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

Matt Arsenault via llvm-commits llvm-commits at lists.llvm.org
Tue Jun 25 07:52:10 PDT 2024


================
@@ -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.
----------------
arsenm wrote:

Testing 8 and 16 alignments will at least show some LDS differences. We can probably get away without the addrspace 4 cases, unless they are specially crafted to use uniform source pointers 

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


More information about the llvm-commits mailing list