[llvm] [CodeGen] Improve ExpandMemCmp for more efficient non-register aligned sizes handling (PR #69942)

[Clement Courbet via llvm-commits]

================
@@ -255,6 +255,31 @@ MemCmpExpansion::MemCmpExpansion(
     }
   }
   assert(LoadSequence.size() <= Options.MaxNumLoads && "broken invariant");
+  // This part of code attempts to optimize the LoadSequence by merging allowed
+  // subsequences into single loads of allowed sizes from
+  // `AllowedTailExpansions`. If it is for zero comparison or if no allowed tail
+  // expansions are specified, we exit early.
+  if (IsUsedForZeroCmp || !Options.AllowedTailExpansions.size())
+    return;
+
+  while (LoadSequence.size() >= 2) {
+    auto Last = LoadSequence[LoadSequence.size() - 1];
+    auto PreLast = LoadSequence[LoadSequence.size() - 2];
+
----------------
legrosbuffle wrote:

Can you move this to a separate function for consistency with the `computeOverlappingLoadSequence` code above ?

https://github.com/llvm/llvm-project/pull/69942