 1000, (i + 1) * 1000,
+                                             true, &grain));
+  }
+  aom_film_grain_table_free(&table);
+}
+
+class FilmGrainTableIOTest : public ::testing::Test {
+ protected:
+  void SetUp() { memset(&error_, 0, sizeof(error_)); }
+  struct aom_internal_error_info error_;
+};
+
+TEST_F(FilmGrainTableIOTest, ReadMissingFile) {
+  aom_film_grain_table_t table;
+  memset(&table, 0, sizeof(table));
+  ASSERT_EQ(AOM_CODEC_ERROR, aom_film_grain_table_read(
+                                 &table, "/path/to/missing/file", &error_));
+}
+
+TEST_F(FilmGrainTableIOTest, ReadTruncatedFile) {
+  aom_film_grain_table_t table;
+  memset(&table, 0, sizeof(table));
+
+  std::string grain_file;
+  FILE *file = libaom_test::GetTempOutFile(&grain_file);
+  fwrite("deadbeef", 8, 1, file);
+  fclose(file);
+  ASSERT_EQ(AOM_CODEC_ERROR,
+            aom_film_grain_table_read(&table, grain_file.c_str(), &error_));
+  EXPECT_EQ(0, remove(grain_file.c_str()));
+}
+
+TEST_F(FilmGrainTableIOTest, RoundTripReadWrite) {
+  aom_film_grain_table_t table;
+  memset(&table, 0, sizeof(table));
+
+  aom_film_grain_t expected_grain[16];
+  const int kNumTestVectors =
+      sizeof(film_grain_test_vectors) / sizeof(film_grain_test_vectors[0]);
+  for (int i = 0; i < kNumTestVectors; ++i) {
+    expected_grain[i] = film_grain_test_vectors[i];
+    expected_grain[i].random_seed = i;
+    expected_grain[i].update_parameters = i % 2;
+    expected_grain[i].apply_grain = (i + 1) % 2;
+    expected_grain[i].bit_depth = 0;
+    aom_film_grain_table_append(&table, i * 1000, (i + 1) * 1000,
+                                expected_grain + i);
+  }
+  std::string grain_file;
+  fclose(libaom_test::GetTempOutFile(&grain_file));
+  ASSERT_EQ(AOM_CODEC_OK,
+            aom_film_grain_table_write(&table, grain_file.c_str(), &error_));
+  aom_film_grain_table_free(&table);
+
+  memset(&table, 0, sizeof(table));
+  ASSERT_EQ(AOM_CODEC_OK,
+            aom_film_grain_table_read(&table, grain_file.c_str(), &error_));
+  for (int i = 0; i < kNumTestVectors; ++i) {
+    aom_film_grain_t grain;
+    EXPECT_TRUE(aom_film_grain_table_lookup(&table, i * 1000, (i + 1) * 1000,
+                                            true, &grain));
+    grain_equal(expected_grain + i, &grain);
+  }
+  aom_film_grain_table_free(&table);
+  EXPECT_EQ(0, remove(grain_file.c_str()));
+}
+
+TEST_F(FilmGrainTableIOTest, RoundTripSplit) {
+  std::string grain_file;
+  fclose(libaom_test::GetTempOutFile(&grain_file));
+
+  aom_film_grain_table_t table;
+  memset(&table, 0, sizeof(table));
+
+  aom_film_grain_t grain = film_grain_test_vectors[0];
+  aom_film_grain_table_append(&table, 0, 3000, &grain);
+  ASSERT_TRUE(aom_film_grain_table_lookup(&table, 1000, 2000, true, &grain));
+  ASSERT_TRUE(aom_film_grain_table_lookup(&table, 0, 1000, false, &grain));
+  EXPECT_FALSE(aom_film_grain_table_lookup(&table, 1000, 2000, false, &grain));
+  ASSERT_TRUE(aom_film_grain_table_lookup(&table, 2000, 3000, false, &grain));
+  ASSERT_EQ(AOM_CODEC_OK,
+            aom_film_grain_table_write(&table, grain_file.c_str(), &error_));
+  aom_film_grain_table_free(&table);
+
+  memset(&table, 0, sizeof(table));
+  ASSERT_EQ(AOM_CODEC_OK,
+            aom_film_grain_table_read(&table, grain_file.c_str(), &error_));
+  ASSERT_TRUE(aom_film_grain_table_lookup(&table, 0, 1000, false, &grain));
+  ASSERT_FALSE(aom_film_grain_table_lookup(&table, 1000, 2000, false, &grain));
+  ASSERT_TRUE(aom_film_grain_table_lookup(&table, 2000, 3000, false, &grain));
+  aom_film_grain_table_free(&table);
+
+  EXPECT_EQ(0, remove(grain_file.c_str()));
+}

third_party/aom/test/filterintra_predictors_test.cc

@@ -1,331 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
-
-#include "./av1_rtcd.h"
-#include "test/acm_random.h"
-#include "test/clear_system_state.h"
-#include "test/register_state_check.h"
-#include "test/util.h"
-#include "av1/common/enums.h"
-
-namespace {
-
-using std::tr1::tuple;
-using libaom_test::ACMRandom;
-
-typedef void (*Predictor)(uint8_t *dst, ptrdiff_t stride, int bs,
-                          const uint8_t *above, const uint8_t *left);
-
-// Note:
-//  Test parameter list:
-//  Reference predictor, optimized predictor, prediction mode, block size
-//
-typedef tuple<Predictor, Predictor, int> PredFuncMode;
-typedef tuple<PredFuncMode, int> PredParams;
-
-#if CONFIG_HIGHBITDEPTH
-typedef void (*HbdPredictor)(uint16_t *dst, ptrdiff_t stride, int bs,
-                             const uint16_t *above, const uint16_t *left,
-                             int bd);
-
-// Note:
-//  Test parameter list:
-//  Reference predictor, optimized predictor, prediction mode, block size,
-//  bit depth
-//
-typedef tuple<HbdPredictor, HbdPredictor, int> HbdPredFuncMode;
-typedef tuple<HbdPredFuncMode, int, int> HbdPredParams;
-#endif
-
-const int MaxBlkSize = 32;
-
-// By default, disable speed test
-#define PREDICTORS_SPEED_TEST (0)
-
-#if PREDICTORS_SPEED_TEST
-const int MaxTestNum = 100000;
-#else
-const int MaxTestNum = 100;
-#endif
-
-class AV1FilterIntraPredOptimzTest
-    : public ::testing::TestWithParam<PredParams> {
- public:
-  virtual ~AV1FilterIntraPredOptimzTest() {}
-  virtual void SetUp() {
-    PredFuncMode funcMode = GET_PARAM(0);
-    predFuncRef_ = std::tr1::get<0>(funcMode);
-    predFunc_ = std::tr1::get<1>(funcMode);
-    mode_ = std::tr1::get<2>(funcMode);
-    blockSize_ = GET_PARAM(1);
-
-    alloc_ = new uint8_t[3 * MaxBlkSize + 2];
-    predRef_ = new uint8_t[MaxBlkSize * MaxBlkSize];
-    pred_ = new uint8_t[MaxBlkSize * MaxBlkSize];
-  }
-
-  virtual void TearDown() {
-    delete[] alloc_;
-    delete[] predRef_;
-    delete[] pred_;
-    libaom_test::ClearSystemState();
-  }
-
- protected:
-  void RunTest() const {
-    int tstIndex = 0;
-    int stride = blockSize_;
-    uint8_t *left = alloc_;
-    uint8_t *above = alloc_ + MaxBlkSize + 1;
-    while (tstIndex < MaxTestNum) {
-      PrepareBuffer();
-      predFuncRef_(predRef_, stride, blockSize_, &above[1], left);
-      ASM_REGISTER_STATE_CHECK(
-          predFunc_(pred_, stride, blockSize_, &above[1], left));
-      DiffPred(tstIndex);
-      tstIndex += 1;
-    }
-  }
-
-  void RunSpeedTestC() const {
-    int tstIndex = 0;
-    int stride = blockSize_;
-    uint8_t *left = alloc_;
-    uint8_t *above = alloc_ + MaxBlkSize + 1;
-    PrepareBuffer();
-    while (tstIndex < MaxTestNum) {
-      predFuncRef_(predRef_, stride, blockSize_, &above[1], left);
-      tstIndex += 1;
-    }
-  }
-
-  void RunSpeedTestSSE() const {
-    int tstIndex = 0;
-    int stride = blockSize_;
-    uint8_t *left = alloc_;
-    uint8_t *above = alloc_ + MaxBlkSize + 1;
-    PrepareBuffer();
-    while (tstIndex < MaxTestNum) {
-      predFunc_(predRef_, stride, blockSize_, &above[1], left);
-      tstIndex += 1;
-    }
-  }
-
- private:
-  void PrepareBuffer() const {
-    ACMRandom rnd(ACMRandom::DeterministicSeed());
-    int i = 0;
-    while (i < (3 * MaxBlkSize + 2)) {
-      alloc_[i] = rnd.Rand8();
-      i += 1;
-    }
-  }
-
-  void DiffPred(int testNum) const {
-    int i = 0;
-    while (i < blockSize_ * blockSize_) {
-      EXPECT_EQ(predRef_[i], pred_[i]) << "Error at position: " << i << " "
-                                       << "Block size: " << blockSize_ << " "
-                                       << "Test number: " << testNum;
-      i += 1;
-    }
-  }
-
-  Predictor predFunc_;
-  Predictor predFuncRef_;
-  int mode_;
-  int blockSize_;
-  uint8_t *alloc_;
-  uint8_t *pred_;
-  uint8_t *predRef_;
-};
-
-#if CONFIG_HIGHBITDEPTH
-class AV1HbdFilterIntraPredOptimzTest
-    : public ::testing::TestWithParam<HbdPredParams> {
- public:
-  virtual ~AV1HbdFilterIntraPredOptimzTest() {}
-  virtual void SetUp() {
-    HbdPredFuncMode funcMode = GET_PARAM(0);
-    predFuncRef_ = std::tr1::get<0>(funcMode);
-    predFunc_ = std::tr1::get<1>(funcMode);
-    mode_ = std::tr1::get<2>(funcMode);
-    blockSize_ = GET_PARAM(1);
-    bd_ = GET_PARAM(2);
-
-    alloc_ = new uint16_t[3 * MaxBlkSize + 2];
-    predRef_ = new uint16_t[MaxBlkSize * MaxBlkSize];
-    pred_ = new uint16_t[MaxBlkSize * MaxBlkSize];
-  }
-
-  virtual void TearDown() {
-    delete[] alloc_;
-    delete[] predRef_;
-    delete[] pred_;
-    libaom_test::ClearSystemState();
-  }
-
- protected:
-  void RunTest() const {
-    int tstIndex = 0;
-    int stride = blockSize_;
-    uint16_t *left = alloc_;
-    uint16_t *above = alloc_ + MaxBlkSize + 1;
-    while (tstIndex < MaxTestNum) {
-      PrepareBuffer();
-      predFuncRef_(predRef_, stride, blockSize_, &above[1], left, bd_);
-      ASM_REGISTER_STATE_CHECK(
-          predFunc_(pred_, stride, blockSize_, &above[1], left, bd_));
-      DiffPred(tstIndex);
-      tstIndex += 1;
-    }
-  }
-
-  void RunSpeedTestC() const {
-    int tstIndex = 0;
-    int stride = blockSize_;
-    uint16_t *left = alloc_;
-    uint16_t *above = alloc_ + MaxBlkSize + 1;
-    PrepareBuffer();
-    while (tstIndex < MaxTestNum) {
-      predFuncRef_(predRef_, stride, blockSize_, &above[1], left, bd_);
-      tstIndex += 1;
-    }
-  }
-
-  void RunSpeedTestSSE() const {
-    int tstIndex = 0;
-    int stride = blockSize_;
-    uint16_t *left = alloc_;
-    uint16_t *above = alloc_ + MaxBlkSize + 1;
-    PrepareBuffer();
-    while (tstIndex < MaxTestNum) {
-      predFunc_(predRef_, stride, blockSize_, &above[1], left, bd_);
-      tstIndex += 1;
-    }
-  }
-
- private:
-  void PrepareBuffer() const {
-    ACMRandom rnd(ACMRandom::DeterministicSeed());
-    int i = 0;
-    while (i < (3 * MaxBlkSize + 2)) {
-      alloc_[i] = rnd.Rand16() & ((1 << bd_) - 1);
-      i += 1;
-    }
-  }
-
-  void DiffPred(int testNum) const {
-    int i = 0;
-    while (i < blockSize_ * blockSize_) {
-      EXPECT_EQ(predRef_[i], pred_[i]) << "Error at position: " << i << " "
-                                       << "Block size: " << blockSize_ << " "
-                                       << "Bit depth: " << bd_ << " "
-                                       << "Test number: " << testNum;
-      i += 1;
-    }
-  }
-
-  HbdPredictor predFunc_;
-  HbdPredictor predFuncRef_;
-  int mode_;
-  int blockSize_;
-  int bd_;
-  uint16_t *alloc_;
-  uint16_t *pred_;
-  uint16_t *predRef_;
-};
-#endif  // CONFIG_HIGHBITDEPTH
-
-TEST_P(AV1FilterIntraPredOptimzTest, BitExactCheck) { RunTest(); }
-
-#if PREDICTORS_SPEED_TEST
-TEST_P(AV1FilterIntraPredOptimzTest, SpeedCheckC) { RunSpeedTestC(); }
-
-TEST_P(AV1FilterIntraPredOptimzTest, SpeedCheckSSE) { RunSpeedTestSSE(); }
-#endif
-
-#if CONFIG_HIGHBITDEPTH
-TEST_P(AV1HbdFilterIntraPredOptimzTest, BitExactCheck) { RunTest(); }
-
-#if PREDICTORS_SPEED_TEST
-TEST_P(AV1HbdFilterIntraPredOptimzTest, SpeedCheckC) { RunSpeedTestC(); }
-
-TEST_P(AV1HbdFilterIntraPredOptimzTest, SpeedCheckSSE) { RunSpeedTestSSE(); }
-#endif  // PREDICTORS_SPEED_TEST
-#endif  // CONFIG_HIGHBITDEPTH
-
-using std::tr1::make_tuple;
-
-const PredFuncMode kPredFuncMdArray[] = {
-  make_tuple(av1_dc_filter_predictor_c, av1_dc_filter_predictor_sse4_1,
-             DC_PRED),
-  make_tuple(av1_v_filter_predictor_c, av1_v_filter_predictor_sse4_1, V_PRED),
-  make_tuple(av1_h_filter_predictor_c, av1_h_filter_predictor_sse4_1, H_PRED),
-  make_tuple(av1_d45_filter_predictor_c, av1_d45_filter_predictor_sse4_1,
-             D45_PRED),
-  make_tuple(av1_d135_filter_predictor_c, av1_d135_filter_predictor_sse4_1,
-             D135_PRED),
-  make_tuple(av1_d117_filter_predictor_c, av1_d117_filter_predictor_sse4_1,
-             D117_PRED),
-  make_tuple(av1_d153_filter_predictor_c, av1_d153_filter_predictor_sse4_1,
-             D153_PRED),
-  make_tuple(av1_d207_filter_predictor_c, av1_d207_filter_predictor_sse4_1,
-             D207_PRED),
-  make_tuple(av1_d63_filter_predictor_c, av1_d63_filter_predictor_sse4_1,
-             D63_PRED),
-  make_tuple(av1_tm_filter_predictor_c, av1_tm_filter_predictor_sse4_1,
-             TM_PRED),
-};
-
-const int kBlkSize[] = { 4, 8, 16, 32 };
-
-INSTANTIATE_TEST_CASE_P(
-    SSE4_1, AV1FilterIntraPredOptimzTest,
-    ::testing::Combine(::testing::ValuesIn(kPredFuncMdArray),
-                       ::testing::ValuesIn(kBlkSize)));
-
-#if CONFIG_HIGHBITDEPTH
-const HbdPredFuncMode kHbdPredFuncMdArray[] = {
-  make_tuple(av1_highbd_dc_filter_predictor_c,
-             av1_highbd_dc_filter_predictor_sse4_1, DC_PRED),
-  make_tuple(av1_highbd_v_filter_predictor_c,
-             av1_highbd_v_filter_predictor_sse4_1, V_PRED),
-  make_tuple(av1_highbd_h_filter_predictor_c,
-             av1_highbd_h_filter_predictor_sse4_1, H_PRED),
-  make_tuple(av1_highbd_d45_filter_predictor_c,
-             av1_highbd_d45_filter_predictor_sse4_1, D45_PRED),
-  make_tuple(av1_highbd_d135_filter_predictor_c,
-             av1_highbd_d135_filter_predictor_sse4_1, D135_PRED),
-  make_tuple(av1_highbd_d117_filter_predictor_c,
-             av1_highbd_d117_filter_predictor_sse4_1, D117_PRED),
-  make_tuple(av1_highbd_d153_filter_predictor_c,
-             av1_highbd_d153_filter_predictor_sse4_1, D153_PRED),
-  make_tuple(av1_highbd_d207_filter_predictor_c,
-             av1_highbd_d207_filter_predictor_sse4_1, D207_PRED),
-  make_tuple(av1_highbd_d63_filter_predictor_c,
-             av1_highbd_d63_filter_predictor_sse4_1, D63_PRED),
-  make_tuple(av1_highbd_tm_filter_predictor_c,
-             av1_highbd_tm_filter_predictor_sse4_1, TM_PRED),
-};
-
-const int kBd[] = { 10, 12 };
-
-INSTANTIATE_TEST_CASE_P(
-    SSE4_1, AV1HbdFilterIntraPredOptimzTest,
-    ::testing::Combine(::testing::ValuesIn(kHbdPredFuncMdArray),
-                       ::testing::ValuesIn(kBlkSize),
-                       ::testing::ValuesIn(kBd)));
-#endif  // CONFIG_HIGHBITDEPTH
-
-}  // namespace

third_party/aom/test/filterintra_test.cc

@@ -0,0 +1,134 @@
+/*
+ * Copyright (c) 2016, Alliance for Open Media. All rights reserved
+ *
+ * This source code is subject to the terms of the BSD 2 Clause License and
+ * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
+ * was not distributed with this source code in the LICENSE file, you can
+ * obtain it at www.aomedia.org/license/software. If the Alliance for Open
+ * Media Patent License 1.0 was not distributed with this source code in the
+ * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
+ */
+
+#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
+
+#include "config/av1_rtcd.h"
+
+#include "test/acm_random.h"
+#include "test/clear_system_state.h"
+#include "test/register_state_check.h"
+#include "test/util.h"
+#include "av1/common/enums.h"
+
+namespace {
+
+using ::testing::tuple;
+using libaom_test::ACMRandom;
+
+typedef void (*Predictor)(uint8_t *dst, ptrdiff_t stride, TX_SIZE tx_size,
+                          const uint8_t *above, const uint8_t *left, int mode);
+
+// Note:
+//  Test parameter list:
+//  Reference predictor, optimized predictor, prediction mode, tx size
+//
+typedef tuple<Predictor, Predictor, int> PredFuncMode;
+typedef tuple<PredFuncMode, TX_SIZE> PredParams;
+
+const int MaxTxSize = 32;
+
+const int MaxTestNum = 100;
+
+class AV1FilterIntraPredTest : public ::testing::TestWithParam<PredParams> {
+ public:
+  virtual ~AV1FilterIntraPredTest() {}
+  virtual void SetUp() {
+    PredFuncMode funcMode = GET_PARAM(0);
+    predFuncRef_ = ::testing::get<0>(funcMode);
+    predFunc_ = ::testing::get<1>(funcMode);
+    mode_ = ::testing::get<2>(funcMode);
+    txSize_ = GET_PARAM(1);
+
+    alloc_ = new uint8_t[2 * MaxTxSize + 1];
+    predRef_ = new uint8_t[MaxTxSize * MaxTxSize];
+    pred_ = new uint8_t[MaxTxSize * MaxTxSize];
+  }
+
+  virtual void TearDown() {
+    delete[] alloc_;
+    delete[] predRef_;
+    delete[] pred_;
+    libaom_test::ClearSystemState();
+  }
+
+ protected:
+  void RunTest() const {
+    int tstIndex = 0;
+    int stride = tx_size_wide[txSize_];
+    uint8_t *left = alloc_;
+    uint8_t *above = alloc_ + MaxTxSize;
+    while (tstIndex < MaxTestNum) {
+      PrepareBuffer();
+      predFuncRef_(predRef_, stride, txSize_, &above[1], left, mode_);
+      ASM_REGISTER_STATE_CHECK(
+          predFunc_(pred_, stride, txSize_, &above[1], left, mode_));
+      DiffPred(tstIndex);
+      tstIndex += 1;
+    }
+  }
+
+ private:
+  void PrepareBuffer() const {
+    ACMRandom rnd(ACMRandom::DeterministicSeed());
+    int i = 0;
+    while (i < (2 * MaxTxSize + 1)) {
+      alloc_[i] = rnd.Rand8();
+      i++;
+    }
+  }
+
+  void DiffPred(int testNum) const {
+    int i = 0;
+    while (i < tx_size_wide[txSize_] * tx_size_high[txSize_]) {
+      EXPECT_EQ(predRef_[i], pred_[i]) << "Error at position: " << i << " "
+                                       << "Tx size: " << tx_size_wide[txSize_]
+                                       << "x" << tx_size_high[txSize_] << " "
+                                       << "Test number: " << testNum;
+      i++;
+    }
+  }
+
+  Predictor predFunc_;
+  Predictor predFuncRef_;
+  int mode_;
+  TX_SIZE txSize_;
+  uint8_t *alloc_;
+  uint8_t *pred_;
+  uint8_t *predRef_;
+};
+
+TEST_P(AV1FilterIntraPredTest, BitExactCheck) { RunTest(); }
+
+using ::testing::make_tuple;
+
+const PredFuncMode kPredFuncMdArray[] = {
+  make_tuple(&av1_filter_intra_predictor_c, &av1_filter_intra_predictor_sse4_1,
+             FILTER_DC_PRED),
+  make_tuple(&av1_filter_intra_predictor_c, &av1_filter_intra_predictor_sse4_1,
+             FILTER_V_PRED),
+  make_tuple(&av1_filter_intra_predictor_c, &av1_filter_intra_predictor_sse4_1,
+             FILTER_H_PRED),
+  make_tuple(&av1_filter_intra_predictor_c, &av1_filter_intra_predictor_sse4_1,
+             FILTER_D157_PRED),
+  make_tuple(&av1_filter_intra_predictor_c, &av1_filter_intra_predictor_sse4_1,
+             FILTER_PAETH_PRED),
+};
+
+const TX_SIZE kTxSize[] = { TX_4X4,  TX_8X8,  TX_16X16, TX_32X32, TX_4X8,
+                            TX_8X4,  TX_8X16, TX_16X8,  TX_16X32, TX_32X16,
+                            TX_4X16, TX_16X4, TX_8X32,  TX_32X8 };
+
+INSTANTIATE_TEST_CASE_P(
+    SSE4_1, AV1FilterIntraPredTest,
+    ::testing::Combine(::testing::ValuesIn(kPredFuncMdArray),
+                       ::testing::ValuesIn(kTxSize)));
+}  // namespace

third_party/aom/test/frame_size_tests.cc

@@ -7,7 +7,7 @@
  * obtain it at www.aomedia.org/license/software. If the Alliance for Open
  * Media Patent License 1.0 was not distributed with this source code in the
  * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-*/
+ */
 
 #include "third_party/googletest/src/googletest/include/gtest/gtest.h"
 #include "test/codec_factory.h"

third_party/aom/test/fwht4x4_test.cc

@@ -0,0 +1,98 @@
+/*
+ * Copyright (c) 2016, Alliance for Open Media. All rights reserved
+ *
+ * This source code is subject to the terms of the BSD 2 Clause License and
+ * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
+ * was not distributed with this source code in the LICENSE file, you can
+ * obtain it at www.aomedia.org/license/software. If the Alliance for Open
+ * Media Patent License 1.0 was not distributed with this source code in the
+ * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
+ */
+
+#include <math.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include "third_party/googletest/src/googletest/include/gtest/gtest.h"
+
+#include "config/av1_rtcd.h"
+#include "config/aom_dsp_rtcd.h"
+#include "test/acm_random.h"
+#include "test/clear_system_state.h"
+#include "test/register_state_check.h"
+#include "test/transform_test_base.h"
+#include "test/util.h"
+#include "av1/common/entropy.h"
+#include "aom/aom_codec.h"
+#include "aom/aom_integer.h"
+#include "aom_ports/mem.h"
+
+using libaom_test::ACMRandom;
+
+namespace {
+typedef void (*FdctFunc)(const int16_t *in, tran_low_t *out, int stride);
+typedef void (*IdctFunc)(const tran_low_t *in, uint8_t *out, int stride);
+
+using libaom_test::FhtFunc;
+
+typedef ::testing::tuple<FdctFunc, IdctFunc, TX_TYPE, aom_bit_depth_t, int>
+    Dct4x4Param;
+
+void fwht4x4_ref(const int16_t *in, tran_low_t *out, int stride,
+                 TxfmParam * /*txfm_param*/) {
+  av1_fwht4x4_c(in, out, stride);
+}
+
+void iwht4x4_10(const tran_low_t *in, uint8_t *out, int stride) {
+  av1_highbd_iwht4x4_16_add_c(in, out, stride, 10);
+}
+
+void iwht4x4_12(const tran_low_t *in, uint8_t *out, int stride) {
+  av1_highbd_iwht4x4_16_add_c(in, out, stride, 12);
+}
+
+class Trans4x4WHT : public libaom_test::TransformTestBase,
+                    public ::testing::TestWithParam<Dct4x4Param> {
+ public:
+  virtual ~Trans4x4WHT() {}
+
+  virtual void SetUp() {
+    fwd_txfm_ = GET_PARAM(0);
+    inv_txfm_ = GET_PARAM(1);
+    pitch_ = 4;
+    height_ = 4;
+    fwd_txfm_ref = fwht4x4_ref;
+    bit_depth_ = GET_PARAM(3);
+    mask_ = (1 << bit_depth_) - 1;
+    num_coeffs_ = GET_PARAM(4);
+  }
+  virtual void TearDown() { libaom_test::ClearSystemState(); }
+
+ protected:
+  void RunFwdTxfm(const int16_t *in, tran_low_t *out, int stride) {
+    fwd_txfm_(in, out, stride);
+  }
+  void RunInvTxfm(const tran_low_t *out, uint8_t *dst, int stride) {
+    inv_txfm_(out, dst, stride);
+  }
+
+  FdctFunc fwd_txfm_;
+  IdctFunc inv_txfm_;
+};
+
+TEST_P(Trans4x4WHT, AccuracyCheck) { RunAccuracyCheck(0, 0.00001); }
+
+TEST_P(Trans4x4WHT, CoeffCheck) { RunCoeffCheck(); }
+
+TEST_P(Trans4x4WHT, MemCheck) { RunMemCheck(); }
+
+TEST_P(Trans4x4WHT, InvAccuracyCheck) { RunInvAccuracyCheck(0); }
+using ::testing::make_tuple;
+
+INSTANTIATE_TEST_CASE_P(
+    C, Trans4x4WHT,
+    ::testing::Values(make_tuple(&av1_highbd_fwht4x4_c, &iwht4x4_10, DCT_DCT,
+                                 AOM_BITS_10, 16),
+                      make_tuple(&av1_highbd_fwht4x4_c, &iwht4x4_12, DCT_DCT,
+                                 AOM_BITS_12, 16)));
+}  // namespace