3 ms·
That's disappointing. It may be a bad interaction between OpenMP and AVX512 (cache pressure etc)? I've also seen reliable increases in performance up through A
by marmaduke 9y ago
That's disappointing. It may be a bad interaction between OpenMP and AVX512 (cache pressure etc)?
I've also seen reliable increases in performance up through AVX2 but when I tried to run same code on a Xeon Phi, it fell short of the plain Xeon.