[llvm] [AMDGPU][DAG] Enable ganging up of memcpy loads/stores for AMDGPU (PR #96185)

[Fabian Ritter via llvm-commits]

================
@@ -67,6 +67,9 @@ AMDGPUTargetLowering::AMDGPUTargetLowering(const TargetMachine &TM,
   MaxStoresPerMemcpy = MaxStoresPerMemcpyOptSize = ~0U;
   MaxStoresPerMemmove = MaxStoresPerMemmoveOptSize = ~0U;
 
+  // Enable ganging up loads and stores in the memcpy DAG lowering.
----------------
ritter-x2a wrote:

Looking more into the code, I noticed that the DAG lowering of memcpy only takes the smaller one when different alignments are specified for source and destination. So, there is no point in testing combinations of different alignments, which leaves us with ~20k lines of test cases (the now updated state in this PR).

Do you have objections against merging the PR like this, @arsenm ?

https://github.com/llvm/llvm-project/pull/96185