colorspacedsp_init.c 4.4 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119
  1. /*
  2. * Copyright (c) 2016 Ronald S. Bultje <rsbultje@gmail.com>
  3. *
  4. * This file is part of FFmpeg.
  5. *
  6. * FFmpeg is free software; you can redistribute it and/or
  7. * modify it under the terms of the GNU Lesser General Public
  8. * License as published by the Free Software Foundation; either
  9. * version 2.1 of the License, or (at your option) any later version.
  10. *
  11. * FFmpeg 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 GNU
  14. * Lesser General Public License for more details.
  15. *
  16. * You should have received a copy of the GNU Lesser General Public
  17. * License along with FFmpeg; if not, write to the Free Software
  18. * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  19. */
  20. #include "libavutil/x86/cpu.h"
  21. #include "libavfilter/colorspacedsp.h"
  22. #define decl_yuv2yuv_fn(t) \
  23. void ff_yuv2yuv_##t##_sse2(uint8_t *yuv_out[3], const ptrdiff_t yuv_out_stride[3], \
  24. uint8_t *yuv_in[3], const ptrdiff_t yuv_in_stride[3], \
  25. int w, int h, const int16_t yuv2yuv_coeffs[3][3][8], \
  26. const int16_t yuv_offset[2][8])
  27. #define decl_yuv2yuv_fns(ss) \
  28. decl_yuv2yuv_fn(ss##p8to8); \
  29. decl_yuv2yuv_fn(ss##p10to8); \
  30. decl_yuv2yuv_fn(ss##p12to8); \
  31. decl_yuv2yuv_fn(ss##p8to10); \
  32. decl_yuv2yuv_fn(ss##p10to10); \
  33. decl_yuv2yuv_fn(ss##p12to10); \
  34. decl_yuv2yuv_fn(ss##p8to12); \
  35. decl_yuv2yuv_fn(ss##p10to12); \
  36. decl_yuv2yuv_fn(ss##p12to12)
  37. decl_yuv2yuv_fns(420);
  38. decl_yuv2yuv_fns(422);
  39. decl_yuv2yuv_fns(444);
  40. #define decl_yuv2rgb_fn(t) \
  41. void ff_yuv2rgb_##t##_sse2(int16_t *rgb_out[3], ptrdiff_t rgb_stride, \
  42. uint8_t *yuv_in[3], const ptrdiff_t yuv_stride[3], \
  43. int w, int h, const int16_t coeff[3][3][8], \
  44. const int16_t yuv_offset[8])
  45. #define decl_yuv2rgb_fns(ss) \
  46. decl_yuv2rgb_fn(ss##p8); \
  47. decl_yuv2rgb_fn(ss##p10); \
  48. decl_yuv2rgb_fn(ss##p12)
  49. decl_yuv2rgb_fns(420);
  50. decl_yuv2rgb_fns(422);
  51. decl_yuv2rgb_fns(444);
  52. #define decl_rgb2yuv_fn(t) \
  53. void ff_rgb2yuv_##t##_sse2(uint8_t *yuv_out[3], const ptrdiff_t yuv_stride[3], \
  54. int16_t *rgb_in[3], ptrdiff_t rgb_stride, \
  55. int w, int h, const int16_t coeff[3][3][8], \
  56. const int16_t yuv_offset[8])
  57. #define decl_rgb2yuv_fns(ss) \
  58. decl_rgb2yuv_fn(ss##p8); \
  59. decl_rgb2yuv_fn(ss##p10); \
  60. decl_rgb2yuv_fn(ss##p12)
  61. decl_rgb2yuv_fns(420);
  62. decl_rgb2yuv_fns(422);
  63. decl_rgb2yuv_fns(444);
  64. void ff_multiply3x3_sse2(int16_t *data[3], ptrdiff_t stride, int w, int h,
  65. const int16_t coeff[3][3][8]);
  66. void ff_colorspacedsp_x86_init(ColorSpaceDSPContext *dsp)
  67. {
  68. int cpu_flags = av_get_cpu_flags();
  69. if (ARCH_X86_64 && EXTERNAL_SSE2(cpu_flags)) {
  70. #define assign_yuv2yuv_fns(ss) \
  71. dsp->yuv2yuv[BPP_8 ][BPP_8 ][SS_##ss] = ff_yuv2yuv_##ss##p8to8_sse2; \
  72. dsp->yuv2yuv[BPP_8 ][BPP_10][SS_##ss] = ff_yuv2yuv_##ss##p8to10_sse2; \
  73. dsp->yuv2yuv[BPP_8 ][BPP_12][SS_##ss] = ff_yuv2yuv_##ss##p8to12_sse2; \
  74. dsp->yuv2yuv[BPP_10][BPP_8 ][SS_##ss] = ff_yuv2yuv_##ss##p10to8_sse2; \
  75. dsp->yuv2yuv[BPP_10][BPP_10][SS_##ss] = ff_yuv2yuv_##ss##p10to10_sse2; \
  76. dsp->yuv2yuv[BPP_10][BPP_12][SS_##ss] = ff_yuv2yuv_##ss##p10to12_sse2; \
  77. dsp->yuv2yuv[BPP_12][BPP_8 ][SS_##ss] = ff_yuv2yuv_##ss##p12to8_sse2; \
  78. dsp->yuv2yuv[BPP_12][BPP_10][SS_##ss] = ff_yuv2yuv_##ss##p12to10_sse2; \
  79. dsp->yuv2yuv[BPP_12][BPP_12][SS_##ss] = ff_yuv2yuv_##ss##p12to12_sse2
  80. assign_yuv2yuv_fns(420);
  81. assign_yuv2yuv_fns(422);
  82. assign_yuv2yuv_fns(444);
  83. #define assign_yuv2rgb_fns(ss) \
  84. dsp->yuv2rgb[BPP_8 ][SS_##ss] = ff_yuv2rgb_##ss##p8_sse2; \
  85. dsp->yuv2rgb[BPP_10][SS_##ss] = ff_yuv2rgb_##ss##p10_sse2; \
  86. dsp->yuv2rgb[BPP_12][SS_##ss] = ff_yuv2rgb_##ss##p12_sse2
  87. assign_yuv2rgb_fns(420);
  88. assign_yuv2rgb_fns(422);
  89. assign_yuv2rgb_fns(444);
  90. #define assign_rgb2yuv_fns(ss) \
  91. dsp->rgb2yuv[BPP_8 ][SS_##ss] = ff_rgb2yuv_##ss##p8_sse2; \
  92. dsp->rgb2yuv[BPP_10][SS_##ss] = ff_rgb2yuv_##ss##p10_sse2; \
  93. dsp->rgb2yuv[BPP_12][SS_##ss] = ff_rgb2yuv_##ss##p12_sse2
  94. assign_rgb2yuv_fns(420);
  95. assign_rgb2yuv_fns(422);
  96. assign_rgb2yuv_fns(444);
  97. dsp->multiply3x3 = ff_multiply3x3_sse2;
  98. }
  99. }