dany/libwebp
1
0
Fork 0
mirror of https://github.com/webmproject/libwebp.git synced 2024-11-20 04:18:26 +01:00
Code Issues Actions Packages Projects Releases Wiki Activity

Merge "SSE2: slightly (~2%) faster Predictor #1"

Browse Source
This commit is contained in:
Pascal Massimino 2016-12-12 14:46:21 +00:00 committed by Gerrit Code Review
commit 9ae0b3f65a
1 changed files with 3 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
src/dsp/lossless_sse2.c
View File

@ -193,6 +193,7 @@ static void PredictorAdd0_SSE2(const uint32_t* in, const uint32_t* upper,
static void PredictorAdd1_SSE2(const uint32_t* in, const uint32_t* upper,
int num_pixels, uint32_t* out) {
int i;
__m128i prev = _mm_set1_epi32(out[-1]);
for (i = 0; i + 4 <= num_pixels; i += 4) {
// a | b | c | d
const __m128i src = _mm_loadu_si128((const __m128i*)&in[i]);
@ -204,9 +205,10 @@ static void PredictorAdd1_SSE2(const uint32_t* in, const uint32_t* upper,
const __m128i shift1 = _mm_slli_si128(sum0, 8);
// a | a + b | a + b + c | a + b + c + d
const __m128i sum1 = _mm_add_epi8(sum0, shift1);
const __m128i prev = _mm_set1_epi32(out[i - 1]);
const __m128i res = _mm_add_epi8(sum1, prev);
_mm_storeu_si128((__m128i*)&out[i], res);
// replicate prev output on the four lanes
prev = _mm_shuffle_epi32(res, (3 << 0) | (3 << 2) | (3 << 4) | (3 << 6));
}
VP8LPredictorsAdd_C[1](in + i, upper + i, num_pixels - i, out + i);
}