cpu_avx512_knm.c 1.1 KB

123456789101112131415161718192021222324252627282930
  1. #if defined(DETECT_FEATURES) && defined(__INTEL_COMPILER)
  2. /*
  3. * Unlike GCC and CLANG, Intel Compiler exposes all supported intrinsics,
  4. * whether or not the build options for those features are specified.
  5. * Therefore, we must test #definitions of CPU features when option native/host
  6. * is enabled via `--cpu-baseline` or through env var `CFLAGS` otherwise
  7. * the test will be broken and leads to enable all possible features.
  8. */
  9. #if !defined(__AVX5124FMAPS__) || !defined(__AVX5124VNNIW__) || !defined(__AVX512VPOPCNTDQ__)
  10. #error "HOST/ARCH doesn't support Knights Mill AVX512 features"
  11. #endif
  12. #endif
  13. #include <immintrin.h>
  14. int main(int argc, char **argv)
  15. {
  16. __m512i a = _mm512_loadu_si512((const __m512i*)argv[argc-1]);
  17. __m512 b = _mm512_loadu_ps((const __m512*)argv[argc-2]);
  18. /* 4FMAPS */
  19. b = _mm512_4fmadd_ps(b, b, b, b, b, NULL);
  20. /* 4VNNIW */
  21. a = _mm512_4dpwssd_epi32(a, a, a, a, a, NULL);
  22. /* VPOPCNTDQ */
  23. a = _mm512_popcnt_epi64(a);
  24. a = _mm512_add_epi32(a, _mm512_castps_si512(b));
  25. return _mm_cvtsi128_si32(_mm512_castsi512_si128(a));
  26. }