libjpeg-turbo: update to 2.1.x rev 0734e34f6

This commit is contained in:
roytam1 2024-10-11 12:01:54 +08:00
commit 19dc62b0a2
139 changed files with 989 additions and 962 deletions

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jcolsamp.inc"

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jcolsamp.inc"

View file

@ -7,11 +7,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jcolsamp.inc"

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jsimdext.inc"

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jsimdext.inc"

View file

@ -7,11 +7,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jsimdext.inc"

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jsimdext.inc"

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jsimdext.inc"

View file

@ -7,11 +7,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jsimdext.inc"

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jcolsamp.inc"

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jcolsamp.inc"

View file

@ -7,11 +7,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jcolsamp.inc"

View file

@ -1,7 +1,7 @@
;
; jchuff-sse2.asm - Huffman entropy encoding (SSE2)
;
; Copyright (C) 2009-2011, 2014-2017, 2019, D. R. Commander.
; Copyright (C) 2009-2011, 2014-2017, 2019, 2024, D. R. Commander.
; Copyright (C) 2015, Matthieu Darbois.
; Copyright (C) 2018, Matthias Räncker.
;
@ -9,11 +9,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
;
; This file contains an SSE2 implementation for Huffman coding of one block.
; The following code is based on jchuff.c; see jchuff.c for more details.
@ -65,7 +61,8 @@ times 1 << 2 db 3
times 1 << 1 db 2
times 1 << 0 db 1
times 1 db 0
jpeg_nbits_table:
GLOBAL_DATA(jpeg_nbits_table)
EXTN(jpeg_nbits_table):
times 1 db 0
times 1 << 0 db 1
times 1 << 1 db 2
@ -88,9 +85,9 @@ times 1 << 14 db 15
%ifdef PIC
%define NBITS(x) nbits_base + x
%else
%define NBITS(x) jpeg_nbits_table + x
%define NBITS(x) EXTN(jpeg_nbits_table) + x
%endif
%define MASK_BITS(x) NBITS((x) * 8) + (jpeg_mask_bits - jpeg_nbits_table)
%define MASK_BITS(x) NBITS((x) * 8) + (jpeg_mask_bits - EXTN(jpeg_nbits_table))
; --------------------------------------------------------------------------
SECTION SEG_TEXT
@ -469,7 +466,7 @@ EXTN(jsimd_huff_encode_one_block_sse2):
pcmpeqw mm_all_0xff, mm_all_0xff ;Z: all_0xff[i] = 0xFF;
%endmacro
GET_SYM nbits_base, jpeg_nbits_table, GET_SYM_BEFORE, GET_SYM_AFTER
GET_SYM nbits_base, EXTN(jpeg_nbits_table), GET_SYM_BEFORE, GET_SYM_AFTER
psrldq xmm4, 1 * SIZEOF_WORD ;G: w4 = 37 44 45 38 39 46 47 --
shufpd xmm1, xmm5, 10b ;F: w1 = 36 37 44 45 50 51 58 59

View file

@ -7,11 +7,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
;
; This file contains an SSE2 implementation of data preparation for progressive
; Huffman encoding. See jcphuff.c for more details.

View file

@ -9,11 +9,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jsimdext.inc"

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jsimdext.inc"

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jsimdext.inc"

View file

@ -9,11 +9,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jcolsamp.inc"

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jcolsamp.inc"

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jcolsamp.inc"

View file

@ -9,11 +9,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jsimdext.inc"

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jsimdext.inc"

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jsimdext.inc"

View file

@ -9,11 +9,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jsimdext.inc"

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jsimdext.inc"

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jsimdext.inc"

View file

@ -9,11 +9,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jcolsamp.inc"

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jcolsamp.inc"

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jcolsamp.inc"

View file

@ -9,11 +9,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jsimdext.inc"

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jsimdext.inc"

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jsimdext.inc"

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
;
; This file contains a floating-point implementation of the forward DCT
; (Discrete Cosine Transform). The following code is based directly on

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
;
; This file contains a floating-point implementation of the forward DCT
; (Discrete Cosine Transform). The following code is based directly on

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
;
; This file contains a fast, not so accurate integer implementation of
; the forward DCT (Discrete Cosine Transform). The following code is

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
;
; This file contains a fast, not so accurate integer implementation of
; the forward DCT (Discrete Cosine Transform). The following code is

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
;
; This file contains a slower but more accurate integer implementation of the
; forward DCT (Discrete Cosine Transform). The following code is based

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
;
; This file contains a slower but more accurate integer implementation of the
; forward DCT (Discrete Cosine Transform). The following code is based

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
;
; This file contains a slower but more accurate integer implementation of the
; forward DCT (Discrete Cosine Transform). The following code is based

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
;
; This file contains a floating-point implementation of the inverse DCT
; (Discrete Cosine Transform). The following code is based directly on

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
;
; This file contains a floating-point implementation of the inverse DCT
; (Discrete Cosine Transform). The following code is based directly on

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
;
; This file contains a floating-point implementation of the inverse DCT
; (Discrete Cosine Transform). The following code is based directly on

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
;
; This file contains a fast, not so accurate integer implementation of
; the inverse DCT (Discrete Cosine Transform). The following code is

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
;
; This file contains a fast, not so accurate integer implementation of
; the inverse DCT (Discrete Cosine Transform). The following code is

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
;
; This file contains a slower but more accurate integer implementation of the
; inverse DCT (Discrete Cosine Transform). The following code is based

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
;
; This file contains a slower but more accurate integer implementation of the
; inverse DCT (Discrete Cosine Transform). The following code is based

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
;
; This file contains a slower but more accurate integer implementation of the
; inverse DCT (Discrete Cosine Transform). The following code is based

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
;
; This file contains inverse-DCT routines that produce reduced-size
; output: either 4x4 or 2x2 pixels from an 8x8 DCT block.

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
;
; This file contains inverse-DCT routines that produce reduced-size
; output: either 4x4 or 2x2 pixels from an 8x8 DCT block.

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jsimdext.inc"
%include "jdct.inc"

View file

@ -2,17 +2,13 @@
; jquant.asm - sample data conversion and quantization (MMX)
;
; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
; Copyright (C) 2016, D. R. Commander.
; Copyright (C) 2016, 2024, D. R. Commander.
;
; Based on the x86 SIMD extension for IJG JPEG library
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jsimdext.inc"
%include "jdct.inc"
@ -120,8 +116,8 @@ EXTN(jsimd_convsamp_mmx):
; Quantize/descale the coefficients, and store into coef_block
;
; This implementation is based on an algorithm described in
; "How to optimize for the Pentium family of microprocessors"
; (http://www.agner.org/assem/).
; "Optimizing subroutines in assembly language:
; An optimization guide for x86 platforms" (https://agner.org/optimize).
;
; GLOBAL(void)
; jsimd_quantize_mmx(JCOEFPTR coef_block, DCTELEM *divisors,

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jsimdext.inc"
%include "jdct.inc"

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jsimdext.inc"
%include "jdct.inc"

View file

@ -2,18 +2,14 @@
; jquanti.asm - sample data conversion and quantization (AVX2)
;
; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
; Copyright (C) 2016, 2018, D. R. Commander.
; Copyright (C) 2016, 2018, 2024, D. R. Commander.
; Copyright (C) 2016, Matthieu Darbois.
;
; Based on the x86 SIMD extension for IJG JPEG library
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jsimdext.inc"
%include "jdct.inc"
@ -107,8 +103,8 @@ EXTN(jsimd_convsamp_avx2):
; Quantize/descale the coefficients, and store into coef_block
;
; This implementation is based on an algorithm described in
; "How to optimize for the Pentium family of microprocessors"
; (http://www.agner.org/assem/).
; "Optimizing subroutines in assembly language:
; An optimization guide for x86 platforms" (https://agner.org/optimize).
;
; GLOBAL(void)
; jsimd_quantize_avx2(JCOEFPTR coef_block, DCTELEM *divisors,

View file

@ -2,17 +2,13 @@
; jquanti.asm - sample data conversion and quantization (SSE2)
;
; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
; Copyright (C) 2016, D. R. Commander.
; Copyright (C) 2016, 2024, D. R. Commander.
;
; Based on the x86 SIMD extension for IJG JPEG library
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jsimdext.inc"
%include "jdct.inc"
@ -98,8 +94,8 @@ EXTN(jsimd_convsamp_sse2):
; Quantize/descale the coefficients, and store into coef_block
;
; This implementation is based on an algorithm described in
; "How to optimize for the Pentium family of microprocessors"
; (http://www.agner.org/assem/).
; "Optimizing subroutines in assembly language:
; An optimization guide for x86 platforms" (https://agner.org/optimize).
;
; GLOBAL(void)
; jsimd_quantize_sse2(JCOEFPTR coef_block, DCTELEM *divisors,

View file

@ -2,8 +2,8 @@
* jsimd_i386.c
*
* Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
* Copyright (C) 2009-2011, 2013-2014, 2016, 2018, 2022, D. R. Commander.
* Copyright (C) 2015-2016, 2018, Matthieu Darbois.
* Copyright (C) 2009-2011, 2013-2014, 2016, 2018, 2022-2023, D. R. Commander.
* Copyright (C) 2015-2016, 2018, 2022, Matthieu Darbois.
*
* Based on the x86 SIMD extension for IJG JPEG library,
* Copyright (C) 1999-2006, MIYASAKA Masaru.
@ -21,7 +21,6 @@
#include "../../jdct.h"
#include "../../jsimddct.h"
#include "../jsimd.h"
#include "jconfigint.h"
/*
* In the PIC cases, we have no guarantee that constants will keep
@ -32,13 +31,11 @@
#define IS_ALIGNED_SSE(ptr) (IS_ALIGNED(ptr, 4)) /* 16 byte alignment */
#define IS_ALIGNED_AVX(ptr) (IS_ALIGNED(ptr, 5)) /* 32 byte alignment */
static unsigned int simd_support = (unsigned int)(~0);
static unsigned int simd_huffman = 1;
static THREAD_LOCAL unsigned int simd_support = (unsigned int)(~0);
static THREAD_LOCAL unsigned int simd_huffman = 1;
/*
* Check what SIMD accelerations are supported.
*
* FIXME: This code is racy under a multi-threaded environment.
*/
LOCAL(void)
init_simd(void)
@ -161,6 +158,9 @@ jsimd_rgb_ycc_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
void (*sse2fct) (JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
void (*mmxfct) (JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
if (simd_support == ~0U)
init_simd();
switch (cinfo->in_color_space) {
case JCS_EXT_RGB:
avx2fct = jsimd_extrgb_ycc_convert_avx2;
@ -220,6 +220,9 @@ jsimd_rgb_gray_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
void (*sse2fct) (JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
void (*mmxfct) (JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
if (simd_support == ~0U)
init_simd();
switch (cinfo->in_color_space) {
case JCS_EXT_RGB:
avx2fct = jsimd_extrgb_gray_convert_avx2;
@ -279,6 +282,9 @@ jsimd_ycc_rgb_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
void (*sse2fct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int);
void (*mmxfct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int);
if (simd_support == ~0U)
init_simd();
switch (cinfo->out_color_space) {
case JCS_EXT_RGB:
avx2fct = jsimd_ycc_extrgb_convert_avx2;
@ -382,6 +388,9 @@ GLOBAL(void)
jsimd_h2v2_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
JSAMPARRAY input_data, JSAMPARRAY output_data)
{
if (simd_support == ~0U)
init_simd();
if (simd_support & JSIMD_AVX2)
jsimd_h2v2_downsample_avx2(cinfo->image_width, cinfo->max_v_samp_factor,
compptr->v_samp_factor,
@ -402,6 +411,9 @@ GLOBAL(void)
jsimd_h2v1_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
JSAMPARRAY input_data, JSAMPARRAY output_data)
{
if (simd_support == ~0U)
init_simd();
if (simd_support & JSIMD_AVX2)
jsimd_h2v1_downsample_avx2(cinfo->image_width, cinfo->max_v_samp_factor,
compptr->v_samp_factor,
@ -464,6 +476,9 @@ GLOBAL(void)
jsimd_h2v2_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
{
if (simd_support == ~0U)
init_simd();
if (simd_support & JSIMD_AVX2)
jsimd_h2v2_upsample_avx2(cinfo->max_v_samp_factor, cinfo->output_width,
input_data, output_data_ptr);
@ -479,6 +494,9 @@ GLOBAL(void)
jsimd_h2v1_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
{
if (simd_support == ~0U)
init_simd();
if (simd_support & JSIMD_AVX2)
jsimd_h2v1_upsample_avx2(cinfo->max_v_samp_factor, cinfo->output_width,
input_data, output_data_ptr);
@ -540,6 +558,9 @@ GLOBAL(void)
jsimd_h2v2_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
{
if (simd_support == ~0U)
init_simd();
if (simd_support & JSIMD_AVX2)
jsimd_h2v2_fancy_upsample_avx2(cinfo->max_v_samp_factor,
compptr->downsampled_width, input_data,
@ -558,6 +579,9 @@ GLOBAL(void)
jsimd_h2v1_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
{
if (simd_support == ~0U)
init_simd();
if (simd_support & JSIMD_AVX2)
jsimd_h2v1_fancy_upsample_avx2(cinfo->max_v_samp_factor,
compptr->downsampled_width, input_data,
@ -626,6 +650,9 @@ jsimd_h2v2_merged_upsample(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
void (*sse2fct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
void (*mmxfct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
if (simd_support == ~0U)
init_simd();
switch (cinfo->out_color_space) {
case JCS_EXT_RGB:
avx2fct = jsimd_h2v2_extrgb_merged_upsample_avx2;
@ -684,6 +711,9 @@ jsimd_h2v1_merged_upsample(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
void (*sse2fct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
void (*mmxfct) (JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
if (simd_support == ~0U)
init_simd();
switch (cinfo->out_color_space) {
case JCS_EXT_RGB:
avx2fct = jsimd_h2v1_extrgb_merged_upsample_avx2;
@ -788,6 +818,9 @@ GLOBAL(void)
jsimd_convsamp(JSAMPARRAY sample_data, JDIMENSION start_col,
DCTELEM *workspace)
{
if (simd_support == ~0U)
init_simd();
if (simd_support & JSIMD_AVX2)
jsimd_convsamp_avx2(sample_data, start_col, workspace);
else if (simd_support & JSIMD_SSE2)
@ -800,6 +833,9 @@ GLOBAL(void)
jsimd_convsamp_float(JSAMPARRAY sample_data, JDIMENSION start_col,
FAST_FLOAT *workspace)
{
if (simd_support == ~0U)
init_simd();
if (simd_support & JSIMD_SSE2)
jsimd_convsamp_float_sse2(sample_data, start_col, workspace);
else if (simd_support & JSIMD_SSE)
@ -870,6 +906,9 @@ jsimd_can_fdct_float(void)
GLOBAL(void)
jsimd_fdct_islow(DCTELEM *data)
{
if (simd_support == ~0U)
init_simd();
if (simd_support & JSIMD_AVX2)
jsimd_fdct_islow_avx2(data);
else if (simd_support & JSIMD_SSE2)
@ -881,6 +920,9 @@ jsimd_fdct_islow(DCTELEM *data)
GLOBAL(void)
jsimd_fdct_ifast(DCTELEM *data)
{
if (simd_support == ~0U)
init_simd();
if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_fdct_islow_sse2))
jsimd_fdct_ifast_sse2(data);
else
@ -890,6 +932,9 @@ jsimd_fdct_ifast(DCTELEM *data)
GLOBAL(void)
jsimd_fdct_float(FAST_FLOAT *data)
{
if (simd_support == ~0U)
init_simd();
if ((simd_support & JSIMD_SSE) && IS_ALIGNED_SSE(jconst_fdct_float_sse))
jsimd_fdct_float_sse(data);
else if (simd_support & JSIMD_3DNOW)
@ -945,6 +990,9 @@ jsimd_can_quantize_float(void)
GLOBAL(void)
jsimd_quantize(JCOEFPTR coef_block, DCTELEM *divisors, DCTELEM *workspace)
{
if (simd_support == ~0U)
init_simd();
if (simd_support & JSIMD_AVX2)
jsimd_quantize_avx2(coef_block, divisors, workspace);
else if (simd_support & JSIMD_SSE2)
@ -957,6 +1005,9 @@ GLOBAL(void)
jsimd_quantize_float(JCOEFPTR coef_block, FAST_FLOAT *divisors,
FAST_FLOAT *workspace)
{
if (simd_support == ~0U)
init_simd();
if (simd_support & JSIMD_SSE2)
jsimd_quantize_float_sse2(coef_block, divisors, workspace);
else if (simd_support & JSIMD_SSE)
@ -1020,6 +1071,9 @@ jsimd_idct_2x2(j_decompress_ptr cinfo, jpeg_component_info *compptr,
JCOEFPTR coef_block, JSAMPARRAY output_buf,
JDIMENSION output_col)
{
if (simd_support == ~0U)
init_simd();
if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_red_sse2))
jsimd_idct_2x2_sse2(compptr->dct_table, coef_block, output_buf,
output_col);
@ -1032,6 +1086,9 @@ jsimd_idct_4x4(j_decompress_ptr cinfo, jpeg_component_info *compptr,
JCOEFPTR coef_block, JSAMPARRAY output_buf,
JDIMENSION output_col)
{
if (simd_support == ~0U)
init_simd();
if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_red_sse2))
jsimd_idct_4x4_sse2(compptr->dct_table, coef_block, output_buf,
output_col);
@ -1126,6 +1183,9 @@ jsimd_idct_islow(j_decompress_ptr cinfo, jpeg_component_info *compptr,
JCOEFPTR coef_block, JSAMPARRAY output_buf,
JDIMENSION output_col)
{
if (simd_support == ~0U)
init_simd();
if (simd_support & JSIMD_AVX2)
jsimd_idct_islow_avx2(compptr->dct_table, coef_block, output_buf,
output_col);
@ -1142,6 +1202,9 @@ jsimd_idct_ifast(j_decompress_ptr cinfo, jpeg_component_info *compptr,
JCOEFPTR coef_block, JSAMPARRAY output_buf,
JDIMENSION output_col)
{
if (simd_support == ~0U)
init_simd();
if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_ifast_sse2))
jsimd_idct_ifast_sse2(compptr->dct_table, coef_block, output_buf,
output_col);
@ -1155,6 +1218,9 @@ jsimd_idct_float(j_decompress_ptr cinfo, jpeg_component_info *compptr,
JCOEFPTR coef_block, JSAMPARRAY output_buf,
JDIMENSION output_col)
{
if (simd_support == ~0U)
init_simd();
if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_float_sse2))
jsimd_idct_float_sse2(compptr->dct_table, coef_block, output_buf,
output_col);
@ -1212,7 +1278,7 @@ jsimd_can_encode_mcu_AC_first_prepare(void)
GLOBAL(void)
jsimd_encode_mcu_AC_first_prepare(const JCOEF *block,
const int *jpeg_natural_order_start, int Sl,
int Al, JCOEF *values, size_t *zerobits)
int Al, UJCOEF *values, size_t *zerobits)
{
jsimd_encode_mcu_AC_first_prepare_sse2(block, jpeg_natural_order_start,
Sl, Al, values, zerobits);
@ -1238,7 +1304,7 @@ jsimd_can_encode_mcu_AC_refine_prepare(void)
GLOBAL(int)
jsimd_encode_mcu_AC_refine_prepare(const JCOEF *block,
const int *jpeg_natural_order_start, int Sl,
int Al, JCOEF *absvalues, size_t *bits)
int Al, UJCOEF *absvalues, size_t *bits)
{
return jsimd_encode_mcu_AC_refine_prepare_sse2(block,
jpeg_natural_order_start,

View file

@ -8,11 +8,7 @@
; Copyright (C) 1999-2006, MIYASAKA Masaru.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
; This file should be assembled with NASM (Netwide Assembler) or Yasm.
%include "jsimdext.inc"