dany/libwebp
1
0
Fork 0
mirror of https://github.com/webmproject/libwebp.git synced 2025-07-16 22:09:57 +02:00
Code Issues Actions Packages Projects Releases Wiki Activity

neon-implementation for rescaler code

Browse Source 
It's better to stay with a 32b fixed-point precision overall, otherwise
the C-version on ARM gets *slower*.
Actually, gcc ARM compiler optimizes some instructions pretty
well when WEBP_RESCALER_FIX is exactly 32, even in C.

Change-Id: I0eea97f7db5947470f5af355dee098eca81e178d
This commit is contained in:
skal
2015-10-05 11:26:52 +02:00
committed by Pascal Massimino
parent 6dfa5e3e58
commit b4e731cd93
7 changed files with 205 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
src/dsp/rescaler_sse2.c
View File

@ -70,8 +70,15 @@ static WEBP_INLINE void ProcessRow(const __m128i* const A0,
const __m128i C3 = _mm_add_epi64(B3, rounder);
const __m128i D0 = _mm_srli_epi64(C0, WEBP_RESCALER_RFIX);
const __m128i D1 = _mm_srli_epi64(C1, WEBP_RESCALER_RFIX);
#if (WEBP_RESCALER_FIX < 32)
const __m128i D2 =
_mm_and_si128(_mm_slli_epi64(C2, 32 - WEBP_RESCALER_RFIX), mask);
const __m128i D3 =
_mm_and_si128(_mm_slli_epi64(C3, 32 - WEBP_RESCALER_RFIX), mask);
#else
const __m128i D2 = _mm_and_si128(C2, mask);
const __m128i D3 = _mm_and_si128(C3, mask);
#endif
const __m128i E0 = _mm_or_si128(D0, D2);
const __m128i E1 = _mm_or_si128(D1, D3);
const __m128i F = _mm_packs_epi32(E0, E1);