1 ; NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
2 ; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -stop-after=instruction-select -verify-machineinstrs -o - %s | FileCheck -check-prefix=GFX10_GFX11 %s
3 ; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -stop-after=instruction-select -verify-machineinstrs -o - %s | FileCheck -check-prefix=GFX10_GFX11 %s
4 ; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 -stop-after=instruction-select -verify-machineinstrs -o - %s | FileCheck -check-prefix=GFX12 %s
6 define amdgpu_ps float @raw_tbuffer_load_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset(<4 x i32> inreg %rsrc, i32 %voffset, i32 inreg %soffset) {
7 ; GFX10_GFX11-LABEL: name: raw_tbuffer_load_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset
8 ; GFX10_GFX11: bb.1 (%ir-block.0):
9 ; GFX10_GFX11-NEXT: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0
10 ; GFX10_GFX11-NEXT: {{ $}}
11 ; GFX10_GFX11-NEXT: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
12 ; GFX10_GFX11-NEXT: [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
13 ; GFX10_GFX11-NEXT: [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
14 ; GFX10_GFX11-NEXT: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
15 ; GFX10_GFX11-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
16 ; GFX10_GFX11-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
17 ; GFX10_GFX11-NEXT: [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr6
18 ; GFX10_GFX11-NEXT: [[TBUFFER_LOAD_FORMAT_X_OFFEN:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFEN [[COPY4]], [[REG_SEQUENCE]], [[COPY5]], 0, 78, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
19 ; GFX10_GFX11-NEXT: $vgpr0 = COPY [[TBUFFER_LOAD_FORMAT_X_OFFEN]]
20 ; GFX10_GFX11-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
22 ; GFX12-LABEL: name: raw_tbuffer_load_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset
23 ; GFX12: bb.1 (%ir-block.0):
24 ; GFX12-NEXT: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0
26 ; GFX12-NEXT: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
27 ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
28 ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
29 ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
30 ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
31 ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
32 ; GFX12-NEXT: [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr6
33 ; GFX12-NEXT: [[TBUFFER_LOAD_FORMAT_X_VBUFFER_OFFEN:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_VBUFFER_OFFEN [[COPY4]], [[REG_SEQUENCE]], [[COPY5]], 0, 78, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
34 ; GFX12-NEXT: $vgpr0 = COPY [[TBUFFER_LOAD_FORMAT_X_VBUFFER_OFFEN]]
35 ; GFX12-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
36 %val = call float @llvm.amdgcn.raw.tbuffer.load.f32(<4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 78, i32 0)
40 define amdgpu_ps <2 x float> @raw_tbuffer_load_v2f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset(<4 x i32> inreg %rsrc, i32 %voffset, i32 inreg %soffset) {
41 ; GFX10_GFX11-LABEL: name: raw_tbuffer_load_v2f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset
42 ; GFX10_GFX11: bb.1 (%ir-block.0):
43 ; GFX10_GFX11-NEXT: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0
44 ; GFX10_GFX11-NEXT: {{ $}}
45 ; GFX10_GFX11-NEXT: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
46 ; GFX10_GFX11-NEXT: [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
47 ; GFX10_GFX11-NEXT: [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
48 ; GFX10_GFX11-NEXT: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
49 ; GFX10_GFX11-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
50 ; GFX10_GFX11-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
51 ; GFX10_GFX11-NEXT: [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr6
52 ; GFX10_GFX11-NEXT: [[TBUFFER_LOAD_FORMAT_XY_OFFEN:%[0-9]+]]:vreg_64 = TBUFFER_LOAD_FORMAT_XY_OFFEN [[COPY4]], [[REG_SEQUENCE]], [[COPY5]], 0, 78, 0, 0, implicit $exec :: (dereferenceable load (<2 x s32>), align 1, addrspace 8)
53 ; GFX10_GFX11-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XY_OFFEN]].sub0
54 ; GFX10_GFX11-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XY_OFFEN]].sub1
55 ; GFX10_GFX11-NEXT: $vgpr0 = COPY [[COPY6]]
56 ; GFX10_GFX11-NEXT: $vgpr1 = COPY [[COPY7]]
57 ; GFX10_GFX11-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
59 ; GFX12-LABEL: name: raw_tbuffer_load_v2f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset
60 ; GFX12: bb.1 (%ir-block.0):
61 ; GFX12-NEXT: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0
63 ; GFX12-NEXT: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
64 ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
65 ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
66 ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
67 ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
68 ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
69 ; GFX12-NEXT: [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr6
70 ; GFX12-NEXT: [[TBUFFER_LOAD_FORMAT_XY_VBUFFER_OFFEN:%[0-9]+]]:vreg_64 = TBUFFER_LOAD_FORMAT_XY_VBUFFER_OFFEN [[COPY4]], [[REG_SEQUENCE]], [[COPY5]], 0, 78, 0, 0, implicit $exec :: (dereferenceable load (<2 x s32>), align 1, addrspace 8)
71 ; GFX12-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XY_VBUFFER_OFFEN]].sub0
72 ; GFX12-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XY_VBUFFER_OFFEN]].sub1
73 ; GFX12-NEXT: $vgpr0 = COPY [[COPY6]]
74 ; GFX12-NEXT: $vgpr1 = COPY [[COPY7]]
75 ; GFX12-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
76 %val = call <2 x float> @llvm.amdgcn.raw.tbuffer.load.v2f32(<4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 78, i32 0)
80 define amdgpu_ps <3 x float> @raw_tbuffer_load_v3f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset(<4 x i32> inreg %rsrc, i32 %voffset, i32 inreg %soffset) {
81 ; GFX10_GFX11-LABEL: name: raw_tbuffer_load_v3f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset
82 ; GFX10_GFX11: bb.1 (%ir-block.0):
83 ; GFX10_GFX11-NEXT: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0
84 ; GFX10_GFX11-NEXT: {{ $}}
85 ; GFX10_GFX11-NEXT: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
86 ; GFX10_GFX11-NEXT: [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
87 ; GFX10_GFX11-NEXT: [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
88 ; GFX10_GFX11-NEXT: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
89 ; GFX10_GFX11-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
90 ; GFX10_GFX11-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
91 ; GFX10_GFX11-NEXT: [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr6
92 ; GFX10_GFX11-NEXT: [[TBUFFER_LOAD_FORMAT_XYZ_OFFEN:%[0-9]+]]:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_OFFEN [[COPY4]], [[REG_SEQUENCE]], [[COPY5]], 0, 78, 0, 0, implicit $exec :: (dereferenceable load (<3 x s32>), align 1, addrspace 8)
93 ; GFX10_GFX11-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XYZ_OFFEN]].sub0
94 ; GFX10_GFX11-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XYZ_OFFEN]].sub1
95 ; GFX10_GFX11-NEXT: [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XYZ_OFFEN]].sub2
96 ; GFX10_GFX11-NEXT: $vgpr0 = COPY [[COPY6]]
97 ; GFX10_GFX11-NEXT: $vgpr1 = COPY [[COPY7]]
98 ; GFX10_GFX11-NEXT: $vgpr2 = COPY [[COPY8]]
99 ; GFX10_GFX11-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2
101 ; GFX12-LABEL: name: raw_tbuffer_load_v3f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset
102 ; GFX12: bb.1 (%ir-block.0):
103 ; GFX12-NEXT: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0
105 ; GFX12-NEXT: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
106 ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
107 ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
108 ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
109 ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
110 ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
111 ; GFX12-NEXT: [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr6
112 ; GFX12-NEXT: [[TBUFFER_LOAD_FORMAT_XYZ_VBUFFER_OFFEN:%[0-9]+]]:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_VBUFFER_OFFEN [[COPY4]], [[REG_SEQUENCE]], [[COPY5]], 0, 78, 0, 0, implicit $exec :: (dereferenceable load (<3 x s32>), align 1, addrspace 8)
113 ; GFX12-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XYZ_VBUFFER_OFFEN]].sub0
114 ; GFX12-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XYZ_VBUFFER_OFFEN]].sub1
115 ; GFX12-NEXT: [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XYZ_VBUFFER_OFFEN]].sub2
116 ; GFX12-NEXT: $vgpr0 = COPY [[COPY6]]
117 ; GFX12-NEXT: $vgpr1 = COPY [[COPY7]]
118 ; GFX12-NEXT: $vgpr2 = COPY [[COPY8]]
119 ; GFX12-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2
120 %val = call <3 x float> @llvm.amdgcn.raw.tbuffer.load.v3f32(<4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 78, i32 0)
124 define amdgpu_ps <4 x float> @raw_tbuffer_load_v4f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset(<4 x i32> inreg %rsrc, i32 %voffset, i32 inreg %soffset) {
125 ; GFX10_GFX11-LABEL: name: raw_tbuffer_load_v4f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset
126 ; GFX10_GFX11: bb.1 (%ir-block.0):
127 ; GFX10_GFX11-NEXT: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0
128 ; GFX10_GFX11-NEXT: {{ $}}
129 ; GFX10_GFX11-NEXT: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
130 ; GFX10_GFX11-NEXT: [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
131 ; GFX10_GFX11-NEXT: [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
132 ; GFX10_GFX11-NEXT: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
133 ; GFX10_GFX11-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
134 ; GFX10_GFX11-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
135 ; GFX10_GFX11-NEXT: [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr6
136 ; GFX10_GFX11-NEXT: [[TBUFFER_LOAD_FORMAT_XYZW_OFFEN:%[0-9]+]]:vreg_128 = TBUFFER_LOAD_FORMAT_XYZW_OFFEN [[COPY4]], [[REG_SEQUENCE]], [[COPY5]], 0, 78, 0, 0, implicit $exec :: (dereferenceable load (<4 x s32>), align 1, addrspace 8)
137 ; GFX10_GFX11-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XYZW_OFFEN]].sub0
138 ; GFX10_GFX11-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XYZW_OFFEN]].sub1
139 ; GFX10_GFX11-NEXT: [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XYZW_OFFEN]].sub2
140 ; GFX10_GFX11-NEXT: [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XYZW_OFFEN]].sub3
141 ; GFX10_GFX11-NEXT: $vgpr0 = COPY [[COPY6]]
142 ; GFX10_GFX11-NEXT: $vgpr1 = COPY [[COPY7]]
143 ; GFX10_GFX11-NEXT: $vgpr2 = COPY [[COPY8]]
144 ; GFX10_GFX11-NEXT: $vgpr3 = COPY [[COPY9]]
145 ; GFX10_GFX11-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
147 ; GFX12-LABEL: name: raw_tbuffer_load_v4f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset
148 ; GFX12: bb.1 (%ir-block.0):
149 ; GFX12-NEXT: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0
151 ; GFX12-NEXT: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
152 ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
153 ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
154 ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
155 ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
156 ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
157 ; GFX12-NEXT: [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr6
158 ; GFX12-NEXT: [[TBUFFER_LOAD_FORMAT_XYZW_VBUFFER_OFFEN:%[0-9]+]]:vreg_128 = TBUFFER_LOAD_FORMAT_XYZW_VBUFFER_OFFEN [[COPY4]], [[REG_SEQUENCE]], [[COPY5]], 0, 78, 0, 0, implicit $exec :: (dereferenceable load (<4 x s32>), align 1, addrspace 8)
159 ; GFX12-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XYZW_VBUFFER_OFFEN]].sub0
160 ; GFX12-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XYZW_VBUFFER_OFFEN]].sub1
161 ; GFX12-NEXT: [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XYZW_VBUFFER_OFFEN]].sub2
162 ; GFX12-NEXT: [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XYZW_VBUFFER_OFFEN]].sub3
163 ; GFX12-NEXT: $vgpr0 = COPY [[COPY6]]
164 ; GFX12-NEXT: $vgpr1 = COPY [[COPY7]]
165 ; GFX12-NEXT: $vgpr2 = COPY [[COPY8]]
166 ; GFX12-NEXT: $vgpr3 = COPY [[COPY9]]
167 ; GFX12-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
168 %val = call <4 x float> @llvm.amdgcn.raw.tbuffer.load.v4f32(<4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 78, i32 0)
172 define amdgpu_ps float @raw_tbuffer_load_f32__vgpr_rsrc__sgpr_voffset__vgpr_soffset(<4 x i32> %rsrc, i32 inreg %voffset, i32 %soffset) {
173 ; GFX10_GFX11-LABEL: name: raw_tbuffer_load_f32__vgpr_rsrc__sgpr_voffset__vgpr_soffset
174 ; GFX10_GFX11: bb.1 (%ir-block.0):
175 ; GFX10_GFX11-NEXT: successors: %bb.2(0x80000000)
176 ; GFX10_GFX11-NEXT: liveins: $sgpr2, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4
177 ; GFX10_GFX11-NEXT: {{ $}}
178 ; GFX10_GFX11-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
179 ; GFX10_GFX11-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
180 ; GFX10_GFX11-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
181 ; GFX10_GFX11-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr3
182 ; GFX10_GFX11-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
183 ; GFX10_GFX11-NEXT: [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr2
184 ; GFX10_GFX11-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr4
185 ; GFX10_GFX11-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY [[COPY4]]
186 ; GFX10_GFX11-NEXT: [[S_MOV_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_MOV_B32 $exec_lo
187 ; GFX10_GFX11-NEXT: {{ $}}
188 ; GFX10_GFX11-NEXT: bb.2:
189 ; GFX10_GFX11-NEXT: successors: %bb.3(0x80000000)
190 ; GFX10_GFX11-NEXT: {{ $}}
191 ; GFX10_GFX11-NEXT: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY]], implicit $exec
192 ; GFX10_GFX11-NEXT: [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY1]], implicit $exec
193 ; GFX10_GFX11-NEXT: [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY2]], implicit $exec
194 ; GFX10_GFX11-NEXT: [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY3]], implicit $exec
195 ; GFX10_GFX11-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
196 ; GFX10_GFX11-NEXT: [[COPY7:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub0_sub1
197 ; GFX10_GFX11-NEXT: [[COPY8:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub2_sub3
198 ; GFX10_GFX11-NEXT: [[COPY9:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE1]].sub0_sub1
199 ; GFX10_GFX11-NEXT: [[COPY10:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE1]].sub2_sub3
200 ; GFX10_GFX11-NEXT: [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[COPY9]], [[COPY7]], implicit $exec
201 ; GFX10_GFX11-NEXT: [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[COPY10]], [[COPY8]], implicit $exec
202 ; GFX10_GFX11-NEXT: [[S_AND_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[V_CMP_EQ_U64_e64_]], [[V_CMP_EQ_U64_e64_1]], implicit-def dead $scc
203 ; GFX10_GFX11-NEXT: [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY5]], implicit $exec
204 ; GFX10_GFX11-NEXT: [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY5]], implicit $exec
205 ; GFX10_GFX11-NEXT: [[S_AND_B32_1:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[S_AND_B32_]], [[V_CMP_EQ_U32_e64_]], implicit-def dead $scc
206 ; GFX10_GFX11-NEXT: [[S_AND_SAVEEXEC_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_SAVEEXEC_B32 killed [[S_AND_B32_1]], implicit-def $exec, implicit-def $scc, implicit $exec
207 ; GFX10_GFX11-NEXT: {{ $}}
208 ; GFX10_GFX11-NEXT: bb.3:
209 ; GFX10_GFX11-NEXT: successors: %bb.4(0x40000000), %bb.2(0x40000000)
210 ; GFX10_GFX11-NEXT: {{ $}}
211 ; GFX10_GFX11-NEXT: [[TBUFFER_LOAD_FORMAT_X_OFFEN:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFEN [[COPY6]], [[REG_SEQUENCE1]], [[V_READFIRSTLANE_B32_4]], 0, 78, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
212 ; GFX10_GFX11-NEXT: $exec_lo = S_XOR_B32_term $exec_lo, [[S_AND_SAVEEXEC_B32_]], implicit-def $scc
213 ; GFX10_GFX11-NEXT: SI_WATERFALL_LOOP %bb.2, implicit $exec
214 ; GFX10_GFX11-NEXT: {{ $}}
215 ; GFX10_GFX11-NEXT: bb.4:
216 ; GFX10_GFX11-NEXT: successors: %bb.5(0x80000000)
217 ; GFX10_GFX11-NEXT: {{ $}}
218 ; GFX10_GFX11-NEXT: $exec_lo = S_MOV_B32_term [[S_MOV_B32_]]
219 ; GFX10_GFX11-NEXT: {{ $}}
220 ; GFX10_GFX11-NEXT: bb.5:
221 ; GFX10_GFX11-NEXT: $vgpr0 = COPY [[TBUFFER_LOAD_FORMAT_X_OFFEN]]
222 ; GFX10_GFX11-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
224 ; GFX12-LABEL: name: raw_tbuffer_load_f32__vgpr_rsrc__sgpr_voffset__vgpr_soffset
225 ; GFX12: bb.1 (%ir-block.0):
226 ; GFX12-NEXT: successors: %bb.2(0x80000000)
227 ; GFX12-NEXT: liveins: $sgpr2, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4
229 ; GFX12-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
230 ; GFX12-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
231 ; GFX12-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
232 ; GFX12-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr3
233 ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
234 ; GFX12-NEXT: [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr2
235 ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr4
236 ; GFX12-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY [[COPY4]]
237 ; GFX12-NEXT: [[S_MOV_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_MOV_B32 $exec_lo
240 ; GFX12-NEXT: successors: %bb.3(0x80000000)
242 ; GFX12-NEXT: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY]], implicit $exec
243 ; GFX12-NEXT: [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY1]], implicit $exec
244 ; GFX12-NEXT: [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY2]], implicit $exec
245 ; GFX12-NEXT: [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY3]], implicit $exec
246 ; GFX12-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
247 ; GFX12-NEXT: [[COPY7:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub0_sub1
248 ; GFX12-NEXT: [[COPY8:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub2_sub3
249 ; GFX12-NEXT: [[COPY9:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE1]].sub0_sub1
250 ; GFX12-NEXT: [[COPY10:%[0-9]+]]:sreg_64 = COPY [[REG_SEQUENCE1]].sub2_sub3
251 ; GFX12-NEXT: [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[COPY9]], [[COPY7]], implicit $exec
252 ; GFX12-NEXT: [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[COPY10]], [[COPY8]], implicit $exec
253 ; GFX12-NEXT: [[S_AND_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[V_CMP_EQ_U64_e64_]], [[V_CMP_EQ_U64_e64_1]], implicit-def dead $scc
254 ; GFX12-NEXT: [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY5]], implicit $exec
255 ; GFX12-NEXT: [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY5]], implicit $exec
256 ; GFX12-NEXT: [[S_AND_B32_1:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[S_AND_B32_]], [[V_CMP_EQ_U32_e64_]], implicit-def dead $scc
257 ; GFX12-NEXT: [[S_AND_SAVEEXEC_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_SAVEEXEC_B32 killed [[S_AND_B32_1]], implicit-def $exec, implicit-def $scc, implicit $exec
260 ; GFX12-NEXT: successors: %bb.4(0x40000000), %bb.2(0x40000000)
262 ; GFX12-NEXT: [[TBUFFER_LOAD_FORMAT_X_VBUFFER_OFFEN:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_VBUFFER_OFFEN [[COPY6]], [[REG_SEQUENCE1]], [[V_READFIRSTLANE_B32_4]], 0, 78, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
263 ; GFX12-NEXT: $exec_lo = S_XOR_B32_term $exec_lo, [[S_AND_SAVEEXEC_B32_]], implicit-def $scc
264 ; GFX12-NEXT: SI_WATERFALL_LOOP %bb.2, implicit $exec
267 ; GFX12-NEXT: successors: %bb.5(0x80000000)
269 ; GFX12-NEXT: $exec_lo = S_MOV_B32_term [[S_MOV_B32_]]
272 ; GFX12-NEXT: $vgpr0 = COPY [[TBUFFER_LOAD_FORMAT_X_VBUFFER_OFFEN]]
273 ; GFX12-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
274 %val = call float @llvm.amdgcn.raw.tbuffer.load.f32(<4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 78, i32 0)
278 define amdgpu_ps float @raw_tbuffer_load_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_glc(<4 x i32> inreg %rsrc, i32 %voffset, i32 inreg %soffset) {
279 ; GFX10_GFX11-LABEL: name: raw_tbuffer_load_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_glc
280 ; GFX10_GFX11: bb.1 (%ir-block.0):
281 ; GFX10_GFX11-NEXT: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0
282 ; GFX10_GFX11-NEXT: {{ $}}
283 ; GFX10_GFX11-NEXT: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
284 ; GFX10_GFX11-NEXT: [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
285 ; GFX10_GFX11-NEXT: [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
286 ; GFX10_GFX11-NEXT: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
287 ; GFX10_GFX11-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
288 ; GFX10_GFX11-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
289 ; GFX10_GFX11-NEXT: [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr6
290 ; GFX10_GFX11-NEXT: [[TBUFFER_LOAD_FORMAT_X_OFFEN:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFEN [[COPY4]], [[REG_SEQUENCE]], [[COPY5]], 0, 78, 1, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
291 ; GFX10_GFX11-NEXT: $vgpr0 = COPY [[TBUFFER_LOAD_FORMAT_X_OFFEN]]
292 ; GFX10_GFX11-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
294 ; GFX12-LABEL: name: raw_tbuffer_load_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_glc
295 ; GFX12: bb.1 (%ir-block.0):
296 ; GFX12-NEXT: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0
298 ; GFX12-NEXT: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
299 ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
300 ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
301 ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
302 ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
303 ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
304 ; GFX12-NEXT: [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr6
305 ; GFX12-NEXT: [[TBUFFER_LOAD_FORMAT_X_VBUFFER_OFFEN:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_VBUFFER_OFFEN [[COPY4]], [[REG_SEQUENCE]], [[COPY5]], 0, 78, 1, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
306 ; GFX12-NEXT: $vgpr0 = COPY [[TBUFFER_LOAD_FORMAT_X_VBUFFER_OFFEN]]
307 ; GFX12-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
308 %val = call float @llvm.amdgcn.raw.tbuffer.load.f32(<4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 78, i32 1)
312 define amdgpu_ps float @raw_tbuffer_load_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_slc(<4 x i32> inreg %rsrc, i32 %voffset, i32 inreg %soffset) {
313 ; GFX10_GFX11-LABEL: name: raw_tbuffer_load_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_slc
314 ; GFX10_GFX11: bb.1 (%ir-block.0):
315 ; GFX10_GFX11-NEXT: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0
316 ; GFX10_GFX11-NEXT: {{ $}}
317 ; GFX10_GFX11-NEXT: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
318 ; GFX10_GFX11-NEXT: [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
319 ; GFX10_GFX11-NEXT: [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
320 ; GFX10_GFX11-NEXT: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
321 ; GFX10_GFX11-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
322 ; GFX10_GFX11-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
323 ; GFX10_GFX11-NEXT: [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr6
324 ; GFX10_GFX11-NEXT: [[TBUFFER_LOAD_FORMAT_X_OFFEN:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFEN [[COPY4]], [[REG_SEQUENCE]], [[COPY5]], 0, 78, 2, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
325 ; GFX10_GFX11-NEXT: $vgpr0 = COPY [[TBUFFER_LOAD_FORMAT_X_OFFEN]]
326 ; GFX10_GFX11-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
328 ; GFX12-LABEL: name: raw_tbuffer_load_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_slc
329 ; GFX12: bb.1 (%ir-block.0):
330 ; GFX12-NEXT: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0
332 ; GFX12-NEXT: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
333 ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
334 ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
335 ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
336 ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
337 ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
338 ; GFX12-NEXT: [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr6
339 ; GFX12-NEXT: [[TBUFFER_LOAD_FORMAT_X_VBUFFER_OFFEN:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_VBUFFER_OFFEN [[COPY4]], [[REG_SEQUENCE]], [[COPY5]], 0, 78, 2, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
340 ; GFX12-NEXT: $vgpr0 = COPY [[TBUFFER_LOAD_FORMAT_X_VBUFFER_OFFEN]]
341 ; GFX12-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
342 %val = call float @llvm.amdgcn.raw.tbuffer.load.f32(<4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 78, i32 2)
346 define amdgpu_ps float @raw_tbuffer_load_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_slc_glc(<4 x i32> inreg %rsrc, i32 %voffset, i32 inreg %soffset) {
347 ; GFX10_GFX11-LABEL: name: raw_tbuffer_load_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_slc_glc
348 ; GFX10_GFX11: bb.1 (%ir-block.0):
349 ; GFX10_GFX11-NEXT: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0
350 ; GFX10_GFX11-NEXT: {{ $}}
351 ; GFX10_GFX11-NEXT: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
352 ; GFX10_GFX11-NEXT: [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
353 ; GFX10_GFX11-NEXT: [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
354 ; GFX10_GFX11-NEXT: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
355 ; GFX10_GFX11-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
356 ; GFX10_GFX11-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
357 ; GFX10_GFX11-NEXT: [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr6
358 ; GFX10_GFX11-NEXT: [[TBUFFER_LOAD_FORMAT_X_OFFEN:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFEN [[COPY4]], [[REG_SEQUENCE]], [[COPY5]], 0, 78, 3, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
359 ; GFX10_GFX11-NEXT: $vgpr0 = COPY [[TBUFFER_LOAD_FORMAT_X_OFFEN]]
360 ; GFX10_GFX11-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
362 ; GFX12-LABEL: name: raw_tbuffer_load_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_slc_glc
363 ; GFX12: bb.1 (%ir-block.0):
364 ; GFX12-NEXT: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0
366 ; GFX12-NEXT: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
367 ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
368 ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
369 ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
370 ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
371 ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
372 ; GFX12-NEXT: [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr6
373 ; GFX12-NEXT: [[TBUFFER_LOAD_FORMAT_X_VBUFFER_OFFEN:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_VBUFFER_OFFEN [[COPY4]], [[REG_SEQUENCE]], [[COPY5]], 0, 78, 3, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
374 ; GFX12-NEXT: $vgpr0 = COPY [[TBUFFER_LOAD_FORMAT_X_VBUFFER_OFFEN]]
375 ; GFX12-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
376 %val = call float @llvm.amdgcn.raw.tbuffer.load.f32(<4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 78, i32 3)
380 define amdgpu_ps float @raw_tbuffer_load_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_dlc(<4 x i32> inreg %rsrc, i32 %voffset, i32 inreg %soffset) {
381 ; GFX10_GFX11-LABEL: name: raw_tbuffer_load_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_dlc
382 ; GFX10_GFX11: bb.1 (%ir-block.0):
383 ; GFX10_GFX11-NEXT: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0
384 ; GFX10_GFX11-NEXT: {{ $}}
385 ; GFX10_GFX11-NEXT: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
386 ; GFX10_GFX11-NEXT: [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
387 ; GFX10_GFX11-NEXT: [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
388 ; GFX10_GFX11-NEXT: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
389 ; GFX10_GFX11-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
390 ; GFX10_GFX11-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
391 ; GFX10_GFX11-NEXT: [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr6
392 ; GFX10_GFX11-NEXT: [[TBUFFER_LOAD_FORMAT_X_OFFEN:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFEN [[COPY4]], [[REG_SEQUENCE]], [[COPY5]], 0, 78, 4, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
393 ; GFX10_GFX11-NEXT: $vgpr0 = COPY [[TBUFFER_LOAD_FORMAT_X_OFFEN]]
394 ; GFX10_GFX11-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
396 ; GFX12-LABEL: name: raw_tbuffer_load_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_dlc
397 ; GFX12: bb.1 (%ir-block.0):
398 ; GFX12-NEXT: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $vgpr0
400 ; GFX12-NEXT: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr2
401 ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr3
402 ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sreg_32 = COPY $sgpr4
403 ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5
404 ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
405 ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
406 ; GFX12-NEXT: [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr6
407 ; GFX12-NEXT: [[TBUFFER_LOAD_FORMAT_X_VBUFFER_OFFEN:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_VBUFFER_OFFEN [[COPY4]], [[REG_SEQUENCE]], [[COPY5]], 0, 78, 4, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8)
408 ; GFX12-NEXT: $vgpr0 = COPY [[TBUFFER_LOAD_FORMAT_X_VBUFFER_OFFEN]]
409 ; GFX12-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
410 %val = call float @llvm.amdgcn.raw.tbuffer.load.f32(<4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 78, i32 4)
414 declare float @llvm.amdgcn.raw.tbuffer.load.f32(<4 x i32>, i32, i32, i32 immarg, i32 immarg) #0
415 declare <2 x float> @llvm.amdgcn.raw.tbuffer.load.v2f32(<4 x i32>, i32, i32, i32 immarg, i32 immarg) #0
416 declare <3 x float> @llvm.amdgcn.raw.tbuffer.load.v3f32(<4 x i32>, i32, i32, i32 immarg, i32 immarg) #0
417 declare <4 x float> @llvm.amdgcn.raw.tbuffer.load.v4f32(<4 x i32>, i32, i32, i32 immarg, i32 immarg) #0
419 attributes #0 = { nounwind readonly }