[llvm] [AMDGPU] Move kernarg preload logic to separate pass (PR #130434)

[Krzysztof Drewniak via llvm-commits]

================
@@ -523,6 +525,8 @@
 ; GCN-O1-OPTS-NEXT:      Scalarize Masked Memory Intrinsics
 ; GCN-O1-OPTS-NEXT:      Expand reduction intrinsics
 ; GCN-O1-OPTS-NEXT:      Early CSE
+; GCN-O1-OPTS-NEXT:    AMDGPU Preload Kernel Arguments
----------------
krzysz00 wrote:

Quick side question: does this pass need to run before the main CodeGenPrepare?

If not ... I have a plan to move AMDGPULowerBufferFatPointers after LoadStoreVectorizer, and, given the desire for buffer pointer alignment reform, might want to move kernel argument lowering after that so we can see function arguments as themselves.

Not exactly relevant or blocking, just thought I'd check on this

https://github.com/llvm/llvm-project/pull/130434