[llvm] [X86] For minsize, use size for alignment, rather than actual alignment (PR #87003)

[Simon Pilgrim via llvm-commits]

================
@@ -66,11 +66,30 @@ 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 ||
+      (!AlwaysInline &&
+       (Alignment < Align(4) ||
+        ConstantSize->getZExtValue() > Subtarget.getMaxInlineSizeThreshold())))
     return SDValue();
 
+  // If we have minsize, then don't care about the alignment.
+  // On x86, the CPU doesn't care and neither should you.
+  // As long as the count is aligned, we can use the minimum number of
+  // instructions without always having to resort to stosb.
+  //
+  // Because this is a feature specific to x86, we must handle it here.
   uint64_t SizeVal = ConstantSize->getZExtValue();
----------------
RKSimon wrote:

(See #87825 - split the if statement to simplify)

https://github.com/llvm/llvm-project/pull/87003