[PATCH] D72675: Fix -ffast-math/-ffp-contract interaction

[Sanjay Patel via Phabricator via cfe-commits]

spatel added inline comments.


================
Comment at: llvm/test/CodeGen/PowerPC/fmf-propagation.ll:201-203
 ; fma(X, 7.0, X * 42.0) --> X * 49.0
-; This is the minimum FMF needed for this transform - the FMA allows reassociation.
+; This is the minimum FMF needed for this transform - the 'contract' allows the needed reassociation.
+
----------------
I was ok with the reasoning up to here, but this example lost me.
Why does 'contract' alone allow splitting an fma?
Is 'contract' relevant on anything besides fadd (with an fmul operand)?


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D72675/new/

https://reviews.llvm.org/D72675