source/common/x86/pixel-util.h

   1 /*****************************************************************************
   2  * Copyright (C) 2013 x265 project
   3  *
   4  * Authors: Steve Borho <steve@borho.org>
   5  *
   6  * This program is free software; you can redistribute it and/or modify
   7  * it under the terms of the GNU General Public License as published by
   8  * the Free Software Foundation; either version 2 of the License, or
   9  * (at your option) any later version.
  10  *
  11  * This program is distributed in the hope that it will be useful,
  12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  14  * GNU General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU General Public License
  17  * along with this program; if not, write to the Free Software
  18  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02111, USA.
  19  *
  20  * This program is also available under a commercial proprietary license.
  21  * For more information, contact us at license @ x265.com.
  22  *****************************************************************************/
  23
  24 #ifndef X265_PIXEL_UTIL_H
  25 #define X265_PIXEL_UTIL_H
  26
  27 void x265_getResidual4_sse2(const pixel* fenc, const pixel* pred, int16_t* residual, intptr_t stride);
  28 void x265_getResidual8_sse2(const pixel* fenc, const pixel* pred, int16_t* residual, intptr_t stride);
  29 void x265_getResidual16_sse2(const pixel* fenc, const pixel* pred, int16_t* residual, intptr_t stride);
  30 void x265_getResidual16_sse4(const pixel* fenc, const pixel* pred, int16_t* residual, intptr_t stride);
  31 void x265_getResidual32_sse2(const pixel* fenc, const pixel* pred, int16_t* residual, intptr_t stride);
  32 void x265_getResidual32_sse4(const pixel* fenc, const pixel* pred, int16_t* residual, intptr_t stride);
  33
  34 void x265_transpose4_sse2(pixel* dest, const pixel* src, intptr_t stride);
  35 void x265_transpose8_sse2(pixel* dest, const pixel* src, intptr_t stride);
  36 void x265_transpose16_sse2(pixel* dest, const pixel* src, intptr_t stride);
  37 void x265_transpose32_sse2(pixel* dest, const pixel* src, intptr_t stride);
  38 void x265_transpose64_sse2(pixel* dest, const pixel* src, intptr_t stride);
  39
  40 void x265_transpose8_avx2(pixel* dest, const pixel* src, intptr_t stride);
  41 void x265_transpose16_avx2(pixel* dest, const pixel* src, intptr_t stride);
  42 void x265_transpose32_avx2(pixel* dest, const pixel* src, intptr_t stride);
  43 void x265_transpose64_avx2(pixel* dest, const pixel* src, intptr_t stride);
  44
  45 uint32_t x265_quant_sse4(const int16_t* coef, const int32_t* quantCoeff, int32_t* deltaU, int16_t* qCoef, int qBits, int add, int numCoeff);
  46 uint32_t x265_quant_avx2(const int16_t* coef, const int32_t* quantCoeff, int32_t* deltaU, int16_t* qCoef, int qBits, int add, int numCoeff);
  47 uint32_t x265_nquant_sse4(const int16_t* coef, const int32_t* quantCoeff, int16_t* qCoef, int qBits, int add, int numCoeff);
  48 uint32_t x265_nquant_avx2(const int16_t* coef, const int32_t* quantCoeff, int16_t* qCoef, int qBits, int add, int numCoeff);
  49 void x265_dequant_normal_sse4(const int16_t* quantCoef, int16_t* coef, int num, int scale, int shift);
  50 void x265_dequant_normal_avx2(const int16_t* quantCoef, int16_t* coef, int num, int scale, int shift);
  51 int x265_count_nonzero_ssse3(const int16_t* quantCoeff, int numCoeff);
  52
  53 void x265_weight_pp_sse4(const pixel* src, pixel* dst, intptr_t stride, int width, int height, int w0, int round, int shift, int offset);
  54 void x265_weight_pp_avx2(const pixel* src, pixel* dst, intptr_t stride, int width, int height, int w0, int round, int shift, int offset);
  55 void x265_weight_sp_sse4(const int16_t* src, pixel* dst, intptr_t srcStride, intptr_t dstStride, int width, int height, int w0, int round, int shift, int offset);
  56
  57 void x265_pixel_ssim_4x4x2_core_mmx2(const uint8_t* pix1, intptr_t stride1,
  58                                      const uint8_t* pix2, intptr_t stride2, int sums[2][4]);
  59 void x265_pixel_ssim_4x4x2_core_sse2(const pixel* pix1, intptr_t stride1,
  60                                      const pixel* pix2, intptr_t stride2, int sums[2][4]);
  61 void x265_pixel_ssim_4x4x2_core_avx(const pixel* pix1, intptr_t stride1,
  62                                     const pixel* pix2, intptr_t stride2, int sums[2][4]);
  63 float x265_pixel_ssim_end4_sse2(int sum0[5][4], int sum1[5][4], int width);
  64 float x265_pixel_ssim_end4_avx(int sum0[5][4], int sum1[5][4], int width);
  65
  66 void x265_scale1D_128to64_ssse3(pixel*, const pixel*, intptr_t);
  67 void x265_scale1D_128to64_avx2(pixel*, const pixel*, intptr_t);
  68 void x265_scale2D_64to32_ssse3(pixel*, const pixel*, intptr_t);
  69
  70 #define SETUP_CHROMA_PIXELSUB_PS_FUNC(W, H, cpu) \
  71     void x265_pixel_sub_ps_ ## W ## x ## H ## cpu(int16_t*  dest, intptr_t destride, const pixel* src0, const pixel* src1, intptr_t srcstride0, intptr_t srcstride1); \
  72     void x265_pixel_add_ps_ ## W ## x ## H ## cpu(pixel* dest, intptr_t destride, const pixel* src0, const int16_t*  scr1, intptr_t srcStride0, intptr_t srcStride1);
  73
  74 #define CHROMA_PIXELSUB_DEF(cpu) \
  75     SETUP_CHROMA_PIXELSUB_PS_FUNC(4, 4, cpu); \
  76     SETUP_CHROMA_PIXELSUB_PS_FUNC(8, 8, cpu); \
  77     SETUP_CHROMA_PIXELSUB_PS_FUNC(16, 16, cpu); \
  78     SETUP_CHROMA_PIXELSUB_PS_FUNC(32, 32, cpu);
  79
  80 #define CHROMA_PIXELSUB_DEF_422(cpu) \
  81     SETUP_CHROMA_PIXELSUB_PS_FUNC(4, 8, cpu); \
  82     SETUP_CHROMA_PIXELSUB_PS_FUNC(8, 16, cpu); \
  83     SETUP_CHROMA_PIXELSUB_PS_FUNC(16, 32, cpu); \
  84     SETUP_CHROMA_PIXELSUB_PS_FUNC(32, 64, cpu);
  85
  86 #define SETUP_LUMA_PIXELSUB_PS_FUNC(W, H, cpu) \
  87     void x265_pixel_sub_ps_ ## W ## x ## H ## cpu(int16_t*  dest, intptr_t destride, const pixel* src0, const pixel* src1, intptr_t srcstride0, intptr_t srcstride1); \
  88     void x265_pixel_add_ps_ ## W ## x ## H ## cpu(pixel* dest, intptr_t destride, const pixel* src0, const int16_t*  scr1, intptr_t srcStride0, intptr_t srcStride1);
  89
  90 #define LUMA_PIXELSUB_DEF(cpu) \
  91     SETUP_LUMA_PIXELSUB_PS_FUNC(8,   8, cpu); \
  92     SETUP_LUMA_PIXELSUB_PS_FUNC(16, 16, cpu); \
  93     SETUP_LUMA_PIXELSUB_PS_FUNC(32, 32, cpu); \
  94     SETUP_LUMA_PIXELSUB_PS_FUNC(64, 64, cpu);
  95
  96 CHROMA_PIXELSUB_DEF(_sse4);
  97 LUMA_PIXELSUB_DEF(_sse4);
  98 CHROMA_PIXELSUB_DEF(_sse2);
  99 LUMA_PIXELSUB_DEF(_sse2);
 100
 101 CHROMA_PIXELSUB_DEF_422(_sse4);
 102 CHROMA_PIXELSUB_DEF_422(_sse2);
 103
 104 #define SETUP_LUMA_PIXELVAR_FUNC(W, H, cpu) \
 105     uint64_t x265_pixel_var_ ## W ## x ## H ## cpu(const pixel* pix, intptr_t pixstride);
 106
 107 #define LUMA_PIXELVAR_DEF(cpu) \
 108     SETUP_LUMA_PIXELVAR_FUNC(8,   8, cpu); \
 109     SETUP_LUMA_PIXELVAR_FUNC(16, 16, cpu); \
 110     SETUP_LUMA_PIXELVAR_FUNC(32, 32, cpu); \
 111     SETUP_LUMA_PIXELVAR_FUNC(64, 64, cpu);
 112
 113 LUMA_PIXELVAR_DEF(_sse2);
 114
 115 #undef CHROMA_PIXELSUB_DEF
 116 #undef CHROMA_PIXELSUB_DEF_422
 117 #undef LUMA_PIXELSUB_DEF
 118 #undef LUMA_PIXELVAR_DEF
 119 #undef SETUP_CHROMA_PIXELSUB_PS_FUNC
 120 #undef SETUP_LUMA_PIXELSUB_PS_FUNC
 121 #undef SETUP_LUMA_PIXELVAR_FUNC
 122
 123 #endif // ifndef X265_PIXEL_UTIL_H