[PATCH] D97667: [loop-idiom] Hoist loop memcpys to loop preheader

[Han Zhu via Phabricator via llvm-commits]

zhuhan0 added a comment.

In D97667#2725626 <https://reviews.llvm.org/D97667#2725626>, @dstenb wrote:

> With this patch we got the following assertion:
>
>   bool llvm::APInt::operator==(const llvm::APInt &) const: Assertion `BitWidth == RHS.BitWidth && "Comparison requires equal bit widths"' failed.
>
> in `LoopIdiomRecognize::processLoopMemCpy()` at the following comparison:
>
>   // Check if the load stride matches the store stride.
>   if (StrIntStride != LoadIntStride && StrIntStride != -LoadIntStride)
>     return false;
>
> for a memcpy done between two address spaces with different pointer sizes.
>
> I don't have a upstream reproducer ready for this, but I'll see if I can create one.

The new version should fix the assertion I think. Let me know if you still see it.



================
Comment at: llvm/lib/Transforms/Scalar/LoopIdiomRecognize.cpp:865
+  APInt LoadIntStride = LoadStride->getAPInt();
+  if (SizeInBytes != StoreIntStride && SizeInBytes != -StoreIntStride) {
+    ORE.emit([&]() {
----------------
lebedev.ri wrote:
> I don't think this should handle the case of different directions.
The different directions case is handled below in 
```
  // Check if the load stride matches the store stride.
  if (StoreStrideInt != LoadStrideInt)
    return false;
```
Or do you refer to something else?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D97667/new/

https://reviews.llvm.org/D97667