1; RUN: llc < %s -march=nvptx -mcpu=sm_20 -verify-machineinstrs | FileCheck %s
2
3target triple = "nvptx-unknown-cuda"
4
5declare void @bar(<4 x i32>)
6
7; CHECK-LABEL: .func foo(
8; CHECK-DAG: ld.param.v4.u32 {[[E0:%r[0-9]+]], [[E1:%r[0-9]+]], [[E2:%r[0-9]+]], [[E3:%r[0-9]+]]}, [foo_param_0];
9; CHECK: .param .align 16 .b8 param0[16];
10; CHECK-DAG: st.param.v4.b32  [param0+0],  {[[E0]], [[E1]], [[E2]], [[E3]]};
11; CHECK:     call.uni
12; CHECK:     ret;
13define void @foo(<4 x i32> %a) {
14  tail call void @bar(<4 x i32> %a)
15  ret void
16}
17
18; CHECK-LABEL: .func foo3(
19; CHECK-DAG: ld.param.v2.u32 {[[E0:%r[0-9]+]], [[E1:%r[0-9]+]]}, [foo3_param_0];
20; CHECK-DAG: ld.param.u32 [[E2:%r[0-9]+]], [foo3_param_0+8];
21; CHECK: .param .align 16 .b8 param0[16];
22; CHECK-DAG: st.param.v2.b32  [param0+0],  {[[E0]], [[E1]]};
23; CHECK-DAG: st.param.b32     [param0+8],  [[E2]];
24; CHECK:     call.uni
25; CHECK:     ret;
26declare void @bar3(<3 x i32>)
27define void @foo3(<3 x i32> %a) {
28  tail call void @bar3(<3 x i32> %a)
29  ret void
30}
31