RFR: 8280120: [IR Framework] Add attribute to @IR to enable/disable IR matching based on the architecture [v2]
Daniel Skantz
duke at openjdk.org
Wed Oct 4 11:51:37 UTC 2023
On Mon, 2 Oct 2023 09:02:00 GMT, Emanuel Peter <epeter at openjdk.org> wrote:
>>> > Appreciate the review @robcasloz. I rewrote parts of the documentation and added more tests.
>>>
>>> The new version looks good. Maybe I missed something, but does any of the newly added tests address my suggestion (https://github.com/openjdk/jdk/pull/15938/files#r1338489724) to check that platform preconditions take precedence over VM flags, i.e. VM flags are not even validated if platform preconditions do not hold?
>>
>> Thanks for the review. The intention is that the last four tests in `TestPreconditions.java` test precedence through platform-specific VM flags. As an example `testApplyPlatformSVE` will fail with `Could not find VM flag "UseSVE"` if we remove the `applyIfPlatform` precondition on x64.
>
> Thanks for doing this, @danielogh !
> I thinks this patch is good. But please add the rules requested here https://github.com/openjdk/jdk/pull/15938#discussion_r1338489724.
> And make sure to file the follow-up RFE and link it to this issue here https://github.com/openjdk/jdk/pull/15938#discussion_r1338137686.
>
> Update: And I think you don't need to add more IR rules. You already have tests that use VM flags that only exist on certain platforms. Platform features need to be evaluated before VM flags, but not necessarily before CPU features. (just repeating what we discussed offline)
>
> Update 2: I filed the follow-up RFE: [JDK-8317341](https://bugs.openjdk.org/browse/JDK-8317341)
Thanks @eme64 and @TobiHartmann for reviews
-------------
PR Comment: https://git.openjdk.org/jdk/pull/15938#issuecomment-1746713486
More information about the hotspot-compiler-dev
mailing list