[PATCH] D70951: [Matrix] Use fmuladd when lowering matrix.multiply.

[Florian Hahn via Phabricator via llvm-commits]

fhahn updated this revision to Diff 234600.
fhahn added a comment.

> I am concerned about the stability of the result in the presence of the FMA and non-FMA targets. We ought to generate the same result in IEEE fashion but with this we wouldn't. I think that this should still be driven by the fp-contract fp flag and perhaps we should have command-line option that enables this specifically for matmul?

Agreed! I've added a matrix-allow-contraction option. I'll add support for using fast-math flags as follow up


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D70951/new/

https://reviews.llvm.org/D70951

Files:
  llvm/lib/Transforms/Scalar/LowerMatrixIntrinsics.cpp
  llvm/test/Transforms/LowerMatrixIntrinsics/bigger-expressions-double-contraction.ll
  llvm/test/Transforms/LowerMatrixIntrinsics/multiply-double-contraction.ll
  llvm/test/Transforms/LowerMatrixIntrinsics/multiply-float-contraction.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D70951.234600.patch
Type: text/x-patch
Size: 97007 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20191218/e076d0ca/attachment.bin>