AMDGPU: Move LICM after AMDGPUCodeGenPrepare

The commit that added the run says it's to hoist uniform parts of
integer division expansion. That expansion is performed later, so this
didn't do anything in that case. Move this later so the original test
shows the improvement.

This also saves a run of "Canonicalize natural loops". Not sure why
this appears to be still getting a separate loop PM run. Also feels a
bit heavy to run this just for divide. Is there a way to specifically
hoist the divide sequence when it expands?

llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp

@@ -967,7 +967,6 @@ void AMDGPUPassConfig::addEarlyCSEOrGVNPass() {
 }
 
 void AMDGPUPassConfig::addStraightLineScalarOptimizationPasses() {
-  addPass(createLICMPass());
   addPass(createSeparateConstOffsetFromGEPPass());
   // ReassociateGEPs exposes more opportunities for SLSR. See
   // the example in reassociate-geps-and-slsr.ll.
@@ -1039,6 +1038,11 @@ void AMDGPUPassConfig::addIRPasses() {
       // TODO: May want to move later or split into an early and late one.
       addPass(createAMDGPUCodeGenPreparePass());
     }
+
+    // Try to hoist loop invariant parts of divisions AMDGPUCodeGenPrepare may
+    // have expanded.
+    if (TM.getOptLevel() > CodeGenOpt::Less)
+      addPass(createLICMPass());
   }
 
   TargetPassConfig::addIRPasses();