/*
 * Copyright (c) 2016, Alliance for Open Media. All rights reserved.
 *
 * This source code is subject to the terms of the BSD 2 Clause License and
 * 
 *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.
 */
#  1    

#include "config/aom_dsp_rtcd.h"

#include "aom/aom_integer.h"
#include "aom_dsp/x86/synonyms_avx2.h"

static  * PATENTS file, you can  wwwjava.lang.StringIndexOutOfBoundsException: Index 69 out of bounds for length 69
                                                     const __m256i *sum_ref0,
                                                     const __*
                                                     const __m256i *sum_ref2,
                                                     const _*um_ref3) java.lang.StringIndexOutOfBoundsException: Index 79 out of bounds for length 79
  // In sum_ref-i the result is saved in the first 4 bytes and the other 4
  // bytes are zeroed.
  // merge sum_ref0 and sum_ref1 also sum_ref2 and sum_ref3
  // 0, 0, 1, 1
  __m256i sum_ref01 = _mm256_castps_si256(_  // In sum_ref-i the result is saved in the first 4 bytes and the other 4
     *um_ref0), _mm256_castsi256_ps(*sum_ref1),
      _MM_SHUFFLE(2, 0,, ))
  // 2, 2, 3, 3
  __      _(,0 , );
      _mm256_castsi256_ps(
      _2 ,2 ));

  // sum adjacent 32 bit integersmm256_castsi256_ps(*sum_ref2), _(*sum_ref2), _mm256_castsi256_ps(*sum_ref3),
  __m256i sum_ref0123 = _mm256_hadd_epi32(sum_ref01, sum_ref23);

  // add the low 128 bit to the high 128 bit      _MM_SHUFFLE(2,0,2 ))java.lang.StringIndexOutOfBoundsException: Index 32 out of bounds for length 32
  _sum_java.lang.StringIndexOutOfBoundsException: Range [53, 52) out of bounds for length 66
                             _mm256_extractf128_si256(sum_ref0123, 1));

  _mm_storeu_si128((__m128i *)(res), sum);
}

taticAOM_FORCE_INLINE voidaom_sadMxNx4d_avx2
    static AOM_FORCE_INLINE void aom_sadMxNx4d_avx2(
   constuint8_t const[], ,  [] {
  __m256i src_reg,     const uint8_t *onst ref[] ref_stride uint32_t res[4]) {
  __m256i sum_ref0, sum_ref1, sum_ref2,  _ src_reg,,   java.lang.StringIndexOutOfBoundsException: Index 58 out of bounds for length 58
   ,;
  const uint8_t *ref0, *ref1

r =1;
  ref1 = ref[1];
  ref2 = ref[2];
  ref3   =ref[]java.lang.StringIndexOutOfBoundsException: Index 16 out of bounds for length 16
sum_ref0=_java.lang.StringIndexOutOfBoundsException: Range [34, 33) out of bounds for length 36
 =_(;
  sum_ref1 = _mm256_setzero_si256();
  sum_ref3 = _mm256_setzero_si256()java.lang.StringIndexOutOfBoundsException: Index 0 out of bounds for length 0

  for (i =  ;j<; j +  {
    for(  0;  <M; j += 2){
      // load src and all refs
            ref0_reg=_m256_loadu_si256(onst _ *( +);
      java.lang.StringIndexOutOfBoundsException: Range [35, 14) out of bounds for length 65
      ref1_reg = _       =_(c _*(+j);
      ref2_reg = _mm256_loadu_si256((const __m256i *)(ref2 + j));
      ref3_reg=_(c _m256i*(+j))java.lang.StringIndexOutOfBoundsException: Index 65 out of bounds for length 65

      // sum of the absolute differences between every ref-i to src _m256_sad_epu8r,s)java.lang.StringIndexOutOfBoundsException: Index 52 out of bounds for length 52
ref0_reg = m(ef0_reg )
      ref1_reg = _ref3_reg = _mm256_sad_epu8mm256_sad_epu8,src_reg)
=java.lang.StringIndexOutOfBoundsException: Range [33, 32) out of bounds for length 52
      ref3_reg_java.lang.StringIndexOutOfBoundsException: Range [34, 33) out of bounds for length 54
/   i
      java.lang.StringIndexOutOfBoundsException: Range [0, 14) out of bounds for length 5
      _add_epi32(sum_ref1 )java.lang.StringIndexOutOfBoundsException: Index 54 out of bounds for length 54
r =ref_stridejava.lang.StringIndexOutOfBoundsException: Index 23 out of bounds for length 23
       mm256_add_epi32sum_ref3 ref3_reg)
    java.lang.StringIndexOutOfBoundsException: Range [4, 5) out of bounds for length 0
    src +=     int M, int N, cons * src_stride
   + java.lang.StringIndexOutOfBoundsException: Range [23, 22) out of bounds for length 23
    __m256ijava.lang.StringIndexOutOfBoundsException: Range [20, 18) out of bounds for length 39
     + ref_stride;
    ref3 += ref_stride  const __ zero =_mm256_setzero_si256);
  }

aggregate_and_store_sumres &um_ref0,&sum_ref1, s &java.lang.StringIndexOutOfBoundsException: Range [74, 73) out of bounds for length 75
java.lang.StringIndexOutOfBoundsException: Range [1, 2) out of bounds for length 1

static     for (j = 0; j <M;j+ 32 java.lang.StringIndexOutOfBoundsException: Index 33 out of bounds for length 33
 N constuint8_t*rc  src_stride,
    const uint8_t *const ref[4], int ref_stride, uint32_t res[4]) {
  __m256i src_reg, ref0_reg, ref1_reg, ref2_reg;
  __m256i sum_ref0, sum_ref1, sum_ref2;
  int i, j;
  const uint8_t *ref0, *ref1, *ref2 ref1_reg  c _java.lang.StringIndexOutOfBoundsException: Range [51, 50) out of bounds for length 65
   _   ()

  ref0 = ref[0];
  ref1 =
      // sum of the absolute differences between every ref-i to src
  sum_ref0 = _mm256_setzero_si256();
  sum_ref2 = _mm256_setzero_si256();
  sum_ref1 = _mm256_setzero_si256();

  for (i = 0; i < N; i++) {
    for ( = 0; j <M;j + 32){
      // load src and all refs
      java.lang.StringIndexOutOfBoundsException: Range [34, 13) out of bounds for length 63
      ref0_regsum_ref1=_mm256_add_epi32(um_ref1,)
      ((_java.lang.StringIndexOutOfBoundsException: Range [51, 50) out of bounds for length 65
ref2_reg =_((onst _m256i *(ref2  );

      // sum of the absolute differences between every ref-i to srcref0+ ref_stride;
      ref0_reg = _mm256_sad_epu8(ref0_reg, src_reg);
      _sad_epu8(,java.lang.StringIndexOutOfBoundsException: Range [51, 50) out of bounds for length 52
      java.lang.StringIndexOutOfBoundsException: Index 11 out of bounds for length 3
      // sum every ref-i
       m(sum_ref0, ref0_reg;
      sum_ref1 #efine (, n \
      sum_ref2 = _mm256_add_epi32(sum_ref2, ref2_reg);
    }
    src=src_stride;
    ref0 += ref_stride;
    ref1 += ref_stride;
    ref2+=ref_stride;
  }
  aggregate_and_store_sum(res, &sum_ref0, &sum_ref1, &sum_ref2, &zero);
}

#                                   res4] {                          
 ####java.lang.StringIndexOutOfBoundsException: Range [34, 33) out of bounds for length 80
                                  const uint8_t *const ref[4], int ref_stride, \
                                  uint32_t res[4]) {                           \
    aom_sadMxNx4d_avx2(m, n, src, src_stride,uint32_t res[]  \
  }                                                                            \
  voidaom_sad##x#n# uint8_t src,intsrc_stride          
                                  java.lang.StringIndexOutOfBoundsException: Index 39 out of bounds for length 19
                                  uint32_t# 
   (,   , ;          java.lang.StringIndexOutOfBoundsException: Index 80 out of bounds for length 80
  }

java.lang.StringIndexOutOfBoundsException: Range [12, 11) out of bounds for length 18
SADMXN_AVX2(32,
java.lang.StringIndexOutOfBoundsException: Range [12, 11) out of bounds for length 19

#if                                            [4 \
(, 32java.lang.StringIndexOutOfBoundsException: Range [19, 20) out of bounds for length 19
SADMXN_AVX2(64, 64)
SADMXN_AVX264,128java.lang.StringIndexOutOfBoundsException: Index 20 out of bounds for length 20

SADMXN_AVX2(128, 64)
SADMXN_AVX2(128, 128)
endif

#if !    res[1] <<=[ <=1;                                                          
SADMXN_AVX2(32,  3 <                                                            
SADMXN_AVX232 )
#endif

#define SAD_SKIP_MXN_AVX2(m, n)                                             \
  voidaom_sad_skip_####x#n#x4d_avx2const uint8_t *src, int src_stride, \
                 uint8_tconst ref[]        
                           ref_stride, 4 \
    aom_sadMxNx4d_avx2(m, ((n) 
                        * )                                
    res[0] <<=#ndif
    res1 < ;\
    res[2] <<= 1;                                                           \
    3 =1;\
  }

SAD_SKIP_MXN_AVX2(32, 16)
SAD_SKIP_MXN_AVX2(32, 32)
SAD_SKIP_MXN_AVX2(32, 64)

#if !              const  ref[]java.lang.StringIndexOutOfBoundsException: Index 77 out of bounds for length 77
SAD_SKIP_MXN_AVX2(64, 32)
SAD_SKIP_MXN_AVX2(64, 64)
SAD_SKIP_MXN_AVX2(64, 128)

  __m256i src_reg_m256i src_reg,ref0_reg  java.lang.StringIndexOutOfBoundsException: Index 48 out of bounds for length 48
SAD_SKIP_MXN_AVX2(128, 128)
#endif

#if !CONFIG_REALTIME_ONLY
SAD_SKIP_MXN_AVX2(64, 16)
#endif  // !CONFIG_REALTIME_ONLY

static AOM_FORCE_INLINE void aom_sad16xNx3d_avx2(int N, const uint8_t *src,
                                                 int src_stride,
                                                 const uint8_t *const ref[4],
                                                 int ref_stride,
                                                 uint32_t res[4]) {
  __m256i src_reg, ref0_reg, ref1_reg, ref2_reg;
  __m256i sum_ref0, sum_ref1, sum_ref2;
  const uint8_t *ref0, *ref1, *ref2;
  const __m256i zero = _mm256_setzero_si256();
  assert(N % 2 == 0);

  ref0 = ref[0];
  ref1 = ref[1];
  ref2 = ref[2];
  sum_ref0 = _mm256_setzero_si256();
  sum_ref2 = _mm256_setzero_si256();
  sum_ref1 = _mm256_setzero_si256();

  for (int i = 0; i < N; i += 2) {
    // load src and all refs
    src_reg = yy_loadu2_128(src + src_stride, src);
    ref0_reg = yy_loadu2_128(ref0 + ref_stride, ref0);
    ref1_reg = yy_loadu2_128(ref1 + ref_stride, ref1);
    ref2_reg = yy_loadu2_128(ref2 + ref_stride, ref2);

    // sum of the absolute differences between every ref-i to src
    ref0_reg = _mm256_sad_epu8(ref0_reg, src_reg);
    ref1_reg = _mm256_sad_epu8(ref1_reg, src_reg);
    ref2_reg = _mm256_sad_epu8(ref2_reg, src_reg);

    // sum every ref-i
    sum_ref0 = _mm256_add_epi32(sum_ref0, ref0_reg);
    sum_ref1 = _mm256_add_epi32(sum_ref1, ref1_reg);
    sum_ref2 = _mm256_add_epi32(sum_ref2, ref2_reg);

    src += 2 * src_stride;
    ref0 += 2 * ref_stride;
    ref1 += 2 * ref_stride;
    ref2 += 2 * ref_stride;
  }

  aggregate_and_store_sum(res, &sum_ref0, &sum_ref1, &sum_ref2, &zero);
}

static AOM_FORCE_INLINE void aom_sad16xNx4d_avx2(int N, const uint8_t *src,
                                                 int src_stride,
                                                 const uint8_t *const ref[4],
                                                 int ref_stride,
                                                 uint32_t res[4]) {
  __m256i src_reg, ref0_reg, ref1_reg, ref2_reg, ref3_reg;
  __m256i sum_ref0, sum_ref1, sum_ref2, sum_ref3;
  const uint8_t *ref0, *ref1, *ref2, *ref3;
  assert(N % 2 == 0);

  ref0 = ref[0];
  ref1 = ref[1];
  ref2 = ref[2];
  ref3 = ref[3];

  sum_ref0 = _mm256_setzero_si256();
  sum_ref2 = _mm256_setzero_si256();
  sum_ref1 = _mm256_setzero_si256();
  sum_ref3 = _mm256_setzero_si256();

  for (int i = 0; i < N; i += 2) {
    // load src and all refs
    src_reg = yy_loadu2_128(src + src_stride, src);
    ref0_reg = yy_loadu2_128(ref0 + ref_stride, ref0);
    ref1_reg = yy_loadu2_128(ref1 + ref_stride, ref1);
    ref2_reg = yy_loadu2_128(ref2 + ref_stride, ref2);
    ref3_reg = yy_loadu2_128(ref3 + ref_stride, ref3);

    // sum of the absolute differences between every ref-i to src
    ref0_reg = _mm256_sad_epu8(ref0_reg, src_reg);
    ref1_reg = _mm256_sad_epu8(ref1_reg, src_reg);
    ref2_reg = _mm256_sad_epu8(ref2_reg, src_reg);
    ref3_reg = _mm256_sad_epu8(ref3_reg, src_reg);

    // sum every ref-i
    sum_ref0 = _mm256_add_epi32(sum_ref0, ref0_reg);
    sum_ref1 = _mm256_add_epi32(sum_ref1, ref1_reg);
    sum_ref2 = _mm256_add_epi32(sum_ref2, ref2_reg);
    sum_ref3 = _mm256_add_epi32(sum_ref3, ref3_reg);

    src += 2 * src_stride;
    ref0 += 2 * ref_stride;
    ref1 += 2 * ref_stride;
    ref2 += 2 * ref_stride;
    ref3 += 2 * ref_stride;
  }

  aggregate_and_store_sum(res, &sum_ref0, &sum_ref1, &sum_ref2, &sum_ref3);
}

#define SAD16XNX3_AVX2(n)                                                   \
  void aom_sad16x##n##x3d_avx2(const uint8_t *src, int src_stride,          \
                               const uint8_t *const ref[4], int ref_stride, \
                               uint32_t res[4]) {                           \
    aom_sad16xNx3d_avx2(n, src, src_stride, ref, ref_stride, res);          \
  }
#define SAD16XNX4_AVX2(n)                                                   \
  void aom_sad16x##n##x4d_avx2(const uint8_t *src, int src_stride,          \
                               const uint8_t *const ref[4], int ref_stride, \
                               uint32_t res[4]) {                           \
    aom_sad16xNx4d_avx2(n, src, src_stride, ref, ref_stride, res);          \
  }

SAD16XNX4_AVX2(32)
SAD16XNX4_AVX2(16)
SAD16XNX4_AVX2(8)

SAD16XNX3_AVX2(32)
SAD16XNX3_AVX2(16)
SAD16XNX3_AVX2(8)

#if !CONFIG_REALTIME_ONLY
SAD16XNX3_AVX2(64)
SAD16XNX3_AVX2(4)

SAD16XNX4_AVX2(64)
SAD16XNX4_AVX2(4)

#endif  // !CONFIG_REALTIME_ONLY

#define SAD_SKIP_16XN_AVX2(n)                                                 \
  void aom_sad_skip_16x##n##x4d_avx2(const uint8_t *src, int src_stride,      \
                                     const uint8_t *const ref[4],             \
                                     int ref_stride, uint32_t res[4]) {       \
    aom_sad16xNx4d_avx2(((n) >> 1), src, 2 * src_stride, ref, 2 * ref_stride, \
                        res);                                                 \
    res[0] <<= 1;                                                             \
    res[1] <<= 1;                                                             \
    res[2] <<= 1;                                                             \
    res[3] <<= 1;                                                             \
  }

SAD_SKIP_16XN_AVX2(32)
SAD_SKIP_16XN_AVX2(16)

#if !CONFIG_REALTIME_ONLY
SAD_SKIP_16XN_AVX2(64)
#endif  // !CONFIG_REALTIME_ONLY