python-pillow/Pillow
1
1
Fork 0
mirror of https://github.com/python-pillow/Pillow.git synced 2025-08-21 04:34:47 +03:00
Code Issues Packages Projects Releases Wiki Activity

SIMD Filter. use macros in 3x3

Browse Source
This commit is contained in:
Alexander 2017-08-13 21:49:45 +03:00 committed by Alexander Karpinsky
parent f747e7efc8
commit 4da1f70050
1 changed files with 51 additions and 84 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

135
src/libImaging/Filter.c
View File

@ -109,13 +109,15 @@ ImagingFilter3x3(Imaging imOut, Imaging im, const float *kernel, float offset) {
(_i2f((UINT8)in0[x - d]) * (kernel)[0] + _i2f((UINT8)in0[x]) * (kernel)[1] + \ (_i2f((UINT8)in0[x - d]) * (kernel)[0] + _i2f((UINT8)in0[x]) * (kernel)[1] + \
_i2f((UINT8)in0[x + d]) * (kernel)[2]) _i2f((UINT8)in0[x + d]) * (kernel)[2])
#define MM_KERNEL1x3(ss, in0, x, kernel, d) \ #define MM_KERNEL1x3_LOAD(row, x) \
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps((kernel)[0]), \ pix0##row = _mm_cvtepi32_ps(_mm_cvtepu8_epi32(*(__m128i*) &in1[x])); \
_mm_cvtepi32_ps(_mm_cvtepu8_epi32(*(__m128i *) &(in0)[x-d])))); \ pix1##row = _mm_cvtepi32_ps(_mm_cvtepu8_epi32(*(__m128i*) &in0[x])); \
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps((kernel)[1]), \ pix2##row = _mm_cvtepi32_ps(_mm_cvtepu8_epi32(*(__m128i*) &in_1[x]));
_mm_cvtepi32_ps(_mm_cvtepu8_epi32(*(__m128i *) &(in0)[x+0])))); \
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps((kernel)[2]), \ #define MM_KERNEL1x3_SUM(row, kindex) \
_mm_cvtepi32_ps(_mm_cvtepu8_epi32(*(__m128i *) &(in0)[x+d])))); ss = _mm_add_ps(ss, _mm_mul_ps(pix0##row, _mm_set1_ps(kernel[0 + kindex]))); \
ss = _mm_add_ps(ss, _mm_mul_ps(pix1##row, _mm_set1_ps(kernel[3 + kindex]))); \
ss = _mm_add_ps(ss, _mm_mul_ps(pix2##row, _mm_set1_ps(kernel[6 + kindex])));
int x = 0, y = 0; int x = 0, y = 0;
@ -145,13 +147,11 @@ ImagingFilter3x3(Imaging imOut, Imaging im, const float *kernel, float offset) {
UINT32* in0 = (UINT32*) im->image[y]; UINT32* in0 = (UINT32*) im->image[y];
UINT32* in1 = (UINT32*) im->image[y+1]; UINT32* in1 = (UINT32*) im->image[y+1];
UINT32* out = (UINT32*) imOut->image[y]; UINT32* out = (UINT32*) imOut->image[y];
__m128 pix00 = _mm_cvtepi32_ps(_mm_cvtepu8_epi32(*(__m128i *) &(in1)[0])); __m128 pix00, pix10, pix20;
__m128 pix01 = _mm_cvtepi32_ps(_mm_cvtepu8_epi32(*(__m128i *) &(in1)[1])); __m128 pix01, pix11, pix21;
__m128 pix10 = _mm_cvtepi32_ps(_mm_cvtepu8_epi32(*(__m128i *) &(in0)[0]));
__m128 pix11 = _mm_cvtepi32_ps(_mm_cvtepu8_epi32(*(__m128i *) &(in0)[1]));
__m128 pix20 = _mm_cvtepi32_ps(_mm_cvtepu8_epi32(*(__m128i *) &(in_1)[0]));
__m128 pix21 = _mm_cvtepi32_ps(_mm_cvtepu8_epi32(*(__m128i *) &(in_1)[1]));
__m128 pix02, pix12, pix22; __m128 pix02, pix12, pix22;
MM_KERNEL1x3_LOAD(0, 0);
MM_KERNEL1x3_LOAD(1, 1);
out[0] = in0[0]; out[0] = in0[0];
x = 1; x = 1;
@ -160,49 +160,24 @@ ImagingFilter3x3(Imaging imOut, Imaging im, const float *kernel, float offset) {
__m128i ssi0, ssi1, ssi2; __m128i ssi0, ssi1, ssi2;
ss = _mm_set1_ps(offset); ss = _mm_set1_ps(offset);
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps(kernel[0]), pix00)); MM_KERNEL1x3_SUM(0, 0);
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps(kernel[3]), pix10)); MM_KERNEL1x3_SUM(1, 1);
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps(kernel[6]), pix20)); MM_KERNEL1x3_LOAD(2, x+1);
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps(kernel[1]), pix01)); MM_KERNEL1x3_SUM(2, 2);
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps(kernel[4]), pix11));
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps(kernel[7]), pix21));
pix02 = _mm_cvtepi32_ps(_mm_cvtepu8_epi32(*(__m128i *) &(in1)[x+1]));
pix12 = _mm_cvtepi32_ps(_mm_cvtepu8_epi32(*(__m128i *) &(in0)[x+1]));
pix22 = _mm_cvtepi32_ps(_mm_cvtepu8_epi32(*(__m128i *) &(in_1)[x+1]));
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps(kernel[2]), pix02));
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps(kernel[5]), pix12));
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps(kernel[8]), pix22));
ssi0 = _mm_cvtps_epi32(ss); ssi0 = _mm_cvtps_epi32(ss);
ss = _mm_set1_ps(offset); ss = _mm_set1_ps(offset);
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps(kernel[0]), pix01)); MM_KERNEL1x3_SUM(1, 0);
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps(kernel[3]), pix11)); MM_KERNEL1x3_SUM(2, 1);
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps(kernel[6]), pix21)); MM_KERNEL1x3_LOAD(0, x+2);
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps(kernel[1]), pix02)); MM_KERNEL1x3_SUM(0, 2);
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps(kernel[4]), pix12));
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps(kernel[7]), pix22));
pix00 = _mm_cvtepi32_ps(_mm_cvtepu8_epi32(*(__m128i *) &(in1)[x+2]));
pix10 = _mm_cvtepi32_ps(_mm_cvtepu8_epi32(*(__m128i *) &(in0)[x+2]));
pix20 = _mm_cvtepi32_ps(_mm_cvtepu8_epi32(*(__m128i *) &(in_1)[x+2]));
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps(kernel[2]), pix00));
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps(kernel[5]), pix10));
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps(kernel[8]), pix20));
ssi1 = _mm_cvtps_epi32(ss); ssi1 = _mm_cvtps_epi32(ss);
ss = _mm_set1_ps(offset); ss = _mm_set1_ps(offset);
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps(kernel[0]), pix02)); MM_KERNEL1x3_SUM(2, 0);
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps(kernel[3]), pix12)); MM_KERNEL1x3_SUM(0, 1);
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps(kernel[6]), pix22)); MM_KERNEL1x3_LOAD(1, x+3);
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps(kernel[1]), pix00)); MM_KERNEL1x3_SUM(1, 2);
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps(kernel[4]), pix10));
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps(kernel[7]), pix20));
pix01 = _mm_cvtepi32_ps(_mm_cvtepu8_epi32(*(__m128i *) &(in1)[x+3]));
pix11 = _mm_cvtepi32_ps(_mm_cvtepu8_epi32(*(__m128i *) &(in0)[x+3]));
pix21 = _mm_cvtepi32_ps(_mm_cvtepu8_epi32(*(__m128i *) &(in_1)[x+3]));
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps(kernel[2]), pix01));
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps(kernel[5]), pix11));
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps(kernel[8]), pix21));
ssi2 = _mm_cvtps_epi32(ss); ssi2 = _mm_cvtps_epi32(ss);
ssi0 = _mm_packs_epi32(ssi0, ssi1); ssi0 = _mm_packs_epi32(ssi0, ssi1);
@ -212,16 +187,17 @@ ImagingFilter3x3(Imaging imOut, Imaging im, const float *kernel, float offset) {
} }
for (; x < im->xsize-1; x++) { for (; x < im->xsize-1; x++) {
__m128 ss = _mm_set1_ps(offset); __m128 ss = _mm_set1_ps(offset);
__m128i ssi; __m128i ssi0;
MM_KERNEL1x3_LOAD(0, x-1);
MM_KERNEL1x3(ss, in1, x, &kernel[0], 1); MM_KERNEL1x3_SUM(0, 0);
MM_KERNEL1x3(ss, in0, x, &kernel[3], 1); MM_KERNEL1x3_LOAD(0, x+0);
MM_KERNEL1x3(ss, in_1, x, &kernel[6], 1); MM_KERNEL1x3_SUM(0, 1);
MM_KERNEL1x3_LOAD(0, x+1);
ssi = _mm_cvtps_epi32(ss); MM_KERNEL1x3_SUM(0, 2);
ssi = _mm_packs_epi32(ssi, ssi); ssi0 = _mm_cvtps_epi32(ss);
ssi = _mm_packus_epi16(ssi, ssi); ssi0 = _mm_packs_epi32(ssi0, ssi0);
out[x] = _mm_cvtsi128_si32(ssi); ssi0 = _mm_packus_epi16(ssi0, ssi0);
out[x] = _mm_cvtsi128_si32(ssi0);
} }
out[x] = in0[x]; out[x] = in0[x];
} }
@ -237,18 +213,6 @@ ImagingFilter5x5(Imaging imOut, Imaging im, const float *kernel, float offset) {
_i2f((UINT8)in0[x + d]) * (kernel)[3] + \ _i2f((UINT8)in0[x + d]) * (kernel)[3] + \
_i2f((UINT8)in0[x + d + d]) * (kernel)[4]) _i2f((UINT8)in0[x + d + d]) * (kernel)[4])
#define MM_KERNEL1x5(ss, in0, x, kernel, d) \
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps((kernel)[0]), \
_mm_cvtepi32_ps(_mm_cvtepu8_epi32(*(__m128i *) &(in0)[x-d-d])))); \
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps((kernel)[1]), \
_mm_cvtepi32_ps(_mm_cvtepu8_epi32(*(__m128i *) &(in0)[x-d])))); \
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps((kernel)[2]), \
_mm_cvtepi32_ps(_mm_cvtepu8_epi32(*(__m128i *) &(in0)[x+0])))); \
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps((kernel)[3]), \
_mm_cvtepi32_ps(_mm_cvtepu8_epi32(*(__m128i *) &(in0)[x+d])))); \
ss = _mm_add_ps(ss, _mm_mul_ps(_mm_set1_ps((kernel)[4]), \
_mm_cvtepi32_ps(_mm_cvtepu8_epi32(*(__m128i *) &(in0)[x+d+d]))));
#define MM_KERNEL1x5_LOAD(row, x) \ #define MM_KERNEL1x5_LOAD(row, x) \
pix0##row = _mm_cvtepi32_ps(_mm_cvtepu8_epi32(*(__m128i*) &in2[x])); \ pix0##row = _mm_cvtepi32_ps(_mm_cvtepu8_epi32(*(__m128i*) &in2[x])); \
pix1##row = _mm_cvtepi32_ps(_mm_cvtepu8_epi32(*(__m128i*) &in1[x])); \ pix1##row = _mm_cvtepi32_ps(_mm_cvtepu8_epi32(*(__m128i*) &in1[x])); \
@ -372,18 +336,21 @@ ImagingFilter5x5(Imaging imOut, Imaging im, const float *kernel, float offset) {
} }
for (; x < im->xsize-2; x++) { for (; x < im->xsize-2; x++) {
__m128 ss = _mm_set1_ps(offset); __m128 ss = _mm_set1_ps(offset);
__m128i ssi; __m128i ssi0;
MM_KERNEL1x5_LOAD(0, x-2);
MM_KERNEL1x5(ss, in2, x, &kernel[0], 1); MM_KERNEL1x5_SUM(0, 0);
MM_KERNEL1x5(ss, in1, x, &kernel[5], 1); MM_KERNEL1x5_LOAD(0, x-1);
MM_KERNEL1x5(ss, in0, x, &kernel[10], 1); MM_KERNEL1x5_SUM(0, 1);
MM_KERNEL1x5(ss, in_1, x, &kernel[15], 1); MM_KERNEL1x5_LOAD(0, x+0);
MM_KERNEL1x5(ss, in_2, x, &kernel[20], 1); MM_KERNEL1x5_SUM(0, 2);
MM_KERNEL1x5_LOAD(0, x+1);
ssi = _mm_cvtps_epi32(ss); MM_KERNEL1x5_SUM(0, 3);
ssi = _mm_packs_epi32(ssi, ssi); MM_KERNEL1x5_LOAD(0, x+2);
ssi = _mm_packus_epi16(ssi, ssi); MM_KERNEL1x5_SUM(0, 4);
out[x] = _mm_cvtsi128_si32(ssi); ssi0 = _mm_cvtps_epi32(ss);
ssi0 = _mm_packs_epi32(ssi0, ssi0);
ssi0 = _mm_packus_epi16(ssi0, ssi0);
out[x] = _mm_cvtsi128_si32(ssi0);
} }
out[x] = in0[x]; out[x] = in0[x];
out[x+1] = in0[x+1]; out[x+1] = in0[x+1];