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

[Amara Emerson 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)))) &&
----------------
aemerson wrote:

Maybe I'm not understanding what you're suggesting. `isInTailCallPosition` is called in multiple places so adding in an extra parameter doesn't make it any cleaner IMO. getMemset() is called later, by that point we've already decided to tail-call. We could downgrade the tail call in getMemset() but then we also lose the tail-call when the parent function's return type was void or it returned undef.

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