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

[Matt Arsenault via llvm-commits]

================
@@ -595,7 +944,38 @@ static void scalarize(Instruction *I, SmallVectorImpl<Instruction *> &Replace) {
   I->eraseFromParent();
 }
 
-static bool runImpl(Function &F, const TargetLowering &TLI) {
+// This covers all floating point types; more than we need here.
+// TODO Move somewhere else for general use?
+/// Return the Libcall for a frem instruction of
+/// type \p Ty.
+static RTLIB::Libcall fremToLibcall(Type *Ty) {
+  assert(Ty->isFloatingPointTy());
+  if (Ty->isFloatTy() || Ty->is16bitFPTy())
+    return RTLIB::REM_F32;
+  if (Ty->isDoubleTy())
+    return RTLIB::REM_F64;
+  if (Ty->isFP128Ty())
+    return RTLIB::REM_F128;
+  if (Ty->isX86_FP80Ty())
+    return RTLIB::REM_F80;
+  if (Ty->isPPC_FP128Ty())
+    return RTLIB::REM_PPCF128;
+
+  llvm_unreachable("Unknown floating point type");
+}
+
+/* Return true if, according to \p LibInfo, the target either directly
+   supports the frem instruction for the \p Ty, has a custom lowering,
+   or uses a libcall. */
+static bool targetSupportsFrem(const TargetLowering &TLI, Type *Ty) {
+  if (!TLI.isOperationExpand(ISD::FREM, EVT::getEVT(Ty)))
+    return true;
----------------
arsenm wrote:

Check if it's libcall or expand, using explicit libcall actions is another forever stalled migration 

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