1; RUN: llc -march=amdgcn -verify-machineinstrs -amdgpu-scalar-ir-passes=false < %s | FileCheck -check-prefix=GCN -check-prefix=SICIVI -check-prefix=SI %s
2; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs -amdgpu-scalar-ir-passes=false < %s | FileCheck -check-prefix=GCN -check-prefix=SICIVI -check-prefix=VI %s
3; RUN: llc -march=amdgcn -mcpu=gfx900 -mattr=-flat-for-global -verify-machineinstrs -amdgpu-scalar-ir-passes=false < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 %s
4
5; These tests are split out from umed3.ll and smed3.ll and use the
6; -amdgpu-scalar-ir-passes=false flag, because InstSimplify would constant
7; fold these functions otherwise.
8
9declare i32 @llvm.amdgcn.workitem.id.x() nounwind readnone
10
11; GCN-LABEL: {{^}}v_test_umed3_r_i_i_constant_order_i32:
12; GCN: v_max_u32_e32 v{{[0-9]+}}, 17, v{{[0-9]+}}
13; GCN: v_min_u32_e32 v{{[0-9]+}}, 12, v{{[0-9]+}}
14define amdgpu_kernel void @v_test_umed3_r_i_i_constant_order_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %aptr) #1 {
15  %tid = call i32 @llvm.amdgcn.workitem.id.x()
16  %gep0 = getelementptr i32, i32 addrspace(1)* %aptr, i32 %tid
17  %outgep = getelementptr i32, i32 addrspace(1)* %out, i32 %tid
18  %a = load i32, i32 addrspace(1)* %gep0
19
20  %icmp0 = icmp ugt i32 %a, 17
21  %i0 = select i1 %icmp0, i32 %a, i32 17
22
23  %icmp1 = icmp ult i32 %i0, 12
24  %i1 = select i1 %icmp1, i32 %i0, i32 12
25
26  store i32 %i1, i32 addrspace(1)* %outgep
27  ret void
28}
29
30; GCN-LABEL: {{^}}v_test_smed3_r_i_i_constant_order_i32:
31; GCN: v_max_i32_e32 v{{[0-9]+}}, 17, v{{[0-9]+}}
32; GCN: v_min_i32_e32 v{{[0-9]+}}, 12, v{{[0-9]+}}
33define amdgpu_kernel void @v_test_smed3_r_i_i_constant_order_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %aptr) #1 {
34  %tid = call i32 @llvm.amdgcn.workitem.id.x()
35  %gep0 = getelementptr i32, i32 addrspace(1)* %aptr, i32 %tid
36  %outgep = getelementptr i32, i32 addrspace(1)* %out, i32 %tid
37  %a = load i32, i32 addrspace(1)* %gep0
38
39  %icmp0 = icmp sgt i32 %a, 17
40  %i0 = select i1 %icmp0, i32 %a, i32 17
41
42  %icmp1 = icmp slt i32 %i0, 12
43  %i1 = select i1 %icmp1, i32 %i0, i32 12
44
45  store i32 %i1, i32 addrspace(1)* %outgep
46  ret void
47}
48
49