python-pillow/Pillow
1
1
Fork 0
mirror of https://github.com/python-pillow/Pillow.git synced 2025-08-20 20:24:45 +03:00
Code Issues Packages Projects Releases Wiki Activity

SIMD Filter. better macros

Browse Source
This commit is contained in:
Alexander 2017-09-13 01:58:43 +03:00 committed by Alexander Karpinsky
parent eb57b5eca7
commit 0e17b8313b
1 changed files with 17 additions and 21 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

38
libImaging/FilterSIMD_3x3i_4u8.c
View File

@ -7,25 +7,16 @@ ImagingFilter3x3i_4u8(Imaging imOut, Imaging im, const INT16* kernel,
pix1##row = _mm_shuffle_epi8(_mm_loadu_si128((__m128i*) &in0[x]), shuffle); \ pix1##row = _mm_shuffle_epi8(_mm_loadu_si128((__m128i*) &in0[x]), shuffle); \
pix2##row = _mm_shuffle_epi8(_mm_loadu_si128((__m128i*) &in_1[x]), shuffle); pix2##row = _mm_shuffle_epi8(_mm_loadu_si128((__m128i*) &in_1[x]), shuffle);
#define MM_KERNEL_SUM(ss, row, krow) \ #define MM_KERNEL_SUM(ss, row, krow, unpack_epi8, unpack_epi32) \
ss = _mm_add_epi32(ss, _mm_madd_epi16( \ ss = _mm_add_epi32(ss, _mm_madd_epi16( \
_mm_unpacklo_epi8(pix0##row, _mm_setzero_si128()), \ unpack_epi8(pix0##row, _mm_setzero_si128()), \
_mm_unpacklo_epi32(kernel0##krow, kernel0##krow))); \ unpack_epi32(kernel0##krow, kernel0##krow))); \
ss = _mm_add_epi32(ss, _mm_madd_epi16( \ ss = _mm_add_epi32(ss, _mm_madd_epi16( \
_mm_unpacklo_epi8(pix1##row, _mm_setzero_si128()), \ unpack_epi8(pix1##row, _mm_setzero_si128()), \
_mm_unpacklo_epi32(kernel1##krow, kernel1##krow))); \ unpack_epi32(kernel1##krow, kernel1##krow))); \
ss = _mm_add_epi32(ss, _mm_madd_epi16( \ ss = _mm_add_epi32(ss, _mm_madd_epi16( \
_mm_unpacklo_epi8(pix2##row, _mm_setzero_si128()), \ unpack_epi8(pix2##row, _mm_setzero_si128()), \
_mm_unpacklo_epi32(kernel2##krow, kernel2##krow))); \ unpack_epi32(kernel2##krow, kernel2##krow)));
ss = _mm_add_epi32(ss, _mm_madd_epi16( \
_mm_unpackhi_epi8(pix0##row, _mm_setzero_si128()), \
_mm_unpackhi_epi32(kernel0##krow, kernel0##krow))); \
ss = _mm_add_epi32(ss, _mm_madd_epi16( \
_mm_unpackhi_epi8(pix1##row, _mm_setzero_si128()), \
_mm_unpackhi_epi32(kernel1##krow, kernel1##krow))); \
ss = _mm_add_epi32(ss, _mm_madd_epi16( \
_mm_unpackhi_epi8(pix2##row, _mm_setzero_si128()), \
_mm_unpackhi_epi32(kernel2##krow, kernel2##krow)));
int x, y; int x, y;
@ -70,16 +61,20 @@ ImagingFilter3x3i_4u8(Imaging imOut, Imaging im, const INT16* kernel,
__m128i ss3 = _mm_set1_epi32(offset); __m128i ss3 = _mm_set1_epi32(offset);
MM_KERNEL_LOAD(0, x-1); MM_KERNEL_LOAD(0, x-1);
MM_KERNEL_SUM(ss0, 0, 0); MM_KERNEL_SUM(ss0, 0, 0, _mm_unpacklo_epi8, _mm_unpacklo_epi32);
MM_KERNEL_SUM(ss0, 0, 0, _mm_unpackhi_epi8, _mm_unpackhi_epi32);
ss0 = _mm_srai_epi32(ss0, PRECISION_BITS); ss0 = _mm_srai_epi32(ss0, PRECISION_BITS);
MM_KERNEL_SUM(ss1, 0, 1); MM_KERNEL_SUM(ss1, 0, 1, _mm_unpacklo_epi8, _mm_unpacklo_epi32);
MM_KERNEL_SUM(ss1, 0, 1, _mm_unpackhi_epi8, _mm_unpackhi_epi32);
ss1 = _mm_srai_epi32(ss1, PRECISION_BITS); ss1 = _mm_srai_epi32(ss1, PRECISION_BITS);
ss0 = _mm_packs_epi32(ss0, ss1); ss0 = _mm_packs_epi32(ss0, ss1);
MM_KERNEL_LOAD(0, x+1); MM_KERNEL_LOAD(0, x+1);
MM_KERNEL_SUM(ss2, 0, 0); MM_KERNEL_SUM(ss2, 0, 0, _mm_unpacklo_epi8, _mm_unpacklo_epi32);
MM_KERNEL_SUM(ss2, 0, 0, _mm_unpackhi_epi8, _mm_unpackhi_epi32);
ss2 = _mm_srai_epi32(ss2, PRECISION_BITS); ss2 = _mm_srai_epi32(ss2, PRECISION_BITS);
MM_KERNEL_SUM(ss3, 0, 1); MM_KERNEL_SUM(ss3, 0, 1, _mm_unpacklo_epi8, _mm_unpacklo_epi32);
MM_KERNEL_SUM(ss3, 0, 1, _mm_unpackhi_epi8, _mm_unpackhi_epi32);
ss3 = _mm_srai_epi32(ss3, PRECISION_BITS); ss3 = _mm_srai_epi32(ss3, PRECISION_BITS);
ss2 = _mm_packs_epi32(ss2, ss3); ss2 = _mm_packs_epi32(ss2, ss3);
@ -91,7 +86,8 @@ ImagingFilter3x3i_4u8(Imaging imOut, Imaging im, const INT16* kernel,
MM_KERNEL_LOAD(0, x-1); MM_KERNEL_LOAD(0, x-1);
MM_KERNEL_SUM(ss, 0, 0); MM_KERNEL_SUM(ss, 0, 0, _mm_unpacklo_epi8, _mm_unpacklo_epi32);
MM_KERNEL_SUM(ss, 0, 0, _mm_unpackhi_epi8, _mm_unpackhi_epi32);
ss = _mm_srai_epi32(ss, PRECISION_BITS); ss = _mm_srai_epi32(ss, PRECISION_BITS);
ss = _mm_packs_epi32(ss, ss); ss = _mm_packs_epi32(ss, ss);