[PATCH] D122477: [X86][tablgen] Consider the mnemonic when auto-generating memory folding table

Kan Shengchen via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Fri Mar 25 22:53:51 PDT 2022


skan added a comment.

In D122477#3409420 <https://reviews.llvm.org/D122477#3409420>, @craig.topper wrote:

> In D122477#3409418 <https://reviews.llvm.org/D122477#3409418>, @skan wrote:
>
>> In D122477#3409416 <https://reviews.llvm.org/D122477#3409416>, @craig.topper wrote:
>>
>>> In D122477#3409403 <https://reviews.llvm.org/D122477#3409403>, @skan wrote:
>>>
>>>> In D122477#3409372 <https://reviews.llvm.org/D122477#3409372>, @craig.topper wrote:
>>>>
>>>>> Do you intend to replace the other cases that use NotMemoryFoldable in the .td file which was how we used to fix this issue. See the multiple cases already in X86InstrVMX.td
>>>>
>>>> I can see you added them in https://reviews.llvm.org/rG66572df76eb9acf5c6b84c573541065cb29ce2ff
>>>> I think most of `NotMemoryFoldable` can be saved by checking the `SchedRW` is not `WriteNop` or `WriteSystem`. Do you mind I do it in another NFC patch?
>>>
>>> As I said I was primarily asking about the ones in X86InstrVMX.td that are the exact same issue as VMXON and SENDUIPI. I just want us to solve the same problems in the same way.
>>
>> Okay, I will update the patch soon.
>
> Maybe it's not InstrVMX that has the overlap. But I remember something overlapped. It might be TPAUSE and CLWB

You're right! Done.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D122477/new/

https://reviews.llvm.org/D122477



More information about the llvm-commits mailing list