[PATCH] D109131: [GlobalISel] Add a store-merging optimization pass and enable for AArch64.

[Jessica Paquette via Phabricator via llvm-commits]

paquette added inline comments.


================
Comment at: llvm/lib/CodeGen/GlobalISel/LoadStoreOpt.cpp:157
+
+  if (Base0Def->getOpcode() == TargetOpcode::G_FRAME_INDEX) {
+    if (Base1Def->getOpcode() == TargetOpcode::G_FRAME_INDEX) {
----------------
Will there ever be a combination of opcodes where `Base0Def->getOpcode() != Base1Def->getOpcode()`?

Would it be simpler to check something like...

```
unsigned Opc = Base0Def->getOpcode();
if (Opc !=  Base1Def->getOpcode())
  return false;

if (Opc == TargetOpcode::G_FRAME_INDEX)
   ...
```

?


================
Comment at: llvm/lib/CodeGen/GlobalISel/LoadStoreOpt.cpp:403
+      Builder.buildStore(WideReg, Stores[0]->getPointerReg(), *WideMMO);
+  LLVM_DEBUG(dbgs() << "Created merged store: " << *NewStore);
+  NumStoresMerged += Stores.size();
----------------
Wonder if we should actually have remarks for these as well? IIRC people sometimes want remarks from SDAG. Maybe this is something we can improve on with GISel.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D109131/new/

https://reviews.llvm.org/D109131