On Wed, 27 May 2026 17:50:12 GMT, Jatin Bhateja <[email protected]> wrote:
>>> @PaulSandoz @jatin-bhateja I saw that there is already a PR for the >>> template fuzzer, so I'd like to review that one too, so we can integrate >>> the two things in close succession. As we have seen with the rest of the >>> Vector API before: fuzzing revealed a good number of bugs that our >>> hand-written tests did not cover. Integrating tests later means we discover >>> bugs later, and that puts more pressure on the compiler team - I'd like to >>> avoid that ;) >>> >>> #30997 >> >> Understood. We are getting very close to RDP1, after we still have time to >> add more tests but not features. >> >> @jatin-bhateja is #30997 ready for testing with this PR? > >> > @PaulSandoz @jatin-bhateja I saw that there is already a PR for the >> > template fuzzer, so I'd like to review that one too, so we can integrate >> > the two things in close succession. As we have seen with the rest of the >> > Vector API before: fuzzing revealed a good number of bugs that our >> > hand-written tests did not cover. Integrating tests later means we >> > discover bugs later, and that puts more pressure on the compiler team - >> > I'd like to avoid that ;) >> > #30997 >> >> Understood. We are getting very close to RDP1, after we still have time to >> add more tests but not features. >> >> @jatin-bhateja is #30997 ready for testing with this PR? > > Hi @PaulSandoz , Yes I was able to use VectorExpressionFuzzer to generate > random tests involving Float16Vector operations. Will focus on addressing > @eme64's review comments once this patch gets integrated. @jatin-bhateja I'm running the internal mach5 testing on pr-28002 + pr30997. It might take some time to complete. ------------- PR Comment: https://git.openjdk.org/jdk/pull/28002#issuecomment-4560897488
