[llvm-branch-commits] [llvm] [AMDGPU][SwLowerLDS] Lower non-kernels with LDS instructions and assign kernel IDs to their callers (PR #228043)

[via llvm-branch-commits]

================
@@ -1073,13 +1107,18 @@ void AMDGPUSwLowerLDS::buildNonKernelLDSBaseTable(
   if (Kernels.empty())
     return;
   const size_t NumberKernels = Kernels.size();
-  ArrayType *AllKernelsOffsetsType =
-      ArrayType::get(IRB.getPtrTy(AMDGPUAS::LOCAL_ADDRESS), NumberKernels);
+  PointerType *LDSPtrTy = IRB.getPtrTy(AMDGPUAS::LOCAL_ADDRESS);
+  ArrayType *AllKernelsOffsetsType = ArrayType::get(LDSPtrTy, NumberKernels);
   std::vector<Constant *> OverallConstantExprElts(NumberKernels);
   for (size_t i = 0; i < NumberKernels; i++) {
-    Function *Func = Kernels[i];
-    auto &LDSParams = FuncLDSAccessInfo.KernelToLDSParametersMap[Func];
-    OverallConstantExprElts[i] = LDSParams.SwLDS;
+    // Null for kernels without LDS, they reach LDS instructions only on
+    // invalid paths.
----------------
skc7 wrote:

Updated the comment. Intended, the result is unused on the flat path.

https://github.com/llvm/llvm-project/pull/228043