On Tue, 25 Feb 2025 00:00:23 GMT, Paul Sandoz <psan...@openjdk.org> wrote:
>> Thanks for pointing that out. Typically, ARRAYLEN is almost always a POT >> value, which is also assumed by many other benchmarks. Are we realistically >> going to test with an ARRAYLEN of 30? >> >> I think the POT assumption is reasonable for our purposes. > > It's a reasonable assumption. Since `ARRAYLEN` is a parameter of the > benchmark we should enforce that constraint in benchmark initialization > method, checking if the value is POT and failing otherwise. Hi, @PaulSandoz, thanks for your suggestions. In @jatin-bhateja's latest updates, non-POT `ARRAYLEN` values have been supported by processing only the elements up to the nearest multiple of the vector length. In that case, we can use a wider range of `ARRAYLEN` values while focusing on the main loop for vector functionality. ------------- PR Review Comment: https://git.openjdk.org/jdk/pull/22963#discussion_r1971372152
