[GlobalISel] Remove dead code. (NFC)

[petechou]

LegalizerHelper::lowerMemCpyFamily only execpts G_MEMCPY, G_MEMMOVE, and
G_MMSET.

[llvmbot]

@llvm/pr-subscribers-llvm-globalisel

Author: Pete Chou (petechou)

Changes

LegalizerHelper::lowerMemCpyFamily only execpts G_MEMCPY, G_MEMMOVE, and
G_MMSET.

---

Full diff: https://github.com/llvm/llvm-project/pull/145811.diff

1 Files Affected:

• (modified) llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp (-4)

diff --git a/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp b/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
index 461fc35337eac..b87b029d01632 100644
--- a/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
@@ -10099,10 +10099,6 @@ LegalizerHelper::lowerMemCpyFamily(MachineInstr &MI, unsigned MaxLen) {
   }
 
   bool IsVolatile = MemOp->isVolatile();
-  if (Opc == TargetOpcode::G_MEMCPY_INLINE)
-    return lowerMemcpyInline(MI, Dst, Src, KnownLen, DstAlign, SrcAlign,
-                             IsVolatile);
-
   // Don't try to optimize volatile.
   if (IsVolatile)
     return UnableToLegalize;

[petechou]

Please help review the pr and merge it if it looks good. Thanks.

The assertion around L10073 should already test no G_MEMCPY_INLINE is expected. Actually, the code looks dead when it's first introduced in a642872, and is still kept in a later refactoring change 36527cb.

[arsenm]

This lowerMemcpyInline overload is also probably dead

[petechou]

No, I guess it's not. It's used in LegalizerHelper::lowerMemcpyInline(MachineInstr &MI) which is called by CombinerHelper::tryEmitMemcpyInline(MachineInstr &MI) from target pre-legalizer combiner.