[PATCH] D70901: [Matrix] Update shape propagation to iterate until done.

[Adam Nemet via Phabricator via llvm-commits]

anemet accepted this revision.
anemet added a comment.
This revision is now accepted and ready to land.

LGTM.



================
Comment at: llvm/lib/Transforms/Scalar/LowerMatrixIntrinsics.cpp:391
+    auto pushInstruction = [&WorkList](Value *V) {
+      Instruction *I = dyn_cast<Instruction>(V);
+      if (I)
----------------
I think that this should take the WorkList as a parameter so that it's clear it's not pushing to the NewWorkList


================
Comment at: llvm/lib/Transforms/Scalar/LowerMatrixIntrinsics.cpp:404
+      size_t WSize = WorkList.size();
       if (!isa<Instruction>(V))
         continue;
----------------
WSize -> BeforeProcessingV or something like that


================
Comment at: llvm/lib/Transforms/Scalar/LowerMatrixIntrinsics.cpp:447
+      // After we discovered new shape info for new instructions in the
+      // worklist, we use them as seeds for the next round of forward
+      // propagation.
----------------
... we use their users ...


================
Comment at: llvm/lib/Transforms/Scalar/LowerMatrixIntrinsics.cpp:451
+        for (User *U : WorkList[I]->users())
+          if (isa<Instruction>(U))
+            NewWorkList.push_back(cast<Instruction>(U));
----------------
Should this exclude V?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D70901/new/

https://reviews.llvm.org/D70901