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.

144 lines
6.0KB

  1. /*
  2. * Assembly testing and benchmarking tool
  3. * Copyright (c) 2015 Henrik Gramner
  4. * Copyright (c) 2008 Loren Merritt
  5. *
  6. * This file is part of FFmpeg.
  7. *
  8. * FFmpeg is free software; you can redistribute it and/or modify
  9. * it under the terms of the GNU General Public License as published by
  10. * the Free Software Foundation; either version 2 of the License, or
  11. * (at your option) any later version.
  12. *
  13. * FFmpeg 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
  16. * GNU General Public License for more details.
  17. *
  18. * You should have received a copy of the GNU General Public License along
  19. * with FFmpeg; if not, write to the Free Software Foundation, Inc.,
  20. * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
  21. */
  22. #ifndef CHECKASM_H
  23. #define CHECKASM_H
  24. #include <stdint.h>
  25. #include "config.h"
  26. #include "libavutil/avstring.h"
  27. #include "libavutil/cpu.h"
  28. #include "libavutil/lfg.h"
  29. #include "libavutil/timer.h"
  30. void checkasm_check_alacdsp(void);
  31. void checkasm_check_bswapdsp(void);
  32. void checkasm_check_flacdsp(void);
  33. void checkasm_check_h264pred(void);
  34. void checkasm_check_h264qpel(void);
  35. void checkasm_check_jpeg2000dsp(void);
  36. void checkasm_check_pixblockdsp(void);
  37. void checkasm_check_v210enc(void);
  38. void checkasm_check_vp9dsp(void);
  39. void *checkasm_check_func(void *func, const char *name, ...) av_printf_format(2, 3);
  40. int checkasm_bench_func(void);
  41. void checkasm_fail_func(const char *msg, ...) av_printf_format(1, 2);
  42. void checkasm_update_bench(int iterations, uint64_t cycles);
  43. void checkasm_report(const char *name, ...) av_printf_format(1, 2);
  44. extern AVLFG checkasm_lfg;
  45. #define rnd() av_lfg_get(&checkasm_lfg)
  46. static av_unused void *func_ref, *func_new;
  47. #define BENCH_RUNS 1000 /* Trade-off between accuracy and speed */
  48. /* Decide whether or not the specified function needs to be tested */
  49. #define check_func(func, ...) (func_ref = checkasm_check_func((func_new = func), __VA_ARGS__))
  50. /* Declare the function prototype. The first argument is the return value, the remaining
  51. * arguments are the function parameters. Naming parameters is optional. */
  52. #define declare_func(ret, ...) declare_new(ret, __VA_ARGS__) typedef ret func_type(__VA_ARGS__)
  53. #define declare_func_emms(cpu_flags, ret, ...) declare_new_emms(cpu_flags, ret, __VA_ARGS__) typedef ret func_type(__VA_ARGS__)
  54. /* Indicate that the current test has failed */
  55. #define fail() checkasm_fail_func("%s:%d", av_basename(__FILE__), __LINE__)
  56. /* Print the test outcome */
  57. #define report checkasm_report
  58. /* Call the reference function */
  59. #define call_ref(...) ((func_type *)func_ref)(__VA_ARGS__)
  60. #if ARCH_X86 && HAVE_YASM
  61. /* Verifies that clobbered callee-saved registers are properly saved and restored
  62. * and that either no MMX registers are touched or emms is issued */
  63. void checkasm_checked_call(void *func, ...);
  64. /* Verifies that clobbered callee-saved registers are properly saved and restored
  65. * and issues emms for asm functions which are not required to do so */
  66. void checkasm_checked_call_emms(void *func, ...);
  67. #if ARCH_X86_64
  68. /* Evil hack: detect incorrect assumptions that 32-bit ints are zero-extended to 64-bit.
  69. * This is done by clobbering the stack with junk around the stack pointer and calling the
  70. * assembly function through checked_call() with added dummy arguments which forces all
  71. * real arguments to be passed on the stack and not in registers. For 32-bit arguments the
  72. * upper half of the 64-bit register locations on the stack will now contain junk which will
  73. * cause misbehaving functions to either produce incorrect output or segfault. Note that
  74. * even though this works extremely well in practice, it's technically not guaranteed
  75. * and false negatives is theoretically possible, but there can never be any false positives.
  76. */
  77. void checkasm_stack_clobber(uint64_t clobber, ...);
  78. #define declare_new(ret, ...) ret (*checked_call)(void *, int, int, int, int, int, __VA_ARGS__)\
  79. = (void *)checkasm_checked_call;
  80. #define declare_new_emms(cpu_flags, ret, ...) \
  81. ret (*checked_call)(void *, int, int, int, int, int, __VA_ARGS__) = \
  82. ((cpu_flags) & av_get_cpu_flags()) ? (void *)checkasm_checked_call_emms : \
  83. (void *)checkasm_checked_call;
  84. #define CLOB (UINT64_C(0xdeadbeefdeadbeef))
  85. #define call_new(...) (checkasm_stack_clobber(CLOB,CLOB,CLOB,CLOB,CLOB,CLOB,CLOB,CLOB,CLOB,CLOB,CLOB,\
  86. CLOB,CLOB,CLOB,CLOB,CLOB,CLOB,CLOB,CLOB,CLOB,CLOB),\
  87. checked_call(func_new, 0, 0, 0, 0, 0, __VA_ARGS__))
  88. #elif ARCH_X86_32
  89. #define declare_new(ret, ...) ret (*checked_call)(void *, __VA_ARGS__) = (void *)checkasm_checked_call;
  90. #define declare_new_emms(cpu_flags, ret, ...) ret (*checked_call)(void *, __VA_ARGS__) = \
  91. ((cpu_flags) & av_get_cpu_flags()) ? (void *)checkasm_checked_call_emms : \
  92. (void *)checkasm_checked_call;
  93. #define call_new(...) checked_call(func_new, __VA_ARGS__)
  94. #endif
  95. #else
  96. #define declare_new(ret, ...)
  97. #define declare_new_emms(cpu_flags, ret, ...)
  98. /* Call the function */
  99. #define call_new(...) ((func_type *)func_new)(__VA_ARGS__)
  100. #endif
  101. /* Benchmark the function */
  102. #ifdef AV_READ_TIME
  103. #define bench_new(...)\
  104. do {\
  105. if (checkasm_bench_func()) {\
  106. func_type *tfunc = func_new;\
  107. uint64_t tsum = 0;\
  108. int ti, tcount = 0;\
  109. for (ti = 0; ti < BENCH_RUNS; ti++) {\
  110. uint64_t t = AV_READ_TIME();\
  111. tfunc(__VA_ARGS__);\
  112. tfunc(__VA_ARGS__);\
  113. tfunc(__VA_ARGS__);\
  114. tfunc(__VA_ARGS__);\
  115. t = AV_READ_TIME() - t;\
  116. if (t*tcount <= tsum*4 && ti > 0) {\
  117. tsum += t;\
  118. tcount++;\
  119. }\
  120. }\
  121. checkasm_update_bench(tcount, tsum);\
  122. }\
  123. } while (0)
  124. #else
  125. #define bench_new(...) while(0)
  126. #endif
  127. #endif