RFR: 8309660: C2: failed: XMM register should be 0-15 (UseKNLSetting and ConvF2HF)
Emanuel Peter
epeter at openjdk.org
Mon Jun 19 13:46:11 UTC 2023
On Thu, 15 Jun 2023 02:30:53 GMT, Quan Anh Mai <qamai at openjdk.org> wrote:
>> @merykitty @sviswa7 @fg1417 Is there a way to stress-test the registers? It seems this bug only triggered because we had a moderately large unrolling factor, and then did not vectorize, leaving lots of instructions with probably a higher register pressure. Would be nice to have some sort of testing where we generate more (all?) of the possible register combinations. What do you think?
>
> @eme64 Yes that was my mistake, that node requires AVX512VL so `vlRegF` and `regF` are the same.
>
>> Is there a way to stress-test the registers?
>
> Can we randomise the allocated register during register allocation?
>
> Thanks.
@merykitty Yes, randomization would be great. I don't know much about the register allocator, so feel free to do something like that if you want and have time ;)
@sviswa7 Is there something you want me to change still?
-------------
PR Comment: https://git.openjdk.org/jdk/pull/14379#issuecomment-1597219792
More information about the hotspot-compiler-dev
mailing list