[PATCH] D156249: [RISCV] Implement getOptimalMemOpType for memcpy/memset lowering

[Kito Cheng via Phabricator via llvm-commits]

kito-cheng added inline comments.


================
Comment at: llvm/lib/Target/RISCV/RISCVISelLowering.cpp:17002-17009
+  // to byte aligned accesses.
+  if (PreferredVT != MVT::i8 && !Subtarget.enableUnalignedVectorMem()) {
+    if (Op.isFixedDstAlign() &&
+        PreferredVT.getStoreSize() > Op.getDstAlign().value())
+      PreferredVT = MVT::i8;
+    if (Op.isMemcpy() &&
+        PreferredVT.getStoreSize() > Op.getSrcAlign().value())
----------------
Just a minor optimization thought here, we might use i16 or i32 (on RV64) if alignment is OK?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D156249/new/

https://reviews.llvm.org/D156249