[clang] [Clang][OpenCL][AMDGPU] Allow a kernel to call another kernel (PR #115821)

[Aniket Lal via cfe-commits]

================
@@ -0,0 +1,43 @@
+// RUN: %clang_cc1 -triple amdgcn-unknown-unknown -emit-llvm -o - %s | FileCheck %s
+
+// CHECK: define dso_local amdgpu_kernel void @callee_kern({{.*}})
+__attribute__((noinline)) kernel void callee_kern(global int *A){
+  *A = 1;
+}
+
+__attribute__((noinline)) kernel void ext_callee_kern(global int *A);
+
+// CHECK: define dso_local void @callee_func({{.*}})
+__attribute__((noinline)) void callee_func(global int *A){
+  *A = 2;
+}
+
+// CHECK: define dso_local amdgpu_kernel void @caller_kern({{.*}})
+kernel void caller_kern(global int* A){
+  callee_kern(A);
+  // CHECK: tail call void @__clang_ocl_kern_imp_callee_kern({{.*}})
+  ext_callee_kern(A);
+  // CHECK: tail call void @__clang_ocl_kern_imp_ext_callee_kern({{.*}})
+  callee_func(A);
+  // CHECK: tail call void @callee_func({{.*}})
+
+}
+
+// CHECK: define dso_local void @__clang_ocl_kern_imp_callee_kern({{.*}})
+
+// CHECK: declare void @__clang_ocl_kern_imp_ext_callee_kern({{.*}})
+
+// CHECK: define dso_local void @caller_func({{.*}})
+void caller_func(global int* A){
+  callee_kern(A);
+  // CHECK: tail call void @__clang_ocl_kern_imp_callee_kern({{.*}}) #7
+  ext_callee_kern(A);
+  // CHECK: tail call void @__clang_ocl_kern_imp_ext_callee_kern({{.*}}) #8
+  callee_func(A);
+  // CHECK: tail call void @callee_func({{.*}})
+}
+
----------------
lalaniket8 wrote:

[addr-space-struct-arg.cl](https://github.com/llvm/llvm-project/blob/12d17535057faf07b39fe2f59b3952f7b817b780/clang/test/CodeGenOpenCL/addr-space-struct-arg.cl), [amdgpu-abi-struct-arg-byref.cl](https://github.com/llvm/llvm-project/blob/12d17535057faf07b39fe2f59b3952f7b817b780/clang/test/CodeGenOpenCL/amdgpu-abi-struct-arg-byref.cl) tests for different aggregates passed byref and byval.

Do let me know if we need to account for some other cases not in these littests.

https://github.com/llvm/llvm-project/pull/115821