You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

109 lines
4.5KB

  1. /*
  2. * BlackFin DSPUTILS COMMON OPTIMIZATIONS HEADER
  3. *
  4. * Copyright (C) 2007 Marc Hoffman <mmh@pleasantst.com>
  5. *
  6. * This file is part of Libav.
  7. *
  8. * Libav is free software; you can redistribute it and/or
  9. * modify it under the terms of the GNU Lesser General Public
  10. * License as published by the Free Software Foundation; either
  11. * version 2.1 of the License, or (at your option) any later version.
  12. *
  13. * Libav is distributed in the hope that it will be useful,
  14. * but WITHOUT ANY WARRANTY; without even the implied warranty of
  15. * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  16. * Lesser General Public License for more details.
  17. *
  18. * You should have received a copy of the GNU Lesser General Public
  19. * License along with Libav; if not, write to the Free Software
  20. * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  21. */
  22. #ifndef AVCODEC_BFIN_DSPUTIL_BFIN_H
  23. #define AVCODEC_BFIN_DSPUTIL_BFIN_H
  24. #include "config.h"
  25. #include "libavcodec/dsputil.h"
  26. #if defined(__FDPIC__) && CONFIG_SRAM
  27. #define attribute_l1_text __attribute__ ((l1_text))
  28. #define attribute_l1_data_b __attribute__((l1_data_B))
  29. #else
  30. #define attribute_l1_text
  31. #define attribute_l1_data_b
  32. #endif
  33. void ff_bfin_idct (DCTELEM *block) attribute_l1_text;
  34. void ff_bfin_fdct (DCTELEM *block) attribute_l1_text;
  35. void ff_bfin_vp3_idct (DCTELEM *block);
  36. void ff_bfin_vp3_idct_put (uint8_t *dest, int line_size, DCTELEM *block);
  37. void ff_bfin_vp3_idct_add (uint8_t *dest, int line_size, DCTELEM *block);
  38. void ff_bfin_add_pixels_clamped (const DCTELEM *block, uint8_t *dest, int line_size) attribute_l1_text;
  39. void ff_bfin_put_pixels_clamped (const DCTELEM *block, uint8_t *dest, int line_size) attribute_l1_text;
  40. void ff_bfin_diff_pixels (DCTELEM *block, const uint8_t *s1, const uint8_t *s2, int stride) attribute_l1_text;
  41. void ff_bfin_get_pixels (DCTELEM *restrict block, const uint8_t *pixels, int line_size) attribute_l1_text;
  42. int ff_bfin_pix_norm1 (uint8_t * pix, int line_size) attribute_l1_text;
  43. int ff_bfin_z_sad8x8 (uint8_t *blk1, uint8_t *blk2, int dsz, int line_size, int h) attribute_l1_text;
  44. int ff_bfin_z_sad16x16 (uint8_t *blk1, uint8_t *blk2, int dsz, int line_size, int h) attribute_l1_text;
  45. void ff_bfin_z_put_pixels16_xy2 (uint8_t *block, const uint8_t *s0, int dest_size, int line_size, int h) attribute_l1_text;
  46. void ff_bfin_z_put_pixels8_xy2 (uint8_t *block, const uint8_t *s0, int dest_size, int line_size, int h) attribute_l1_text;
  47. void ff_bfin_put_pixels16_xy2_nornd (uint8_t *block, const uint8_t *s0, int line_size, int h) attribute_l1_text;
  48. void ff_bfin_put_pixels8_xy2_nornd (uint8_t *block, const uint8_t *s0, int line_size, int h) attribute_l1_text;
  49. int ff_bfin_pix_sum (uint8_t *p, int stride) attribute_l1_text;
  50. void ff_bfin_put_pixels8uc (uint8_t *block, const uint8_t *s0, const uint8_t *s1, int dest_size, int line_size, int h) attribute_l1_text;
  51. void ff_bfin_put_pixels16uc (uint8_t *block, const uint8_t *s0, const uint8_t *s1, int dest_size, int line_size, int h) attribute_l1_text;
  52. void ff_bfin_put_pixels8uc_nornd (uint8_t *block, const uint8_t *s0, const uint8_t *s1, int line_size, int h) attribute_l1_text;
  53. void ff_bfin_put_pixels16uc_nornd (uint8_t *block, const uint8_t *s0, const uint8_t *s1, int line_size, int h) attribute_l1_text;
  54. int ff_bfin_sse4 (void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h) attribute_l1_text;
  55. int ff_bfin_sse8 (void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h) attribute_l1_text;
  56. int ff_bfin_sse16 (void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h) attribute_l1_text;
  57. #ifdef BFIN_PROFILE
  58. static double Telem[16];
  59. static char *TelemNames[16];
  60. static int TelemCnt;
  61. #define PROF(lab,e) { int xx_e = e; char*xx_lab = lab; uint64_t xx_t0 = read_time();
  62. #define EPROF() xx_t0 = read_time()-xx_t0; Telem[xx_e] = Telem[xx_e] + xx_t0; TelemNames[xx_e] = xx_lab; }
  63. static void prof_report (void)
  64. {
  65. int i;
  66. double s = 0;
  67. for (i=0;i<16;i++) {
  68. double v;
  69. if (TelemNames[i]) {
  70. v = Telem[i]/TelemCnt;
  71. av_log (NULL,AV_LOG_DEBUG,"%-20s: %12.4f\t%12.4f\n", TelemNames[i],v,v/64);
  72. s = s + Telem[i];
  73. }
  74. }
  75. av_log (NULL,AV_LOG_DEBUG,"%-20s: %12.4f\t%12.4f\n%20.4f\t%d\n",
  76. "total",s/TelemCnt,s/TelemCnt/64,s,TelemCnt);
  77. }
  78. static void bfprof (void)
  79. {
  80. static int init;
  81. if (!init) atexit (prof_report);
  82. init=1;
  83. TelemCnt++;
  84. }
  85. #else
  86. #define PROF(a,b)
  87. #define EPROF()
  88. #define bfprof()
  89. #endif
  90. #endif /* AVCODEC_BFIN_DSPUTIL_BFIN_H */