[PATCH] D19076: AMDGPU/SI: Re-implement the lowering for 32-bit floating point division

[Matt Arsenault via llvm-commits]

arsenm added a comment.

The fdiv test must be modified.

We should not simply replace the division lowering this way. This should be controlled by precision options, so the existing lowering should still be able to work


http://reviews.llvm.org/D19076