[PATCH] Allow speculating llvm.sqrt, fma and fmuladd
listmail at philipreames.com
Wed Dec 18 19:17:18 PST 2013
On 12/17/13 11:12 AM, Matt Arsenault wrote:
> This should be OK since the sqrt intrinsic isn't supposed to be able to set errno, and the standard fma function doesn't anyway. The comment mentions others could be, but the documentation for these is very unclear. sqrt specifically says it doesn't set errno, but the others don't, though it probably shouldn't be different.
The change itself looked fine, but your comment bothers me. Can we get
a patch for the documentation which clarifies this? That might generate
some discussion that effects the correctness of this optimization.
More information about the llvm-commits