[PATCH] D28508: [NVPTX] Implement NVPTXTargetLowering::getSqrtEstimate.

[Justin Lebar via Phabricator via llvm-commits]

jlebar added a comment.

In https://reviews.llvm.org/D28508#661626, @escha wrote:

> Don't be too embarrassed; when we switched internally from an rcp(rsqrt(x)) expansion to x * rsqrt(x), we *also* completely missed this.


:)  I switched to rcp(rsqrt(x)) in r293713.  I tested and it's significantly faster than x*rsqrt(x), to say nothing of adding an extra select around that.

Thanks again for catching this.


Repository:
  rL LLVM

https://reviews.llvm.org/D28508