godot/thirdparty/libwebp/src/dsp/common_sse2.h

// Copyright 2016 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.
// -----------------------------------------------------------------------------
//
// SSE2 code common to several files.
//
// Author: Vincent Rabaud ([email protected])

#ifndef WEBP_DSP_COMMON_SSE2_H_
#define WEBP_DSP_COMMON_SSE2_H_

#ifdef __cplusplus
extern "C" {
#endif

#if defined(WEBP_USE_SSE2)

#include <emmintrin.h>

//------------------------------------------------------------------------------
// Quite useful macro for debugging. Left here for convenience.

#if 0
#include <stdio.h>
static WEBP_INLINE void PrintReg(const __m128i r, const char* const name,
                                 int size) {
  int n;
  union {
    __m128i r;
    uint8_t i8[16];
    uint16_t i16[8];
    uint32_t i32[4];
    uint64_t i64[2];
  } tmp;
  tmp.r = r;
  fprintf(stderr, "%s\t: ", name);
  if (size == 8) {
    for (n = 0; n < 16; ++n) fprintf(stderr, "%.2x ", tmp.i8[n]);
  } else if (size == 16) {
    for (n = 0; n < 8; ++n) fprintf(stderr, "%.4x ", tmp.i16[n]);
  } else if (size == 32) {
    for (n = 0; n < 4; ++n) fprintf(stderr, "%.8x ", tmp.i32[n]);
  } else {
    for (n = 0; n < 2; ++n) fprintf(stderr, "%.16lx ", tmp.i64[n]);
  }
  fprintf(stderr, "\n");
}
#endif

//------------------------------------------------------------------------------
// Math functions.

// Return the sum of all the 8b in the register.
static WEBP_INLINE int VP8HorizontalAdd8b(const __m128i* const a) { … }

// Transpose two 4x4 16b matrices horizontally stored in registers.
static WEBP_INLINE void VP8Transpose_2_4x4_16b(
    const __m128i* const in0, const __m128i* const in1,
    const __m128i* const in2, const __m128i* const in3, __m128i* const out0,
    __m128i* const out1, __m128i* const out2, __m128i* const out3) { … }

//------------------------------------------------------------------------------
// Channel mixing.

// Function used several times in VP8PlanarTo24b.
// It samples the in buffer as follows: one every two unsigned char is stored
// at the beginning of the buffer, while the other half is stored at the end.
#define VP8PlanarTo24bHelper …

// Pack the planar buffers
// rrrr... rrrr... gggg... gggg... bbbb... bbbb....
// triplet by triplet in the output buffer rgb as rgbrgbrgbrgb ...
static WEBP_INLINE void VP8PlanarTo24b_SSE2(
    __m128i* const in0, __m128i* const in1, __m128i* const in2,
    __m128i* const in3, __m128i* const in4, __m128i* const in5) { … }

#undef VP8PlanarTo24bHelper

// Convert four packed four-channel buffers like argbargbargbargb... into the
// split channels aaaaa ... rrrr ... gggg .... bbbbb ......
static WEBP_INLINE void VP8L32bToPlanar_SSE2(__m128i* const in0,
                                             __m128i* const in1,
                                             __m128i* const in2,
                                             __m128i* const in3) { … }

#endif  // WEBP_USE_SSE2

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

#endif  // WEBP_DSP_COMMON_SSE2_H_