[PATCH] D80416: [RFC][OpenCL] Set fp contract flag on -cl-mad-enable

[Matt Arsenault via Phabricator via cfe-commits]

arsenm added a comment.

The langref wording makes me think this isn't quite right. This depends on your definition of floating point contraction. I've always assumed it meant allow FMA, potentially increasing precision. Is contracting into something less precise allowed? If not, that's stricter / the opposite of what -cl-mad-enable implies. My interpretation of the CL spec description would be to use fmuladd with an afn flag (although that still can allow for increasing precision)

For AMDGPU I've thought about interpreting less-precise-fpmad as allowing denormal flushing that would otherwise be illegal. Currently it doesn't do anything, but somehow interpreting the flags for this would be better.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D80416/new/

https://reviews.llvm.org/D80416