[llvm] [AMDGPU] Use table strategy for LowerModuleLDSPass at O0 (PR #160181)

[Matt Arsenault via llvm-commits]

================
@@ -1188,13 +1191,17 @@ class AMDGPULowerModuleLDS {
           // Allocated at zero, recorded once on construction, not once per
           // kernel
           Offset += DL.getTypeAllocSize(MaybeModuleScopeStruct->getValueType());
+          LLVM_DEBUG(dbgs() << "amdgpu-lds-size after ModuleScopeStruct"
+                            << Offset << "\n");
         }
 
         if (AllocateKernelScopeStruct) {
           GlobalVariable *KernelStruct = Replacement->second.SGV;
           Offset = alignTo(Offset, AMDGPU::getAlign(DL, KernelStruct));
           recordLDSAbsoluteAddress(&M, KernelStruct, Offset);
           Offset += DL.getTypeAllocSize(KernelStruct->getValueType());
+          LLVM_DEBUG(dbgs()
+                     << "amdgpu-lds-size after KernelStruct" << Offset << "\n");
----------------
arsenm wrote:

```suggestion
                     << "amdgpu-lds-size after KernelStruct" << Offset << '\n');
```

https://github.com/llvm/llvm-project/pull/160181