[PATCH] D131773: [AArch64] Add support for 256-bit non temporal loads

Tue Aug 16 01:44:52 PDT 2022

dmgreen accepted this revision.
dmgreen added a comment.
This revision is now accepted and ready to land.

LGTM if @fhahn doesn't know of a reason to add the isKnownEven.

================
Comment at: llvm/lib/Target/AArch64/AArch64ISelLowering.cpp:20391
+    if (LoadNode->isNonTemporal() && MemVT.getSizeInBits() == 256u &&
+        EC.isKnownEven() &&
+        ((MemVT.getScalarSizeInBits() == 8u ||
----------------
zjaffal wrote:
> dmgreen wrote:
> > Why is EC.isKnownEven needed?
> I followed the same code that is used to lower `STNP` but I think it is not really necessary here. 
> I think we can remove it.
Yeah - I agree so long as it is checking the size of the MemVT and it's scalar element size.

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D131773/new/

https://reviews.llvm.org/D131773