aom_dsp/x86/masked_sad_intrin_avx2.c - aom - Git at Google

 /*
  * Copyright (c) 2018, Alliance for Open Media. All rights reserved
  *
  * This source code is subject to the terms of the BSD 2 Clause License and
  * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
  * was not distributed with this source code in the LICENSE file, you can
  * obtain it at www.aomedia.org/license/software. If the Alliance for Open
  * Media Patent License 1.0 was not distributed with this source code in the
  * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
  */

 #include <tmmintrin.h>

 #include "config/aom_config.h"
 #include "config/aom_dsp_rtcd.h"

 #include "aom_dsp/blend.h"
 #include "aom/aom_integer.h"
 #include "aom_dsp/x86/synonyms.h"
 #include "aom_dsp/x86/masked_sad_intrin_ssse3.h"

 static INLINE unsigned int masked_sad32xh_avx2(
     const uint8_t *src_ptr, int src_stride, const uint8_t *a_ptr, int a_stride,
     const uint8_t *b_ptr, int b_stride, const uint8_t *m_ptr, int m_stride,
     int width, int height) {
   int x, y;
   __m256i res = _mm256_setzero_si256();
   const __m256i mask_max = _mm256_set1_epi8((1 << AOM_BLEND_A64_ROUND_BITS));
   const __m256i round_scale =
       _mm256_set1_epi16(1 << (15 - AOM_BLEND_A64_ROUND_BITS));
   for (y = 0; y < height; y++) {
     for (x = 0; x < width; x += 32) {
       const __m256i src = _mm256_lddqu_si256((const __m256i *)&src_ptr[x]);
       const __m256i a = _mm256_lddqu_si256((const __m256i *)&a_ptr[x]);
       const __m256i b = _mm256_lddqu_si256((const __m256i *)&b_ptr[x]);
       const __m256i m = _mm256_lddqu_si256((const __m256i *)&m_ptr[x]);
       const __m256i m_inv = _mm256_sub_epi8(mask_max, m);

       // Calculate 16 predicted pixels.
       // Note that the maximum value of any entry of 'pred_l' or 'pred_r'
       // is 64 * 255, so we have plenty of space to add rounding constants.
       const __m256i data_l = _mm256_unpacklo_epi8(a, b);
       const __m256i mask_l = _mm256_unpacklo_epi8(m, m_inv);
       __m256i pred_l = _mm256_maddubs_epi16(data_l, mask_l);
       pred_l = _mm256_mulhrs_epi16(pred_l, round_scale);

       const __m256i data_r = _mm256_unpackhi_epi8(a, b);
       const __m256i mask_r = _mm256_unpackhi_epi8(m, m_inv);
       __m256i pred_r = _mm256_maddubs_epi16(data_r, mask_r);
       pred_r = _mm256_mulhrs_epi16(pred_r, round_scale);

       const __m256i pred = _mm256_packus_epi16(pred_l, pred_r);
       res = _mm256_add_epi32(res, _mm256_sad_epu8(pred, src));
     }

     src_ptr += src_stride;
     a_ptr += a_stride;
     b_ptr += b_stride;
     m_ptr += m_stride;
   }
   // At this point, we have two 32-bit partial SADs in lanes 0 and 2 of 'res'.
   res = _mm256_shuffle_epi32(res, 0xd8);
   res = _mm256_permute4x64_epi64(res, 0xd8);
   res = _mm256_hadd_epi32(res, res);
   res = _mm256_hadd_epi32(res, res);
   int32_t sad = _mm256_extract_epi32(res, 0);
   return sad;
 }

 static INLINE __m256i xx_loadu2_m128i(const void *hi, const void *lo) {
   __m128i a0 = _mm_lddqu_si128((const __m128i *)(lo));
   __m128i a1 = _mm_lddqu_si128((const __m128i *)(hi));
   __m256i a = _mm256_castsi128_si256(a0);
   return _mm256_inserti128_si256(a, a1, 1);
 }

 static INLINE unsigned int masked_sad16xh_avx2(
     const uint8_t *src_ptr, int src_stride, const uint8_t *a_ptr, int a_stride,
     const uint8_t *b_ptr, int b_stride, const uint8_t *m_ptr, int m_stride,
     int height) {
   int y;
   __m256i res = _mm256_setzero_si256();
   const __m256i mask_max = _mm256_set1_epi8((1 << AOM_BLEND_A64_ROUND_BITS));
   const __m256i round_scale =
       _mm256_set1_epi16(1 << (15 - AOM_BLEND_A64_ROUND_BITS));
   for (y = 0; y < height; y += 2) {
     const __m256i src = xx_loadu2_m128i(src_ptr + src_stride, src_ptr);
     const __m256i a = xx_loadu2_m128i(a_ptr + a_stride, a_ptr);
     const __m256i b = xx_loadu2_m128i(b_ptr + b_stride, b_ptr);
     const __m256i m = xx_loadu2_m128i(m_ptr + m_stride, m_ptr);
     const __m256i m_inv = _mm256_sub_epi8(mask_max, m);

     // Calculate 16 predicted pixels.
     // Note that the maximum value of any entry of 'pred_l' or 'pred_r'
     // is 64 * 255, so we have plenty of space to add rounding constants.
     const __m256i data_l = _mm256_unpacklo_epi8(a, b);
     const __m256i mask_l = _mm256_unpacklo_epi8(m, m_inv);
     __m256i pred_l = _mm256_maddubs_epi16(data_l, mask_l);
     pred_l = _mm256_mulhrs_epi16(pred_l, round_scale);

     const __m256i data_r = _mm256_unpackhi_epi8(a, b);
     const __m256i mask_r = _mm256_unpackhi_epi8(m, m_inv);
     __m256i pred_r = _mm256_maddubs_epi16(data_r, mask_r);
     pred_r = _mm256_mulhrs_epi16(pred_r, round_scale);

     const __m256i pred = _mm256_packus_epi16(pred_l, pred_r);
     res = _mm256_add_epi32(res, _mm256_sad_epu8(pred, src));

     src_ptr += src_stride << 1;
     a_ptr += a_stride << 1;
     b_ptr += b_stride << 1;
     m_ptr += m_stride << 1;
   }
   // At this point, we have two 32-bit partial SADs in lanes 0 and 2 of 'res'.
   res = _mm256_shuffle_epi32(res, 0xd8);
   res = _mm256_permute4x64_epi64(res, 0xd8);
   res = _mm256_hadd_epi32(res, res);
   res = _mm256_hadd_epi32(res, res);
   int32_t sad = _mm256_extract_epi32(res, 0);
   return sad;
 }

 static INLINE unsigned int aom_masked_sad_avx2(
     const uint8_t *src, int src_stride, const uint8_t *ref, int ref_stride,
     const uint8_t *second_pred, const uint8_t *msk, int msk_stride,
     int invert_mask, int m, int n) {
   unsigned int sad;
   if (!invert_mask) {
     switch (m) {
       case 4:
         sad = aom_masked_sad4xh_ssse3(src, src_stride, ref, ref_stride,
                                       second_pred, m, msk, msk_stride, n);
         break;
       case 8:
         sad = aom_masked_sad8xh_ssse3(src, src_stride, ref, ref_stride,
                                       second_pred, m, msk, msk_stride, n);
         break;
       case 16:
         sad = masked_sad16xh_avx2(src, src_stride, ref, ref_stride, second_pred,
                                   m, msk, msk_stride, n);
         break;
       default:
         sad = masked_sad32xh_avx2(src, src_stride, ref, ref_stride, second_pred,
                                   m, msk, msk_stride, m, n);
         break;
     }
   } else {
     switch (m) {
       case 4:
         sad = aom_masked_sad4xh_ssse3(src, src_stride, second_pred, m, ref,
                                       ref_stride, msk, msk_stride, n);
         break;
       case 8:
         sad = aom_masked_sad8xh_ssse3(src, src_stride, second_pred, m, ref,
                                       ref_stride, msk, msk_stride, n);
         break;
       case 16:
         sad = masked_sad16xh_avx2(src, src_stride, second_pred, m, ref,
                                   ref_stride, msk, msk_stride, n);
         break;
       default:
         sad = masked_sad32xh_avx2(src, src_stride, second_pred, m, ref,
                                   ref_stride, msk, msk_stride, m, n);
         break;
     }
   }
   return sad;
 }

 #define MASKSADMXN_AVX2(m, n)                                                 \
   unsigned int aom_masked_sad##m##x##n##_avx2(                                \
       const uint8_t *src, int src_stride, const uint8_t *ref, int ref_stride, \
       const uint8_t *second_pred, const uint8_t *msk, int msk_stride,         \
       int invert_mask) {                                                      \
     return aom_masked_sad_avx2(src, src_stride, ref, ref_stride, second_pred, \
                                msk, msk_stride, invert_mask, m, n);           \
   }

 MASKSADMXN_AVX2(4, 4)
 MASKSADMXN_AVX2(4, 8)
 MASKSADMXN_AVX2(8, 4)
 MASKSADMXN_AVX2(8, 8)
 MASKSADMXN_AVX2(8, 16)
 MASKSADMXN_AVX2(16, 8)
 MASKSADMXN_AVX2(16, 16)
 MASKSADMXN_AVX2(16, 32)
 MASKSADMXN_AVX2(32, 16)
 MASKSADMXN_AVX2(32, 32)
 MASKSADMXN_AVX2(32, 64)
 MASKSADMXN_AVX2(64, 32)
 MASKSADMXN_AVX2(64, 64)
 MASKSADMXN_AVX2(64, 128)
 MASKSADMXN_AVX2(128, 64)
 MASKSADMXN_AVX2(128, 128)
 MASKSADMXN_AVX2(4, 16)
 MASKSADMXN_AVX2(16, 4)
 MASKSADMXN_AVX2(8, 32)
 MASKSADMXN_AVX2(32, 8)
 MASKSADMXN_AVX2(16, 64)
 MASKSADMXN_AVX2(64, 16)

 static INLINE unsigned int highbd_masked_sad8xh_avx2(
     const uint8_t *src8, int src_stride, const uint8_t *a8, int a_stride,
     const uint8_t *b8, int b_stride, const uint8_t *m_ptr, int m_stride,
     int height) {
   const uint16_t *src_ptr = CONVERT_TO_SHORTPTR(src8);
   const uint16_t *a_ptr = CONVERT_TO_SHORTPTR(a8);
   const uint16_t *b_ptr = CONVERT_TO_SHORTPTR(b8);
   int y;
   __m256i res = _mm256_setzero_si256();
   const __m256i mask_max = _mm256_set1_epi16((1 << AOM_BLEND_A64_ROUND_BITS));
   const __m256i round_const =
       _mm256_set1_epi32((1 << AOM_BLEND_A64_ROUND_BITS) >> 1);
   const __m256i one = _mm256_set1_epi16(1);

   for (y = 0; y < height; y += 2) {
     const __m256i src = xx_loadu2_m128i(src_ptr + src_stride, src_ptr);
     const __m256i a = xx_loadu2_m128i(a_ptr + a_stride, a_ptr);
     const __m256i b = xx_loadu2_m128i(b_ptr + b_stride, b_ptr);
     // Zero-extend mask to 16 bits
     const __m256i m = _mm256_cvtepu8_epi16(_mm_unpacklo_epi64(
         _mm_loadl_epi64((const __m128i *)(m_ptr)),
         _mm_loadl_epi64((const __m128i *)(m_ptr + m_stride))));
     const __m256i m_inv = _mm256_sub_epi16(mask_max, m);

     const __m256i data_l = _mm256_unpacklo_epi16(a, b);
     const __m256i mask_l = _mm256_unpacklo_epi16(m, m_inv);
     __m256i pred_l = _mm256_madd_epi16(data_l, mask_l);
     pred_l = _mm256_srai_epi32(_mm256_add_epi32(pred_l, round_const),
                                AOM_BLEND_A64_ROUND_BITS);

     const __m256i data_r = _mm256_unpackhi_epi16(a, b);
     const __m256i mask_r = _mm256_unpackhi_epi16(m, m_inv);
     __m256i pred_r = _mm256_madd_epi16(data_r, mask_r);
     pred_r = _mm256_srai_epi32(_mm256_add_epi32(pred_r, round_const),
                                AOM_BLEND_A64_ROUND_BITS);

     // Note: the maximum value in pred_l/r is (2^bd)-1 < 2^15,
     // so it is safe to do signed saturation here.
     const __m256i pred = _mm256_packs_epi32(pred_l, pred_r);
     // There is no 16-bit SAD instruction, so we have to synthesize
     // an 8-element SAD. We do this by storing 4 32-bit partial SADs,
     // and accumulating them at the end
     const __m256i diff = _mm256_abs_epi16(_mm256_sub_epi16(pred, src));
     res = _mm256_add_epi32(res, _mm256_madd_epi16(diff, one));

     src_ptr += src_stride << 1;
     a_ptr += a_stride << 1;
     b_ptr += b_stride << 1;
     m_ptr += m_stride << 1;
   }
   // At this point, we have four 32-bit partial SADs stored in 'res'.
   res = _mm256_hadd_epi32(res, res);
   res = _mm256_hadd_epi32(res, res);
   int sad = _mm256_extract_epi32(res, 0) + _mm256_extract_epi32(res, 4);
   return sad;
 }

 static INLINE unsigned int highbd_masked_sad16xh_avx2(
     const uint8_t *src8, int src_stride, const uint8_t *a8, int a_stride,
     const uint8_t *b8, int b_stride, const uint8_t *m_ptr, int m_stride,
     int width, int height) {
   const uint16_t *src_ptr = CONVERT_TO_SHORTPTR(src8);
   const uint16_t *a_ptr = CONVERT_TO_SHORTPTR(a8);
   const uint16_t *b_ptr = CONVERT_TO_SHORTPTR(b8);
   int x, y;
   __m256i res = _mm256_setzero_si256();
   const __m256i mask_max = _mm256_set1_epi16((1 << AOM_BLEND_A64_ROUND_BITS));
   const __m256i round_const =
       _mm256_set1_epi32((1 << AOM_BLEND_A64_ROUND_BITS) >> 1);
   const __m256i one = _mm256_set1_epi16(1);

   for (y = 0; y < height; y++) {
     for (x = 0; x < width; x += 16) {
       const __m256i src = _mm256_lddqu_si256((const __m256i *)&src_ptr[x]);
       const __m256i a = _mm256_lddqu_si256((const __m256i *)&a_ptr[x]);
       const __m256i b = _mm256_lddqu_si256((const __m256i *)&b_ptr[x]);
       // Zero-extend mask to 16 bits
       const __m256i m =
           _mm256_cvtepu8_epi16(_mm_lddqu_si128((const __m128i *)&m_ptr[x]));
       const __m256i m_inv = _mm256_sub_epi16(mask_max, m);

       const __m256i data_l = _mm256_unpacklo_epi16(a, b);
       const __m256i mask_l = _mm256_unpacklo_epi16(m, m_inv);
       __m256i pred_l = _mm256_madd_epi16(data_l, mask_l);
       pred_l = _mm256_srai_epi32(_mm256_add_epi32(pred_l, round_const),
                                  AOM_BLEND_A64_ROUND_BITS);

       const __m256i data_r = _mm256_unpackhi_epi16(a, b);
       const __m256i mask_r = _mm256_unpackhi_epi16(m, m_inv);
       __m256i pred_r = _mm256_madd_epi16(data_r, mask_r);
       pred_r = _mm256_srai_epi32(_mm256_add_epi32(pred_r, round_const),
                                  AOM_BLEND_A64_ROUND_BITS);

       // Note: the maximum value in pred_l/r is (2^bd)-1 < 2^15,
       // so it is safe to do signed saturation here.
       const __m256i pred = _mm256_packs_epi32(pred_l, pred_r);
       // There is no 16-bit SAD instruction, so we have to synthesize
       // an 8-element SAD. We do this by storing 4 32-bit partial SADs,
       // and accumulating them at the end
       const __m256i diff = _mm256_abs_epi16(_mm256_sub_epi16(pred, src));
       res = _mm256_add_epi32(res, _mm256_madd_epi16(diff, one));
     }

     src_ptr += src_stride;
     a_ptr += a_stride;
     b_ptr += b_stride;
     m_ptr += m_stride;
   }
   // At this point, we have four 32-bit partial SADs stored in 'res'.
   res = _mm256_hadd_epi32(res, res);
   res = _mm256_hadd_epi32(res, res);
   int sad = _mm256_extract_epi32(res, 0) + _mm256_extract_epi32(res, 4);
   return sad;
 }

 static INLINE unsigned int aom_highbd_masked_sad_avx2(
     const uint8_t *src, int src_stride, const uint8_t *ref, int ref_stride,
     const uint8_t *second_pred, const uint8_t *msk, int msk_stride,
     int invert_mask, int m, int n) {
   unsigned int sad;
   if (!invert_mask) {
     switch (m) {
       case 4:
         sad =
             aom_highbd_masked_sad4xh_ssse3(src, src_stride, ref, ref_stride,
                                            second_pred, m, msk, msk_stride, n);
         break;
       case 8:
         sad = highbd_masked_sad8xh_avx2(src, src_stride, ref, ref_stride,
                                         second_pred, m, msk, msk_stride, n);
         break;
       default:
         sad = highbd_masked_sad16xh_avx2(src, src_stride, ref, ref_stride,
                                          second_pred, m, msk, msk_stride, m, n);
         break;
     }
   } else {
     switch (m) {
       case 4:
         sad =
             aom_highbd_masked_sad4xh_ssse3(src, src_stride, second_pred, m, ref,
                                            ref_stride, msk, msk_stride, n);
         break;
       case 8:
         sad = highbd_masked_sad8xh_avx2(src, src_stride, second_pred, m, ref,
                                         ref_stride, msk, msk_stride, n);
         break;
       default:
         sad = highbd_masked_sad16xh_avx2(src, src_stride, second_pred, m, ref,
                                          ref_stride, msk, msk_stride, m, n);
         break;
     }
   }
   return sad;
 }

 #define HIGHBD_MASKSADMXN_AVX2(m, n)                                      \
   unsigned int aom_highbd_masked_sad##m##x##n##_avx2(                     \
       const uint8_t *src8, int src_stride, const uint8_t *ref8,           \
       int ref_stride, const uint8_t *second_pred8, const uint8_t *msk,    \
       int msk_stride, int invert_mask) {                                  \
     return aom_highbd_masked_sad_avx2(src8, src_stride, ref8, ref_stride, \
                                       second_pred8, msk, msk_stride,      \
                                       invert_mask, m, n);                 \
   }

 HIGHBD_MASKSADMXN_AVX2(4, 4);
 HIGHBD_MASKSADMXN_AVX2(4, 8);
 HIGHBD_MASKSADMXN_AVX2(8, 4);
 HIGHBD_MASKSADMXN_AVX2(8, 8);
 HIGHBD_MASKSADMXN_AVX2(8, 16);
 HIGHBD_MASKSADMXN_AVX2(16, 8);
 HIGHBD_MASKSADMXN_AVX2(16, 16);
 HIGHBD_MASKSADMXN_AVX2(16, 32);
 HIGHBD_MASKSADMXN_AVX2(32, 16);
 HIGHBD_MASKSADMXN_AVX2(32, 32);
 HIGHBD_MASKSADMXN_AVX2(32, 64);
 HIGHBD_MASKSADMXN_AVX2(64, 32);
 HIGHBD_MASKSADMXN_AVX2(64, 64);
 HIGHBD_MASKSADMXN_AVX2(64, 128);
 HIGHBD_MASKSADMXN_AVX2(128, 64);
 HIGHBD_MASKSADMXN_AVX2(128, 128);
 HIGHBD_MASKSADMXN_AVX2(4, 16);
 HIGHBD_MASKSADMXN_AVX2(16, 4);
 HIGHBD_MASKSADMXN_AVX2(8, 32);
 HIGHBD_MASKSADMXN_AVX2(32, 8);
 HIGHBD_MASKSADMXN_AVX2(16, 64);
 HIGHBD_MASKSADMXN_AVX2(64, 16);
	/*
	* Copyright (c) 2018, Alliance for Open Media. All rights reserved
	*
	* This source code is subject to the terms of the BSD 2 Clause License and
	* the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
	* was not distributed with this source code in the LICENSE file, you can
	* obtain it at www.aomedia.org/license/software. If the Alliance for Open
	* Media Patent License 1.0 was not distributed with this source code in the
	* PATENTS file, you can obtain it at www.aomedia.org/license/patent.
	*/

	#include <tmmintrin.h>

	#include "config/aom_config.h"
	#include "config/aom_dsp_rtcd.h"

	#include "aom_dsp/blend.h"
	#include "aom/aom_integer.h"
	#include "aom_dsp/x86/synonyms.h"
	#include "aom_dsp/x86/masked_sad_intrin_ssse3.h"

	static INLINE unsigned int masked_sad32xh_avx2(
	const uint8_t src_ptr, int src_stride, const uint8_t a_ptr, int a_stride,
	const uint8_t b_ptr, int b_stride, const uint8_t m_ptr, int m_stride,
	int width, int height) {
	int x, y;
	__m256i res = _mm256_setzero_si256();
	const __m256i mask_max = _mm256_set1_epi8((1 << AOM_BLEND_A64_ROUND_BITS));
	const __m256i round_scale =
	_mm256_set1_epi16(1 << (15 - AOM_BLEND_A64_ROUND_BITS));
	for (y = 0; y < height; y++) {
	for (x = 0; x < width; x += 32) {
	const __m256i src = _mm256_lddqu_si256((const __m256i *)&src_ptr[x]);
	const __m256i a = _mm256_lddqu_si256((const __m256i *)&a_ptr[x]);
	const __m256i b = _mm256_lddqu_si256((const __m256i *)&b_ptr[x]);
	const __m256i m = _mm256_lddqu_si256((const __m256i *)&m_ptr[x]);
	const __m256i m_inv = _mm256_sub_epi8(mask_max, m);

	// Calculate 16 predicted pixels.
	// Note that the maximum value of any entry of 'pred_l' or 'pred_r'
	// is 64 * 255, so we have plenty of space to add rounding constants.
	const __m256i data_l = _mm256_unpacklo_epi8(a, b);
	const __m256i mask_l = _mm256_unpacklo_epi8(m, m_inv);
	__m256i pred_l = _mm256_maddubs_epi16(data_l, mask_l);
	pred_l = _mm256_mulhrs_epi16(pred_l, round_scale);

	const __m256i data_r = _mm256_unpackhi_epi8(a, b);
	const __m256i mask_r = _mm256_unpackhi_epi8(m, m_inv);
	__m256i pred_r = _mm256_maddubs_epi16(data_r, mask_r);
	pred_r = _mm256_mulhrs_epi16(pred_r, round_scale);

	const __m256i pred = _mm256_packus_epi16(pred_l, pred_r);
	res = _mm256_add_epi32(res, _mm256_sad_epu8(pred, src));
	}

	src_ptr += src_stride;
	a_ptr += a_stride;
	b_ptr += b_stride;
	m_ptr += m_stride;
	}
	// At this point, we have two 32-bit partial SADs in lanes 0 and 2 of 'res'.
	res = _mm256_shuffle_epi32(res, 0xd8);
	res = _mm256_permute4x64_epi64(res, 0xd8);
	res = _mm256_hadd_epi32(res, res);
	res = _mm256_hadd_epi32(res, res);
	int32_t sad = _mm256_extract_epi32(res, 0);
	return sad;
	}

	static INLINE __m256i xx_loadu2_m128i(const void hi, const void lo) {
	__m128i a0 = _mm_lddqu_si128((const __m128i *)(lo));
	__m128i a1 = _mm_lddqu_si128((const __m128i *)(hi));
	__m256i a = _mm256_castsi128_si256(a0);
	return _mm256_inserti128_si256(a, a1, 1);
	}

	static INLINE unsigned int masked_sad16xh_avx2(
	const uint8_t src_ptr, int src_stride, const uint8_t a_ptr, int a_stride,
	const uint8_t b_ptr, int b_stride, const uint8_t m_ptr, int m_stride,
	int height) {
	int y;
	__m256i res = _mm256_setzero_si256();
	const __m256i mask_max = _mm256_set1_epi8((1 << AOM_BLEND_A64_ROUND_BITS));
	const __m256i round_scale =
	_mm256_set1_epi16(1 << (15 - AOM_BLEND_A64_ROUND_BITS));
	for (y = 0; y < height; y += 2) {
	const __m256i src = xx_loadu2_m128i(src_ptr + src_stride, src_ptr);
	const __m256i a = xx_loadu2_m128i(a_ptr + a_stride, a_ptr);
	const __m256i b = xx_loadu2_m128i(b_ptr + b_stride, b_ptr);
	const __m256i m = xx_loadu2_m128i(m_ptr + m_stride, m_ptr);
	const __m256i m_inv = _mm256_sub_epi8(mask_max, m);

	// Calculate 16 predicted pixels.
	// Note that the maximum value of any entry of 'pred_l' or 'pred_r'
	// is 64 * 255, so we have plenty of space to add rounding constants.
	const __m256i data_l = _mm256_unpacklo_epi8(a, b);
	const __m256i mask_l = _mm256_unpacklo_epi8(m, m_inv);
	__m256i pred_l = _mm256_maddubs_epi16(data_l, mask_l);
	pred_l = _mm256_mulhrs_epi16(pred_l, round_scale);

	const __m256i data_r = _mm256_unpackhi_epi8(a, b);
	const __m256i mask_r = _mm256_unpackhi_epi8(m, m_inv);
	__m256i pred_r = _mm256_maddubs_epi16(data_r, mask_r);
	pred_r = _mm256_mulhrs_epi16(pred_r, round_scale);

	const __m256i pred = _mm256_packus_epi16(pred_l, pred_r);
	res = _mm256_add_epi32(res, _mm256_sad_epu8(pred, src));

	src_ptr += src_stride << 1;
	a_ptr += a_stride << 1;
	b_ptr += b_stride << 1;
	m_ptr += m_stride << 1;
	}
	// At this point, we have two 32-bit partial SADs in lanes 0 and 2 of 'res'.
	res = _mm256_shuffle_epi32(res, 0xd8);
	res = _mm256_permute4x64_epi64(res, 0xd8);
	res = _mm256_hadd_epi32(res, res);
	res = _mm256_hadd_epi32(res, res);
	int32_t sad = _mm256_extract_epi32(res, 0);
	return sad;
	}

	static INLINE unsigned int aom_masked_sad_avx2(
	const uint8_t src, int src_stride, const uint8_t ref, int ref_stride,
	const uint8_t second_pred, const uint8_t msk, int msk_stride,
	int invert_mask, int m, int n) {
	unsigned int sad;
	if (!invert_mask) {
	switch (m) {
	case 4:
	sad = aom_masked_sad4xh_ssse3(src, src_stride, ref, ref_stride,
	second_pred, m, msk, msk_stride, n);
	break;
	case 8:
	sad = aom_masked_sad8xh_ssse3(src, src_stride, ref, ref_stride,
	second_pred, m, msk, msk_stride, n);
	break;
	case 16:
	sad = masked_sad16xh_avx2(src, src_stride, ref, ref_stride, second_pred,
	m, msk, msk_stride, n);
	break;
	default:
	sad = masked_sad32xh_avx2(src, src_stride, ref, ref_stride, second_pred,
	m, msk, msk_stride, m, n);
	break;
	}
	} else {
	switch (m) {
	case 4:
	sad = aom_masked_sad4xh_ssse3(src, src_stride, second_pred, m, ref,
	ref_stride, msk, msk_stride, n);
	break;
	case 8:
	sad = aom_masked_sad8xh_ssse3(src, src_stride, second_pred, m, ref,
	ref_stride, msk, msk_stride, n);
	break;
	case 16:
	sad = masked_sad16xh_avx2(src, src_stride, second_pred, m, ref,
	ref_stride, msk, msk_stride, n);
	break;
	default:
	sad = masked_sad32xh_avx2(src, src_stride, second_pred, m, ref,
	ref_stride, msk, msk_stride, m, n);
	break;
	}
	}
	return sad;
	}

	#define MASKSADMXN_AVX2(m, n) \
	unsigned int aom_masked_sad##m##x##n##_avx2( \
	const uint8_t src, int src_stride, const uint8_t ref, int ref_stride, \
	const uint8_t second_pred, const uint8_t msk, int msk_stride, \
	int invert_mask) { \
	return aom_masked_sad_avx2(src, src_stride, ref, ref_stride, second_pred, \
	msk, msk_stride, invert_mask, m, n); \
	}

	MASKSADMXN_AVX2(4, 4)
	MASKSADMXN_AVX2(4, 8)
	MASKSADMXN_AVX2(8, 4)
	MASKSADMXN_AVX2(8, 8)
	MASKSADMXN_AVX2(8, 16)
	MASKSADMXN_AVX2(16, 8)
	MASKSADMXN_AVX2(16, 16)
	MASKSADMXN_AVX2(16, 32)
	MASKSADMXN_AVX2(32, 16)
	MASKSADMXN_AVX2(32, 32)
	MASKSADMXN_AVX2(32, 64)
	MASKSADMXN_AVX2(64, 32)
	MASKSADMXN_AVX2(64, 64)
	MASKSADMXN_AVX2(64, 128)
	MASKSADMXN_AVX2(128, 64)
	MASKSADMXN_AVX2(128, 128)
	MASKSADMXN_AVX2(4, 16)
	MASKSADMXN_AVX2(16, 4)
	MASKSADMXN_AVX2(8, 32)
	MASKSADMXN_AVX2(32, 8)
	MASKSADMXN_AVX2(16, 64)
	MASKSADMXN_AVX2(64, 16)

	static INLINE unsigned int highbd_masked_sad8xh_avx2(
	const uint8_t src8, int src_stride, const uint8_t a8, int a_stride,
	const uint8_t b8, int b_stride, const uint8_t m_ptr, int m_stride,
	int height) {
	const uint16_t *src_ptr = CONVERT_TO_SHORTPTR(src8);
	const uint16_t *a_ptr = CONVERT_TO_SHORTPTR(a8);
	const uint16_t *b_ptr = CONVERT_TO_SHORTPTR(b8);
	int y;
	__m256i res = _mm256_setzero_si256();
	const __m256i mask_max = _mm256_set1_epi16((1 << AOM_BLEND_A64_ROUND_BITS));
	const __m256i round_const =
	_mm256_set1_epi32((1 << AOM_BLEND_A64_ROUND_BITS) >> 1);
	const __m256i one = _mm256_set1_epi16(1);

	for (y = 0; y < height; y += 2) {
	const __m256i src = xx_loadu2_m128i(src_ptr + src_stride, src_ptr);
	const __m256i a = xx_loadu2_m128i(a_ptr + a_stride, a_ptr);
	const __m256i b = xx_loadu2_m128i(b_ptr + b_stride, b_ptr);
	// Zero-extend mask to 16 bits
	const __m256i m = _mm256_cvtepu8_epi16(_mm_unpacklo_epi64(
	_mm_loadl_epi64((const __m128i *)(m_ptr)),
	_mm_loadl_epi64((const __m128i *)(m_ptr + m_stride))));
	const __m256i m_inv = _mm256_sub_epi16(mask_max, m);

	const __m256i data_l = _mm256_unpacklo_epi16(a, b);
	const __m256i mask_l = _mm256_unpacklo_epi16(m, m_inv);
	__m256i pred_l = _mm256_madd_epi16(data_l, mask_l);
	pred_l = _mm256_srai_epi32(_mm256_add_epi32(pred_l, round_const),
	AOM_BLEND_A64_ROUND_BITS);

	const __m256i data_r = _mm256_unpackhi_epi16(a, b);
	const __m256i mask_r = _mm256_unpackhi_epi16(m, m_inv);
	__m256i pred_r = _mm256_madd_epi16(data_r, mask_r);
	pred_r = _mm256_srai_epi32(_mm256_add_epi32(pred_r, round_const),
	AOM_BLEND_A64_ROUND_BITS);

	// Note: the maximum value in pred_l/r is (2^bd)-1 < 2^15,
	// so it is safe to do signed saturation here.
	const __m256i pred = _mm256_packs_epi32(pred_l, pred_r);
	// There is no 16-bit SAD instruction, so we have to synthesize
	// an 8-element SAD. We do this by storing 4 32-bit partial SADs,
	// and accumulating them at the end
	const __m256i diff = _mm256_abs_epi16(_mm256_sub_epi16(pred, src));
	res = _mm256_add_epi32(res, _mm256_madd_epi16(diff, one));

	src_ptr += src_stride << 1;
	a_ptr += a_stride << 1;
	b_ptr += b_stride << 1;
	m_ptr += m_stride << 1;
	}
	// At this point, we have four 32-bit partial SADs stored in 'res'.
	res = _mm256_hadd_epi32(res, res);
	res = _mm256_hadd_epi32(res, res);
	int sad = _mm256_extract_epi32(res, 0) + _mm256_extract_epi32(res, 4);
	return sad;
	}

	static INLINE unsigned int highbd_masked_sad16xh_avx2(
	const uint8_t src8, int src_stride, const uint8_t a8, int a_stride,
	const uint8_t b8, int b_stride, const uint8_t m_ptr, int m_stride,
	int width, int height) {
	const uint16_t *src_ptr = CONVERT_TO_SHORTPTR(src8);
	const uint16_t *a_ptr = CONVERT_TO_SHORTPTR(a8);
	const uint16_t *b_ptr = CONVERT_TO_SHORTPTR(b8);
	int x, y;
	__m256i res = _mm256_setzero_si256();
	const __m256i mask_max = _mm256_set1_epi16((1 << AOM_BLEND_A64_ROUND_BITS));
	const __m256i round_const =
	_mm256_set1_epi32((1 << AOM_BLEND_A64_ROUND_BITS) >> 1);
	const __m256i one = _mm256_set1_epi16(1);

	for (y = 0; y < height; y++) {
	for (x = 0; x < width; x += 16) {
	const __m256i src = _mm256_lddqu_si256((const __m256i *)&src_ptr[x]);
	const __m256i a = _mm256_lddqu_si256((const __m256i *)&a_ptr[x]);
	const __m256i b = _mm256_lddqu_si256((const __m256i *)&b_ptr[x]);
	// Zero-extend mask to 16 bits
	const __m256i m =
	_mm256_cvtepu8_epi16(_mm_lddqu_si128((const __m128i *)&m_ptr[x]));
	const __m256i m_inv = _mm256_sub_epi16(mask_max, m);

	const __m256i data_l = _mm256_unpacklo_epi16(a, b);
	const __m256i mask_l = _mm256_unpacklo_epi16(m, m_inv);
	__m256i pred_l = _mm256_madd_epi16(data_l, mask_l);
	pred_l = _mm256_srai_epi32(_mm256_add_epi32(pred_l, round_const),
	AOM_BLEND_A64_ROUND_BITS);

	const __m256i data_r = _mm256_unpackhi_epi16(a, b);
	const __m256i mask_r = _mm256_unpackhi_epi16(m, m_inv);
	__m256i pred_r = _mm256_madd_epi16(data_r, mask_r);
	pred_r = _mm256_srai_epi32(_mm256_add_epi32(pred_r, round_const),
	AOM_BLEND_A64_ROUND_BITS);

	// Note: the maximum value in pred_l/r is (2^bd)-1 < 2^15,
	// so it is safe to do signed saturation here.
	const __m256i pred = _mm256_packs_epi32(pred_l, pred_r);
	// There is no 16-bit SAD instruction, so we have to synthesize
	// an 8-element SAD. We do this by storing 4 32-bit partial SADs,
	// and accumulating them at the end
	const __m256i diff = _mm256_abs_epi16(_mm256_sub_epi16(pred, src));
	res = _mm256_add_epi32(res, _mm256_madd_epi16(diff, one));
	}

	src_ptr += src_stride;
	a_ptr += a_stride;
	b_ptr += b_stride;
	m_ptr += m_stride;
	}
	// At this point, we have four 32-bit partial SADs stored in 'res'.
	res = _mm256_hadd_epi32(res, res);
	res = _mm256_hadd_epi32(res, res);
	int sad = _mm256_extract_epi32(res, 0) + _mm256_extract_epi32(res, 4);
	return sad;
	}

	static INLINE unsigned int aom_highbd_masked_sad_avx2(
	const uint8_t src, int src_stride, const uint8_t ref, int ref_stride,
	const uint8_t second_pred, const uint8_t msk, int msk_stride,
	int invert_mask, int m, int n) {
	unsigned int sad;
	if (!invert_mask) {
	switch (m) {
	case 4:
	sad =
	aom_highbd_masked_sad4xh_ssse3(src, src_stride, ref, ref_stride,
	second_pred, m, msk, msk_stride, n);
	break;
	case 8:
	sad = highbd_masked_sad8xh_avx2(src, src_stride, ref, ref_stride,
	second_pred, m, msk, msk_stride, n);
	break;
	default:
	sad = highbd_masked_sad16xh_avx2(src, src_stride, ref, ref_stride,
	second_pred, m, msk, msk_stride, m, n);
	break;
	}
	} else {
	switch (m) {
	case 4:
	sad =
	aom_highbd_masked_sad4xh_ssse3(src, src_stride, second_pred, m, ref,
	ref_stride, msk, msk_stride, n);
	break;
	case 8:
	sad = highbd_masked_sad8xh_avx2(src, src_stride, second_pred, m, ref,
	ref_stride, msk, msk_stride, n);
	break;
	default:
	sad = highbd_masked_sad16xh_avx2(src, src_stride, second_pred, m, ref,
	ref_stride, msk, msk_stride, m, n);
	break;
	}
	}
	return sad;
	}

	#define HIGHBD_MASKSADMXN_AVX2(m, n) \
	unsigned int aom_highbd_masked_sad##m##x##n##_avx2( \
	const uint8_t src8, int src_stride, const uint8_t ref8, \
	int ref_stride, const uint8_t second_pred8, const uint8_t msk, \
	int msk_stride, int invert_mask) { \
	return aom_highbd_masked_sad_avx2(src8, src_stride, ref8, ref_stride, \
	second_pred8, msk, msk_stride, \
	invert_mask, m, n); \
	}

	HIGHBD_MASKSADMXN_AVX2(4, 4);
	HIGHBD_MASKSADMXN_AVX2(4, 8);
	HIGHBD_MASKSADMXN_AVX2(8, 4);
	HIGHBD_MASKSADMXN_AVX2(8, 8);
	HIGHBD_MASKSADMXN_AVX2(8, 16);
	HIGHBD_MASKSADMXN_AVX2(16, 8);
	HIGHBD_MASKSADMXN_AVX2(16, 16);
	HIGHBD_MASKSADMXN_AVX2(16, 32);
	HIGHBD_MASKSADMXN_AVX2(32, 16);
	HIGHBD_MASKSADMXN_AVX2(32, 32);
	HIGHBD_MASKSADMXN_AVX2(32, 64);
	HIGHBD_MASKSADMXN_AVX2(64, 32);
	HIGHBD_MASKSADMXN_AVX2(64, 64);
	HIGHBD_MASKSADMXN_AVX2(64, 128);
	HIGHBD_MASKSADMXN_AVX2(128, 64);
	HIGHBD_MASKSADMXN_AVX2(128, 128);
	HIGHBD_MASKSADMXN_AVX2(4, 16);
	HIGHBD_MASKSADMXN_AVX2(16, 4);
	HIGHBD_MASKSADMXN_AVX2(8, 32);
	HIGHBD_MASKSADMXN_AVX2(32, 8);
	HIGHBD_MASKSADMXN_AVX2(16, 64);
	HIGHBD_MASKSADMXN_AVX2(64, 16);