[PATCH] Allow FMAs in safe math mode in some cases when one operand of the fmul is either exactly 0.0 or exactly 1.0.

Stephen Lin swlin at post.harvard.edu
Tue Jul 9 16:29:02 PDT 2013


> Which one?
>
> In any case, I suspect that handling UNDEFs here is much more important for vectors than for scalars.
>
>  -Hal

test/CodeGen/X86/2012-04-26-sdglue.ll . Not really sure what's going
on with that.

I see what you mean about handling undef for promoted vectors but I'd
rather test specifically for the cases that matter rather than
allowing undef in a blanket fashion, just in case.

Stephen



More information about the llvm-commits mailing list