ref: fe8cce2e36ff22d5426afe3271f9c7e32d8dc9ac
dir: /vpx_dsp/x86/vpx_subpixel_8t_intrin_avx2.c/
/*
* Copyright (c) 2010 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include <immintrin.h>
#include <stdio.h>
#include "./vpx_dsp_rtcd.h"
#include "vpx_dsp/x86/convolve.h"
#include "vpx_dsp/x86/convolve_avx2.h"
#include "vpx_dsp/x86/convolve_sse2.h"
#include "vpx_ports/mem.h"
// filters for 16_h8
DECLARE_ALIGNED(32, static const uint8_t,
filt1_global_avx2[32]) = { 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5,
6, 6, 7, 7, 8, 0, 1, 1, 2, 2, 3,
3, 4, 4, 5, 5, 6, 6, 7, 7, 8 };
DECLARE_ALIGNED(32, static const uint8_t,
filt2_global_avx2[32]) = { 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7,
8, 8, 9, 9, 10, 2, 3, 3, 4, 4, 5,
5, 6, 6, 7, 7, 8, 8, 9, 9, 10 };
DECLARE_ALIGNED(32, static const uint8_t, filt3_global_avx2[32]) = {
4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9, 10, 10, 11, 11, 12,
4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9, 10, 10, 11, 11, 12
};
DECLARE_ALIGNED(32, static const uint8_t, filt4_global_avx2[32]) = {
6, 7, 7, 8, 8, 9, 9, 10, 10, 11, 11, 12, 12, 13, 13, 14,
6, 7, 7, 8, 8, 9, 9, 10, 10, 11, 11, 12, 12, 13, 13, 14
};
static INLINE void vpx_filter_block1d16_h8_x_avx2(
const uint8_t *src_ptr, ptrdiff_t src_pixels_per_line, uint8_t *output_ptr,
ptrdiff_t output_pitch, uint32_t output_height, const int16_t *filter,
const int avg) {
__m128i outReg1, outReg2;
__m256i outReg32b1, outReg32b2;
unsigned int i;
ptrdiff_t src_stride, dst_stride;
__m256i f[4], filt[4], s[4];
shuffle_filter_avx2(filter, f);
filt[0] = _mm256_load_si256((__m256i const *)filt1_global_avx2);
filt[1] = _mm256_load_si256((__m256i const *)filt2_global_avx2);
filt[2] = _mm256_load_si256((__m256i const *)filt3_global_avx2);
filt[3] = _mm256_load_si256((__m256i const *)filt4_global_avx2);
// multiple the size of the source and destination stride by two
src_stride = src_pixels_per_line << 1;
dst_stride = output_pitch << 1;
for (i = output_height; i > 1; i -= 2) {
__m256i srcReg;
// load the 2 strides of source
srcReg =
_mm256_castsi128_si256(_mm_loadu_si128((const __m128i *)(src_ptr - 3)));
srcReg = _mm256_inserti128_si256(
srcReg,
_mm_loadu_si128((const __m128i *)(src_ptr + src_pixels_per_line - 3)),
1);
// filter the source buffer
s[0] = _mm256_shuffle_epi8(srcReg, filt[0]);
s[1] = _mm256_shuffle_epi8(srcReg, filt[1]);
s[2] = _mm256_shuffle_epi8(srcReg, filt[2]);
s[3] = _mm256_shuffle_epi8(srcReg, filt[3]);
outReg32b1 = convolve8_16_avx2(s, f);
// reading 2 strides of the next 16 bytes
// (part of it was being read by earlier read)
srcReg =
_mm256_castsi128_si256(_mm_loadu_si128((const __m128i *)(src_ptr + 5)));
srcReg = _mm256_inserti128_si256(
srcReg,
_mm_loadu_si128((const __m128i *)(src_ptr + src_pixels_per_line + 5)),
1);
// filter the source buffer
s[0] = _mm256_shuffle_epi8(srcReg, filt[0]);
s[1] = _mm256_shuffle_epi8(srcReg, filt[1]);
s[2] = _mm256_shuffle_epi8(srcReg, filt[2]);
s[3] = _mm256_shuffle_epi8(srcReg, filt[3]);
outReg32b2 = convolve8_16_avx2(s, f);
// shrink to 8 bit each 16 bits, the low and high 64-bits of each lane
// contain the first and second convolve result respectively
outReg32b1 = _mm256_packus_epi16(outReg32b1, outReg32b2);
src_ptr += src_stride;
// average if necessary
outReg1 = _mm256_castsi256_si128(outReg32b1);
outReg2 = _mm256_extractf128_si256(outReg32b1, 1);
if (avg) {
outReg1 = _mm_avg_epu8(outReg1, _mm_load_si128((__m128i *)output_ptr));
outReg2 = _mm_avg_epu8(
outReg2, _mm_load_si128((__m128i *)(output_ptr + output_pitch)));
}
// save 16 bytes
_mm_store_si128((__m128i *)output_ptr, outReg1);
// save the next 16 bits
_mm_store_si128((__m128i *)(output_ptr + output_pitch), outReg2);
output_ptr += dst_stride;
}
// if the number of strides is odd.
// process only 16 bytes
if (i > 0) {
__m128i srcReg;
// load the first 16 bytes of the last row
srcReg = _mm_loadu_si128((const __m128i *)(src_ptr - 3));
// filter the source buffer
s[0] = _mm256_castsi128_si256(
_mm_shuffle_epi8(srcReg, _mm256_castsi256_si128(filt[0])));
s[1] = _mm256_castsi128_si256(
_mm_shuffle_epi8(srcReg, _mm256_castsi256_si128(filt[1])));
s[2] = _mm256_castsi128_si256(
_mm_shuffle_epi8(srcReg, _mm256_castsi256_si128(filt[2])));
s[3] = _mm256_castsi128_si256(
_mm_shuffle_epi8(srcReg, _mm256_castsi256_si128(filt[3])));
outReg1 = convolve8_8_avx2(s, f);
// reading the next 16 bytes
// (part of it was being read by earlier read)
srcReg = _mm_loadu_si128((const __m128i *)(src_ptr + 5));
// filter the source buffer
s[0] = _mm256_castsi128_si256(
_mm_shuffle_epi8(srcReg, _mm256_castsi256_si128(filt[0])));
s[1] = _mm256_castsi128_si256(
_mm_shuffle_epi8(srcReg, _mm256_castsi256_si128(filt[1])));
s[2] = _mm256_castsi128_si256(
_mm_shuffle_epi8(srcReg, _mm256_castsi256_si128(filt[2])));
s[3] = _mm256_castsi128_si256(
_mm_shuffle_epi8(srcReg, _mm256_castsi256_si128(filt[3])));
outReg2 = convolve8_8_avx2(s, f);
// shrink to 8 bit each 16 bits, the low and high 64-bits of each lane
// contain the first and second convolve result respectively
outReg1 = _mm_packus_epi16(outReg1, outReg2);
// average if necessary
if (avg) {
outReg1 = _mm_avg_epu8(outReg1, _mm_load_si128((__m128i *)output_ptr));
}
// save 16 bytes
_mm_store_si128((__m128i *)output_ptr, outReg1);
}
}
static void vpx_filter_block1d16_h8_avx2(
const uint8_t *src_ptr, ptrdiff_t src_stride, uint8_t *output_ptr,
ptrdiff_t dst_stride, uint32_t output_height, const int16_t *filter) {
vpx_filter_block1d16_h8_x_avx2(src_ptr, src_stride, output_ptr, dst_stride,
output_height, filter, 0);
}
static void vpx_filter_block1d16_h8_avg_avx2(
const uint8_t *src_ptr, ptrdiff_t src_stride, uint8_t *output_ptr,
ptrdiff_t dst_stride, uint32_t output_height, const int16_t *filter) {
vpx_filter_block1d16_h8_x_avx2(src_ptr, src_stride, output_ptr, dst_stride,
output_height, filter, 1);
}
static INLINE void vpx_filter_block1d16_v8_x_avx2(
const uint8_t *src_ptr, ptrdiff_t src_pitch, uint8_t *output_ptr,
ptrdiff_t out_pitch, uint32_t output_height, const int16_t *filter,
const int avg) {
__m128i outReg1, outReg2;
__m256i srcRegHead1;
unsigned int i;
ptrdiff_t src_stride, dst_stride;
__m256i f[4], s1[4], s2[4];
shuffle_filter_avx2(filter, f);
// multiple the size of the source and destination stride by two
src_stride = src_pitch << 1;
dst_stride = out_pitch << 1;
{
__m128i s[6];
__m256i s32b[6];
// load 16 bytes 7 times in stride of src_pitch
s[0] = _mm_loadu_si128((const __m128i *)(src_ptr + 0 * src_pitch));
s[1] = _mm_loadu_si128((const __m128i *)(src_ptr + 1 * src_pitch));
s[2] = _mm_loadu_si128((const __m128i *)(src_ptr + 2 * src_pitch));
s[3] = _mm_loadu_si128((const __m128i *)(src_ptr + 3 * src_pitch));
s[4] = _mm_loadu_si128((const __m128i *)(src_ptr + 4 * src_pitch));
s[5] = _mm_loadu_si128((const __m128i *)(src_ptr + 5 * src_pitch));
srcRegHead1 = _mm256_castsi128_si256(
_mm_loadu_si128((const __m128i *)(src_ptr + 6 * src_pitch)));
// have each consecutive loads on the same 256 register
s32b[0] = _mm256_inserti128_si256(_mm256_castsi128_si256(s[0]), s[1], 1);
s32b[1] = _mm256_inserti128_si256(_mm256_castsi128_si256(s[1]), s[2], 1);
s32b[2] = _mm256_inserti128_si256(_mm256_castsi128_si256(s[2]), s[3], 1);
s32b[3] = _mm256_inserti128_si256(_mm256_castsi128_si256(s[3]), s[4], 1);
s32b[4] = _mm256_inserti128_si256(_mm256_castsi128_si256(s[4]), s[5], 1);
s32b[5] = _mm256_inserti128_si256(_mm256_castsi128_si256(s[5]),
_mm256_castsi256_si128(srcRegHead1), 1);
// merge every two consecutive registers except the last one
// the first lanes contain values for filtering odd rows (1,3,5...) and
// the second lanes contain values for filtering even rows (2,4,6...)
s1[0] = _mm256_unpacklo_epi8(s32b[0], s32b[1]);
s2[0] = _mm256_unpackhi_epi8(s32b[0], s32b[1]);
s1[1] = _mm256_unpacklo_epi8(s32b[2], s32b[3]);
s2[1] = _mm256_unpackhi_epi8(s32b[2], s32b[3]);
s1[2] = _mm256_unpacklo_epi8(s32b[4], s32b[5]);
s2[2] = _mm256_unpackhi_epi8(s32b[4], s32b[5]);
}
for (i = output_height; i > 1; i -= 2) {
__m256i srcRegHead2, srcRegHead3;
// load the next 2 loads of 16 bytes and have every two
// consecutive loads in the same 256 bit register
srcRegHead2 = _mm256_castsi128_si256(
_mm_loadu_si128((const __m128i *)(src_ptr + 7 * src_pitch)));
srcRegHead1 = _mm256_inserti128_si256(
srcRegHead1, _mm256_castsi256_si128(srcRegHead2), 1);
srcRegHead3 = _mm256_castsi128_si256(
_mm_loadu_si128((const __m128i *)(src_ptr + 8 * src_pitch)));
srcRegHead2 = _mm256_inserti128_si256(
srcRegHead2, _mm256_castsi256_si128(srcRegHead3), 1);
// merge the two new consecutive registers
// the first lane contain values for filtering odd rows (1,3,5...) and
// the second lane contain values for filtering even rows (2,4,6...)
s1[3] = _mm256_unpacklo_epi8(srcRegHead1, srcRegHead2);
s2[3] = _mm256_unpackhi_epi8(srcRegHead1, srcRegHead2);
s1[0] = convolve8_16_avx2(s1, f);
s2[0] = convolve8_16_avx2(s2, f);
// shrink to 8 bit each 16 bits, the low and high 64-bits of each lane
// contain the first and second convolve result respectively
s1[0] = _mm256_packus_epi16(s1[0], s2[0]);
src_ptr += src_stride;
// average if necessary
outReg1 = _mm256_castsi256_si128(s1[0]);
outReg2 = _mm256_extractf128_si256(s1[0], 1);
if (avg) {
outReg1 = _mm_avg_epu8(outReg1, _mm_load_si128((__m128i *)output_ptr));
outReg2 = _mm_avg_epu8(
outReg2, _mm_load_si128((__m128i *)(output_ptr + out_pitch)));
}
// save 16 bytes
_mm_store_si128((__m128i *)output_ptr, outReg1);
// save the next 16 bits
_mm_store_si128((__m128i *)(output_ptr + out_pitch), outReg2);
output_ptr += dst_stride;
// shift down by two rows
s1[0] = s1[1];
s2[0] = s2[1];
s1[1] = s1[2];
s2[1] = s2[2];
s1[2] = s1[3];
s2[2] = s2[3];
srcRegHead1 = srcRegHead3;
}
// if the number of strides is odd.
// process only 16 bytes
if (i > 0) {
// load the last 16 bytes
const __m128i srcRegHead2 =
_mm_loadu_si128((const __m128i *)(src_ptr + src_pitch * 7));
// merge the last 2 results together
s1[0] = _mm256_castsi128_si256(
_mm_unpacklo_epi8(_mm256_castsi256_si128(srcRegHead1), srcRegHead2));
s2[0] = _mm256_castsi128_si256(
_mm_unpackhi_epi8(_mm256_castsi256_si128(srcRegHead1), srcRegHead2));
outReg1 = convolve8_8_avx2(s1, f);
outReg2 = convolve8_8_avx2(s2, f);
// shrink to 8 bit each 16 bits, the low and high 64-bits of each lane
// contain the first and second convolve result respectively
outReg1 = _mm_packus_epi16(outReg1, outReg2);
// average if necessary
if (avg) {
outReg1 = _mm_avg_epu8(outReg1, _mm_load_si128((__m128i *)output_ptr));
}
// save 16 bytes
_mm_store_si128((__m128i *)output_ptr, outReg1);
}
}
static void vpx_filter_block1d16_v8_avx2(const uint8_t *src_ptr,
ptrdiff_t src_stride, uint8_t *dst_ptr,
ptrdiff_t dst_stride, uint32_t height,
const int16_t *filter) {
vpx_filter_block1d16_v8_x_avx2(src_ptr, src_stride, dst_ptr, dst_stride,
height, filter, 0);
}
static void vpx_filter_block1d16_v8_avg_avx2(
const uint8_t *src_ptr, ptrdiff_t src_stride, uint8_t *dst_ptr,
ptrdiff_t dst_stride, uint32_t height, const int16_t *filter) {
vpx_filter_block1d16_v8_x_avx2(src_ptr, src_stride, dst_ptr, dst_stride,
height, filter, 1);
}
static void vpx_filter_block1d16_h4_avx2(const uint8_t *src_ptr,
ptrdiff_t src_stride, uint8_t *dst_ptr,
ptrdiff_t dst_stride, uint32_t height,
const int16_t *kernel) {
// We will cast the kernel from 16-bit words to 8-bit words, and then extract
// the middle four elements of the kernel into two registers in the form
// ... k[3] k[2] k[3] k[2]
// ... k[5] k[4] k[5] k[4]
// Then we shuffle the source into
// ... s[1] s[0] s[0] s[-1]
// ... s[3] s[2] s[2] s[1]
// Calling multiply and add gives us half of the sum. Calling add gives us
// first half of the output. Repeat again to get the second half of the
// output. Finally we shuffle again to combine the two outputs.
// Since avx2 allows us to use 256-bit buffer, we can do this two rows at a
// time.
__m128i kernel_reg; // Kernel
__m256i kernel_reg_256, kernel_reg_23,
kernel_reg_45; // Segments of the kernel used
const __m256i reg_32 = _mm256_set1_epi16(32); // Used for rounding
const ptrdiff_t unrolled_src_stride = src_stride << 1;
const ptrdiff_t unrolled_dst_stride = dst_stride << 1;
int h;
__m256i src_reg, src_reg_shift_0, src_reg_shift_2;
__m256i dst_first, dst_second;
__m256i tmp_0, tmp_1;
__m256i idx_shift_0 =
_mm256_setr_epi8(0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 0, 1, 1,
2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8);
__m256i idx_shift_2 =
_mm256_setr_epi8(2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9, 10, 2, 3, 3,
4, 4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9, 10);
// Start one pixel before as we need tap/2 - 1 = 1 sample from the past
src_ptr -= 1;
// Load Kernel
kernel_reg = _mm_loadu_si128((const __m128i *)kernel);
kernel_reg = _mm_srai_epi16(kernel_reg, 1);
kernel_reg = _mm_packs_epi16(kernel_reg, kernel_reg);
kernel_reg_256 = _mm256_broadcastsi128_si256(kernel_reg);
kernel_reg_23 =
_mm256_shuffle_epi8(kernel_reg_256, _mm256_set1_epi16(0x0302u));
kernel_reg_45 =
_mm256_shuffle_epi8(kernel_reg_256, _mm256_set1_epi16(0x0504u));
for (h = height; h >= 2; h -= 2) {
// Load the source
src_reg = mm256_loadu2_si128(src_ptr, src_ptr + src_stride);
src_reg_shift_0 = _mm256_shuffle_epi8(src_reg, idx_shift_0);
src_reg_shift_2 = _mm256_shuffle_epi8(src_reg, idx_shift_2);
// Partial result for first half
tmp_0 = _mm256_maddubs_epi16(src_reg_shift_0, kernel_reg_23);
tmp_1 = _mm256_maddubs_epi16(src_reg_shift_2, kernel_reg_45);
dst_first = _mm256_adds_epi16(tmp_0, tmp_1);
// Do again to get the second half of dst
// Load the source
src_reg = mm256_loadu2_si128(src_ptr + 8, src_ptr + src_stride + 8);
src_reg_shift_0 = _mm256_shuffle_epi8(src_reg, idx_shift_0);
src_reg_shift_2 = _mm256_shuffle_epi8(src_reg, idx_shift_2);
// Partial result for second half
tmp_0 = _mm256_maddubs_epi16(src_reg_shift_0, kernel_reg_23);
tmp_1 = _mm256_maddubs_epi16(src_reg_shift_2, kernel_reg_45);
dst_second = _mm256_adds_epi16(tmp_0, tmp_1);
// Round each result
dst_first = mm256_round_epi16(&dst_first, ®_32, 6);
dst_second = mm256_round_epi16(&dst_second, ®_32, 6);
// Finally combine to get the final dst
dst_first = _mm256_packus_epi16(dst_first, dst_second);
mm256_store2_si128((__m128i *)dst_ptr, (__m128i *)(dst_ptr + dst_stride),
&dst_first);
src_ptr += unrolled_src_stride;
dst_ptr += unrolled_dst_stride;
}
// Repeat for the last row if needed
if (h > 0) {
src_reg = _mm256_loadu_si256((const __m256i *)src_ptr);
// Reorder into 2 1 1 2
src_reg = _mm256_permute4x64_epi64(src_reg, 0x94);
src_reg_shift_0 = _mm256_shuffle_epi8(src_reg, idx_shift_0);
src_reg_shift_2 = _mm256_shuffle_epi8(src_reg, idx_shift_2);
tmp_0 = _mm256_maddubs_epi16(src_reg_shift_0, kernel_reg_23);
tmp_1 = _mm256_maddubs_epi16(src_reg_shift_2, kernel_reg_45);
dst_first = _mm256_adds_epi16(tmp_0, tmp_1);
dst_first = mm256_round_epi16(&dst_first, ®_32, 6);
dst_first = _mm256_packus_epi16(dst_first, dst_first);
dst_first = _mm256_permute4x64_epi64(dst_first, 0x8);
_mm_store_si128((__m128i *)dst_ptr, _mm256_castsi256_si128(dst_first));
}
}
static void vpx_filter_block1d16_v4_avx2(const uint8_t *src_ptr,
ptrdiff_t src_stride, uint8_t *dst_ptr,
ptrdiff_t dst_stride, uint32_t height,
const int16_t *kernel) {
// We will load two rows of pixels as 8-bit words, rearrange them into the
// form
// ... s[1,0] s[0,0] s[0,0] s[-1,0]
// so that we can call multiply and add with the kernel partial output. Then
// we can call add with another row to get the output.
// Register for source s[-1:3, :]
__m256i src_reg_1, src_reg_2, src_reg_3;
// Interleaved rows of the source. lo is first half, hi second
__m256i src_reg_m10, src_reg_01, src_reg_12, src_reg_23;
__m256i src_reg_m1001_lo, src_reg_m1001_hi, src_reg_1223_lo, src_reg_1223_hi;
__m128i kernel_reg; // Kernel
__m256i kernel_reg_256, kernel_reg_23,
kernel_reg_45; // Segments of the kernel used
// Result after multiply and add
__m256i res_reg_m1001_lo, res_reg_1223_lo, res_reg_m1001_hi, res_reg_1223_hi;
__m256i res_reg, res_reg_lo, res_reg_hi;
const __m256i reg_32 = _mm256_set1_epi16(32); // Used for rounding
// We will compute the result two rows at a time
const ptrdiff_t src_stride_unrolled = src_stride << 1;
const ptrdiff_t dst_stride_unrolled = dst_stride << 1;
int h;
// Load Kernel
kernel_reg = _mm_loadu_si128((const __m128i *)kernel);
kernel_reg = _mm_srai_epi16(kernel_reg, 1);
kernel_reg = _mm_packs_epi16(kernel_reg, kernel_reg);
kernel_reg_256 = _mm256_broadcastsi128_si256(kernel_reg);
kernel_reg_23 =
_mm256_shuffle_epi8(kernel_reg_256, _mm256_set1_epi16(0x0302u));
kernel_reg_45 =
_mm256_shuffle_epi8(kernel_reg_256, _mm256_set1_epi16(0x0504u));
// Row -1 to row 0
src_reg_m10 = mm256_loadu2_si128((const __m128i *)src_ptr,
(const __m128i *)(src_ptr + src_stride));
// Row 0 to row 1
src_reg_1 = _mm256_castsi128_si256(
_mm_loadu_si128((const __m128i *)(src_ptr + src_stride * 2)));
src_reg_01 = _mm256_permute2x128_si256(src_reg_m10, src_reg_1, 0x21);
// First three rows
src_reg_m1001_lo = _mm256_unpacklo_epi8(src_reg_m10, src_reg_01);
src_reg_m1001_hi = _mm256_unpackhi_epi8(src_reg_m10, src_reg_01);
for (h = height; h > 1; h -= 2) {
src_reg_2 = _mm256_castsi128_si256(
_mm_loadu_si128((const __m128i *)(src_ptr + src_stride * 3)));
src_reg_12 = _mm256_inserti128_si256(src_reg_1,
_mm256_castsi256_si128(src_reg_2), 1);
src_reg_3 = _mm256_castsi128_si256(
_mm_loadu_si128((const __m128i *)(src_ptr + src_stride * 4)));
src_reg_23 = _mm256_inserti128_si256(src_reg_2,
_mm256_castsi256_si128(src_reg_3), 1);
// Last three rows
src_reg_1223_lo = _mm256_unpacklo_epi8(src_reg_12, src_reg_23);
src_reg_1223_hi = _mm256_unpackhi_epi8(src_reg_12, src_reg_23);
// Output from first half
res_reg_m1001_lo = _mm256_maddubs_epi16(src_reg_m1001_lo, kernel_reg_23);
res_reg_1223_lo = _mm256_maddubs_epi16(src_reg_1223_lo, kernel_reg_45);
res_reg_lo = _mm256_adds_epi16(res_reg_m1001_lo, res_reg_1223_lo);
// Output from second half
res_reg_m1001_hi = _mm256_maddubs_epi16(src_reg_m1001_hi, kernel_reg_23);
res_reg_1223_hi = _mm256_maddubs_epi16(src_reg_1223_hi, kernel_reg_45);
res_reg_hi = _mm256_adds_epi16(res_reg_m1001_hi, res_reg_1223_hi);
// Round the words
res_reg_lo = mm256_round_epi16(&res_reg_lo, ®_32, 6);
res_reg_hi = mm256_round_epi16(&res_reg_hi, ®_32, 6);
// Combine to get the result
res_reg = _mm256_packus_epi16(res_reg_lo, res_reg_hi);
// Save the result
mm256_store2_si128((__m128i *)dst_ptr, (__m128i *)(dst_ptr + dst_stride),
&res_reg);
// Update the source by two rows
src_ptr += src_stride_unrolled;
dst_ptr += dst_stride_unrolled;
src_reg_m1001_lo = src_reg_1223_lo;
src_reg_m1001_hi = src_reg_1223_hi;
src_reg_1 = src_reg_3;
}
}
static void vpx_filter_block1d8_h4_avx2(const uint8_t *src_ptr,
ptrdiff_t src_stride, uint8_t *dst_ptr,
ptrdiff_t dst_stride, uint32_t height,
const int16_t *kernel) {
// We will cast the kernel from 16-bit words to 8-bit words, and then extract
// the middle four elements of the kernel into two registers in the form
// ... k[3] k[2] k[3] k[2]
// ... k[5] k[4] k[5] k[4]
// Then we shuffle the source into
// ... s[1] s[0] s[0] s[-1]
// ... s[3] s[2] s[2] s[1]
// Calling multiply and add gives us half of the sum. Calling add gives us
// first half of the output. Repeat again to get the second half of the
// output. Finally we shuffle again to combine the two outputs.
// Since avx2 allows us to use 256-bit buffer, we can do this two rows at a
// time.
__m128i kernel_reg_128; // Kernel
__m256i kernel_reg, kernel_reg_23,
kernel_reg_45; // Segments of the kernel used
const __m256i reg_32 = _mm256_set1_epi16(32); // Used for rounding
const ptrdiff_t unrolled_src_stride = src_stride << 1;
const ptrdiff_t unrolled_dst_stride = dst_stride << 1;
int h;
__m256i src_reg, src_reg_shift_0, src_reg_shift_2;
__m256i dst_reg;
__m256i tmp_0, tmp_1;
__m256i idx_shift_0 =
_mm256_setr_epi8(0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 0, 1, 1,
2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8);
__m256i idx_shift_2 =
_mm256_setr_epi8(2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9, 10, 2, 3, 3,
4, 4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9, 10);
// Start one pixel before as we need tap/2 - 1 = 1 sample from the past
src_ptr -= 1;
// Load Kernel
kernel_reg_128 = _mm_loadu_si128((const __m128i *)kernel);
kernel_reg_128 = _mm_srai_epi16(kernel_reg_128, 1);
kernel_reg_128 = _mm_packs_epi16(kernel_reg_128, kernel_reg_128);
kernel_reg = _mm256_broadcastsi128_si256(kernel_reg_128);
kernel_reg_23 = _mm256_shuffle_epi8(kernel_reg, _mm256_set1_epi16(0x0302u));
kernel_reg_45 = _mm256_shuffle_epi8(kernel_reg, _mm256_set1_epi16(0x0504u));
for (h = height; h >= 2; h -= 2) {
// Load the source
src_reg = mm256_loadu2_si128(src_ptr, src_ptr + src_stride);
src_reg_shift_0 = _mm256_shuffle_epi8(src_reg, idx_shift_0);
src_reg_shift_2 = _mm256_shuffle_epi8(src_reg, idx_shift_2);
// Get the output
tmp_0 = _mm256_maddubs_epi16(src_reg_shift_0, kernel_reg_23);
tmp_1 = _mm256_maddubs_epi16(src_reg_shift_2, kernel_reg_45);
dst_reg = _mm256_adds_epi16(tmp_0, tmp_1);
// Round the result
dst_reg = mm256_round_epi16(&dst_reg, ®_32, 6);
// Finally combine to get the final dst
dst_reg = _mm256_packus_epi16(dst_reg, dst_reg);
mm256_storeu2_epi64((__m128i *)dst_ptr, (__m128i *)(dst_ptr + dst_stride),
&dst_reg);
src_ptr += unrolled_src_stride;
dst_ptr += unrolled_dst_stride;
}
// Repeat for the last row if needed
if (h > 0) {
__m128i src_reg = _mm_loadu_si128((const __m128i *)src_ptr);
__m128i dst_reg;
const __m128i reg_32 = _mm_set1_epi16(32); // Used for rounding
__m128i tmp_0, tmp_1;
__m128i src_reg_shift_0 =
_mm_shuffle_epi8(src_reg, _mm256_castsi256_si128(idx_shift_0));
__m128i src_reg_shift_2 =
_mm_shuffle_epi8(src_reg, _mm256_castsi256_si128(idx_shift_2));
tmp_0 = _mm_maddubs_epi16(src_reg_shift_0,
_mm256_castsi256_si128(kernel_reg_23));
tmp_1 = _mm_maddubs_epi16(src_reg_shift_2,
_mm256_castsi256_si128(kernel_reg_45));
dst_reg = _mm_adds_epi16(tmp_0, tmp_1);
dst_reg = mm_round_epi16_sse2(&dst_reg, ®_32, 6);
dst_reg = _mm_packus_epi16(dst_reg, _mm_setzero_si128());
_mm_storel_epi64((__m128i *)dst_ptr, dst_reg);
}
}
static void vpx_filter_block1d8_v4_avx2(const uint8_t *src_ptr,
ptrdiff_t src_stride, uint8_t *dst_ptr,
ptrdiff_t dst_stride, uint32_t height,
const int16_t *kernel) {
// We will load two rows of pixels as 8-bit words, rearrange them into the
// form
// ... s[1,0] s[0,0] s[0,0] s[-1,0]
// so that we can call multiply and add with the kernel partial output. Then
// we can call add with another row to get the output.
// Register for source s[-1:3, :]
__m256i src_reg_1, src_reg_2, src_reg_3;
// Interleaved rows of the source. lo is first half, hi second
__m256i src_reg_m10, src_reg_01, src_reg_12, src_reg_23;
__m256i src_reg_m1001, src_reg_1223;
__m128i kernel_reg_128; // Kernel
__m256i kernel_reg, kernel_reg_23,
kernel_reg_45; // Segments of the kernel used
// Result after multiply and add
__m256i res_reg_m1001, res_reg_1223;
__m256i res_reg;
const __m256i reg_32 = _mm256_set1_epi16(32); // Used for rounding
// We will compute the result two rows at a time
const ptrdiff_t src_stride_unrolled = src_stride << 1;
const ptrdiff_t dst_stride_unrolled = dst_stride << 1;
int h;
// Load Kernel
kernel_reg_128 = _mm_loadu_si128((const __m128i *)kernel);
kernel_reg_128 = _mm_srai_epi16(kernel_reg_128, 1);
kernel_reg_128 = _mm_packs_epi16(kernel_reg_128, kernel_reg_128);
kernel_reg = _mm256_broadcastsi128_si256(kernel_reg_128);
kernel_reg_23 = _mm256_shuffle_epi8(kernel_reg, _mm256_set1_epi16(0x0302u));
kernel_reg_45 = _mm256_shuffle_epi8(kernel_reg, _mm256_set1_epi16(0x0504u));
// Row -1 to row 0
src_reg_m10 = mm256_loadu2_epi64((const __m128i *)src_ptr,
(const __m128i *)(src_ptr + src_stride));
// Row 0 to row 1
src_reg_1 = _mm256_castsi128_si256(
_mm_loadu_si128((const __m128i *)(src_ptr + src_stride * 2)));
src_reg_01 = _mm256_permute2x128_si256(src_reg_m10, src_reg_1, 0x21);
// First three rows
src_reg_m1001 = _mm256_unpacklo_epi8(src_reg_m10, src_reg_01);
for (h = height; h > 1; h -= 2) {
src_reg_2 = _mm256_castsi128_si256(
_mm_loadl_epi64((const __m128i *)(src_ptr + src_stride * 3)));
src_reg_12 = _mm256_inserti128_si256(src_reg_1,
_mm256_castsi256_si128(src_reg_2), 1);
src_reg_3 = _mm256_castsi128_si256(
_mm_loadl_epi64((const __m128i *)(src_ptr + src_stride * 4)));
src_reg_23 = _mm256_inserti128_si256(src_reg_2,
_mm256_castsi256_si128(src_reg_3), 1);
// Last three rows
src_reg_1223 = _mm256_unpacklo_epi8(src_reg_12, src_reg_23);
// Output
res_reg_m1001 = _mm256_maddubs_epi16(src_reg_m1001, kernel_reg_23);
res_reg_1223 = _mm256_maddubs_epi16(src_reg_1223, kernel_reg_45);
res_reg = _mm256_adds_epi16(res_reg_m1001, res_reg_1223);
// Round the words
res_reg = mm256_round_epi16(&res_reg, ®_32, 6);
// Combine to get the result
res_reg = _mm256_packus_epi16(res_reg, res_reg);
// Save the result
mm256_storeu2_epi64((__m128i *)dst_ptr, (__m128i *)(dst_ptr + dst_stride),
&res_reg);
// Update the source by two rows
src_ptr += src_stride_unrolled;
dst_ptr += dst_stride_unrolled;
src_reg_m1001 = src_reg_1223;
src_reg_1 = src_reg_3;
}
}
static void vpx_filter_block1d4_h4_avx2(const uint8_t *src_ptr,
ptrdiff_t src_stride, uint8_t *dst_ptr,
ptrdiff_t dst_stride, uint32_t height,
const int16_t *kernel) {
// We will cast the kernel from 16-bit words to 8-bit words, and then extract
// the middle four elements of the kernel into a single register in the form
// k[5:2] k[5:2] k[5:2] k[5:2]
// Then we shuffle the source into
// s[5:2] s[4:1] s[3:0] s[2:-1]
// Calling multiply and add gives us half of the sum next to each other.
// Calling horizontal add then gives us the output.
// Since avx2 has 256-bit register, we can do 2 rows at a time.
__m128i kernel_reg_128; // Kernel
__m256i kernel_reg;
const __m256i reg_32 = _mm256_set1_epi16(32); // Used for rounding
int h;
const ptrdiff_t unrolled_src_stride = src_stride << 1;
const ptrdiff_t unrolled_dst_stride = dst_stride << 1;
__m256i src_reg, src_reg_shuf;
__m256i dst;
__m256i shuf_idx =
_mm256_setr_epi8(0, 1, 2, 3, 1, 2, 3, 4, 2, 3, 4, 5, 3, 4, 5, 6, 0, 1, 2,
3, 1, 2, 3, 4, 2, 3, 4, 5, 3, 4, 5, 6);
// Start one pixel before as we need tap/2 - 1 = 1 sample from the past
src_ptr -= 1;
// Load Kernel
kernel_reg_128 = _mm_loadu_si128((const __m128i *)kernel);
kernel_reg_128 = _mm_srai_epi16(kernel_reg_128, 1);
kernel_reg_128 = _mm_packs_epi16(kernel_reg_128, kernel_reg_128);
kernel_reg = _mm256_broadcastsi128_si256(kernel_reg_128);
kernel_reg = _mm256_shuffle_epi8(kernel_reg, _mm256_set1_epi32(0x05040302u));
for (h = height; h > 1; h -= 2) {
// Load the source
src_reg = mm256_loadu2_epi64((const __m128i *)src_ptr,
(const __m128i *)(src_ptr + src_stride));
src_reg_shuf = _mm256_shuffle_epi8(src_reg, shuf_idx);
// Get the result
dst = _mm256_maddubs_epi16(src_reg_shuf, kernel_reg);
dst = _mm256_hadds_epi16(dst, _mm256_setzero_si256());
// Round result
dst = mm256_round_epi16(&dst, ®_32, 6);
// Pack to 8-bits
dst = _mm256_packus_epi16(dst, _mm256_setzero_si256());
// Save
mm256_storeu2_epi32((__m128i *const)dst_ptr,
(__m128i *const)(dst_ptr + dst_stride), &dst);
src_ptr += unrolled_src_stride;
dst_ptr += unrolled_dst_stride;
}
if (h > 0) {
// Load the source
const __m128i reg_32 = _mm_set1_epi16(32); // Used for rounding
__m128i src_reg = _mm_loadl_epi64((const __m128i *)src_ptr);
__m128i src_reg_shuf =
_mm_shuffle_epi8(src_reg, _mm256_castsi256_si128(shuf_idx));
// Get the result
__m128i dst =
_mm_maddubs_epi16(src_reg_shuf, _mm256_castsi256_si128(kernel_reg));
dst = _mm_hadds_epi16(dst, _mm_setzero_si128());
// Round result
dst = mm_round_epi16_sse2(&dst, ®_32, 6);
// Pack to 8-bits
dst = _mm_packus_epi16(dst, _mm_setzero_si128());
*((uint32_t *)(dst_ptr)) = _mm_cvtsi128_si32(dst);
}
}
static void vpx_filter_block1d4_v4_avx2(const uint8_t *src_ptr,
ptrdiff_t src_stride, uint8_t *dst_ptr,
ptrdiff_t dst_stride, uint32_t height,
const int16_t *kernel) {
// We will load two rows of pixels as 8-bit words, rearrange them into the
// form
// ... s[3,0] s[2,0] s[1,0] s[0,0] s[2,0] s[1,0] s[0,0] s[-1,0]
// so that we can call multiply and add with the kernel to get partial output.
// Calling horizontal add then gives us the completely output
// Register for source s[-1:3, :]
__m256i src_reg_1, src_reg_2, src_reg_3;
// Interleaved rows of the source. lo is first half, hi second
__m256i src_reg_m10, src_reg_01, src_reg_12, src_reg_23;
__m256i src_reg_m1001, src_reg_1223, src_reg_m1012_1023;
__m128i kernel_reg_128; // Kernel
__m256i kernel_reg;
// Result after multiply and add
__m256i res_reg;
const __m256i reg_32 = _mm256_set1_epi16(32); // Used for rounding
// We will compute the result two rows at a time
const ptrdiff_t src_stride_unrolled = src_stride << 1;
const ptrdiff_t dst_stride_unrolled = dst_stride << 1;
int h;
// Load Kernel
kernel_reg_128 = _mm_loadu_si128((const __m128i *)kernel);
kernel_reg_128 = _mm_srai_epi16(kernel_reg_128, 1);
kernel_reg_128 = _mm_packs_epi16(kernel_reg_128, kernel_reg_128);
kernel_reg = _mm256_broadcastsi128_si256(kernel_reg_128);
kernel_reg = _mm256_shuffle_epi8(kernel_reg, _mm256_set1_epi32(0x05040302u));
// Row -1 to row 0
src_reg_m10 = mm256_loadu2_si128((const __m128i *)src_ptr,
(const __m128i *)(src_ptr + src_stride));
// Row 0 to row 1
src_reg_1 = _mm256_castsi128_si256(
_mm_loadu_si128((const __m128i *)(src_ptr + src_stride * 2)));
src_reg_01 = _mm256_permute2x128_si256(src_reg_m10, src_reg_1, 0x21);
// First three rows
src_reg_m1001 = _mm256_unpacklo_epi8(src_reg_m10, src_reg_01);
for (h = height; h > 1; h -= 2) {
src_reg_2 = _mm256_castsi128_si256(
_mm_loadl_epi64((const __m128i *)(src_ptr + src_stride * 3)));
src_reg_12 = _mm256_inserti128_si256(src_reg_1,
_mm256_castsi256_si128(src_reg_2), 1);
src_reg_3 = _mm256_castsi128_si256(
_mm_loadl_epi64((const __m128i *)(src_ptr + src_stride * 4)));
src_reg_23 = _mm256_inserti128_si256(src_reg_2,
_mm256_castsi256_si128(src_reg_3), 1);
// Last three rows
src_reg_1223 = _mm256_unpacklo_epi8(src_reg_12, src_reg_23);
// Combine all the rows
src_reg_m1012_1023 = _mm256_unpacklo_epi16(src_reg_m1001, src_reg_1223);
// Output
res_reg = _mm256_maddubs_epi16(src_reg_m1012_1023, kernel_reg);
res_reg = _mm256_hadds_epi16(res_reg, _mm256_setzero_si256());
// Round the words
res_reg = mm256_round_epi16(&res_reg, ®_32, 6);
// Combine to get the result
res_reg = _mm256_packus_epi16(res_reg, res_reg);
// Save the result
mm256_storeu2_epi32((__m128i *)dst_ptr, (__m128i *)(dst_ptr + dst_stride),
&res_reg);
// Update the source by two rows
src_ptr += src_stride_unrolled;
dst_ptr += dst_stride_unrolled;
src_reg_m1001 = src_reg_1223;
src_reg_1 = src_reg_3;
}
}
#if HAVE_AVX2 && HAVE_SSSE3
filter8_1dfunction vpx_filter_block1d4_v8_ssse3;
#if VPX_ARCH_X86_64
filter8_1dfunction vpx_filter_block1d8_v8_intrin_ssse3;
filter8_1dfunction vpx_filter_block1d8_h8_intrin_ssse3;
filter8_1dfunction vpx_filter_block1d4_h8_intrin_ssse3;
#define vpx_filter_block1d8_v8_avx2 vpx_filter_block1d8_v8_intrin_ssse3
#define vpx_filter_block1d8_h8_avx2 vpx_filter_block1d8_h8_intrin_ssse3
#define vpx_filter_block1d4_h8_avx2 vpx_filter_block1d4_h8_intrin_ssse3
#else // VPX_ARCH_X86
filter8_1dfunction vpx_filter_block1d8_v8_ssse3;
filter8_1dfunction vpx_filter_block1d8_h8_ssse3;
filter8_1dfunction vpx_filter_block1d4_h8_ssse3;
#define vpx_filter_block1d8_v8_avx2 vpx_filter_block1d8_v8_ssse3
#define vpx_filter_block1d8_h8_avx2 vpx_filter_block1d8_h8_ssse3
#define vpx_filter_block1d4_h8_avx2 vpx_filter_block1d4_h8_ssse3
#endif // VPX_ARCH_X86_64
filter8_1dfunction vpx_filter_block1d8_v8_avg_ssse3;
filter8_1dfunction vpx_filter_block1d8_h8_avg_ssse3;
filter8_1dfunction vpx_filter_block1d4_v8_avg_ssse3;
filter8_1dfunction vpx_filter_block1d4_h8_avg_ssse3;
#define vpx_filter_block1d8_v8_avg_avx2 vpx_filter_block1d8_v8_avg_ssse3
#define vpx_filter_block1d8_h8_avg_avx2 vpx_filter_block1d8_h8_avg_ssse3
#define vpx_filter_block1d4_v8_avg_avx2 vpx_filter_block1d4_v8_avg_ssse3
#define vpx_filter_block1d4_h8_avg_avx2 vpx_filter_block1d4_h8_avg_ssse3
filter8_1dfunction vpx_filter_block1d16_v2_ssse3;
filter8_1dfunction vpx_filter_block1d16_h2_ssse3;
filter8_1dfunction vpx_filter_block1d8_v2_ssse3;
filter8_1dfunction vpx_filter_block1d8_h2_ssse3;
filter8_1dfunction vpx_filter_block1d4_v2_ssse3;
filter8_1dfunction vpx_filter_block1d4_h2_ssse3;
#define vpx_filter_block1d4_v8_avx2 vpx_filter_block1d4_v8_ssse3
#define vpx_filter_block1d16_v2_avx2 vpx_filter_block1d16_v2_ssse3
#define vpx_filter_block1d16_h2_avx2 vpx_filter_block1d16_h2_ssse3
#define vpx_filter_block1d8_v2_avx2 vpx_filter_block1d8_v2_ssse3
#define vpx_filter_block1d8_h2_avx2 vpx_filter_block1d8_h2_ssse3
#define vpx_filter_block1d4_v2_avx2 vpx_filter_block1d4_v2_ssse3
#define vpx_filter_block1d4_h2_avx2 vpx_filter_block1d4_h2_ssse3
filter8_1dfunction vpx_filter_block1d16_v2_avg_ssse3;
filter8_1dfunction vpx_filter_block1d16_h2_avg_ssse3;
filter8_1dfunction vpx_filter_block1d8_v2_avg_ssse3;
filter8_1dfunction vpx_filter_block1d8_h2_avg_ssse3;
filter8_1dfunction vpx_filter_block1d4_v2_avg_ssse3;
filter8_1dfunction vpx_filter_block1d4_h2_avg_ssse3;
#define vpx_filter_block1d16_v2_avg_avx2 vpx_filter_block1d16_v2_avg_ssse3
#define vpx_filter_block1d16_h2_avg_avx2 vpx_filter_block1d16_h2_avg_ssse3
#define vpx_filter_block1d8_v2_avg_avx2 vpx_filter_block1d8_v2_avg_ssse3
#define vpx_filter_block1d8_h2_avg_avx2 vpx_filter_block1d8_h2_avg_ssse3
#define vpx_filter_block1d4_v2_avg_avx2 vpx_filter_block1d4_v2_avg_ssse3
#define vpx_filter_block1d4_h2_avg_avx2 vpx_filter_block1d4_h2_avg_ssse3
#define vpx_filter_block1d16_v4_avg_avx2 vpx_filter_block1d16_v8_avg_avx2
#define vpx_filter_block1d16_h4_avg_avx2 vpx_filter_block1d16_h8_avg_avx2
#define vpx_filter_block1d8_v4_avg_avx2 vpx_filter_block1d8_v8_avg_avx2
#define vpx_filter_block1d8_h4_avg_avx2 vpx_filter_block1d8_h8_avg_avx2
#define vpx_filter_block1d4_v4_avg_avx2 vpx_filter_block1d4_v8_avg_avx2
#define vpx_filter_block1d4_h4_avg_avx2 vpx_filter_block1d4_h8_avg_avx2
// void vpx_convolve8_horiz_avx2(const uint8_t *src, ptrdiff_t src_stride,
// uint8_t *dst, ptrdiff_t dst_stride,
// const InterpKernel *filter, int x0_q4,
// int32_t x_step_q4, int y0_q4, int y_step_q4,
// int w, int h);
// void vpx_convolve8_vert_avx2(const uint8_t *src, ptrdiff_t src_stride,
// uint8_t *dst, ptrdiff_t dst_stride,
// const InterpKernel *filter, int x0_q4,
// int32_t x_step_q4, int y0_q4, int y_step_q4,
// int w, int h);
// void vpx_convolve8_avg_horiz_avx2(const uint8_t *src, ptrdiff_t src_stride,
// uint8_t *dst, ptrdiff_t dst_stride,
// const InterpKernel *filter, int x0_q4,
// int32_t x_step_q4, int y0_q4,
// int y_step_q4, int w, int h);
// void vpx_convolve8_avg_vert_avx2(const uint8_t *src, ptrdiff_t src_stride,
// uint8_t *dst, ptrdiff_t dst_stride,
// const InterpKernel *filter, int x0_q4,
// int32_t x_step_q4, int y0_q4,
// int y_step_q4, int w, int h);
FUN_CONV_1D(horiz, x0_q4, x_step_q4, h, src, , avx2, 0);
FUN_CONV_1D(vert, y0_q4, y_step_q4, v, src - src_stride * (num_taps / 2 - 1), ,
avx2, 0);
FUN_CONV_1D(avg_horiz, x0_q4, x_step_q4, h, src, avg_, avx2, 1);
FUN_CONV_1D(avg_vert, y0_q4, y_step_q4, v,
src - src_stride * (num_taps / 2 - 1), avg_, avx2, 1);
// void vpx_convolve8_avx2(const uint8_t *src, ptrdiff_t src_stride,
// uint8_t *dst, ptrdiff_t dst_stride,
// const InterpKernel *filter, int x0_q4,
// int32_t x_step_q4, int y0_q4, int y_step_q4,
// int w, int h);
// void vpx_convolve8_avg_avx2(const uint8_t *src, ptrdiff_t src_stride,
// uint8_t *dst, ptrdiff_t dst_stride,
// const InterpKernel *filter, int x0_q4,
// int32_t x_step_q4, int y0_q4, int y_step_q4,
// int w, int h);
FUN_CONV_2D(, avx2, 0);
FUN_CONV_2D(avg_, avx2, 1);
#endif // HAVE_AX2 && HAVE_SSSE3