[PATCH] D149589: AMDGPU: Drop and auto-upgrade llvm.amdgcn.ldexp to llvm.ldexp

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added inline comments.


================
Comment at: llvm/test/CodeGen/AMDGPU/known-never-snan.ll:661
 declare float @llvm.fmuladd.f32(float, float, float) #1
-declare float @llvm.amdgcn.ldexp.f32(float, i32) #1
+declare float @llvm.ldexp.f32.i32(float, i32) #1
 declare float @llvm.amdgcn.fmul.legacy(float, float) #1
----------------
Joe_Nash wrote:
> Why did this case need a manual update rather than use the autoupgrade?
The auto-upgrade would work functionality wise. But there are 2 issues. For the optimization purposes, the old form wasn't recognized in isCanonicalized so the optimization broke. Second I think it's good practice to upgrade tests to the new intrinsics and only test the old forms in upgrade tests 


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D149589/new/

https://reviews.llvm.org/D149589