[flang][cuda] Fix GPULaunchKernelConversion to generate correct kernel launch parameters (#119431)

For the call to _FortranACUFLaunchKernel, we store the pointer to a
member of a temporary structure in a parameter array. However, when we
obtain an element pointer from the parameter array, its address is
calculated based on the type of the structure. This PR properly treats
the parameter array as an array of pointers.

Example:

```mlir
%30 = llvm.load %29 : !llvm.ptr -> i32
%31 = llvm.mlir.constant(1 : i32) : i32
%32 = llvm.alloca %31 x !llvm.struct<(i64, i64, i32, ptr)> : (i32) -> !llvm.ptr
%33 = llvm.mlir.constant(4 : i32) : i32
%34 = llvm.alloca %33 x !llvm.ptr : (i32) -> !llvm.ptr
%35 = llvm.mlir.constant(0 : i32) : i32
%36 = llvm.getelementptr %32[%35] : (!llvm.ptr, i32) -> !llvm.ptr, !llvm.struct<(i64, i64, i32, ptr)>
llvm.store %8, %36 : i64, !llvm.ptr
%37 = llvm.getelementptr %34[%35] : (!llvm.ptr, i32) -> !llvm.ptr, !llvm.struct<(i64, i64, i32, ptr)>
llvm.store %36, %37 : !llvm.ptr, !llvm.ptr
...
llvm.call @_FortranACUFLaunchKernel(%47, %8, %8, %8, %2, %8, %8, %7, %34, %48) : (!llvm.ptr, i64, i64, i64, i64, i64, i64, i32, !llvm.ptr, !llvm.ptr) -> () 
```
In this example, `%37 = llvm.getelementptr %34[%35] : (!llvm.ptr, i32)
-> !llvm.ptr, !llvm.struct<(i64, i64, i32, ptr)>` will be `%37 =
llvm.getelementptr %34[%35] : (!llvm.ptr, i32) -> !llvm.ptr, !llvm.ptr`.

Author: khaki3

flang/lib/Optimizer/Transforms/CUFGPUToLLVMConversion.cpp

@@ -58,7 +58,7 @@ static mlir::Value createKernelArgArray(mlir::Location loc,
         loc, ptrTy, structTy, argStruct, mlir::ArrayRef<mlir::Value>({indice}));
     rewriter.create<LLVM::StoreOp>(loc, arg, structMember);
     mlir::Value arrayMember = rewriter.create<LLVM::GEPOp>(
-        loc, ptrTy, structTy, argArray, mlir::ArrayRef<mlir::Value>({indice}));
+        loc, ptrTy, ptrTy, argArray, mlir::ArrayRef<mlir::Value>({indice}));
     rewriter.create<LLVM::StoreOp>(loc, structMember, arrayMember);
   }
   return argArray;

flang/test/Fir/CUDA/cuda-gpu-launch-func.mlir

@@ -99,9 +99,16 @@ module attributes {dlti.dl_spec = #dlti.dl_spec<#dlti.dl_entry<i1, dense<8> : ve
 }
 
 // CHECK-LABEL: _QMmod1Phost_sub
-
+// CHECK: %[[STRUCT:.*]] = llvm.alloca %{{.*}} x !llvm.struct<(ptr)> : (i32) -> !llvm.ptr
+// CHECK: %[[PARAMS:.*]] = llvm.alloca %{{.*}} x !llvm.ptr : (i32) -> !llvm.ptr
+// CHECK: %[[ZERO:.*]] = llvm.mlir.constant(0 : i32) : i32
+// CHECK: %[[STRUCT_PTR:.*]] = llvm.getelementptr %[[STRUCT]][%[[ZERO]]] : (!llvm.ptr, i32) -> !llvm.ptr, !llvm.struct<(ptr)>
+// CHECK: llvm.store %{{.*}}, %[[STRUCT_PTR]] : !llvm.ptr, !llvm.ptr
+// CHECK: %[[PARAM_PTR:.*]] = llvm.getelementptr %[[PARAMS]][%[[ZERO]]] : (!llvm.ptr, i32) -> !llvm.ptr, !llvm.ptr
+// CHECK: llvm.store %[[STRUCT_PTR]], %[[PARAM_PTR]] : !llvm.ptr, !llvm.ptr
 // CHECK: %[[KERNEL_PTR:.*]] = llvm.mlir.addressof @_QMmod1Psub1 : !llvm.ptr
-// CHECK: llvm.call @_FortranACUFLaunchKernel(%[[KERNEL_PTR]], {{.*}})
+// CHECK: %[[NULL:.*]] = llvm.mlir.zero : !llvm.ptr
+// CHECK: llvm.call @_FortranACUFLaunchKernel(%[[KERNEL_PTR]], {{.*}}, %[[PARAMS]], %[[NULL]])
 
 // -----