On Wed, 19 Aug 2026 08:44:44 GMT, Andrew Dinn <[email protected]> wrote:
> > With these changes, @mcpowers has also confirmed improved benchmark results > > on a Neoverse-N1 > > I'm a bit surprised by that. Did we actually generate any different code? > (rather than just make the generator code clearer?). Sorry for the confusion, but what I meant by "With these changes..." is "With the original set of changes...". Previously I had only tested the original change-set on an M3 (with the -UseSIMDForSHA3Intrinsic option) and wanted to confirm similar performance results on a Neoverse-N1 system, in which we did. ------------- PR Comment: https://git.openjdk.org/jdk/pull/32049#issuecomment-5349193379
