Skip to content

Commit

Permalink
optimization
Browse files Browse the repository at this point in the history
  • Loading branch information
@itzpr3d4t0r
itzpr3d4t0r committed Dec 29, 2023
1 parent afe94b5 commit 634aef9
Showing 1 changed file with 28 additions and 33 deletions.
61 changes: 28 additions & 33 deletions src_c/simd_blitters_avx2.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -1530,38 +1530,33 @@ blit_blend_premultiplied_avx2(SDL_BlitInfo *info)
#endif /* defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
!defined(SDL_DISABLE_IMMINTRIN_H) */

#define PREMUL_ALPHA_CODE \
/* extract the alpha */ \
mm_alpha_in = _mm256_and_si256(mm_src, mm256_amask); \
/*generate the mask for choosing which pixel to use with blendv*/ \
alpha_eq_mask = _mm256_cmpeq_epi32(mm_alpha_in, mm256_amask); \
\
/*redistribute the alphas across the R, G, B channels*/ \
alphaA = _mm256_shuffle_epi8(mm_src, shuffle_maskA); \
alphaB = _mm256_shuffle_epi8(mm_src, shuffle_maskB); \
\
/*prep the pixels for 16-bit math*/ \
mm_srcA = _mm256_unpacklo_epi8(mm_src, mm_zero); \
mm_srcB = _mm256_unpackhi_epi8(mm_src, mm_zero); \
\
mm_srcA = _mm256_add_epi16(mm_srcA, mm256_ones); \
mm_srcB = _mm256_add_epi16(mm_srcB, mm256_ones); \
\
/*multiply the pixels by the alphas*/ \
mm_srcA = _mm256_mullo_epi16(mm_srcA, alphaA); \
mm_srcB = _mm256_mullo_epi16(mm_srcB, alphaB); \
\
/*shift the pixels back down to 8-bit*/ \
mm_srcA = _mm256_srli_epi16(mm_srcA, 8); \
mm_srcB = _mm256_srli_epi16(mm_srcB, 8); \
\
/*pack the pixels back together*/ \
mm_dst = _mm256_packus_epi16(mm_srcA, mm_srcB); \
/*add the original alpha back in*/ \
mm_dst = _mm256_or_si256(mm_dst, mm_alpha_in); \
/*if the original alpha=255, use base pixel, else use \
* the premultiplied one*/ \
mm_dst = _mm256_blendv_epi8(mm_dst, mm_src, alpha_eq_mask);
#define PREMUL_ALPHA_CODE \
/* extract the alpha */ \
mm_alpha_in = _mm256_and_si256(mm_src, mm256_amask); \
\
/*redistribute the alphas across the R, G, B channels*/ \
alphaA = _mm256_shuffle_epi8(mm_src, shuffle_maskA); \
alphaB = _mm256_shuffle_epi8(mm_src, shuffle_maskB); \
\
/*prep the pixels for 16-bit math*/ \
mm_srcA = _mm256_unpacklo_epi8(mm_src, mm_zero); \
mm_srcB = _mm256_unpackhi_epi8(mm_src, mm_zero); \
\
mm_srcA = _mm256_add_epi16(mm_srcA, mm256_ones); \
mm_srcB = _mm256_add_epi16(mm_srcB, mm256_ones); \
\
/*multiply the pixels by the alphas*/ \
mm_srcA = _mm256_mullo_epi16(mm_srcA, alphaA); \
mm_srcB = _mm256_mullo_epi16(mm_srcB, alphaB); \
\
/*shift the pixels back down to 8-bit*/ \
mm_srcA = _mm256_srli_epi16(mm_srcA, 8); \
mm_srcB = _mm256_srli_epi16(mm_srcB, 8); \
\
/*pack the pixels back together*/ \
mm_dst = _mm256_packus_epi16(mm_srcA, mm_srcB); \
/*add the original alpha back in*/ \
mm_dst = _mm256_or_si256(mm_dst, mm_alpha_in);

#if defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
!defined(SDL_DISABLE_IMMINTRIN_H)
Expand All @@ -1581,7 +1576,7 @@ premul_surf_color_by_alpha_avx2(SDL_Surface *src, SDL_Surface *dst)
Uint32 *dstp = (Uint32 *)dst->pixels;

__m256i mm_src, mm_dst, alphaA, alphaB, mm_alpha_in;
__m256i alpha_eq_mask, mm_srcA, mm_srcB;
__m256i mm_srcA, mm_srcB;

const __m256i mm256_amask = _mm256_set1_epi32(src->format->Amask);
const __m256i mm_zero = _mm256_setzero_si256();
Expand Down

0 comments on commit 634aef9

Please sign in to comment.