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
4 define amdgpu_ps void @struct_ptr_buffer_store_format_f32__vgpr_val__sgpr_rsrc__vgpr_vindex__vgpr_voffset__sgpr_soffset(float %val, ptr addrspace(8) inreg %rsrc, i32 %vindex, i32 %voffset, i32 inreg %soffset) {
5 ; CHECK-LABEL: name: struct_ptr_buffer_store_format_f32__vgpr_val__sgpr_rsrc__vgpr_vindex__vgpr_voffset__sgpr_soffset
6 ; CHECK: bb.1 (%ir-block.0):
7 ; CHECK-NEXT: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1, $vgpr2
9 ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
10 ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
11 ; CHECK-NEXT: [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
12 ; CHECK-NEXT: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
13 ; CHECK-NEXT: [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
14 ; CHECK-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
15 ; CHECK-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr2
16 ; CHECK-NEXT: [[COPY7:%[0-9]+]]:sreg_32 = COPY $sgpr6
17 ; CHECK-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
18 ; CHECK-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY5]], %subreg.sub0, [[COPY6]], %subreg.sub1
19 ; CHECK-NEXT: BUFFER_STORE_FORMAT_X_BOTHEN_exact [[COPY]], [[REG_SEQUENCE1]], [[REG_SEQUENCE]], [[COPY7]], 0, 0, 0, implicit $exec :: (dereferenceable store (s32) into %ir.rsrc, align 1, addrspace 8)
20 ; CHECK-NEXT: S_ENDPGM 0
21 call void @llvm.amdgcn.struct.ptr.buffer.store.format.f32(float %val, ptr addrspace(8) %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
25 define amdgpu_ps void @struct_ptr_buffer_store_format_v2f32__vgpr_val__sgpr_rsrc__vgpr_vindex__vgpr_voffset__sgpr_soffset(<2 x float> %val, ptr addrspace(8) inreg %rsrc, i32 %vindex, i32 %voffset, i32 inreg %soffset) {
26 ; CHECK-LABEL: name: struct_ptr_buffer_store_format_v2f32__vgpr_val__sgpr_rsrc__vgpr_vindex__vgpr_voffset__sgpr_soffset
27 ; CHECK: bb.1 (%ir-block.0):
28 ; CHECK-NEXT: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1, $vgpr2, $vgpr3
30 ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
31 ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
32 ; CHECK-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1
33 ; CHECK-NEXT: [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr2
34 ; CHECK-NEXT: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr3
35 ; CHECK-NEXT: [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr4
36 ; CHECK-NEXT: [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr5
37 ; CHECK-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr2
38 ; CHECK-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY $vgpr3
39 ; CHECK-NEXT: [[COPY8:%[0-9]+]]:sreg_32 = COPY $sgpr6
40 ; CHECK-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY2]], %subreg.sub0, [[COPY3]], %subreg.sub1, [[COPY4]], %subreg.sub2, [[COPY5]], %subreg.sub3
41 ; CHECK-NEXT: [[REG_SEQUENCE2:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY6]], %subreg.sub0, [[COPY7]], %subreg.sub1
42 ; CHECK-NEXT: BUFFER_STORE_FORMAT_XY_BOTHEN_exact [[REG_SEQUENCE]], [[REG_SEQUENCE2]], [[REG_SEQUENCE1]], [[COPY8]], 0, 0, 0, implicit $exec :: (dereferenceable store (<2 x s32>) into %ir.rsrc, align 1, addrspace 8)
43 ; CHECK-NEXT: S_ENDPGM 0
44 call void @llvm.amdgcn.struct.ptr.buffer.store.format.v2f32(<2 x float> %val, ptr addrspace(8) %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
48 define amdgpu_ps void @struct_ptr_buffer_store_format_v3f32__vgpr_val__sgpr_rsrc__vgpr_vindex__vgpr_voffset__sgpr_soffset(<3 x float> %val, ptr addrspace(8) inreg %rsrc, i32 %vindex, i32 %voffset, i32 inreg %soffset) {
49 ; CHECK-LABEL: name: struct_ptr_buffer_store_format_v3f32__vgpr_val__sgpr_rsrc__vgpr_vindex__vgpr_voffset__sgpr_soffset
50 ; CHECK: bb.1 (%ir-block.0):
51 ; CHECK-NEXT: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4
53 ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
54 ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
55 ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
56 ; CHECK-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_96 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2
57 ; CHECK-NEXT: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr2
58 ; CHECK-NEXT: [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr3
59 ; CHECK-NEXT: [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr4
60 ; CHECK-NEXT: [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr5
61 ; CHECK-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY $vgpr3
62 ; CHECK-NEXT: [[COPY8:%[0-9]+]]:vgpr_32 = COPY $vgpr4
63 ; CHECK-NEXT: [[COPY9:%[0-9]+]]:sreg_32 = COPY $sgpr6
64 ; CHECK-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY3]], %subreg.sub0, [[COPY4]], %subreg.sub1, [[COPY5]], %subreg.sub2, [[COPY6]], %subreg.sub3
65 ; CHECK-NEXT: [[REG_SEQUENCE2:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY7]], %subreg.sub0, [[COPY8]], %subreg.sub1
66 ; CHECK-NEXT: BUFFER_STORE_FORMAT_XYZ_BOTHEN_exact [[REG_SEQUENCE]], [[REG_SEQUENCE2]], [[REG_SEQUENCE1]], [[COPY9]], 0, 0, 0, implicit $exec :: (dereferenceable store (<3 x s32>) into %ir.rsrc, align 1, addrspace 8)
67 ; CHECK-NEXT: S_ENDPGM 0
68 call void @llvm.amdgcn.struct.ptr.buffer.store.format.v3f32(<3 x float> %val, ptr addrspace(8) %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
72 define amdgpu_ps void @struct_ptr_buffer_store_format_v4f32__vgpr_val__sgpr_rsrc__vgpr_vindex__vgpr_voffset__sgpr_soffset(<4 x float> %val, ptr addrspace(8) inreg %rsrc, i32 %vindex, i32 %voffset, i32 inreg %soffset) {
73 ; CHECK-LABEL: name: struct_ptr_buffer_store_format_v4f32__vgpr_val__sgpr_rsrc__vgpr_vindex__vgpr_voffset__sgpr_soffset
74 ; CHECK: bb.1 (%ir-block.0):
75 ; CHECK-NEXT: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5
77 ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
78 ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
79 ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
80 ; CHECK-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr3
81 ; CHECK-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
82 ; CHECK-NEXT: [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr2
83 ; CHECK-NEXT: [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr3
84 ; CHECK-NEXT: [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr4
85 ; CHECK-NEXT: [[COPY7:%[0-9]+]]:sreg_32 = COPY $sgpr5
86 ; CHECK-NEXT: [[COPY8:%[0-9]+]]:vgpr_32 = COPY $vgpr4
87 ; CHECK-NEXT: [[COPY9:%[0-9]+]]:vgpr_32 = COPY $vgpr5
88 ; CHECK-NEXT: [[COPY10:%[0-9]+]]:sreg_32 = COPY $sgpr6
89 ; CHECK-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[COPY5]], %subreg.sub1, [[COPY6]], %subreg.sub2, [[COPY7]], %subreg.sub3
90 ; CHECK-NEXT: [[REG_SEQUENCE2:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY8]], %subreg.sub0, [[COPY9]], %subreg.sub1
91 ; CHECK-NEXT: BUFFER_STORE_FORMAT_XYZW_BOTHEN_exact [[REG_SEQUENCE]], [[REG_SEQUENCE2]], [[REG_SEQUENCE1]], [[COPY10]], 0, 0, 0, implicit $exec :: (dereferenceable store (<4 x s32>) into %ir.rsrc, align 1, addrspace 8)
92 ; CHECK-NEXT: S_ENDPGM 0
93 call void @llvm.amdgcn.struct.ptr.buffer.store.format.v4f32(<4 x float> %val, ptr addrspace(8) %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
97 define amdgpu_ps void @struct_ptr_buffer_store_format_f32__sgpr_val__vgpr_rsrc__sgpr_vindex__sgpr_voffset__vgpr_soffset(float inreg %val, ptr addrspace(8) %rsrc, i32 inreg %vindex, i32 inreg %voffset, i32 %soffset) {
98 ; CHECK-LABEL: name: struct_ptr_buffer_store_format_f32__sgpr_val__vgpr_rsrc__sgpr_vindex__sgpr_voffset__vgpr_soffset
99 ; CHECK: bb.1 (%ir-block.0):
100 ; CHECK-NEXT: successors: %bb.2(0x80000000)
101 ; CHECK-NEXT: liveins: $sgpr2, $sgpr3, $sgpr4, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4
103 ; CHECK-NEXT: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
104 ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr0
105 ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr1
106 ; CHECK-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr2
107 ; CHECK-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr3
108 ; CHECK-NEXT: [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr3
109 ; CHECK-NEXT: [[COPY6:%[0-9]+]]:sreg_32 = COPY $sgpr4
110 ; CHECK-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY $vgpr4
111 ; CHECK-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
112 ; CHECK-NEXT: [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[COPY]]
113 ; CHECK-NEXT: [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[COPY5]]
114 ; CHECK-NEXT: [[COPY10:%[0-9]+]]:vgpr_32 = COPY [[COPY6]]
115 ; CHECK-NEXT: [[S_MOV_B64_:%[0-9]+]]:sreg_64_xexec = S_MOV_B64 $exec
118 ; CHECK-NEXT: successors: %bb.3(0x80000000)
120 ; CHECK-NEXT: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY1]], implicit $exec
121 ; CHECK-NEXT: [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY2]], implicit $exec
122 ; CHECK-NEXT: [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY3]], implicit $exec
123 ; CHECK-NEXT: [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY4]], implicit $exec
124 ; CHECK-NEXT: [[REG_SEQUENCE1:%[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
125 ; CHECK-NEXT: [[COPY11:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub0_sub1
126 ; CHECK-NEXT: [[COPY12:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub2_sub3
127 ; CHECK-NEXT: [[COPY13:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE1]].sub0_sub1
128 ; CHECK-NEXT: [[COPY14:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE1]].sub2_sub3
129 ; CHECK-NEXT: [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[COPY13]], [[COPY11]], implicit $exec
130 ; CHECK-NEXT: [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[COPY14]], [[COPY12]], implicit $exec
131 ; CHECK-NEXT: [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_]], [[V_CMP_EQ_U64_e64_1]], implicit-def dead $scc
132 ; CHECK-NEXT: [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY7]], implicit $exec
133 ; CHECK-NEXT: [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY7]], implicit $exec
134 ; CHECK-NEXT: [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[S_AND_B64_]], [[V_CMP_EQ_U32_e64_]], implicit-def dead $scc
135 ; CHECK-NEXT: [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
138 ; CHECK-NEXT: successors: %bb.4(0x40000000), %bb.2(0x40000000)
140 ; CHECK-NEXT: [[REG_SEQUENCE2:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY9]], %subreg.sub0, [[COPY10]], %subreg.sub1
141 ; CHECK-NEXT: BUFFER_STORE_FORMAT_X_BOTHEN_exact [[COPY8]], [[REG_SEQUENCE2]], [[REG_SEQUENCE1]], [[V_READFIRSTLANE_B32_4]], 0, 0, 0, implicit $exec :: (dereferenceable store (s32) into %ir.rsrc, align 1, addrspace 8)
142 ; CHECK-NEXT: $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
143 ; CHECK-NEXT: SI_WATERFALL_LOOP %bb.2, implicit $exec
146 ; CHECK-NEXT: successors: %bb.5(0x80000000)
148 ; CHECK-NEXT: $exec = S_MOV_B64_term [[S_MOV_B64_]]
151 ; CHECK-NEXT: S_ENDPGM 0
152 call void @llvm.amdgcn.struct.ptr.buffer.store.format.f32(float %val, ptr addrspace(8) %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
156 define amdgpu_ps void @struct_ptr_buffer_store_format_i32__vgpr_val__sgpr_rsrc__vgpr_vindex__vgpr_voffset__sgpr_soffset(i32 %val, ptr addrspace(8) inreg %rsrc, i32 %vindex, i32 %voffset, i32 inreg %soffset) {
157 ; CHECK-LABEL: name: struct_ptr_buffer_store_format_i32__vgpr_val__sgpr_rsrc__vgpr_vindex__vgpr_voffset__sgpr_soffset
158 ; CHECK: bb.1 (%ir-block.0):
159 ; CHECK-NEXT: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0, $vgpr1, $vgpr2
161 ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
162 ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
163 ; CHECK-NEXT: [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr3
164 ; CHECK-NEXT: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr4
165 ; CHECK-NEXT: [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr5
166 ; CHECK-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
167 ; CHECK-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr2
168 ; CHECK-NEXT: [[COPY7:%[0-9]+]]:sreg_32 = COPY $sgpr6
169 ; CHECK-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
170 ; CHECK-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY5]], %subreg.sub0, [[COPY6]], %subreg.sub1
171 ; CHECK-NEXT: BUFFER_STORE_FORMAT_X_BOTHEN_exact [[COPY]], [[REG_SEQUENCE1]], [[REG_SEQUENCE]], [[COPY7]], 0, 0, 0, implicit $exec :: (dereferenceable store (s32) into %ir.rsrc, align 1, addrspace 8)
172 ; CHECK-NEXT: S_ENDPGM 0
173 call void @llvm.amdgcn.struct.ptr.buffer.store.format.i32(i32 %val, ptr addrspace(8) %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
177 declare void @llvm.amdgcn.struct.ptr.buffer.store.format.f32(float, ptr addrspace(8), i32, i32, i32, i32 immarg)
178 declare void @llvm.amdgcn.struct.ptr.buffer.store.format.v2f32(<2 x float>, ptr addrspace(8), i32, i32, i32, i32 immarg)
179 declare void @llvm.amdgcn.struct.ptr.buffer.store.format.v3f32(<3 x float>, ptr addrspace(8), i32, i32, i32, i32 immarg)
180 declare void @llvm.amdgcn.struct.ptr.buffer.store.format.v4f32(<4 x float>, ptr addrspace(8), i32, i32, i32, i32 immarg)
181 declare void @llvm.amdgcn.struct.ptr.buffer.store.format.i32(i32, ptr addrspace(8), i32, i32, i32, i32 immarg)