1;RUN: llc < %s -march=amdgcn -mcpu=verde -verify-machineinstrs | FileCheck %s -check-prefix=CHECK -check-prefix=SICI
2;RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs | FileCheck %s -check-prefix=CHECK -check-prefix=VI
3
4;CHECK-LABEL: {{^}}buffer_load:
5;CHECK: buffer_load_dwordx4 v[0:3], {{v[0-9]+}}, s[0:3], 0 idxen
6;CHECK: buffer_load_dwordx4 v[4:7], {{v[0-9]+}}, s[0:3], 0 idxen glc
7;CHECK: buffer_load_dwordx4 v[8:11], {{v[0-9]+}}, s[0:3], 0 idxen slc
8;CHECK: s_waitcnt
9define amdgpu_ps {<4 x float>, <4 x float>, <4 x float>} @buffer_load(<4 x i32> inreg) {
10main_body:
11  %data = call <4 x float> @llvm.amdgcn.struct.buffer.load.v4f32(<4 x i32> %0, i32 0, i32 0, i32 0, i32 0)
12  %data_glc = call <4 x float> @llvm.amdgcn.struct.buffer.load.v4f32(<4 x i32> %0, i32 0, i32 0, i32 0, i32 1)
13  %data_slc = call <4 x float> @llvm.amdgcn.struct.buffer.load.v4f32(<4 x i32> %0, i32 0, i32 0, i32 0, i32 2)
14  %r0 = insertvalue {<4 x float>, <4 x float>, <4 x float>} undef, <4 x float> %data, 0
15  %r1 = insertvalue {<4 x float>, <4 x float>, <4 x float>} %r0, <4 x float> %data_glc, 1
16  %r2 = insertvalue {<4 x float>, <4 x float>, <4 x float>} %r1, <4 x float> %data_slc, 2
17  ret {<4 x float>, <4 x float>, <4 x float>} %r2
18}
19
20;CHECK-LABEL: {{^}}buffer_load_immoffs:
21;CHECK: buffer_load_dwordx4 v[0:3], {{v[0-9]+}}, s[0:3], 0 idxen offset:40
22;CHECK: s_waitcnt
23define amdgpu_ps <4 x float> @buffer_load_immoffs(<4 x i32> inreg) {
24main_body:
25  %data = call <4 x float> @llvm.amdgcn.struct.buffer.load.v4f32(<4 x i32> %0, i32 0, i32 40, i32 0, i32 0)
26  ret <4 x float> %data
27}
28
29;CHECK-LABEL: {{^}}buffer_load_immoffs_large:
30;CHECK: s_movk_i32 [[OFFSET:s[0-9]+]], 0x1ffc
31;CHECK: buffer_load_dwordx4 v[0:3], {{v[0-9]+}}, s[0:3], [[OFFSET]] idxen offset:4
32;CHECK: s_waitcnt
33define amdgpu_ps <4 x float> @buffer_load_immoffs_large(<4 x i32> inreg) {
34main_body:
35  %data = call <4 x float> @llvm.amdgcn.struct.buffer.load.v4f32(<4 x i32> %0, i32 0, i32 4, i32 8188, i32 0)
36  ret <4 x float> %data
37}
38
39;CHECK-LABEL: {{^}}buffer_load_idx:
40;CHECK: buffer_load_dwordx4 v[0:3], v0, s[0:3], 0 idxen
41;CHECK: s_waitcnt
42define amdgpu_ps <4 x float> @buffer_load_idx(<4 x i32> inreg, i32) {
43main_body:
44  %data = call <4 x float> @llvm.amdgcn.struct.buffer.load.v4f32(<4 x i32> %0, i32 %1, i32 0, i32 0, i32 0)
45  ret <4 x float> %data
46}
47
48;CHECK-LABEL: {{^}}buffer_load_ofs:
49;CHECK: buffer_load_dwordx4 v[0:3], v[0:1], s[0:3], 0 idxen offen
50;CHECK: s_waitcnt
51define amdgpu_ps <4 x float> @buffer_load_ofs(<4 x i32> inreg, i32) {
52main_body:
53  %data = call <4 x float> @llvm.amdgcn.struct.buffer.load.v4f32(<4 x i32> %0, i32 0, i32 %1, i32 0, i32 0)
54  ret <4 x float> %data
55}
56
57;CHECK-LABEL: {{^}}buffer_load_ofs_imm:
58;CHECK: buffer_load_dwordx4 v[0:3], v[0:1], s[0:3], 0 idxen offen offset:60
59;CHECK: s_waitcnt
60define amdgpu_ps <4 x float> @buffer_load_ofs_imm(<4 x i32> inreg, i32) {
61main_body:
62  %ofs = add i32 %1, 60
63  %data = call <4 x float> @llvm.amdgcn.struct.buffer.load.v4f32(<4 x i32> %0, i32 0, i32 %ofs, i32 0, i32 0)
64  ret <4 x float> %data
65}
66
67;CHECK-LABEL: {{^}}buffer_load_both:
68;CHECK: buffer_load_dwordx4 v[0:3], v[0:1], s[0:3], 0 idxen offen
69;CHECK: s_waitcnt
70define amdgpu_ps <4 x float> @buffer_load_both(<4 x i32> inreg, i32, i32) {
71main_body:
72  %data = call <4 x float> @llvm.amdgcn.struct.buffer.load.v4f32(<4 x i32> %0, i32 %1, i32 %2, i32 0, i32 0)
73  ret <4 x float> %data
74}
75
76;CHECK-LABEL: {{^}}buffer_load_both_reversed:
77;CHECK: v_mov_b32_e32 v2, v0
78;CHECK: buffer_load_dwordx4 v[0:3], v[1:2], s[0:3], 0 idxen offen
79;CHECK: s_waitcnt
80define amdgpu_ps <4 x float> @buffer_load_both_reversed(<4 x i32> inreg, i32, i32) {
81main_body:
82  %data = call <4 x float> @llvm.amdgcn.struct.buffer.load.v4f32(<4 x i32> %0, i32 %2, i32 %1, i32 0, i32 0)
83  ret <4 x float> %data
84}
85
86;CHECK-LABEL: {{^}}buffer_load_x1:
87;CHECK: buffer_load_dword v0, v[0:1], s[0:3], 0 idxen offen
88;CHECK: s_waitcnt
89define amdgpu_ps float @buffer_load_x1(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs) {
90main_body:
91  %data = call float @llvm.amdgcn.struct.buffer.load.f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 0, i32 0)
92  ret float %data
93}
94
95;CHECK-LABEL: {{^}}buffer_load_x2:
96;CHECK: buffer_load_dwordx2 v[0:1], v[0:1], s[0:3], 0 idxen offen
97;CHECK: s_waitcnt
98define amdgpu_ps <2 x float> @buffer_load_x2(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs) {
99main_body:
100  %data = call <2 x float> @llvm.amdgcn.struct.buffer.load.v2f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 0, i32 0)
101  ret <2 x float> %data
102}
103
104;CHECK-LABEL: {{^}}buffer_load_negative_offset:
105;CHECK: v_add_{{[iu]}}32_e32 {{v[0-9]+}}, vcc, 0xfffff000, v0
106;CHECK: buffer_load_dwordx4 v[0:3], {{v\[[0-9]+:[0-9]+\]}}, s[0:3], 0 idxen offen offset:4080
107define amdgpu_ps <4 x float> @buffer_load_negative_offset(<4 x i32> inreg, i32 %ofs) {
108main_body:
109  %ofs.1 = add i32 %ofs, -16
110  %data = call <4 x float> @llvm.amdgcn.struct.buffer.load.v4f32(<4 x i32> %0, i32 0, i32 %ofs.1, i32 0, i32 0)
111  ret <4 x float> %data
112}
113
114; SI won't merge ds memory operations, because of the signed offset bug, so
115; we only have check lines for VI.
116; CHECK-LABEL: buffer_load_mmo:
117; VI: v_mov_b32_e32 [[ZERO:v[0-9]+]], 0
118; VI: ds_write2_b32 v{{[0-9]+}}, [[ZERO]], [[ZERO]] offset1:4
119define amdgpu_ps float @buffer_load_mmo(<4 x i32> inreg %rsrc, float addrspace(3)* %lds) {
120entry:
121  store float 0.0, float addrspace(3)* %lds
122  %val = call float @llvm.amdgcn.struct.buffer.load.f32(<4 x i32> %rsrc, i32 0, i32 0, i32 0, i32 0)
123  %tmp2 = getelementptr float, float addrspace(3)* %lds, i32 4
124  store float 0.0, float addrspace(3)* %tmp2
125  ret float %val
126}
127
128;CHECK-LABEL: {{^}}buffer_load_int:
129;CHECK: buffer_load_dwordx4 v[0:3], {{v[0-9]+}}, s[0:3], 0 idxen
130;CHECK: buffer_load_dwordx2 v[4:5], {{v[0-9]+}}, s[0:3], 0 idxen glc
131;CHECK: buffer_load_dword v6, {{v[0-9]+}}, s[0:3], 0 idxen slc
132;CHECK: s_waitcnt
133define amdgpu_ps {<4 x float>, <2 x float>, float} @buffer_load_int(<4 x i32> inreg) {
134main_body:
135  %data = call <4 x i32> @llvm.amdgcn.struct.buffer.load.v4i32(<4 x i32> %0, i32 0, i32 0, i32 0, i32 0)
136  %data_glc = call <2 x i32> @llvm.amdgcn.struct.buffer.load.v2i32(<4 x i32> %0, i32 0, i32 0, i32 0, i32 1)
137  %data_slc = call i32 @llvm.amdgcn.struct.buffer.load.i32(<4 x i32> %0, i32 0, i32 0, i32 0, i32 2)
138  %fdata = bitcast <4 x i32> %data to <4 x float>
139  %fdata_glc = bitcast <2 x i32> %data_glc to <2 x float>
140  %fdata_slc = bitcast i32 %data_slc to float
141  %r0 = insertvalue {<4 x float>, <2 x float>, float} undef, <4 x float> %fdata, 0
142  %r1 = insertvalue {<4 x float>, <2 x float>, float} %r0, <2 x float> %fdata_glc, 1
143  %r2 = insertvalue {<4 x float>, <2 x float>, float} %r1, float %fdata_slc, 2
144  ret {<4 x float>, <2 x float>, float} %r2
145}
146
147declare float @llvm.amdgcn.struct.buffer.load.f32(<4 x i32>, i32, i32, i32, i32) #0
148declare <2 x float> @llvm.amdgcn.struct.buffer.load.v2f32(<4 x i32>, i32, i32, i32, i32) #0
149declare <4 x float> @llvm.amdgcn.struct.buffer.load.v4f32(<4 x i32>, i32, i32, i32, i32) #0
150declare i32 @llvm.amdgcn.struct.buffer.load.i32(<4 x i32>, i32, i32, i32, i32) #0
151declare <2 x i32> @llvm.amdgcn.struct.buffer.load.v2i32(<4 x i32>, i32, i32, i32, i32) #0
152declare <4 x i32> @llvm.amdgcn.struct.buffer.load.v4i32(<4 x i32>, i32, i32, i32, i32) #0
153declare void @llvm.amdgcn.exp.f32(i32, i32, float, float, float, float, i1, i1) #0
154
155attributes #0 = { nounwind readonly }
156