4 ms·
> for a problem that doesn't immediately jump out as SIMD. It's a single transformation applied to elements of a large array. Parallelization is obvious, and m
by einpoklum 5y ago
> for a problem that doesn't immediately jump out as SIMD.
It's a single transformation applied to elements of a large array. Parallelization is obvious, and maybe the exact SIMDization is not obvious, one is certainly motivated to formulate it. And a scatter-like approach does spring to mind right away, I believe.