1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc -mtriple=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN %s
4 define amdgpu_kernel void @lsh8_or_and(ptr addrspace(1) nocapture %arg, i32 %arg1) {
5 ; GCN-LABEL: lsh8_or_and:
7 ; GCN-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
8 ; GCN-NEXT: s_load_dword s2, s[4:5], 0x2c
9 ; GCN-NEXT: v_lshlrev_b32_e32 v0, 2, v0
10 ; GCN-NEXT: v_mov_b32_e32 v3, 0x6050400
11 ; GCN-NEXT: s_waitcnt lgkmcnt(0)
12 ; GCN-NEXT: v_mov_b32_e32 v1, s1
13 ; GCN-NEXT: v_add_u32_e32 v0, vcc, s0, v0
14 ; GCN-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
15 ; GCN-NEXT: flat_load_dword v2, v[0:1]
16 ; GCN-NEXT: s_waitcnt vmcnt(0)
17 ; GCN-NEXT: v_perm_b32 v2, v2, s2, v3
18 ; GCN-NEXT: flat_store_dword v[0:1], v2
21 %id = tail call i32 @llvm.amdgcn.workitem.id.x()
22 %gep = getelementptr i32, ptr addrspace(1) %arg, i32 %id
23 %tmp = load i32, ptr addrspace(1) %gep, align 4
24 %tmp2 = shl i32 %tmp, 8
25 %tmp3 = and i32 %arg1, 255
26 %tmp4 = or i32 %tmp2, %tmp3
27 store i32 %tmp4, ptr addrspace(1) %gep, align 4
31 define amdgpu_kernel void @lsr24_or_and(ptr addrspace(1) nocapture %arg, i32 %arg1) {
32 ; GCN-LABEL: lsr24_or_and:
34 ; GCN-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
35 ; GCN-NEXT: s_load_dword s2, s[4:5], 0x2c
36 ; GCN-NEXT: v_lshlrev_b32_e32 v0, 2, v0
37 ; GCN-NEXT: v_mov_b32_e32 v3, 0x7060503
38 ; GCN-NEXT: s_waitcnt lgkmcnt(0)
39 ; GCN-NEXT: v_mov_b32_e32 v1, s1
40 ; GCN-NEXT: v_add_u32_e32 v0, vcc, s0, v0
41 ; GCN-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
42 ; GCN-NEXT: flat_load_dword v2, v[0:1]
43 ; GCN-NEXT: s_waitcnt vmcnt(0)
44 ; GCN-NEXT: v_perm_b32 v2, s2, v2, v3
45 ; GCN-NEXT: flat_store_dword v[0:1], v2
48 %id = tail call i32 @llvm.amdgcn.workitem.id.x()
49 %gep = getelementptr i32, ptr addrspace(1) %arg, i32 %id
50 %tmp = load i32, ptr addrspace(1) %gep, align 4
51 %tmp2 = lshr i32 %tmp, 24
52 %tmp3 = and i32 %arg1, 4294967040 ; 0xffffff00
53 %tmp4 = or i32 %tmp2, %tmp3
54 store i32 %tmp4, ptr addrspace(1) %gep, align 4
58 define amdgpu_kernel void @and_or_lsr24(ptr addrspace(1) nocapture %arg, i32 %arg1) {
59 ; GCN-LABEL: and_or_lsr24:
61 ; GCN-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
62 ; GCN-NEXT: s_load_dword s2, s[4:5], 0x2c
63 ; GCN-NEXT: v_lshlrev_b32_e32 v0, 2, v0
64 ; GCN-NEXT: v_mov_b32_e32 v3, 0x7060503
65 ; GCN-NEXT: s_waitcnt lgkmcnt(0)
66 ; GCN-NEXT: v_mov_b32_e32 v1, s1
67 ; GCN-NEXT: v_add_u32_e32 v0, vcc, s0, v0
68 ; GCN-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
69 ; GCN-NEXT: flat_load_dword v2, v[0:1]
70 ; GCN-NEXT: s_waitcnt vmcnt(0)
71 ; GCN-NEXT: v_perm_b32 v2, v2, s2, v3
72 ; GCN-NEXT: v_xor_b32_e32 v2, 0x80000000, v2
73 ; GCN-NEXT: flat_store_dword v[0:1], v2
76 %id = tail call i32 @llvm.amdgcn.workitem.id.x()
77 %gep = getelementptr i32, ptr addrspace(1) %arg, i32 %id
78 %tmp = load i32, ptr addrspace(1) %gep, align 4
79 %tmp2 = and i32 %tmp, 4294967040 ; 0xffffff00
80 %tmp3 = lshr i32 %arg1, 24
81 %tmp4 = or i32 %tmp2, %tmp3
82 %tmp5 = xor i32 %tmp4, -2147483648
83 store i32 %tmp5, ptr addrspace(1) %gep, align 4
87 define amdgpu_kernel void @and_or_and(ptr addrspace(1) nocapture %arg, i32 %arg1) {
88 ; GCN-LABEL: and_or_and:
90 ; GCN-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
91 ; GCN-NEXT: s_load_dword s2, s[4:5], 0x2c
92 ; GCN-NEXT: v_lshlrev_b32_e32 v0, 2, v0
93 ; GCN-NEXT: v_mov_b32_e32 v3, 0x7020500
94 ; GCN-NEXT: s_waitcnt lgkmcnt(0)
95 ; GCN-NEXT: v_mov_b32_e32 v1, s1
96 ; GCN-NEXT: v_add_u32_e32 v0, vcc, s0, v0
97 ; GCN-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
98 ; GCN-NEXT: flat_load_dword v2, v[0:1]
99 ; GCN-NEXT: s_waitcnt vmcnt(0)
100 ; GCN-NEXT: v_perm_b32 v2, v2, s2, v3
101 ; GCN-NEXT: flat_store_dword v[0:1], v2
104 %id = tail call i32 @llvm.amdgcn.workitem.id.x()
105 %gep = getelementptr i32, ptr addrspace(1) %arg, i32 %id
106 %tmp = load i32, ptr addrspace(1) %gep, align 4
107 %tmp2 = and i32 %tmp, -16711936
108 %tmp3 = and i32 %arg1, 16711935
109 %tmp4 = or i32 %tmp2, %tmp3
110 store i32 %tmp4, ptr addrspace(1) %gep, align 4
114 ; FIXME: produce v_alignbit_b32 v2, v2, s0, 24 instead of v_perm
115 define amdgpu_kernel void @lsh8_or_lsr24(ptr addrspace(1) nocapture %arg, i32 %arg1) {
116 ; GCN-LABEL: lsh8_or_lsr24:
117 ; GCN: ; %bb.0: ; %bb
118 ; GCN-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
119 ; GCN-NEXT: s_load_dword s2, s[4:5], 0x2c
120 ; GCN-NEXT: v_lshlrev_b32_e32 v0, 2, v0
121 ; GCN-NEXT: v_mov_b32_e32 v3, 0x2010007
122 ; GCN-NEXT: s_waitcnt lgkmcnt(0)
123 ; GCN-NEXT: v_mov_b32_e32 v1, s1
124 ; GCN-NEXT: v_add_u32_e32 v0, vcc, s0, v0
125 ; GCN-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
126 ; GCN-NEXT: flat_load_dword v2, v[0:1]
127 ; GCN-NEXT: s_waitcnt vmcnt(0)
128 ; GCN-NEXT: v_perm_b32 v2, s2, v2, v3
129 ; GCN-NEXT: flat_store_dword v[0:1], v2
132 %id = tail call i32 @llvm.amdgcn.workitem.id.x()
133 %gep = getelementptr i32, ptr addrspace(1) %arg, i32 %id
134 %tmp = load i32, ptr addrspace(1) %gep, align 4
135 %tmp2 = shl i32 %tmp, 8
136 %tmp3 = lshr i32 %arg1, 24
137 %tmp4 = or i32 %tmp2, %tmp3
138 store i32 %tmp4, ptr addrspace(1) %gep, align 4
142 define amdgpu_kernel void @lsh16_or_lsr24(ptr addrspace(1) nocapture %arg, i32 %arg1) {
143 ; GCN-LABEL: lsh16_or_lsr24:
144 ; GCN: ; %bb.0: ; %bb
145 ; GCN-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
146 ; GCN-NEXT: s_load_dword s2, s[4:5], 0x2c
147 ; GCN-NEXT: v_lshlrev_b32_e32 v0, 2, v0
148 ; GCN-NEXT: v_mov_b32_e32 v3, 0x5040c03
149 ; GCN-NEXT: s_waitcnt lgkmcnt(0)
150 ; GCN-NEXT: v_mov_b32_e32 v1, s1
151 ; GCN-NEXT: v_add_u32_e32 v0, vcc, s0, v0
152 ; GCN-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
153 ; GCN-NEXT: flat_load_dword v2, v[0:1]
154 ; GCN-NEXT: s_waitcnt vmcnt(0)
155 ; GCN-NEXT: v_perm_b32 v2, v2, s2, v3
156 ; GCN-NEXT: flat_store_dword v[0:1], v2
159 %id = tail call i32 @llvm.amdgcn.workitem.id.x()
160 %gep = getelementptr i32, ptr addrspace(1) %arg, i32 %id
161 %tmp = load i32, ptr addrspace(1) %gep, align 4
162 %tmp2 = shl i32 %tmp, 16
163 %tmp3 = lshr i32 %arg1, 24
164 %tmp4 = or i32 %tmp2, %tmp3
165 store i32 %tmp4, ptr addrspace(1) %gep, align 4
169 define amdgpu_kernel void @and_xor_and(ptr addrspace(1) nocapture %arg, i32 %arg1) {
170 ; GCN-LABEL: and_xor_and:
171 ; GCN: ; %bb.0: ; %bb
172 ; GCN-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
173 ; GCN-NEXT: s_load_dword s2, s[4:5], 0x2c
174 ; GCN-NEXT: v_lshlrev_b32_e32 v0, 2, v0
175 ; GCN-NEXT: v_mov_b32_e32 v3, 0x7020104
176 ; GCN-NEXT: s_waitcnt lgkmcnt(0)
177 ; GCN-NEXT: v_mov_b32_e32 v1, s1
178 ; GCN-NEXT: v_add_u32_e32 v0, vcc, s0, v0
179 ; GCN-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
180 ; GCN-NEXT: flat_load_dword v2, v[0:1]
181 ; GCN-NEXT: s_waitcnt vmcnt(0)
182 ; GCN-NEXT: v_perm_b32 v2, v2, s2, v3
183 ; GCN-NEXT: flat_store_dword v[0:1], v2
186 %id = tail call i32 @llvm.amdgcn.workitem.id.x()
187 %gep = getelementptr i32, ptr addrspace(1) %arg, i32 %id
188 %tmp = load i32, ptr addrspace(1) %gep, align 4
189 %tmp2 = and i32 %tmp, -16776961
190 %tmp3 = and i32 %arg1, 16776960
191 %tmp4 = xor i32 %tmp2, %tmp3
192 store i32 %tmp4, ptr addrspace(1) %gep, align 4
196 ; FIXME here should have been "v_perm_b32" with 0xffff0500 mask.
197 define amdgpu_kernel void @and_or_or_and(ptr addrspace(1) nocapture %arg, i32 %arg1) {
198 ; GCN-LABEL: and_or_or_and:
199 ; GCN: ; %bb.0: ; %bb
200 ; GCN-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
201 ; GCN-NEXT: s_load_dword s2, s[4:5], 0x2c
202 ; GCN-NEXT: v_lshlrev_b32_e32 v0, 2, v0
203 ; GCN-NEXT: s_waitcnt lgkmcnt(0)
204 ; GCN-NEXT: v_mov_b32_e32 v1, s1
205 ; GCN-NEXT: v_add_u32_e32 v0, vcc, s0, v0
206 ; GCN-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
207 ; GCN-NEXT: flat_load_dword v2, v[0:1]
208 ; GCN-NEXT: s_and_b32 s0, s2, 0xff00
209 ; GCN-NEXT: s_or_b32 s0, s0, 0xffff0000
210 ; GCN-NEXT: s_waitcnt vmcnt(0)
211 ; GCN-NEXT: v_and_b32_e32 v2, 0xff00ff, v2
212 ; GCN-NEXT: v_or_b32_e32 v2, s0, v2
213 ; GCN-NEXT: flat_store_dword v[0:1], v2
216 %id = tail call i32 @llvm.amdgcn.workitem.id.x()
217 %gep = getelementptr i32, ptr addrspace(1) %arg, i32 %id
218 %tmp = load i32, ptr addrspace(1) %gep, align 4
219 %and = and i32 %tmp, 16711935 ; 0x00ff00ff
220 %tmp1 = and i32 %arg1, 4294967040 ; 0xffffff00
221 %tmp2 = or i32 %tmp1, -65536
222 %tmp3 = or i32 %tmp2, %and
223 store i32 %tmp3, ptr addrspace(1) %gep, align 4
227 define amdgpu_kernel void @and_or_and_shl(ptr addrspace(1) nocapture %arg, i32 %arg1) {
228 ; GCN-LABEL: and_or_and_shl:
229 ; GCN: ; %bb.0: ; %bb
230 ; GCN-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
231 ; GCN-NEXT: s_load_dword s2, s[4:5], 0x2c
232 ; GCN-NEXT: v_lshlrev_b32_e32 v0, 2, v0
233 ; GCN-NEXT: v_mov_b32_e32 v3, 0x50c0c00
234 ; GCN-NEXT: s_waitcnt lgkmcnt(0)
235 ; GCN-NEXT: v_mov_b32_e32 v1, s1
236 ; GCN-NEXT: v_add_u32_e32 v0, vcc, s0, v0
237 ; GCN-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
238 ; GCN-NEXT: flat_load_dword v2, v[0:1]
239 ; GCN-NEXT: s_waitcnt vmcnt(0)
240 ; GCN-NEXT: v_perm_b32 v2, v2, s2, v3
241 ; GCN-NEXT: flat_store_dword v[0:1], v2
244 %id = tail call i32 @llvm.amdgcn.workitem.id.x()
245 %gep = getelementptr i32, ptr addrspace(1) %arg, i32 %id
246 %tmp = load i32, ptr addrspace(1) %gep, align 4
247 %tmp2 = shl i32 %tmp, 16
248 %tmp3 = and i32 %arg1, 65535
249 %tmp4 = or i32 %tmp2, %tmp3
250 %and = and i32 %tmp4, 4278190335
251 store i32 %and, ptr addrspace(1) %gep, align 4
255 define amdgpu_kernel void @or_and_or(ptr addrspace(1) nocapture %arg, i32 %arg1) {
256 ; GCN-LABEL: or_and_or:
257 ; GCN: ; %bb.0: ; %bb
258 ; GCN-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
259 ; GCN-NEXT: s_load_dword s2, s[4:5], 0x2c
260 ; GCN-NEXT: v_lshlrev_b32_e32 v0, 2, v0
261 ; GCN-NEXT: v_mov_b32_e32 v3, 0x7020104
262 ; GCN-NEXT: s_waitcnt lgkmcnt(0)
263 ; GCN-NEXT: v_mov_b32_e32 v1, s1
264 ; GCN-NEXT: v_add_u32_e32 v0, vcc, s0, v0
265 ; GCN-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
266 ; GCN-NEXT: flat_load_dword v2, v[0:1]
267 ; GCN-NEXT: s_waitcnt vmcnt(0)
268 ; GCN-NEXT: v_perm_b32 v2, v2, s2, v3
269 ; GCN-NEXT: flat_store_dword v[0:1], v2
272 %id = tail call i32 @llvm.amdgcn.workitem.id.x()
273 %gep = getelementptr i32, ptr addrspace(1) %arg, i32 %id
274 %tmp = load i32, ptr addrspace(1) %gep, align 4
275 %or1 = or i32 %tmp, 16776960 ; 0x00ffff00
276 %or2 = or i32 %arg1, 4278190335 ; 0xff0000ff
277 %and = and i32 %or1, %or2
278 store i32 %and, ptr addrspace(1) %gep, align 4
282 ; FIXME here should have been "v_perm_b32" with 0xffff0500 mask.
283 define amdgpu_kernel void @known_ffff0500(ptr addrspace(1) nocapture %arg, i32 %arg1) {
284 ; GCN-LABEL: known_ffff0500:
285 ; GCN: ; %bb.0: ; %bb
286 ; GCN-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
287 ; GCN-NEXT: s_load_dword s2, s[4:5], 0x2c
288 ; GCN-NEXT: v_lshlrev_b32_e32 v0, 2, v0
289 ; GCN-NEXT: v_mov_b32_e32 v5, 0xffff8004
290 ; GCN-NEXT: s_waitcnt lgkmcnt(0)
291 ; GCN-NEXT: v_mov_b32_e32 v1, s1
292 ; GCN-NEXT: v_add_u32_e32 v0, vcc, s0, v0
293 ; GCN-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
294 ; GCN-NEXT: flat_load_dword v4, v[0:1]
295 ; GCN-NEXT: s_bitset1_b32 s2, 15
296 ; GCN-NEXT: v_mov_b32_e32 v3, s1
297 ; GCN-NEXT: v_mov_b32_e32 v2, s0
298 ; GCN-NEXT: s_and_b32 s0, s2, 0xff00
299 ; GCN-NEXT: s_or_b32 s0, s0, 0xffff0000
300 ; GCN-NEXT: s_waitcnt vmcnt(0)
301 ; GCN-NEXT: v_or_b32_e32 v4, 4, v4
302 ; GCN-NEXT: v_and_b32_e32 v4, 0xff00ff, v4
303 ; GCN-NEXT: v_or_b32_e32 v4, s0, v4
304 ; GCN-NEXT: flat_store_dword v[0:1], v4
305 ; GCN-NEXT: flat_store_dword v[2:3], v5
308 %id = tail call i32 @llvm.amdgcn.workitem.id.x()
309 %gep = getelementptr i32, ptr addrspace(1) %arg, i32 %id
310 %load = load i32, ptr addrspace(1) %gep, align 4
311 %mask1 = or i32 %arg1, 32768 ; 0x8000
312 %mask2 = or i32 %load, 4
313 %and = and i32 %mask2, 16711935 ; 0x00ff00ff
314 %tmp1 = and i32 %mask1, 4294967040 ; 0xffffff00
315 %tmp2 = or i32 %tmp1, 4294901760 ; 0xffff0000
316 %tmp3 = or i32 %tmp2, %and
317 store i32 %tmp3, ptr addrspace(1) %gep, align 4
318 %v = and i32 %tmp3, 4294934532 ; 0xffff8004
319 store i32 %v, ptr addrspace(1) %arg, align 4
323 define amdgpu_kernel void @known_050c0c00(ptr addrspace(1) nocapture %arg, i32 %arg1) {
324 ; GCN-LABEL: known_050c0c00:
325 ; GCN: ; %bb.0: ; %bb
326 ; GCN-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
327 ; GCN-NEXT: s_load_dword s2, s[4:5], 0x2c
328 ; GCN-NEXT: v_lshlrev_b32_e32 v0, 2, v0
329 ; GCN-NEXT: v_mov_b32_e32 v5, 0x50c0c00
330 ; GCN-NEXT: v_mov_b32_e32 v6, 4
331 ; GCN-NEXT: s_waitcnt lgkmcnt(0)
332 ; GCN-NEXT: v_mov_b32_e32 v1, s1
333 ; GCN-NEXT: v_add_u32_e32 v0, vcc, s0, v0
334 ; GCN-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
335 ; GCN-NEXT: flat_load_dword v4, v[0:1]
336 ; GCN-NEXT: s_or_b32 s2, s2, 4
337 ; GCN-NEXT: v_mov_b32_e32 v3, s1
338 ; GCN-NEXT: v_mov_b32_e32 v2, s0
339 ; GCN-NEXT: s_waitcnt vmcnt(0)
340 ; GCN-NEXT: v_perm_b32 v4, v4, s2, v5
341 ; GCN-NEXT: flat_store_dword v[0:1], v4
342 ; GCN-NEXT: flat_store_dword v[2:3], v6
345 %id = tail call i32 @llvm.amdgcn.workitem.id.x()
346 %gep = getelementptr i32, ptr addrspace(1) %arg, i32 %id
347 %tmp = load i32, ptr addrspace(1) %gep, align 4
348 %tmp2 = shl i32 %tmp, 16
349 %mask = or i32 %arg1, 4
350 %tmp3 = and i32 %mask, 65535
351 %tmp4 = or i32 %tmp2, %tmp3
352 %and = and i32 %tmp4, 4278190335
353 store i32 %and, ptr addrspace(1) %gep, align 4
354 %v = and i32 %and, 16776964
355 store i32 %v, ptr addrspace(1) %arg, align 4
359 define amdgpu_kernel void @known_ffff8004(ptr addrspace(1) nocapture %arg, i32 %arg1) {
360 ; GCN-LABEL: known_ffff8004:
361 ; GCN: ; %bb.0: ; %bb
362 ; GCN-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
363 ; GCN-NEXT: s_load_dword s2, s[4:5], 0x2c
364 ; GCN-NEXT: v_lshlrev_b32_e32 v0, 2, v0
365 ; GCN-NEXT: v_mov_b32_e32 v5, 0xffff0500
366 ; GCN-NEXT: v_mov_b32_e32 v6, 0xffff8004
367 ; GCN-NEXT: s_waitcnt lgkmcnt(0)
368 ; GCN-NEXT: v_mov_b32_e32 v1, s1
369 ; GCN-NEXT: v_add_u32_e32 v0, vcc, s0, v0
370 ; GCN-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
371 ; GCN-NEXT: flat_load_dword v4, v[0:1]
372 ; GCN-NEXT: s_or_b32 s2, s2, 4
373 ; GCN-NEXT: v_mov_b32_e32 v3, s1
374 ; GCN-NEXT: v_mov_b32_e32 v2, s0
375 ; GCN-NEXT: s_waitcnt vmcnt(0)
376 ; GCN-NEXT: v_or_b32_e32 v4, 0x8000, v4
377 ; GCN-NEXT: v_perm_b32 v4, v4, s2, v5
378 ; GCN-NEXT: flat_store_dword v[0:1], v4
379 ; GCN-NEXT: flat_store_dword v[2:3], v6
382 %id = tail call i32 @llvm.amdgcn.workitem.id.x()
383 %gep = getelementptr i32, ptr addrspace(1) %arg, i32 %id
384 %load = load i32, ptr addrspace(1) %gep, align 4
385 %mask1 = or i32 %arg1, 4
386 %mask2 = or i32 %load, 32768 ; 0x8000
387 %and = and i32 %mask1, 16711935 ; 0x00ff00ff
388 %tmp1 = and i32 %mask2, 4294967040 ; 0xffffff00
389 %tmp2 = or i32 %tmp1, 4294901760 ; 0xffff0000
390 %tmp3 = or i32 %tmp2, %and
391 store i32 %tmp3, ptr addrspace(1) %gep, align 4
392 %v = and i32 %tmp3, 4294934532 ; 0xffff8004
393 store i32 %v, ptr addrspace(1) %arg, align 4
397 declare i32 @llvm.amdgcn.workitem.id.x()