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 ColorLUT. table_channels==4 case, minor optimizations

Browse Source
This commit is contained in:
Alexander 2018-03-27 10:29:20 +03:00 committed by Alexander Karpinsky
parent c7e3b0ead9
commit 91dcda6b43
1 changed files with 50 additions and 25 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

75
src/libImaging/ColorLUT.c
View File

@ -80,8 +80,10 @@ ImagingColorLUT3D_linear(
(size2D - 1) / 255.0 * (1<<SCALE_BITS),
(size1D - 1) / 255.0 * (1<<SCALE_BITS));
__m128i scale_mask = _mm_set1_epi32(SCALE_MASK);
__m128i index_mul = _mm_set_epi32(0, size1D_2D, size1D, 1);
__m128i shuffle_source = _mm_set_epi8(-1,-1, -1,-1, 11,10, 5,4, 9,8, 3,2, 7,6, 1,0);
__m128i index_mul = _mm_set_epi32(0, size1D_2D*table_channels,
size1D*table_channels, table_channels);
__m128i shuffle3 = _mm_set_epi8(-1,-1, -1,-1, 11,10, 5,4, 9,8, 3,2, 7,6, 1,0);
__m128i shuffle4 = _mm_set_epi8(15,14, 7,6, 13,12, 5,4, 11,10, 3,2, 9,8, 1,0);
__m128i left_mask = _mm_set1_epi32(0x0000ffff);
__m128i right_mask = _mm_set1_epi32(0xffff0000);
int x, y;
@ -108,17 +110,17 @@ ImagingColorLUT3D_linear(
UINT32* rowOut = (UINT32 *)imOut->image[y];
__m128i index = _mm_mullo_epi32(scale,
_mm_cvtepu8_epi32(*(__m128i *) &rowIn[0]));
int idx = table_channels * _mm_extract_epi32(
int idx = _mm_cvtsi128_si32(
_mm_hadd_epi32(_mm_hadd_epi32(
_mm_madd_epi16(index_mul, _mm_srli_epi32(index, SCALE_BITS)),
_mm_setzero_si128()), _mm_setzero_si128()), 0);
_mm_setzero_si128()), _mm_setzero_si128()));
for (x = 0; x < imOut->xsize; x++) {
__m128i next_index = _mm_mullo_epi32(scale,
_mm_cvtepu8_epi32(*(__m128i *) &rowIn[x*4 + 4]));
int next_idx = table_channels * _mm_extract_epi32(
int next_idx = _mm_cvtsi128_si32(
_mm_hadd_epi32(_mm_hadd_epi32(
_mm_madd_epi16(index_mul, _mm_srli_epi32(next_index, SCALE_BITS)),
_mm_setzero_si128()), _mm_setzero_si128()), 0);
_mm_setzero_si128()), _mm_setzero_si128()));
__m128i shift = _mm_srli_epi32(
_mm_and_si128(scale_mask, index), (SCALE_BITS - SHIFT_BITS));
__m128i shift1D, shift2D, shift3D;
@ -135,22 +137,22 @@ ImagingColorLUT3D_linear(
if (table_channels == 3) {
source = _mm_shuffle_epi8(
_mm_loadu_si128((__m128i *) &table[idx + 0]), shuffle_source);
_mm_loadu_si128((__m128i *) &table[idx + 0]), shuffle3);
leftleft = _mm_and_si128(_mm_srai_epi32(_mm_madd_epi16(
source, shift1D), SHIFT_BITS), left_mask);
source = _mm_shuffle_epi8(
_mm_loadu_si128((__m128i *) &table[idx + size1D*3]), shuffle_source);
_mm_loadu_si128((__m128i *) &table[idx + size1D*3]), shuffle3);
leftright = _mm_and_si128(_mm_slli_epi32(_mm_madd_epi16(
source, shift1D), 16 - SHIFT_BITS), right_mask);
source = _mm_shuffle_epi8(
_mm_loadu_si128((__m128i *) &table[idx + size1D_2D*3]), shuffle_source);
_mm_loadu_si128((__m128i *) &table[idx + size1D_2D*3]), shuffle3);
rightleft = _mm_and_si128(_mm_srai_epi32(_mm_madd_epi16(
source, shift1D), SHIFT_BITS), left_mask);
source = _mm_shuffle_epi8(
_mm_loadu_si128((__m128i *) &table[idx + size1D_2D*3 + size1D*3]), shuffle_source);
_mm_loadu_si128((__m128i *) &table[idx + size1D_2D*3 + size1D*3]), shuffle3);
rightright = _mm_and_si128(_mm_slli_epi32(_mm_madd_epi16(
source, shift1D), 16 - SHIFT_BITS), right_mask);
@ -172,24 +174,47 @@ ImagingColorLUT3D_linear(
rowOut[x] = _mm_cvtsi128_si32(_mm_packus_epi16(result, result));
}
// if (table_channels == 4) {
// interpolate4(leftleft, &table[idx + 0], &table[idx + 4], shift1D);
// interpolate4(leftright, &table[idx + size1D*4],
// &table[idx + size1D*4 + 4], shift1D);
// interpolate4(left, leftleft, leftright, shift2D);
if (table_channels == 4) {
source = _mm_shuffle_epi8(
_mm_loadu_si128((__m128i *) &table[idx + 0]), shuffle4);
leftleft = _mm_and_si128(_mm_srai_epi32(_mm_madd_epi16(
source, shift1D), SHIFT_BITS), left_mask);
// interpolate4(rightleft, &table[idx + size1D_2D*4],
// &table[idx + size1D_2D*4 + 4], shift1D);
// interpolate4(rightright, &table[idx + size1D_2D*4 + size1D*4],
// &table[idx + size1D_2D*4 + size1D*4 + 4], shift1D);
// interpolate4(right, rightleft, rightright, shift2D);
source = _mm_shuffle_epi8(
_mm_loadu_si128((__m128i *) &table[idx + size1D*4]), shuffle4);
leftright = _mm_and_si128(_mm_slli_epi32(_mm_madd_epi16(
source, shift1D), 16 - SHIFT_BITS), right_mask);
// interpolate4(result, left, right, shift3D);
source = _mm_shuffle_epi8(
_mm_loadu_si128((__m128i *) &table[idx + size1D_2D*4]), shuffle4);
rightleft = _mm_and_si128(_mm_srai_epi32(_mm_madd_epi16(
source, shift1D), SHIFT_BITS), left_mask);
source = _mm_shuffle_epi8(
_mm_loadu_si128((__m128i *) &table[idx + size1D_2D*4 + size1D*4]), shuffle4);
rightright = _mm_and_si128(_mm_slli_epi32(_mm_madd_epi16(
source, shift1D), 16 - SHIFT_BITS), right_mask);
// rowOut[x] = MAKE_UINT32(
// clip8(result[0]), clip8(result[1]),
// clip8(result[2]), clip8(result[3]));
// }
left = _mm_and_si128(_mm_srai_epi32(_mm_madd_epi16(
_mm_or_si128(leftleft, leftright), shift2D),
SHIFT_BITS), left_mask);
right = _mm_and_si128(_mm_slli_epi32(_mm_madd_epi16(
_mm_or_si128(rightleft, rightright), shift2D),
16 - SHIFT_BITS), right_mask);
result = _mm_madd_epi16(_mm_or_si128(left, right), shift3D);
result = _mm_srai_epi32(_mm_add_epi32(
_mm_set1_epi32(PRECISION_ROUNDING<<SHIFT_BITS), result),
PRECISION_BITS + SHIFT_BITS);
result = _mm_packs_epi32(result, result);
rowOut[x] = _mm_cvtsi128_si32(_mm_packus_epi16(result, result));
}
idx = next_idx;
index = next_index;
}
}
ImagingSectionLeave(&cookie);