Motivation: the getter/setter route seems to be significantly slower [citation needed], so need some baseline benchmarks for comparison and to justify unrolling this stuff for normal ndarrays.
It will prove very difficult to optimize this for general ndarrays with no guarantee of row-major vs. column-major, but we can try at least.
Motivation: the getter/setter route seems to be significantly slower
[citation needed], so need some baseline benchmarks for comparison and to justify unrolling this stuff for normal ndarrays.It will prove very difficult to optimize this for general ndarrays with no guarantee of row-major vs. column-major, but we can try at least.