On Thu, 12 Sep 2024 11:34:44 GMT, Maurizio Cimadamore <mcimadam...@openjdk.org> wrote:
>> In other words, I don't think the goal of this (and related) PR is "improve >> mismatch so that it blows other alternatives - like Unsafe, or array" out of >> the water - as much as it is "make sure that using MemorySegment::mismatch >> is competitive with other offerings". > > Then, an interesting part of these PRs is that we have uncovered quite a lot > of issues both with our intrinsics (e.g. `fill` is not vectorized and works > badly on Windows, mismatch works poorly on aarch64) *and* missed optimization > opportunities - e.g. "short" segment loops are not optimized as tightly as > they could. But it is not the job of these PRs to fix all these issues. The > main focus remain: for small sizes it is not worth going down > intrinsics-lane. Let's stick to it (in the interest of keeping the review > focused). Yepp. So, let us keep these tricks up our sleeves and then maybe come back with a new PR. ------------- PR Review Comment: https://git.openjdk.org/jdk/pull/20848#discussion_r1756793795
