[PATCH] D31350: AMDGPU : Fix common dominator of two incoming blocks terminates with uniform branch issue.

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added inline comments.


================
Comment at: lib/Target/AMDGPU/SIFixSGPRCopies.cpp:335
+  cur.push(MBB);
+  while (!cur.empty())
+  {
----------------
There needs to be a set because there can be loops


================
Comment at: test/CodeGen/AMDGPU/sgprcopies.ll:1
+; RUN: llc -march=amdgcn < %s | FileCheck -check-prefix=SI %s
+
----------------
s/SI GCN.

Also needs -verify-machineinstrs


================
Comment at: test/CodeGen/AMDGPU/sgprcopies.ll:3
+
+; SI-LABEL: {{^}}sgprcopies
+; SI: BB0_2
----------------
Should have better name and secretion of what this is really checking


================
Comment at: test/CodeGen/AMDGPU/sgprcopies.ll:12
+  %conv1 = sitofp i32 %rem to float
+  %0 = tail call float @llvm.fmuladd.f32(float %xStep, float %conv1, float %xPos)
+  %conv2 = sitofp i32 %div to float
----------------
Should have instnamer run on this and the test could be simplified a bit


Repository:
  rL LLVM

https://reviews.llvm.org/D31350