[PATCH] D23583: [AArch64] Add feature has-fast-fma

[Renato Golin via llvm-commits]

rengolin added a comment.

In https://reviews.llvm.org/D23583#518142, @evandro wrote:

> However, if you look at the foldings in `DAGCompiner.cpp` performed only if  `enableAggressiveFMAFusion()` is true you will find that `FMUL` is the pivot of the folding, not `FADD`, in spite of what the comments states.


This makes sense. FMA being faster-than/as-fast-as FADD doesn't make sense.

> For instance, whether it should fold into FMA twice, back to back.


This may be remnants of an ARMv7 issue around OOO cores having a hazard for FMA+FMA...

> As a matter of fact, it seems to me that most "big" targets would benefit from folding into FMA more aggressively.


I tend to agree, but I don't have evidence. The best evidence would be some benchmark results or scheduling descriptions.

I'm not against this change in itself, but I really think this is a matter for the scheduler, not a target feature.


Repository:
  rL LLVM

https://reviews.llvm.org/D23583