wuggy/Dactyloidae
wuggy/Dactyloidae
1
0
Fork 0
mirror of https://repo.dactyloidae.xyz/Dactyloidae/UXP.git synced 2026-10-09 08:47:31 +09:00
Code Issues Projects Releases Packages Wiki Activity

Update aom to commit id f5bdeac22930ff4c6b219be49c843db35970b918

Browse source
This commit is contained in:
trav90 2018-10-18 06:04:57 -05:00 • committed by Roy Tam
commit 2ab8fcbfd1
370 changed files with 56185 additions and 32026 deletions
Download patch file Download diff file Expand all files Collapse all files

141
third_party/aom/aom_dsp/x86/fwd_txfm_sse2.c vendored
View file

@ -85,147 +85,6 @@ void aom_fdct8x8_1_sse2(const int16_t *input, tran_low_t *output, int stride) {
output[0] = (tran_low_t)_mm_cvtsi128_si32(in1);
}
void aom_fdct16x16_1_sse2(const int16_t *input, tran_low_t *output,
int stride) {
__m128i in0, in1, in2, in3;
__m128i u0, u1;
__m128i sum = _mm_setzero_si128();
int i;
for (i = 0; i < 2; ++i) {
in0 = _mm_load_si128((const __m128i *)(input + 0 * stride + 0));
in1 = _mm_load_si128((const __m128i *)(input + 0 * stride + 8));
in2 = _mm_load_si128((const __m128i *)(input + 1 * stride + 0));
in3 = _mm_load_si128((const __m128i *)(input + 1 * stride + 8));
u0 = _mm_add_epi16(in0, in1);
u1 = _mm_add_epi16(in2, in3);
sum = _mm_add_epi16(sum, u0);
in0 = _mm_load_si128((const __m128i *)(input + 2 * stride + 0));
in1 = _mm_load_si128((const __m128i *)(input + 2 * stride + 8));
in2 = _mm_load_si128((const __m128i *)(input + 3 * stride + 0));
in3 = _mm_load_si128((const __m128i *)(input + 3 * stride + 8));
sum = _mm_add_epi16(sum, u1);
u0 = _mm_add_epi16(in0, in1);
u1 = _mm_add_epi16(in2, in3);
sum = _mm_add_epi16(sum, u0);
in0 = _mm_load_si128((const __m128i *)(input + 4 * stride + 0));
in1 = _mm_load_si128((const __m128i *)(input + 4 * stride + 8));
in2 = _mm_load_si128((const __m128i *)(input + 5 * stride + 0));
in3 = _mm_load_si128((const __m128i *)(input + 5 * stride + 8));
sum = _mm_add_epi16(sum, u1);
u0 = _mm_add_epi16(in0, in1);
u1 = _mm_add_epi16(in2, in3);
sum = _mm_add_epi16(sum, u0);
in0 = _mm_load_si128((const __m128i *)(input + 6 * stride + 0));
in1 = _mm_load_si128((const __m128i *)(input + 6 * stride + 8));
in2 = _mm_load_si128((const __m128i *)(input + 7 * stride + 0));
in3 = _mm_load_si128((const __m128i *)(input + 7 * stride + 8));
sum = _mm_add_epi16(sum, u1);
u0 = _mm_add_epi16(in0, in1);
u1 = _mm_add_epi16(in2, in3);
sum = _mm_add_epi16(sum, u0);
sum = _mm_add_epi16(sum, u1);
input += 8 * stride;
}
u0 = _mm_setzero_si128();
in0 = _mm_unpacklo_epi16(u0, sum);
in1 = _mm_unpackhi_epi16(u0, sum);
in0 = _mm_srai_epi32(in0, 16);
in1 = _mm_srai_epi32(in1, 16);
sum = _mm_add_epi32(in0, in1);
in0 = _mm_unpacklo_epi32(sum, u0);
in1 = _mm_unpackhi_epi32(sum, u0);
sum = _mm_add_epi32(in0, in1);
in0 = _mm_srli_si128(sum, 8);
in1 = _mm_add_epi32(sum, in0);
in1 = _mm_srai_epi32(in1, 1);
output[0] = (tran_low_t)_mm_cvtsi128_si32(in1);
}
void aom_fdct32x32_1_sse2(const int16_t *input, tran_low_t *output,
int stride) {
__m128i in0, in1, in2, in3;
__m128i u0, u1;
__m128i sum = _mm_setzero_si128();
int i;
for (i = 0; i < 8; ++i) {
in0 = _mm_load_si128((const __m128i *)(input + 0));
in1 = _mm_load_si128((const __m128i *)(input + 8));
in2 = _mm_load_si128((const __m128i *)(input + 16));
in3 = _mm_load_si128((const __m128i *)(input + 24));
input += stride;
u0 = _mm_add_epi16(in0, in1);
u1 = _mm_add_epi16(in2, in3);
sum = _mm_add_epi16(sum, u0);
in0 = _mm_load_si128((const __m128i *)(input + 0));
in1 = _mm_load_si128((const __m128i *)(input + 8));
in2 = _mm_load_si128((const __m128i *)(input + 16));
in3 = _mm_load_si128((const __m128i *)(input + 24));
input += stride;
sum = _mm_add_epi16(sum, u1);
u0 = _mm_add_epi16(in0, in1);
u1 = _mm_add_epi16(in2, in3);
sum = _mm_add_epi16(sum, u0);
in0 = _mm_load_si128((const __m128i *)(input + 0));
in1 = _mm_load_si128((const __m128i *)(input + 8));
in2 = _mm_load_si128((const __m128i *)(input + 16));
in3 = _mm_load_si128((const __m128i *)(input + 24));
input += stride;
sum = _mm_add_epi16(sum, u1);
u0 = _mm_add_epi16(in0, in1);
u1 = _mm_add_epi16(in2, in3);
sum = _mm_add_epi16(sum, u0);
in0 = _mm_load_si128((const __m128i *)(input + 0));
in1 = _mm_load_si128((const __m128i *)(input + 8));
in2 = _mm_load_si128((const __m128i *)(input + 16));
in3 = _mm_load_si128((const __m128i *)(input + 24));
input += stride;
sum = _mm_add_epi16(sum, u1);
u0 = _mm_add_epi16(in0, in1);
u1 = _mm_add_epi16(in2, in3);
sum = _mm_add_epi16(sum, u0);
sum = _mm_add_epi16(sum, u1);
}
u0 = _mm_setzero_si128();
in0 = _mm_unpacklo_epi16(u0, sum);
in1 = _mm_unpackhi_epi16(u0, sum);
in0 = _mm_srai_epi32(in0, 16);
in1 = _mm_srai_epi32(in1, 16);
sum = _mm_add_epi32(in0, in1);
in0 = _mm_unpacklo_epi32(sum, u0);
in1 = _mm_unpackhi_epi32(sum, u0);
sum = _mm_add_epi32(in0, in1);
in0 = _mm_srli_si128(sum, 8);
in1 = _mm_add_epi32(sum, in0);
in1 = _mm_srai_epi32(in1, 3);
output[0] = (tran_low_t)_mm_cvtsi128_si32(in1);
}
#define DCT_HIGH_BIT_DEPTH 0
#define FDCT4x4_2D aom_fdct4x4_sse2
#define FDCT8x8_2D aom_fdct8x8_sse2