[PATCH] D28508: [NVPTX] Lower to sqrt.approx and rsqrt.approx under more circumstances.

Justin Lebar via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Fri Jan 13 13:49:23 PST 2017


jlebar added a comment.

> I'm spinning a patch to use AutoUpgrade to get rid of some nvvm intrinsics entirely, and to use InstCombine to transform other nvvm intrinsics that we can't unconditionally remove into llvm intrinsics, where possible.

This is coming along, but I think it makes sense to leave the llvm.nvvm.sqrt.f special case in this patch.

The "upgrade patch" will upgrade llvm.nvvm.sqrt.f to llvm.sqrt.f32.  Therefore it needs this patch, which adds additional patterns around llvm.sqrt.f32.  But we can only get rid of the pattern matching 1/llvm.nvvm.sqrt.f after the upgrade patch lands.


https://reviews.llvm.org/D28508





More information about the llvm-commits mailing list