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

[Jeffrey Byrnes via llvm-commits]

================
@@ -1419,6 +1419,10 @@ unsigned TargetTransformInfo::getMaxNumArgs() const {
   return TTIImpl->getMaxNumArgs();
 }
 
+bool TargetTransformInfo::canVectorizei8s() const {
----------------
jrbyrnes wrote:

Not sure about this TTI hook. In general, there is no special about i8s. They just happen to require certain special handling on AMDGPUs. However, for other targets, i8s may be first class citizens / legal, so they are already legalized by default -- x86 for example I believe has legal i8s

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