// Copyright 2010 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING 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.
// -----------------------------------------------------------------------------
//
// inline YUV<->RGB conversion function
//
// The exact naming is Y'CbCr, following the ITU-R BT.601 standard.
// More information at: https://en.wikipedia.org/wiki/YCbCr
// Y = 0.2569 * R + 0.5044 * G + 0.0979 * B + 16
// U = -0.1483 * R - 0.2911 * G + 0.4394 * B + 128
// V = 0.4394 * R - 0.3679 * G - 0.0715 * B + 128
// We use 16bit fixed point operations for RGB->YUV conversion (YUV_FIX).
//
// For the Y'CbCr to RGB conversion, the BT.601 specification reads:
//   R = 1.164 * (Y-16) + 1.596 * (V-128)
//   G = 1.164 * (Y-16) - 0.813 * (V-128) - 0.391 * (U-128)
//   B = 1.164 * (Y-16)                   + 2.018 * (U-128)
// where Y is in the [16,235] range, and U/V in the [16,240] range.
//
// The fixed-point implementation used here is:
//  R = (19077 . y             + 26149 . v - 14234) >> 6
//  G = (19077 . y -  6419 . u - 13320 . v +  8708) >> 6
//  B = (19077 . y + 33050 . u             - 17685) >> 6
// where the '.' operator is the mulhi_epu16 variant:
//   a . b = ((a << 8) * b) >> 16
// that preserves 8 bits of fractional precision before final descaling.

// Author: Skal (pascal.massimino@gmail.com)

#ifndef WEBP_DSP_YUV_H_
#define WEBP_DSP_YUV_H_

#include "src/dsp/dsp.h"
#include "src/dec/vp8_dec.h"

//------------------------------------------------------------------------------
// YUV -> RGB conversion

#ifdef __cplusplus
extern "C" {
#endif

enum {
  YUV_FIX = 16,                    // fixed-point precision for RGB->YUV
  YUV_HALF = 1 << (YUV_FIX - 1),

  YUV_FIX2 = 6,                   // fixed-point precision for YUV->RGB
  YUV_MASK2 = (256 << YUV_FIX2) - 1
};

//------------------------------------------------------------------------------
// slower on x86 by ~7-8%, but bit-exact with the SSE2/NEON version

static WEBP_INLINE int MultHi(int v, int coeff) {   // _mm_mulhi_epu16 emulation
  return (v * coeff) >> 8;
}

static WEBP_INLINE int VP8Clip8(int v) {
  return ((v & ~YUV_MASK2) == 0) ? (v >> YUV_FIX2) : (v < 0) ? 0 : 255;
}

static WEBP_INLINE int VP8YUVToR(int y, int v) {
  return VP8Clip8(MultHi(y, 19077) + MultHi(v, 26149) - 14234);
}

static WEBP_INLINE int VP8YUVToG(int y, int u, int v) {
  return VP8Clip8(MultHi(y, 19077) - MultHi(u, 6419) - MultHi(v, 13320) + 8708);
}

static WEBP_INLINE int VP8YUVToB(int y, int u) {
  return VP8Clip8(MultHi(y, 19077) + MultHi(u, 33050) - 17685);
}

static WEBP_INLINE void VP8YuvToRgb(int y, int u, int v,
                                    uint8_t* const rgb) {
  rgb[0] = VP8YUVToR(y, v);
  rgb[1] = VP8YUVToG(y, u, v);
  rgb[2] = VP8YUVToB(y, u);
}

static WEBP_INLINE void VP8YuvToBgr(int y, int u, int v,
                                    uint8_t* const bgr) {
  bgr[0] = VP8YUVToB(y, u);
  bgr[1] = VP8YUVToG(y, u, v);
  bgr[2] = VP8YUVToR(y, v);
}

static WEBP_INLINE void VP8YuvToRgb565(int y, int u, int v,
                                       uint8_t* const rgb) {
  const int r = VP8YUVToR(y, v);      // 5 usable bits
  const int g = VP8YUVToG(y, u, v);   // 6 usable bits
  const int b = VP8YUVToB(y, u);      // 5 usable bits
  const int rg = (r & 0xf8) | (g >> 5);
  const int gb = ((g << 3) & 0xe0) | (b >> 3);
#if (WEBP_SWAP_16BIT_CSP == 1)
  rgb[0] = gb;
  rgb[1] = rg;
#else
  rgb[0] = rg;
  rgb[1] = gb;
#endif
}

static WEBP_INLINE void VP8YuvToRgba4444(int y, int u, int v,
                                         uint8_t* const argb) {
  const int r = VP8YUVToR(y, v);        // 4 usable bits
  const int g = VP8YUVToG(y, u, v);     // 4 usable bits
  const int b = VP8YUVToB(y, u);        // 4 usable bits
  const int rg = (r & 0xf0) | (g >> 4);
  const int ba = (b & 0xf0) | 0x0f;     // overwrite the lower 4 bits
#if (WEBP_SWAP_16BIT_CSP == 1)
  argb[0] = ba;
  argb[1] = rg;
#else
  argb[0] = rg;
  argb[1] = ba;
#endif
}

//-----------------------------------------------------------------------------
// Alpha handling variants

static WEBP_INLINE void VP8YuvToArgb(uint8_t y, uint8_t u, uint8_t v,
                                     uint8_t* const argb) {
  argb[0] = 0xff;
  VP8YuvToRgb(y, u, v, argb + 1);
}

static WEBP_INLINE void VP8YuvToBgra(uint8_t y, uint8_t u, uint8_t v,
                                     uint8_t* const bgra) {
  VP8YuvToBgr(y, u, v, bgra);
  bgra[3] = 0xff;
}

static WEBP_INLINE void VP8YuvToRgba(uint8_t y, uint8_t u, uint8_t v,
                                     uint8_t* const rgba) {
  VP8YuvToRgb(y, u, v, rgba);
  rgba[3] = 0xff;
}

//-----------------------------------------------------------------------------
// SSE2 extra functions (mostly for upsampling_sse2.c)

#if defined(WEBP_USE_SSE2)

// Process 32 pixels and store the result (16b, 24b or 32b per pixel) in *dst.
void VP8YuvToRgba32_SSE2(const uint8_t* y, const uint8_t* u, const uint8_t* v,
                         uint8_t* dst);
void VP8YuvToRgb32_SSE2(const uint8_t* y, const uint8_t* u, const uint8_t* v,
                        uint8_t* dst);
void VP8YuvToBgra32_SSE2(const uint8_t* y, const uint8_t* u, const uint8_t* v,
                         uint8_t* dst);
void VP8YuvToBgr32_SSE2(const uint8_t* y, const uint8_t* u, const uint8_t* v,
                        uint8_t* dst);
void VP8YuvToArgb32_SSE2(const uint8_t* y, const uint8_t* u, const uint8_t* v,
                         uint8_t* dst);
void VP8YuvToRgba444432_SSE2(const uint8_t* y, const uint8_t* u,
                             const uint8_t* v, uint8_t* dst);
void VP8YuvToRgb56532_SSE2(const uint8_t* y, const uint8_t* u, const uint8_t* v,
                           uint8_t* dst);

#endif    // WEBP_USE_SSE2

//-----------------------------------------------------------------------------
// SSE41 extra functions (mostly for upsampling_sse41.c)

#if defined(WEBP_USE_SSE41)

// Process 32 pixels and store the result (16b, 24b or 32b per pixel) in *dst.
void VP8YuvToRgb32_SSE41(const uint8_t* y, const uint8_t* u, const uint8_t* v,
                         uint8_t* dst);
void VP8YuvToBgr32_SSE41(const uint8_t* y, const uint8_t* u, const uint8_t* v,
                         uint8_t* dst);

#endif    // WEBP_USE_SSE41

//------------------------------------------------------------------------------
// RGB -> YUV conversion

// Stub functions that can be called with various rounding values:
static WEBP_INLINE int VP8ClipUV(int uv, int rounding) {
  uv = (uv + rounding + (128 << (YUV_FIX + 2))) >> (YUV_FIX + 2);
  return ((uv & ~0xff) == 0) ? uv : (uv < 0) ? 0 : 255;
}

static WEBP_INLINE int VP8RGBToY(int r, int g, int b, int rounding) {
  const int luma = 16839 * r + 33059 * g + 6420 * b;
  return (luma + rounding + (16 << YUV_FIX)) >> YUV_FIX;  // no need to clip
}

static WEBP_INLINE int VP8RGBToU(int r, int g, int b, int rounding) {
  const int u = -9719 * r - 19081 * g + 28800 * b;
  return VP8ClipUV(u, rounding);
}

static WEBP_INLINE int VP8RGBToV(int r, int g, int b, int rounding) {
  const int v = +28800 * r - 24116 * g - 4684 * b;
  return VP8ClipUV(v, rounding);
}

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

#endif  // WEBP_DSP_YUV_H_