[PATCH] D140599: AMDGPU: Promote array alloca if used by memmove/memcpy

[Jay Foad via Phabricator via llvm-commits]

foad added inline comments.


================
Comment at: llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp:544
+    Instruction *Inst = DeferredInsts.pop_back_val();
+    if (MemTransferInst *TransferInst = dyn_cast<MemTransferInst>(Inst)) {
+      // TODO: Support the case if the pointers are from different alloca or
----------------
This can be `cast` instead of `dyn_cast`.


================
Comment at: llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp:611
+            Type *VecPtrTy = VectorTy->getPointerTo(Alloca->getAddressSpace());
+            Value *BitCast = Builder.CreateBitCast(Alloca, VecPtrTy);
+            Value *VecValue = Builder.CreateLoad(VectorTy, BitCast);
----------------
ruiling wrote:
> arsenm wrote:
> > No reason to care about type pointers anymore
> Our graphics compiler has not fully switched to opaque pointers yet. @foad am I right?
Yes.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D140599/new/

https://reviews.llvm.org/D140599