[PATCH] D132880: [SVE] Fix SVEDup0 matching -0.0f

[Paul Walker via Phabricator via llvm-commits]

paulwalker-arm added inline comments.


================
Comment at: llvm/lib/Target/AArch64/AArch64ISelDAGToDAG.cpp:187
+      if (isNullConstant(Opnd0))
+        return true;
+      if (isNullFPConstant(Opnd0))
----------------
Allen wrote:
> sorry, I have a question: can we match the -0.0 if the option **-ffast-math**  is added in command ?
That feels a little dangerous to me.  I think such a canonicalisation would be better done at the DAG or perhaps even the IR level.  For instruction selection it seems safer to be explicit.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D132880/new/

https://reviews.llvm.org/D132880