[llvm] [AMDGPU] Avoid resource propagation for recursion through multiple functions (PR #111004)

[Janek van Oirschot via llvm-commits]

================
@@ -91,13 +91,66 @@ MCSymbol *MCResourceInfo::getMaxSGPRSymbol(MCContext &OutContext) {
   return OutContext.getOrCreateSymbol("amdgpu.max_num_sgpr");
 }
 
+static bool findSymbolInExpr(MCSymbol *Sym, const MCExpr *Expr,
+                             SmallVectorImpl<const MCExpr *> &Exprs) {
+  switch (Expr->getKind()) {
+  default:
+    return false;
+  case MCExpr::ExprKind::SymbolRef: {
+    const MCSymbolRefExpr *SymRefExpr = cast<MCSymbolRefExpr>(Expr);
+    const MCSymbol &SymRef = SymRefExpr->getSymbol();
+    if (Sym == &SymRef)
+      return true;
+    if (SymRef.isVariable())
+      Exprs.push_back(SymRef.getVariableValue(/*isUsed=*/false));
+    return false;
+  }
+  case MCExpr::ExprKind::Binary: {
+    const MCBinaryExpr *BExpr = cast<MCBinaryExpr>(Expr);
+    return findSymbolInExpr(Sym, BExpr->getLHS(), Exprs) ||
+           findSymbolInExpr(Sym, BExpr->getRHS(), Exprs);
----------------
JanekvO wrote:

I assume the `getConstantAccess` example walks over an already completed expression. My approach here is more preventative in terms of recursion so no recursion should exist, at all. I've added a Visited set but more so to ensure the invariant of no recursion to be upheld + added more comments that hopefully explain that this walk is to avoid recursion rather than detect and remove recursion.

https://github.com/llvm/llvm-project/pull/111004