[llvm] [AMDGPU] Enable vectorization of i8 values. (PR #134934)

[Jeffrey Byrnes via llvm-commits]

================
@@ -562,6 +563,7 @@ InstructionCost GCNTTIImpl::getArithmeticInstrCost(
     if (ST->has16BitInsts() && SLT == MVT::i16)
       NElts = (NElts + 1) / 2;
 
+    // i32
----------------
jrbyrnes wrote:

Accidental comment? Should be full sentence.

https://github.com/llvm/llvm-project/pull/134934