[PATCH] D25722: Improved cost model for FDIV and FSQRT
Michael Kuperstein via llvm-commits
llvm-commits at lists.llvm.org
Fri Oct 28 09:20:22 PDT 2016
mkuper added a comment.
In https://reviews.llvm.org/D25722#582086, @RKSimon wrote:
> I think this is almost done now - please can you add the AVX2/Haswell costs:
> FDIV f32/4f32 = 7, 8f32 = 14, f64/2f64 = 14, 4f64 = 28
> FSQRT f32/4f32 = 7, 8f32 = 14, f64/2f64 = 14, 4f64 = 28
> Other than that does anyone have any additional feedback?
No, LGTM (modulo what Simon said about Haswell)
Thanks for fixing this, Andrew.
More information about the llvm-commits