[PATCH] D94395: [X86] AMD Zen 3 Scheduler Model

Roman Lebedev via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Mon Apr 26 13:11:37 PDT 2021


lebedev.ri added a comment.

In D94395#2717476 <https://reviews.llvm.org/D94395#2717476>, @GGanesh wrote:

> In D94395#2717365 <https://reviews.llvm.org/D94395#2717365>, @lebedev.ri wrote:
>
>> In D94395#2717286 <https://reviews.llvm.org/D94395#2717286>, @GGanesh wrote:
>>
>>> Regarding the pipe events, the patch in kernel.org had just copy pasted events from zen1 and zen2. Apparently, I learnt that these events are restricted and at present not really ready for public for znver3. That is what is reflected in the kernel patch as well. If we notice, PMCx000 is incorrectly mapped and doesn't really count the 6 pipes.
>>
>> Is that something that might be fixed via/in next microcode updates, or it's a hardware bug?
>
> I have very limited info. However, definitely the solution is not near term. That's why I am inclined to check on the workarounds.

I see.
*Please* do tell them that it would be good to also have these counters for Int pipes :)

>>> Is there any workaround that you guys suggest. I understand that these are pfm counters and are dynamic in nature. However, for the model, (at least for the throughput numbers) shall I publish our internal numbers running the internally enabled libpfm4 library? This is restricted in its immediate purpose which I understand however at present it may act as a work around for the throughput and llvm-mca numbers.
>>
>> The answer depends on an answer to the previous question.

Then, what do you mean by "internal numbers running the internally enabled libpfm4 library"?
I'm guessing you aren't talking about just patching libpfm4, and running it on a real cpu?

I think we can improve this afterwards, the initial version doesn't have to be perfect,
it won't ever be.

>>> I am going through the modifications\corrections made in the by @lebedev.ri  in the meantime.




Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D94395/new/

https://reviews.llvm.org/D94395



More information about the llvm-commits mailing list