[clang] [llvm] [AArch64][SME] Allow more inlining when SME attributes are incompatible. (PR #223393)

[Sander de Smalen via cfe-commits]

================
@@ -230,30 +230,96 @@ static cl::opt<bool> EnableFixedwidthAutovecInStreamingMode(
 static cl::opt<bool> EnableScalableAutovecInStreamingMode(
     "enable-scalable-autovec-in-streaming-mode", cl::init(false), cl::Hidden);
 
-static bool isSMEABIRoutineCall(const CallInst &CI,
-                                const AArch64TargetLowering &TLI) {
-  const auto *F = CI.getCalledFunction();
-  return F &&
-         SMEAttrs(F->getName(), TLI.getRuntimeLibcallsInfo()).isSMEABIRoutine();
+/// Returns true if \p I is an intrinsic that may not be compatible with a
+/// different streaming mode (because it depends on vscale).
+static bool isPossiblyIncompatibleIntrinsic(const Instruction *I) {
+  if (I->isDebugOrPseudoInst())
+    return false;
+
+  if (auto *II = dyn_cast<IntrinsicInst>(I)) {
+    unsigned IID = II->getIntrinsicID();
+    switch (IID) {
+    default:
+      return Intrinsic::isTargetIntrinsic(IID);
----------------
sdesmalen-arm wrote:

For now, we keep this behaviour conservative, in the sense that we assume any target intrinsic is possibly invalid in streaming mode (even if [we can assume that] vscale is equivalent).

https://github.com/llvm/llvm-project/pull/223393