[PATCH] D61447: [FPEnv] WIP on threading fneg through llvm

[Cameron McInally via Phabricator via llvm-commits]

cameron.mcinally added a comment.

Thanks for working on this, Kevin. It looks like the general direction is good. A couple small comments inline. I also noticed some of the tests still need updating to replace fsub's, but I assume you know that.

@spatel I'm not a PatternMatch expert, but I think that change is what we wanted. That is, to continue to match `FSub(-0.0, X)` as an `FNeg(X)`.

I'm currently working on ConstantExpr folding for FNeg in D61419 <https://reviews.llvm.org/D61419>. And soon to be looking at FNeg folding in InstructionSimplify. Just FYI...



================
Comment at: lib/Transforms/InstCombine/InstCombineAddSub.cpp:1938
+    }
+
+  return nullptr;
----------------
We should also do:

`-undef -> undef`


================
Comment at: lib/Transforms/Scalar/Reassociate.cpp:887
   // Insert a 'neg' instruction that subtracts the value from zero to get the
   // negation.
+  Instruction *NewNeg = CreateNeg(V, V->getName() + ".neg", BI, BI);
----------------
Comment needs to be updated.


Repository:
  rL LLVM

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D61447/new/

https://reviews.llvm.org/D61447