AMDGPU: Mark all unspecified CC functions in tests as amdgpu_kernel

Currently the default C calling convention functions are treated
the same as compute kernels. Make this explicit so the default
calling convention can be changed to a non-kernel.

Converted with perl -pi -e 's/define void/define amdgpu_kernel void/'
on the relevant test directories (and undoing in one place that actually
wanted a non-kernel).

llvm-svn: 298444
diff --git a/llvm/test/CodeGen/AMDGPU/atomic_load_sub.ll b/llvm/test/CodeGen/AMDGPU/atomic_load_sub.ll
index c6e5b11..a027589 100644
--- a/llvm/test/CodeGen/AMDGPU/atomic_load_sub.ll
+++ b/llvm/test/CodeGen/AMDGPU/atomic_load_sub.ll
@@ -5,7 +5,7 @@
 ; FUNC-LABEL: {{^}}atomic_sub_local:
 ; R600: LDS_SUB *
 ; SI: ds_sub_u32
-define void @atomic_sub_local(i32 addrspace(3)* %local) {
+define amdgpu_kernel void @atomic_sub_local(i32 addrspace(3)* %local) {
    %unused = atomicrmw volatile sub i32 addrspace(3)* %local, i32 5 seq_cst
    ret void
 }
@@ -13,7 +13,7 @@
 ; FUNC-LABEL: {{^}}atomic_sub_local_const_offset:
 ; R600: LDS_SUB *
 ; SI: ds_sub_u32 v{{[0-9]+}}, v{{[0-9]+}} offset:16
-define void @atomic_sub_local_const_offset(i32 addrspace(3)* %local) {
+define amdgpu_kernel void @atomic_sub_local_const_offset(i32 addrspace(3)* %local) {
   %gep = getelementptr i32, i32 addrspace(3)* %local, i32 4
   %val = atomicrmw volatile sub i32 addrspace(3)* %gep, i32 5 seq_cst
   ret void
@@ -22,7 +22,7 @@
 ; FUNC-LABEL: {{^}}atomic_sub_ret_local:
 ; R600: LDS_SUB_RET *
 ; SI: ds_sub_rtn_u32
-define void @atomic_sub_ret_local(i32 addrspace(1)* %out, i32 addrspace(3)* %local) {
+define amdgpu_kernel void @atomic_sub_ret_local(i32 addrspace(1)* %out, i32 addrspace(3)* %local) {
   %val = atomicrmw volatile sub i32 addrspace(3)* %local, i32 5 seq_cst
   store i32 %val, i32 addrspace(1)* %out
   ret void
@@ -31,7 +31,7 @@
 ; FUNC-LABEL: {{^}}atomic_sub_ret_local_const_offset:
 ; R600: LDS_SUB_RET *
 ; SI: ds_sub_rtn_u32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:20
-define void @atomic_sub_ret_local_const_offset(i32 addrspace(1)* %out, i32 addrspace(3)* %local) {
+define amdgpu_kernel void @atomic_sub_ret_local_const_offset(i32 addrspace(1)* %out, i32 addrspace(3)* %local) {
   %gep = getelementptr i32, i32 addrspace(3)* %local, i32 5
   %val = atomicrmw volatile sub i32 addrspace(3)* %gep, i32 5 seq_cst
   store i32 %val, i32 addrspace(1)* %out