[llvm] [AMDGPU] Implement IR expansion for frem instruction (PR #130988)

[Matt Arsenault via llvm-commits]

================
@@ -696,13 +1104,19 @@ class ExpandFpLegacyPass : public FunctionPass {
 
 PreservedAnalyses ExpandFpPass::run(Function &F, FunctionAnalysisManager &FAM) {
   const TargetSubtargetInfo *STI = TM->getSubtargetImpl(F);
-  return runImpl(F, *STI->getTargetLowering()) ? PreservedAnalyses::none()
-                                               : PreservedAnalyses::all();
+  auto &TLI = *STI->getTargetLowering();
+  AssumptionCache *AC = nullptr;
+  if (TM->getOptLevel() != CodeGenOptLevel::None)
----------------
arsenm wrote:

I'm not sure, this area is kind of a mess. We have mixed uses of both 

https://github.com/llvm/llvm-project/pull/130988