dany/libwebp
1
0
Fork 0
mirror of https://github.com/webmproject/libwebp.git synced 2024-12-27 22:28:22 +01:00
Code Issues Actions Packages Projects Releases Wiki Activity

Improve SSE4.1 implementation of TTransform.

Browse Source 
SSE4.1 is slower than the SSE2 implementation and this seems to
be due to a slow _mm_loadl_epi64 implementation by gcc
(hence a bug with my gcc 4.8) and a very slow _mm_hadd_epi32. Both
got confirmed by IACA and experiments.

Change-Id: I05607f66b7ccd8f4f42e000693aea583ffd5768f
This commit is contained in:
Vincent Rabaud 2016-02-19 09:11:53 +01:00
parent 7ccdb734c2
commit 6583bb1a42
1 changed files with 13 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

23
src/dsp/enc_sse41.c
View File

@ -72,17 +72,22 @@ static void CollectHistogram(const uint8_t* ref, const uint8_t* pred,
// w[] contains a row-major 4 by 4 symmetric matrix. // w[] contains a row-major 4 by 4 symmetric matrix.
static int TTransform(const uint8_t* inA, const uint8_t* inB, static int TTransform(const uint8_t* inA, const uint8_t* inB,
const uint16_t* const w) { const uint16_t* const w) {
int32_t sum[4];
__m128i tmp_0, tmp_1, tmp_2, tmp_3; __m128i tmp_0, tmp_1, tmp_2, tmp_3;
// Load and combine inputs. // Load and combine inputs.
{ {
const __m128i inA_0 = _mm_loadl_epi64((const __m128i*)&inA[BPS * 0]); const __m128i inA_0 = _mm_loadu_si128((const __m128i*)&inA[BPS * 0]);
const __m128i inA_1 = _mm_loadl_epi64((const __m128i*)&inA[BPS * 1]); const __m128i inA_1 = _mm_loadu_si128((const __m128i*)&inA[BPS * 1]);
const __m128i inA_2 = _mm_loadl_epi64((const __m128i*)&inA[BPS * 2]); const __m128i inA_2 = _mm_loadu_si128((const __m128i*)&inA[BPS * 2]);
// In SSE4.1, with gcc 4.8 at least (maybe other versions),
// _mm_loadu_si128 is faster than _mm_loadl_epi64. But for the last lump
// of inA and inB, _mm_loadl_epi64 is still used not to have an out of
// bound read.
const __m128i inA_3 = _mm_loadl_epi64((const __m128i*)&inA[BPS * 3]); const __m128i inA_3 = _mm_loadl_epi64((const __m128i*)&inA[BPS * 3]);
const __m128i inB_0 = _mm_loadl_epi64((const __m128i*)&inB[BPS * 0]); const __m128i inB_0 = _mm_loadu_si128((const __m128i*)&inB[BPS * 0]);
const __m128i inB_1 = _mm_loadl_epi64((const __m128i*)&inB[BPS * 1]); const __m128i inB_1 = _mm_loadu_si128((const __m128i*)&inB[BPS * 1]);
const __m128i inB_2 = _mm_loadl_epi64((const __m128i*)&inB[BPS * 2]); const __m128i inB_2 = _mm_loadu_si128((const __m128i*)&inB[BPS * 2]);
const __m128i inB_3 = _mm_loadl_epi64((const __m128i*)&inB[BPS * 3]); const __m128i inB_3 = _mm_loadl_epi64((const __m128i*)&inB[BPS * 3]);
// Combine inA and inB (we'll do two transforms in parallel). // Combine inA and inB (we'll do two transforms in parallel).
@ -158,11 +163,9 @@ static int TTransform(const uint8_t* inA, const uint8_t* inB,
// difference of weighted sums // difference of weighted sums
A_b2 = _mm_sub_epi32(A_b0, B_b0); A_b2 = _mm_sub_epi32(A_b0, B_b0);
// cascading summation of the differences _mm_storeu_si128((__m128i*)&sum[0], A_b2);
B_b0 = _mm_hadd_epi32(A_b2, A_b2);
B_b2 = _mm_hadd_epi32(B_b0, B_b0);
return _mm_cvtsi128_si32(B_b2);
} }
return sum[0] + sum[1] + sum[2] + sum[3];
} }
static int Disto4x4(const uint8_t* const a, const uint8_t* const b, static int Disto4x4(const uint8_t* const a, const uint8_t* const b,