[llvm] [AArch64] Don't tail call memset if it would convert to a bzero. (PR #98969)

[Eli Friedman via llvm-commits]

================
@@ -685,7 +687,10 @@ bool llvm::returnTypeIsEligibleForTailCall(const Function *F,
          (IID == Intrinsic::memmove &&
           TLI.getLibcallName(RTLIB::MEMMOVE) == StringRef("memmove")) ||
          (IID == Intrinsic::memset &&
-          TLI.getLibcallName(RTLIB::MEMSET) == StringRef("memset"))) &&
+          TLI.getLibcallName(RTLIB::MEMSET) == StringRef("memset") &&
+          (!isa<ConstantInt>(Call->getOperand(1)) ||
+           !cast<ConstantInt>(Call->getOperand(1))->isZero() ||
+           !TLI.getLibcallName(RTLIB::BZERO)))) &&
----------------
efriedma-quic wrote:

Currently, isTailCall is a parameter to SelectionDAG::getMemset(), but it doesn't really need to be; SelectionDAGBuilder::visitIntrinsicCall just calls isInTailCallPosition so it can pass the result into getMemset().  Instead, we can just call isInTailCallPosition from getMemset(), once we know whether we're calling bzero or memset, and we can pass that information down.

https://github.com/llvm/llvm-project/pull/98969