// Copyright (c) 2012 The Chromium Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.

// This webpage shows layout of YV12 and other YUV formats
// http://www.fourcc.org/yuv.php
// The actual conversion is best described here
// http://en.wikipedia.org/wiki/YUV
// An article on optimizing YUV conversion using tables instead of multiplies
// http://lestourtereaux.free.fr/papers/data/yuvrgb.pdf
//
// YV12 is a full plane of Y and a half height, half width chroma planes
// YV16 is a full plane of Y and a full height, half width chroma planes
//
// ARGB pixel format is output, which on little endian is stored as BGRA.
// The alpha is set to 255, allowing the application to use RGBA or RGB32.

#include "cobalt/media/base/yuv_convert.h"

#include <stddef.h>

#include <algorithm>

#include "base/basictypes.h"
#include "base/cpu.h"
#include "base/lazy_instance.h"
#include "base/logging.h"
#include "base/memory/aligned_memory.h"
#include "base/third_party/dynamic_annotations/dynamic_annotations.h"
#include "build/build_config.h"
#include "cobalt/media/base/simd/convert_rgb_to_yuv.h"
#include "cobalt/media/base/simd/convert_yuv_to_rgb.h"
#include "cobalt/media/base/simd/filter_yuv.h"

#if defined(ARCH_CPU_X86_FAMILY)
#if defined(COMPILER_MSVC)
#include <intrin.h>
#else
#include <mmintrin.h>
#endif
#endif

// Assembly functions are declared without namespace.
extern "C" {
void EmptyRegisterState_MMX();
}  // extern "C"

namespace media {

typedef void (*FilterYUVRowsProc)(uint8_t*, const uint8_t*, const uint8_t*, int,
                                  uint8_t);

typedef void (*ConvertRGBToYUVProc)(const uint8_t*, uint8_t*, uint8_t*,
                                    uint8_t*, int, int, int, int, int);

typedef void (*ConvertYUVToRGB32Proc)(const uint8_t*, const uint8_t*,
                                      const uint8_t*, uint8_t*, int, int, int,
                                      int, int, YUVType);

typedef void (*ConvertYUVAToARGBProc)(const uint8_t*, const uint8_t*,
                                      const uint8_t*, const uint8_t*, uint8_t*,
                                      int, int, int, int, int, int, YUVType);

typedef void (*ConvertYUVToRGB32RowProc)(const uint8_t*, const uint8_t*,
                                         const uint8_t*, uint8_t*, ptrdiff_t,
                                         const int16_t*);

typedef void (*ConvertYUVAToARGBRowProc)(const uint8_t*, const uint8_t*,
                                         const uint8_t*, const uint8_t*,
                                         uint8_t*, ptrdiff_t, const int16_t*);

typedef void (*ScaleYUVToRGB32RowProc)(const uint8_t*, const uint8_t*,
                                       const uint8_t*, uint8_t*, ptrdiff_t,
                                       ptrdiff_t, const int16_t*);

static FilterYUVRowsProc g_filter_yuv_rows_proc_ = NULL;
static ConvertYUVToRGB32RowProc g_convert_yuv_to_rgb32_row_proc_ = NULL;
static ScaleYUVToRGB32RowProc g_scale_yuv_to_rgb32_row_proc_ = NULL;
static ScaleYUVToRGB32RowProc g_linear_scale_yuv_to_rgb32_row_proc_ = NULL;
static ConvertRGBToYUVProc g_convert_rgb32_to_yuv_proc_ = NULL;
static ConvertRGBToYUVProc g_convert_rgb24_to_yuv_proc_ = NULL;
static ConvertYUVToRGB32Proc g_convert_yuv_to_rgb32_proc_ = NULL;
static ConvertYUVAToARGBProc g_convert_yuva_to_argb_proc_ = NULL;

static const int kYUVToRGBTableSize = 256 * 4 * 4 * sizeof(int16_t);

// base::AlignedMemory has a private operator new(), so wrap it in a struct so
// that we can put it in a LazyInstance::Leaky.
struct YUVToRGBTableWrapper {
  base::AlignedMemory<kYUVToRGBTableSize, 16> table;
};

typedef base::LazyInstance<YUVToRGBTableWrapper>::Leaky YUVToRGBTable;
static YUVToRGBTable g_table_rec601 = LAZY_INSTANCE_INITIALIZER;
static YUVToRGBTable g_table_jpeg = LAZY_INSTANCE_INITIALIZER;
static YUVToRGBTable g_table_rec709 = LAZY_INSTANCE_INITIALIZER;
static const int16_t* g_table_rec601_ptr = NULL;
static const int16_t* g_table_jpeg_ptr = NULL;
static const int16_t* g_table_rec709_ptr = NULL;

// Empty SIMD registers state after using them.
void EmptyRegisterStateStub() {}
#if defined(MEDIA_MMX_INTRINSICS_AVAILABLE)
void EmptyRegisterStateIntrinsic() { _mm_empty(); }
#endif
typedef void (*EmptyRegisterStateProc)();
static EmptyRegisterStateProc g_empty_register_state_proc_ = NULL;

// Get the appropriate value to bitshift by for vertical indices.
int GetVerticalShift(YUVType type) {
  switch (type) {
    case YV16:
      return 0;
    case YV12:
    case YV12J:
    case YV12HD:
      return 1;
  }
  NOTREACHED();
  return 0;
}

const int16_t* GetLookupTable(YUVType type) {
  switch (type) {
    case YV12:
    case YV16:
      return g_table_rec601_ptr;
    case YV12J:
      return g_table_jpeg_ptr;
    case YV12HD:
      return g_table_rec709_ptr;
  }
  NOTREACHED();
  return NULL;
}

// Populates a pre-allocated lookup table from a YUV->RGB matrix.
const int16_t* PopulateYUVToRGBTable(const double matrix[3][3], bool full_range,
                                     int16_t* table) {
  // We'll have 4 sub-tables that lie contiguous in memory, one for each of Y,
  // U, V and A.
  const int kNumTables = 4;
  // Each table has 256 rows (for all possible 8-bit values).
  const int kNumRows = 256;
  // Each row has 4 columns, for contributions to each of R, G, B and A.
  const int kNumColumns = 4;
  // Each element is a fixed-point (10.6) 16-bit signed value.
  const int kElementSize = sizeof(int16_t);

  // Sanity check that our constants here match the size of the statically
  // allocated tables.
  static_assert(
      kNumTables * kNumRows * kNumColumns * kElementSize == kYUVToRGBTableSize,
      "YUV lookup table size doesn't match expectation.");

  // Y needs an offset of -16 for color ranges that ignore the lower 16 values,
  // U and V get -128 to put them in [-128, 127] from [0, 255].
  int offsets[3] = {(full_range ? 0 : -16), -128, -128};

  for (int i = 0; i < kNumRows; ++i) {
    // Y, U, and V contributions to each of R, G, B and A.
    for (int j = 0; j < 3; ++j) {
#if defined(OS_ANDROID)
      // Android is RGBA.
      table[(j * kNumRows + i) * kNumColumns + 0] =
          matrix[j][0] * 64 * (i + offsets[j]) + 0.5;
      table[(j * kNumRows + i) * kNumColumns + 1] =
          matrix[j][1] * 64 * (i + offsets[j]) + 0.5;
      table[(j * kNumRows + i) * kNumColumns + 2] =
          matrix[j][2] * 64 * (i + offsets[j]) + 0.5;
#else
      // Other platforms are BGRA.
      table[(j * kNumRows + i) * kNumColumns + 0] =
          matrix[j][2] * 64 * (i + offsets[j]) + 0.5;
      table[(j * kNumRows + i) * kNumColumns + 1] =
          matrix[j][1] * 64 * (i + offsets[j]) + 0.5;
      table[(j * kNumRows + i) * kNumColumns + 2] =
          matrix[j][0] * 64 * (i + offsets[j]) + 0.5;
#endif
      // Alpha contributions from Y and V are always 0. U is set such that
      // all values result in a full '255' alpha value.
      table[(j * kNumRows + i) * kNumColumns + 3] = (j == 1) ? 256 * 64 - 1 : 0;
    }
    // And YUVA alpha is passed through as-is.
    for (int k = 0; k < kNumTables; ++k)
      table[((kNumTables - 1) * kNumRows + i) * kNumColumns + k] = i;
  }

  return table;
}

void InitializeCPUSpecificYUVConversions() {
  CHECK(!g_filter_yuv_rows_proc_);
  CHECK(!g_convert_yuv_to_rgb32_row_proc_);
  CHECK(!g_scale_yuv_to_rgb32_row_proc_);
  CHECK(!g_linear_scale_yuv_to_rgb32_row_proc_);
  CHECK(!g_convert_rgb32_to_yuv_proc_);
  CHECK(!g_convert_rgb24_to_yuv_proc_);
  CHECK(!g_convert_yuv_to_rgb32_proc_);
  CHECK(!g_convert_yuva_to_argb_proc_);
  CHECK(!g_empty_register_state_proc_);

  g_filter_yuv_rows_proc_ = FilterYUVRows_C;
  g_convert_yuv_to_rgb32_row_proc_ = ConvertYUVToRGB32Row_C;
  g_scale_yuv_to_rgb32_row_proc_ = ScaleYUVToRGB32Row_C;
  g_linear_scale_yuv_to_rgb32_row_proc_ = LinearScaleYUVToRGB32Row_C;
  g_convert_rgb32_to_yuv_proc_ = ConvertRGB32ToYUV_C;
  g_convert_rgb24_to_yuv_proc_ = ConvertRGB24ToYUV_C;
  g_convert_yuv_to_rgb32_proc_ = ConvertYUVToRGB32_C;
  g_convert_yuva_to_argb_proc_ = ConvertYUVAToARGB_C;
  g_empty_register_state_proc_ = EmptyRegisterStateStub;

// Assembly code confuses MemorySanitizer. Also not available in iOS builds.
#if defined(ARCH_CPU_X86_FAMILY) && !defined(MEMORY_SANITIZER) && \
    !defined(OS_IOS)
  g_convert_yuva_to_argb_proc_ = ConvertYUVAToARGB_MMX;

#if defined(MEDIA_MMX_INTRINSICS_AVAILABLE)
  g_empty_register_state_proc_ = EmptyRegisterStateIntrinsic;
#else
  g_empty_register_state_proc_ = EmptyRegisterState_MMX;
#endif

  g_convert_yuv_to_rgb32_row_proc_ = ConvertYUVToRGB32Row_SSE;
  g_convert_yuv_to_rgb32_proc_ = ConvertYUVToRGB32_SSE;

  g_filter_yuv_rows_proc_ = FilterYUVRows_SSE2;
  g_convert_rgb32_to_yuv_proc_ = ConvertRGB32ToYUV_SSE2;

#if defined(ARCH_CPU_X86_64)
  g_scale_yuv_to_rgb32_row_proc_ = ScaleYUVToRGB32Row_SSE2_X64;

  // Technically this should be in the MMX section, but MSVC will optimize out
  // the export of LinearScaleYUVToRGB32Row_MMX, which is required by the unit
  // tests, if that decision can be made at compile time.  Since all X64 CPUs
  // have SSE2, we can hack around this by making the selection here.
  g_linear_scale_yuv_to_rgb32_row_proc_ = LinearScaleYUVToRGB32Row_MMX_X64;
#else
  g_scale_yuv_to_rgb32_row_proc_ = ScaleYUVToRGB32Row_SSE;
  g_linear_scale_yuv_to_rgb32_row_proc_ = LinearScaleYUVToRGB32Row_SSE;
#endif

  base::CPU cpu;
  if (cpu.has_ssse3()) {
    g_convert_rgb24_to_yuv_proc_ = &ConvertRGB24ToYUV_SSSE3;

    // TODO(hclam): Add ConvertRGB32ToYUV_SSSE3 when the cyan problem is solved.
    // See: crbug.com/100462
  }
#endif

  // Initialize YUV conversion lookup tables.

  // SD Rec601 YUV->RGB matrix, see http://www.fourcc.org/fccyvrgb.php
  const double kRec601ConvertMatrix[3][3] = {
      {1.164, 1.164, 1.164}, {0.0, -0.391, 2.018}, {1.596, -0.813, 0.0},
  };

  // JPEG table, values from above link.
  const double kJPEGConvertMatrix[3][3] = {
      {1.0, 1.0, 1.0}, {0.0, -0.34414, 1.772}, {1.402, -0.71414, 0.0},
  };

  // Rec709 "HD" color space, values from:
  // http://www.equasys.de/colorconversion.html
  const double kRec709ConvertMatrix[3][3] = {
      {1.164, 1.164, 1.164}, {0.0, -0.213, 2.112}, {1.793, -0.533, 0.0},
  };

  PopulateYUVToRGBTable(kRec601ConvertMatrix, false,
                        g_table_rec601.Get().table.data_as<int16_t>());
  PopulateYUVToRGBTable(kJPEGConvertMatrix, true,
                        g_table_jpeg.Get().table.data_as<int16_t>());
  PopulateYUVToRGBTable(kRec709ConvertMatrix, false,
                        g_table_rec709.Get().table.data_as<int16_t>());
  g_table_rec601_ptr = g_table_rec601.Get().table.data_as<int16_t>();
  g_table_rec709_ptr = g_table_rec709.Get().table.data_as<int16_t>();
  g_table_jpeg_ptr = g_table_jpeg.Get().table.data_as<int16_t>();
}

// Empty SIMD registers state after using them.
void EmptyRegisterState() { g_empty_register_state_proc_(); }

// 16.16 fixed point arithmetic
const int kFractionBits = 16;
const int kFractionMax = 1 << kFractionBits;
const int kFractionMask = ((1 << kFractionBits) - 1);

// Scale a frame of YUV to 32 bit ARGB.
void ScaleYUVToRGB32(const uint8_t* y_buf, const uint8_t* u_buf,
                     const uint8_t* v_buf, uint8_t* rgb_buf, int source_width,
                     int source_height, int width, int height, int y_pitch,
                     int uv_pitch, int rgb_pitch, YUVType yuv_type,
                     Rotate view_rotate, ScaleFilter filter) {
  // Handle zero sized sources and destinations.
  if ((yuv_type == YV12 && (source_width < 2 || source_height < 2)) ||
      (yuv_type == YV16 && (source_width < 2 || source_height < 1)) ||
      width == 0 || height == 0)
    return;

  const int16_t* lookup_table = GetLookupTable(yuv_type);

  // 4096 allows 3 buffers to fit in 12k.
  // Helps performance on CPU with 16K L1 cache.
  // Large enough for 3830x2160 and 30" displays which are 2560x1600.
  const int kFilterBufferSize = 4096;
  // Disable filtering if the screen is too big (to avoid buffer overflows).
  // This should never happen to regular users: they don't have monitors
  // wider than 4096 pixels.
  // TODO(fbarchard): Allow rotated videos to filter.
  if (source_width > kFilterBufferSize || view_rotate) filter = FILTER_NONE;

  unsigned int y_shift = GetVerticalShift(yuv_type);
  // Diagram showing origin and direction of source sampling.
  // ->0   4<-
  // 7       3
  //
  // 6       5
  // ->1   2<-
  // Rotations that start at right side of image.
  if ((view_rotate == ROTATE_180) || (view_rotate == ROTATE_270) ||
      (view_rotate == MIRROR_ROTATE_0) || (view_rotate == MIRROR_ROTATE_90)) {
    y_buf += source_width - 1;
    u_buf += source_width / 2 - 1;
    v_buf += source_width / 2 - 1;
    source_width = -source_width;
  }
  // Rotations that start at bottom of image.
  if ((view_rotate == ROTATE_90) || (view_rotate == ROTATE_180) ||
      (view_rotate == MIRROR_ROTATE_90) || (view_rotate == MIRROR_ROTATE_180)) {
    y_buf += (source_height - 1) * y_pitch;
    u_buf += ((source_height >> y_shift) - 1) * uv_pitch;
    v_buf += ((source_height >> y_shift) - 1) * uv_pitch;
    source_height = -source_height;
  }

  int source_dx = source_width * kFractionMax / width;

  if ((view_rotate == ROTATE_90) || (view_rotate == ROTATE_270)) {
    int tmp = height;
    height = width;
    width = tmp;
    tmp = source_height;
    source_height = source_width;
    source_width = tmp;
    int source_dy = source_height * kFractionMax / height;
    source_dx = ((source_dy >> kFractionBits) * y_pitch) << kFractionBits;
    if (view_rotate == ROTATE_90) {
      y_pitch = -1;
      uv_pitch = -1;
      source_height = -source_height;
    } else {
      y_pitch = 1;
      uv_pitch = 1;
    }
  }

  // Need padding because FilterRows() will write 1 to 16 extra pixels
  // after the end for SSE2 version.
  uint8_t yuvbuf[16 + kFilterBufferSize * 3 + 16];
  uint8_t* ybuf = reinterpret_cast<uint8_t*>(
      reinterpret_cast<uintptr_t>(yuvbuf + 15) & ~15);
  uint8_t* ubuf = ybuf + kFilterBufferSize;
  uint8_t* vbuf = ubuf + kFilterBufferSize;

  // TODO(fbarchard): Fixed point math is off by 1 on negatives.

  // We take a y-coordinate in [0,1] space in the source image space, and
  // transform to a y-coordinate in [0,1] space in the destination image space.
  // Note that the coordinate endpoints lie on pixel boundaries, not on pixel
  // centers: e.g. a two-pixel-high image will have pixel centers at 0.25 and
  // 0.75.  The formula is as follows (in fixed-point arithmetic):
  //   y_dst = dst_height * ((y_src + 0.5) / src_height)
  //   dst_pixel = clamp([0, dst_height - 1], floor(y_dst - 0.5))
  // Implement this here as an accumulator + delta, to avoid expensive math
  // in the loop.
  int source_y_subpixel_accum =
      ((kFractionMax / 2) * source_height) / height - (kFractionMax / 2);
  int source_y_subpixel_delta = ((1 << kFractionBits) * source_height) / height;

  // TODO(fbarchard): Split this into separate function for better efficiency.
  for (int y = 0; y < height; ++y) {
    uint8_t* dest_pixel = rgb_buf + y * rgb_pitch;
    int source_y_subpixel = source_y_subpixel_accum;
    source_y_subpixel_accum += source_y_subpixel_delta;
    if (source_y_subpixel < 0)
      source_y_subpixel = 0;
    else if (source_y_subpixel > ((source_height - 1) << kFractionBits))
      source_y_subpixel = (source_height - 1) << kFractionBits;

    const uint8_t* y_ptr = NULL;
    const uint8_t* u_ptr = NULL;
    const uint8_t* v_ptr = NULL;
    // Apply vertical filtering if necessary.
    // TODO(fbarchard): Remove memcpy when not necessary.
    if (filter & media::FILTER_BILINEAR_V) {
      int source_y = source_y_subpixel >> kFractionBits;
      y_ptr = y_buf + source_y * y_pitch;
      u_ptr = u_buf + (source_y >> y_shift) * uv_pitch;
      v_ptr = v_buf + (source_y >> y_shift) * uv_pitch;

      // Vertical scaler uses 16.8 fixed point.
      uint8_t source_y_fraction = (source_y_subpixel & kFractionMask) >> 8;
      if (source_y_fraction != 0) {
        g_filter_yuv_rows_proc_(ybuf, y_ptr, y_ptr + y_pitch, source_width,
                                source_y_fraction);
      } else {
        memcpy(ybuf, y_ptr, source_width);
      }
      y_ptr = ybuf;
      ybuf[source_width] = ybuf[source_width - 1];

      int uv_source_width = (source_width + 1) / 2;
      uint8_t source_uv_fraction;

      // For formats with half-height UV planes, each even-numbered pixel row
      // should not interpolate, since the next row to interpolate from should
      // be a duplicate of the current row.
      if (y_shift && (source_y & 0x1) == 0)
        source_uv_fraction = 0;
      else
        source_uv_fraction = source_y_fraction;

      if (source_uv_fraction != 0) {
        g_filter_yuv_rows_proc_(ubuf, u_ptr, u_ptr + uv_pitch, uv_source_width,
                                source_uv_fraction);
        g_filter_yuv_rows_proc_(vbuf, v_ptr, v_ptr + uv_pitch, uv_source_width,
                                source_uv_fraction);
      } else {
        memcpy(ubuf, u_ptr, uv_source_width);
        memcpy(vbuf, v_ptr, uv_source_width);
      }
      u_ptr = ubuf;
      v_ptr = vbuf;
      ubuf[uv_source_width] = ubuf[uv_source_width - 1];
      vbuf[uv_source_width] = vbuf[uv_source_width - 1];
    } else {
      // Offset by 1/2 pixel for center sampling.
      int source_y = (source_y_subpixel + (kFractionMax / 2)) >> kFractionBits;
      y_ptr = y_buf + source_y * y_pitch;
      u_ptr = u_buf + (source_y >> y_shift) * uv_pitch;
      v_ptr = v_buf + (source_y >> y_shift) * uv_pitch;
    }
    if (source_dx == kFractionMax) {  // Not scaled
      g_convert_yuv_to_rgb32_row_proc_(y_ptr, u_ptr, v_ptr, dest_pixel, width,
                                       lookup_table);
    } else {
      if (filter & FILTER_BILINEAR_H) {
        g_linear_scale_yuv_to_rgb32_row_proc_(y_ptr, u_ptr, v_ptr, dest_pixel,
                                              width, source_dx, lookup_table);
      } else {
        g_scale_yuv_to_rgb32_row_proc_(y_ptr, u_ptr, v_ptr, dest_pixel, width,
                                       source_dx, lookup_table);
      }
    }
  }

  g_empty_register_state_proc_();
}

// Scale a frame of YV12 to 32 bit ARGB for a specific rectangle.
void ScaleYUVToRGB32WithRect(const uint8_t* y_buf, const uint8_t* u_buf,
                             const uint8_t* v_buf, uint8_t* rgb_buf,
                             int source_width, int source_height,
                             int dest_width, int dest_height,
                             int dest_rect_left, int dest_rect_top,
                             int dest_rect_right, int dest_rect_bottom,
                             int y_pitch, int uv_pitch, int rgb_pitch) {
  // This routine doesn't currently support up-scaling.
  CHECK_LE(dest_width, source_width);
  CHECK_LE(dest_height, source_height);

  // Sanity-check the destination rectangle.
  DCHECK(dest_rect_left >= 0 && dest_rect_right <= dest_width);
  DCHECK(dest_rect_top >= 0 && dest_rect_bottom <= dest_height);
  DCHECK(dest_rect_right > dest_rect_left);
  DCHECK(dest_rect_bottom > dest_rect_top);

  const int16_t* lookup_table = GetLookupTable(YV12);

  // Fixed-point value of vertical and horizontal scale down factor.
  // Values are in the format 16.16.
  int y_step = kFractionMax * source_height / dest_height;
  int x_step = kFractionMax * source_width / dest_width;

  // Determine the coordinates of the rectangle in 16.16 coords.
  // NB: Our origin is the *center* of the top/left pixel, NOT its top/left.
  // If we're down-scaling by more than a factor of two, we start with a 50%
  // fraction to avoid degenerating to point-sampling - we should really just
  // fix the fraction at 50% for all pixels in that case.
  int source_left = dest_rect_left * x_step;
  int source_right = (dest_rect_right - 1) * x_step;
  if (x_step < kFractionMax * 2) {
    source_left += ((x_step - kFractionMax) / 2);
    source_right += ((x_step - kFractionMax) / 2);
  } else {
    source_left += kFractionMax / 2;
    source_right += kFractionMax / 2;
  }
  int source_top = dest_rect_top * y_step;
  if (y_step < kFractionMax * 2) {
    source_top += ((y_step - kFractionMax) / 2);
  } else {
    source_top += kFractionMax / 2;
  }

  // Determine the parts of the Y, U and V buffers to interpolate.
  int source_y_left = source_left >> kFractionBits;
  int source_y_right =
      std::min((source_right >> kFractionBits) + 2, source_width + 1);

  int source_uv_left = source_y_left / 2;
  int source_uv_right = std::min((source_right >> (kFractionBits + 1)) + 2,
                                 (source_width + 1) / 2);

  int source_y_width = source_y_right - source_y_left;
  int source_uv_width = source_uv_right - source_uv_left;

  // Determine number of pixels in each output row.
  int dest_rect_width = dest_rect_right - dest_rect_left;

  // Intermediate buffer for vertical interpolation.
  // 4096 bytes allows 3 buffers to fit in 12k, which fits in a 16K L1 cache,
  // and is bigger than most users will generally need.
  // The buffer is 16-byte aligned and padded with 16 extra bytes; some of the
  // FilterYUVRowsProcs have alignment requirements, and the SSE version can
  // write up to 16 bytes past the end of the buffer.
  const int kFilterBufferSize = 4096;
  const bool kAvoidUsingOptimizedFilter = source_width > kFilterBufferSize;
  uint8_t yuv_temp[16 + kFilterBufferSize * 3 + 16];
  // memset() yuv_temp to 0 to avoid bogus warnings when running on Valgrind.
  if (RunningOnValgrind()) memset(yuv_temp, 0, sizeof(yuv_temp));
  uint8_t* y_temp = reinterpret_cast<uint8_t*>(
      reinterpret_cast<uintptr_t>(yuv_temp + 15) & ~15);
  uint8_t* u_temp = y_temp + kFilterBufferSize;
  uint8_t* v_temp = u_temp + kFilterBufferSize;

  // Move to the top-left pixel of output.
  rgb_buf += dest_rect_top * rgb_pitch;
  rgb_buf += dest_rect_left * 4;

  // For each destination row perform interpolation and color space
  // conversion to produce the output.
  for (int row = dest_rect_top; row < dest_rect_bottom; ++row) {
    // Round the fixed-point y position to get the current row.
    int source_row = source_top >> kFractionBits;
    int source_uv_row = source_row / 2;
    DCHECK(source_row < source_height);

    // Locate the first row for each plane for interpolation.
    const uint8_t* y0_ptr = y_buf + y_pitch * source_row + source_y_left;
    const uint8_t* u0_ptr = u_buf + uv_pitch * source_uv_row + source_uv_left;
    const uint8_t* v0_ptr = v_buf + uv_pitch * source_uv_row + source_uv_left;
    const uint8_t* y1_ptr = NULL;
    const uint8_t* u1_ptr = NULL;
    const uint8_t* v1_ptr = NULL;

    // Locate the second row for interpolation, being careful not to overrun.
    if (source_row + 1 >= source_height) {
      y1_ptr = y0_ptr;
    } else {
      y1_ptr = y0_ptr + y_pitch;
    }
    if (source_uv_row + 1 >= (source_height + 1) / 2) {
      u1_ptr = u0_ptr;
      v1_ptr = v0_ptr;
    } else {
      u1_ptr = u0_ptr + uv_pitch;
      v1_ptr = v0_ptr + uv_pitch;
    }

    if (!kAvoidUsingOptimizedFilter) {
      // Vertical scaler uses 16.8 fixed point.
      uint8_t fraction = (source_top & kFractionMask) >> 8;
      g_filter_yuv_rows_proc_(y_temp + source_y_left, y0_ptr, y1_ptr,
                              source_y_width, fraction);
      g_filter_yuv_rows_proc_(u_temp + source_uv_left, u0_ptr, u1_ptr,
                              source_uv_width, fraction);
      g_filter_yuv_rows_proc_(v_temp + source_uv_left, v0_ptr, v1_ptr,
                              source_uv_width, fraction);

      // Perform horizontal interpolation and color space conversion.
      // TODO(hclam): Use the MMX version after more testing.
      LinearScaleYUVToRGB32RowWithRange_C(y_temp, u_temp, v_temp, rgb_buf,
                                          dest_rect_width, source_left, x_step,
                                          lookup_table);
    } else {
      // If the frame is too large then we linear scale a single row.
      LinearScaleYUVToRGB32RowWithRange_C(y0_ptr, u0_ptr, v0_ptr, rgb_buf,
                                          dest_rect_width, source_left, x_step,
                                          lookup_table);
    }

    // Advance vertically in the source and destination image.
    source_top += y_step;
    rgb_buf += rgb_pitch;
  }

  g_empty_register_state_proc_();
}

void ConvertRGB32ToYUV(const uint8_t* rgbframe, uint8_t* yplane,
                       uint8_t* uplane, uint8_t* vplane, int width, int height,
                       int rgbstride, int ystride, int uvstride) {
  g_convert_rgb32_to_yuv_proc_(rgbframe, yplane, uplane, vplane, width, height,
                               rgbstride, ystride, uvstride);
}

void ConvertRGB24ToYUV(const uint8_t* rgbframe, uint8_t* yplane,
                       uint8_t* uplane, uint8_t* vplane, int width, int height,
                       int rgbstride, int ystride, int uvstride) {
  g_convert_rgb24_to_yuv_proc_(rgbframe, yplane, uplane, vplane, width, height,
                               rgbstride, ystride, uvstride);
}

void ConvertYUVToRGB32(const uint8_t* yplane, const uint8_t* uplane,
                       const uint8_t* vplane, uint8_t* rgbframe, int width,
                       int height, int ystride, int uvstride, int rgbstride,
                       YUVType yuv_type) {
  g_convert_yuv_to_rgb32_proc_(yplane, uplane, vplane, rgbframe, width, height,
                               ystride, uvstride, rgbstride, yuv_type);
}

void ConvertYUVAToARGB(const uint8_t* yplane, const uint8_t* uplane,
                       const uint8_t* vplane, const uint8_t* aplane,
                       uint8_t* rgbframe, int width, int height, int ystride,
                       int uvstride, int astride, int rgbstride,
                       YUVType yuv_type) {
  g_convert_yuva_to_argb_proc_(yplane, uplane, vplane, aplane, rgbframe, width,
                               height, ystride, uvstride, astride, rgbstride,
                               yuv_type);
}

}  // namespace media