src/liboggplay/x86/oggplay_yuv2rgb_x86.c

   1 /*
   2    Copyright (C) 2003 Commonwealth Scientific and Industrial Research
   3    Organisation (CSIRO) Australia
   4
   5    Redistribution and use in source and binary forms, with or without
   6    modification, are permitted provided that the following conditions
   7    are met:
   8
   9    - Redistributions of source code must retain the above copyright
  10    notice, this list of conditions and the following disclaimer.
  11
  12    - Redistributions in binary form must reproduce the above copyright
  13    notice, this list of conditions and the following disclaimer in the
  14    documentation and/or other materials provided with the distribution.
  15
  16    - Neither the name of CSIRO Australia nor the names of its
  17    contributors may be used to endorse or promote products derived from
  18    this software without specific prior written permission.
  19
  20    THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  21    ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  22    LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
  23    PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE ORGANISATION OR
  24    CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
  25    EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
  26    PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
  27    PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  28    LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  29    NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  30    SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  31 */
  32
  33 /**
  34  * YUV to RGB conversion using x86 CPU extensions
  35  */
  36 #include "oggplay_private.h"
  37 #include "oggplay_yuv2rgb_template.h"
  38 #include "cpu.h"
  39
  40 #if defined(_MSC_VER)
  41 # if defined(_M_AMD64)
  42   /* MSVC with x64 doesn't support inline assembler */
  43 #include <emmintrin.h>
  44 # endif
  45 #include "yuv2rgb_x86_vs.h"
  46 #elif defined(__GNUC__)
  47 #include "yuv2rgb_x86.h"
  48 #endif
  49
  50 typedef union
  51 {
  52         long long               q[2];
  53         unsigned long long      uq[2];
  54         int                     d[4];
  55         unsigned int            ud[4];
  56         short                   w[8];
  57         unsigned short          uw[8];
  58         char                    b[16];
  59         unsigned char           ub[16];
  60         float                   s[4];
  61 } ATTR_ALIGN(16) simd_t;
  62
  63 #define UV_128 0x0080008000800080LL
  64 #define Y_16 0x1010101010101010LL
  65 #define Y_Co 0x253f253f253f253fLL
  66 #define GU_Co 0xf37df37df37df37dLL
  67 #define GV_Co 0xe5fce5fce5fce5fcLL
  68 #define BU_Co 0x4093409340934093LL
  69 #define RV_Co 0x3312331233123312LL
  70 #define Y_MASK 0x00ff00ff00ff00ffLL
  71 #define ALFA 0xffffffffffffffffLL
  72
  73 /**
  74  * coefficients and constants for yuv to rgb SIMD conversion
  75  */
  76 static const simd_t simd_table[9] = {
  77         {{UV_128, UV_128}},
  78         {{GU_Co, GU_Co}},
  79         {{GV_Co, GV_Co}},
  80         {{BU_Co, BU_Co}},
  81         {{RV_Co, RV_Co}},
  82         {{Y_16, Y_16}},
  83         {{Y_Co, Y_Co}},
  84         {{Y_MASK, Y_MASK}},
  85         {{ALFA, ALFA}}
  86 };
  87
  88 /* MMX intristics are not supported by VS in x64 mode, thus disable it */
  89 #if !(defined(_MSC_VER) && defined(_M_AMD64))
  90 /**
  91  *  the conversion functions using MMX instructions
  92  */
  93
  94 /* template for the MMX conversion functions */
  95 #define YUV_CONVERT_MMX(FUNC, CONVERT, CONV_BY_PIXEL, UV_SHIFT, UV_VERT_SUB) YUV_CONVERT(FUNC, CONVERT, CONV_BY_PIXEL, 8, 32, 8, UV_SHIFT, UV_VERT_SUB)
  96
  97 #define CLEANUP emms()
  98 #define OUT_RGBA_32 OUTPUT_RGBA_32(movq, mm, 8, 16, 24)
  99 #define OUT_ARGB_32 OUTPUT_ARGB_32(movq, mm, 8, 16, 24)
 100 #define OUT_BGRA_32 OUTPUT_BGRA_32(movq, mm, 8, 16, 24)
 101 #define MOVNTQ MMX_MOVNTQ
 102
 103 /* yuv420, yuv422 */
 104 #define CONVERT(OUTPUT_FUNC) LOAD_YUV(movq, movd, mm) \
 105                              YUV_2_RGB(movq, mm)      \
 106                              OUTPUT_FUNC
 107
 108 /* conversion function for yuv444 */
 109 #define CONVERT_444(OUTPUT_FUNC) LOAD_YUV444(movq, mm)  \
 110                                  YUV444_2_RGB(movq, mm) \
 111                                  OUTPUT_FUNC
 112
 113 /* MMX yuv420 functions */
 114 YUV_CONVERT_MMX(yuv420_to_rgba_mmx, CONVERT(OUT_RGBA_32), VANILLA_RGBA_OUT, 4, 2)
 115 YUV_CONVERT_MMX(yuv420_to_bgra_mmx, CONVERT(OUT_BGRA_32), VANILLA_BGRA_OUT, 4, 2)
 116 YUV_CONVERT_MMX(yuv420_to_argb_mmx, CONVERT(OUT_ARGB_32), VANILLA_ARGB_OUT, 4, 2)
 117
 118 /* MMX yuv422 functions */
 119 YUV_CONVERT_MMX(yuv422_to_rgba_mmx, CONVERT(OUT_RGBA_32), VANILLA_RGBA_OUT, 4, 1)
 120 YUV_CONVERT_MMX(yuv422_to_bgra_mmx, CONVERT(OUT_BGRA_32), VANILLA_BGRA_OUT, 4, 1)
 121 YUV_CONVERT_MMX(yuv422_to_argb_mmx, CONVERT(OUT_ARGB_32), VANILLA_ARGB_OUT, 4, 1)
 122
 123 /* MMX yuv444 functions */
 124 YUV_CONVERT_MMX(yuv444_to_rgba_mmx, CONVERT_444(OUT_RGBA_32), VANILLA_RGBA_OUT, 8, 1)
 125 YUV_CONVERT_MMX(yuv444_to_bgra_mmx, CONVERT_444(OUT_BGRA_32), VANILLA_BGRA_OUT, 8, 1)
 126 YUV_CONVERT_MMX(yuv444_to_argb_mmx, CONVERT_444(OUT_ARGB_32), VANILLA_ARGB_OUT, 8, 1)
 127
 128 #undef MOVNTQ
 129
 130 /* template for the SSE conversion functions */
 131 #define MOVNTQ SSE_MOVNTQ
 132
 133 /* SSE yuv420 functions */
 134 YUV_CONVERT_MMX(yuv420_to_rgba_sse, CONVERT(OUT_RGBA_32), VANILLA_RGBA_OUT, 4, 2)
 135 YUV_CONVERT_MMX(yuv420_to_bgra_sse, CONVERT(OUT_BGRA_32), VANILLA_BGRA_OUT, 4, 2)
 136 YUV_CONVERT_MMX(yuv420_to_argb_sse, CONVERT(OUT_ARGB_32), VANILLA_ARGB_OUT, 4, 2)
 137
 138 /* SSE yuv422 functions */
 139 YUV_CONVERT_MMX(yuv422_to_rgba_sse, CONVERT(OUT_RGBA_32), VANILLA_RGBA_OUT, 4, 1)
 140 YUV_CONVERT_MMX(yuv422_to_bgra_sse, CONVERT(OUT_BGRA_32), VANILLA_BGRA_OUT, 4, 1)
 141 YUV_CONVERT_MMX(yuv422_to_argb_sse, CONVERT(OUT_ARGB_32), VANILLA_ARGB_OUT, 4, 1)
 142
 143 /* SSE yuv444 functions */
 144 YUV_CONVERT_MMX(yuv444_to_rgba_sse, CONVERT_444(OUT_RGBA_32), VANILLA_RGBA_OUT, 8, 1)
 145 YUV_CONVERT_MMX(yuv444_to_bgra_sse, CONVERT_444(OUT_BGRA_32), VANILLA_BGRA_OUT, 8, 1)
 146 YUV_CONVERT_MMX(yuv444_to_argb_sse, CONVERT_444(OUT_ARGB_32), VANILLA_ARGB_OUT, 8, 1)
 147
 148 #undef CONVERT
 149 #undef CONVERT_444
 150 #undef CLEANUP
 151 #undef OUT_RGBA_32
 152 #undef OUT_ARGB_32
 153 #undef OUT_BGRA_32
 154 #undef MOVNTQ
 155 #endif
 156
 157
 158 /**
 159  *  the conversion functions using SSE2 instructions
 160  */
 161
 162 /* template for the SSE2 conversion functions */
 163 #define YUV_CONVERT_SSE2(FUNC, CONVERT, CONV_BY_PIX, UV_SHIFT, UV_VERT_SUB) YUV_CONVERT(FUNC, CONVERT, CONV_BY_PIX, 16, 64, 16, UV_SHIFT, UV_VERT_SUB)
 164
 165 /* Output functions */
 166 #define OUT_RGBA_32 OUTPUT_RGBA_32(movdqa, xmm, 16, 32, 48)
 167 #define OUT_ARGB_32 OUTPUT_ARGB_32(movdqa, xmm, 16, 32, 48)
 168 #define OUT_BGRA_32 OUTPUT_BGRA_32(movdqa, xmm, 16, 32, 48)
 169 #define MOVNTQ SSE2_MOVNTQ
 170 #define CLEANUP
 171
 172 #if defined(_MSC_VER) && defined(_M_AMD64)
 173 /* yuv420, yuv422 conversion function */
 174 #define CONVERT(OUTPUT_FUNC) __m128i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7; \
 175                                 LOAD_YUV_PLANAR_2(movdqu, xmm) \
 176                                 YUV_2_RGB(movdqa, xmm)  \
 177                                 OUTPUT_FUNC
 178 #else
 179 /* yuv420, yuv422 conversion function */
 180 #define CONVERT(OUTPUT_FUNC) LOAD_YUV(movdqu, movq, xmm)  \
 181                              YUV_2_RGB(movdqa, xmm)       \
 182                                                      OUTPUT_FUNC
 183
 184 /* conversion function for yuv444 */
 185 #define CONVERT_444(OUTPUT_FUNC) LOAD_YUV444(movdqu, xmm)   \
 186                                  YUV444_2_RGB(movdqa, xmm)  \
 187                                  OUTPUT_FUNC
 188 #endif
 189
 190 /* SSE2 yuv420 functions */
 191 YUV_CONVERT_SSE2(yuv420_to_rgba_sse2, CONVERT(OUT_RGBA_32), VANILLA_RGBA_OUT, 8, 2)
 192 YUV_CONVERT_SSE2(yuv420_to_bgra_sse2, CONVERT(OUT_BGRA_32), VANILLA_BGRA_OUT, 8, 2)
 193 YUV_CONVERT_SSE2(yuv420_to_argb_sse2, CONVERT(OUT_ARGB_32), VANILLA_ARGB_OUT, 8, 2)
 194
 195 /* SSE2 yuv422 functions */
 196 YUV_CONVERT_SSE2(yuv422_to_rgba_sse2, CONVERT(OUT_RGBA_32), VANILLA_RGBA_OUT, 8, 1)
 197 YUV_CONVERT_SSE2(yuv422_to_bgra_sse2, CONVERT(OUT_BGRA_32), VANILLA_BGRA_OUT, 8, 1)
 198 YUV_CONVERT_SSE2(yuv422_to_argb_sse2, CONVERT(OUT_ARGB_32), VANILLA_ARGB_OUT, 8, 1)
 199
 200 /* SSE2 yuv444 functions */
 201 YUV_CONVERT_SSE2(yuv444_to_rgba_sse2, CONVERT_444(OUT_RGBA_32), VANILLA_RGBA_OUT, 16, 1)
 202 YUV_CONVERT_SSE2(yuv444_to_bgra_sse2, CONVERT_444(OUT_BGRA_32), VANILLA_BGRA_OUT, 16, 1)
 203 YUV_CONVERT_SSE2(yuv444_to_argb_sse2, CONVERT_444(OUT_ARGB_32), VANILLA_ARGB_OUT, 16, 1)
 204
 205 #undef CONVERT
 206 #undef CONVERT_444
 207 #undef OUT_RGBA_32
 208 #undef OUT_ARGB_32
 209 #undef OUT_BGRA_32
 210 #undef MOVNTQ
 211 #undef CLEANUP
 212