RFR: 8355913: RISC-V: improve hotspot/jtreg/compiler/vectorization/runner/BasicFloatOpTest.java

Fei Yang fyang at openjdk.org
Wed Apr 30 09:03:46 UTC 2025


On Wed, 30 Apr 2025 08:48:18 GMT, Hamlin Li <mli at openjdk.org> wrote:

>> I suggest we don't touch other platforms unless I have to do so (e.g. in https://github.com/openjdk/jdk/pull/24947, I have to touch an existing test in another platform), because we need to test it, and it could introduce potential failures, so I won't risk for it.
>> 
>> For `-XX:+UseFMA`, I suggest we don't add it to the whole test, as there are other IR verifications, which does not depend on it.
>> 
>> How do you think about it?
>
> And, as you can see `"fma", "true", "avx", "true"`, it's doing the similar thing as riscv.

> I suggest we don't touch other platforms unless I have to do so (e.g. in #24947, I have to touch an existing test in another platform), because we need to test it, and it could introduce potential failures, so I won't risk for it.

OK if you want to keep this change RISC-V specific.

> For `-XX:+UseFMA`, I suggest we don't add it to the whole test, as there are other IR verifications, which does not depend on it.

That make sense to me.

-------------

PR Review Comment: https://git.openjdk.org/jdk/pull/24950#discussion_r2068212508


More information about the hotspot-compiler-dev mailing list