[llvm] [AMDGPU] Run early-cse<memssa> at the end of the full-LTO pipeline (PR #208461)

[via llvm-commits]

================
@@ -1111,6 +1111,17 @@ void AMDGPUTargetMachine::registerPassBuilderCallbacks(PassBuilder &PB) {
 
   PB.registerFullLinkTimeOptimizationLastEPCallback(
       [this](ModulePassManager &PM, OptimizationLevel Level) {
+        // The regular (non-LTO) and ThinLTO function simplification pipelines
+        // run early-cse with MemorySSA near the start of the function pass
+        // sequence, but the full-LTO postlink pipeline does not. Without it a
+        // redundant load/store round-trip can survive to codegen (the later
+        // GVN/DSE do not catch this pattern), which costs extra hazard s_nop
+        // padding in some kernels. Run it here, at the end of the full-LTO
+        // middle-end, before codegen.
----------------
michaelselehov wrote:

Sure. Already been doing that. Sorry for the late notice :)

https://github.com/llvm/llvm-project/pull/208461