1; NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
2; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=tonga -stop-after=instruction-select -o - %s | FileCheck %s
3; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx810 -stop-after=instruction-select -o - %s | FileCheck %s
4
5define amdgpu_ps void @raw_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_f32(<4 x i32> inreg %rsrc, float %val, i32 %voffset, i32 inreg %soffset) {
6  ; CHECK-LABEL: name: raw_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_f32
7  ; CHECK: bb.1 (%ir-block.0):
8  ; CHECK-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1
9  ; CHECK-NEXT: {{  $}}
10  ; CHECK-NEXT:   [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
11  ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
12  ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
13  ; CHECK-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
14  ; CHECK-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
15  ; CHECK-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
16  ; CHECK-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
17  ; CHECK-NEXT:   [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr6
18  ; CHECK-NEXT:   BUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY4]], [[COPY5]], [[REG_SEQUENCE]], [[COPY6]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
19  ; CHECK-NEXT:   S_ENDPGM 0
20  call void @llvm.amdgcn.raw.buffer.store.format.f32(float %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 0)
21  ret void
22}
23
24define amdgpu_ps void @raw_buffer_store_format__sgpr_rsrc__vgpr_val__voffset_4095__sgpr_soffset_f32(<4 x i32> inreg %rsrc, float %val, i32 inreg %soffset) {
25  ; CHECK-LABEL: name: raw_buffer_store_format__sgpr_rsrc__vgpr_val__voffset_4095__sgpr_soffset_f32
26  ; CHECK: bb.1 (%ir-block.0):
27  ; CHECK-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0
28  ; CHECK-NEXT: {{  $}}
29  ; CHECK-NEXT:   [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
30  ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
31  ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
32  ; CHECK-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
33  ; CHECK-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
34  ; CHECK-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
35  ; CHECK-NEXT:   [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr6
36  ; CHECK-NEXT:   BUFFER_STORE_FORMAT_X_OFFSET_exact [[COPY4]], [[REG_SEQUENCE]], [[COPY5]], 4095, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
37  ; CHECK-NEXT:   S_ENDPGM 0
38  call void @llvm.amdgcn.raw.buffer.store.format.f32(float %val, <4 x i32> %rsrc, i32 4095, i32 %soffset, i32 0)
39  ret void
40}
41
42define amdgpu_ps void @raw_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v2f32(<4 x i32> inreg %rsrc, <2 x float> %val, i32 %voffset, i32 inreg %soffset) {
43  ; CHECK-LABEL: name: raw_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v2f32
44  ; CHECK: bb.1 (%ir-block.0):
45  ; CHECK-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1, $vgpr2
46  ; CHECK-NEXT: {{  $}}
47  ; CHECK-NEXT:   [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
48  ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
49  ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
50  ; CHECK-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
51  ; CHECK-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
52  ; CHECK-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
53  ; CHECK-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
54  ; CHECK-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[COPY5]], %subreg.sub1
55  ; CHECK-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr2
56  ; CHECK-NEXT:   [[COPY7:%[0-9]+]]:sreg_32 = COPY $sgpr6
57  ; CHECK-NEXT:   BUFFER_STORE_FORMAT_XY_OFFEN_exact [[REG_SEQUENCE1]], [[COPY6]], [[REG_SEQUENCE]], [[COPY7]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (<2 x s32>), align 1, addrspace 4)
58  ; CHECK-NEXT:   S_ENDPGM 0
59  call void @llvm.amdgcn.raw.buffer.store.format.v2f32(<2 x float> %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 0)
60  ret void
61}
62
63define amdgpu_ps void @raw_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v3f32(<4 x i32> inreg %rsrc, <3 x float> %val, i32 %voffset, i32 inreg %soffset) {
64  ; CHECK-LABEL: name: raw_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v3f32
65  ; CHECK: bb.1 (%ir-block.0):
66  ; CHECK-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1, $vgpr2, $vgpr3
67  ; CHECK-NEXT: {{  $}}
68  ; CHECK-NEXT:   [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
69  ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
70  ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
71  ; CHECK-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
72  ; CHECK-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
73  ; CHECK-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
74  ; CHECK-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
75  ; CHECK-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr2
76  ; CHECK-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:vreg_96 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[COPY5]], %subreg.sub1, [[COPY6]], %subreg.sub2
77  ; CHECK-NEXT:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY $vgpr3
78  ; CHECK-NEXT:   [[COPY8:%[0-9]+]]:sreg_32 = COPY $sgpr6
79  ; CHECK-NEXT:   BUFFER_STORE_FORMAT_XYZ_OFFEN_exact [[REG_SEQUENCE1]], [[COPY7]], [[REG_SEQUENCE]], [[COPY8]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (<3 x s32>), align 1, addrspace 4)
80  ; CHECK-NEXT:   S_ENDPGM 0
81  call void @llvm.amdgcn.raw.buffer.store.format.v3f32(<3 x float> %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 0)
82  ret void
83}
84
85define amdgpu_ps void @raw_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v4f32(<4 x i32> inreg %rsrc, <4 x float> %val, i32 %voffset, i32 inreg %soffset) {
86  ; CHECK-LABEL: name: raw_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v4f32
87  ; CHECK: bb.1 (%ir-block.0):
88  ; CHECK-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4
89  ; CHECK-NEXT: {{  $}}
90  ; CHECK-NEXT:   [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
91  ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
92  ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
93  ; CHECK-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
94  ; CHECK-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
95  ; CHECK-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
96  ; CHECK-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
97  ; CHECK-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr2
98  ; CHECK-NEXT:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY $vgpr3
99  ; CHECK-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[COPY5]], %subreg.sub1, [[COPY6]], %subreg.sub2, [[COPY7]], %subreg.sub3
100  ; CHECK-NEXT:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY $vgpr4
101  ; CHECK-NEXT:   [[COPY9:%[0-9]+]]:sreg_32 = COPY $sgpr6
102  ; CHECK-NEXT:   BUFFER_STORE_FORMAT_XYZW_OFFEN_exact [[REG_SEQUENCE1]], [[COPY8]], [[REG_SEQUENCE]], [[COPY9]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (<4 x s32>), align 1, addrspace 4)
103  ; CHECK-NEXT:   S_ENDPGM 0
104  call void @llvm.amdgcn.raw.buffer.store.format.v4f32(<4 x float> %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 0)
105  ret void
106}
107
108define amdgpu_ps void @raw_buffer_store_format__vgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v4f32(<4 x i32> %rsrc, <4 x float> %val, i32 %voffset, i32 inreg %soffset) {
109  ; CHECK-LABEL: name: raw_buffer_store_format__vgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v4f32
110  ; CHECK: bb.1 (%ir-block.0):
111  ; CHECK-NEXT:   successors: %bb.2(0x80000000)
112  ; CHECK-NEXT:   liveins: $sgpr2, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8
113  ; CHECK-NEXT: {{  $}}
114  ; CHECK-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
115  ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
116  ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
117  ; CHECK-NEXT:   [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr3
118  ; CHECK-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
119  ; CHECK-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr4
120  ; CHECK-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr5
121  ; CHECK-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr6
122  ; CHECK-NEXT:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY $vgpr7
123  ; CHECK-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[COPY5]], %subreg.sub1, [[COPY6]], %subreg.sub2, [[COPY7]], %subreg.sub3
124  ; CHECK-NEXT:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY $vgpr8
125  ; CHECK-NEXT:   [[COPY9:%[0-9]+]]:sreg_32 = COPY $sgpr2
126  ; CHECK-NEXT:   [[COPY10:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub0_sub1
127  ; CHECK-NEXT:   [[COPY11:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub2_sub3
128  ; CHECK-NEXT:   [[S_MOV_B64_:%[0-9]+]]:sreg_64_xexec = S_MOV_B64 $exec
129  ; CHECK-NEXT: {{  $}}
130  ; CHECK-NEXT: bb.2:
131  ; CHECK-NEXT:   successors: %bb.3(0x80000000)
132  ; CHECK-NEXT: {{  $}}
133  ; CHECK-NEXT:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY10]].sub0, implicit $exec
134  ; CHECK-NEXT:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY10]].sub1, implicit $exec
135  ; CHECK-NEXT:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
136  ; CHECK-NEXT:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY10]], implicit $exec
137  ; CHECK-NEXT:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY11]].sub0, implicit $exec
138  ; CHECK-NEXT:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY11]].sub1, implicit $exec
139  ; CHECK-NEXT:   [[REG_SEQUENCE3:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
140  ; CHECK-NEXT:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE3]], [[COPY11]], implicit $exec
141  ; CHECK-NEXT:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
142  ; CHECK-NEXT:   [[REG_SEQUENCE4:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
143  ; CHECK-NEXT:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
144  ; CHECK-NEXT: {{  $}}
145  ; CHECK-NEXT: bb.3:
146  ; CHECK-NEXT:   successors: %bb.4(0x40000000), %bb.2(0x40000000)
147  ; CHECK-NEXT: {{  $}}
148  ; CHECK-NEXT:   BUFFER_STORE_FORMAT_XYZW_OFFEN_exact [[REG_SEQUENCE1]], [[COPY8]], [[REG_SEQUENCE4]], [[COPY9]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (<4 x s32>), align 1, addrspace 4)
149  ; CHECK-NEXT:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
150  ; CHECK-NEXT:   SI_WATERFALL_LOOP %bb.2, implicit $exec
151  ; CHECK-NEXT: {{  $}}
152  ; CHECK-NEXT: bb.4:
153  ; CHECK-NEXT:   successors: %bb.5(0x80000000)
154  ; CHECK-NEXT: {{  $}}
155  ; CHECK-NEXT:   $exec = S_MOV_B64_term [[S_MOV_B64_]]
156  ; CHECK-NEXT: {{  $}}
157  ; CHECK-NEXT: bb.5:
158  ; CHECK-NEXT:   S_ENDPGM 0
159  call void @llvm.amdgcn.raw.buffer.store.format.v4f32(<4 x float> %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 0)
160  ret void
161}
162
163define amdgpu_ps void @raw_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v2f32_soffset4095(<4 x i32> inreg %rsrc, <2 x float> %val, i32 %voffset, i32 inreg %soffset) {
164  ; CHECK-LABEL: name: raw_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v2f32_soffset4095
165  ; CHECK: bb.1 (%ir-block.0):
166  ; CHECK-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1, $vgpr2
167  ; CHECK-NEXT: {{  $}}
168  ; CHECK-NEXT:   [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
169  ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
170  ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
171  ; CHECK-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
172  ; CHECK-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
173  ; CHECK-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
174  ; CHECK-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
175  ; CHECK-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[COPY5]], %subreg.sub1
176  ; CHECK-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr2
177  ; CHECK-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 4095
178  ; CHECK-NEXT:   BUFFER_STORE_FORMAT_XY_OFFEN_exact [[REG_SEQUENCE1]], [[COPY6]], [[REG_SEQUENCE]], [[S_MOV_B32_]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (<2 x s32>), align 1, addrspace 4)
179  ; CHECK-NEXT:   S_ENDPGM 0
180  call void @llvm.amdgcn.raw.buffer.store.format.v2f32(<2 x float> %val, <4 x i32> %rsrc, i32 %voffset, i32 4095, i32 0)
181  ret void
182}
183
184define amdgpu_ps void @raw_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v2f32_soffset4096(<4 x i32> inreg %rsrc, <2 x float> %val, i32 %voffset, i32 inreg %soffset) {
185  ; CHECK-LABEL: name: raw_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v2f32_soffset4096
186  ; CHECK: bb.1 (%ir-block.0):
187  ; CHECK-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1, $vgpr2
188  ; CHECK-NEXT: {{  $}}
189  ; CHECK-NEXT:   [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
190  ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
191  ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
192  ; CHECK-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
193  ; CHECK-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
194  ; CHECK-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
195  ; CHECK-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
196  ; CHECK-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[COPY5]], %subreg.sub1
197  ; CHECK-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr2
198  ; CHECK-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 4096
199  ; CHECK-NEXT:   BUFFER_STORE_FORMAT_XY_OFFEN_exact [[REG_SEQUENCE1]], [[COPY6]], [[REG_SEQUENCE]], [[S_MOV_B32_]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (<2 x s32>), align 1, addrspace 4)
200  ; CHECK-NEXT:   S_ENDPGM 0
201  call void @llvm.amdgcn.raw.buffer.store.format.v2f32(<2 x float> %val, <4 x i32> %rsrc, i32 %voffset, i32 4096, i32 0)
202  ret void
203}
204
205define amdgpu_ps void @raw_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v2f32_voffset_add_16(<4 x i32> inreg %rsrc, <2 x float> %val, i32 %voffset, i32 inreg %soffset) {
206  ; CHECK-LABEL: name: raw_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v2f32_voffset_add_16
207  ; CHECK: bb.1 (%ir-block.0):
208  ; CHECK-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1, $vgpr2
209  ; CHECK-NEXT: {{  $}}
210  ; CHECK-NEXT:   [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
211  ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
212  ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
213  ; CHECK-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
214  ; CHECK-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
215  ; CHECK-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
216  ; CHECK-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
217  ; CHECK-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[COPY5]], %subreg.sub1
218  ; CHECK-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr2
219  ; CHECK-NEXT:   [[COPY7:%[0-9]+]]:sreg_32 = COPY $sgpr6
220  ; CHECK-NEXT:   BUFFER_STORE_FORMAT_XY_OFFEN_exact [[REG_SEQUENCE1]], [[COPY6]], [[REG_SEQUENCE]], [[COPY7]], 16, 0, 0, 0, implicit $exec :: (dereferenceable store (<2 x s32>), align 1, addrspace 4)
221  ; CHECK-NEXT:   S_ENDPGM 0
222  %voffset.add = add i32 %voffset, 16
223  call void @llvm.amdgcn.raw.buffer.store.format.v2f32(<2 x float> %val, <4 x i32> %rsrc, i32 %voffset.add, i32 %soffset, i32 0)
224  ret void
225}
226
227define amdgpu_ps void @raw_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v2f32_voffset_add_4095(<4 x i32> inreg %rsrc, <2 x float> %val, i32 %voffset, i32 inreg %soffset) {
228  ; CHECK-LABEL: name: raw_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v2f32_voffset_add_4095
229  ; CHECK: bb.1 (%ir-block.0):
230  ; CHECK-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1, $vgpr2
231  ; CHECK-NEXT: {{  $}}
232  ; CHECK-NEXT:   [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
233  ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
234  ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
235  ; CHECK-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
236  ; CHECK-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
237  ; CHECK-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
238  ; CHECK-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
239  ; CHECK-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[COPY5]], %subreg.sub1
240  ; CHECK-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr2
241  ; CHECK-NEXT:   [[COPY7:%[0-9]+]]:sreg_32 = COPY $sgpr6
242  ; CHECK-NEXT:   BUFFER_STORE_FORMAT_XY_OFFEN_exact [[REG_SEQUENCE1]], [[COPY6]], [[REG_SEQUENCE]], [[COPY7]], 4095, 0, 0, 0, implicit $exec :: (dereferenceable store (<2 x s32>), align 1, addrspace 4)
243  ; CHECK-NEXT:   S_ENDPGM 0
244  %voffset.add = add i32 %voffset, 4095
245  call void @llvm.amdgcn.raw.buffer.store.format.v2f32(<2 x float> %val, <4 x i32> %rsrc, i32 %voffset.add, i32 %soffset, i32 0)
246  ret void
247}
248
249define amdgpu_ps void @raw_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v2f32_voffset_add_4096(<4 x i32> inreg %rsrc, <2 x float> %val, i32 %voffset, i32 inreg %soffset) {
250  ; CHECK-LABEL: name: raw_buffer_store_format__sgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v2f32_voffset_add_4096
251  ; CHECK: bb.1 (%ir-block.0):
252  ; CHECK-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1, $vgpr2
253  ; CHECK-NEXT: {{  $}}
254  ; CHECK-NEXT:   [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
255  ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
256  ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
257  ; CHECK-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
258  ; CHECK-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
259  ; CHECK-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
260  ; CHECK-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
261  ; CHECK-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[COPY5]], %subreg.sub1
262  ; CHECK-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr2
263  ; CHECK-NEXT:   [[COPY7:%[0-9]+]]:sreg_32 = COPY $sgpr6
264  ; CHECK-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 4096
265  ; CHECK-NEXT:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[S_MOV_B32_]]
266  ; CHECK-NEXT:   %13:vgpr_32, dead %17:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY6]], [[COPY8]], 0, implicit $exec
267  ; CHECK-NEXT:   BUFFER_STORE_FORMAT_XY_OFFEN_exact [[REG_SEQUENCE1]], %13, [[REG_SEQUENCE]], [[COPY7]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (<2 x s32>), align 1, addrspace 4)
268  ; CHECK-NEXT:   S_ENDPGM 0
269  %voffset.add = add i32 %voffset, 4096
270  call void @llvm.amdgcn.raw.buffer.store.format.v2f32(<2 x float> %val, <4 x i32> %rsrc, i32 %voffset.add, i32 %soffset, i32 0)
271  ret void
272}
273
274
275; Check what happens with offset add inside a waterfall loop
276define amdgpu_ps void @raw_buffer_store_format__vgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v4f32_add_4096(<4 x i32> %rsrc, <4 x float> %val, i32 %voffset, i32 inreg %soffset) {
277  ; CHECK-LABEL: name: raw_buffer_store_format__vgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v4f32_add_4096
278  ; CHECK: bb.1 (%ir-block.0):
279  ; CHECK-NEXT:   successors: %bb.2(0x80000000)
280  ; CHECK-NEXT:   liveins: $sgpr2, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8
281  ; CHECK-NEXT: {{  $}}
282  ; CHECK-NEXT:   [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
283  ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
284  ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
285  ; CHECK-NEXT:   [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr3
286  ; CHECK-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
287  ; CHECK-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr4
288  ; CHECK-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr5
289  ; CHECK-NEXT:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr6
290  ; CHECK-NEXT:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY $vgpr7
291  ; CHECK-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[COPY5]], %subreg.sub1, [[COPY6]], %subreg.sub2, [[COPY7]], %subreg.sub3
292  ; CHECK-NEXT:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY $vgpr8
293  ; CHECK-NEXT:   [[COPY9:%[0-9]+]]:sreg_32 = COPY $sgpr2
294  ; CHECK-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 4096
295  ; CHECK-NEXT:   [[COPY10:%[0-9]+]]:vgpr_32 = COPY [[S_MOV_B32_]]
296  ; CHECK-NEXT:   %15:vgpr_32, dead %35:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY8]], [[COPY10]], 0, implicit $exec
297  ; CHECK-NEXT:   [[COPY11:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub0_sub1
298  ; CHECK-NEXT:   [[COPY12:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub2_sub3
299  ; CHECK-NEXT:   [[S_MOV_B64_:%[0-9]+]]:sreg_64_xexec = S_MOV_B64 $exec
300  ; CHECK-NEXT: {{  $}}
301  ; CHECK-NEXT: bb.2:
302  ; CHECK-NEXT:   successors: %bb.3(0x80000000)
303  ; CHECK-NEXT: {{  $}}
304  ; CHECK-NEXT:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY11]].sub0, implicit $exec
305  ; CHECK-NEXT:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY11]].sub1, implicit $exec
306  ; CHECK-NEXT:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
307  ; CHECK-NEXT:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY11]], implicit $exec
308  ; CHECK-NEXT:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY12]].sub0, implicit $exec
309  ; CHECK-NEXT:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY12]].sub1, implicit $exec
310  ; CHECK-NEXT:   [[REG_SEQUENCE3:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
311  ; CHECK-NEXT:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE3]], [[COPY12]], implicit $exec
312  ; CHECK-NEXT:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
313  ; CHECK-NEXT:   [[REG_SEQUENCE4:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
314  ; CHECK-NEXT:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
315  ; CHECK-NEXT: {{  $}}
316  ; CHECK-NEXT: bb.3:
317  ; CHECK-NEXT:   successors: %bb.4(0x40000000), %bb.2(0x40000000)
318  ; CHECK-NEXT: {{  $}}
319  ; CHECK-NEXT:   BUFFER_STORE_FORMAT_XYZW_OFFEN_exact [[REG_SEQUENCE1]], %15, [[REG_SEQUENCE4]], [[COPY9]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (<4 x s32>), align 1, addrspace 4)
320  ; CHECK-NEXT:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
321  ; CHECK-NEXT:   SI_WATERFALL_LOOP %bb.2, implicit $exec
322  ; CHECK-NEXT: {{  $}}
323  ; CHECK-NEXT: bb.4:
324  ; CHECK-NEXT:   successors: %bb.5(0x80000000)
325  ; CHECK-NEXT: {{  $}}
326  ; CHECK-NEXT:   $exec = S_MOV_B64_term [[S_MOV_B64_]]
327  ; CHECK-NEXT: {{  $}}
328  ; CHECK-NEXT: bb.5:
329  ; CHECK-NEXT:   S_ENDPGM 0
330  %voffset.add = add i32 %voffset, 4096
331  call void @llvm.amdgcn.raw.buffer.store.format.v4f32(<4 x float> %val, <4 x i32> %rsrc, i32 %voffset.add, i32 %soffset, i32 0)
332  ret void
333}
334
335declare void @llvm.amdgcn.raw.buffer.store.format.f32(float, <4 x i32>, i32, i32, i32 immarg)
336declare void @llvm.amdgcn.raw.buffer.store.format.v2f32(<2 x float>, <4 x i32>, i32, i32, i32 immarg)
337declare void @llvm.amdgcn.raw.buffer.store.format.v3f32(<3 x float>, <4 x i32>, i32, i32, i32 immarg)
338declare void @llvm.amdgcn.raw.buffer.store.format.v4f32(<4 x float>, <4 x i32>, i32, i32, i32 immarg)
339