chromium/third_party/libyuv/source/compare_gcc.cc

/*
 *  Copyright 2012 The LibYuv 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 "libyuv/basic_types.h"

#include "libyuv/compare_row.h"
#include "libyuv/row.h"

#ifdef __cplusplus
namespace libyuv {
extern "C" {
#endif

// This module is for GCC x86 and x64.
#if !defined(LIBYUV_DISABLE_X86) && (defined(__x86_64__) || defined(__i386__))

// "memory" clobber prevents the reads from being removed

#if defined(__x86_64__)
uint32_t HammingDistance_SSE42(const uint8_t* src_a,
                               const uint8_t* src_b,
                               int count) { … }
#else
uint32_t HammingDistance_SSE42(const uint8_t* src_a,
                               const uint8_t* src_b,
                               int count) {
  uint32_t diff = 0u;

  asm volatile (
      // Process 16 bytes per loop.
      LABELALIGN
      "1:                                        \n"
      "mov         (%0),%%ecx                    \n"
      "mov         0x4(%0),%%edx                 \n"
      "xor         (%1),%%ecx                    \n"
      "xor         0x4(%1),%%edx                 \n"
      "popcnt      %%ecx,%%ecx                   \n"
      "add         %%ecx,%3                      \n"
      "popcnt      %%edx,%%edx                   \n"
      "add         %%edx,%3                      \n"
      "mov         0x8(%0),%%ecx                 \n"
      "mov         0xc(%0),%%edx                 \n"
      "xor         0x8(%1),%%ecx                 \n"
      "xor         0xc(%1),%%edx                 \n"
      "popcnt      %%ecx,%%ecx                   \n"
      "add         %%ecx,%3                      \n"
      "popcnt      %%edx,%%edx                   \n"
      "add         %%edx,%3                      \n"
      "add         $0x10,%0                      \n"
      "add         $0x10,%1                      \n"
      "sub         $0x10,%2                      \n"
      "jg          1b                            \n"
      : "+r"(src_a),  // %0
        "+r"(src_b),  // %1
        "+r"(count),  // %2
        "+r"(diff)    // %3
      :
      : "cc", "memory", "ecx", "edx");

  return diff;
}
#endif

static const vec8 kNibbleMask = …;
static const vec8 kBitCount = …;

uint32_t HammingDistance_SSSE3(const uint8_t* src_a,
                               const uint8_t* src_b,
                               int count) { … }

#ifdef HAS_HAMMINGDISTANCE_AVX2
uint32_t HammingDistance_AVX2(const uint8_t* src_a,
                              const uint8_t* src_b,
                              int count) { … }
#endif  // HAS_HAMMINGDISTANCE_AVX2

uint32_t SumSquareError_SSE2(const uint8_t* src_a,
                             const uint8_t* src_b,
                             int count) { … }

static const uvec32 kHash16x33 = …;  // 33 ^ 16
static const uvec32 kHashMul0 = …;
static const uvec32 kHashMul1 = …;
static const uvec32 kHashMul2 = …;
static const uvec32 kHashMul3 = …;

uint32_t HashDjb2_SSE41(const uint8_t* src, int count, uint32_t seed) { … }
#endif  // defined(__x86_64__) || (defined(__i386__) && !defined(__pic__)))

#ifdef __cplusplus
}  // extern "C"
}  // namespace libyuv
#endif