[PATCH] D128481: Add expensive/extensive pass to set Load/Store alignment

[Alexander Richardson via Phabricator via llvm-commits]

arichardson added inline comments.


================
Comment at: llvm/lib/Transforms/Scalar/SetLoadStoreAlignment.cpp:128
+        }
+      } else if (StoreInst *SI = dyn_cast<StoreInst>(&I)) {
+        Value *ptr = SI->getPointerOperand();
----------------
It probably also makes sense to include memtransferinst here. While looking at poor codegen for CHERI-enabled RISC-V, I've noticed that missing alignment propagation on llvm.memcpy/memmove resulted in memcpy calls instead of inline expansion.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D128481/new/

https://reviews.llvm.org/D128481