1; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -strict-whitespace %s --check-prefix=DEFAULT
2; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -strict-whitespace %s --check-prefix=DEFAULT
3; RUN: llc -march=amdgcn --misched=ilpmax -verify-machineinstrs < %s | FileCheck -strict-whitespace %s --check-prefix=ILPMAX
4; RUN: llc -march=amdgcn --misched=ilpmax -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -strict-whitespace %s --check-prefix=ILPMAX
5; The ilpmax scheduler is used for the second test to get the ordering we want for the test.
6
7; DEFAULT-LABEL: {{^}}main:
8; DEFAULT: s_load_dwordx4
9; DEFAULT: s_load_dwordx4
10; DEFAULT: s_waitcnt vmcnt(0)
11; DEFAULT: exp
12; DEFAULT: s_waitcnt lgkmcnt(0)
13; DEFAULT: s_endpgm
14define amdgpu_vs void @main(<16 x i8> addrspace(2)* inreg %arg, <16 x i8> addrspace(2)* inreg %arg1, <32 x i8> addrspace(2)* inreg %arg2, <16 x i8> addrspace(2)* inreg %arg3, <16 x i8> addrspace(2)* inreg %arg4, i32 inreg %arg5, i32 %arg6, i32 %arg7, i32 %arg8, i32 %arg9, float addrspace(2)* inreg %constptr) #0 {
15main_body:
16  %tmp = getelementptr <16 x i8>, <16 x i8> addrspace(2)* %arg3, i32 0
17  %tmp10 = load <16 x i8>, <16 x i8> addrspace(2)* %tmp, !tbaa !0
18  %tmp11 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %tmp10, i32 0, i32 %arg6)
19  %tmp12 = extractelement <4 x float> %tmp11, i32 0
20  %tmp13 = extractelement <4 x float> %tmp11, i32 1
21  call void @llvm.amdgcn.s.barrier() #1
22  %tmp14 = extractelement <4 x float> %tmp11, i32 2
23  %tmp15 = load float, float addrspace(2)* %constptr, align 4
24  %tmp16 = getelementptr <16 x i8>, <16 x i8> addrspace(2)* %arg3, i32 1
25  %tmp17 = load <16 x i8>, <16 x i8> addrspace(2)* %tmp16, !tbaa !0
26  %tmp18 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %tmp17, i32 0, i32 %arg6)
27  %tmp19 = extractelement <4 x float> %tmp18, i32 0
28  %tmp20 = extractelement <4 x float> %tmp18, i32 1
29  %tmp21 = extractelement <4 x float> %tmp18, i32 2
30  %tmp22 = extractelement <4 x float> %tmp18, i32 3
31  call void @llvm.amdgcn.exp.f32(i32 32, i32 15, float %tmp19, float %tmp20, float %tmp21, float %tmp22, i1 false, i1 false) #0
32  call void @llvm.amdgcn.exp.f32(i32 12, i32 15, float %tmp12, float %tmp13, float %tmp14, float %tmp15, i1 true, i1 false) #0
33  ret void
34}
35
36; ILPMAX-LABEL: {{^}}main2:
37; ILPMAX: s_load_dwordx4
38; ILPMAX: s_waitcnt lgkmcnt(0)
39; ILPMAX: buffer_load
40; ILPMAX: s_load_dwordx4
41; ILPMAX: s_waitcnt lgkmcnt(0)
42; ILPMAX: buffer_load
43; ILPMAX: s_waitcnt vmcnt(1)
44; ILPMAX: s_waitcnt vmcnt(0)
45; ILPMAX: s_endpgm
46define amdgpu_vs void @main2([6 x <16 x i8>] addrspace(2)* byval %arg, [17 x <16 x i8>] addrspace(2)* byval %arg1, [17 x <4 x i32>] addrspace(2)* byval %arg2, [34 x <8 x i32>] addrspace(2)* byval %arg3, [16 x <16 x i8>] addrspace(2)* byval %arg4, i32 inreg %arg5, i32 inreg %arg6, i32 %arg7, i32 %arg8, i32 %arg9, i32 %arg10) #0 {
47main_body:
48  %tmp = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %arg4, i64 0, i64 0
49  %tmp11 = load <16 x i8>, <16 x i8> addrspace(2)* %tmp, align 16, !tbaa !0
50  %tmp12 = add i32 %arg5, %arg7
51  %tmp13 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %tmp11, i32 0, i32 %tmp12)
52  %tmp14 = extractelement <4 x float> %tmp13, i32 0
53  %tmp15 = extractelement <4 x float> %tmp13, i32 1
54  %tmp16 = extractelement <4 x float> %tmp13, i32 2
55  %tmp17 = extractelement <4 x float> %tmp13, i32 3
56  %tmp18 = getelementptr [16 x <16 x i8>], [16 x <16 x i8>] addrspace(2)* %arg4, i64 0, i64 1
57  %tmp19 = load <16 x i8>, <16 x i8> addrspace(2)* %tmp18, align 16, !tbaa !0
58  %tmp20 = add i32 %arg5, %arg7
59  %tmp21 = call <4 x float> @llvm.SI.vs.load.input(<16 x i8> %tmp19, i32 0, i32 %tmp20)
60  %tmp22 = extractelement <4 x float> %tmp21, i32 0
61  %tmp23 = extractelement <4 x float> %tmp21, i32 1
62  %tmp24 = extractelement <4 x float> %tmp21, i32 2
63  %tmp25 = extractelement <4 x float> %tmp21, i32 3
64  call void @llvm.amdgcn.exp.f32(i32 12, i32 15, float %tmp14, float %tmp15, float %tmp16, float %tmp17, i1 true, i1 false) #0
65  call void @llvm.amdgcn.exp.f32(i32 32, i32 15, float %tmp22, float %tmp23, float %tmp24, float %tmp25, i1 false, i1 false) #0
66  ret void
67}
68
69declare void @llvm.amdgcn.s.barrier() #1
70declare <4 x float> @llvm.SI.vs.load.input(<16 x i8>, i32, i32) #2
71declare void @llvm.amdgcn.exp.f32(i32, i32, float, float, float, float, i1, i1) #0
72
73attributes #0 = { nounwind }
74attributes #1 = { convergent nounwind }
75attributes #2 = { nounwind readnone }
76
77!0 = !{!1, !1, i64 0, i32 1}
78!1 = !{!"const", !2}
79!2 = !{!"tbaa root"}
80