RFR: 8302681: [IR Framework] Only allow cpuFeatures from a verified list [v6]

Tobias Hartmann thartmann at openjdk.org
Wed Feb 22 10:10:28 UTC 2023


On Wed, 22 Feb 2023 08:03:22 GMT, Emanuel Peter <epeter at openjdk.org> wrote:

>> Wrongly typed cpuFeatures can lead to an IR rule being ignored.
>> Example: https://github.com/openjdk/jdk/pull/12601 [JDK-8302668](https://bugs.openjdk.org/browse/JDK-8302668)
>> 
>> We should have a list of verified cpuFeatures, and assert if one is used that is not in that list. That way, typos and bad copies can be avoided, and we make sure the tests run when intended.
>> 
>> I verified the `intel` cpuFeatures by hand on my laptop. And for `asimd` and `sve` I asked @tobiasholenstein and @nick-arm .
>> 
>> `sve1` seems to have been a typo, and I am changing it to `sve`.
>> 
>> Note: cpu-features can be displayed like this: `./java -Xlog:os+cpu --version`
>> 
>> On my machine, I get:
>> 
>> cmov, fxsr, ht, mmx, 3dnowpref, sse, sse2, sse3, ssse3, sse4.1, sse4.2, popcnt, lzcnt, tsc, tscinvbit, avx, avx2, aes, erms, clmul, bmi1, bmi2, adx, avx512f, avx512dq, avx512cd, avx512bw, avx512vl, sha, fma, vzeroupper, avx512_vpopcntdq, avx512_vpclmulqdq, avx512_vaes, avx512_vnni, clflush, clflushopt, clwb, avx512_vbmi2, avx512_vbmi, rdtscp, rdpid, fsrm, gfni, avx512_bitalg, f16c, pku, ospke, cet_ibt, cet_ss, avx512_ifma
>> 
>> 
>> @nick-arm got this on one of his machines:
>> 
>> fp, asimd, evtstrm, aes, pmull, sha1, sha256, crc32, lse, dcpop, sha3, sha512, sve, paca
>> 
>> 
>> @tobiasholenstein got this:
>> 
>> fp, asimd, aes, pmull, sha1, sha256, crc32, lse, sha3, sha512
>> 
>> 
>> I had to adapt a test that used a non-existent cpu feature. I replaced it with an impossible combination.
>
> Emanuel Peter has updated the pull request incrementally with one additional commit since the last revision:
> 
>   more fixes after Tobias review

Thanks for making these changes. Looks good to me.

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

Marked as reviewed by thartmann (Reviewer).

PR: https://git.openjdk.org/jdk/pull/12669


More information about the hotspot-compiler-dev mailing list