[llvm] [SDAG] Heed enableAggressiveFMAFusion when folding fma(a,b,c*d)+e (PR #94209)

[Jay Foad via llvm-commits]

jayfoad wrote:

> In a separate patch does the visitFSUBForFMACombine need a similar update?

Yes, similar but different. Currently the equivalent folds in `visitFSUBForFMACombine` are not performed at all without enableAggressiveFMAFusion, even if the intermediate nodes are only used once.

https://github.com/llvm/llvm-project/pull/94209