Skip to content

Instantly share code, notes, and snippets.

@andersx

andersx/fastexp.h

Created February 16, 2015 01:27
Show Gist options
  • Save andersx/8057b2a6fd3d715d35eb to your computer and use it in GitHub Desktop.
Save andersx/8057b2a6fd3d715d35eb to your computer and use it in GitHub Desktop.
Download ZIP
Very fast EXP(x) for AVX2+FMA instructions
Raw
fastexp.h
// Approximation for EXP(x) -- very fast, but not super accurate
static inline __m256 _mm256_expfaster_ps(const __m256 &q) {
const __m256 C1 = _mm256_set1_ps(1064872507.1541044f);
const __m256 C2 = _mm256_set1_ps(12102203.161561485f);
return _mm256_castsi256_ps(_mm256_cvttps_epi32(_mm256_fmadd_ps(C2, q, C1)));
}
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment