[X86] Do not ignore AlwaysInline when determining if inlining is worth it

[AtariDreams]

AlwaysInline parameter would be ignored if the outlining exceeded the threshold or simply was not aligned enough.

[llvmbot]

@llvm/pr-subscribers-backend-x86

Author: AtariDreams (AtariDreams)

Changes

AlwaysInline parameter would be ignored if the outlining exceeded the threshold or simply was not aligned enough.

---

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

1 Files Affected:

• (modified) llvm/lib/Target/X86/X86SelectionDAGInfo.cpp (+4-2)

diff --git a/llvm/lib/Target/X86/X86SelectionDAGInfo.cpp b/llvm/lib/Target/X86/X86SelectionDAGInfo.cpp
index 7c630a2b0da080..d1e7c9046fdc17 100644
--- a/llvm/lib/Target/X86/X86SelectionDAGInfo.cpp
+++ b/llvm/lib/Target/X86/X86SelectionDAGInfo.cpp
@@ -66,8 +66,10 @@ 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();
 
   uint64_t SizeVal = ConstantSize->getZExtValue();

[phoebewang]

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.