3 ms·
The title gives the impression that handwritten assembly code is the cause of this performance improvement, when in fact it was use of AVX-512 SIMD instructions
by HarHarVeryFunny 2y ago
The title gives the impression that handwritten assembly code is the cause of this performance improvement, when in fact it was use of AVX-512 SIMD instructions that really made the difference.
I've got to wonder how gcc's AVX-512 vectorization would compare ?