1 ; NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
2 ; RUN: llc -global-isel -stop-after=irtranslator -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -verify-machineinstrs -o - %s | FileCheck -enable-var-scope %s
4 ; Test that we don't insert code to pass implicit arguments we know
5 ; the callee does not need.
7 declare hidden void @extern()
9 define amdgpu_kernel void @kernel_call_no_workitem_ids() {
10 ; CHECK-LABEL: name: kernel_call_no_workitem_ids
11 ; CHECK: bb.1 (%ir-block.0):
12 ; CHECK-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
14 ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr16
15 ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr15
16 ; CHECK-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr14
17 ; CHECK-NEXT: [[COPY3:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
18 ; CHECK-NEXT: [[COPY4:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
19 ; CHECK-NEXT: [[COPY5:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
20 ; CHECK-NEXT: [[COPY6:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
21 ; CHECK-NEXT: ADJCALLSTACKUP 0, 0, implicit-def $scc
22 ; CHECK-NEXT: [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @extern
23 ; CHECK-NEXT: [[COPY7:%[0-9]+]]:_(p4) = COPY [[COPY5]]
24 ; CHECK-NEXT: [[COPY8:%[0-9]+]]:_(p4) = COPY [[COPY4]]
25 ; CHECK-NEXT: [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY6]](p4)
26 ; CHECK-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
27 ; CHECK-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY9]], [[C]](s64)
28 ; CHECK-NEXT: [[COPY10:%[0-9]+]]:_(s64) = COPY [[COPY3]]
29 ; CHECK-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY [[COPY2]]
30 ; CHECK-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY1]]
31 ; CHECK-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY]]
32 ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
33 ; CHECK-NEXT: [[COPY14:%[0-9]+]]:_(<4 x s32>) = COPY $private_rsrc_reg
34 ; CHECK-NEXT: $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY14]](<4 x s32>)
35 ; CHECK-NEXT: $sgpr4_sgpr5 = COPY [[COPY7]](p4)
36 ; CHECK-NEXT: $sgpr6_sgpr7 = COPY [[COPY8]](p4)
37 ; CHECK-NEXT: $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
38 ; CHECK-NEXT: $sgpr10_sgpr11 = COPY [[COPY10]](s64)
39 ; CHECK-NEXT: $sgpr12 = COPY [[COPY11]](s32)
40 ; CHECK-NEXT: $sgpr13 = COPY [[COPY12]](s32)
41 ; CHECK-NEXT: $sgpr14 = COPY [[COPY13]](s32)
42 ; CHECK-NEXT: $sgpr15 = COPY [[DEF]](s32)
43 ; CHECK-NEXT: $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @extern, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15
44 ; CHECK-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def $scc
45 ; CHECK-NEXT: S_ENDPGM 0
46 call void @extern() "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z"
50 define amdgpu_kernel void @kernel_call_no_workgroup_ids() {
51 ; CHECK-LABEL: name: kernel_call_no_workgroup_ids
52 ; CHECK: bb.1 (%ir-block.0):
53 ; CHECK-NEXT: liveins: $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
55 ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
56 ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
57 ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
58 ; CHECK-NEXT: [[COPY3:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
59 ; CHECK-NEXT: [[COPY4:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
60 ; CHECK-NEXT: [[COPY5:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
61 ; CHECK-NEXT: [[COPY6:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
62 ; CHECK-NEXT: ADJCALLSTACKUP 0, 0, implicit-def $scc
63 ; CHECK-NEXT: [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @extern
64 ; CHECK-NEXT: [[COPY7:%[0-9]+]]:_(p4) = COPY [[COPY5]]
65 ; CHECK-NEXT: [[COPY8:%[0-9]+]]:_(p4) = COPY [[COPY4]]
66 ; CHECK-NEXT: [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY6]](p4)
67 ; CHECK-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
68 ; CHECK-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY9]], [[C]](s64)
69 ; CHECK-NEXT: [[COPY10:%[0-9]+]]:_(s64) = COPY [[COPY3]]
70 ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
71 ; CHECK-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY [[COPY2]](s32)
72 ; CHECK-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32)
73 ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 10
74 ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[COPY12]], [[C1]](s32)
75 ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[COPY11]], [[SHL]]
76 ; CHECK-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
77 ; CHECK-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
78 ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY13]], [[C2]](s32)
79 ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
80 ; CHECK-NEXT: [[COPY14:%[0-9]+]]:_(<4 x s32>) = COPY $private_rsrc_reg
81 ; CHECK-NEXT: $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY14]](<4 x s32>)
82 ; CHECK-NEXT: $sgpr4_sgpr5 = COPY [[COPY7]](p4)
83 ; CHECK-NEXT: $sgpr6_sgpr7 = COPY [[COPY8]](p4)
84 ; CHECK-NEXT: $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
85 ; CHECK-NEXT: $sgpr10_sgpr11 = COPY [[COPY10]](s64)
86 ; CHECK-NEXT: $sgpr15 = COPY [[DEF]](s32)
87 ; CHECK-NEXT: $vgpr31 = COPY [[OR1]](s32)
88 ; CHECK-NEXT: $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @extern, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr15, implicit $vgpr31
89 ; CHECK-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def $scc
90 ; CHECK-NEXT: S_ENDPGM 0
91 call void @extern() "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z"
95 define amdgpu_kernel void @kernel_call_no_other_sgprs() {
96 ; CHECK-LABEL: name: kernel_call_no_other_sgprs
97 ; CHECK: bb.1 (%ir-block.0):
98 ; CHECK-NEXT: liveins: $vgpr0, $vgpr1, $vgpr2, $sgpr8_sgpr9
100 ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
101 ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
102 ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
103 ; CHECK-NEXT: [[COPY3:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
104 ; CHECK-NEXT: ADJCALLSTACKUP 0, 0, implicit-def $scc
105 ; CHECK-NEXT: [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @extern
106 ; CHECK-NEXT: [[COPY4:%[0-9]+]]:_(p4) = COPY [[COPY3]](p4)
107 ; CHECK-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
108 ; CHECK-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY4]], [[C]](s64)
109 ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
110 ; CHECK-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY [[COPY2]](s32)
111 ; CHECK-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32)
112 ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 10
113 ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[COPY6]], [[C1]](s32)
114 ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[COPY5]], [[SHL]]
115 ; CHECK-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
116 ; CHECK-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
117 ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY7]], [[C2]](s32)
118 ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
119 ; CHECK-NEXT: [[COPY8:%[0-9]+]]:_(<4 x s32>) = COPY $private_rsrc_reg
120 ; CHECK-NEXT: $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY8]](<4 x s32>)
121 ; CHECK-NEXT: $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
122 ; CHECK-NEXT: $sgpr15 = COPY [[DEF]](s32)
123 ; CHECK-NEXT: $vgpr31 = COPY [[OR1]](s32)
124 ; CHECK-NEXT: $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @extern, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr8_sgpr9, implicit $sgpr15, implicit $vgpr31
125 ; CHECK-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def $scc
126 ; CHECK-NEXT: S_ENDPGM 0
127 call void @extern() "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z"
131 define void @func_call_no_workitem_ids() {
132 ; CHECK-LABEL: name: func_call_no_workitem_ids
133 ; CHECK: bb.1 (%ir-block.0):
134 ; CHECK-NEXT: liveins: $sgpr12, $sgpr13, $sgpr14, $sgpr15, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
136 ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr15
137 ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr14
138 ; CHECK-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr13
139 ; CHECK-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr12
140 ; CHECK-NEXT: [[COPY4:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
141 ; CHECK-NEXT: [[COPY5:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
142 ; CHECK-NEXT: [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
143 ; CHECK-NEXT: [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
144 ; CHECK-NEXT: ADJCALLSTACKUP 0, 0, implicit-def $scc
145 ; CHECK-NEXT: [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @extern
146 ; CHECK-NEXT: [[COPY8:%[0-9]+]]:_(p4) = COPY [[COPY7]]
147 ; CHECK-NEXT: [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY6]]
148 ; CHECK-NEXT: [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY5]]
149 ; CHECK-NEXT: [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY4]]
150 ; CHECK-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY3]]
151 ; CHECK-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY2]]
152 ; CHECK-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY1]]
153 ; CHECK-NEXT: [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY]]
154 ; CHECK-NEXT: [[COPY16:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
155 ; CHECK-NEXT: $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY16]](<4 x s32>)
156 ; CHECK-NEXT: $sgpr4_sgpr5 = COPY [[COPY8]](p4)
157 ; CHECK-NEXT: $sgpr6_sgpr7 = COPY [[COPY9]](p4)
158 ; CHECK-NEXT: $sgpr8_sgpr9 = COPY [[COPY10]](p4)
159 ; CHECK-NEXT: $sgpr10_sgpr11 = COPY [[COPY11]](s64)
160 ; CHECK-NEXT: $sgpr12 = COPY [[COPY12]](s32)
161 ; CHECK-NEXT: $sgpr13 = COPY [[COPY13]](s32)
162 ; CHECK-NEXT: $sgpr14 = COPY [[COPY14]](s32)
163 ; CHECK-NEXT: $sgpr15 = COPY [[COPY15]](s32)
164 ; CHECK-NEXT: $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @extern, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15
165 ; CHECK-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def $scc
166 ; CHECK-NEXT: SI_RETURN
167 call void @extern() "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z"
171 define void @func_call_no_workgroup_ids() {
172 ; CHECK-LABEL: name: func_call_no_workgroup_ids
173 ; CHECK: bb.1 (%ir-block.0):
174 ; CHECK-NEXT: liveins: $sgpr15, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
176 ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr31
177 ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr15
178 ; CHECK-NEXT: [[COPY2:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
179 ; CHECK-NEXT: [[COPY3:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
180 ; CHECK-NEXT: [[COPY4:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
181 ; CHECK-NEXT: [[COPY5:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
182 ; CHECK-NEXT: ADJCALLSTACKUP 0, 0, implicit-def $scc
183 ; CHECK-NEXT: [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @extern
184 ; CHECK-NEXT: [[COPY6:%[0-9]+]]:_(p4) = COPY [[COPY5]]
185 ; CHECK-NEXT: [[COPY7:%[0-9]+]]:_(p4) = COPY [[COPY4]]
186 ; CHECK-NEXT: [[COPY8:%[0-9]+]]:_(p4) = COPY [[COPY3]]
187 ; CHECK-NEXT: [[COPY9:%[0-9]+]]:_(s64) = COPY [[COPY2]]
188 ; CHECK-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY [[COPY1]]
189 ; CHECK-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
190 ; CHECK-NEXT: [[COPY12:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
191 ; CHECK-NEXT: $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY12]](<4 x s32>)
192 ; CHECK-NEXT: $sgpr4_sgpr5 = COPY [[COPY6]](p4)
193 ; CHECK-NEXT: $sgpr6_sgpr7 = COPY [[COPY7]](p4)
194 ; CHECK-NEXT: $sgpr8_sgpr9 = COPY [[COPY8]](p4)
195 ; CHECK-NEXT: $sgpr10_sgpr11 = COPY [[COPY9]](s64)
196 ; CHECK-NEXT: $sgpr15 = COPY [[COPY10]](s32)
197 ; CHECK-NEXT: $vgpr31 = COPY [[COPY11]](s32)
198 ; CHECK-NEXT: $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @extern, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr15, implicit $vgpr31
199 ; CHECK-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def $scc
200 ; CHECK-NEXT: SI_RETURN
201 call void @extern() "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z"
205 define void @func_call_no_other_sgprs() {
206 ; CHECK-LABEL: name: func_call_no_other_sgprs
207 ; CHECK: bb.1 (%ir-block.0):
208 ; CHECK-NEXT: liveins: $sgpr15, $vgpr31, $sgpr8_sgpr9
210 ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr31
211 ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr15
212 ; CHECK-NEXT: [[COPY2:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
213 ; CHECK-NEXT: ADJCALLSTACKUP 0, 0, implicit-def $scc
214 ; CHECK-NEXT: [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @extern
215 ; CHECK-NEXT: [[COPY3:%[0-9]+]]:_(p4) = COPY [[COPY2]]
216 ; CHECK-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY [[COPY1]]
217 ; CHECK-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
218 ; CHECK-NEXT: [[COPY6:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
219 ; CHECK-NEXT: $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY6]](<4 x s32>)
220 ; CHECK-NEXT: $sgpr8_sgpr9 = COPY [[COPY3]](p4)
221 ; CHECK-NEXT: $sgpr15 = COPY [[COPY4]](s32)
222 ; CHECK-NEXT: $vgpr31 = COPY [[COPY5]](s32)
223 ; CHECK-NEXT: $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @extern, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr8_sgpr9, implicit $sgpr15, implicit $vgpr31
224 ; CHECK-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def $scc
225 ; CHECK-NEXT: SI_RETURN
226 call void @extern() "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z"