[llvm] [X86] Do not ignore AlwaysInline when determining if inlining is worth it (PR #87825)

[Phoebe Wang via llvm-commits]

================
@@ -66,11 +66,14 @@ SDValue X86SelectionDAGInfo::EmitTargetCodeForMemset(
   // If not DWORD aligned or size is more than the threshold, call the library.
   // The libc version is likely to be faster for these cases. It can use the
   // address value and run time information about the CPU.
-  if (Alignment < Align(4) || !ConstantSize ||
-      ConstantSize->getZExtValue() > Subtarget.getMaxInlineSizeThreshold()) 
+  if (!ConstantSize)
     return SDValue();
 
   uint64_t SizeVal = ConstantSize->getZExtValue();
+  if (!AlwaysInline &&
----------------
phoebewang wrote:

`AlwaysInline` is taken care in https://github.com/llvm/llvm-project/blob/main/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp#L8260 anyway.
I'm not sure if `stosb` is better than sequential stores.

https://github.com/llvm/llvm-project/pull/87825