xref: /llvm-project/llvm/test/CodeGen/AMDGPU/shufflevector.v3i64.v3i64.ll (revision 585858aeb6247b3892218edb9d353c63f1c33186) 
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
2; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 < %s | FileCheck -check-prefixes=GFX9,GFX900 %s
3; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx90a < %s | FileCheck -check-prefixes=GFX9,GFX90APLUS,GFX90A %s
4; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx940 < %s | FileCheck -check-prefixes=GFX9,GFX90APLUS,GFX940 %s
5
6
7define void @v_shuffle_v3i64_v3i64__u_u_u(ptr addrspace(1) inreg %ptr) {
8; GFX9-LABEL: v_shuffle_v3i64_v3i64__u_u_u:
9; GFX9:       ; %bb.0:
10; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
11; GFX9-NEXT:    s_setpc_b64 s[30:31]
12  %vec0 = call <3 x i64> asm "; def $0", "=v"()
13  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> poison
14  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
15  ret void
16}
17
18define void @v_shuffle_v3i64_v3i64__0_u_u(ptr addrspace(1) inreg %ptr) {
19; GFX900-LABEL: v_shuffle_v3i64_v3i64__0_u_u:
20; GFX900:       ; %bb.0:
21; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
22; GFX900-NEXT:    v_mov_b32_e32 v6, 0
23; GFX900-NEXT:    ;;#ASMSTART
24; GFX900-NEXT:    ; def v[0:5]
25; GFX900-NEXT:    ;;#ASMEND
26; GFX900-NEXT:    global_store_dwordx2 v6, v[4:5], s[16:17] offset:16
27; GFX900-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
28; GFX900-NEXT:    s_waitcnt vmcnt(0)
29; GFX900-NEXT:    s_setpc_b64 s[30:31]
30;
31; GFX90A-LABEL: v_shuffle_v3i64_v3i64__0_u_u:
32; GFX90A:       ; %bb.0:
33; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
34; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
35; GFX90A-NEXT:    ;;#ASMSTART
36; GFX90A-NEXT:    ; def v[0:5]
37; GFX90A-NEXT:    ;;#ASMEND
38; GFX90A-NEXT:    global_store_dwordx2 v6, v[4:5], s[16:17] offset:16
39; GFX90A-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
40; GFX90A-NEXT:    s_waitcnt vmcnt(0)
41; GFX90A-NEXT:    s_setpc_b64 s[30:31]
42;
43; GFX940-LABEL: v_shuffle_v3i64_v3i64__0_u_u:
44; GFX940:       ; %bb.0:
45; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
46; GFX940-NEXT:    v_mov_b32_e32 v6, 0
47; GFX940-NEXT:    ;;#ASMSTART
48; GFX940-NEXT:    ; def v[0:5]
49; GFX940-NEXT:    ;;#ASMEND
50; GFX940-NEXT:    global_store_dwordx2 v6, v[4:5], s[0:1] offset:16 sc0 sc1
51; GFX940-NEXT:    global_store_dwordx4 v6, v[0:3], s[0:1] sc0 sc1
52; GFX940-NEXT:    s_waitcnt vmcnt(0)
53; GFX940-NEXT:    s_setpc_b64 s[30:31]
54  %vec0 = call <3 x i64> asm "; def $0", "=v"()
55  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 0, i32 poison, i32 poison>
56  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
57  ret void
58}
59
60define void @v_shuffle_v3i64_v3i64__1_u_u(ptr addrspace(1) inreg %ptr) {
61; GFX900-LABEL: v_shuffle_v3i64_v3i64__1_u_u:
62; GFX900:       ; %bb.0:
63; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
64; GFX900-NEXT:    v_mov_b32_e32 v6, 0
65; GFX900-NEXT:    ;;#ASMSTART
66; GFX900-NEXT:    ; def v[0:5]
67; GFX900-NEXT:    ;;#ASMEND
68; GFX900-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
69; GFX900-NEXT:    s_waitcnt vmcnt(0)
70; GFX900-NEXT:    s_setpc_b64 s[30:31]
71;
72; GFX90A-LABEL: v_shuffle_v3i64_v3i64__1_u_u:
73; GFX90A:       ; %bb.0:
74; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
75; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
76; GFX90A-NEXT:    ;;#ASMSTART
77; GFX90A-NEXT:    ; def v[0:5]
78; GFX90A-NEXT:    ;;#ASMEND
79; GFX90A-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
80; GFX90A-NEXT:    s_waitcnt vmcnt(0)
81; GFX90A-NEXT:    s_setpc_b64 s[30:31]
82;
83; GFX940-LABEL: v_shuffle_v3i64_v3i64__1_u_u:
84; GFX940:       ; %bb.0:
85; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
86; GFX940-NEXT:    v_mov_b32_e32 v6, 0
87; GFX940-NEXT:    ;;#ASMSTART
88; GFX940-NEXT:    ; def v[0:5]
89; GFX940-NEXT:    ;;#ASMEND
90; GFX940-NEXT:    global_store_dwordx4 v6, v[2:5], s[0:1] sc0 sc1
91; GFX940-NEXT:    s_waitcnt vmcnt(0)
92; GFX940-NEXT:    s_setpc_b64 s[30:31]
93  %vec0 = call <3 x i64> asm "; def $0", "=v"()
94  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 1, i32 poison, i32 poison>
95  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
96  ret void
97}
98
99define void @v_shuffle_v3i64_v3i64__2_u_u(ptr addrspace(1) inreg %ptr) {
100; GFX900-LABEL: v_shuffle_v3i64_v3i64__2_u_u:
101; GFX900:       ; %bb.0:
102; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
103; GFX900-NEXT:    ;;#ASMSTART
104; GFX900-NEXT:    ; def v[0:5]
105; GFX900-NEXT:    ;;#ASMEND
106; GFX900-NEXT:    v_mov_b32_e32 v6, 0
107; GFX900-NEXT:    v_mov_b32_e32 v0, v4
108; GFX900-NEXT:    v_mov_b32_e32 v1, v5
109; GFX900-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
110; GFX900-NEXT:    s_waitcnt vmcnt(0)
111; GFX900-NEXT:    s_setpc_b64 s[30:31]
112;
113; GFX90A-LABEL: v_shuffle_v3i64_v3i64__2_u_u:
114; GFX90A:       ; %bb.0:
115; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
116; GFX90A-NEXT:    ;;#ASMSTART
117; GFX90A-NEXT:    ; def v[0:5]
118; GFX90A-NEXT:    ;;#ASMEND
119; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
120; GFX90A-NEXT:    v_mov_b32_e32 v0, v4
121; GFX90A-NEXT:    v_mov_b32_e32 v1, v5
122; GFX90A-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
123; GFX90A-NEXT:    s_waitcnt vmcnt(0)
124; GFX90A-NEXT:    s_setpc_b64 s[30:31]
125;
126; GFX940-LABEL: v_shuffle_v3i64_v3i64__2_u_u:
127; GFX940:       ; %bb.0:
128; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
129; GFX940-NEXT:    ;;#ASMSTART
130; GFX940-NEXT:    ; def v[0:5]
131; GFX940-NEXT:    ;;#ASMEND
132; GFX940-NEXT:    v_mov_b32_e32 v6, 0
133; GFX940-NEXT:    v_mov_b32_e32 v0, v4
134; GFX940-NEXT:    v_mov_b32_e32 v1, v5
135; GFX940-NEXT:    global_store_dwordx4 v6, v[0:3], s[0:1] sc0 sc1
136; GFX940-NEXT:    s_waitcnt vmcnt(0)
137; GFX940-NEXT:    s_setpc_b64 s[30:31]
138  %vec0 = call <3 x i64> asm "; def $0", "=v"()
139  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 2, i32 poison, i32 poison>
140  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
141  ret void
142}
143
144define void @v_shuffle_v3i64_v3i64__3_u_u(ptr addrspace(1) inreg %ptr) {
145; GFX9-LABEL: v_shuffle_v3i64_v3i64__3_u_u:
146; GFX9:       ; %bb.0:
147; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
148; GFX9-NEXT:    s_setpc_b64 s[30:31]
149  %vec0 = call <3 x i64> asm "; def $0", "=v"()
150  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 3, i32 poison, i32 poison>
151  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
152  ret void
153}
154
155define void @v_shuffle_v3i64_v3i64__4_u_u(ptr addrspace(1) inreg %ptr) {
156; GFX900-LABEL: v_shuffle_v3i64_v3i64__4_u_u:
157; GFX900:       ; %bb.0:
158; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
159; GFX900-NEXT:    v_mov_b32_e32 v6, 0
160; GFX900-NEXT:    ;;#ASMSTART
161; GFX900-NEXT:    ; def v[0:5]
162; GFX900-NEXT:    ;;#ASMEND
163; GFX900-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
164; GFX900-NEXT:    s_waitcnt vmcnt(0)
165; GFX900-NEXT:    s_setpc_b64 s[30:31]
166;
167; GFX90A-LABEL: v_shuffle_v3i64_v3i64__4_u_u:
168; GFX90A:       ; %bb.0:
169; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
170; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
171; GFX90A-NEXT:    ;;#ASMSTART
172; GFX90A-NEXT:    ; def v[0:5]
173; GFX90A-NEXT:    ;;#ASMEND
174; GFX90A-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
175; GFX90A-NEXT:    s_waitcnt vmcnt(0)
176; GFX90A-NEXT:    s_setpc_b64 s[30:31]
177;
178; GFX940-LABEL: v_shuffle_v3i64_v3i64__4_u_u:
179; GFX940:       ; %bb.0:
180; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
181; GFX940-NEXT:    v_mov_b32_e32 v6, 0
182; GFX940-NEXT:    ;;#ASMSTART
183; GFX940-NEXT:    ; def v[0:5]
184; GFX940-NEXT:    ;;#ASMEND
185; GFX940-NEXT:    global_store_dwordx4 v6, v[2:5], s[0:1] sc0 sc1
186; GFX940-NEXT:    s_waitcnt vmcnt(0)
187; GFX940-NEXT:    s_setpc_b64 s[30:31]
188  %vec0 = call <3 x i64> asm "; def $0", "=v"()
189  %vec1 = call <3 x i64> asm "; def $0", "=v"()
190  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 4, i32 poison, i32 poison>
191  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
192  ret void
193}
194
195define void @v_shuffle_v3i64_v3i64__5_u_u(ptr addrspace(1) inreg %ptr) {
196; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_u_u:
197; GFX900:       ; %bb.0:
198; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
199; GFX900-NEXT:    ;;#ASMSTART
200; GFX900-NEXT:    ; def v[0:5]
201; GFX900-NEXT:    ;;#ASMEND
202; GFX900-NEXT:    v_mov_b32_e32 v6, 0
203; GFX900-NEXT:    v_mov_b32_e32 v0, v4
204; GFX900-NEXT:    v_mov_b32_e32 v1, v5
205; GFX900-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
206; GFX900-NEXT:    s_waitcnt vmcnt(0)
207; GFX900-NEXT:    s_setpc_b64 s[30:31]
208;
209; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_u_u:
210; GFX90A:       ; %bb.0:
211; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
212; GFX90A-NEXT:    ;;#ASMSTART
213; GFX90A-NEXT:    ; def v[0:5]
214; GFX90A-NEXT:    ;;#ASMEND
215; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
216; GFX90A-NEXT:    v_mov_b32_e32 v0, v4
217; GFX90A-NEXT:    v_mov_b32_e32 v1, v5
218; GFX90A-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
219; GFX90A-NEXT:    s_waitcnt vmcnt(0)
220; GFX90A-NEXT:    s_setpc_b64 s[30:31]
221;
222; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_u_u:
223; GFX940:       ; %bb.0:
224; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
225; GFX940-NEXT:    ;;#ASMSTART
226; GFX940-NEXT:    ; def v[0:5]
227; GFX940-NEXT:    ;;#ASMEND
228; GFX940-NEXT:    v_mov_b32_e32 v6, 0
229; GFX940-NEXT:    v_mov_b32_e32 v0, v4
230; GFX940-NEXT:    v_mov_b32_e32 v1, v5
231; GFX940-NEXT:    global_store_dwordx4 v6, v[0:3], s[0:1] sc0 sc1
232; GFX940-NEXT:    s_waitcnt vmcnt(0)
233; GFX940-NEXT:    s_setpc_b64 s[30:31]
234  %vec0 = call <3 x i64> asm "; def $0", "=v"()
235  %vec1 = call <3 x i64> asm "; def $0", "=v"()
236  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 poison, i32 poison>
237  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
238  ret void
239}
240
241define void @v_shuffle_v3i64_v3i64__5_0_u(ptr addrspace(1) inreg %ptr) {
242; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_0_u:
243; GFX900:       ; %bb.0:
244; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
245; GFX900-NEXT:    ;;#ASMSTART
246; GFX900-NEXT:    ; def v[0:5]
247; GFX900-NEXT:    ;;#ASMEND
248; GFX900-NEXT:    ;;#ASMSTART
249; GFX900-NEXT:    ; def v[2:7]
250; GFX900-NEXT:    ;;#ASMEND
251; GFX900-NEXT:    v_mov_b32_e32 v8, 0
252; GFX900-NEXT:    v_mov_b32_e32 v2, v6
253; GFX900-NEXT:    v_mov_b32_e32 v3, v7
254; GFX900-NEXT:    v_mov_b32_e32 v4, v0
255; GFX900-NEXT:    v_mov_b32_e32 v5, v1
256; GFX900-NEXT:    global_store_dwordx4 v8, v[2:5], s[16:17]
257; GFX900-NEXT:    s_waitcnt vmcnt(0)
258; GFX900-NEXT:    s_setpc_b64 s[30:31]
259;
260; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_0_u:
261; GFX90A:       ; %bb.0:
262; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
263; GFX90A-NEXT:    ;;#ASMSTART
264; GFX90A-NEXT:    ; def v[0:5]
265; GFX90A-NEXT:    ;;#ASMEND
266; GFX90A-NEXT:    ;;#ASMSTART
267; GFX90A-NEXT:    ; def v[2:7]
268; GFX90A-NEXT:    ;;#ASMEND
269; GFX90A-NEXT:    v_mov_b32_e32 v8, 0
270; GFX90A-NEXT:    v_mov_b32_e32 v2, v6
271; GFX90A-NEXT:    v_mov_b32_e32 v3, v7
272; GFX90A-NEXT:    v_mov_b32_e32 v4, v0
273; GFX90A-NEXT:    v_mov_b32_e32 v5, v1
274; GFX90A-NEXT:    global_store_dwordx4 v8, v[2:5], s[16:17]
275; GFX90A-NEXT:    s_waitcnt vmcnt(0)
276; GFX90A-NEXT:    s_setpc_b64 s[30:31]
277;
278; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_0_u:
279; GFX940:       ; %bb.0:
280; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
281; GFX940-NEXT:    ;;#ASMSTART
282; GFX940-NEXT:    ; def v[0:5]
283; GFX940-NEXT:    ;;#ASMEND
284; GFX940-NEXT:    v_mov_b32_e32 v8, 0
285; GFX940-NEXT:    ;;#ASMSTART
286; GFX940-NEXT:    ; def v[2:7]
287; GFX940-NEXT:    ;;#ASMEND
288; GFX940-NEXT:    s_nop 0
289; GFX940-NEXT:    v_mov_b32_e32 v2, v6
290; GFX940-NEXT:    v_mov_b32_e32 v3, v7
291; GFX940-NEXT:    v_mov_b32_e32 v4, v0
292; GFX940-NEXT:    v_mov_b32_e32 v5, v1
293; GFX940-NEXT:    global_store_dwordx4 v8, v[2:5], s[0:1] sc0 sc1
294; GFX940-NEXT:    s_waitcnt vmcnt(0)
295; GFX940-NEXT:    s_setpc_b64 s[30:31]
296  %vec0 = call <3 x i64> asm "; def $0", "=v"()
297  %vec1 = call <3 x i64> asm "; def $0", "=v"()
298  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 0, i32 poison>
299  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
300  ret void
301}
302
303define void @v_shuffle_v3i64_v3i64__5_1_u(ptr addrspace(1) inreg %ptr) {
304; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_1_u:
305; GFX900:       ; %bb.0:
306; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
307; GFX900-NEXT:    ;;#ASMSTART
308; GFX900-NEXT:    ; def v[0:5]
309; GFX900-NEXT:    ;;#ASMEND
310; GFX900-NEXT:    v_mov_b32_e32 v10, 0
311; GFX900-NEXT:    ;;#ASMSTART
312; GFX900-NEXT:    ; def v[4:9]
313; GFX900-NEXT:    ;;#ASMEND
314; GFX900-NEXT:    v_mov_b32_e32 v0, v8
315; GFX900-NEXT:    v_mov_b32_e32 v1, v9
316; GFX900-NEXT:    global_store_dwordx4 v10, v[0:3], s[16:17]
317; GFX900-NEXT:    s_waitcnt vmcnt(0)
318; GFX900-NEXT:    s_setpc_b64 s[30:31]
319;
320; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_1_u:
321; GFX90A:       ; %bb.0:
322; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
323; GFX90A-NEXT:    ;;#ASMSTART
324; GFX90A-NEXT:    ; def v[0:5]
325; GFX90A-NEXT:    ;;#ASMEND
326; GFX90A-NEXT:    v_mov_b32_e32 v10, 0
327; GFX90A-NEXT:    ;;#ASMSTART
328; GFX90A-NEXT:    ; def v[4:9]
329; GFX90A-NEXT:    ;;#ASMEND
330; GFX90A-NEXT:    v_mov_b32_e32 v0, v8
331; GFX90A-NEXT:    v_mov_b32_e32 v1, v9
332; GFX90A-NEXT:    global_store_dwordx4 v10, v[0:3], s[16:17]
333; GFX90A-NEXT:    s_waitcnt vmcnt(0)
334; GFX90A-NEXT:    s_setpc_b64 s[30:31]
335;
336; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_1_u:
337; GFX940:       ; %bb.0:
338; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
339; GFX940-NEXT:    ;;#ASMSTART
340; GFX940-NEXT:    ; def v[0:5]
341; GFX940-NEXT:    ;;#ASMEND
342; GFX940-NEXT:    v_mov_b32_e32 v10, 0
343; GFX940-NEXT:    ;;#ASMSTART
344; GFX940-NEXT:    ; def v[4:9]
345; GFX940-NEXT:    ;;#ASMEND
346; GFX940-NEXT:    s_nop 0
347; GFX940-NEXT:    v_mov_b32_e32 v0, v8
348; GFX940-NEXT:    v_mov_b32_e32 v1, v9
349; GFX940-NEXT:    global_store_dwordx4 v10, v[0:3], s[0:1] sc0 sc1
350; GFX940-NEXT:    s_waitcnt vmcnt(0)
351; GFX940-NEXT:    s_setpc_b64 s[30:31]
352  %vec0 = call <3 x i64> asm "; def $0", "=v"()
353  %vec1 = call <3 x i64> asm "; def $0", "=v"()
354  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 1, i32 poison>
355  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
356  ret void
357}
358
359define void @v_shuffle_v3i64_v3i64__5_2_u(ptr addrspace(1) inreg %ptr) {
360; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_2_u:
361; GFX900:       ; %bb.0:
362; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
363; GFX900-NEXT:    ;;#ASMSTART
364; GFX900-NEXT:    ; def v[0:5]
365; GFX900-NEXT:    ;;#ASMEND
366; GFX900-NEXT:    v_mov_b32_e32 v12, 0
367; GFX900-NEXT:    ;;#ASMSTART
368; GFX900-NEXT:    ; def v[6:11]
369; GFX900-NEXT:    ;;#ASMEND
370; GFX900-NEXT:    v_mov_b32_e32 v2, v10
371; GFX900-NEXT:    v_mov_b32_e32 v3, v11
372; GFX900-NEXT:    global_store_dwordx4 v12, v[2:5], s[16:17]
373; GFX900-NEXT:    s_waitcnt vmcnt(0)
374; GFX900-NEXT:    s_setpc_b64 s[30:31]
375;
376; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_2_u:
377; GFX90A:       ; %bb.0:
378; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
379; GFX90A-NEXT:    ;;#ASMSTART
380; GFX90A-NEXT:    ; def v[0:5]
381; GFX90A-NEXT:    ;;#ASMEND
382; GFX90A-NEXT:    v_mov_b32_e32 v12, 0
383; GFX90A-NEXT:    ;;#ASMSTART
384; GFX90A-NEXT:    ; def v[6:11]
385; GFX90A-NEXT:    ;;#ASMEND
386; GFX90A-NEXT:    v_mov_b32_e32 v2, v10
387; GFX90A-NEXT:    v_mov_b32_e32 v3, v11
388; GFX90A-NEXT:    global_store_dwordx4 v12, v[2:5], s[16:17]
389; GFX90A-NEXT:    s_waitcnt vmcnt(0)
390; GFX90A-NEXT:    s_setpc_b64 s[30:31]
391;
392; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_2_u:
393; GFX940:       ; %bb.0:
394; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
395; GFX940-NEXT:    ;;#ASMSTART
396; GFX940-NEXT:    ; def v[0:5]
397; GFX940-NEXT:    ;;#ASMEND
398; GFX940-NEXT:    v_mov_b32_e32 v12, 0
399; GFX940-NEXT:    ;;#ASMSTART
400; GFX940-NEXT:    ; def v[6:11]
401; GFX940-NEXT:    ;;#ASMEND
402; GFX940-NEXT:    s_nop 0
403; GFX940-NEXT:    v_mov_b32_e32 v2, v10
404; GFX940-NEXT:    v_mov_b32_e32 v3, v11
405; GFX940-NEXT:    global_store_dwordx4 v12, v[2:5], s[0:1] sc0 sc1
406; GFX940-NEXT:    s_waitcnt vmcnt(0)
407; GFX940-NEXT:    s_setpc_b64 s[30:31]
408  %vec0 = call <3 x i64> asm "; def $0", "=v"()
409  %vec1 = call <3 x i64> asm "; def $0", "=v"()
410  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 2, i32 poison>
411  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
412  ret void
413}
414
415define void @v_shuffle_v3i64_v3i64__5_3_u(ptr addrspace(1) inreg %ptr) {
416; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_3_u:
417; GFX900:       ; %bb.0:
418; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
419; GFX900-NEXT:    ;;#ASMSTART
420; GFX900-NEXT:    ; def v[0:5]
421; GFX900-NEXT:    ;;#ASMEND
422; GFX900-NEXT:    v_mov_b32_e32 v6, 0
423; GFX900-NEXT:    v_mov_b32_e32 v2, v4
424; GFX900-NEXT:    v_mov_b32_e32 v3, v5
425; GFX900-NEXT:    v_mov_b32_e32 v4, v0
426; GFX900-NEXT:    v_mov_b32_e32 v5, v1
427; GFX900-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
428; GFX900-NEXT:    s_waitcnt vmcnt(0)
429; GFX900-NEXT:    s_setpc_b64 s[30:31]
430;
431; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_3_u:
432; GFX90A:       ; %bb.0:
433; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
434; GFX90A-NEXT:    ;;#ASMSTART
435; GFX90A-NEXT:    ; def v[0:5]
436; GFX90A-NEXT:    ;;#ASMEND
437; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
438; GFX90A-NEXT:    v_mov_b32_e32 v2, v4
439; GFX90A-NEXT:    v_mov_b32_e32 v3, v5
440; GFX90A-NEXT:    v_mov_b32_e32 v4, v0
441; GFX90A-NEXT:    v_mov_b32_e32 v5, v1
442; GFX90A-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
443; GFX90A-NEXT:    s_waitcnt vmcnt(0)
444; GFX90A-NEXT:    s_setpc_b64 s[30:31]
445;
446; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_3_u:
447; GFX940:       ; %bb.0:
448; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
449; GFX940-NEXT:    ;;#ASMSTART
450; GFX940-NEXT:    ; def v[0:5]
451; GFX940-NEXT:    ;;#ASMEND
452; GFX940-NEXT:    v_mov_b32_e32 v6, 0
453; GFX940-NEXT:    v_mov_b32_e32 v2, v4
454; GFX940-NEXT:    v_mov_b32_e32 v3, v5
455; GFX940-NEXT:    v_mov_b32_e32 v4, v0
456; GFX940-NEXT:    v_mov_b32_e32 v5, v1
457; GFX940-NEXT:    global_store_dwordx4 v6, v[2:5], s[0:1] sc0 sc1
458; GFX940-NEXT:    s_waitcnt vmcnt(0)
459; GFX940-NEXT:    s_setpc_b64 s[30:31]
460  %vec0 = call <3 x i64> asm "; def $0", "=v"()
461  %vec1 = call <3 x i64> asm "; def $0", "=v"()
462  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 3, i32 poison>
463  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
464  ret void
465}
466
467define void @v_shuffle_v3i64_v3i64__5_4_u(ptr addrspace(1) inreg %ptr) {
468; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_4_u:
469; GFX900:       ; %bb.0:
470; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
471; GFX900-NEXT:    ;;#ASMSTART
472; GFX900-NEXT:    ; def v[0:5]
473; GFX900-NEXT:    ;;#ASMEND
474; GFX900-NEXT:    v_mov_b32_e32 v6, 0
475; GFX900-NEXT:    v_mov_b32_e32 v0, v4
476; GFX900-NEXT:    v_mov_b32_e32 v1, v5
477; GFX900-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
478; GFX900-NEXT:    s_waitcnt vmcnt(0)
479; GFX900-NEXT:    s_setpc_b64 s[30:31]
480;
481; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_4_u:
482; GFX90A:       ; %bb.0:
483; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
484; GFX90A-NEXT:    ;;#ASMSTART
485; GFX90A-NEXT:    ; def v[0:5]
486; GFX90A-NEXT:    ;;#ASMEND
487; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
488; GFX90A-NEXT:    v_mov_b32_e32 v0, v4
489; GFX90A-NEXT:    v_mov_b32_e32 v1, v5
490; GFX90A-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
491; GFX90A-NEXT:    s_waitcnt vmcnt(0)
492; GFX90A-NEXT:    s_setpc_b64 s[30:31]
493;
494; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_4_u:
495; GFX940:       ; %bb.0:
496; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
497; GFX940-NEXT:    ;;#ASMSTART
498; GFX940-NEXT:    ; def v[0:5]
499; GFX940-NEXT:    ;;#ASMEND
500; GFX940-NEXT:    v_mov_b32_e32 v6, 0
501; GFX940-NEXT:    v_mov_b32_e32 v0, v4
502; GFX940-NEXT:    v_mov_b32_e32 v1, v5
503; GFX940-NEXT:    global_store_dwordx4 v6, v[0:3], s[0:1] sc0 sc1
504; GFX940-NEXT:    s_waitcnt vmcnt(0)
505; GFX940-NEXT:    s_setpc_b64 s[30:31]
506  %vec0 = call <3 x i64> asm "; def $0", "=v"()
507  %vec1 = call <3 x i64> asm "; def $0", "=v"()
508  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 4, i32 poison>
509  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
510  ret void
511}
512
513define void @v_shuffle_v3i64_v3i64__5_5_u(ptr addrspace(1) inreg %ptr) {
514; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_5_u:
515; GFX900:       ; %bb.0:
516; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
517; GFX900-NEXT:    ;;#ASMSTART
518; GFX900-NEXT:    ; def v[0:5]
519; GFX900-NEXT:    ;;#ASMEND
520; GFX900-NEXT:    v_mov_b32_e32 v6, 0
521; GFX900-NEXT:    v_mov_b32_e32 v2, v4
522; GFX900-NEXT:    v_mov_b32_e32 v3, v5
523; GFX900-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
524; GFX900-NEXT:    s_waitcnt vmcnt(0)
525; GFX900-NEXT:    s_setpc_b64 s[30:31]
526;
527; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_5_u:
528; GFX90A:       ; %bb.0:
529; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
530; GFX90A-NEXT:    ;;#ASMSTART
531; GFX90A-NEXT:    ; def v[0:5]
532; GFX90A-NEXT:    ;;#ASMEND
533; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
534; GFX90A-NEXT:    v_mov_b32_e32 v2, v4
535; GFX90A-NEXT:    v_mov_b32_e32 v3, v5
536; GFX90A-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
537; GFX90A-NEXT:    s_waitcnt vmcnt(0)
538; GFX90A-NEXT:    s_setpc_b64 s[30:31]
539;
540; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_5_u:
541; GFX940:       ; %bb.0:
542; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
543; GFX940-NEXT:    ;;#ASMSTART
544; GFX940-NEXT:    ; def v[0:5]
545; GFX940-NEXT:    ;;#ASMEND
546; GFX940-NEXT:    v_mov_b32_e32 v6, 0
547; GFX940-NEXT:    v_mov_b32_e32 v2, v4
548; GFX940-NEXT:    v_mov_b32_e32 v3, v5
549; GFX940-NEXT:    global_store_dwordx4 v6, v[2:5], s[0:1] sc0 sc1
550; GFX940-NEXT:    s_waitcnt vmcnt(0)
551; GFX940-NEXT:    s_setpc_b64 s[30:31]
552  %vec0 = call <3 x i64> asm "; def $0", "=v"()
553  %vec1 = call <3 x i64> asm "; def $0", "=v"()
554  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 5, i32 poison>
555  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
556  ret void
557}
558
559define void @v_shuffle_v3i64_v3i64__5_5_0(ptr addrspace(1) inreg %ptr) {
560; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_5_0:
561; GFX900:       ; %bb.0:
562; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
563; GFX900-NEXT:    ;;#ASMSTART
564; GFX900-NEXT:    ; def v[0:5]
565; GFX900-NEXT:    ;;#ASMEND
566; GFX900-NEXT:    ;;#ASMSTART
567; GFX900-NEXT:    ; def v[2:7]
568; GFX900-NEXT:    ;;#ASMEND
569; GFX900-NEXT:    v_mov_b32_e32 v8, 0
570; GFX900-NEXT:    v_mov_b32_e32 v4, v6
571; GFX900-NEXT:    v_mov_b32_e32 v5, v7
572; GFX900-NEXT:    global_store_dwordx2 v8, v[0:1], s[16:17] offset:16
573; GFX900-NEXT:    global_store_dwordx4 v8, v[4:7], s[16:17]
574; GFX900-NEXT:    s_waitcnt vmcnt(0)
575; GFX900-NEXT:    s_setpc_b64 s[30:31]
576;
577; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_5_0:
578; GFX90A:       ; %bb.0:
579; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
580; GFX90A-NEXT:    ;;#ASMSTART
581; GFX90A-NEXT:    ; def v[0:5]
582; GFX90A-NEXT:    ;;#ASMEND
583; GFX90A-NEXT:    ;;#ASMSTART
584; GFX90A-NEXT:    ; def v[2:7]
585; GFX90A-NEXT:    ;;#ASMEND
586; GFX90A-NEXT:    v_mov_b32_e32 v8, 0
587; GFX90A-NEXT:    v_mov_b32_e32 v4, v6
588; GFX90A-NEXT:    v_mov_b32_e32 v5, v7
589; GFX90A-NEXT:    global_store_dwordx2 v8, v[0:1], s[16:17] offset:16
590; GFX90A-NEXT:    global_store_dwordx4 v8, v[4:7], s[16:17]
591; GFX90A-NEXT:    s_waitcnt vmcnt(0)
592; GFX90A-NEXT:    s_setpc_b64 s[30:31]
593;
594; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_5_0:
595; GFX940:       ; %bb.0:
596; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
597; GFX940-NEXT:    ;;#ASMSTART
598; GFX940-NEXT:    ; def v[0:5]
599; GFX940-NEXT:    ;;#ASMEND
600; GFX940-NEXT:    v_mov_b32_e32 v8, 0
601; GFX940-NEXT:    ;;#ASMSTART
602; GFX940-NEXT:    ; def v[2:7]
603; GFX940-NEXT:    ;;#ASMEND
604; GFX940-NEXT:    global_store_dwordx2 v8, v[0:1], s[0:1] offset:16 sc0 sc1
605; GFX940-NEXT:    v_mov_b32_e32 v4, v6
606; GFX940-NEXT:    v_mov_b32_e32 v5, v7
607; GFX940-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] sc0 sc1
608; GFX940-NEXT:    s_waitcnt vmcnt(0)
609; GFX940-NEXT:    s_setpc_b64 s[30:31]
610  %vec0 = call <3 x i64> asm "; def $0", "=v"()
611  %vec1 = call <3 x i64> asm "; def $0", "=v"()
612  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 5, i32 0>
613  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
614  ret void
615}
616
617define void @v_shuffle_v3i64_v3i64__5_5_1(ptr addrspace(1) inreg %ptr) {
618; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_5_1:
619; GFX900:       ; %bb.0:
620; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
621; GFX900-NEXT:    ;;#ASMSTART
622; GFX900-NEXT:    ; def v[0:5]
623; GFX900-NEXT:    ;;#ASMEND
624; GFX900-NEXT:    ;;#ASMSTART
625; GFX900-NEXT:    ; def v[4:9]
626; GFX900-NEXT:    ;;#ASMEND
627; GFX900-NEXT:    v_mov_b32_e32 v10, 0
628; GFX900-NEXT:    v_mov_b32_e32 v6, v8
629; GFX900-NEXT:    v_mov_b32_e32 v7, v9
630; GFX900-NEXT:    global_store_dwordx2 v10, v[2:3], s[16:17] offset:16
631; GFX900-NEXT:    global_store_dwordx4 v10, v[6:9], s[16:17]
632; GFX900-NEXT:    s_waitcnt vmcnt(0)
633; GFX900-NEXT:    s_setpc_b64 s[30:31]
634;
635; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_5_1:
636; GFX90A:       ; %bb.0:
637; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
638; GFX90A-NEXT:    ;;#ASMSTART
639; GFX90A-NEXT:    ; def v[0:5]
640; GFX90A-NEXT:    ;;#ASMEND
641; GFX90A-NEXT:    ;;#ASMSTART
642; GFX90A-NEXT:    ; def v[4:9]
643; GFX90A-NEXT:    ;;#ASMEND
644; GFX90A-NEXT:    v_mov_b32_e32 v10, 0
645; GFX90A-NEXT:    v_mov_b32_e32 v6, v8
646; GFX90A-NEXT:    v_mov_b32_e32 v7, v9
647; GFX90A-NEXT:    global_store_dwordx2 v10, v[2:3], s[16:17] offset:16
648; GFX90A-NEXT:    global_store_dwordx4 v10, v[6:9], s[16:17]
649; GFX90A-NEXT:    s_waitcnt vmcnt(0)
650; GFX90A-NEXT:    s_setpc_b64 s[30:31]
651;
652; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_5_1:
653; GFX940:       ; %bb.0:
654; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
655; GFX940-NEXT:    ;;#ASMSTART
656; GFX940-NEXT:    ; def v[0:5]
657; GFX940-NEXT:    ;;#ASMEND
658; GFX940-NEXT:    v_mov_b32_e32 v10, 0
659; GFX940-NEXT:    ;;#ASMSTART
660; GFX940-NEXT:    ; def v[4:9]
661; GFX940-NEXT:    ;;#ASMEND
662; GFX940-NEXT:    global_store_dwordx2 v10, v[2:3], s[0:1] offset:16 sc0 sc1
663; GFX940-NEXT:    v_mov_b32_e32 v6, v8
664; GFX940-NEXT:    v_mov_b32_e32 v7, v9
665; GFX940-NEXT:    global_store_dwordx4 v10, v[6:9], s[0:1] sc0 sc1
666; GFX940-NEXT:    s_waitcnt vmcnt(0)
667; GFX940-NEXT:    s_setpc_b64 s[30:31]
668  %vec0 = call <3 x i64> asm "; def $0", "=v"()
669  %vec1 = call <3 x i64> asm "; def $0", "=v"()
670  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 5, i32 1>
671  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
672  ret void
673}
674
675define void @v_shuffle_v3i64_v3i64__5_5_2(ptr addrspace(1) inreg %ptr) {
676; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_5_2:
677; GFX900:       ; %bb.0:
678; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
679; GFX900-NEXT:    ;;#ASMSTART
680; GFX900-NEXT:    ; def v[6:11]
681; GFX900-NEXT:    ;;#ASMEND
682; GFX900-NEXT:    v_mov_b32_e32 v12, 0
683; GFX900-NEXT:    v_mov_b32_e32 v8, v10
684; GFX900-NEXT:    v_mov_b32_e32 v9, v11
685; GFX900-NEXT:    ;;#ASMSTART
686; GFX900-NEXT:    ; def v[0:5]
687; GFX900-NEXT:    ;;#ASMEND
688; GFX900-NEXT:    global_store_dwordx2 v12, v[4:5], s[16:17] offset:16
689; GFX900-NEXT:    global_store_dwordx4 v12, v[8:11], s[16:17]
690; GFX900-NEXT:    s_waitcnt vmcnt(0)
691; GFX900-NEXT:    s_setpc_b64 s[30:31]
692;
693; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_5_2:
694; GFX90A:       ; %bb.0:
695; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
696; GFX90A-NEXT:    ;;#ASMSTART
697; GFX90A-NEXT:    ; def v[6:11]
698; GFX90A-NEXT:    ;;#ASMEND
699; GFX90A-NEXT:    v_mov_b32_e32 v12, 0
700; GFX90A-NEXT:    v_mov_b32_e32 v8, v10
701; GFX90A-NEXT:    v_mov_b32_e32 v9, v11
702; GFX90A-NEXT:    ;;#ASMSTART
703; GFX90A-NEXT:    ; def v[0:5]
704; GFX90A-NEXT:    ;;#ASMEND
705; GFX90A-NEXT:    global_store_dwordx2 v12, v[4:5], s[16:17] offset:16
706; GFX90A-NEXT:    global_store_dwordx4 v12, v[8:11], s[16:17]
707; GFX90A-NEXT:    s_waitcnt vmcnt(0)
708; GFX90A-NEXT:    s_setpc_b64 s[30:31]
709;
710; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_5_2:
711; GFX940:       ; %bb.0:
712; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
713; GFX940-NEXT:    ;;#ASMSTART
714; GFX940-NEXT:    ; def v[6:11]
715; GFX940-NEXT:    ;;#ASMEND
716; GFX940-NEXT:    v_mov_b32_e32 v12, 0
717; GFX940-NEXT:    v_mov_b32_e32 v8, v10
718; GFX940-NEXT:    v_mov_b32_e32 v9, v11
719; GFX940-NEXT:    ;;#ASMSTART
720; GFX940-NEXT:    ; def v[0:5]
721; GFX940-NEXT:    ;;#ASMEND
722; GFX940-NEXT:    global_store_dwordx2 v12, v[4:5], s[0:1] offset:16 sc0 sc1
723; GFX940-NEXT:    global_store_dwordx4 v12, v[8:11], s[0:1] sc0 sc1
724; GFX940-NEXT:    s_waitcnt vmcnt(0)
725; GFX940-NEXT:    s_setpc_b64 s[30:31]
726  %vec0 = call <3 x i64> asm "; def $0", "=v"()
727  %vec1 = call <3 x i64> asm "; def $0", "=v"()
728  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 5, i32 2>
729  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
730  ret void
731}
732
733define void @v_shuffle_v3i64_v3i64__5_5_3(ptr addrspace(1) inreg %ptr) {
734; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_5_3:
735; GFX900:       ; %bb.0:
736; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
737; GFX900-NEXT:    ;;#ASMSTART
738; GFX900-NEXT:    ; def v[0:5]
739; GFX900-NEXT:    ;;#ASMEND
740; GFX900-NEXT:    v_mov_b32_e32 v6, 0
741; GFX900-NEXT:    v_mov_b32_e32 v2, v4
742; GFX900-NEXT:    v_mov_b32_e32 v3, v5
743; GFX900-NEXT:    global_store_dwordx2 v6, v[0:1], s[16:17] offset:16
744; GFX900-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
745; GFX900-NEXT:    s_waitcnt vmcnt(0)
746; GFX900-NEXT:    s_setpc_b64 s[30:31]
747;
748; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_5_3:
749; GFX90A:       ; %bb.0:
750; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
751; GFX90A-NEXT:    ;;#ASMSTART
752; GFX90A-NEXT:    ; def v[0:5]
753; GFX90A-NEXT:    ;;#ASMEND
754; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
755; GFX90A-NEXT:    v_mov_b32_e32 v2, v4
756; GFX90A-NEXT:    v_mov_b32_e32 v3, v5
757; GFX90A-NEXT:    global_store_dwordx2 v6, v[0:1], s[16:17] offset:16
758; GFX90A-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
759; GFX90A-NEXT:    s_waitcnt vmcnt(0)
760; GFX90A-NEXT:    s_setpc_b64 s[30:31]
761;
762; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_5_3:
763; GFX940:       ; %bb.0:
764; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
765; GFX940-NEXT:    ;;#ASMSTART
766; GFX940-NEXT:    ; def v[0:5]
767; GFX940-NEXT:    ;;#ASMEND
768; GFX940-NEXT:    v_mov_b32_e32 v6, 0
769; GFX940-NEXT:    v_mov_b32_e32 v2, v4
770; GFX940-NEXT:    v_mov_b32_e32 v3, v5
771; GFX940-NEXT:    global_store_dwordx2 v6, v[0:1], s[0:1] offset:16 sc0 sc1
772; GFX940-NEXT:    global_store_dwordx4 v6, v[2:5], s[0:1] sc0 sc1
773; GFX940-NEXT:    s_waitcnt vmcnt(0)
774; GFX940-NEXT:    s_setpc_b64 s[30:31]
775  %vec0 = call <3 x i64> asm "; def $0", "=v"()
776  %vec1 = call <3 x i64> asm "; def $0", "=v"()
777  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 5, i32 3>
778  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
779  ret void
780}
781
782define void @v_shuffle_v3i64_v3i64__5_5_4(ptr addrspace(1) inreg %ptr) {
783; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_5_4:
784; GFX900:       ; %bb.0:
785; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
786; GFX900-NEXT:    v_mov_b32_e32 v6, 0
787; GFX900-NEXT:    ;;#ASMSTART
788; GFX900-NEXT:    ; def v[0:5]
789; GFX900-NEXT:    ;;#ASMEND
790; GFX900-NEXT:    global_store_dwordx2 v6, v[2:3], s[16:17] offset:16
791; GFX900-NEXT:    v_mov_b32_e32 v2, v4
792; GFX900-NEXT:    v_mov_b32_e32 v3, v5
793; GFX900-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
794; GFX900-NEXT:    s_waitcnt vmcnt(0)
795; GFX900-NEXT:    s_setpc_b64 s[30:31]
796;
797; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_5_4:
798; GFX90A:       ; %bb.0:
799; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
800; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
801; GFX90A-NEXT:    ;;#ASMSTART
802; GFX90A-NEXT:    ; def v[0:5]
803; GFX90A-NEXT:    ;;#ASMEND
804; GFX90A-NEXT:    global_store_dwordx2 v6, v[2:3], s[16:17] offset:16
805; GFX90A-NEXT:    v_mov_b32_e32 v2, v4
806; GFX90A-NEXT:    v_mov_b32_e32 v3, v5
807; GFX90A-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
808; GFX90A-NEXT:    s_waitcnt vmcnt(0)
809; GFX90A-NEXT:    s_setpc_b64 s[30:31]
810;
811; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_5_4:
812; GFX940:       ; %bb.0:
813; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
814; GFX940-NEXT:    v_mov_b32_e32 v6, 0
815; GFX940-NEXT:    ;;#ASMSTART
816; GFX940-NEXT:    ; def v[0:5]
817; GFX940-NEXT:    ;;#ASMEND
818; GFX940-NEXT:    global_store_dwordx2 v6, v[2:3], s[0:1] offset:16 sc0 sc1
819; GFX940-NEXT:    v_mov_b32_e32 v2, v4
820; GFX940-NEXT:    v_mov_b32_e32 v3, v5
821; GFX940-NEXT:    global_store_dwordx4 v6, v[2:5], s[0:1] sc0 sc1
822; GFX940-NEXT:    s_waitcnt vmcnt(0)
823; GFX940-NEXT:    s_setpc_b64 s[30:31]
824  %vec0 = call <3 x i64> asm "; def $0", "=v"()
825  %vec1 = call <3 x i64> asm "; def $0", "=v"()
826  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 5, i32 4>
827  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
828  ret void
829}
830
831define void @v_shuffle_v3i64_v3i64__5_5_5(ptr addrspace(1) inreg %ptr) {
832; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_5_5:
833; GFX900:       ; %bb.0:
834; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
835; GFX900-NEXT:    ;;#ASMSTART
836; GFX900-NEXT:    ; def v[0:5]
837; GFX900-NEXT:    ;;#ASMEND
838; GFX900-NEXT:    v_mov_b32_e32 v6, 0
839; GFX900-NEXT:    v_mov_b32_e32 v2, v4
840; GFX900-NEXT:    v_mov_b32_e32 v3, v5
841; GFX900-NEXT:    global_store_dwordx2 v6, v[4:5], s[16:17] offset:16
842; GFX900-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
843; GFX900-NEXT:    s_waitcnt vmcnt(0)
844; GFX900-NEXT:    s_setpc_b64 s[30:31]
845;
846; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_5_5:
847; GFX90A:       ; %bb.0:
848; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
849; GFX90A-NEXT:    ;;#ASMSTART
850; GFX90A-NEXT:    ; def v[0:5]
851; GFX90A-NEXT:    ;;#ASMEND
852; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
853; GFX90A-NEXT:    v_mov_b32_e32 v2, v4
854; GFX90A-NEXT:    v_mov_b32_e32 v3, v5
855; GFX90A-NEXT:    global_store_dwordx2 v6, v[4:5], s[16:17] offset:16
856; GFX90A-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
857; GFX90A-NEXT:    s_waitcnt vmcnt(0)
858; GFX90A-NEXT:    s_setpc_b64 s[30:31]
859;
860; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_5_5:
861; GFX940:       ; %bb.0:
862; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
863; GFX940-NEXT:    ;;#ASMSTART
864; GFX940-NEXT:    ; def v[0:5]
865; GFX940-NEXT:    ;;#ASMEND
866; GFX940-NEXT:    v_mov_b32_e32 v6, 0
867; GFX940-NEXT:    v_mov_b32_e32 v2, v4
868; GFX940-NEXT:    v_mov_b32_e32 v3, v5
869; GFX940-NEXT:    global_store_dwordx2 v6, v[4:5], s[0:1] offset:16 sc0 sc1
870; GFX940-NEXT:    global_store_dwordx4 v6, v[2:5], s[0:1] sc0 sc1
871; GFX940-NEXT:    s_waitcnt vmcnt(0)
872; GFX940-NEXT:    s_setpc_b64 s[30:31]
873  %vec0 = call <3 x i64> asm "; def $0", "=v"()
874  %vec1 = call <3 x i64> asm "; def $0", "=v"()
875  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 5, i32 5>
876  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
877  ret void
878}
879
880define void @v_shuffle_v3i64_v3i64__u_0_0(ptr addrspace(1) inreg %ptr) {
881; GFX900-LABEL: v_shuffle_v3i64_v3i64__u_0_0:
882; GFX900:       ; %bb.0:
883; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
884; GFX900-NEXT:    ;;#ASMSTART
885; GFX900-NEXT:    ; def v[0:5]
886; GFX900-NEXT:    ;;#ASMEND
887; GFX900-NEXT:    v_mov_b32_e32 v6, 0
888; GFX900-NEXT:    v_mov_b32_e32 v2, v0
889; GFX900-NEXT:    v_mov_b32_e32 v3, v1
890; GFX900-NEXT:    global_store_dwordx2 v6, v[0:1], s[16:17] offset:16
891; GFX900-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
892; GFX900-NEXT:    s_waitcnt vmcnt(0)
893; GFX900-NEXT:    s_setpc_b64 s[30:31]
894;
895; GFX90A-LABEL: v_shuffle_v3i64_v3i64__u_0_0:
896; GFX90A:       ; %bb.0:
897; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
898; GFX90A-NEXT:    ;;#ASMSTART
899; GFX90A-NEXT:    ; def v[0:5]
900; GFX90A-NEXT:    ;;#ASMEND
901; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
902; GFX90A-NEXT:    v_mov_b32_e32 v2, v0
903; GFX90A-NEXT:    v_mov_b32_e32 v3, v1
904; GFX90A-NEXT:    global_store_dwordx2 v6, v[0:1], s[16:17] offset:16
905; GFX90A-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
906; GFX90A-NEXT:    s_waitcnt vmcnt(0)
907; GFX90A-NEXT:    s_setpc_b64 s[30:31]
908;
909; GFX940-LABEL: v_shuffle_v3i64_v3i64__u_0_0:
910; GFX940:       ; %bb.0:
911; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
912; GFX940-NEXT:    ;;#ASMSTART
913; GFX940-NEXT:    ; def v[0:5]
914; GFX940-NEXT:    ;;#ASMEND
915; GFX940-NEXT:    v_mov_b32_e32 v6, 0
916; GFX940-NEXT:    v_mov_b32_e32 v2, v0
917; GFX940-NEXT:    v_mov_b32_e32 v3, v1
918; GFX940-NEXT:    global_store_dwordx2 v6, v[0:1], s[0:1] offset:16 sc0 sc1
919; GFX940-NEXT:    global_store_dwordx4 v6, v[0:3], s[0:1] sc0 sc1
920; GFX940-NEXT:    s_waitcnt vmcnt(0)
921; GFX940-NEXT:    s_setpc_b64 s[30:31]
922  %vec0 = call <3 x i64> asm "; def $0", "=v"()
923  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 poison, i32 0, i32 0>
924  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
925  ret void
926}
927
928define void @v_shuffle_v3i64_v3i64__0_0_0(ptr addrspace(1) inreg %ptr) {
929; GFX900-LABEL: v_shuffle_v3i64_v3i64__0_0_0:
930; GFX900:       ; %bb.0:
931; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
932; GFX900-NEXT:    ;;#ASMSTART
933; GFX900-NEXT:    ; def v[0:5]
934; GFX900-NEXT:    ;;#ASMEND
935; GFX900-NEXT:    v_mov_b32_e32 v6, 0
936; GFX900-NEXT:    v_mov_b32_e32 v2, v0
937; GFX900-NEXT:    v_mov_b32_e32 v3, v1
938; GFX900-NEXT:    global_store_dwordx2 v6, v[0:1], s[16:17] offset:16
939; GFX900-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
940; GFX900-NEXT:    s_waitcnt vmcnt(0)
941; GFX900-NEXT:    s_setpc_b64 s[30:31]
942;
943; GFX90A-LABEL: v_shuffle_v3i64_v3i64__0_0_0:
944; GFX90A:       ; %bb.0:
945; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
946; GFX90A-NEXT:    ;;#ASMSTART
947; GFX90A-NEXT:    ; def v[0:5]
948; GFX90A-NEXT:    ;;#ASMEND
949; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
950; GFX90A-NEXT:    v_mov_b32_e32 v2, v0
951; GFX90A-NEXT:    v_mov_b32_e32 v3, v1
952; GFX90A-NEXT:    global_store_dwordx2 v6, v[0:1], s[16:17] offset:16
953; GFX90A-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
954; GFX90A-NEXT:    s_waitcnt vmcnt(0)
955; GFX90A-NEXT:    s_setpc_b64 s[30:31]
956;
957; GFX940-LABEL: v_shuffle_v3i64_v3i64__0_0_0:
958; GFX940:       ; %bb.0:
959; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
960; GFX940-NEXT:    ;;#ASMSTART
961; GFX940-NEXT:    ; def v[0:5]
962; GFX940-NEXT:    ;;#ASMEND
963; GFX940-NEXT:    v_mov_b32_e32 v6, 0
964; GFX940-NEXT:    v_mov_b32_e32 v2, v0
965; GFX940-NEXT:    v_mov_b32_e32 v3, v1
966; GFX940-NEXT:    global_store_dwordx2 v6, v[0:1], s[0:1] offset:16 sc0 sc1
967; GFX940-NEXT:    global_store_dwordx4 v6, v[0:3], s[0:1] sc0 sc1
968; GFX940-NEXT:    s_waitcnt vmcnt(0)
969; GFX940-NEXT:    s_setpc_b64 s[30:31]
970  %vec0 = call <3 x i64> asm "; def $0", "=v"()
971  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> zeroinitializer
972  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
973  ret void
974}
975
976define void @v_shuffle_v3i64_v3i64__1_0_0(ptr addrspace(1) inreg %ptr) {
977; GFX900-LABEL: v_shuffle_v3i64_v3i64__1_0_0:
978; GFX900:       ; %bb.0:
979; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
980; GFX900-NEXT:    ;;#ASMSTART
981; GFX900-NEXT:    ; def v[0:5]
982; GFX900-NEXT:    ;;#ASMEND
983; GFX900-NEXT:    v_mov_b32_e32 v6, 0
984; GFX900-NEXT:    v_mov_b32_e32 v4, v0
985; GFX900-NEXT:    v_mov_b32_e32 v5, v1
986; GFX900-NEXT:    global_store_dwordx2 v6, v[0:1], s[16:17] offset:16
987; GFX900-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
988; GFX900-NEXT:    s_waitcnt vmcnt(0)
989; GFX900-NEXT:    s_setpc_b64 s[30:31]
990;
991; GFX90A-LABEL: v_shuffle_v3i64_v3i64__1_0_0:
992; GFX90A:       ; %bb.0:
993; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
994; GFX90A-NEXT:    ;;#ASMSTART
995; GFX90A-NEXT:    ; def v[0:5]
996; GFX90A-NEXT:    ;;#ASMEND
997; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
998; GFX90A-NEXT:    v_mov_b32_e32 v4, v0
999; GFX90A-NEXT:    v_mov_b32_e32 v5, v1
1000; GFX90A-NEXT:    global_store_dwordx2 v6, v[0:1], s[16:17] offset:16
1001; GFX90A-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
1002; GFX90A-NEXT:    s_waitcnt vmcnt(0)
1003; GFX90A-NEXT:    s_setpc_b64 s[30:31]
1004;
1005; GFX940-LABEL: v_shuffle_v3i64_v3i64__1_0_0:
1006; GFX940:       ; %bb.0:
1007; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1008; GFX940-NEXT:    ;;#ASMSTART
1009; GFX940-NEXT:    ; def v[0:5]
1010; GFX940-NEXT:    ;;#ASMEND
1011; GFX940-NEXT:    v_mov_b32_e32 v6, 0
1012; GFX940-NEXT:    v_mov_b32_e32 v4, v0
1013; GFX940-NEXT:    v_mov_b32_e32 v5, v1
1014; GFX940-NEXT:    global_store_dwordx2 v6, v[0:1], s[0:1] offset:16 sc0 sc1
1015; GFX940-NEXT:    global_store_dwordx4 v6, v[2:5], s[0:1] sc0 sc1
1016; GFX940-NEXT:    s_waitcnt vmcnt(0)
1017; GFX940-NEXT:    s_setpc_b64 s[30:31]
1018  %vec0 = call <3 x i64> asm "; def $0", "=v"()
1019  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 1, i32 0, i32 0>
1020  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
1021  ret void
1022}
1023
1024define void @v_shuffle_v3i64_v3i64__2_0_0(ptr addrspace(1) inreg %ptr) {
1025; GFX900-LABEL: v_shuffle_v3i64_v3i64__2_0_0:
1026; GFX900:       ; %bb.0:
1027; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1028; GFX900-NEXT:    ;;#ASMSTART
1029; GFX900-NEXT:    ; def v[0:5]
1030; GFX900-NEXT:    ;;#ASMEND
1031; GFX900-NEXT:    v_mov_b32_e32 v6, 0
1032; GFX900-NEXT:    v_mov_b32_e32 v2, v4
1033; GFX900-NEXT:    v_mov_b32_e32 v3, v5
1034; GFX900-NEXT:    v_mov_b32_e32 v4, v0
1035; GFX900-NEXT:    v_mov_b32_e32 v5, v1
1036; GFX900-NEXT:    global_store_dwordx2 v6, v[0:1], s[16:17] offset:16
1037; GFX900-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
1038; GFX900-NEXT:    s_waitcnt vmcnt(0)
1039; GFX900-NEXT:    s_setpc_b64 s[30:31]
1040;
1041; GFX90A-LABEL: v_shuffle_v3i64_v3i64__2_0_0:
1042; GFX90A:       ; %bb.0:
1043; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1044; GFX90A-NEXT:    ;;#ASMSTART
1045; GFX90A-NEXT:    ; def v[0:5]
1046; GFX90A-NEXT:    ;;#ASMEND
1047; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
1048; GFX90A-NEXT:    v_mov_b32_e32 v2, v4
1049; GFX90A-NEXT:    v_mov_b32_e32 v3, v5
1050; GFX90A-NEXT:    v_mov_b32_e32 v4, v0
1051; GFX90A-NEXT:    v_mov_b32_e32 v5, v1
1052; GFX90A-NEXT:    global_store_dwordx2 v6, v[0:1], s[16:17] offset:16
1053; GFX90A-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
1054; GFX90A-NEXT:    s_waitcnt vmcnt(0)
1055; GFX90A-NEXT:    s_setpc_b64 s[30:31]
1056;
1057; GFX940-LABEL: v_shuffle_v3i64_v3i64__2_0_0:
1058; GFX940:       ; %bb.0:
1059; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1060; GFX940-NEXT:    ;;#ASMSTART
1061; GFX940-NEXT:    ; def v[0:5]
1062; GFX940-NEXT:    ;;#ASMEND
1063; GFX940-NEXT:    v_mov_b32_e32 v6, 0
1064; GFX940-NEXT:    v_mov_b32_e32 v2, v4
1065; GFX940-NEXT:    v_mov_b32_e32 v3, v5
1066; GFX940-NEXT:    v_mov_b32_e32 v4, v0
1067; GFX940-NEXT:    v_mov_b32_e32 v5, v1
1068; GFX940-NEXT:    global_store_dwordx2 v6, v[0:1], s[0:1] offset:16 sc0 sc1
1069; GFX940-NEXT:    global_store_dwordx4 v6, v[2:5], s[0:1] sc0 sc1
1070; GFX940-NEXT:    s_waitcnt vmcnt(0)
1071; GFX940-NEXT:    s_setpc_b64 s[30:31]
1072  %vec0 = call <3 x i64> asm "; def $0", "=v"()
1073  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 2, i32 0, i32 0>
1074  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
1075  ret void
1076}
1077
1078define void @v_shuffle_v3i64_v3i64__3_0_0(ptr addrspace(1) inreg %ptr) {
1079; GFX900-LABEL: v_shuffle_v3i64_v3i64__3_0_0:
1080; GFX900:       ; %bb.0:
1081; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1082; GFX900-NEXT:    ;;#ASMSTART
1083; GFX900-NEXT:    ; def v[0:5]
1084; GFX900-NEXT:    ;;#ASMEND
1085; GFX900-NEXT:    v_mov_b32_e32 v6, 0
1086; GFX900-NEXT:    v_mov_b32_e32 v2, v0
1087; GFX900-NEXT:    v_mov_b32_e32 v3, v1
1088; GFX900-NEXT:    global_store_dwordx2 v6, v[0:1], s[16:17] offset:16
1089; GFX900-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
1090; GFX900-NEXT:    s_waitcnt vmcnt(0)
1091; GFX900-NEXT:    s_setpc_b64 s[30:31]
1092;
1093; GFX90A-LABEL: v_shuffle_v3i64_v3i64__3_0_0:
1094; GFX90A:       ; %bb.0:
1095; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1096; GFX90A-NEXT:    ;;#ASMSTART
1097; GFX90A-NEXT:    ; def v[0:5]
1098; GFX90A-NEXT:    ;;#ASMEND
1099; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
1100; GFX90A-NEXT:    v_mov_b32_e32 v2, v0
1101; GFX90A-NEXT:    v_mov_b32_e32 v3, v1
1102; GFX90A-NEXT:    global_store_dwordx2 v6, v[0:1], s[16:17] offset:16
1103; GFX90A-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
1104; GFX90A-NEXT:    s_waitcnt vmcnt(0)
1105; GFX90A-NEXT:    s_setpc_b64 s[30:31]
1106;
1107; GFX940-LABEL: v_shuffle_v3i64_v3i64__3_0_0:
1108; GFX940:       ; %bb.0:
1109; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1110; GFX940-NEXT:    ;;#ASMSTART
1111; GFX940-NEXT:    ; def v[0:5]
1112; GFX940-NEXT:    ;;#ASMEND
1113; GFX940-NEXT:    v_mov_b32_e32 v6, 0
1114; GFX940-NEXT:    v_mov_b32_e32 v2, v0
1115; GFX940-NEXT:    v_mov_b32_e32 v3, v1
1116; GFX940-NEXT:    global_store_dwordx2 v6, v[0:1], s[0:1] offset:16 sc0 sc1
1117; GFX940-NEXT:    global_store_dwordx4 v6, v[0:3], s[0:1] sc0 sc1
1118; GFX940-NEXT:    s_waitcnt vmcnt(0)
1119; GFX940-NEXT:    s_setpc_b64 s[30:31]
1120  %vec0 = call <3 x i64> asm "; def $0", "=v"()
1121  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 3, i32 0, i32 0>
1122  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
1123  ret void
1124}
1125
1126define void @v_shuffle_v3i64_v3i64__4_0_0(ptr addrspace(1) inreg %ptr) {
1127; GFX900-LABEL: v_shuffle_v3i64_v3i64__4_0_0:
1128; GFX900:       ; %bb.0:
1129; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1130; GFX900-NEXT:    ;;#ASMSTART
1131; GFX900-NEXT:    ; def v[0:5]
1132; GFX900-NEXT:    ;;#ASMEND
1133; GFX900-NEXT:    ;;#ASMSTART
1134; GFX900-NEXT:    ; def v[2:7]
1135; GFX900-NEXT:    ;;#ASMEND
1136; GFX900-NEXT:    v_mov_b32_e32 v8, 0
1137; GFX900-NEXT:    v_mov_b32_e32 v6, v0
1138; GFX900-NEXT:    v_mov_b32_e32 v7, v1
1139; GFX900-NEXT:    global_store_dwordx2 v8, v[0:1], s[16:17] offset:16
1140; GFX900-NEXT:    global_store_dwordx4 v8, v[4:7], s[16:17]
1141; GFX900-NEXT:    s_waitcnt vmcnt(0)
1142; GFX900-NEXT:    s_setpc_b64 s[30:31]
1143;
1144; GFX90A-LABEL: v_shuffle_v3i64_v3i64__4_0_0:
1145; GFX90A:       ; %bb.0:
1146; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1147; GFX90A-NEXT:    ;;#ASMSTART
1148; GFX90A-NEXT:    ; def v[0:5]
1149; GFX90A-NEXT:    ;;#ASMEND
1150; GFX90A-NEXT:    ;;#ASMSTART
1151; GFX90A-NEXT:    ; def v[2:7]
1152; GFX90A-NEXT:    ;;#ASMEND
1153; GFX90A-NEXT:    v_mov_b32_e32 v8, 0
1154; GFX90A-NEXT:    v_mov_b32_e32 v6, v0
1155; GFX90A-NEXT:    v_mov_b32_e32 v7, v1
1156; GFX90A-NEXT:    global_store_dwordx2 v8, v[0:1], s[16:17] offset:16
1157; GFX90A-NEXT:    global_store_dwordx4 v8, v[4:7], s[16:17]
1158; GFX90A-NEXT:    s_waitcnt vmcnt(0)
1159; GFX90A-NEXT:    s_setpc_b64 s[30:31]
1160;
1161; GFX940-LABEL: v_shuffle_v3i64_v3i64__4_0_0:
1162; GFX940:       ; %bb.0:
1163; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1164; GFX940-NEXT:    ;;#ASMSTART
1165; GFX940-NEXT:    ; def v[0:5]
1166; GFX940-NEXT:    ;;#ASMEND
1167; GFX940-NEXT:    v_mov_b32_e32 v8, 0
1168; GFX940-NEXT:    ;;#ASMSTART
1169; GFX940-NEXT:    ; def v[2:7]
1170; GFX940-NEXT:    ;;#ASMEND
1171; GFX940-NEXT:    global_store_dwordx2 v8, v[0:1], s[0:1] offset:16 sc0 sc1
1172; GFX940-NEXT:    v_mov_b32_e32 v6, v0
1173; GFX940-NEXT:    v_mov_b32_e32 v7, v1
1174; GFX940-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] sc0 sc1
1175; GFX940-NEXT:    s_waitcnt vmcnt(0)
1176; GFX940-NEXT:    s_setpc_b64 s[30:31]
1177  %vec0 = call <3 x i64> asm "; def $0", "=v"()
1178  %vec1 = call <3 x i64> asm "; def $0", "=v"()
1179  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 4, i32 0, i32 0>
1180  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
1181  ret void
1182}
1183
1184define void @v_shuffle_v3i64_v3i64__5_0_0(ptr addrspace(1) inreg %ptr) {
1185; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_0_0:
1186; GFX900:       ; %bb.0:
1187; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1188; GFX900-NEXT:    ;;#ASMSTART
1189; GFX900-NEXT:    ; def v[0:5]
1190; GFX900-NEXT:    ;;#ASMEND
1191; GFX900-NEXT:    ;;#ASMSTART
1192; GFX900-NEXT:    ; def v[2:7]
1193; GFX900-NEXT:    ;;#ASMEND
1194; GFX900-NEXT:    v_mov_b32_e32 v8, 0
1195; GFX900-NEXT:    v_mov_b32_e32 v2, v6
1196; GFX900-NEXT:    v_mov_b32_e32 v3, v7
1197; GFX900-NEXT:    v_mov_b32_e32 v4, v0
1198; GFX900-NEXT:    v_mov_b32_e32 v5, v1
1199; GFX900-NEXT:    global_store_dwordx2 v8, v[0:1], s[16:17] offset:16
1200; GFX900-NEXT:    global_store_dwordx4 v8, v[2:5], s[16:17]
1201; GFX900-NEXT:    s_waitcnt vmcnt(0)
1202; GFX900-NEXT:    s_setpc_b64 s[30:31]
1203;
1204; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_0_0:
1205; GFX90A:       ; %bb.0:
1206; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1207; GFX90A-NEXT:    ;;#ASMSTART
1208; GFX90A-NEXT:    ; def v[0:5]
1209; GFX90A-NEXT:    ;;#ASMEND
1210; GFX90A-NEXT:    ;;#ASMSTART
1211; GFX90A-NEXT:    ; def v[2:7]
1212; GFX90A-NEXT:    ;;#ASMEND
1213; GFX90A-NEXT:    v_mov_b32_e32 v8, 0
1214; GFX90A-NEXT:    v_mov_b32_e32 v2, v6
1215; GFX90A-NEXT:    v_mov_b32_e32 v3, v7
1216; GFX90A-NEXT:    v_mov_b32_e32 v4, v0
1217; GFX90A-NEXT:    v_mov_b32_e32 v5, v1
1218; GFX90A-NEXT:    global_store_dwordx2 v8, v[0:1], s[16:17] offset:16
1219; GFX90A-NEXT:    global_store_dwordx4 v8, v[2:5], s[16:17]
1220; GFX90A-NEXT:    s_waitcnt vmcnt(0)
1221; GFX90A-NEXT:    s_setpc_b64 s[30:31]
1222;
1223; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_0_0:
1224; GFX940:       ; %bb.0:
1225; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1226; GFX940-NEXT:    ;;#ASMSTART
1227; GFX940-NEXT:    ; def v[0:5]
1228; GFX940-NEXT:    ;;#ASMEND
1229; GFX940-NEXT:    v_mov_b32_e32 v8, 0
1230; GFX940-NEXT:    ;;#ASMSTART
1231; GFX940-NEXT:    ; def v[2:7]
1232; GFX940-NEXT:    ;;#ASMEND
1233; GFX940-NEXT:    global_store_dwordx2 v8, v[0:1], s[0:1] offset:16 sc0 sc1
1234; GFX940-NEXT:    v_mov_b32_e32 v2, v6
1235; GFX940-NEXT:    v_mov_b32_e32 v3, v7
1236; GFX940-NEXT:    v_mov_b32_e32 v4, v0
1237; GFX940-NEXT:    v_mov_b32_e32 v5, v1
1238; GFX940-NEXT:    global_store_dwordx4 v8, v[2:5], s[0:1] sc0 sc1
1239; GFX940-NEXT:    s_waitcnt vmcnt(0)
1240; GFX940-NEXT:    s_setpc_b64 s[30:31]
1241  %vec0 = call <3 x i64> asm "; def $0", "=v"()
1242  %vec1 = call <3 x i64> asm "; def $0", "=v"()
1243  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 0, i32 0>
1244  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
1245  ret void
1246}
1247
1248define void @v_shuffle_v3i64_v3i64__5_u_0(ptr addrspace(1) inreg %ptr) {
1249; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_u_0:
1250; GFX900:       ; %bb.0:
1251; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1252; GFX900-NEXT:    v_mov_b32_e32 v8, 0
1253; GFX900-NEXT:    ;;#ASMSTART
1254; GFX900-NEXT:    ; def v[0:5]
1255; GFX900-NEXT:    ;;#ASMEND
1256; GFX900-NEXT:    ;;#ASMSTART
1257; GFX900-NEXT:    ; def v[2:7]
1258; GFX900-NEXT:    ;;#ASMEND
1259; GFX900-NEXT:    global_store_dwordx2 v8, v[0:1], s[16:17] offset:16
1260; GFX900-NEXT:    v_mov_b32_e32 v0, v6
1261; GFX900-NEXT:    v_mov_b32_e32 v1, v7
1262; GFX900-NEXT:    global_store_dwordx4 v8, v[0:3], s[16:17]
1263; GFX900-NEXT:    s_waitcnt vmcnt(0)
1264; GFX900-NEXT:    s_setpc_b64 s[30:31]
1265;
1266; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_u_0:
1267; GFX90A:       ; %bb.0:
1268; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1269; GFX90A-NEXT:    v_mov_b32_e32 v8, 0
1270; GFX90A-NEXT:    ;;#ASMSTART
1271; GFX90A-NEXT:    ; def v[0:5]
1272; GFX90A-NEXT:    ;;#ASMEND
1273; GFX90A-NEXT:    ;;#ASMSTART
1274; GFX90A-NEXT:    ; def v[2:7]
1275; GFX90A-NEXT:    ;;#ASMEND
1276; GFX90A-NEXT:    global_store_dwordx2 v8, v[0:1], s[16:17] offset:16
1277; GFX90A-NEXT:    v_mov_b32_e32 v0, v6
1278; GFX90A-NEXT:    v_mov_b32_e32 v1, v7
1279; GFX90A-NEXT:    global_store_dwordx4 v8, v[0:3], s[16:17]
1280; GFX90A-NEXT:    s_waitcnt vmcnt(0)
1281; GFX90A-NEXT:    s_setpc_b64 s[30:31]
1282;
1283; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_u_0:
1284; GFX940:       ; %bb.0:
1285; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1286; GFX940-NEXT:    v_mov_b32_e32 v8, 0
1287; GFX940-NEXT:    ;;#ASMSTART
1288; GFX940-NEXT:    ; def v[0:5]
1289; GFX940-NEXT:    ;;#ASMEND
1290; GFX940-NEXT:    global_store_dwordx2 v8, v[0:1], s[0:1] offset:16 sc0 sc1
1291; GFX940-NEXT:    ;;#ASMSTART
1292; GFX940-NEXT:    ; def v[2:7]
1293; GFX940-NEXT:    ;;#ASMEND
1294; GFX940-NEXT:    s_nop 0
1295; GFX940-NEXT:    v_mov_b32_e32 v0, v6
1296; GFX940-NEXT:    v_mov_b32_e32 v1, v7
1297; GFX940-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] sc0 sc1
1298; GFX940-NEXT:    s_waitcnt vmcnt(0)
1299; GFX940-NEXT:    s_setpc_b64 s[30:31]
1300  %vec0 = call <3 x i64> asm "; def $0", "=v"()
1301  %vec1 = call <3 x i64> asm "; def $0", "=v"()
1302  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 poison, i32 0>
1303  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
1304  ret void
1305}
1306
1307define void @v_shuffle_v3i64_v3i64__5_1_0(ptr addrspace(1) inreg %ptr) {
1308; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_1_0:
1309; GFX900:       ; %bb.0:
1310; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1311; GFX900-NEXT:    v_mov_b32_e32 v10, 0
1312; GFX900-NEXT:    ;;#ASMSTART
1313; GFX900-NEXT:    ; def v[0:5]
1314; GFX900-NEXT:    ;;#ASMEND
1315; GFX900-NEXT:    ;;#ASMSTART
1316; GFX900-NEXT:    ; def v[4:9]
1317; GFX900-NEXT:    ;;#ASMEND
1318; GFX900-NEXT:    global_store_dwordx2 v10, v[0:1], s[16:17] offset:16
1319; GFX900-NEXT:    v_mov_b32_e32 v0, v8
1320; GFX900-NEXT:    v_mov_b32_e32 v1, v9
1321; GFX900-NEXT:    global_store_dwordx4 v10, v[0:3], s[16:17]
1322; GFX900-NEXT:    s_waitcnt vmcnt(0)
1323; GFX900-NEXT:    s_setpc_b64 s[30:31]
1324;
1325; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_1_0:
1326; GFX90A:       ; %bb.0:
1327; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1328; GFX90A-NEXT:    v_mov_b32_e32 v10, 0
1329; GFX90A-NEXT:    ;;#ASMSTART
1330; GFX90A-NEXT:    ; def v[0:5]
1331; GFX90A-NEXT:    ;;#ASMEND
1332; GFX90A-NEXT:    ;;#ASMSTART
1333; GFX90A-NEXT:    ; def v[4:9]
1334; GFX90A-NEXT:    ;;#ASMEND
1335; GFX90A-NEXT:    global_store_dwordx2 v10, v[0:1], s[16:17] offset:16
1336; GFX90A-NEXT:    v_mov_b32_e32 v0, v8
1337; GFX90A-NEXT:    v_mov_b32_e32 v1, v9
1338; GFX90A-NEXT:    global_store_dwordx4 v10, v[0:3], s[16:17]
1339; GFX90A-NEXT:    s_waitcnt vmcnt(0)
1340; GFX90A-NEXT:    s_setpc_b64 s[30:31]
1341;
1342; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_1_0:
1343; GFX940:       ; %bb.0:
1344; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1345; GFX940-NEXT:    v_mov_b32_e32 v10, 0
1346; GFX940-NEXT:    ;;#ASMSTART
1347; GFX940-NEXT:    ; def v[0:5]
1348; GFX940-NEXT:    ;;#ASMEND
1349; GFX940-NEXT:    global_store_dwordx2 v10, v[0:1], s[0:1] offset:16 sc0 sc1
1350; GFX940-NEXT:    ;;#ASMSTART
1351; GFX940-NEXT:    ; def v[4:9]
1352; GFX940-NEXT:    ;;#ASMEND
1353; GFX940-NEXT:    s_nop 0
1354; GFX940-NEXT:    v_mov_b32_e32 v0, v8
1355; GFX940-NEXT:    v_mov_b32_e32 v1, v9
1356; GFX940-NEXT:    global_store_dwordx4 v10, v[0:3], s[0:1] sc0 sc1
1357; GFX940-NEXT:    s_waitcnt vmcnt(0)
1358; GFX940-NEXT:    s_setpc_b64 s[30:31]
1359  %vec0 = call <3 x i64> asm "; def $0", "=v"()
1360  %vec1 = call <3 x i64> asm "; def $0", "=v"()
1361  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 1, i32 0>
1362  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
1363  ret void
1364}
1365
1366define void @v_shuffle_v3i64_v3i64__5_2_0(ptr addrspace(1) inreg %ptr) {
1367; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_2_0:
1368; GFX900:       ; %bb.0:
1369; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1370; GFX900-NEXT:    ;;#ASMSTART
1371; GFX900-NEXT:    ; def v[0:5]
1372; GFX900-NEXT:    ;;#ASMEND
1373; GFX900-NEXT:    v_mov_b32_e32 v12, 0
1374; GFX900-NEXT:    ;;#ASMSTART
1375; GFX900-NEXT:    ; def v[6:11]
1376; GFX900-NEXT:    ;;#ASMEND
1377; GFX900-NEXT:    v_mov_b32_e32 v2, v10
1378; GFX900-NEXT:    v_mov_b32_e32 v3, v11
1379; GFX900-NEXT:    global_store_dwordx2 v12, v[0:1], s[16:17] offset:16
1380; GFX900-NEXT:    global_store_dwordx4 v12, v[2:5], s[16:17]
1381; GFX900-NEXT:    s_waitcnt vmcnt(0)
1382; GFX900-NEXT:    s_setpc_b64 s[30:31]
1383;
1384; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_2_0:
1385; GFX90A:       ; %bb.0:
1386; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1387; GFX90A-NEXT:    ;;#ASMSTART
1388; GFX90A-NEXT:    ; def v[0:5]
1389; GFX90A-NEXT:    ;;#ASMEND
1390; GFX90A-NEXT:    v_mov_b32_e32 v12, 0
1391; GFX90A-NEXT:    ;;#ASMSTART
1392; GFX90A-NEXT:    ; def v[6:11]
1393; GFX90A-NEXT:    ;;#ASMEND
1394; GFX90A-NEXT:    v_mov_b32_e32 v2, v10
1395; GFX90A-NEXT:    v_mov_b32_e32 v3, v11
1396; GFX90A-NEXT:    global_store_dwordx2 v12, v[0:1], s[16:17] offset:16
1397; GFX90A-NEXT:    global_store_dwordx4 v12, v[2:5], s[16:17]
1398; GFX90A-NEXT:    s_waitcnt vmcnt(0)
1399; GFX90A-NEXT:    s_setpc_b64 s[30:31]
1400;
1401; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_2_0:
1402; GFX940:       ; %bb.0:
1403; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1404; GFX940-NEXT:    ;;#ASMSTART
1405; GFX940-NEXT:    ; def v[0:5]
1406; GFX940-NEXT:    ;;#ASMEND
1407; GFX940-NEXT:    v_mov_b32_e32 v12, 0
1408; GFX940-NEXT:    ;;#ASMSTART
1409; GFX940-NEXT:    ; def v[6:11]
1410; GFX940-NEXT:    ;;#ASMEND
1411; GFX940-NEXT:    global_store_dwordx2 v12, v[0:1], s[0:1] offset:16 sc0 sc1
1412; GFX940-NEXT:    v_mov_b32_e32 v2, v10
1413; GFX940-NEXT:    v_mov_b32_e32 v3, v11
1414; GFX940-NEXT:    global_store_dwordx4 v12, v[2:5], s[0:1] sc0 sc1
1415; GFX940-NEXT:    s_waitcnt vmcnt(0)
1416; GFX940-NEXT:    s_setpc_b64 s[30:31]
1417  %vec0 = call <3 x i64> asm "; def $0", "=v"()
1418  %vec1 = call <3 x i64> asm "; def $0", "=v"()
1419  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 2, i32 0>
1420  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
1421  ret void
1422}
1423
1424define void @v_shuffle_v3i64_v3i64__5_3_0(ptr addrspace(1) inreg %ptr) {
1425; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_3_0:
1426; GFX900:       ; %bb.0:
1427; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1428; GFX900-NEXT:    v_mov_b32_e32 v8, 0
1429; GFX900-NEXT:    ;;#ASMSTART
1430; GFX900-NEXT:    ; def v[0:5]
1431; GFX900-NEXT:    ;;#ASMEND
1432; GFX900-NEXT:    ;;#ASMSTART
1433; GFX900-NEXT:    ; def v[2:7]
1434; GFX900-NEXT:    ;;#ASMEND
1435; GFX900-NEXT:    global_store_dwordx2 v8, v[0:1], s[16:17] offset:16
1436; GFX900-NEXT:    v_mov_b32_e32 v0, v6
1437; GFX900-NEXT:    v_mov_b32_e32 v1, v7
1438; GFX900-NEXT:    global_store_dwordx4 v8, v[0:3], s[16:17]
1439; GFX900-NEXT:    s_waitcnt vmcnt(0)
1440; GFX900-NEXT:    s_setpc_b64 s[30:31]
1441;
1442; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_3_0:
1443; GFX90A:       ; %bb.0:
1444; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1445; GFX90A-NEXT:    v_mov_b32_e32 v8, 0
1446; GFX90A-NEXT:    ;;#ASMSTART
1447; GFX90A-NEXT:    ; def v[0:5]
1448; GFX90A-NEXT:    ;;#ASMEND
1449; GFX90A-NEXT:    ;;#ASMSTART
1450; GFX90A-NEXT:    ; def v[2:7]
1451; GFX90A-NEXT:    ;;#ASMEND
1452; GFX90A-NEXT:    global_store_dwordx2 v8, v[0:1], s[16:17] offset:16
1453; GFX90A-NEXT:    v_mov_b32_e32 v0, v6
1454; GFX90A-NEXT:    v_mov_b32_e32 v1, v7
1455; GFX90A-NEXT:    global_store_dwordx4 v8, v[0:3], s[16:17]
1456; GFX90A-NEXT:    s_waitcnt vmcnt(0)
1457; GFX90A-NEXT:    s_setpc_b64 s[30:31]
1458;
1459; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_3_0:
1460; GFX940:       ; %bb.0:
1461; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1462; GFX940-NEXT:    v_mov_b32_e32 v8, 0
1463; GFX940-NEXT:    ;;#ASMSTART
1464; GFX940-NEXT:    ; def v[0:5]
1465; GFX940-NEXT:    ;;#ASMEND
1466; GFX940-NEXT:    global_store_dwordx2 v8, v[0:1], s[0:1] offset:16 sc0 sc1
1467; GFX940-NEXT:    ;;#ASMSTART
1468; GFX940-NEXT:    ; def v[2:7]
1469; GFX940-NEXT:    ;;#ASMEND
1470; GFX940-NEXT:    s_nop 0
1471; GFX940-NEXT:    v_mov_b32_e32 v0, v6
1472; GFX940-NEXT:    v_mov_b32_e32 v1, v7
1473; GFX940-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] sc0 sc1
1474; GFX940-NEXT:    s_waitcnt vmcnt(0)
1475; GFX940-NEXT:    s_setpc_b64 s[30:31]
1476  %vec0 = call <3 x i64> asm "; def $0", "=v"()
1477  %vec1 = call <3 x i64> asm "; def $0", "=v"()
1478  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 3, i32 0>
1479  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
1480  ret void
1481}
1482
1483define void @v_shuffle_v3i64_v3i64__5_4_0(ptr addrspace(1) inreg %ptr) {
1484; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_4_0:
1485; GFX900:       ; %bb.0:
1486; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1487; GFX900-NEXT:    ;;#ASMSTART
1488; GFX900-NEXT:    ; def v[0:5]
1489; GFX900-NEXT:    ;;#ASMEND
1490; GFX900-NEXT:    ;;#ASMSTART
1491; GFX900-NEXT:    ; def v[2:7]
1492; GFX900-NEXT:    ;;#ASMEND
1493; GFX900-NEXT:    v_mov_b32_e32 v8, 0
1494; GFX900-NEXT:    v_mov_b32_e32 v2, v6
1495; GFX900-NEXT:    v_mov_b32_e32 v3, v7
1496; GFX900-NEXT:    global_store_dwordx2 v8, v[0:1], s[16:17] offset:16
1497; GFX900-NEXT:    global_store_dwordx4 v8, v[2:5], s[16:17]
1498; GFX900-NEXT:    s_waitcnt vmcnt(0)
1499; GFX900-NEXT:    s_setpc_b64 s[30:31]
1500;
1501; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_4_0:
1502; GFX90A:       ; %bb.0:
1503; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1504; GFX90A-NEXT:    ;;#ASMSTART
1505; GFX90A-NEXT:    ; def v[0:5]
1506; GFX90A-NEXT:    ;;#ASMEND
1507; GFX90A-NEXT:    ;;#ASMSTART
1508; GFX90A-NEXT:    ; def v[2:7]
1509; GFX90A-NEXT:    ;;#ASMEND
1510; GFX90A-NEXT:    v_mov_b32_e32 v8, 0
1511; GFX90A-NEXT:    v_mov_b32_e32 v2, v6
1512; GFX90A-NEXT:    v_mov_b32_e32 v3, v7
1513; GFX90A-NEXT:    global_store_dwordx2 v8, v[0:1], s[16:17] offset:16
1514; GFX90A-NEXT:    global_store_dwordx4 v8, v[2:5], s[16:17]
1515; GFX90A-NEXT:    s_waitcnt vmcnt(0)
1516; GFX90A-NEXT:    s_setpc_b64 s[30:31]
1517;
1518; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_4_0:
1519; GFX940:       ; %bb.0:
1520; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1521; GFX940-NEXT:    ;;#ASMSTART
1522; GFX940-NEXT:    ; def v[0:5]
1523; GFX940-NEXT:    ;;#ASMEND
1524; GFX940-NEXT:    v_mov_b32_e32 v8, 0
1525; GFX940-NEXT:    ;;#ASMSTART
1526; GFX940-NEXT:    ; def v[2:7]
1527; GFX940-NEXT:    ;;#ASMEND
1528; GFX940-NEXT:    global_store_dwordx2 v8, v[0:1], s[0:1] offset:16 sc0 sc1
1529; GFX940-NEXT:    v_mov_b32_e32 v2, v6
1530; GFX940-NEXT:    v_mov_b32_e32 v3, v7
1531; GFX940-NEXT:    global_store_dwordx4 v8, v[2:5], s[0:1] sc0 sc1
1532; GFX940-NEXT:    s_waitcnt vmcnt(0)
1533; GFX940-NEXT:    s_setpc_b64 s[30:31]
1534  %vec0 = call <3 x i64> asm "; def $0", "=v"()
1535  %vec1 = call <3 x i64> asm "; def $0", "=v"()
1536  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 4, i32 0>
1537  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
1538  ret void
1539}
1540
1541define void @v_shuffle_v3i64_v3i64__u_1_1(ptr addrspace(1) inreg %ptr) {
1542; GFX900-LABEL: v_shuffle_v3i64_v3i64__u_1_1:
1543; GFX900:       ; %bb.0:
1544; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1545; GFX900-NEXT:    v_mov_b32_e32 v6, 0
1546; GFX900-NEXT:    ;;#ASMSTART
1547; GFX900-NEXT:    ; def v[0:5]
1548; GFX900-NEXT:    ;;#ASMEND
1549; GFX900-NEXT:    global_store_dwordx2 v6, v[2:3], s[16:17] offset:16
1550; GFX900-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
1551; GFX900-NEXT:    s_waitcnt vmcnt(0)
1552; GFX900-NEXT:    s_setpc_b64 s[30:31]
1553;
1554; GFX90A-LABEL: v_shuffle_v3i64_v3i64__u_1_1:
1555; GFX90A:       ; %bb.0:
1556; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1557; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
1558; GFX90A-NEXT:    ;;#ASMSTART
1559; GFX90A-NEXT:    ; def v[0:5]
1560; GFX90A-NEXT:    ;;#ASMEND
1561; GFX90A-NEXT:    global_store_dwordx2 v6, v[2:3], s[16:17] offset:16
1562; GFX90A-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
1563; GFX90A-NEXT:    s_waitcnt vmcnt(0)
1564; GFX90A-NEXT:    s_setpc_b64 s[30:31]
1565;
1566; GFX940-LABEL: v_shuffle_v3i64_v3i64__u_1_1:
1567; GFX940:       ; %bb.0:
1568; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1569; GFX940-NEXT:    v_mov_b32_e32 v6, 0
1570; GFX940-NEXT:    ;;#ASMSTART
1571; GFX940-NEXT:    ; def v[0:5]
1572; GFX940-NEXT:    ;;#ASMEND
1573; GFX940-NEXT:    global_store_dwordx2 v6, v[2:3], s[0:1] offset:16 sc0 sc1
1574; GFX940-NEXT:    global_store_dwordx4 v6, v[0:3], s[0:1] sc0 sc1
1575; GFX940-NEXT:    s_waitcnt vmcnt(0)
1576; GFX940-NEXT:    s_setpc_b64 s[30:31]
1577  %vec0 = call <3 x i64> asm "; def $0", "=v"()
1578  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 poison, i32 1, i32 1>
1579  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
1580  ret void
1581}
1582
1583define void @v_shuffle_v3i64_v3i64__0_1_1(ptr addrspace(1) inreg %ptr) {
1584; GFX900-LABEL: v_shuffle_v3i64_v3i64__0_1_1:
1585; GFX900:       ; %bb.0:
1586; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1587; GFX900-NEXT:    v_mov_b32_e32 v6, 0
1588; GFX900-NEXT:    ;;#ASMSTART
1589; GFX900-NEXT:    ; def v[0:5]
1590; GFX900-NEXT:    ;;#ASMEND
1591; GFX900-NEXT:    global_store_dwordx2 v6, v[2:3], s[16:17] offset:16
1592; GFX900-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
1593; GFX900-NEXT:    s_waitcnt vmcnt(0)
1594; GFX900-NEXT:    s_setpc_b64 s[30:31]
1595;
1596; GFX90A-LABEL: v_shuffle_v3i64_v3i64__0_1_1:
1597; GFX90A:       ; %bb.0:
1598; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1599; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
1600; GFX90A-NEXT:    ;;#ASMSTART
1601; GFX90A-NEXT:    ; def v[0:5]
1602; GFX90A-NEXT:    ;;#ASMEND
1603; GFX90A-NEXT:    global_store_dwordx2 v6, v[2:3], s[16:17] offset:16
1604; GFX90A-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
1605; GFX90A-NEXT:    s_waitcnt vmcnt(0)
1606; GFX90A-NEXT:    s_setpc_b64 s[30:31]
1607;
1608; GFX940-LABEL: v_shuffle_v3i64_v3i64__0_1_1:
1609; GFX940:       ; %bb.0:
1610; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1611; GFX940-NEXT:    v_mov_b32_e32 v6, 0
1612; GFX940-NEXT:    ;;#ASMSTART
1613; GFX940-NEXT:    ; def v[0:5]
1614; GFX940-NEXT:    ;;#ASMEND
1615; GFX940-NEXT:    global_store_dwordx2 v6, v[2:3], s[0:1] offset:16 sc0 sc1
1616; GFX940-NEXT:    global_store_dwordx4 v6, v[0:3], s[0:1] sc0 sc1
1617; GFX940-NEXT:    s_waitcnt vmcnt(0)
1618; GFX940-NEXT:    s_setpc_b64 s[30:31]
1619  %vec0 = call <3 x i64> asm "; def $0", "=v"()
1620  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 0, i32 1, i32 1>
1621  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
1622  ret void
1623}
1624
1625define void @v_shuffle_v3i64_v3i64__1_1_1(ptr addrspace(1) inreg %ptr) {
1626; GFX900-LABEL: v_shuffle_v3i64_v3i64__1_1_1:
1627; GFX900:       ; %bb.0:
1628; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1629; GFX900-NEXT:    ;;#ASMSTART
1630; GFX900-NEXT:    ; def v[0:5]
1631; GFX900-NEXT:    ;;#ASMEND
1632; GFX900-NEXT:    v_mov_b32_e32 v6, 0
1633; GFX900-NEXT:    v_mov_b32_e32 v4, v2
1634; GFX900-NEXT:    v_mov_b32_e32 v5, v3
1635; GFX900-NEXT:    global_store_dwordx2 v6, v[2:3], s[16:17] offset:16
1636; GFX900-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
1637; GFX900-NEXT:    s_waitcnt vmcnt(0)
1638; GFX900-NEXT:    s_setpc_b64 s[30:31]
1639;
1640; GFX90A-LABEL: v_shuffle_v3i64_v3i64__1_1_1:
1641; GFX90A:       ; %bb.0:
1642; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1643; GFX90A-NEXT:    ;;#ASMSTART
1644; GFX90A-NEXT:    ; def v[0:5]
1645; GFX90A-NEXT:    ;;#ASMEND
1646; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
1647; GFX90A-NEXT:    v_mov_b32_e32 v4, v2
1648; GFX90A-NEXT:    v_mov_b32_e32 v5, v3
1649; GFX90A-NEXT:    global_store_dwordx2 v6, v[2:3], s[16:17] offset:16
1650; GFX90A-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
1651; GFX90A-NEXT:    s_waitcnt vmcnt(0)
1652; GFX90A-NEXT:    s_setpc_b64 s[30:31]
1653;
1654; GFX940-LABEL: v_shuffle_v3i64_v3i64__1_1_1:
1655; GFX940:       ; %bb.0:
1656; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1657; GFX940-NEXT:    ;;#ASMSTART
1658; GFX940-NEXT:    ; def v[0:5]
1659; GFX940-NEXT:    ;;#ASMEND
1660; GFX940-NEXT:    v_mov_b32_e32 v6, 0
1661; GFX940-NEXT:    v_mov_b32_e32 v4, v2
1662; GFX940-NEXT:    v_mov_b32_e32 v5, v3
1663; GFX940-NEXT:    global_store_dwordx2 v6, v[2:3], s[0:1] offset:16 sc0 sc1
1664; GFX940-NEXT:    global_store_dwordx4 v6, v[2:5], s[0:1] sc0 sc1
1665; GFX940-NEXT:    s_waitcnt vmcnt(0)
1666; GFX940-NEXT:    s_setpc_b64 s[30:31]
1667  %vec0 = call <3 x i64> asm "; def $0", "=v"()
1668  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 1, i32 1, i32 1>
1669  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
1670  ret void
1671}
1672
1673define void @v_shuffle_v3i64_v3i64__2_1_1(ptr addrspace(1) inreg %ptr) {
1674; GFX900-LABEL: v_shuffle_v3i64_v3i64__2_1_1:
1675; GFX900:       ; %bb.0:
1676; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1677; GFX900-NEXT:    ;;#ASMSTART
1678; GFX900-NEXT:    ; def v[0:5]
1679; GFX900-NEXT:    ;;#ASMEND
1680; GFX900-NEXT:    v_mov_b32_e32 v6, 0
1681; GFX900-NEXT:    v_mov_b32_e32 v0, v4
1682; GFX900-NEXT:    v_mov_b32_e32 v1, v5
1683; GFX900-NEXT:    global_store_dwordx2 v6, v[2:3], s[16:17] offset:16
1684; GFX900-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
1685; GFX900-NEXT:    s_waitcnt vmcnt(0)
1686; GFX900-NEXT:    s_setpc_b64 s[30:31]
1687;
1688; GFX90A-LABEL: v_shuffle_v3i64_v3i64__2_1_1:
1689; GFX90A:       ; %bb.0:
1690; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1691; GFX90A-NEXT:    ;;#ASMSTART
1692; GFX90A-NEXT:    ; def v[0:5]
1693; GFX90A-NEXT:    ;;#ASMEND
1694; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
1695; GFX90A-NEXT:    v_mov_b32_e32 v0, v4
1696; GFX90A-NEXT:    v_mov_b32_e32 v1, v5
1697; GFX90A-NEXT:    global_store_dwordx2 v6, v[2:3], s[16:17] offset:16
1698; GFX90A-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
1699; GFX90A-NEXT:    s_waitcnt vmcnt(0)
1700; GFX90A-NEXT:    s_setpc_b64 s[30:31]
1701;
1702; GFX940-LABEL: v_shuffle_v3i64_v3i64__2_1_1:
1703; GFX940:       ; %bb.0:
1704; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1705; GFX940-NEXT:    ;;#ASMSTART
1706; GFX940-NEXT:    ; def v[0:5]
1707; GFX940-NEXT:    ;;#ASMEND
1708; GFX940-NEXT:    v_mov_b32_e32 v6, 0
1709; GFX940-NEXT:    v_mov_b32_e32 v0, v4
1710; GFX940-NEXT:    v_mov_b32_e32 v1, v5
1711; GFX940-NEXT:    global_store_dwordx2 v6, v[2:3], s[0:1] offset:16 sc0 sc1
1712; GFX940-NEXT:    global_store_dwordx4 v6, v[0:3], s[0:1] sc0 sc1
1713; GFX940-NEXT:    s_waitcnt vmcnt(0)
1714; GFX940-NEXT:    s_setpc_b64 s[30:31]
1715  %vec0 = call <3 x i64> asm "; def $0", "=v"()
1716  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 2, i32 1, i32 1>
1717  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
1718  ret void
1719}
1720
1721define void @v_shuffle_v3i64_v3i64__3_1_1(ptr addrspace(1) inreg %ptr) {
1722; GFX900-LABEL: v_shuffle_v3i64_v3i64__3_1_1:
1723; GFX900:       ; %bb.0:
1724; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1725; GFX900-NEXT:    v_mov_b32_e32 v6, 0
1726; GFX900-NEXT:    ;;#ASMSTART
1727; GFX900-NEXT:    ; def v[0:5]
1728; GFX900-NEXT:    ;;#ASMEND
1729; GFX900-NEXT:    global_store_dwordx2 v6, v[2:3], s[16:17] offset:16
1730; GFX900-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
1731; GFX900-NEXT:    s_waitcnt vmcnt(0)
1732; GFX900-NEXT:    s_setpc_b64 s[30:31]
1733;
1734; GFX90A-LABEL: v_shuffle_v3i64_v3i64__3_1_1:
1735; GFX90A:       ; %bb.0:
1736; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1737; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
1738; GFX90A-NEXT:    ;;#ASMSTART
1739; GFX90A-NEXT:    ; def v[0:5]
1740; GFX90A-NEXT:    ;;#ASMEND
1741; GFX90A-NEXT:    global_store_dwordx2 v6, v[2:3], s[16:17] offset:16
1742; GFX90A-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
1743; GFX90A-NEXT:    s_waitcnt vmcnt(0)
1744; GFX90A-NEXT:    s_setpc_b64 s[30:31]
1745;
1746; GFX940-LABEL: v_shuffle_v3i64_v3i64__3_1_1:
1747; GFX940:       ; %bb.0:
1748; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1749; GFX940-NEXT:    v_mov_b32_e32 v6, 0
1750; GFX940-NEXT:    ;;#ASMSTART
1751; GFX940-NEXT:    ; def v[0:5]
1752; GFX940-NEXT:    ;;#ASMEND
1753; GFX940-NEXT:    global_store_dwordx2 v6, v[2:3], s[0:1] offset:16 sc0 sc1
1754; GFX940-NEXT:    global_store_dwordx4 v6, v[0:3], s[0:1] sc0 sc1
1755; GFX940-NEXT:    s_waitcnt vmcnt(0)
1756; GFX940-NEXT:    s_setpc_b64 s[30:31]
1757  %vec0 = call <3 x i64> asm "; def $0", "=v"()
1758  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 3, i32 1, i32 1>
1759  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
1760  ret void
1761}
1762
1763define void @v_shuffle_v3i64_v3i64__4_1_1(ptr addrspace(1) inreg %ptr) {
1764; GFX900-LABEL: v_shuffle_v3i64_v3i64__4_1_1:
1765; GFX900:       ; %bb.0:
1766; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1767; GFX900-NEXT:    ;;#ASMSTART
1768; GFX900-NEXT:    ; def v[0:5]
1769; GFX900-NEXT:    ;;#ASMEND
1770; GFX900-NEXT:    ;;#ASMSTART
1771; GFX900-NEXT:    ; def v[4:9]
1772; GFX900-NEXT:    ;;#ASMEND
1773; GFX900-NEXT:    v_mov_b32_e32 v10, 0
1774; GFX900-NEXT:    v_mov_b32_e32 v8, v2
1775; GFX900-NEXT:    v_mov_b32_e32 v9, v3
1776; GFX900-NEXT:    global_store_dwordx2 v10, v[2:3], s[16:17] offset:16
1777; GFX900-NEXT:    global_store_dwordx4 v10, v[6:9], s[16:17]
1778; GFX900-NEXT:    s_waitcnt vmcnt(0)
1779; GFX900-NEXT:    s_setpc_b64 s[30:31]
1780;
1781; GFX90A-LABEL: v_shuffle_v3i64_v3i64__4_1_1:
1782; GFX90A:       ; %bb.0:
1783; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1784; GFX90A-NEXT:    ;;#ASMSTART
1785; GFX90A-NEXT:    ; def v[0:5]
1786; GFX90A-NEXT:    ;;#ASMEND
1787; GFX90A-NEXT:    ;;#ASMSTART
1788; GFX90A-NEXT:    ; def v[4:9]
1789; GFX90A-NEXT:    ;;#ASMEND
1790; GFX90A-NEXT:    v_mov_b32_e32 v10, 0
1791; GFX90A-NEXT:    v_mov_b32_e32 v8, v2
1792; GFX90A-NEXT:    v_mov_b32_e32 v9, v3
1793; GFX90A-NEXT:    global_store_dwordx2 v10, v[2:3], s[16:17] offset:16
1794; GFX90A-NEXT:    global_store_dwordx4 v10, v[6:9], s[16:17]
1795; GFX90A-NEXT:    s_waitcnt vmcnt(0)
1796; GFX90A-NEXT:    s_setpc_b64 s[30:31]
1797;
1798; GFX940-LABEL: v_shuffle_v3i64_v3i64__4_1_1:
1799; GFX940:       ; %bb.0:
1800; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1801; GFX940-NEXT:    ;;#ASMSTART
1802; GFX940-NEXT:    ; def v[0:5]
1803; GFX940-NEXT:    ;;#ASMEND
1804; GFX940-NEXT:    v_mov_b32_e32 v10, 0
1805; GFX940-NEXT:    ;;#ASMSTART
1806; GFX940-NEXT:    ; def v[4:9]
1807; GFX940-NEXT:    ;;#ASMEND
1808; GFX940-NEXT:    global_store_dwordx2 v10, v[2:3], s[0:1] offset:16 sc0 sc1
1809; GFX940-NEXT:    v_mov_b32_e32 v8, v2
1810; GFX940-NEXT:    v_mov_b32_e32 v9, v3
1811; GFX940-NEXT:    global_store_dwordx4 v10, v[6:9], s[0:1] sc0 sc1
1812; GFX940-NEXT:    s_waitcnt vmcnt(0)
1813; GFX940-NEXT:    s_setpc_b64 s[30:31]
1814  %vec0 = call <3 x i64> asm "; def $0", "=v"()
1815  %vec1 = call <3 x i64> asm "; def $0", "=v"()
1816  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 4, i32 1, i32 1>
1817  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
1818  ret void
1819}
1820
1821define void @v_shuffle_v3i64_v3i64__5_1_1(ptr addrspace(1) inreg %ptr) {
1822; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_1_1:
1823; GFX900:       ; %bb.0:
1824; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1825; GFX900-NEXT:    ;;#ASMSTART
1826; GFX900-NEXT:    ; def v[0:5]
1827; GFX900-NEXT:    ;;#ASMEND
1828; GFX900-NEXT:    v_mov_b32_e32 v10, 0
1829; GFX900-NEXT:    ;;#ASMSTART
1830; GFX900-NEXT:    ; def v[4:9]
1831; GFX900-NEXT:    ;;#ASMEND
1832; GFX900-NEXT:    v_mov_b32_e32 v0, v8
1833; GFX900-NEXT:    v_mov_b32_e32 v1, v9
1834; GFX900-NEXT:    global_store_dwordx2 v10, v[2:3], s[16:17] offset:16
1835; GFX900-NEXT:    global_store_dwordx4 v10, v[0:3], s[16:17]
1836; GFX900-NEXT:    s_waitcnt vmcnt(0)
1837; GFX900-NEXT:    s_setpc_b64 s[30:31]
1838;
1839; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_1_1:
1840; GFX90A:       ; %bb.0:
1841; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1842; GFX90A-NEXT:    ;;#ASMSTART
1843; GFX90A-NEXT:    ; def v[0:5]
1844; GFX90A-NEXT:    ;;#ASMEND
1845; GFX90A-NEXT:    v_mov_b32_e32 v10, 0
1846; GFX90A-NEXT:    ;;#ASMSTART
1847; GFX90A-NEXT:    ; def v[4:9]
1848; GFX90A-NEXT:    ;;#ASMEND
1849; GFX90A-NEXT:    v_mov_b32_e32 v0, v8
1850; GFX90A-NEXT:    v_mov_b32_e32 v1, v9
1851; GFX90A-NEXT:    global_store_dwordx2 v10, v[2:3], s[16:17] offset:16
1852; GFX90A-NEXT:    global_store_dwordx4 v10, v[0:3], s[16:17]
1853; GFX90A-NEXT:    s_waitcnt vmcnt(0)
1854; GFX90A-NEXT:    s_setpc_b64 s[30:31]
1855;
1856; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_1_1:
1857; GFX940:       ; %bb.0:
1858; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1859; GFX940-NEXT:    ;;#ASMSTART
1860; GFX940-NEXT:    ; def v[0:5]
1861; GFX940-NEXT:    ;;#ASMEND
1862; GFX940-NEXT:    v_mov_b32_e32 v10, 0
1863; GFX940-NEXT:    ;;#ASMSTART
1864; GFX940-NEXT:    ; def v[4:9]
1865; GFX940-NEXT:    ;;#ASMEND
1866; GFX940-NEXT:    global_store_dwordx2 v10, v[2:3], s[0:1] offset:16 sc0 sc1
1867; GFX940-NEXT:    v_mov_b32_e32 v0, v8
1868; GFX940-NEXT:    v_mov_b32_e32 v1, v9
1869; GFX940-NEXT:    global_store_dwordx4 v10, v[0:3], s[0:1] sc0 sc1
1870; GFX940-NEXT:    s_waitcnt vmcnt(0)
1871; GFX940-NEXT:    s_setpc_b64 s[30:31]
1872  %vec0 = call <3 x i64> asm "; def $0", "=v"()
1873  %vec1 = call <3 x i64> asm "; def $0", "=v"()
1874  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 1, i32 1>
1875  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
1876  ret void
1877}
1878
1879define void @v_shuffle_v3i64_v3i64__5_u_1(ptr addrspace(1) inreg %ptr) {
1880; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_u_1:
1881; GFX900:       ; %bb.0:
1882; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1883; GFX900-NEXT:    ;;#ASMSTART
1884; GFX900-NEXT:    ; def v[0:5]
1885; GFX900-NEXT:    ;;#ASMEND
1886; GFX900-NEXT:    v_mov_b32_e32 v10, 0
1887; GFX900-NEXT:    ;;#ASMSTART
1888; GFX900-NEXT:    ; def v[4:9]
1889; GFX900-NEXT:    ;;#ASMEND
1890; GFX900-NEXT:    v_mov_b32_e32 v0, v8
1891; GFX900-NEXT:    v_mov_b32_e32 v1, v9
1892; GFX900-NEXT:    global_store_dwordx2 v10, v[2:3], s[16:17] offset:16
1893; GFX900-NEXT:    global_store_dwordx4 v10, v[0:3], s[16:17]
1894; GFX900-NEXT:    s_waitcnt vmcnt(0)
1895; GFX900-NEXT:    s_setpc_b64 s[30:31]
1896;
1897; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_u_1:
1898; GFX90A:       ; %bb.0:
1899; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1900; GFX90A-NEXT:    ;;#ASMSTART
1901; GFX90A-NEXT:    ; def v[0:5]
1902; GFX90A-NEXT:    ;;#ASMEND
1903; GFX90A-NEXT:    v_mov_b32_e32 v10, 0
1904; GFX90A-NEXT:    ;;#ASMSTART
1905; GFX90A-NEXT:    ; def v[4:9]
1906; GFX90A-NEXT:    ;;#ASMEND
1907; GFX90A-NEXT:    v_mov_b32_e32 v0, v8
1908; GFX90A-NEXT:    v_mov_b32_e32 v1, v9
1909; GFX90A-NEXT:    global_store_dwordx2 v10, v[2:3], s[16:17] offset:16
1910; GFX90A-NEXT:    global_store_dwordx4 v10, v[0:3], s[16:17]
1911; GFX90A-NEXT:    s_waitcnt vmcnt(0)
1912; GFX90A-NEXT:    s_setpc_b64 s[30:31]
1913;
1914; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_u_1:
1915; GFX940:       ; %bb.0:
1916; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1917; GFX940-NEXT:    ;;#ASMSTART
1918; GFX940-NEXT:    ; def v[0:5]
1919; GFX940-NEXT:    ;;#ASMEND
1920; GFX940-NEXT:    v_mov_b32_e32 v10, 0
1921; GFX940-NEXT:    ;;#ASMSTART
1922; GFX940-NEXT:    ; def v[4:9]
1923; GFX940-NEXT:    ;;#ASMEND
1924; GFX940-NEXT:    global_store_dwordx2 v10, v[2:3], s[0:1] offset:16 sc0 sc1
1925; GFX940-NEXT:    v_mov_b32_e32 v0, v8
1926; GFX940-NEXT:    v_mov_b32_e32 v1, v9
1927; GFX940-NEXT:    global_store_dwordx4 v10, v[0:3], s[0:1] sc0 sc1
1928; GFX940-NEXT:    s_waitcnt vmcnt(0)
1929; GFX940-NEXT:    s_setpc_b64 s[30:31]
1930  %vec0 = call <3 x i64> asm "; def $0", "=v"()
1931  %vec1 = call <3 x i64> asm "; def $0", "=v"()
1932  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 poison, i32 1>
1933  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
1934  ret void
1935}
1936
1937define void @v_shuffle_v3i64_v3i64__5_0_1(ptr addrspace(1) inreg %ptr) {
1938; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_0_1:
1939; GFX900:       ; %bb.0:
1940; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1941; GFX900-NEXT:    ;;#ASMSTART
1942; GFX900-NEXT:    ; def v[0:5]
1943; GFX900-NEXT:    ;;#ASMEND
1944; GFX900-NEXT:    v_mov_b32_e32 v10, 0
1945; GFX900-NEXT:    ;;#ASMSTART
1946; GFX900-NEXT:    ; def v[4:9]
1947; GFX900-NEXT:    ;;#ASMEND
1948; GFX900-NEXT:    global_store_dwordx2 v10, v[2:3], s[16:17] offset:16
1949; GFX900-NEXT:    v_mov_b32_e32 v2, v8
1950; GFX900-NEXT:    v_mov_b32_e32 v3, v9
1951; GFX900-NEXT:    v_mov_b32_e32 v4, v0
1952; GFX900-NEXT:    v_mov_b32_e32 v5, v1
1953; GFX900-NEXT:    global_store_dwordx4 v10, v[2:5], s[16:17]
1954; GFX900-NEXT:    s_waitcnt vmcnt(0)
1955; GFX900-NEXT:    s_setpc_b64 s[30:31]
1956;
1957; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_0_1:
1958; GFX90A:       ; %bb.0:
1959; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1960; GFX90A-NEXT:    ;;#ASMSTART
1961; GFX90A-NEXT:    ; def v[0:5]
1962; GFX90A-NEXT:    ;;#ASMEND
1963; GFX90A-NEXT:    v_mov_b32_e32 v10, 0
1964; GFX90A-NEXT:    ;;#ASMSTART
1965; GFX90A-NEXT:    ; def v[4:9]
1966; GFX90A-NEXT:    ;;#ASMEND
1967; GFX90A-NEXT:    global_store_dwordx2 v10, v[2:3], s[16:17] offset:16
1968; GFX90A-NEXT:    v_mov_b32_e32 v2, v8
1969; GFX90A-NEXT:    v_mov_b32_e32 v3, v9
1970; GFX90A-NEXT:    v_mov_b32_e32 v4, v0
1971; GFX90A-NEXT:    v_mov_b32_e32 v5, v1
1972; GFX90A-NEXT:    global_store_dwordx4 v10, v[2:5], s[16:17]
1973; GFX90A-NEXT:    s_waitcnt vmcnt(0)
1974; GFX90A-NEXT:    s_setpc_b64 s[30:31]
1975;
1976; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_0_1:
1977; GFX940:       ; %bb.0:
1978; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
1979; GFX940-NEXT:    ;;#ASMSTART
1980; GFX940-NEXT:    ; def v[0:5]
1981; GFX940-NEXT:    ;;#ASMEND
1982; GFX940-NEXT:    v_mov_b32_e32 v10, 0
1983; GFX940-NEXT:    ;;#ASMSTART
1984; GFX940-NEXT:    ; def v[4:9]
1985; GFX940-NEXT:    ;;#ASMEND
1986; GFX940-NEXT:    global_store_dwordx2 v10, v[2:3], s[0:1] offset:16 sc0 sc1
1987; GFX940-NEXT:    v_mov_b32_e32 v2, v8
1988; GFX940-NEXT:    v_mov_b32_e32 v3, v9
1989; GFX940-NEXT:    v_mov_b32_e32 v4, v0
1990; GFX940-NEXT:    v_mov_b32_e32 v5, v1
1991; GFX940-NEXT:    global_store_dwordx4 v10, v[2:5], s[0:1] sc0 sc1
1992; GFX940-NEXT:    s_waitcnt vmcnt(0)
1993; GFX940-NEXT:    s_setpc_b64 s[30:31]
1994  %vec0 = call <3 x i64> asm "; def $0", "=v"()
1995  %vec1 = call <3 x i64> asm "; def $0", "=v"()
1996  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 0, i32 1>
1997  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
1998  ret void
1999}
2000
2001define void @v_shuffle_v3i64_v3i64__5_2_1(ptr addrspace(1) inreg %ptr) {
2002; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_2_1:
2003; GFX900:       ; %bb.0:
2004; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2005; GFX900-NEXT:    v_mov_b32_e32 v12, 0
2006; GFX900-NEXT:    ;;#ASMSTART
2007; GFX900-NEXT:    ; def v[0:5]
2008; GFX900-NEXT:    ;;#ASMEND
2009; GFX900-NEXT:    ;;#ASMSTART
2010; GFX900-NEXT:    ; def v[6:11]
2011; GFX900-NEXT:    ;;#ASMEND
2012; GFX900-NEXT:    global_store_dwordx2 v12, v[2:3], s[16:17] offset:16
2013; GFX900-NEXT:    v_mov_b32_e32 v2, v10
2014; GFX900-NEXT:    v_mov_b32_e32 v3, v11
2015; GFX900-NEXT:    global_store_dwordx4 v12, v[2:5], s[16:17]
2016; GFX900-NEXT:    s_waitcnt vmcnt(0)
2017; GFX900-NEXT:    s_setpc_b64 s[30:31]
2018;
2019; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_2_1:
2020; GFX90A:       ; %bb.0:
2021; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2022; GFX90A-NEXT:    v_mov_b32_e32 v12, 0
2023; GFX90A-NEXT:    ;;#ASMSTART
2024; GFX90A-NEXT:    ; def v[0:5]
2025; GFX90A-NEXT:    ;;#ASMEND
2026; GFX90A-NEXT:    ;;#ASMSTART
2027; GFX90A-NEXT:    ; def v[6:11]
2028; GFX90A-NEXT:    ;;#ASMEND
2029; GFX90A-NEXT:    global_store_dwordx2 v12, v[2:3], s[16:17] offset:16
2030; GFX90A-NEXT:    v_mov_b32_e32 v2, v10
2031; GFX90A-NEXT:    v_mov_b32_e32 v3, v11
2032; GFX90A-NEXT:    global_store_dwordx4 v12, v[2:5], s[16:17]
2033; GFX90A-NEXT:    s_waitcnt vmcnt(0)
2034; GFX90A-NEXT:    s_setpc_b64 s[30:31]
2035;
2036; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_2_1:
2037; GFX940:       ; %bb.0:
2038; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2039; GFX940-NEXT:    v_mov_b32_e32 v12, 0
2040; GFX940-NEXT:    ;;#ASMSTART
2041; GFX940-NEXT:    ; def v[0:5]
2042; GFX940-NEXT:    ;;#ASMEND
2043; GFX940-NEXT:    ;;#ASMSTART
2044; GFX940-NEXT:    ; def v[6:11]
2045; GFX940-NEXT:    ;;#ASMEND
2046; GFX940-NEXT:    global_store_dwordx2 v12, v[2:3], s[0:1] offset:16 sc0 sc1
2047; GFX940-NEXT:    v_mov_b32_e32 v2, v10
2048; GFX940-NEXT:    v_mov_b32_e32 v3, v11
2049; GFX940-NEXT:    global_store_dwordx4 v12, v[2:5], s[0:1] sc0 sc1
2050; GFX940-NEXT:    s_waitcnt vmcnt(0)
2051; GFX940-NEXT:    s_setpc_b64 s[30:31]
2052  %vec0 = call <3 x i64> asm "; def $0", "=v"()
2053  %vec1 = call <3 x i64> asm "; def $0", "=v"()
2054  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 2, i32 1>
2055  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
2056  ret void
2057}
2058
2059define void @v_shuffle_v3i64_v3i64__5_3_1(ptr addrspace(1) inreg %ptr) {
2060; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_3_1:
2061; GFX900:       ; %bb.0:
2062; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2063; GFX900-NEXT:    v_mov_b32_e32 v10, 0
2064; GFX900-NEXT:    ;;#ASMSTART
2065; GFX900-NEXT:    ; def v[0:5]
2066; GFX900-NEXT:    ;;#ASMEND
2067; GFX900-NEXT:    ;;#ASMSTART
2068; GFX900-NEXT:    ; def v[4:9]
2069; GFX900-NEXT:    ;;#ASMEND
2070; GFX900-NEXT:    global_store_dwordx2 v10, v[2:3], s[16:17] offset:16
2071; GFX900-NEXT:    v_mov_b32_e32 v0, v8
2072; GFX900-NEXT:    v_mov_b32_e32 v1, v9
2073; GFX900-NEXT:    v_mov_b32_e32 v2, v4
2074; GFX900-NEXT:    v_mov_b32_e32 v3, v5
2075; GFX900-NEXT:    global_store_dwordx4 v10, v[0:3], s[16:17]
2076; GFX900-NEXT:    s_waitcnt vmcnt(0)
2077; GFX900-NEXT:    s_setpc_b64 s[30:31]
2078;
2079; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_3_1:
2080; GFX90A:       ; %bb.0:
2081; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2082; GFX90A-NEXT:    v_mov_b32_e32 v10, 0
2083; GFX90A-NEXT:    ;;#ASMSTART
2084; GFX90A-NEXT:    ; def v[0:5]
2085; GFX90A-NEXT:    ;;#ASMEND
2086; GFX90A-NEXT:    ;;#ASMSTART
2087; GFX90A-NEXT:    ; def v[4:9]
2088; GFX90A-NEXT:    ;;#ASMEND
2089; GFX90A-NEXT:    global_store_dwordx2 v10, v[2:3], s[16:17] offset:16
2090; GFX90A-NEXT:    v_mov_b32_e32 v0, v8
2091; GFX90A-NEXT:    v_mov_b32_e32 v1, v9
2092; GFX90A-NEXT:    v_mov_b32_e32 v2, v4
2093; GFX90A-NEXT:    v_mov_b32_e32 v3, v5
2094; GFX90A-NEXT:    global_store_dwordx4 v10, v[0:3], s[16:17]
2095; GFX90A-NEXT:    s_waitcnt vmcnt(0)
2096; GFX90A-NEXT:    s_setpc_b64 s[30:31]
2097;
2098; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_3_1:
2099; GFX940:       ; %bb.0:
2100; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2101; GFX940-NEXT:    v_mov_b32_e32 v10, 0
2102; GFX940-NEXT:    ;;#ASMSTART
2103; GFX940-NEXT:    ; def v[0:5]
2104; GFX940-NEXT:    ;;#ASMEND
2105; GFX940-NEXT:    global_store_dwordx2 v10, v[2:3], s[0:1] offset:16 sc0 sc1
2106; GFX940-NEXT:    ;;#ASMSTART
2107; GFX940-NEXT:    ; def v[4:9]
2108; GFX940-NEXT:    ;;#ASMEND
2109; GFX940-NEXT:    s_nop 0
2110; GFX940-NEXT:    v_mov_b32_e32 v0, v8
2111; GFX940-NEXT:    v_mov_b32_e32 v1, v9
2112; GFX940-NEXT:    v_mov_b32_e32 v2, v4
2113; GFX940-NEXT:    v_mov_b32_e32 v3, v5
2114; GFX940-NEXT:    global_store_dwordx4 v10, v[0:3], s[0:1] sc0 sc1
2115; GFX940-NEXT:    s_waitcnt vmcnt(0)
2116; GFX940-NEXT:    s_setpc_b64 s[30:31]
2117  %vec0 = call <3 x i64> asm "; def $0", "=v"()
2118  %vec1 = call <3 x i64> asm "; def $0", "=v"()
2119  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 3, i32 1>
2120  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
2121  ret void
2122}
2123
2124define void @v_shuffle_v3i64_v3i64__5_4_1(ptr addrspace(1) inreg %ptr) {
2125; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_4_1:
2126; GFX900:       ; %bb.0:
2127; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2128; GFX900-NEXT:    ;;#ASMSTART
2129; GFX900-NEXT:    ; def v[0:5]
2130; GFX900-NEXT:    ;;#ASMEND
2131; GFX900-NEXT:    ;;#ASMSTART
2132; GFX900-NEXT:    ; def v[4:9]
2133; GFX900-NEXT:    ;;#ASMEND
2134; GFX900-NEXT:    v_mov_b32_e32 v10, 0
2135; GFX900-NEXT:    v_mov_b32_e32 v4, v8
2136; GFX900-NEXT:    v_mov_b32_e32 v5, v9
2137; GFX900-NEXT:    global_store_dwordx2 v10, v[2:3], s[16:17] offset:16
2138; GFX900-NEXT:    global_store_dwordx4 v10, v[4:7], s[16:17]
2139; GFX900-NEXT:    s_waitcnt vmcnt(0)
2140; GFX900-NEXT:    s_setpc_b64 s[30:31]
2141;
2142; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_4_1:
2143; GFX90A:       ; %bb.0:
2144; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2145; GFX90A-NEXT:    ;;#ASMSTART
2146; GFX90A-NEXT:    ; def v[0:5]
2147; GFX90A-NEXT:    ;;#ASMEND
2148; GFX90A-NEXT:    ;;#ASMSTART
2149; GFX90A-NEXT:    ; def v[4:9]
2150; GFX90A-NEXT:    ;;#ASMEND
2151; GFX90A-NEXT:    v_mov_b32_e32 v10, 0
2152; GFX90A-NEXT:    v_mov_b32_e32 v4, v8
2153; GFX90A-NEXT:    v_mov_b32_e32 v5, v9
2154; GFX90A-NEXT:    global_store_dwordx2 v10, v[2:3], s[16:17] offset:16
2155; GFX90A-NEXT:    global_store_dwordx4 v10, v[4:7], s[16:17]
2156; GFX90A-NEXT:    s_waitcnt vmcnt(0)
2157; GFX90A-NEXT:    s_setpc_b64 s[30:31]
2158;
2159; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_4_1:
2160; GFX940:       ; %bb.0:
2161; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2162; GFX940-NEXT:    ;;#ASMSTART
2163; GFX940-NEXT:    ; def v[0:5]
2164; GFX940-NEXT:    ;;#ASMEND
2165; GFX940-NEXT:    v_mov_b32_e32 v10, 0
2166; GFX940-NEXT:    ;;#ASMSTART
2167; GFX940-NEXT:    ; def v[4:9]
2168; GFX940-NEXT:    ;;#ASMEND
2169; GFX940-NEXT:    global_store_dwordx2 v10, v[2:3], s[0:1] offset:16 sc0 sc1
2170; GFX940-NEXT:    v_mov_b32_e32 v4, v8
2171; GFX940-NEXT:    v_mov_b32_e32 v5, v9
2172; GFX940-NEXT:    global_store_dwordx4 v10, v[4:7], s[0:1] sc0 sc1
2173; GFX940-NEXT:    s_waitcnt vmcnt(0)
2174; GFX940-NEXT:    s_setpc_b64 s[30:31]
2175  %vec0 = call <3 x i64> asm "; def $0", "=v"()
2176  %vec1 = call <3 x i64> asm "; def $0", "=v"()
2177  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 4, i32 1>
2178  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
2179  ret void
2180}
2181
2182define void @v_shuffle_v3i64_v3i64__u_2_2(ptr addrspace(1) inreg %ptr) {
2183; GFX900-LABEL: v_shuffle_v3i64_v3i64__u_2_2:
2184; GFX900:       ; %bb.0:
2185; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2186; GFX900-NEXT:    v_mov_b32_e32 v6, 0
2187; GFX900-NEXT:    ;;#ASMSTART
2188; GFX900-NEXT:    ; def v[0:5]
2189; GFX900-NEXT:    ;;#ASMEND
2190; GFX900-NEXT:    global_store_dwordx2 v6, v[4:5], s[16:17] offset:16
2191; GFX900-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
2192; GFX900-NEXT:    s_waitcnt vmcnt(0)
2193; GFX900-NEXT:    s_setpc_b64 s[30:31]
2194;
2195; GFX90A-LABEL: v_shuffle_v3i64_v3i64__u_2_2:
2196; GFX90A:       ; %bb.0:
2197; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2198; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
2199; GFX90A-NEXT:    ;;#ASMSTART
2200; GFX90A-NEXT:    ; def v[0:5]
2201; GFX90A-NEXT:    ;;#ASMEND
2202; GFX90A-NEXT:    global_store_dwordx2 v6, v[4:5], s[16:17] offset:16
2203; GFX90A-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
2204; GFX90A-NEXT:    s_waitcnt vmcnt(0)
2205; GFX90A-NEXT:    s_setpc_b64 s[30:31]
2206;
2207; GFX940-LABEL: v_shuffle_v3i64_v3i64__u_2_2:
2208; GFX940:       ; %bb.0:
2209; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2210; GFX940-NEXT:    v_mov_b32_e32 v6, 0
2211; GFX940-NEXT:    ;;#ASMSTART
2212; GFX940-NEXT:    ; def v[0:5]
2213; GFX940-NEXT:    ;;#ASMEND
2214; GFX940-NEXT:    global_store_dwordx2 v6, v[4:5], s[0:1] offset:16 sc0 sc1
2215; GFX940-NEXT:    global_store_dwordx4 v6, v[2:5], s[0:1] sc0 sc1
2216; GFX940-NEXT:    s_waitcnt vmcnt(0)
2217; GFX940-NEXT:    s_setpc_b64 s[30:31]
2218  %vec0 = call <3 x i64> asm "; def $0", "=v"()
2219  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 poison, i32 2, i32 2>
2220  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
2221  ret void
2222}
2223
2224define void @v_shuffle_v3i64_v3i64__0_2_2(ptr addrspace(1) inreg %ptr) {
2225; GFX900-LABEL: v_shuffle_v3i64_v3i64__0_2_2:
2226; GFX900:       ; %bb.0:
2227; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2228; GFX900-NEXT:    ;;#ASMSTART
2229; GFX900-NEXT:    ; def v[0:5]
2230; GFX900-NEXT:    ;;#ASMEND
2231; GFX900-NEXT:    v_mov_b32_e32 v6, 0
2232; GFX900-NEXT:    v_mov_b32_e32 v2, v4
2233; GFX900-NEXT:    v_mov_b32_e32 v3, v5
2234; GFX900-NEXT:    global_store_dwordx2 v6, v[4:5], s[16:17] offset:16
2235; GFX900-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
2236; GFX900-NEXT:    s_waitcnt vmcnt(0)
2237; GFX900-NEXT:    s_setpc_b64 s[30:31]
2238;
2239; GFX90A-LABEL: v_shuffle_v3i64_v3i64__0_2_2:
2240; GFX90A:       ; %bb.0:
2241; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2242; GFX90A-NEXT:    ;;#ASMSTART
2243; GFX90A-NEXT:    ; def v[0:5]
2244; GFX90A-NEXT:    ;;#ASMEND
2245; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
2246; GFX90A-NEXT:    v_mov_b32_e32 v2, v4
2247; GFX90A-NEXT:    v_mov_b32_e32 v3, v5
2248; GFX90A-NEXT:    global_store_dwordx2 v6, v[4:5], s[16:17] offset:16
2249; GFX90A-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
2250; GFX90A-NEXT:    s_waitcnt vmcnt(0)
2251; GFX90A-NEXT:    s_setpc_b64 s[30:31]
2252;
2253; GFX940-LABEL: v_shuffle_v3i64_v3i64__0_2_2:
2254; GFX940:       ; %bb.0:
2255; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2256; GFX940-NEXT:    ;;#ASMSTART
2257; GFX940-NEXT:    ; def v[0:5]
2258; GFX940-NEXT:    ;;#ASMEND
2259; GFX940-NEXT:    v_mov_b32_e32 v6, 0
2260; GFX940-NEXT:    v_mov_b32_e32 v2, v4
2261; GFX940-NEXT:    v_mov_b32_e32 v3, v5
2262; GFX940-NEXT:    global_store_dwordx2 v6, v[4:5], s[0:1] offset:16 sc0 sc1
2263; GFX940-NEXT:    global_store_dwordx4 v6, v[0:3], s[0:1] sc0 sc1
2264; GFX940-NEXT:    s_waitcnt vmcnt(0)
2265; GFX940-NEXT:    s_setpc_b64 s[30:31]
2266  %vec0 = call <3 x i64> asm "; def $0", "=v"()
2267  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 0, i32 2, i32 2>
2268  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
2269  ret void
2270}
2271
2272define void @v_shuffle_v3i64_v3i64__1_2_2(ptr addrspace(1) inreg %ptr) {
2273; GFX900-LABEL: v_shuffle_v3i64_v3i64__1_2_2:
2274; GFX900:       ; %bb.0:
2275; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2276; GFX900-NEXT:    v_mov_b32_e32 v6, 0
2277; GFX900-NEXT:    ;;#ASMSTART
2278; GFX900-NEXT:    ; def v[0:5]
2279; GFX900-NEXT:    ;;#ASMEND
2280; GFX900-NEXT:    global_store_dwordx2 v6, v[4:5], s[16:17] offset:16
2281; GFX900-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
2282; GFX900-NEXT:    s_waitcnt vmcnt(0)
2283; GFX900-NEXT:    s_setpc_b64 s[30:31]
2284;
2285; GFX90A-LABEL: v_shuffle_v3i64_v3i64__1_2_2:
2286; GFX90A:       ; %bb.0:
2287; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2288; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
2289; GFX90A-NEXT:    ;;#ASMSTART
2290; GFX90A-NEXT:    ; def v[0:5]
2291; GFX90A-NEXT:    ;;#ASMEND
2292; GFX90A-NEXT:    global_store_dwordx2 v6, v[4:5], s[16:17] offset:16
2293; GFX90A-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
2294; GFX90A-NEXT:    s_waitcnt vmcnt(0)
2295; GFX90A-NEXT:    s_setpc_b64 s[30:31]
2296;
2297; GFX940-LABEL: v_shuffle_v3i64_v3i64__1_2_2:
2298; GFX940:       ; %bb.0:
2299; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2300; GFX940-NEXT:    v_mov_b32_e32 v6, 0
2301; GFX940-NEXT:    ;;#ASMSTART
2302; GFX940-NEXT:    ; def v[0:5]
2303; GFX940-NEXT:    ;;#ASMEND
2304; GFX940-NEXT:    global_store_dwordx2 v6, v[4:5], s[0:1] offset:16 sc0 sc1
2305; GFX940-NEXT:    global_store_dwordx4 v6, v[2:5], s[0:1] sc0 sc1
2306; GFX940-NEXT:    s_waitcnt vmcnt(0)
2307; GFX940-NEXT:    s_setpc_b64 s[30:31]
2308  %vec0 = call <3 x i64> asm "; def $0", "=v"()
2309  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 1, i32 2, i32 2>
2310  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
2311  ret void
2312}
2313
2314define void @v_shuffle_v3i64_v3i64__2_2_2(ptr addrspace(1) inreg %ptr) {
2315; GFX900-LABEL: v_shuffle_v3i64_v3i64__2_2_2:
2316; GFX900:       ; %bb.0:
2317; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2318; GFX900-NEXT:    ;;#ASMSTART
2319; GFX900-NEXT:    ; def v[0:5]
2320; GFX900-NEXT:    ;;#ASMEND
2321; GFX900-NEXT:    v_mov_b32_e32 v6, 0
2322; GFX900-NEXT:    v_mov_b32_e32 v2, v4
2323; GFX900-NEXT:    v_mov_b32_e32 v3, v5
2324; GFX900-NEXT:    global_store_dwordx2 v6, v[4:5], s[16:17] offset:16
2325; GFX900-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
2326; GFX900-NEXT:    s_waitcnt vmcnt(0)
2327; GFX900-NEXT:    s_setpc_b64 s[30:31]
2328;
2329; GFX90A-LABEL: v_shuffle_v3i64_v3i64__2_2_2:
2330; GFX90A:       ; %bb.0:
2331; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2332; GFX90A-NEXT:    ;;#ASMSTART
2333; GFX90A-NEXT:    ; def v[0:5]
2334; GFX90A-NEXT:    ;;#ASMEND
2335; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
2336; GFX90A-NEXT:    v_mov_b32_e32 v2, v4
2337; GFX90A-NEXT:    v_mov_b32_e32 v3, v5
2338; GFX90A-NEXT:    global_store_dwordx2 v6, v[4:5], s[16:17] offset:16
2339; GFX90A-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
2340; GFX90A-NEXT:    s_waitcnt vmcnt(0)
2341; GFX90A-NEXT:    s_setpc_b64 s[30:31]
2342;
2343; GFX940-LABEL: v_shuffle_v3i64_v3i64__2_2_2:
2344; GFX940:       ; %bb.0:
2345; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2346; GFX940-NEXT:    ;;#ASMSTART
2347; GFX940-NEXT:    ; def v[0:5]
2348; GFX940-NEXT:    ;;#ASMEND
2349; GFX940-NEXT:    v_mov_b32_e32 v6, 0
2350; GFX940-NEXT:    v_mov_b32_e32 v2, v4
2351; GFX940-NEXT:    v_mov_b32_e32 v3, v5
2352; GFX940-NEXT:    global_store_dwordx2 v6, v[4:5], s[0:1] offset:16 sc0 sc1
2353; GFX940-NEXT:    global_store_dwordx4 v6, v[2:5], s[0:1] sc0 sc1
2354; GFX940-NEXT:    s_waitcnt vmcnt(0)
2355; GFX940-NEXT:    s_setpc_b64 s[30:31]
2356  %vec0 = call <3 x i64> asm "; def $0", "=v"()
2357  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 2, i32 2, i32 2>
2358  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
2359  ret void
2360}
2361
2362define void @v_shuffle_v3i64_v3i64__3_2_2(ptr addrspace(1) inreg %ptr) {
2363; GFX900-LABEL: v_shuffle_v3i64_v3i64__3_2_2:
2364; GFX900:       ; %bb.0:
2365; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2366; GFX900-NEXT:    v_mov_b32_e32 v6, 0
2367; GFX900-NEXT:    ;;#ASMSTART
2368; GFX900-NEXT:    ; def v[0:5]
2369; GFX900-NEXT:    ;;#ASMEND
2370; GFX900-NEXT:    global_store_dwordx2 v6, v[4:5], s[16:17] offset:16
2371; GFX900-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
2372; GFX900-NEXT:    s_waitcnt vmcnt(0)
2373; GFX900-NEXT:    s_setpc_b64 s[30:31]
2374;
2375; GFX90A-LABEL: v_shuffle_v3i64_v3i64__3_2_2:
2376; GFX90A:       ; %bb.0:
2377; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2378; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
2379; GFX90A-NEXT:    ;;#ASMSTART
2380; GFX90A-NEXT:    ; def v[0:5]
2381; GFX90A-NEXT:    ;;#ASMEND
2382; GFX90A-NEXT:    global_store_dwordx2 v6, v[4:5], s[16:17] offset:16
2383; GFX90A-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
2384; GFX90A-NEXT:    s_waitcnt vmcnt(0)
2385; GFX90A-NEXT:    s_setpc_b64 s[30:31]
2386;
2387; GFX940-LABEL: v_shuffle_v3i64_v3i64__3_2_2:
2388; GFX940:       ; %bb.0:
2389; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2390; GFX940-NEXT:    v_mov_b32_e32 v6, 0
2391; GFX940-NEXT:    ;;#ASMSTART
2392; GFX940-NEXT:    ; def v[0:5]
2393; GFX940-NEXT:    ;;#ASMEND
2394; GFX940-NEXT:    global_store_dwordx2 v6, v[4:5], s[0:1] offset:16 sc0 sc1
2395; GFX940-NEXT:    global_store_dwordx4 v6, v[2:5], s[0:1] sc0 sc1
2396; GFX940-NEXT:    s_waitcnt vmcnt(0)
2397; GFX940-NEXT:    s_setpc_b64 s[30:31]
2398  %vec0 = call <3 x i64> asm "; def $0", "=v"()
2399  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 3, i32 2, i32 2>
2400  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
2401  ret void
2402}
2403
2404define void @v_shuffle_v3i64_v3i64__4_2_2(ptr addrspace(1) inreg %ptr) {
2405; GFX900-LABEL: v_shuffle_v3i64_v3i64__4_2_2:
2406; GFX900:       ; %bb.0:
2407; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2408; GFX900-NEXT:    ;;#ASMSTART
2409; GFX900-NEXT:    ; def v[6:11]
2410; GFX900-NEXT:    ;;#ASMEND
2411; GFX900-NEXT:    v_mov_b32_e32 v12, 0
2412; GFX900-NEXT:    ;;#ASMSTART
2413; GFX900-NEXT:    ; def v[0:5]
2414; GFX900-NEXT:    ;;#ASMEND
2415; GFX900-NEXT:    v_mov_b32_e32 v10, v4
2416; GFX900-NEXT:    v_mov_b32_e32 v11, v5
2417; GFX900-NEXT:    global_store_dwordx2 v12, v[4:5], s[16:17] offset:16
2418; GFX900-NEXT:    global_store_dwordx4 v12, v[8:11], s[16:17]
2419; GFX900-NEXT:    s_waitcnt vmcnt(0)
2420; GFX900-NEXT:    s_setpc_b64 s[30:31]
2421;
2422; GFX90A-LABEL: v_shuffle_v3i64_v3i64__4_2_2:
2423; GFX90A:       ; %bb.0:
2424; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2425; GFX90A-NEXT:    ;;#ASMSTART
2426; GFX90A-NEXT:    ; def v[6:11]
2427; GFX90A-NEXT:    ;;#ASMEND
2428; GFX90A-NEXT:    v_mov_b32_e32 v12, 0
2429; GFX90A-NEXT:    ;;#ASMSTART
2430; GFX90A-NEXT:    ; def v[0:5]
2431; GFX90A-NEXT:    ;;#ASMEND
2432; GFX90A-NEXT:    v_mov_b32_e32 v10, v4
2433; GFX90A-NEXT:    v_mov_b32_e32 v11, v5
2434; GFX90A-NEXT:    global_store_dwordx2 v12, v[4:5], s[16:17] offset:16
2435; GFX90A-NEXT:    global_store_dwordx4 v12, v[8:11], s[16:17]
2436; GFX90A-NEXT:    s_waitcnt vmcnt(0)
2437; GFX90A-NEXT:    s_setpc_b64 s[30:31]
2438;
2439; GFX940-LABEL: v_shuffle_v3i64_v3i64__4_2_2:
2440; GFX940:       ; %bb.0:
2441; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2442; GFX940-NEXT:    ;;#ASMSTART
2443; GFX940-NEXT:    ; def v[6:11]
2444; GFX940-NEXT:    ;;#ASMEND
2445; GFX940-NEXT:    v_mov_b32_e32 v12, 0
2446; GFX940-NEXT:    ;;#ASMSTART
2447; GFX940-NEXT:    ; def v[0:5]
2448; GFX940-NEXT:    ;;#ASMEND
2449; GFX940-NEXT:    global_store_dwordx2 v12, v[4:5], s[0:1] offset:16 sc0 sc1
2450; GFX940-NEXT:    v_mov_b32_e32 v10, v4
2451; GFX940-NEXT:    v_mov_b32_e32 v11, v5
2452; GFX940-NEXT:    global_store_dwordx4 v12, v[8:11], s[0:1] sc0 sc1
2453; GFX940-NEXT:    s_waitcnt vmcnt(0)
2454; GFX940-NEXT:    s_setpc_b64 s[30:31]
2455  %vec0 = call <3 x i64> asm "; def $0", "=v"()
2456  %vec1 = call <3 x i64> asm "; def $0", "=v"()
2457  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 4, i32 2, i32 2>
2458  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
2459  ret void
2460}
2461
2462define void @v_shuffle_v3i64_v3i64__5_2_2(ptr addrspace(1) inreg %ptr) {
2463; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_2_2:
2464; GFX900:       ; %bb.0:
2465; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2466; GFX900-NEXT:    ;;#ASMSTART
2467; GFX900-NEXT:    ; def v[0:5]
2468; GFX900-NEXT:    ;;#ASMEND
2469; GFX900-NEXT:    v_mov_b32_e32 v12, 0
2470; GFX900-NEXT:    ;;#ASMSTART
2471; GFX900-NEXT:    ; def v[6:11]
2472; GFX900-NEXT:    ;;#ASMEND
2473; GFX900-NEXT:    v_mov_b32_e32 v2, v10
2474; GFX900-NEXT:    v_mov_b32_e32 v3, v11
2475; GFX900-NEXT:    global_store_dwordx2 v12, v[4:5], s[16:17] offset:16
2476; GFX900-NEXT:    global_store_dwordx4 v12, v[2:5], s[16:17]
2477; GFX900-NEXT:    s_waitcnt vmcnt(0)
2478; GFX900-NEXT:    s_setpc_b64 s[30:31]
2479;
2480; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_2_2:
2481; GFX90A:       ; %bb.0:
2482; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2483; GFX90A-NEXT:    ;;#ASMSTART
2484; GFX90A-NEXT:    ; def v[0:5]
2485; GFX90A-NEXT:    ;;#ASMEND
2486; GFX90A-NEXT:    v_mov_b32_e32 v12, 0
2487; GFX90A-NEXT:    ;;#ASMSTART
2488; GFX90A-NEXT:    ; def v[6:11]
2489; GFX90A-NEXT:    ;;#ASMEND
2490; GFX90A-NEXT:    v_mov_b32_e32 v2, v10
2491; GFX90A-NEXT:    v_mov_b32_e32 v3, v11
2492; GFX90A-NEXT:    global_store_dwordx2 v12, v[4:5], s[16:17] offset:16
2493; GFX90A-NEXT:    global_store_dwordx4 v12, v[2:5], s[16:17]
2494; GFX90A-NEXT:    s_waitcnt vmcnt(0)
2495; GFX90A-NEXT:    s_setpc_b64 s[30:31]
2496;
2497; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_2_2:
2498; GFX940:       ; %bb.0:
2499; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2500; GFX940-NEXT:    ;;#ASMSTART
2501; GFX940-NEXT:    ; def v[0:5]
2502; GFX940-NEXT:    ;;#ASMEND
2503; GFX940-NEXT:    v_mov_b32_e32 v12, 0
2504; GFX940-NEXT:    ;;#ASMSTART
2505; GFX940-NEXT:    ; def v[6:11]
2506; GFX940-NEXT:    ;;#ASMEND
2507; GFX940-NEXT:    global_store_dwordx2 v12, v[4:5], s[0:1] offset:16 sc0 sc1
2508; GFX940-NEXT:    v_mov_b32_e32 v2, v10
2509; GFX940-NEXT:    v_mov_b32_e32 v3, v11
2510; GFX940-NEXT:    global_store_dwordx4 v12, v[2:5], s[0:1] sc0 sc1
2511; GFX940-NEXT:    s_waitcnt vmcnt(0)
2512; GFX940-NEXT:    s_setpc_b64 s[30:31]
2513  %vec0 = call <3 x i64> asm "; def $0", "=v"()
2514  %vec1 = call <3 x i64> asm "; def $0", "=v"()
2515  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 2, i32 2>
2516  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
2517  ret void
2518}
2519
2520define void @v_shuffle_v3i64_v3i64__5_u_2(ptr addrspace(1) inreg %ptr) {
2521; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_u_2:
2522; GFX900:       ; %bb.0:
2523; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2524; GFX900-NEXT:    ;;#ASMSTART
2525; GFX900-NEXT:    ; def v[0:5]
2526; GFX900-NEXT:    ;;#ASMEND
2527; GFX900-NEXT:    v_mov_b32_e32 v12, 0
2528; GFX900-NEXT:    ;;#ASMSTART
2529; GFX900-NEXT:    ; def v[6:11]
2530; GFX900-NEXT:    ;;#ASMEND
2531; GFX900-NEXT:    v_mov_b32_e32 v0, v10
2532; GFX900-NEXT:    v_mov_b32_e32 v1, v11
2533; GFX900-NEXT:    global_store_dwordx2 v12, v[4:5], s[16:17] offset:16
2534; GFX900-NEXT:    global_store_dwordx4 v12, v[0:3], s[16:17]
2535; GFX900-NEXT:    s_waitcnt vmcnt(0)
2536; GFX900-NEXT:    s_setpc_b64 s[30:31]
2537;
2538; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_u_2:
2539; GFX90A:       ; %bb.0:
2540; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2541; GFX90A-NEXT:    ;;#ASMSTART
2542; GFX90A-NEXT:    ; def v[0:5]
2543; GFX90A-NEXT:    ;;#ASMEND
2544; GFX90A-NEXT:    v_mov_b32_e32 v12, 0
2545; GFX90A-NEXT:    ;;#ASMSTART
2546; GFX90A-NEXT:    ; def v[6:11]
2547; GFX90A-NEXT:    ;;#ASMEND
2548; GFX90A-NEXT:    v_mov_b32_e32 v0, v10
2549; GFX90A-NEXT:    v_mov_b32_e32 v1, v11
2550; GFX90A-NEXT:    global_store_dwordx2 v12, v[4:5], s[16:17] offset:16
2551; GFX90A-NEXT:    global_store_dwordx4 v12, v[0:3], s[16:17]
2552; GFX90A-NEXT:    s_waitcnt vmcnt(0)
2553; GFX90A-NEXT:    s_setpc_b64 s[30:31]
2554;
2555; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_u_2:
2556; GFX940:       ; %bb.0:
2557; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2558; GFX940-NEXT:    ;;#ASMSTART
2559; GFX940-NEXT:    ; def v[0:5]
2560; GFX940-NEXT:    ;;#ASMEND
2561; GFX940-NEXT:    v_mov_b32_e32 v12, 0
2562; GFX940-NEXT:    ;;#ASMSTART
2563; GFX940-NEXT:    ; def v[6:11]
2564; GFX940-NEXT:    ;;#ASMEND
2565; GFX940-NEXT:    global_store_dwordx2 v12, v[4:5], s[0:1] offset:16 sc0 sc1
2566; GFX940-NEXT:    v_mov_b32_e32 v0, v10
2567; GFX940-NEXT:    v_mov_b32_e32 v1, v11
2568; GFX940-NEXT:    global_store_dwordx4 v12, v[0:3], s[0:1] sc0 sc1
2569; GFX940-NEXT:    s_waitcnt vmcnt(0)
2570; GFX940-NEXT:    s_setpc_b64 s[30:31]
2571  %vec0 = call <3 x i64> asm "; def $0", "=v"()
2572  %vec1 = call <3 x i64> asm "; def $0", "=v"()
2573  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 poison, i32 2>
2574  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
2575  ret void
2576}
2577
2578define void @v_shuffle_v3i64_v3i64__5_0_2(ptr addrspace(1) inreg %ptr) {
2579; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_0_2:
2580; GFX900:       ; %bb.0:
2581; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2582; GFX900-NEXT:    v_mov_b32_e32 v12, 0
2583; GFX900-NEXT:    ;;#ASMSTART
2584; GFX900-NEXT:    ; def v[0:5]
2585; GFX900-NEXT:    ;;#ASMEND
2586; GFX900-NEXT:    ;;#ASMSTART
2587; GFX900-NEXT:    ; def v[6:11]
2588; GFX900-NEXT:    ;;#ASMEND
2589; GFX900-NEXT:    global_store_dwordx2 v12, v[4:5], s[16:17] offset:16
2590; GFX900-NEXT:    v_mov_b32_e32 v2, v10
2591; GFX900-NEXT:    v_mov_b32_e32 v3, v11
2592; GFX900-NEXT:    v_mov_b32_e32 v4, v0
2593; GFX900-NEXT:    v_mov_b32_e32 v5, v1
2594; GFX900-NEXT:    global_store_dwordx4 v12, v[2:5], s[16:17]
2595; GFX900-NEXT:    s_waitcnt vmcnt(0)
2596; GFX900-NEXT:    s_setpc_b64 s[30:31]
2597;
2598; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_0_2:
2599; GFX90A:       ; %bb.0:
2600; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2601; GFX90A-NEXT:    v_mov_b32_e32 v12, 0
2602; GFX90A-NEXT:    ;;#ASMSTART
2603; GFX90A-NEXT:    ; def v[0:5]
2604; GFX90A-NEXT:    ;;#ASMEND
2605; GFX90A-NEXT:    ;;#ASMSTART
2606; GFX90A-NEXT:    ; def v[6:11]
2607; GFX90A-NEXT:    ;;#ASMEND
2608; GFX90A-NEXT:    global_store_dwordx2 v12, v[4:5], s[16:17] offset:16
2609; GFX90A-NEXT:    v_mov_b32_e32 v2, v10
2610; GFX90A-NEXT:    v_mov_b32_e32 v3, v11
2611; GFX90A-NEXT:    v_mov_b32_e32 v4, v0
2612; GFX90A-NEXT:    v_mov_b32_e32 v5, v1
2613; GFX90A-NEXT:    global_store_dwordx4 v12, v[2:5], s[16:17]
2614; GFX90A-NEXT:    s_waitcnt vmcnt(0)
2615; GFX90A-NEXT:    s_setpc_b64 s[30:31]
2616;
2617; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_0_2:
2618; GFX940:       ; %bb.0:
2619; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2620; GFX940-NEXT:    v_mov_b32_e32 v12, 0
2621; GFX940-NEXT:    ;;#ASMSTART
2622; GFX940-NEXT:    ; def v[0:5]
2623; GFX940-NEXT:    ;;#ASMEND
2624; GFX940-NEXT:    ;;#ASMSTART
2625; GFX940-NEXT:    ; def v[6:11]
2626; GFX940-NEXT:    ;;#ASMEND
2627; GFX940-NEXT:    global_store_dwordx2 v12, v[4:5], s[0:1] offset:16 sc0 sc1
2628; GFX940-NEXT:    v_mov_b32_e32 v2, v10
2629; GFX940-NEXT:    v_mov_b32_e32 v3, v11
2630; GFX940-NEXT:    v_mov_b32_e32 v4, v0
2631; GFX940-NEXT:    v_mov_b32_e32 v5, v1
2632; GFX940-NEXT:    global_store_dwordx4 v12, v[2:5], s[0:1] sc0 sc1
2633; GFX940-NEXT:    s_waitcnt vmcnt(0)
2634; GFX940-NEXT:    s_setpc_b64 s[30:31]
2635  %vec0 = call <3 x i64> asm "; def $0", "=v"()
2636  %vec1 = call <3 x i64> asm "; def $0", "=v"()
2637  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 0, i32 2>
2638  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
2639  ret void
2640}
2641
2642define void @v_shuffle_v3i64_v3i64__5_1_2(ptr addrspace(1) inreg %ptr) {
2643; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_1_2:
2644; GFX900:       ; %bb.0:
2645; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2646; GFX900-NEXT:    ;;#ASMSTART
2647; GFX900-NEXT:    ; def v[0:5]
2648; GFX900-NEXT:    ;;#ASMEND
2649; GFX900-NEXT:    v_mov_b32_e32 v12, 0
2650; GFX900-NEXT:    ;;#ASMSTART
2651; GFX900-NEXT:    ; def v[6:11]
2652; GFX900-NEXT:    ;;#ASMEND
2653; GFX900-NEXT:    v_mov_b32_e32 v0, v10
2654; GFX900-NEXT:    v_mov_b32_e32 v1, v11
2655; GFX900-NEXT:    global_store_dwordx2 v12, v[4:5], s[16:17] offset:16
2656; GFX900-NEXT:    global_store_dwordx4 v12, v[0:3], s[16:17]
2657; GFX900-NEXT:    s_waitcnt vmcnt(0)
2658; GFX900-NEXT:    s_setpc_b64 s[30:31]
2659;
2660; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_1_2:
2661; GFX90A:       ; %bb.0:
2662; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2663; GFX90A-NEXT:    ;;#ASMSTART
2664; GFX90A-NEXT:    ; def v[0:5]
2665; GFX90A-NEXT:    ;;#ASMEND
2666; GFX90A-NEXT:    v_mov_b32_e32 v12, 0
2667; GFX90A-NEXT:    ;;#ASMSTART
2668; GFX90A-NEXT:    ; def v[6:11]
2669; GFX90A-NEXT:    ;;#ASMEND
2670; GFX90A-NEXT:    v_mov_b32_e32 v0, v10
2671; GFX90A-NEXT:    v_mov_b32_e32 v1, v11
2672; GFX90A-NEXT:    global_store_dwordx2 v12, v[4:5], s[16:17] offset:16
2673; GFX90A-NEXT:    global_store_dwordx4 v12, v[0:3], s[16:17]
2674; GFX90A-NEXT:    s_waitcnt vmcnt(0)
2675; GFX90A-NEXT:    s_setpc_b64 s[30:31]
2676;
2677; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_1_2:
2678; GFX940:       ; %bb.0:
2679; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2680; GFX940-NEXT:    ;;#ASMSTART
2681; GFX940-NEXT:    ; def v[0:5]
2682; GFX940-NEXT:    ;;#ASMEND
2683; GFX940-NEXT:    v_mov_b32_e32 v12, 0
2684; GFX940-NEXT:    ;;#ASMSTART
2685; GFX940-NEXT:    ; def v[6:11]
2686; GFX940-NEXT:    ;;#ASMEND
2687; GFX940-NEXT:    global_store_dwordx2 v12, v[4:5], s[0:1] offset:16 sc0 sc1
2688; GFX940-NEXT:    v_mov_b32_e32 v0, v10
2689; GFX940-NEXT:    v_mov_b32_e32 v1, v11
2690; GFX940-NEXT:    global_store_dwordx4 v12, v[0:3], s[0:1] sc0 sc1
2691; GFX940-NEXT:    s_waitcnt vmcnt(0)
2692; GFX940-NEXT:    s_setpc_b64 s[30:31]
2693  %vec0 = call <3 x i64> asm "; def $0", "=v"()
2694  %vec1 = call <3 x i64> asm "; def $0", "=v"()
2695  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 1, i32 2>
2696  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
2697  ret void
2698}
2699
2700define void @v_shuffle_v3i64_v3i64__5_3_2(ptr addrspace(1) inreg %ptr) {
2701; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_3_2:
2702; GFX900:       ; %bb.0:
2703; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2704; GFX900-NEXT:    ;;#ASMSTART
2705; GFX900-NEXT:    ; def v[0:5]
2706; GFX900-NEXT:    ;;#ASMEND
2707; GFX900-NEXT:    v_mov_b32_e32 v12, 0
2708; GFX900-NEXT:    ;;#ASMSTART
2709; GFX900-NEXT:    ; def v[6:11]
2710; GFX900-NEXT:    ;;#ASMEND
2711; GFX900-NEXT:    v_mov_b32_e32 v0, v10
2712; GFX900-NEXT:    v_mov_b32_e32 v1, v11
2713; GFX900-NEXT:    v_mov_b32_e32 v2, v6
2714; GFX900-NEXT:    v_mov_b32_e32 v3, v7
2715; GFX900-NEXT:    global_store_dwordx2 v12, v[4:5], s[16:17] offset:16
2716; GFX900-NEXT:    global_store_dwordx4 v12, v[0:3], s[16:17]
2717; GFX900-NEXT:    s_waitcnt vmcnt(0)
2718; GFX900-NEXT:    s_setpc_b64 s[30:31]
2719;
2720; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_3_2:
2721; GFX90A:       ; %bb.0:
2722; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2723; GFX90A-NEXT:    ;;#ASMSTART
2724; GFX90A-NEXT:    ; def v[0:5]
2725; GFX90A-NEXT:    ;;#ASMEND
2726; GFX90A-NEXT:    v_mov_b32_e32 v12, 0
2727; GFX90A-NEXT:    ;;#ASMSTART
2728; GFX90A-NEXT:    ; def v[6:11]
2729; GFX90A-NEXT:    ;;#ASMEND
2730; GFX90A-NEXT:    v_mov_b32_e32 v0, v10
2731; GFX90A-NEXT:    v_mov_b32_e32 v1, v11
2732; GFX90A-NEXT:    v_mov_b32_e32 v2, v6
2733; GFX90A-NEXT:    v_mov_b32_e32 v3, v7
2734; GFX90A-NEXT:    global_store_dwordx2 v12, v[4:5], s[16:17] offset:16
2735; GFX90A-NEXT:    global_store_dwordx4 v12, v[0:3], s[16:17]
2736; GFX90A-NEXT:    s_waitcnt vmcnt(0)
2737; GFX90A-NEXT:    s_setpc_b64 s[30:31]
2738;
2739; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_3_2:
2740; GFX940:       ; %bb.0:
2741; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2742; GFX940-NEXT:    ;;#ASMSTART
2743; GFX940-NEXT:    ; def v[0:5]
2744; GFX940-NEXT:    ;;#ASMEND
2745; GFX940-NEXT:    v_mov_b32_e32 v12, 0
2746; GFX940-NEXT:    ;;#ASMSTART
2747; GFX940-NEXT:    ; def v[6:11]
2748; GFX940-NEXT:    ;;#ASMEND
2749; GFX940-NEXT:    global_store_dwordx2 v12, v[4:5], s[0:1] offset:16 sc0 sc1
2750; GFX940-NEXT:    v_mov_b32_e32 v0, v10
2751; GFX940-NEXT:    v_mov_b32_e32 v1, v11
2752; GFX940-NEXT:    v_mov_b32_e32 v2, v6
2753; GFX940-NEXT:    v_mov_b32_e32 v3, v7
2754; GFX940-NEXT:    global_store_dwordx4 v12, v[0:3], s[0:1] sc0 sc1
2755; GFX940-NEXT:    s_waitcnt vmcnt(0)
2756; GFX940-NEXT:    s_setpc_b64 s[30:31]
2757  %vec0 = call <3 x i64> asm "; def $0", "=v"()
2758  %vec1 = call <3 x i64> asm "; def $0", "=v"()
2759  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 3, i32 2>
2760  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
2761  ret void
2762}
2763
2764define void @v_shuffle_v3i64_v3i64__5_4_2(ptr addrspace(1) inreg %ptr) {
2765; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_4_2:
2766; GFX900:       ; %bb.0:
2767; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2768; GFX900-NEXT:    ;;#ASMSTART
2769; GFX900-NEXT:    ; def v[6:11]
2770; GFX900-NEXT:    ;;#ASMEND
2771; GFX900-NEXT:    v_mov_b32_e32 v12, 0
2772; GFX900-NEXT:    v_mov_b32_e32 v6, v10
2773; GFX900-NEXT:    v_mov_b32_e32 v7, v11
2774; GFX900-NEXT:    ;;#ASMSTART
2775; GFX900-NEXT:    ; def v[0:5]
2776; GFX900-NEXT:    ;;#ASMEND
2777; GFX900-NEXT:    global_store_dwordx2 v12, v[4:5], s[16:17] offset:16
2778; GFX900-NEXT:    global_store_dwordx4 v12, v[6:9], s[16:17]
2779; GFX900-NEXT:    s_waitcnt vmcnt(0)
2780; GFX900-NEXT:    s_setpc_b64 s[30:31]
2781;
2782; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_4_2:
2783; GFX90A:       ; %bb.0:
2784; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2785; GFX90A-NEXT:    ;;#ASMSTART
2786; GFX90A-NEXT:    ; def v[6:11]
2787; GFX90A-NEXT:    ;;#ASMEND
2788; GFX90A-NEXT:    v_mov_b32_e32 v12, 0
2789; GFX90A-NEXT:    v_mov_b32_e32 v6, v10
2790; GFX90A-NEXT:    v_mov_b32_e32 v7, v11
2791; GFX90A-NEXT:    ;;#ASMSTART
2792; GFX90A-NEXT:    ; def v[0:5]
2793; GFX90A-NEXT:    ;;#ASMEND
2794; GFX90A-NEXT:    global_store_dwordx2 v12, v[4:5], s[16:17] offset:16
2795; GFX90A-NEXT:    global_store_dwordx4 v12, v[6:9], s[16:17]
2796; GFX90A-NEXT:    s_waitcnt vmcnt(0)
2797; GFX90A-NEXT:    s_setpc_b64 s[30:31]
2798;
2799; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_4_2:
2800; GFX940:       ; %bb.0:
2801; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2802; GFX940-NEXT:    ;;#ASMSTART
2803; GFX940-NEXT:    ; def v[6:11]
2804; GFX940-NEXT:    ;;#ASMEND
2805; GFX940-NEXT:    v_mov_b32_e32 v12, 0
2806; GFX940-NEXT:    v_mov_b32_e32 v6, v10
2807; GFX940-NEXT:    v_mov_b32_e32 v7, v11
2808; GFX940-NEXT:    ;;#ASMSTART
2809; GFX940-NEXT:    ; def v[0:5]
2810; GFX940-NEXT:    ;;#ASMEND
2811; GFX940-NEXT:    global_store_dwordx2 v12, v[4:5], s[0:1] offset:16 sc0 sc1
2812; GFX940-NEXT:    global_store_dwordx4 v12, v[6:9], s[0:1] sc0 sc1
2813; GFX940-NEXT:    s_waitcnt vmcnt(0)
2814; GFX940-NEXT:    s_setpc_b64 s[30:31]
2815  %vec0 = call <3 x i64> asm "; def $0", "=v"()
2816  %vec1 = call <3 x i64> asm "; def $0", "=v"()
2817  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 4, i32 2>
2818  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
2819  ret void
2820}
2821
2822define void @v_shuffle_v3i64_v3i64__u_3_3(ptr addrspace(1) inreg %ptr) {
2823; GFX9-LABEL: v_shuffle_v3i64_v3i64__u_3_3:
2824; GFX9:       ; %bb.0:
2825; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2826; GFX9-NEXT:    s_setpc_b64 s[30:31]
2827  %vec0 = call <3 x i64> asm "; def $0", "=v"()
2828  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 poison, i32 3, i32 3>
2829  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
2830  ret void
2831}
2832
2833define void @v_shuffle_v3i64_v3i64__0_3_3(ptr addrspace(1) inreg %ptr) {
2834; GFX900-LABEL: v_shuffle_v3i64_v3i64__0_3_3:
2835; GFX900:       ; %bb.0:
2836; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2837; GFX900-NEXT:    v_mov_b32_e32 v6, 0
2838; GFX900-NEXT:    ;;#ASMSTART
2839; GFX900-NEXT:    ; def v[0:5]
2840; GFX900-NEXT:    ;;#ASMEND
2841; GFX900-NEXT:    global_store_dwordx2 v6, v[4:5], s[16:17] offset:16
2842; GFX900-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
2843; GFX900-NEXT:    s_waitcnt vmcnt(0)
2844; GFX900-NEXT:    s_setpc_b64 s[30:31]
2845;
2846; GFX90A-LABEL: v_shuffle_v3i64_v3i64__0_3_3:
2847; GFX90A:       ; %bb.0:
2848; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2849; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
2850; GFX90A-NEXT:    ;;#ASMSTART
2851; GFX90A-NEXT:    ; def v[0:5]
2852; GFX90A-NEXT:    ;;#ASMEND
2853; GFX90A-NEXT:    global_store_dwordx2 v6, v[4:5], s[16:17] offset:16
2854; GFX90A-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
2855; GFX90A-NEXT:    s_waitcnt vmcnt(0)
2856; GFX90A-NEXT:    s_setpc_b64 s[30:31]
2857;
2858; GFX940-LABEL: v_shuffle_v3i64_v3i64__0_3_3:
2859; GFX940:       ; %bb.0:
2860; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2861; GFX940-NEXT:    v_mov_b32_e32 v6, 0
2862; GFX940-NEXT:    ;;#ASMSTART
2863; GFX940-NEXT:    ; def v[0:5]
2864; GFX940-NEXT:    ;;#ASMEND
2865; GFX940-NEXT:    global_store_dwordx2 v6, v[4:5], s[0:1] offset:16 sc0 sc1
2866; GFX940-NEXT:    global_store_dwordx4 v6, v[0:3], s[0:1] sc0 sc1
2867; GFX940-NEXT:    s_waitcnt vmcnt(0)
2868; GFX940-NEXT:    s_setpc_b64 s[30:31]
2869  %vec0 = call <3 x i64> asm "; def $0", "=v"()
2870  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 0, i32 3, i32 3>
2871  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
2872  ret void
2873}
2874
2875define void @v_shuffle_v3i64_v3i64__1_3_3(ptr addrspace(1) inreg %ptr) {
2876; GFX900-LABEL: v_shuffle_v3i64_v3i64__1_3_3:
2877; GFX900:       ; %bb.0:
2878; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2879; GFX900-NEXT:    v_mov_b32_e32 v6, 0
2880; GFX900-NEXT:    ;;#ASMSTART
2881; GFX900-NEXT:    ; def v[0:5]
2882; GFX900-NEXT:    ;;#ASMEND
2883; GFX900-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
2884; GFX900-NEXT:    s_waitcnt vmcnt(0)
2885; GFX900-NEXT:    s_setpc_b64 s[30:31]
2886;
2887; GFX90A-LABEL: v_shuffle_v3i64_v3i64__1_3_3:
2888; GFX90A:       ; %bb.0:
2889; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2890; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
2891; GFX90A-NEXT:    ;;#ASMSTART
2892; GFX90A-NEXT:    ; def v[0:5]
2893; GFX90A-NEXT:    ;;#ASMEND
2894; GFX90A-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
2895; GFX90A-NEXT:    s_waitcnt vmcnt(0)
2896; GFX90A-NEXT:    s_setpc_b64 s[30:31]
2897;
2898; GFX940-LABEL: v_shuffle_v3i64_v3i64__1_3_3:
2899; GFX940:       ; %bb.0:
2900; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2901; GFX940-NEXT:    v_mov_b32_e32 v6, 0
2902; GFX940-NEXT:    ;;#ASMSTART
2903; GFX940-NEXT:    ; def v[0:5]
2904; GFX940-NEXT:    ;;#ASMEND
2905; GFX940-NEXT:    global_store_dwordx4 v6, v[2:5], s[0:1] sc0 sc1
2906; GFX940-NEXT:    s_waitcnt vmcnt(0)
2907; GFX940-NEXT:    s_setpc_b64 s[30:31]
2908  %vec0 = call <3 x i64> asm "; def $0", "=v"()
2909  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 1, i32 3, i32 3>
2910  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
2911  ret void
2912}
2913
2914define void @v_shuffle_v3i64_v3i64__2_3_3(ptr addrspace(1) inreg %ptr) {
2915; GFX900-LABEL: v_shuffle_v3i64_v3i64__2_3_3:
2916; GFX900:       ; %bb.0:
2917; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2918; GFX900-NEXT:    ;;#ASMSTART
2919; GFX900-NEXT:    ; def v[0:5]
2920; GFX900-NEXT:    ;;#ASMEND
2921; GFX900-NEXT:    v_mov_b32_e32 v6, 0
2922; GFX900-NEXT:    v_mov_b32_e32 v0, v4
2923; GFX900-NEXT:    v_mov_b32_e32 v1, v5
2924; GFX900-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
2925; GFX900-NEXT:    s_waitcnt vmcnt(0)
2926; GFX900-NEXT:    s_setpc_b64 s[30:31]
2927;
2928; GFX90A-LABEL: v_shuffle_v3i64_v3i64__2_3_3:
2929; GFX90A:       ; %bb.0:
2930; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2931; GFX90A-NEXT:    ;;#ASMSTART
2932; GFX90A-NEXT:    ; def v[0:5]
2933; GFX90A-NEXT:    ;;#ASMEND
2934; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
2935; GFX90A-NEXT:    v_mov_b32_e32 v0, v4
2936; GFX90A-NEXT:    v_mov_b32_e32 v1, v5
2937; GFX90A-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
2938; GFX90A-NEXT:    s_waitcnt vmcnt(0)
2939; GFX90A-NEXT:    s_setpc_b64 s[30:31]
2940;
2941; GFX940-LABEL: v_shuffle_v3i64_v3i64__2_3_3:
2942; GFX940:       ; %bb.0:
2943; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2944; GFX940-NEXT:    ;;#ASMSTART
2945; GFX940-NEXT:    ; def v[0:5]
2946; GFX940-NEXT:    ;;#ASMEND
2947; GFX940-NEXT:    v_mov_b32_e32 v6, 0
2948; GFX940-NEXT:    v_mov_b32_e32 v0, v4
2949; GFX940-NEXT:    v_mov_b32_e32 v1, v5
2950; GFX940-NEXT:    global_store_dwordx4 v6, v[0:3], s[0:1] sc0 sc1
2951; GFX940-NEXT:    s_waitcnt vmcnt(0)
2952; GFX940-NEXT:    s_setpc_b64 s[30:31]
2953  %vec0 = call <3 x i64> asm "; def $0", "=v"()
2954  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 2, i32 3, i32 3>
2955  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
2956  ret void
2957}
2958
2959define void @v_shuffle_v3i64_v3i64__3_3_3(ptr addrspace(1) inreg %ptr) {
2960; GFX9-LABEL: v_shuffle_v3i64_v3i64__3_3_3:
2961; GFX9:       ; %bb.0:
2962; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2963; GFX9-NEXT:    s_setpc_b64 s[30:31]
2964  %vec0 = call <3 x i64> asm "; def $0", "=v"()
2965  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 3, i32 3, i32 3>
2966  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
2967  ret void
2968}
2969
2970define void @v_shuffle_v3i64_v3i64__4_3_3(ptr addrspace(1) inreg %ptr) {
2971; GFX900-LABEL: v_shuffle_v3i64_v3i64__4_3_3:
2972; GFX900:       ; %bb.0:
2973; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2974; GFX900-NEXT:    ;;#ASMSTART
2975; GFX900-NEXT:    ; def v[0:5]
2976; GFX900-NEXT:    ;;#ASMEND
2977; GFX900-NEXT:    v_mov_b32_e32 v6, 0
2978; GFX900-NEXT:    v_mov_b32_e32 v4, v0
2979; GFX900-NEXT:    v_mov_b32_e32 v5, v1
2980; GFX900-NEXT:    global_store_dwordx2 v6, v[0:1], s[16:17] offset:16
2981; GFX900-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
2982; GFX900-NEXT:    s_waitcnt vmcnt(0)
2983; GFX900-NEXT:    s_setpc_b64 s[30:31]
2984;
2985; GFX90A-LABEL: v_shuffle_v3i64_v3i64__4_3_3:
2986; GFX90A:       ; %bb.0:
2987; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2988; GFX90A-NEXT:    ;;#ASMSTART
2989; GFX90A-NEXT:    ; def v[0:5]
2990; GFX90A-NEXT:    ;;#ASMEND
2991; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
2992; GFX90A-NEXT:    v_mov_b32_e32 v4, v0
2993; GFX90A-NEXT:    v_mov_b32_e32 v5, v1
2994; GFX90A-NEXT:    global_store_dwordx2 v6, v[0:1], s[16:17] offset:16
2995; GFX90A-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
2996; GFX90A-NEXT:    s_waitcnt vmcnt(0)
2997; GFX90A-NEXT:    s_setpc_b64 s[30:31]
2998;
2999; GFX940-LABEL: v_shuffle_v3i64_v3i64__4_3_3:
3000; GFX940:       ; %bb.0:
3001; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3002; GFX940-NEXT:    ;;#ASMSTART
3003; GFX940-NEXT:    ; def v[0:5]
3004; GFX940-NEXT:    ;;#ASMEND
3005; GFX940-NEXT:    v_mov_b32_e32 v6, 0
3006; GFX940-NEXT:    v_mov_b32_e32 v4, v0
3007; GFX940-NEXT:    v_mov_b32_e32 v5, v1
3008; GFX940-NEXT:    global_store_dwordx2 v6, v[0:1], s[0:1] offset:16 sc0 sc1
3009; GFX940-NEXT:    global_store_dwordx4 v6, v[2:5], s[0:1] sc0 sc1
3010; GFX940-NEXT:    s_waitcnt vmcnt(0)
3011; GFX940-NEXT:    s_setpc_b64 s[30:31]
3012  %vec0 = call <3 x i64> asm "; def $0", "=v"()
3013  %vec1 = call <3 x i64> asm "; def $0", "=v"()
3014  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 4, i32 3, i32 3>
3015  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
3016  ret void
3017}
3018
3019define void @v_shuffle_v3i64_v3i64__5_3_3(ptr addrspace(1) inreg %ptr) {
3020; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_3_3:
3021; GFX900:       ; %bb.0:
3022; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3023; GFX900-NEXT:    ;;#ASMSTART
3024; GFX900-NEXT:    ; def v[0:5]
3025; GFX900-NEXT:    ;;#ASMEND
3026; GFX900-NEXT:    v_mov_b32_e32 v6, 0
3027; GFX900-NEXT:    v_mov_b32_e32 v2, v4
3028; GFX900-NEXT:    v_mov_b32_e32 v3, v5
3029; GFX900-NEXT:    v_mov_b32_e32 v4, v0
3030; GFX900-NEXT:    v_mov_b32_e32 v5, v1
3031; GFX900-NEXT:    global_store_dwordx2 v6, v[0:1], s[16:17] offset:16
3032; GFX900-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
3033; GFX900-NEXT:    s_waitcnt vmcnt(0)
3034; GFX900-NEXT:    s_setpc_b64 s[30:31]
3035;
3036; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_3_3:
3037; GFX90A:       ; %bb.0:
3038; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3039; GFX90A-NEXT:    ;;#ASMSTART
3040; GFX90A-NEXT:    ; def v[0:5]
3041; GFX90A-NEXT:    ;;#ASMEND
3042; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
3043; GFX90A-NEXT:    v_mov_b32_e32 v2, v4
3044; GFX90A-NEXT:    v_mov_b32_e32 v3, v5
3045; GFX90A-NEXT:    v_mov_b32_e32 v4, v0
3046; GFX90A-NEXT:    v_mov_b32_e32 v5, v1
3047; GFX90A-NEXT:    global_store_dwordx2 v6, v[0:1], s[16:17] offset:16
3048; GFX90A-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
3049; GFX90A-NEXT:    s_waitcnt vmcnt(0)
3050; GFX90A-NEXT:    s_setpc_b64 s[30:31]
3051;
3052; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_3_3:
3053; GFX940:       ; %bb.0:
3054; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3055; GFX940-NEXT:    ;;#ASMSTART
3056; GFX940-NEXT:    ; def v[0:5]
3057; GFX940-NEXT:    ;;#ASMEND
3058; GFX940-NEXT:    v_mov_b32_e32 v6, 0
3059; GFX940-NEXT:    v_mov_b32_e32 v2, v4
3060; GFX940-NEXT:    v_mov_b32_e32 v3, v5
3061; GFX940-NEXT:    v_mov_b32_e32 v4, v0
3062; GFX940-NEXT:    v_mov_b32_e32 v5, v1
3063; GFX940-NEXT:    global_store_dwordx2 v6, v[0:1], s[0:1] offset:16 sc0 sc1
3064; GFX940-NEXT:    global_store_dwordx4 v6, v[2:5], s[0:1] sc0 sc1
3065; GFX940-NEXT:    s_waitcnt vmcnt(0)
3066; GFX940-NEXT:    s_setpc_b64 s[30:31]
3067  %vec0 = call <3 x i64> asm "; def $0", "=v"()
3068  %vec1 = call <3 x i64> asm "; def $0", "=v"()
3069  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 3, i32 3>
3070  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
3071  ret void
3072}
3073
3074define void @v_shuffle_v3i64_v3i64__5_u_3(ptr addrspace(1) inreg %ptr) {
3075; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_u_3:
3076; GFX900:       ; %bb.0:
3077; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3078; GFX900-NEXT:    v_mov_b32_e32 v6, 0
3079; GFX900-NEXT:    ;;#ASMSTART
3080; GFX900-NEXT:    ; def v[0:5]
3081; GFX900-NEXT:    ;;#ASMEND
3082; GFX900-NEXT:    global_store_dwordx2 v6, v[0:1], s[16:17] offset:16
3083; GFX900-NEXT:    v_mov_b32_e32 v0, v4
3084; GFX900-NEXT:    v_mov_b32_e32 v1, v5
3085; GFX900-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
3086; GFX900-NEXT:    s_waitcnt vmcnt(0)
3087; GFX900-NEXT:    s_setpc_b64 s[30:31]
3088;
3089; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_u_3:
3090; GFX90A:       ; %bb.0:
3091; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3092; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
3093; GFX90A-NEXT:    ;;#ASMSTART
3094; GFX90A-NEXT:    ; def v[0:5]
3095; GFX90A-NEXT:    ;;#ASMEND
3096; GFX90A-NEXT:    global_store_dwordx2 v6, v[0:1], s[16:17] offset:16
3097; GFX90A-NEXT:    v_mov_b32_e32 v0, v4
3098; GFX90A-NEXT:    v_mov_b32_e32 v1, v5
3099; GFX90A-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
3100; GFX90A-NEXT:    s_waitcnt vmcnt(0)
3101; GFX90A-NEXT:    s_setpc_b64 s[30:31]
3102;
3103; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_u_3:
3104; GFX940:       ; %bb.0:
3105; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3106; GFX940-NEXT:    v_mov_b32_e32 v6, 0
3107; GFX940-NEXT:    ;;#ASMSTART
3108; GFX940-NEXT:    ; def v[0:5]
3109; GFX940-NEXT:    ;;#ASMEND
3110; GFX940-NEXT:    global_store_dwordx2 v6, v[0:1], s[0:1] offset:16 sc0 sc1
3111; GFX940-NEXT:    v_mov_b32_e32 v0, v4
3112; GFX940-NEXT:    v_mov_b32_e32 v1, v5
3113; GFX940-NEXT:    global_store_dwordx4 v6, v[0:3], s[0:1] sc0 sc1
3114; GFX940-NEXT:    s_waitcnt vmcnt(0)
3115; GFX940-NEXT:    s_setpc_b64 s[30:31]
3116  %vec0 = call <3 x i64> asm "; def $0", "=v"()
3117  %vec1 = call <3 x i64> asm "; def $0", "=v"()
3118  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 poison, i32 3>
3119  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
3120  ret void
3121}
3122
3123define void @v_shuffle_v3i64_v3i64__5_0_3(ptr addrspace(1) inreg %ptr) {
3124; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_0_3:
3125; GFX900:       ; %bb.0:
3126; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3127; GFX900-NEXT:    ;;#ASMSTART
3128; GFX900-NEXT:    ; def v[0:5]
3129; GFX900-NEXT:    ;;#ASMEND
3130; GFX900-NEXT:    v_mov_b32_e32 v8, 0
3131; GFX900-NEXT:    ;;#ASMSTART
3132; GFX900-NEXT:    ; def v[2:7]
3133; GFX900-NEXT:    ;;#ASMEND
3134; GFX900-NEXT:    global_store_dwordx2 v8, v[2:3], s[16:17] offset:16
3135; GFX900-NEXT:    v_mov_b32_e32 v2, v6
3136; GFX900-NEXT:    v_mov_b32_e32 v3, v7
3137; GFX900-NEXT:    v_mov_b32_e32 v4, v0
3138; GFX900-NEXT:    v_mov_b32_e32 v5, v1
3139; GFX900-NEXT:    global_store_dwordx4 v8, v[2:5], s[16:17]
3140; GFX900-NEXT:    s_waitcnt vmcnt(0)
3141; GFX900-NEXT:    s_setpc_b64 s[30:31]
3142;
3143; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_0_3:
3144; GFX90A:       ; %bb.0:
3145; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3146; GFX90A-NEXT:    ;;#ASMSTART
3147; GFX90A-NEXT:    ; def v[0:5]
3148; GFX90A-NEXT:    ;;#ASMEND
3149; GFX90A-NEXT:    v_mov_b32_e32 v8, 0
3150; GFX90A-NEXT:    ;;#ASMSTART
3151; GFX90A-NEXT:    ; def v[2:7]
3152; GFX90A-NEXT:    ;;#ASMEND
3153; GFX90A-NEXT:    global_store_dwordx2 v8, v[2:3], s[16:17] offset:16
3154; GFX90A-NEXT:    v_mov_b32_e32 v2, v6
3155; GFX90A-NEXT:    v_mov_b32_e32 v3, v7
3156; GFX90A-NEXT:    v_mov_b32_e32 v4, v0
3157; GFX90A-NEXT:    v_mov_b32_e32 v5, v1
3158; GFX90A-NEXT:    global_store_dwordx4 v8, v[2:5], s[16:17]
3159; GFX90A-NEXT:    s_waitcnt vmcnt(0)
3160; GFX90A-NEXT:    s_setpc_b64 s[30:31]
3161;
3162; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_0_3:
3163; GFX940:       ; %bb.0:
3164; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3165; GFX940-NEXT:    ;;#ASMSTART
3166; GFX940-NEXT:    ; def v[0:5]
3167; GFX940-NEXT:    ;;#ASMEND
3168; GFX940-NEXT:    v_mov_b32_e32 v8, 0
3169; GFX940-NEXT:    ;;#ASMSTART
3170; GFX940-NEXT:    ; def v[2:7]
3171; GFX940-NEXT:    ;;#ASMEND
3172; GFX940-NEXT:    global_store_dwordx2 v8, v[2:3], s[0:1] offset:16 sc0 sc1
3173; GFX940-NEXT:    v_mov_b32_e32 v2, v6
3174; GFX940-NEXT:    v_mov_b32_e32 v3, v7
3175; GFX940-NEXT:    v_mov_b32_e32 v4, v0
3176; GFX940-NEXT:    v_mov_b32_e32 v5, v1
3177; GFX940-NEXT:    global_store_dwordx4 v8, v[2:5], s[0:1] sc0 sc1
3178; GFX940-NEXT:    s_waitcnt vmcnt(0)
3179; GFX940-NEXT:    s_setpc_b64 s[30:31]
3180  %vec0 = call <3 x i64> asm "; def $0", "=v"()
3181  %vec1 = call <3 x i64> asm "; def $0", "=v"()
3182  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 0, i32 3>
3183  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
3184  ret void
3185}
3186
3187define void @v_shuffle_v3i64_v3i64__5_1_3(ptr addrspace(1) inreg %ptr) {
3188; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_1_3:
3189; GFX900:       ; %bb.0:
3190; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3191; GFX900-NEXT:    ;;#ASMSTART
3192; GFX900-NEXT:    ; def v[0:5]
3193; GFX900-NEXT:    ;;#ASMEND
3194; GFX900-NEXT:    v_mov_b32_e32 v10, 0
3195; GFX900-NEXT:    ;;#ASMSTART
3196; GFX900-NEXT:    ; def v[4:9]
3197; GFX900-NEXT:    ;;#ASMEND
3198; GFX900-NEXT:    v_mov_b32_e32 v0, v8
3199; GFX900-NEXT:    v_mov_b32_e32 v1, v9
3200; GFX900-NEXT:    global_store_dwordx2 v10, v[4:5], s[16:17] offset:16
3201; GFX900-NEXT:    global_store_dwordx4 v10, v[0:3], s[16:17]
3202; GFX900-NEXT:    s_waitcnt vmcnt(0)
3203; GFX900-NEXT:    s_setpc_b64 s[30:31]
3204;
3205; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_1_3:
3206; GFX90A:       ; %bb.0:
3207; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3208; GFX90A-NEXT:    ;;#ASMSTART
3209; GFX90A-NEXT:    ; def v[0:5]
3210; GFX90A-NEXT:    ;;#ASMEND
3211; GFX90A-NEXT:    v_mov_b32_e32 v10, 0
3212; GFX90A-NEXT:    ;;#ASMSTART
3213; GFX90A-NEXT:    ; def v[4:9]
3214; GFX90A-NEXT:    ;;#ASMEND
3215; GFX90A-NEXT:    v_mov_b32_e32 v0, v8
3216; GFX90A-NEXT:    v_mov_b32_e32 v1, v9
3217; GFX90A-NEXT:    global_store_dwordx2 v10, v[4:5], s[16:17] offset:16
3218; GFX90A-NEXT:    global_store_dwordx4 v10, v[0:3], s[16:17]
3219; GFX90A-NEXT:    s_waitcnt vmcnt(0)
3220; GFX90A-NEXT:    s_setpc_b64 s[30:31]
3221;
3222; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_1_3:
3223; GFX940:       ; %bb.0:
3224; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3225; GFX940-NEXT:    ;;#ASMSTART
3226; GFX940-NEXT:    ; def v[0:5]
3227; GFX940-NEXT:    ;;#ASMEND
3228; GFX940-NEXT:    v_mov_b32_e32 v10, 0
3229; GFX940-NEXT:    ;;#ASMSTART
3230; GFX940-NEXT:    ; def v[4:9]
3231; GFX940-NEXT:    ;;#ASMEND
3232; GFX940-NEXT:    global_store_dwordx2 v10, v[4:5], s[0:1] offset:16 sc0 sc1
3233; GFX940-NEXT:    v_mov_b32_e32 v0, v8
3234; GFX940-NEXT:    v_mov_b32_e32 v1, v9
3235; GFX940-NEXT:    global_store_dwordx4 v10, v[0:3], s[0:1] sc0 sc1
3236; GFX940-NEXT:    s_waitcnt vmcnt(0)
3237; GFX940-NEXT:    s_setpc_b64 s[30:31]
3238  %vec0 = call <3 x i64> asm "; def $0", "=v"()
3239  %vec1 = call <3 x i64> asm "; def $0", "=v"()
3240  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 1, i32 3>
3241  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
3242  ret void
3243}
3244
3245define void @v_shuffle_v3i64_v3i64__5_2_3(ptr addrspace(1) inreg %ptr) {
3246; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_2_3:
3247; GFX900:       ; %bb.0:
3248; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3249; GFX900-NEXT:    ;;#ASMSTART
3250; GFX900-NEXT:    ; def v[0:5]
3251; GFX900-NEXT:    ;;#ASMEND
3252; GFX900-NEXT:    v_mov_b32_e32 v12, 0
3253; GFX900-NEXT:    ;;#ASMSTART
3254; GFX900-NEXT:    ; def v[6:11]
3255; GFX900-NEXT:    ;;#ASMEND
3256; GFX900-NEXT:    v_mov_b32_e32 v2, v10
3257; GFX900-NEXT:    v_mov_b32_e32 v3, v11
3258; GFX900-NEXT:    global_store_dwordx2 v12, v[6:7], s[16:17] offset:16
3259; GFX900-NEXT:    global_store_dwordx4 v12, v[2:5], s[16:17]
3260; GFX900-NEXT:    s_waitcnt vmcnt(0)
3261; GFX900-NEXT:    s_setpc_b64 s[30:31]
3262;
3263; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_2_3:
3264; GFX90A:       ; %bb.0:
3265; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3266; GFX90A-NEXT:    ;;#ASMSTART
3267; GFX90A-NEXT:    ; def v[0:5]
3268; GFX90A-NEXT:    ;;#ASMEND
3269; GFX90A-NEXT:    v_mov_b32_e32 v12, 0
3270; GFX90A-NEXT:    ;;#ASMSTART
3271; GFX90A-NEXT:    ; def v[6:11]
3272; GFX90A-NEXT:    ;;#ASMEND
3273; GFX90A-NEXT:    v_mov_b32_e32 v2, v10
3274; GFX90A-NEXT:    v_mov_b32_e32 v3, v11
3275; GFX90A-NEXT:    global_store_dwordx2 v12, v[6:7], s[16:17] offset:16
3276; GFX90A-NEXT:    global_store_dwordx4 v12, v[2:5], s[16:17]
3277; GFX90A-NEXT:    s_waitcnt vmcnt(0)
3278; GFX90A-NEXT:    s_setpc_b64 s[30:31]
3279;
3280; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_2_3:
3281; GFX940:       ; %bb.0:
3282; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3283; GFX940-NEXT:    ;;#ASMSTART
3284; GFX940-NEXT:    ; def v[0:5]
3285; GFX940-NEXT:    ;;#ASMEND
3286; GFX940-NEXT:    v_mov_b32_e32 v12, 0
3287; GFX940-NEXT:    ;;#ASMSTART
3288; GFX940-NEXT:    ; def v[6:11]
3289; GFX940-NEXT:    ;;#ASMEND
3290; GFX940-NEXT:    global_store_dwordx2 v12, v[6:7], s[0:1] offset:16 sc0 sc1
3291; GFX940-NEXT:    v_mov_b32_e32 v2, v10
3292; GFX940-NEXT:    v_mov_b32_e32 v3, v11
3293; GFX940-NEXT:    global_store_dwordx4 v12, v[2:5], s[0:1] sc0 sc1
3294; GFX940-NEXT:    s_waitcnt vmcnt(0)
3295; GFX940-NEXT:    s_setpc_b64 s[30:31]
3296  %vec0 = call <3 x i64> asm "; def $0", "=v"()
3297  %vec1 = call <3 x i64> asm "; def $0", "=v"()
3298  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 2, i32 3>
3299  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
3300  ret void
3301}
3302
3303define void @v_shuffle_v3i64_v3i64__5_4_3(ptr addrspace(1) inreg %ptr) {
3304; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_4_3:
3305; GFX900:       ; %bb.0:
3306; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3307; GFX900-NEXT:    v_mov_b32_e32 v6, 0
3308; GFX900-NEXT:    ;;#ASMSTART
3309; GFX900-NEXT:    ; def v[0:5]
3310; GFX900-NEXT:    ;;#ASMEND
3311; GFX900-NEXT:    global_store_dwordx2 v6, v[0:1], s[16:17] offset:16
3312; GFX900-NEXT:    v_mov_b32_e32 v0, v4
3313; GFX900-NEXT:    v_mov_b32_e32 v1, v5
3314; GFX900-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
3315; GFX900-NEXT:    s_waitcnt vmcnt(0)
3316; GFX900-NEXT:    s_setpc_b64 s[30:31]
3317;
3318; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_4_3:
3319; GFX90A:       ; %bb.0:
3320; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3321; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
3322; GFX90A-NEXT:    ;;#ASMSTART
3323; GFX90A-NEXT:    ; def v[0:5]
3324; GFX90A-NEXT:    ;;#ASMEND
3325; GFX90A-NEXT:    global_store_dwordx2 v6, v[0:1], s[16:17] offset:16
3326; GFX90A-NEXT:    v_mov_b32_e32 v0, v4
3327; GFX90A-NEXT:    v_mov_b32_e32 v1, v5
3328; GFX90A-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
3329; GFX90A-NEXT:    s_waitcnt vmcnt(0)
3330; GFX90A-NEXT:    s_setpc_b64 s[30:31]
3331;
3332; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_4_3:
3333; GFX940:       ; %bb.0:
3334; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3335; GFX940-NEXT:    v_mov_b32_e32 v6, 0
3336; GFX940-NEXT:    ;;#ASMSTART
3337; GFX940-NEXT:    ; def v[0:5]
3338; GFX940-NEXT:    ;;#ASMEND
3339; GFX940-NEXT:    global_store_dwordx2 v6, v[0:1], s[0:1] offset:16 sc0 sc1
3340; GFX940-NEXT:    v_mov_b32_e32 v0, v4
3341; GFX940-NEXT:    v_mov_b32_e32 v1, v5
3342; GFX940-NEXT:    global_store_dwordx4 v6, v[0:3], s[0:1] sc0 sc1
3343; GFX940-NEXT:    s_waitcnt vmcnt(0)
3344; GFX940-NEXT:    s_setpc_b64 s[30:31]
3345  %vec0 = call <3 x i64> asm "; def $0", "=v"()
3346  %vec1 = call <3 x i64> asm "; def $0", "=v"()
3347  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 4, i32 3>
3348  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
3349  ret void
3350}
3351
3352define void @v_shuffle_v3i64_v3i64__u_4_4(ptr addrspace(1) inreg %ptr) {
3353; GFX900-LABEL: v_shuffle_v3i64_v3i64__u_4_4:
3354; GFX900:       ; %bb.0:
3355; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3356; GFX900-NEXT:    v_mov_b32_e32 v6, 0
3357; GFX900-NEXT:    ;;#ASMSTART
3358; GFX900-NEXT:    ; def v[0:5]
3359; GFX900-NEXT:    ;;#ASMEND
3360; GFX900-NEXT:    global_store_dwordx2 v6, v[2:3], s[16:17] offset:16
3361; GFX900-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
3362; GFX900-NEXT:    s_waitcnt vmcnt(0)
3363; GFX900-NEXT:    s_setpc_b64 s[30:31]
3364;
3365; GFX90A-LABEL: v_shuffle_v3i64_v3i64__u_4_4:
3366; GFX90A:       ; %bb.0:
3367; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3368; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
3369; GFX90A-NEXT:    ;;#ASMSTART
3370; GFX90A-NEXT:    ; def v[0:5]
3371; GFX90A-NEXT:    ;;#ASMEND
3372; GFX90A-NEXT:    global_store_dwordx2 v6, v[2:3], s[16:17] offset:16
3373; GFX90A-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
3374; GFX90A-NEXT:    s_waitcnt vmcnt(0)
3375; GFX90A-NEXT:    s_setpc_b64 s[30:31]
3376;
3377; GFX940-LABEL: v_shuffle_v3i64_v3i64__u_4_4:
3378; GFX940:       ; %bb.0:
3379; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3380; GFX940-NEXT:    v_mov_b32_e32 v6, 0
3381; GFX940-NEXT:    ;;#ASMSTART
3382; GFX940-NEXT:    ; def v[0:5]
3383; GFX940-NEXT:    ;;#ASMEND
3384; GFX940-NEXT:    global_store_dwordx2 v6, v[2:3], s[0:1] offset:16 sc0 sc1
3385; GFX940-NEXT:    global_store_dwordx4 v6, v[0:3], s[0:1] sc0 sc1
3386; GFX940-NEXT:    s_waitcnt vmcnt(0)
3387; GFX940-NEXT:    s_setpc_b64 s[30:31]
3388  %vec0 = call <3 x i64> asm "; def $0", "=v"()
3389  %vec1 = call <3 x i64> asm "; def $0", "=v"()
3390  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 poison, i32 4, i32 4>
3391  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
3392  ret void
3393}
3394
3395define void @v_shuffle_v3i64_v3i64__0_4_4(ptr addrspace(1) inreg %ptr) {
3396; GFX900-LABEL: v_shuffle_v3i64_v3i64__0_4_4:
3397; GFX900:       ; %bb.0:
3398; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3399; GFX900-NEXT:    ;;#ASMSTART
3400; GFX900-NEXT:    ; def v[0:5]
3401; GFX900-NEXT:    ;;#ASMEND
3402; GFX900-NEXT:    ;;#ASMSTART
3403; GFX900-NEXT:    ; def v[2:7]
3404; GFX900-NEXT:    ;;#ASMEND
3405; GFX900-NEXT:    v_mov_b32_e32 v8, 0
3406; GFX900-NEXT:    v_mov_b32_e32 v2, v4
3407; GFX900-NEXT:    v_mov_b32_e32 v3, v5
3408; GFX900-NEXT:    global_store_dwordx2 v8, v[4:5], s[16:17] offset:16
3409; GFX900-NEXT:    global_store_dwordx4 v8, v[0:3], s[16:17]
3410; GFX900-NEXT:    s_waitcnt vmcnt(0)
3411; GFX900-NEXT:    s_setpc_b64 s[30:31]
3412;
3413; GFX90A-LABEL: v_shuffle_v3i64_v3i64__0_4_4:
3414; GFX90A:       ; %bb.0:
3415; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3416; GFX90A-NEXT:    ;;#ASMSTART
3417; GFX90A-NEXT:    ; def v[0:5]
3418; GFX90A-NEXT:    ;;#ASMEND
3419; GFX90A-NEXT:    ;;#ASMSTART
3420; GFX90A-NEXT:    ; def v[2:7]
3421; GFX90A-NEXT:    ;;#ASMEND
3422; GFX90A-NEXT:    v_mov_b32_e32 v8, 0
3423; GFX90A-NEXT:    v_mov_b32_e32 v2, v4
3424; GFX90A-NEXT:    v_mov_b32_e32 v3, v5
3425; GFX90A-NEXT:    global_store_dwordx2 v8, v[4:5], s[16:17] offset:16
3426; GFX90A-NEXT:    global_store_dwordx4 v8, v[0:3], s[16:17]
3427; GFX90A-NEXT:    s_waitcnt vmcnt(0)
3428; GFX90A-NEXT:    s_setpc_b64 s[30:31]
3429;
3430; GFX940-LABEL: v_shuffle_v3i64_v3i64__0_4_4:
3431; GFX940:       ; %bb.0:
3432; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3433; GFX940-NEXT:    ;;#ASMSTART
3434; GFX940-NEXT:    ; def v[0:5]
3435; GFX940-NEXT:    ;;#ASMEND
3436; GFX940-NEXT:    v_mov_b32_e32 v8, 0
3437; GFX940-NEXT:    ;;#ASMSTART
3438; GFX940-NEXT:    ; def v[2:7]
3439; GFX940-NEXT:    ;;#ASMEND
3440; GFX940-NEXT:    global_store_dwordx2 v8, v[4:5], s[0:1] offset:16 sc0 sc1
3441; GFX940-NEXT:    v_mov_b32_e32 v2, v4
3442; GFX940-NEXT:    v_mov_b32_e32 v3, v5
3443; GFX940-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] sc0 sc1
3444; GFX940-NEXT:    s_waitcnt vmcnt(0)
3445; GFX940-NEXT:    s_setpc_b64 s[30:31]
3446  %vec0 = call <3 x i64> asm "; def $0", "=v"()
3447  %vec1 = call <3 x i64> asm "; def $0", "=v"()
3448  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 0, i32 4, i32 4>
3449  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
3450  ret void
3451}
3452
3453define void @v_shuffle_v3i64_v3i64__1_4_4(ptr addrspace(1) inreg %ptr) {
3454; GFX900-LABEL: v_shuffle_v3i64_v3i64__1_4_4:
3455; GFX900:       ; %bb.0:
3456; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3457; GFX900-NEXT:    ;;#ASMSTART
3458; GFX900-NEXT:    ; def v[0:5]
3459; GFX900-NEXT:    ;;#ASMEND
3460; GFX900-NEXT:    ;;#ASMSTART
3461; GFX900-NEXT:    ; def v[4:9]
3462; GFX900-NEXT:    ;;#ASMEND
3463; GFX900-NEXT:    v_mov_b32_e32 v10, 0
3464; GFX900-NEXT:    v_mov_b32_e32 v4, v6
3465; GFX900-NEXT:    v_mov_b32_e32 v5, v7
3466; GFX900-NEXT:    global_store_dwordx2 v10, v[6:7], s[16:17] offset:16
3467; GFX900-NEXT:    global_store_dwordx4 v10, v[2:5], s[16:17]
3468; GFX900-NEXT:    s_waitcnt vmcnt(0)
3469; GFX900-NEXT:    s_setpc_b64 s[30:31]
3470;
3471; GFX90A-LABEL: v_shuffle_v3i64_v3i64__1_4_4:
3472; GFX90A:       ; %bb.0:
3473; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3474; GFX90A-NEXT:    ;;#ASMSTART
3475; GFX90A-NEXT:    ; def v[0:5]
3476; GFX90A-NEXT:    ;;#ASMEND
3477; GFX90A-NEXT:    ;;#ASMSTART
3478; GFX90A-NEXT:    ; def v[4:9]
3479; GFX90A-NEXT:    ;;#ASMEND
3480; GFX90A-NEXT:    v_mov_b32_e32 v10, 0
3481; GFX90A-NEXT:    v_mov_b32_e32 v4, v6
3482; GFX90A-NEXT:    v_mov_b32_e32 v5, v7
3483; GFX90A-NEXT:    global_store_dwordx2 v10, v[6:7], s[16:17] offset:16
3484; GFX90A-NEXT:    global_store_dwordx4 v10, v[2:5], s[16:17]
3485; GFX90A-NEXT:    s_waitcnt vmcnt(0)
3486; GFX90A-NEXT:    s_setpc_b64 s[30:31]
3487;
3488; GFX940-LABEL: v_shuffle_v3i64_v3i64__1_4_4:
3489; GFX940:       ; %bb.0:
3490; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3491; GFX940-NEXT:    ;;#ASMSTART
3492; GFX940-NEXT:    ; def v[0:5]
3493; GFX940-NEXT:    ;;#ASMEND
3494; GFX940-NEXT:    v_mov_b32_e32 v10, 0
3495; GFX940-NEXT:    ;;#ASMSTART
3496; GFX940-NEXT:    ; def v[4:9]
3497; GFX940-NEXT:    ;;#ASMEND
3498; GFX940-NEXT:    global_store_dwordx2 v10, v[6:7], s[0:1] offset:16 sc0 sc1
3499; GFX940-NEXT:    v_mov_b32_e32 v4, v6
3500; GFX940-NEXT:    v_mov_b32_e32 v5, v7
3501; GFX940-NEXT:    global_store_dwordx4 v10, v[2:5], s[0:1] sc0 sc1
3502; GFX940-NEXT:    s_waitcnt vmcnt(0)
3503; GFX940-NEXT:    s_setpc_b64 s[30:31]
3504  %vec0 = call <3 x i64> asm "; def $0", "=v"()
3505  %vec1 = call <3 x i64> asm "; def $0", "=v"()
3506  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 1, i32 4, i32 4>
3507  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
3508  ret void
3509}
3510
3511define void @v_shuffle_v3i64_v3i64__2_4_4(ptr addrspace(1) inreg %ptr) {
3512; GFX900-LABEL: v_shuffle_v3i64_v3i64__2_4_4:
3513; GFX900:       ; %bb.0:
3514; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3515; GFX900-NEXT:    ;;#ASMSTART
3516; GFX900-NEXT:    ; def v[6:11]
3517; GFX900-NEXT:    ;;#ASMEND
3518; GFX900-NEXT:    v_mov_b32_e32 v12, 0
3519; GFX900-NEXT:    ;;#ASMSTART
3520; GFX900-NEXT:    ; def v[0:5]
3521; GFX900-NEXT:    ;;#ASMEND
3522; GFX900-NEXT:    v_mov_b32_e32 v6, v4
3523; GFX900-NEXT:    v_mov_b32_e32 v7, v5
3524; GFX900-NEXT:    global_store_dwordx2 v12, v[8:9], s[16:17] offset:16
3525; GFX900-NEXT:    global_store_dwordx4 v12, v[6:9], s[16:17]
3526; GFX900-NEXT:    s_waitcnt vmcnt(0)
3527; GFX900-NEXT:    s_setpc_b64 s[30:31]
3528;
3529; GFX90A-LABEL: v_shuffle_v3i64_v3i64__2_4_4:
3530; GFX90A:       ; %bb.0:
3531; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3532; GFX90A-NEXT:    ;;#ASMSTART
3533; GFX90A-NEXT:    ; def v[6:11]
3534; GFX90A-NEXT:    ;;#ASMEND
3535; GFX90A-NEXT:    v_mov_b32_e32 v12, 0
3536; GFX90A-NEXT:    ;;#ASMSTART
3537; GFX90A-NEXT:    ; def v[0:5]
3538; GFX90A-NEXT:    ;;#ASMEND
3539; GFX90A-NEXT:    v_mov_b32_e32 v6, v4
3540; GFX90A-NEXT:    v_mov_b32_e32 v7, v5
3541; GFX90A-NEXT:    global_store_dwordx2 v12, v[8:9], s[16:17] offset:16
3542; GFX90A-NEXT:    global_store_dwordx4 v12, v[6:9], s[16:17]
3543; GFX90A-NEXT:    s_waitcnt vmcnt(0)
3544; GFX90A-NEXT:    s_setpc_b64 s[30:31]
3545;
3546; GFX940-LABEL: v_shuffle_v3i64_v3i64__2_4_4:
3547; GFX940:       ; %bb.0:
3548; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3549; GFX940-NEXT:    ;;#ASMSTART
3550; GFX940-NEXT:    ; def v[6:11]
3551; GFX940-NEXT:    ;;#ASMEND
3552; GFX940-NEXT:    v_mov_b32_e32 v12, 0
3553; GFX940-NEXT:    ;;#ASMSTART
3554; GFX940-NEXT:    ; def v[0:5]
3555; GFX940-NEXT:    ;;#ASMEND
3556; GFX940-NEXT:    global_store_dwordx2 v12, v[8:9], s[0:1] offset:16 sc0 sc1
3557; GFX940-NEXT:    v_mov_b32_e32 v6, v4
3558; GFX940-NEXT:    v_mov_b32_e32 v7, v5
3559; GFX940-NEXT:    global_store_dwordx4 v12, v[6:9], s[0:1] sc0 sc1
3560; GFX940-NEXT:    s_waitcnt vmcnt(0)
3561; GFX940-NEXT:    s_setpc_b64 s[30:31]
3562  %vec0 = call <3 x i64> asm "; def $0", "=v"()
3563  %vec1 = call <3 x i64> asm "; def $0", "=v"()
3564  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 2, i32 4, i32 4>
3565  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
3566  ret void
3567}
3568
3569define void @v_shuffle_v3i64_v3i64__3_4_4(ptr addrspace(1) inreg %ptr) {
3570; GFX900-LABEL: v_shuffle_v3i64_v3i64__3_4_4:
3571; GFX900:       ; %bb.0:
3572; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3573; GFX900-NEXT:    v_mov_b32_e32 v6, 0
3574; GFX900-NEXT:    ;;#ASMSTART
3575; GFX900-NEXT:    ; def v[0:5]
3576; GFX900-NEXT:    ;;#ASMEND
3577; GFX900-NEXT:    global_store_dwordx2 v6, v[2:3], s[16:17] offset:16
3578; GFX900-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
3579; GFX900-NEXT:    s_waitcnt vmcnt(0)
3580; GFX900-NEXT:    s_setpc_b64 s[30:31]
3581;
3582; GFX90A-LABEL: v_shuffle_v3i64_v3i64__3_4_4:
3583; GFX90A:       ; %bb.0:
3584; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3585; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
3586; GFX90A-NEXT:    ;;#ASMSTART
3587; GFX90A-NEXT:    ; def v[0:5]
3588; GFX90A-NEXT:    ;;#ASMEND
3589; GFX90A-NEXT:    global_store_dwordx2 v6, v[2:3], s[16:17] offset:16
3590; GFX90A-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
3591; GFX90A-NEXT:    s_waitcnt vmcnt(0)
3592; GFX90A-NEXT:    s_setpc_b64 s[30:31]
3593;
3594; GFX940-LABEL: v_shuffle_v3i64_v3i64__3_4_4:
3595; GFX940:       ; %bb.0:
3596; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3597; GFX940-NEXT:    v_mov_b32_e32 v6, 0
3598; GFX940-NEXT:    ;;#ASMSTART
3599; GFX940-NEXT:    ; def v[0:5]
3600; GFX940-NEXT:    ;;#ASMEND
3601; GFX940-NEXT:    global_store_dwordx2 v6, v[2:3], s[0:1] offset:16 sc0 sc1
3602; GFX940-NEXT:    global_store_dwordx4 v6, v[0:3], s[0:1] sc0 sc1
3603; GFX940-NEXT:    s_waitcnt vmcnt(0)
3604; GFX940-NEXT:    s_setpc_b64 s[30:31]
3605  %vec0 = call <3 x i64> asm "; def $0", "=v"()
3606  %vec1 = call <3 x i64> asm "; def $0", "=v"()
3607  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 3, i32 4, i32 4>
3608  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
3609  ret void
3610}
3611
3612define void @v_shuffle_v3i64_v3i64__4_4_4(ptr addrspace(1) inreg %ptr) {
3613; GFX900-LABEL: v_shuffle_v3i64_v3i64__4_4_4:
3614; GFX900:       ; %bb.0:
3615; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3616; GFX900-NEXT:    ;;#ASMSTART
3617; GFX900-NEXT:    ; def v[0:5]
3618; GFX900-NEXT:    ;;#ASMEND
3619; GFX900-NEXT:    v_mov_b32_e32 v6, 0
3620; GFX900-NEXT:    v_mov_b32_e32 v4, v2
3621; GFX900-NEXT:    v_mov_b32_e32 v5, v3
3622; GFX900-NEXT:    global_store_dwordx2 v6, v[2:3], s[16:17] offset:16
3623; GFX900-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
3624; GFX900-NEXT:    s_waitcnt vmcnt(0)
3625; GFX900-NEXT:    s_setpc_b64 s[30:31]
3626;
3627; GFX90A-LABEL: v_shuffle_v3i64_v3i64__4_4_4:
3628; GFX90A:       ; %bb.0:
3629; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3630; GFX90A-NEXT:    ;;#ASMSTART
3631; GFX90A-NEXT:    ; def v[0:5]
3632; GFX90A-NEXT:    ;;#ASMEND
3633; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
3634; GFX90A-NEXT:    v_mov_b32_e32 v4, v2
3635; GFX90A-NEXT:    v_mov_b32_e32 v5, v3
3636; GFX90A-NEXT:    global_store_dwordx2 v6, v[2:3], s[16:17] offset:16
3637; GFX90A-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
3638; GFX90A-NEXT:    s_waitcnt vmcnt(0)
3639; GFX90A-NEXT:    s_setpc_b64 s[30:31]
3640;
3641; GFX940-LABEL: v_shuffle_v3i64_v3i64__4_4_4:
3642; GFX940:       ; %bb.0:
3643; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3644; GFX940-NEXT:    ;;#ASMSTART
3645; GFX940-NEXT:    ; def v[0:5]
3646; GFX940-NEXT:    ;;#ASMEND
3647; GFX940-NEXT:    v_mov_b32_e32 v6, 0
3648; GFX940-NEXT:    v_mov_b32_e32 v4, v2
3649; GFX940-NEXT:    v_mov_b32_e32 v5, v3
3650; GFX940-NEXT:    global_store_dwordx2 v6, v[2:3], s[0:1] offset:16 sc0 sc1
3651; GFX940-NEXT:    global_store_dwordx4 v6, v[2:5], s[0:1] sc0 sc1
3652; GFX940-NEXT:    s_waitcnt vmcnt(0)
3653; GFX940-NEXT:    s_setpc_b64 s[30:31]
3654  %vec0 = call <3 x i64> asm "; def $0", "=v"()
3655  %vec1 = call <3 x i64> asm "; def $0", "=v"()
3656  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 4, i32 4, i32 4>
3657  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
3658  ret void
3659}
3660
3661define void @v_shuffle_v3i64_v3i64__5_4_4(ptr addrspace(1) inreg %ptr) {
3662; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_4_4:
3663; GFX900:       ; %bb.0:
3664; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3665; GFX900-NEXT:    ;;#ASMSTART
3666; GFX900-NEXT:    ; def v[0:5]
3667; GFX900-NEXT:    ;;#ASMEND
3668; GFX900-NEXT:    v_mov_b32_e32 v6, 0
3669; GFX900-NEXT:    v_mov_b32_e32 v0, v4
3670; GFX900-NEXT:    v_mov_b32_e32 v1, v5
3671; GFX900-NEXT:    global_store_dwordx2 v6, v[2:3], s[16:17] offset:16
3672; GFX900-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
3673; GFX900-NEXT:    s_waitcnt vmcnt(0)
3674; GFX900-NEXT:    s_setpc_b64 s[30:31]
3675;
3676; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_4_4:
3677; GFX90A:       ; %bb.0:
3678; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3679; GFX90A-NEXT:    ;;#ASMSTART
3680; GFX90A-NEXT:    ; def v[0:5]
3681; GFX90A-NEXT:    ;;#ASMEND
3682; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
3683; GFX90A-NEXT:    v_mov_b32_e32 v0, v4
3684; GFX90A-NEXT:    v_mov_b32_e32 v1, v5
3685; GFX90A-NEXT:    global_store_dwordx2 v6, v[2:3], s[16:17] offset:16
3686; GFX90A-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
3687; GFX90A-NEXT:    s_waitcnt vmcnt(0)
3688; GFX90A-NEXT:    s_setpc_b64 s[30:31]
3689;
3690; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_4_4:
3691; GFX940:       ; %bb.0:
3692; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3693; GFX940-NEXT:    ;;#ASMSTART
3694; GFX940-NEXT:    ; def v[0:5]
3695; GFX940-NEXT:    ;;#ASMEND
3696; GFX940-NEXT:    v_mov_b32_e32 v6, 0
3697; GFX940-NEXT:    v_mov_b32_e32 v0, v4
3698; GFX940-NEXT:    v_mov_b32_e32 v1, v5
3699; GFX940-NEXT:    global_store_dwordx2 v6, v[2:3], s[0:1] offset:16 sc0 sc1
3700; GFX940-NEXT:    global_store_dwordx4 v6, v[0:3], s[0:1] sc0 sc1
3701; GFX940-NEXT:    s_waitcnt vmcnt(0)
3702; GFX940-NEXT:    s_setpc_b64 s[30:31]
3703  %vec0 = call <3 x i64> asm "; def $0", "=v"()
3704  %vec1 = call <3 x i64> asm "; def $0", "=v"()
3705  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 4, i32 4>
3706  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
3707  ret void
3708}
3709
3710define void @v_shuffle_v3i64_v3i64__5_u_4(ptr addrspace(1) inreg %ptr) {
3711; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_u_4:
3712; GFX900:       ; %bb.0:
3713; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3714; GFX900-NEXT:    ;;#ASMSTART
3715; GFX900-NEXT:    ; def v[0:5]
3716; GFX900-NEXT:    ;;#ASMEND
3717; GFX900-NEXT:    v_mov_b32_e32 v6, 0
3718; GFX900-NEXT:    v_mov_b32_e32 v0, v4
3719; GFX900-NEXT:    v_mov_b32_e32 v1, v5
3720; GFX900-NEXT:    global_store_dwordx2 v6, v[2:3], s[16:17] offset:16
3721; GFX900-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
3722; GFX900-NEXT:    s_waitcnt vmcnt(0)
3723; GFX900-NEXT:    s_setpc_b64 s[30:31]
3724;
3725; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_u_4:
3726; GFX90A:       ; %bb.0:
3727; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3728; GFX90A-NEXT:    ;;#ASMSTART
3729; GFX90A-NEXT:    ; def v[0:5]
3730; GFX90A-NEXT:    ;;#ASMEND
3731; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
3732; GFX90A-NEXT:    v_mov_b32_e32 v0, v4
3733; GFX90A-NEXT:    v_mov_b32_e32 v1, v5
3734; GFX90A-NEXT:    global_store_dwordx2 v6, v[2:3], s[16:17] offset:16
3735; GFX90A-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
3736; GFX90A-NEXT:    s_waitcnt vmcnt(0)
3737; GFX90A-NEXT:    s_setpc_b64 s[30:31]
3738;
3739; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_u_4:
3740; GFX940:       ; %bb.0:
3741; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3742; GFX940-NEXT:    ;;#ASMSTART
3743; GFX940-NEXT:    ; def v[0:5]
3744; GFX940-NEXT:    ;;#ASMEND
3745; GFX940-NEXT:    v_mov_b32_e32 v6, 0
3746; GFX940-NEXT:    v_mov_b32_e32 v0, v4
3747; GFX940-NEXT:    v_mov_b32_e32 v1, v5
3748; GFX940-NEXT:    global_store_dwordx2 v6, v[2:3], s[0:1] offset:16 sc0 sc1
3749; GFX940-NEXT:    global_store_dwordx4 v6, v[0:3], s[0:1] sc0 sc1
3750; GFX940-NEXT:    s_waitcnt vmcnt(0)
3751; GFX940-NEXT:    s_setpc_b64 s[30:31]
3752  %vec0 = call <3 x i64> asm "; def $0", "=v"()
3753  %vec1 = call <3 x i64> asm "; def $0", "=v"()
3754  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 poison, i32 4>
3755  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
3756  ret void
3757}
3758
3759define void @v_shuffle_v3i64_v3i64__5_0_4(ptr addrspace(1) inreg %ptr) {
3760; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_0_4:
3761; GFX900:       ; %bb.0:
3762; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3763; GFX900-NEXT:    ;;#ASMSTART
3764; GFX900-NEXT:    ; def v[0:5]
3765; GFX900-NEXT:    ;;#ASMEND
3766; GFX900-NEXT:    v_mov_b32_e32 v8, 0
3767; GFX900-NEXT:    ;;#ASMSTART
3768; GFX900-NEXT:    ; def v[2:7]
3769; GFX900-NEXT:    ;;#ASMEND
3770; GFX900-NEXT:    global_store_dwordx2 v8, v[4:5], s[16:17] offset:16
3771; GFX900-NEXT:    v_mov_b32_e32 v2, v6
3772; GFX900-NEXT:    v_mov_b32_e32 v3, v7
3773; GFX900-NEXT:    v_mov_b32_e32 v4, v0
3774; GFX900-NEXT:    v_mov_b32_e32 v5, v1
3775; GFX900-NEXT:    global_store_dwordx4 v8, v[2:5], s[16:17]
3776; GFX900-NEXT:    s_waitcnt vmcnt(0)
3777; GFX900-NEXT:    s_setpc_b64 s[30:31]
3778;
3779; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_0_4:
3780; GFX90A:       ; %bb.0:
3781; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3782; GFX90A-NEXT:    ;;#ASMSTART
3783; GFX90A-NEXT:    ; def v[0:5]
3784; GFX90A-NEXT:    ;;#ASMEND
3785; GFX90A-NEXT:    v_mov_b32_e32 v8, 0
3786; GFX90A-NEXT:    ;;#ASMSTART
3787; GFX90A-NEXT:    ; def v[2:7]
3788; GFX90A-NEXT:    ;;#ASMEND
3789; GFX90A-NEXT:    global_store_dwordx2 v8, v[4:5], s[16:17] offset:16
3790; GFX90A-NEXT:    v_mov_b32_e32 v2, v6
3791; GFX90A-NEXT:    v_mov_b32_e32 v3, v7
3792; GFX90A-NEXT:    v_mov_b32_e32 v4, v0
3793; GFX90A-NEXT:    v_mov_b32_e32 v5, v1
3794; GFX90A-NEXT:    global_store_dwordx4 v8, v[2:5], s[16:17]
3795; GFX90A-NEXT:    s_waitcnt vmcnt(0)
3796; GFX90A-NEXT:    s_setpc_b64 s[30:31]
3797;
3798; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_0_4:
3799; GFX940:       ; %bb.0:
3800; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3801; GFX940-NEXT:    ;;#ASMSTART
3802; GFX940-NEXT:    ; def v[0:5]
3803; GFX940-NEXT:    ;;#ASMEND
3804; GFX940-NEXT:    v_mov_b32_e32 v8, 0
3805; GFX940-NEXT:    ;;#ASMSTART
3806; GFX940-NEXT:    ; def v[2:7]
3807; GFX940-NEXT:    ;;#ASMEND
3808; GFX940-NEXT:    global_store_dwordx2 v8, v[4:5], s[0:1] offset:16 sc0 sc1
3809; GFX940-NEXT:    v_mov_b32_e32 v2, v6
3810; GFX940-NEXT:    v_mov_b32_e32 v3, v7
3811; GFX940-NEXT:    v_mov_b32_e32 v4, v0
3812; GFX940-NEXT:    v_mov_b32_e32 v5, v1
3813; GFX940-NEXT:    global_store_dwordx4 v8, v[2:5], s[0:1] sc0 sc1
3814; GFX940-NEXT:    s_waitcnt vmcnt(0)
3815; GFX940-NEXT:    s_setpc_b64 s[30:31]
3816  %vec0 = call <3 x i64> asm "; def $0", "=v"()
3817  %vec1 = call <3 x i64> asm "; def $0", "=v"()
3818  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 0, i32 4>
3819  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
3820  ret void
3821}
3822
3823define void @v_shuffle_v3i64_v3i64__5_1_4(ptr addrspace(1) inreg %ptr) {
3824; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_1_4:
3825; GFX900:       ; %bb.0:
3826; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3827; GFX900-NEXT:    ;;#ASMSTART
3828; GFX900-NEXT:    ; def v[0:5]
3829; GFX900-NEXT:    ;;#ASMEND
3830; GFX900-NEXT:    v_mov_b32_e32 v10, 0
3831; GFX900-NEXT:    ;;#ASMSTART
3832; GFX900-NEXT:    ; def v[4:9]
3833; GFX900-NEXT:    ;;#ASMEND
3834; GFX900-NEXT:    v_mov_b32_e32 v0, v8
3835; GFX900-NEXT:    v_mov_b32_e32 v1, v9
3836; GFX900-NEXT:    global_store_dwordx2 v10, v[6:7], s[16:17] offset:16
3837; GFX900-NEXT:    global_store_dwordx4 v10, v[0:3], s[16:17]
3838; GFX900-NEXT:    s_waitcnt vmcnt(0)
3839; GFX900-NEXT:    s_setpc_b64 s[30:31]
3840;
3841; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_1_4:
3842; GFX90A:       ; %bb.0:
3843; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3844; GFX90A-NEXT:    ;;#ASMSTART
3845; GFX90A-NEXT:    ; def v[0:5]
3846; GFX90A-NEXT:    ;;#ASMEND
3847; GFX90A-NEXT:    v_mov_b32_e32 v10, 0
3848; GFX90A-NEXT:    ;;#ASMSTART
3849; GFX90A-NEXT:    ; def v[4:9]
3850; GFX90A-NEXT:    ;;#ASMEND
3851; GFX90A-NEXT:    v_mov_b32_e32 v0, v8
3852; GFX90A-NEXT:    v_mov_b32_e32 v1, v9
3853; GFX90A-NEXT:    global_store_dwordx2 v10, v[6:7], s[16:17] offset:16
3854; GFX90A-NEXT:    global_store_dwordx4 v10, v[0:3], s[16:17]
3855; GFX90A-NEXT:    s_waitcnt vmcnt(0)
3856; GFX90A-NEXT:    s_setpc_b64 s[30:31]
3857;
3858; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_1_4:
3859; GFX940:       ; %bb.0:
3860; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3861; GFX940-NEXT:    ;;#ASMSTART
3862; GFX940-NEXT:    ; def v[0:5]
3863; GFX940-NEXT:    ;;#ASMEND
3864; GFX940-NEXT:    v_mov_b32_e32 v10, 0
3865; GFX940-NEXT:    ;;#ASMSTART
3866; GFX940-NEXT:    ; def v[4:9]
3867; GFX940-NEXT:    ;;#ASMEND
3868; GFX940-NEXT:    global_store_dwordx2 v10, v[6:7], s[0:1] offset:16 sc0 sc1
3869; GFX940-NEXT:    v_mov_b32_e32 v0, v8
3870; GFX940-NEXT:    v_mov_b32_e32 v1, v9
3871; GFX940-NEXT:    global_store_dwordx4 v10, v[0:3], s[0:1] sc0 sc1
3872; GFX940-NEXT:    s_waitcnt vmcnt(0)
3873; GFX940-NEXT:    s_setpc_b64 s[30:31]
3874  %vec0 = call <3 x i64> asm "; def $0", "=v"()
3875  %vec1 = call <3 x i64> asm "; def $0", "=v"()
3876  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 1, i32 4>
3877  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
3878  ret void
3879}
3880
3881define void @v_shuffle_v3i64_v3i64__5_2_4(ptr addrspace(1) inreg %ptr) {
3882; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_2_4:
3883; GFX900:       ; %bb.0:
3884; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3885; GFX900-NEXT:    ;;#ASMSTART
3886; GFX900-NEXT:    ; def v[0:5]
3887; GFX900-NEXT:    ;;#ASMEND
3888; GFX900-NEXT:    v_mov_b32_e32 v12, 0
3889; GFX900-NEXT:    ;;#ASMSTART
3890; GFX900-NEXT:    ; def v[6:11]
3891; GFX900-NEXT:    ;;#ASMEND
3892; GFX900-NEXT:    v_mov_b32_e32 v2, v10
3893; GFX900-NEXT:    v_mov_b32_e32 v3, v11
3894; GFX900-NEXT:    global_store_dwordx2 v12, v[8:9], s[16:17] offset:16
3895; GFX900-NEXT:    global_store_dwordx4 v12, v[2:5], s[16:17]
3896; GFX900-NEXT:    s_waitcnt vmcnt(0)
3897; GFX900-NEXT:    s_setpc_b64 s[30:31]
3898;
3899; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_2_4:
3900; GFX90A:       ; %bb.0:
3901; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3902; GFX90A-NEXT:    ;;#ASMSTART
3903; GFX90A-NEXT:    ; def v[0:5]
3904; GFX90A-NEXT:    ;;#ASMEND
3905; GFX90A-NEXT:    v_mov_b32_e32 v12, 0
3906; GFX90A-NEXT:    ;;#ASMSTART
3907; GFX90A-NEXT:    ; def v[6:11]
3908; GFX90A-NEXT:    ;;#ASMEND
3909; GFX90A-NEXT:    v_mov_b32_e32 v2, v10
3910; GFX90A-NEXT:    v_mov_b32_e32 v3, v11
3911; GFX90A-NEXT:    global_store_dwordx2 v12, v[8:9], s[16:17] offset:16
3912; GFX90A-NEXT:    global_store_dwordx4 v12, v[2:5], s[16:17]
3913; GFX90A-NEXT:    s_waitcnt vmcnt(0)
3914; GFX90A-NEXT:    s_setpc_b64 s[30:31]
3915;
3916; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_2_4:
3917; GFX940:       ; %bb.0:
3918; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3919; GFX940-NEXT:    ;;#ASMSTART
3920; GFX940-NEXT:    ; def v[0:5]
3921; GFX940-NEXT:    ;;#ASMEND
3922; GFX940-NEXT:    v_mov_b32_e32 v12, 0
3923; GFX940-NEXT:    ;;#ASMSTART
3924; GFX940-NEXT:    ; def v[6:11]
3925; GFX940-NEXT:    ;;#ASMEND
3926; GFX940-NEXT:    global_store_dwordx2 v12, v[8:9], s[0:1] offset:16 sc0 sc1
3927; GFX940-NEXT:    v_mov_b32_e32 v2, v10
3928; GFX940-NEXT:    v_mov_b32_e32 v3, v11
3929; GFX940-NEXT:    global_store_dwordx4 v12, v[2:5], s[0:1] sc0 sc1
3930; GFX940-NEXT:    s_waitcnt vmcnt(0)
3931; GFX940-NEXT:    s_setpc_b64 s[30:31]
3932  %vec0 = call <3 x i64> asm "; def $0", "=v"()
3933  %vec1 = call <3 x i64> asm "; def $0", "=v"()
3934  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 2, i32 4>
3935  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
3936  ret void
3937}
3938
3939define void @v_shuffle_v3i64_v3i64__5_3_4(ptr addrspace(1) inreg %ptr) {
3940; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_3_4:
3941; GFX900:       ; %bb.0:
3942; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3943; GFX900-NEXT:    v_mov_b32_e32 v6, 0
3944; GFX900-NEXT:    ;;#ASMSTART
3945; GFX900-NEXT:    ; def v[0:5]
3946; GFX900-NEXT:    ;;#ASMEND
3947; GFX900-NEXT:    global_store_dwordx2 v6, v[2:3], s[16:17] offset:16
3948; GFX900-NEXT:    v_mov_b32_e32 v2, v4
3949; GFX900-NEXT:    v_mov_b32_e32 v3, v5
3950; GFX900-NEXT:    v_mov_b32_e32 v4, v0
3951; GFX900-NEXT:    v_mov_b32_e32 v5, v1
3952; GFX900-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
3953; GFX900-NEXT:    s_waitcnt vmcnt(0)
3954; GFX900-NEXT:    s_setpc_b64 s[30:31]
3955;
3956; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_3_4:
3957; GFX90A:       ; %bb.0:
3958; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3959; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
3960; GFX90A-NEXT:    ;;#ASMSTART
3961; GFX90A-NEXT:    ; def v[0:5]
3962; GFX90A-NEXT:    ;;#ASMEND
3963; GFX90A-NEXT:    global_store_dwordx2 v6, v[2:3], s[16:17] offset:16
3964; GFX90A-NEXT:    v_mov_b32_e32 v2, v4
3965; GFX90A-NEXT:    v_mov_b32_e32 v3, v5
3966; GFX90A-NEXT:    v_mov_b32_e32 v4, v0
3967; GFX90A-NEXT:    v_mov_b32_e32 v5, v1
3968; GFX90A-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
3969; GFX90A-NEXT:    s_waitcnt vmcnt(0)
3970; GFX90A-NEXT:    s_setpc_b64 s[30:31]
3971;
3972; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_3_4:
3973; GFX940:       ; %bb.0:
3974; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3975; GFX940-NEXT:    v_mov_b32_e32 v6, 0
3976; GFX940-NEXT:    ;;#ASMSTART
3977; GFX940-NEXT:    ; def v[0:5]
3978; GFX940-NEXT:    ;;#ASMEND
3979; GFX940-NEXT:    global_store_dwordx2 v6, v[2:3], s[0:1] offset:16 sc0 sc1
3980; GFX940-NEXT:    v_mov_b32_e32 v2, v4
3981; GFX940-NEXT:    v_mov_b32_e32 v3, v5
3982; GFX940-NEXT:    v_mov_b32_e32 v4, v0
3983; GFX940-NEXT:    v_mov_b32_e32 v5, v1
3984; GFX940-NEXT:    global_store_dwordx4 v6, v[2:5], s[0:1] sc0 sc1
3985; GFX940-NEXT:    s_waitcnt vmcnt(0)
3986; GFX940-NEXT:    s_setpc_b64 s[30:31]
3987  %vec0 = call <3 x i64> asm "; def $0", "=v"()
3988  %vec1 = call <3 x i64> asm "; def $0", "=v"()
3989  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 3, i32 4>
3990  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
3991  ret void
3992}
3993
3994define void @v_shuffle_v3i64_v3i64__u_5_5(ptr addrspace(1) inreg %ptr) {
3995; GFX900-LABEL: v_shuffle_v3i64_v3i64__u_5_5:
3996; GFX900:       ; %bb.0:
3997; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3998; GFX900-NEXT:    v_mov_b32_e32 v6, 0
3999; GFX900-NEXT:    ;;#ASMSTART
4000; GFX900-NEXT:    ; def v[0:5]
4001; GFX900-NEXT:    ;;#ASMEND
4002; GFX900-NEXT:    global_store_dwordx2 v6, v[4:5], s[16:17] offset:16
4003; GFX900-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
4004; GFX900-NEXT:    s_waitcnt vmcnt(0)
4005; GFX900-NEXT:    s_setpc_b64 s[30:31]
4006;
4007; GFX90A-LABEL: v_shuffle_v3i64_v3i64__u_5_5:
4008; GFX90A:       ; %bb.0:
4009; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4010; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
4011; GFX90A-NEXT:    ;;#ASMSTART
4012; GFX90A-NEXT:    ; def v[0:5]
4013; GFX90A-NEXT:    ;;#ASMEND
4014; GFX90A-NEXT:    global_store_dwordx2 v6, v[4:5], s[16:17] offset:16
4015; GFX90A-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
4016; GFX90A-NEXT:    s_waitcnt vmcnt(0)
4017; GFX90A-NEXT:    s_setpc_b64 s[30:31]
4018;
4019; GFX940-LABEL: v_shuffle_v3i64_v3i64__u_5_5:
4020; GFX940:       ; %bb.0:
4021; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4022; GFX940-NEXT:    v_mov_b32_e32 v6, 0
4023; GFX940-NEXT:    ;;#ASMSTART
4024; GFX940-NEXT:    ; def v[0:5]
4025; GFX940-NEXT:    ;;#ASMEND
4026; GFX940-NEXT:    global_store_dwordx2 v6, v[4:5], s[0:1] offset:16 sc0 sc1
4027; GFX940-NEXT:    global_store_dwordx4 v6, v[2:5], s[0:1] sc0 sc1
4028; GFX940-NEXT:    s_waitcnt vmcnt(0)
4029; GFX940-NEXT:    s_setpc_b64 s[30:31]
4030  %vec0 = call <3 x i64> asm "; def $0", "=v"()
4031  %vec1 = call <3 x i64> asm "; def $0", "=v"()
4032  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 poison, i32 5, i32 5>
4033  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
4034  ret void
4035}
4036
4037define void @v_shuffle_v3i64_v3i64__0_5_5(ptr addrspace(1) inreg %ptr) {
4038; GFX900-LABEL: v_shuffle_v3i64_v3i64__0_5_5:
4039; GFX900:       ; %bb.0:
4040; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4041; GFX900-NEXT:    ;;#ASMSTART
4042; GFX900-NEXT:    ; def v[0:5]
4043; GFX900-NEXT:    ;;#ASMEND
4044; GFX900-NEXT:    ;;#ASMSTART
4045; GFX900-NEXT:    ; def v[2:7]
4046; GFX900-NEXT:    ;;#ASMEND
4047; GFX900-NEXT:    v_mov_b32_e32 v8, 0
4048; GFX900-NEXT:    v_mov_b32_e32 v2, v6
4049; GFX900-NEXT:    v_mov_b32_e32 v3, v7
4050; GFX900-NEXT:    global_store_dwordx2 v8, v[6:7], s[16:17] offset:16
4051; GFX900-NEXT:    global_store_dwordx4 v8, v[0:3], s[16:17]
4052; GFX900-NEXT:    s_waitcnt vmcnt(0)
4053; GFX900-NEXT:    s_setpc_b64 s[30:31]
4054;
4055; GFX90A-LABEL: v_shuffle_v3i64_v3i64__0_5_5:
4056; GFX90A:       ; %bb.0:
4057; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4058; GFX90A-NEXT:    ;;#ASMSTART
4059; GFX90A-NEXT:    ; def v[0:5]
4060; GFX90A-NEXT:    ;;#ASMEND
4061; GFX90A-NEXT:    ;;#ASMSTART
4062; GFX90A-NEXT:    ; def v[2:7]
4063; GFX90A-NEXT:    ;;#ASMEND
4064; GFX90A-NEXT:    v_mov_b32_e32 v8, 0
4065; GFX90A-NEXT:    v_mov_b32_e32 v2, v6
4066; GFX90A-NEXT:    v_mov_b32_e32 v3, v7
4067; GFX90A-NEXT:    global_store_dwordx2 v8, v[6:7], s[16:17] offset:16
4068; GFX90A-NEXT:    global_store_dwordx4 v8, v[0:3], s[16:17]
4069; GFX90A-NEXT:    s_waitcnt vmcnt(0)
4070; GFX90A-NEXT:    s_setpc_b64 s[30:31]
4071;
4072; GFX940-LABEL: v_shuffle_v3i64_v3i64__0_5_5:
4073; GFX940:       ; %bb.0:
4074; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4075; GFX940-NEXT:    ;;#ASMSTART
4076; GFX940-NEXT:    ; def v[0:5]
4077; GFX940-NEXT:    ;;#ASMEND
4078; GFX940-NEXT:    v_mov_b32_e32 v8, 0
4079; GFX940-NEXT:    ;;#ASMSTART
4080; GFX940-NEXT:    ; def v[2:7]
4081; GFX940-NEXT:    ;;#ASMEND
4082; GFX940-NEXT:    global_store_dwordx2 v8, v[6:7], s[0:1] offset:16 sc0 sc1
4083; GFX940-NEXT:    v_mov_b32_e32 v2, v6
4084; GFX940-NEXT:    v_mov_b32_e32 v3, v7
4085; GFX940-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] sc0 sc1
4086; GFX940-NEXT:    s_waitcnt vmcnt(0)
4087; GFX940-NEXT:    s_setpc_b64 s[30:31]
4088  %vec0 = call <3 x i64> asm "; def $0", "=v"()
4089  %vec1 = call <3 x i64> asm "; def $0", "=v"()
4090  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 0, i32 5, i32 5>
4091  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
4092  ret void
4093}
4094
4095define void @v_shuffle_v3i64_v3i64__1_5_5(ptr addrspace(1) inreg %ptr) {
4096; GFX900-LABEL: v_shuffle_v3i64_v3i64__1_5_5:
4097; GFX900:       ; %bb.0:
4098; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4099; GFX900-NEXT:    ;;#ASMSTART
4100; GFX900-NEXT:    ; def v[0:5]
4101; GFX900-NEXT:    ;;#ASMEND
4102; GFX900-NEXT:    ;;#ASMSTART
4103; GFX900-NEXT:    ; def v[4:9]
4104; GFX900-NEXT:    ;;#ASMEND
4105; GFX900-NEXT:    v_mov_b32_e32 v10, 0
4106; GFX900-NEXT:    v_mov_b32_e32 v4, v8
4107; GFX900-NEXT:    v_mov_b32_e32 v5, v9
4108; GFX900-NEXT:    global_store_dwordx2 v10, v[8:9], s[16:17] offset:16
4109; GFX900-NEXT:    global_store_dwordx4 v10, v[2:5], s[16:17]
4110; GFX900-NEXT:    s_waitcnt vmcnt(0)
4111; GFX900-NEXT:    s_setpc_b64 s[30:31]
4112;
4113; GFX90A-LABEL: v_shuffle_v3i64_v3i64__1_5_5:
4114; GFX90A:       ; %bb.0:
4115; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4116; GFX90A-NEXT:    ;;#ASMSTART
4117; GFX90A-NEXT:    ; def v[0:5]
4118; GFX90A-NEXT:    ;;#ASMEND
4119; GFX90A-NEXT:    ;;#ASMSTART
4120; GFX90A-NEXT:    ; def v[4:9]
4121; GFX90A-NEXT:    ;;#ASMEND
4122; GFX90A-NEXT:    v_mov_b32_e32 v10, 0
4123; GFX90A-NEXT:    v_mov_b32_e32 v4, v8
4124; GFX90A-NEXT:    v_mov_b32_e32 v5, v9
4125; GFX90A-NEXT:    global_store_dwordx2 v10, v[8:9], s[16:17] offset:16
4126; GFX90A-NEXT:    global_store_dwordx4 v10, v[2:5], s[16:17]
4127; GFX90A-NEXT:    s_waitcnt vmcnt(0)
4128; GFX90A-NEXT:    s_setpc_b64 s[30:31]
4129;
4130; GFX940-LABEL: v_shuffle_v3i64_v3i64__1_5_5:
4131; GFX940:       ; %bb.0:
4132; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4133; GFX940-NEXT:    ;;#ASMSTART
4134; GFX940-NEXT:    ; def v[0:5]
4135; GFX940-NEXT:    ;;#ASMEND
4136; GFX940-NEXT:    v_mov_b32_e32 v10, 0
4137; GFX940-NEXT:    ;;#ASMSTART
4138; GFX940-NEXT:    ; def v[4:9]
4139; GFX940-NEXT:    ;;#ASMEND
4140; GFX940-NEXT:    global_store_dwordx2 v10, v[8:9], s[0:1] offset:16 sc0 sc1
4141; GFX940-NEXT:    v_mov_b32_e32 v4, v8
4142; GFX940-NEXT:    v_mov_b32_e32 v5, v9
4143; GFX940-NEXT:    global_store_dwordx4 v10, v[2:5], s[0:1] sc0 sc1
4144; GFX940-NEXT:    s_waitcnt vmcnt(0)
4145; GFX940-NEXT:    s_setpc_b64 s[30:31]
4146  %vec0 = call <3 x i64> asm "; def $0", "=v"()
4147  %vec1 = call <3 x i64> asm "; def $0", "=v"()
4148  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 1, i32 5, i32 5>
4149  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
4150  ret void
4151}
4152
4153define void @v_shuffle_v3i64_v3i64__2_5_5(ptr addrspace(1) inreg %ptr) {
4154; GFX900-LABEL: v_shuffle_v3i64_v3i64__2_5_5:
4155; GFX900:       ; %bb.0:
4156; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4157; GFX900-NEXT:    ;;#ASMSTART
4158; GFX900-NEXT:    ; def v[6:11]
4159; GFX900-NEXT:    ;;#ASMEND
4160; GFX900-NEXT:    v_mov_b32_e32 v12, 0
4161; GFX900-NEXT:    ;;#ASMSTART
4162; GFX900-NEXT:    ; def v[0:5]
4163; GFX900-NEXT:    ;;#ASMEND
4164; GFX900-NEXT:    v_mov_b32_e32 v8, v4
4165; GFX900-NEXT:    v_mov_b32_e32 v9, v5
4166; GFX900-NEXT:    global_store_dwordx2 v12, v[10:11], s[16:17] offset:16
4167; GFX900-NEXT:    global_store_dwordx4 v12, v[8:11], s[16:17]
4168; GFX900-NEXT:    s_waitcnt vmcnt(0)
4169; GFX900-NEXT:    s_setpc_b64 s[30:31]
4170;
4171; GFX90A-LABEL: v_shuffle_v3i64_v3i64__2_5_5:
4172; GFX90A:       ; %bb.0:
4173; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4174; GFX90A-NEXT:    ;;#ASMSTART
4175; GFX90A-NEXT:    ; def v[6:11]
4176; GFX90A-NEXT:    ;;#ASMEND
4177; GFX90A-NEXT:    v_mov_b32_e32 v12, 0
4178; GFX90A-NEXT:    ;;#ASMSTART
4179; GFX90A-NEXT:    ; def v[0:5]
4180; GFX90A-NEXT:    ;;#ASMEND
4181; GFX90A-NEXT:    v_mov_b32_e32 v8, v4
4182; GFX90A-NEXT:    v_mov_b32_e32 v9, v5
4183; GFX90A-NEXT:    global_store_dwordx2 v12, v[10:11], s[16:17] offset:16
4184; GFX90A-NEXT:    global_store_dwordx4 v12, v[8:11], s[16:17]
4185; GFX90A-NEXT:    s_waitcnt vmcnt(0)
4186; GFX90A-NEXT:    s_setpc_b64 s[30:31]
4187;
4188; GFX940-LABEL: v_shuffle_v3i64_v3i64__2_5_5:
4189; GFX940:       ; %bb.0:
4190; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4191; GFX940-NEXT:    ;;#ASMSTART
4192; GFX940-NEXT:    ; def v[6:11]
4193; GFX940-NEXT:    ;;#ASMEND
4194; GFX940-NEXT:    v_mov_b32_e32 v12, 0
4195; GFX940-NEXT:    ;;#ASMSTART
4196; GFX940-NEXT:    ; def v[0:5]
4197; GFX940-NEXT:    ;;#ASMEND
4198; GFX940-NEXT:    global_store_dwordx2 v12, v[10:11], s[0:1] offset:16 sc0 sc1
4199; GFX940-NEXT:    v_mov_b32_e32 v8, v4
4200; GFX940-NEXT:    v_mov_b32_e32 v9, v5
4201; GFX940-NEXT:    global_store_dwordx4 v12, v[8:11], s[0:1] sc0 sc1
4202; GFX940-NEXT:    s_waitcnt vmcnt(0)
4203; GFX940-NEXT:    s_setpc_b64 s[30:31]
4204  %vec0 = call <3 x i64> asm "; def $0", "=v"()
4205  %vec1 = call <3 x i64> asm "; def $0", "=v"()
4206  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 2, i32 5, i32 5>
4207  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
4208  ret void
4209}
4210
4211define void @v_shuffle_v3i64_v3i64__3_5_5(ptr addrspace(1) inreg %ptr) {
4212; GFX900-LABEL: v_shuffle_v3i64_v3i64__3_5_5:
4213; GFX900:       ; %bb.0:
4214; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4215; GFX900-NEXT:    ;;#ASMSTART
4216; GFX900-NEXT:    ; def v[0:5]
4217; GFX900-NEXT:    ;;#ASMEND
4218; GFX900-NEXT:    v_mov_b32_e32 v6, 0
4219; GFX900-NEXT:    v_mov_b32_e32 v2, v4
4220; GFX900-NEXT:    v_mov_b32_e32 v3, v5
4221; GFX900-NEXT:    global_store_dwordx2 v6, v[4:5], s[16:17] offset:16
4222; GFX900-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
4223; GFX900-NEXT:    s_waitcnt vmcnt(0)
4224; GFX900-NEXT:    s_setpc_b64 s[30:31]
4225;
4226; GFX90A-LABEL: v_shuffle_v3i64_v3i64__3_5_5:
4227; GFX90A:       ; %bb.0:
4228; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4229; GFX90A-NEXT:    ;;#ASMSTART
4230; GFX90A-NEXT:    ; def v[0:5]
4231; GFX90A-NEXT:    ;;#ASMEND
4232; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
4233; GFX90A-NEXT:    v_mov_b32_e32 v2, v4
4234; GFX90A-NEXT:    v_mov_b32_e32 v3, v5
4235; GFX90A-NEXT:    global_store_dwordx2 v6, v[4:5], s[16:17] offset:16
4236; GFX90A-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
4237; GFX90A-NEXT:    s_waitcnt vmcnt(0)
4238; GFX90A-NEXT:    s_setpc_b64 s[30:31]
4239;
4240; GFX940-LABEL: v_shuffle_v3i64_v3i64__3_5_5:
4241; GFX940:       ; %bb.0:
4242; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4243; GFX940-NEXT:    ;;#ASMSTART
4244; GFX940-NEXT:    ; def v[0:5]
4245; GFX940-NEXT:    ;;#ASMEND
4246; GFX940-NEXT:    v_mov_b32_e32 v6, 0
4247; GFX940-NEXT:    v_mov_b32_e32 v2, v4
4248; GFX940-NEXT:    v_mov_b32_e32 v3, v5
4249; GFX940-NEXT:    global_store_dwordx2 v6, v[4:5], s[0:1] offset:16 sc0 sc1
4250; GFX940-NEXT:    global_store_dwordx4 v6, v[0:3], s[0:1] sc0 sc1
4251; GFX940-NEXT:    s_waitcnt vmcnt(0)
4252; GFX940-NEXT:    s_setpc_b64 s[30:31]
4253  %vec0 = call <3 x i64> asm "; def $0", "=v"()
4254  %vec1 = call <3 x i64> asm "; def $0", "=v"()
4255  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 3, i32 5, i32 5>
4256  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
4257  ret void
4258}
4259
4260define void @v_shuffle_v3i64_v3i64__4_5_5(ptr addrspace(1) inreg %ptr) {
4261; GFX900-LABEL: v_shuffle_v3i64_v3i64__4_5_5:
4262; GFX900:       ; %bb.0:
4263; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4264; GFX900-NEXT:    v_mov_b32_e32 v6, 0
4265; GFX900-NEXT:    ;;#ASMSTART
4266; GFX900-NEXT:    ; def v[0:5]
4267; GFX900-NEXT:    ;;#ASMEND
4268; GFX900-NEXT:    global_store_dwordx2 v6, v[4:5], s[16:17] offset:16
4269; GFX900-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
4270; GFX900-NEXT:    s_waitcnt vmcnt(0)
4271; GFX900-NEXT:    s_setpc_b64 s[30:31]
4272;
4273; GFX90A-LABEL: v_shuffle_v3i64_v3i64__4_5_5:
4274; GFX90A:       ; %bb.0:
4275; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4276; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
4277; GFX90A-NEXT:    ;;#ASMSTART
4278; GFX90A-NEXT:    ; def v[0:5]
4279; GFX90A-NEXT:    ;;#ASMEND
4280; GFX90A-NEXT:    global_store_dwordx2 v6, v[4:5], s[16:17] offset:16
4281; GFX90A-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
4282; GFX90A-NEXT:    s_waitcnt vmcnt(0)
4283; GFX90A-NEXT:    s_setpc_b64 s[30:31]
4284;
4285; GFX940-LABEL: v_shuffle_v3i64_v3i64__4_5_5:
4286; GFX940:       ; %bb.0:
4287; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4288; GFX940-NEXT:    v_mov_b32_e32 v6, 0
4289; GFX940-NEXT:    ;;#ASMSTART
4290; GFX940-NEXT:    ; def v[0:5]
4291; GFX940-NEXT:    ;;#ASMEND
4292; GFX940-NEXT:    global_store_dwordx2 v6, v[4:5], s[0:1] offset:16 sc0 sc1
4293; GFX940-NEXT:    global_store_dwordx4 v6, v[2:5], s[0:1] sc0 sc1
4294; GFX940-NEXT:    s_waitcnt vmcnt(0)
4295; GFX940-NEXT:    s_setpc_b64 s[30:31]
4296  %vec0 = call <3 x i64> asm "; def $0", "=v"()
4297  %vec1 = call <3 x i64> asm "; def $0", "=v"()
4298  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 4, i32 5, i32 5>
4299  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
4300  ret void
4301}
4302
4303define void @v_shuffle_v3i64_v3i64__5_u_5(ptr addrspace(1) inreg %ptr) {
4304; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_u_5:
4305; GFX900:       ; %bb.0:
4306; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4307; GFX900-NEXT:    ;;#ASMSTART
4308; GFX900-NEXT:    ; def v[0:5]
4309; GFX900-NEXT:    ;;#ASMEND
4310; GFX900-NEXT:    v_mov_b32_e32 v6, 0
4311; GFX900-NEXT:    v_mov_b32_e32 v0, v4
4312; GFX900-NEXT:    v_mov_b32_e32 v1, v5
4313; GFX900-NEXT:    global_store_dwordx2 v6, v[4:5], s[16:17] offset:16
4314; GFX900-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
4315; GFX900-NEXT:    s_waitcnt vmcnt(0)
4316; GFX900-NEXT:    s_setpc_b64 s[30:31]
4317;
4318; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_u_5:
4319; GFX90A:       ; %bb.0:
4320; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4321; GFX90A-NEXT:    ;;#ASMSTART
4322; GFX90A-NEXT:    ; def v[0:5]
4323; GFX90A-NEXT:    ;;#ASMEND
4324; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
4325; GFX90A-NEXT:    v_mov_b32_e32 v0, v4
4326; GFX90A-NEXT:    v_mov_b32_e32 v1, v5
4327; GFX90A-NEXT:    global_store_dwordx2 v6, v[4:5], s[16:17] offset:16
4328; GFX90A-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
4329; GFX90A-NEXT:    s_waitcnt vmcnt(0)
4330; GFX90A-NEXT:    s_setpc_b64 s[30:31]
4331;
4332; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_u_5:
4333; GFX940:       ; %bb.0:
4334; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4335; GFX940-NEXT:    ;;#ASMSTART
4336; GFX940-NEXT:    ; def v[0:5]
4337; GFX940-NEXT:    ;;#ASMEND
4338; GFX940-NEXT:    v_mov_b32_e32 v6, 0
4339; GFX940-NEXT:    v_mov_b32_e32 v0, v4
4340; GFX940-NEXT:    v_mov_b32_e32 v1, v5
4341; GFX940-NEXT:    global_store_dwordx2 v6, v[4:5], s[0:1] offset:16 sc0 sc1
4342; GFX940-NEXT:    global_store_dwordx4 v6, v[0:3], s[0:1] sc0 sc1
4343; GFX940-NEXT:    s_waitcnt vmcnt(0)
4344; GFX940-NEXT:    s_setpc_b64 s[30:31]
4345  %vec0 = call <3 x i64> asm "; def $0", "=v"()
4346  %vec1 = call <3 x i64> asm "; def $0", "=v"()
4347  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 poison, i32 5>
4348  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
4349  ret void
4350}
4351
4352define void @v_shuffle_v3i64_v3i64__5_0_5(ptr addrspace(1) inreg %ptr) {
4353; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_0_5:
4354; GFX900:       ; %bb.0:
4355; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4356; GFX900-NEXT:    ;;#ASMSTART
4357; GFX900-NEXT:    ; def v[0:5]
4358; GFX900-NEXT:    ;;#ASMEND
4359; GFX900-NEXT:    ;;#ASMSTART
4360; GFX900-NEXT:    ; def v[2:7]
4361; GFX900-NEXT:    ;;#ASMEND
4362; GFX900-NEXT:    v_mov_b32_e32 v8, 0
4363; GFX900-NEXT:    v_mov_b32_e32 v2, v6
4364; GFX900-NEXT:    v_mov_b32_e32 v3, v7
4365; GFX900-NEXT:    v_mov_b32_e32 v4, v0
4366; GFX900-NEXT:    v_mov_b32_e32 v5, v1
4367; GFX900-NEXT:    global_store_dwordx2 v8, v[6:7], s[16:17] offset:16
4368; GFX900-NEXT:    global_store_dwordx4 v8, v[2:5], s[16:17]
4369; GFX900-NEXT:    s_waitcnt vmcnt(0)
4370; GFX900-NEXT:    s_setpc_b64 s[30:31]
4371;
4372; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_0_5:
4373; GFX90A:       ; %bb.0:
4374; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4375; GFX90A-NEXT:    ;;#ASMSTART
4376; GFX90A-NEXT:    ; def v[0:5]
4377; GFX90A-NEXT:    ;;#ASMEND
4378; GFX90A-NEXT:    ;;#ASMSTART
4379; GFX90A-NEXT:    ; def v[2:7]
4380; GFX90A-NEXT:    ;;#ASMEND
4381; GFX90A-NEXT:    v_mov_b32_e32 v8, 0
4382; GFX90A-NEXT:    v_mov_b32_e32 v2, v6
4383; GFX90A-NEXT:    v_mov_b32_e32 v3, v7
4384; GFX90A-NEXT:    v_mov_b32_e32 v4, v0
4385; GFX90A-NEXT:    v_mov_b32_e32 v5, v1
4386; GFX90A-NEXT:    global_store_dwordx2 v8, v[6:7], s[16:17] offset:16
4387; GFX90A-NEXT:    global_store_dwordx4 v8, v[2:5], s[16:17]
4388; GFX90A-NEXT:    s_waitcnt vmcnt(0)
4389; GFX90A-NEXT:    s_setpc_b64 s[30:31]
4390;
4391; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_0_5:
4392; GFX940:       ; %bb.0:
4393; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4394; GFX940-NEXT:    ;;#ASMSTART
4395; GFX940-NEXT:    ; def v[0:5]
4396; GFX940-NEXT:    ;;#ASMEND
4397; GFX940-NEXT:    v_mov_b32_e32 v8, 0
4398; GFX940-NEXT:    ;;#ASMSTART
4399; GFX940-NEXT:    ; def v[2:7]
4400; GFX940-NEXT:    ;;#ASMEND
4401; GFX940-NEXT:    global_store_dwordx2 v8, v[6:7], s[0:1] offset:16 sc0 sc1
4402; GFX940-NEXT:    v_mov_b32_e32 v2, v6
4403; GFX940-NEXT:    v_mov_b32_e32 v3, v7
4404; GFX940-NEXT:    v_mov_b32_e32 v4, v0
4405; GFX940-NEXT:    v_mov_b32_e32 v5, v1
4406; GFX940-NEXT:    global_store_dwordx4 v8, v[2:5], s[0:1] sc0 sc1
4407; GFX940-NEXT:    s_waitcnt vmcnt(0)
4408; GFX940-NEXT:    s_setpc_b64 s[30:31]
4409  %vec0 = call <3 x i64> asm "; def $0", "=v"()
4410  %vec1 = call <3 x i64> asm "; def $0", "=v"()
4411  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 0, i32 5>
4412  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
4413  ret void
4414}
4415
4416define void @v_shuffle_v3i64_v3i64__5_1_5(ptr addrspace(1) inreg %ptr) {
4417; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_1_5:
4418; GFX900:       ; %bb.0:
4419; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4420; GFX900-NEXT:    ;;#ASMSTART
4421; GFX900-NEXT:    ; def v[0:5]
4422; GFX900-NEXT:    ;;#ASMEND
4423; GFX900-NEXT:    v_mov_b32_e32 v10, 0
4424; GFX900-NEXT:    ;;#ASMSTART
4425; GFX900-NEXT:    ; def v[4:9]
4426; GFX900-NEXT:    ;;#ASMEND
4427; GFX900-NEXT:    v_mov_b32_e32 v0, v8
4428; GFX900-NEXT:    v_mov_b32_e32 v1, v9
4429; GFX900-NEXT:    global_store_dwordx2 v10, v[8:9], s[16:17] offset:16
4430; GFX900-NEXT:    global_store_dwordx4 v10, v[0:3], s[16:17]
4431; GFX900-NEXT:    s_waitcnt vmcnt(0)
4432; GFX900-NEXT:    s_setpc_b64 s[30:31]
4433;
4434; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_1_5:
4435; GFX90A:       ; %bb.0:
4436; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4437; GFX90A-NEXT:    ;;#ASMSTART
4438; GFX90A-NEXT:    ; def v[0:5]
4439; GFX90A-NEXT:    ;;#ASMEND
4440; GFX90A-NEXT:    v_mov_b32_e32 v10, 0
4441; GFX90A-NEXT:    ;;#ASMSTART
4442; GFX90A-NEXT:    ; def v[4:9]
4443; GFX90A-NEXT:    ;;#ASMEND
4444; GFX90A-NEXT:    v_mov_b32_e32 v0, v8
4445; GFX90A-NEXT:    v_mov_b32_e32 v1, v9
4446; GFX90A-NEXT:    global_store_dwordx2 v10, v[8:9], s[16:17] offset:16
4447; GFX90A-NEXT:    global_store_dwordx4 v10, v[0:3], s[16:17]
4448; GFX90A-NEXT:    s_waitcnt vmcnt(0)
4449; GFX90A-NEXT:    s_setpc_b64 s[30:31]
4450;
4451; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_1_5:
4452; GFX940:       ; %bb.0:
4453; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4454; GFX940-NEXT:    ;;#ASMSTART
4455; GFX940-NEXT:    ; def v[0:5]
4456; GFX940-NEXT:    ;;#ASMEND
4457; GFX940-NEXT:    v_mov_b32_e32 v10, 0
4458; GFX940-NEXT:    ;;#ASMSTART
4459; GFX940-NEXT:    ; def v[4:9]
4460; GFX940-NEXT:    ;;#ASMEND
4461; GFX940-NEXT:    global_store_dwordx2 v10, v[8:9], s[0:1] offset:16 sc0 sc1
4462; GFX940-NEXT:    v_mov_b32_e32 v0, v8
4463; GFX940-NEXT:    v_mov_b32_e32 v1, v9
4464; GFX940-NEXT:    global_store_dwordx4 v10, v[0:3], s[0:1] sc0 sc1
4465; GFX940-NEXT:    s_waitcnt vmcnt(0)
4466; GFX940-NEXT:    s_setpc_b64 s[30:31]
4467  %vec0 = call <3 x i64> asm "; def $0", "=v"()
4468  %vec1 = call <3 x i64> asm "; def $0", "=v"()
4469  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 1, i32 5>
4470  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
4471  ret void
4472}
4473
4474define void @v_shuffle_v3i64_v3i64__5_2_5(ptr addrspace(1) inreg %ptr) {
4475; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_2_5:
4476; GFX900:       ; %bb.0:
4477; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4478; GFX900-NEXT:    ;;#ASMSTART
4479; GFX900-NEXT:    ; def v[0:5]
4480; GFX900-NEXT:    ;;#ASMEND
4481; GFX900-NEXT:    v_mov_b32_e32 v12, 0
4482; GFX900-NEXT:    ;;#ASMSTART
4483; GFX900-NEXT:    ; def v[6:11]
4484; GFX900-NEXT:    ;;#ASMEND
4485; GFX900-NEXT:    v_mov_b32_e32 v2, v10
4486; GFX900-NEXT:    v_mov_b32_e32 v3, v11
4487; GFX900-NEXT:    global_store_dwordx2 v12, v[10:11], s[16:17] offset:16
4488; GFX900-NEXT:    global_store_dwordx4 v12, v[2:5], s[16:17]
4489; GFX900-NEXT:    s_waitcnt vmcnt(0)
4490; GFX900-NEXT:    s_setpc_b64 s[30:31]
4491;
4492; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_2_5:
4493; GFX90A:       ; %bb.0:
4494; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4495; GFX90A-NEXT:    ;;#ASMSTART
4496; GFX90A-NEXT:    ; def v[0:5]
4497; GFX90A-NEXT:    ;;#ASMEND
4498; GFX90A-NEXT:    v_mov_b32_e32 v12, 0
4499; GFX90A-NEXT:    ;;#ASMSTART
4500; GFX90A-NEXT:    ; def v[6:11]
4501; GFX90A-NEXT:    ;;#ASMEND
4502; GFX90A-NEXT:    v_mov_b32_e32 v2, v10
4503; GFX90A-NEXT:    v_mov_b32_e32 v3, v11
4504; GFX90A-NEXT:    global_store_dwordx2 v12, v[10:11], s[16:17] offset:16
4505; GFX90A-NEXT:    global_store_dwordx4 v12, v[2:5], s[16:17]
4506; GFX90A-NEXT:    s_waitcnt vmcnt(0)
4507; GFX90A-NEXT:    s_setpc_b64 s[30:31]
4508;
4509; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_2_5:
4510; GFX940:       ; %bb.0:
4511; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4512; GFX940-NEXT:    ;;#ASMSTART
4513; GFX940-NEXT:    ; def v[0:5]
4514; GFX940-NEXT:    ;;#ASMEND
4515; GFX940-NEXT:    v_mov_b32_e32 v12, 0
4516; GFX940-NEXT:    ;;#ASMSTART
4517; GFX940-NEXT:    ; def v[6:11]
4518; GFX940-NEXT:    ;;#ASMEND
4519; GFX940-NEXT:    global_store_dwordx2 v12, v[10:11], s[0:1] offset:16 sc0 sc1
4520; GFX940-NEXT:    v_mov_b32_e32 v2, v10
4521; GFX940-NEXT:    v_mov_b32_e32 v3, v11
4522; GFX940-NEXT:    global_store_dwordx4 v12, v[2:5], s[0:1] sc0 sc1
4523; GFX940-NEXT:    s_waitcnt vmcnt(0)
4524; GFX940-NEXT:    s_setpc_b64 s[30:31]
4525  %vec0 = call <3 x i64> asm "; def $0", "=v"()
4526  %vec1 = call <3 x i64> asm "; def $0", "=v"()
4527  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 2, i32 5>
4528  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
4529  ret void
4530}
4531
4532define void @v_shuffle_v3i64_v3i64__5_3_5(ptr addrspace(1) inreg %ptr) {
4533; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_3_5:
4534; GFX900:       ; %bb.0:
4535; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4536; GFX900-NEXT:    v_mov_b32_e32 v6, 0
4537; GFX900-NEXT:    ;;#ASMSTART
4538; GFX900-NEXT:    ; def v[0:5]
4539; GFX900-NEXT:    ;;#ASMEND
4540; GFX900-NEXT:    global_store_dwordx2 v6, v[4:5], s[16:17] offset:16
4541; GFX900-NEXT:    v_mov_b32_e32 v2, v4
4542; GFX900-NEXT:    v_mov_b32_e32 v3, v5
4543; GFX900-NEXT:    v_mov_b32_e32 v4, v0
4544; GFX900-NEXT:    v_mov_b32_e32 v5, v1
4545; GFX900-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
4546; GFX900-NEXT:    s_waitcnt vmcnt(0)
4547; GFX900-NEXT:    s_setpc_b64 s[30:31]
4548;
4549; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_3_5:
4550; GFX90A:       ; %bb.0:
4551; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4552; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
4553; GFX90A-NEXT:    ;;#ASMSTART
4554; GFX90A-NEXT:    ; def v[0:5]
4555; GFX90A-NEXT:    ;;#ASMEND
4556; GFX90A-NEXT:    global_store_dwordx2 v6, v[4:5], s[16:17] offset:16
4557; GFX90A-NEXT:    v_mov_b32_e32 v2, v4
4558; GFX90A-NEXT:    v_mov_b32_e32 v3, v5
4559; GFX90A-NEXT:    v_mov_b32_e32 v4, v0
4560; GFX90A-NEXT:    v_mov_b32_e32 v5, v1
4561; GFX90A-NEXT:    global_store_dwordx4 v6, v[2:5], s[16:17]
4562; GFX90A-NEXT:    s_waitcnt vmcnt(0)
4563; GFX90A-NEXT:    s_setpc_b64 s[30:31]
4564;
4565; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_3_5:
4566; GFX940:       ; %bb.0:
4567; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4568; GFX940-NEXT:    v_mov_b32_e32 v6, 0
4569; GFX940-NEXT:    ;;#ASMSTART
4570; GFX940-NEXT:    ; def v[0:5]
4571; GFX940-NEXT:    ;;#ASMEND
4572; GFX940-NEXT:    global_store_dwordx2 v6, v[4:5], s[0:1] offset:16 sc0 sc1
4573; GFX940-NEXT:    v_mov_b32_e32 v2, v4
4574; GFX940-NEXT:    v_mov_b32_e32 v3, v5
4575; GFX940-NEXT:    v_mov_b32_e32 v4, v0
4576; GFX940-NEXT:    v_mov_b32_e32 v5, v1
4577; GFX940-NEXT:    global_store_dwordx4 v6, v[2:5], s[0:1] sc0 sc1
4578; GFX940-NEXT:    s_waitcnt vmcnt(0)
4579; GFX940-NEXT:    s_setpc_b64 s[30:31]
4580  %vec0 = call <3 x i64> asm "; def $0", "=v"()
4581  %vec1 = call <3 x i64> asm "; def $0", "=v"()
4582  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 3, i32 5>
4583  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
4584  ret void
4585}
4586
4587define void @v_shuffle_v3i64_v3i64__5_4_5(ptr addrspace(1) inreg %ptr) {
4588; GFX900-LABEL: v_shuffle_v3i64_v3i64__5_4_5:
4589; GFX900:       ; %bb.0:
4590; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4591; GFX900-NEXT:    ;;#ASMSTART
4592; GFX900-NEXT:    ; def v[0:5]
4593; GFX900-NEXT:    ;;#ASMEND
4594; GFX900-NEXT:    v_mov_b32_e32 v6, 0
4595; GFX900-NEXT:    v_mov_b32_e32 v0, v4
4596; GFX900-NEXT:    v_mov_b32_e32 v1, v5
4597; GFX900-NEXT:    global_store_dwordx2 v6, v[4:5], s[16:17] offset:16
4598; GFX900-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
4599; GFX900-NEXT:    s_waitcnt vmcnt(0)
4600; GFX900-NEXT:    s_setpc_b64 s[30:31]
4601;
4602; GFX90A-LABEL: v_shuffle_v3i64_v3i64__5_4_5:
4603; GFX90A:       ; %bb.0:
4604; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4605; GFX90A-NEXT:    ;;#ASMSTART
4606; GFX90A-NEXT:    ; def v[0:5]
4607; GFX90A-NEXT:    ;;#ASMEND
4608; GFX90A-NEXT:    v_mov_b32_e32 v6, 0
4609; GFX90A-NEXT:    v_mov_b32_e32 v0, v4
4610; GFX90A-NEXT:    v_mov_b32_e32 v1, v5
4611; GFX90A-NEXT:    global_store_dwordx2 v6, v[4:5], s[16:17] offset:16
4612; GFX90A-NEXT:    global_store_dwordx4 v6, v[0:3], s[16:17]
4613; GFX90A-NEXT:    s_waitcnt vmcnt(0)
4614; GFX90A-NEXT:    s_setpc_b64 s[30:31]
4615;
4616; GFX940-LABEL: v_shuffle_v3i64_v3i64__5_4_5:
4617; GFX940:       ; %bb.0:
4618; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4619; GFX940-NEXT:    ;;#ASMSTART
4620; GFX940-NEXT:    ; def v[0:5]
4621; GFX940-NEXT:    ;;#ASMEND
4622; GFX940-NEXT:    v_mov_b32_e32 v6, 0
4623; GFX940-NEXT:    v_mov_b32_e32 v0, v4
4624; GFX940-NEXT:    v_mov_b32_e32 v1, v5
4625; GFX940-NEXT:    global_store_dwordx2 v6, v[4:5], s[0:1] offset:16 sc0 sc1
4626; GFX940-NEXT:    global_store_dwordx4 v6, v[0:3], s[0:1] sc0 sc1
4627; GFX940-NEXT:    s_waitcnt vmcnt(0)
4628; GFX940-NEXT:    s_setpc_b64 s[30:31]
4629  %vec0 = call <3 x i64> asm "; def $0", "=v"()
4630  %vec1 = call <3 x i64> asm "; def $0", "=v"()
4631  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 4, i32 5>
4632  store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
4633  ret void
4634}
4635
4636define void @s_shuffle_v3i64_v3i64__u_u_u() {
4637; GFX9-LABEL: s_shuffle_v3i64_v3i64__u_u_u:
4638; GFX9:       ; %bb.0:
4639; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4640; GFX9-NEXT:    ;;#ASMSTART
4641; GFX9-NEXT:    ; use s[8:13]
4642; GFX9-NEXT:    ;;#ASMEND
4643; GFX9-NEXT:    s_setpc_b64 s[30:31]
4644  %vec0 = call <3 x i64> asm "; def $0", "=s"()
4645  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> poison
4646  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
4647  ret void
4648}
4649
4650define void @s_shuffle_v3i64_v3i64__0_u_u() {
4651; GFX900-LABEL: s_shuffle_v3i64_v3i64__0_u_u:
4652; GFX900:       ; %bb.0:
4653; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4654; GFX900-NEXT:    ;;#ASMSTART
4655; GFX900-NEXT:    ; def s[8:13]
4656; GFX900-NEXT:    ;;#ASMEND
4657; GFX900-NEXT:    ;;#ASMSTART
4658; GFX900-NEXT:    ; use s[8:13]
4659; GFX900-NEXT:    ;;#ASMEND
4660; GFX900-NEXT:    s_setpc_b64 s[30:31]
4661;
4662; GFX90A-LABEL: s_shuffle_v3i64_v3i64__0_u_u:
4663; GFX90A:       ; %bb.0:
4664; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4665; GFX90A-NEXT:    ;;#ASMSTART
4666; GFX90A-NEXT:    ; def s[8:13]
4667; GFX90A-NEXT:    ;;#ASMEND
4668; GFX90A-NEXT:    ;;#ASMSTART
4669; GFX90A-NEXT:    ; use s[8:13]
4670; GFX90A-NEXT:    ;;#ASMEND
4671; GFX90A-NEXT:    s_setpc_b64 s[30:31]
4672;
4673; GFX940-LABEL: s_shuffle_v3i64_v3i64__0_u_u:
4674; GFX940:       ; %bb.0:
4675; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4676; GFX940-NEXT:    ;;#ASMSTART
4677; GFX940-NEXT:    ; def s[8:13]
4678; GFX940-NEXT:    ;;#ASMEND
4679; GFX940-NEXT:    s_nop 0
4680; GFX940-NEXT:    ;;#ASMSTART
4681; GFX940-NEXT:    ; use s[8:13]
4682; GFX940-NEXT:    ;;#ASMEND
4683; GFX940-NEXT:    s_setpc_b64 s[30:31]
4684  %vec0 = call <3 x i64> asm "; def $0", "=s"()
4685  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 0, i32 poison, i32 poison>
4686  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
4687  ret void
4688}
4689
4690define void @s_shuffle_v3i64_v3i64__1_u_u() {
4691; GFX900-LABEL: s_shuffle_v3i64_v3i64__1_u_u:
4692; GFX900:       ; %bb.0:
4693; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4694; GFX900-NEXT:    ;;#ASMSTART
4695; GFX900-NEXT:    ; def s[4:9]
4696; GFX900-NEXT:    ;;#ASMEND
4697; GFX900-NEXT:    s_mov_b32 s8, s6
4698; GFX900-NEXT:    s_mov_b32 s9, s7
4699; GFX900-NEXT:    ;;#ASMSTART
4700; GFX900-NEXT:    ; use s[8:13]
4701; GFX900-NEXT:    ;;#ASMEND
4702; GFX900-NEXT:    s_setpc_b64 s[30:31]
4703;
4704; GFX90A-LABEL: s_shuffle_v3i64_v3i64__1_u_u:
4705; GFX90A:       ; %bb.0:
4706; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4707; GFX90A-NEXT:    ;;#ASMSTART
4708; GFX90A-NEXT:    ; def s[4:9]
4709; GFX90A-NEXT:    ;;#ASMEND
4710; GFX90A-NEXT:    s_mov_b32 s8, s6
4711; GFX90A-NEXT:    s_mov_b32 s9, s7
4712; GFX90A-NEXT:    ;;#ASMSTART
4713; GFX90A-NEXT:    ; use s[8:13]
4714; GFX90A-NEXT:    ;;#ASMEND
4715; GFX90A-NEXT:    s_setpc_b64 s[30:31]
4716;
4717; GFX940-LABEL: s_shuffle_v3i64_v3i64__1_u_u:
4718; GFX940:       ; %bb.0:
4719; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4720; GFX940-NEXT:    ;;#ASMSTART
4721; GFX940-NEXT:    ; def s[0:5]
4722; GFX940-NEXT:    ;;#ASMEND
4723; GFX940-NEXT:    s_mov_b32 s8, s2
4724; GFX940-NEXT:    s_mov_b32 s9, s3
4725; GFX940-NEXT:    ;;#ASMSTART
4726; GFX940-NEXT:    ; use s[8:13]
4727; GFX940-NEXT:    ;;#ASMEND
4728; GFX940-NEXT:    s_setpc_b64 s[30:31]
4729  %vec0 = call <3 x i64> asm "; def $0", "=s"()
4730  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 1, i32 poison, i32 poison>
4731  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
4732  ret void
4733}
4734
4735define void @s_shuffle_v3i64_v3i64__2_u_u() {
4736; GFX900-LABEL: s_shuffle_v3i64_v3i64__2_u_u:
4737; GFX900:       ; %bb.0:
4738; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4739; GFX900-NEXT:    ;;#ASMSTART
4740; GFX900-NEXT:    ; def s[4:9]
4741; GFX900-NEXT:    ;;#ASMEND
4742; GFX900-NEXT:    ;;#ASMSTART
4743; GFX900-NEXT:    ; use s[8:13]
4744; GFX900-NEXT:    ;;#ASMEND
4745; GFX900-NEXT:    s_setpc_b64 s[30:31]
4746;
4747; GFX90A-LABEL: s_shuffle_v3i64_v3i64__2_u_u:
4748; GFX90A:       ; %bb.0:
4749; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4750; GFX90A-NEXT:    ;;#ASMSTART
4751; GFX90A-NEXT:    ; def s[4:9]
4752; GFX90A-NEXT:    ;;#ASMEND
4753; GFX90A-NEXT:    ;;#ASMSTART
4754; GFX90A-NEXT:    ; use s[8:13]
4755; GFX90A-NEXT:    ;;#ASMEND
4756; GFX90A-NEXT:    s_setpc_b64 s[30:31]
4757;
4758; GFX940-LABEL: s_shuffle_v3i64_v3i64__2_u_u:
4759; GFX940:       ; %bb.0:
4760; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4761; GFX940-NEXT:    ;;#ASMSTART
4762; GFX940-NEXT:    ; def s[0:5]
4763; GFX940-NEXT:    ;;#ASMEND
4764; GFX940-NEXT:    s_mov_b32 s8, s4
4765; GFX940-NEXT:    s_mov_b32 s9, s5
4766; GFX940-NEXT:    ;;#ASMSTART
4767; GFX940-NEXT:    ; use s[8:13]
4768; GFX940-NEXT:    ;;#ASMEND
4769; GFX940-NEXT:    s_setpc_b64 s[30:31]
4770  %vec0 = call <3 x i64> asm "; def $0", "=s"()
4771  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 2, i32 poison, i32 poison>
4772  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
4773  ret void
4774}
4775
4776define void @s_shuffle_v3i64_v3i64__3_u_u() {
4777; GFX9-LABEL: s_shuffle_v3i64_v3i64__3_u_u:
4778; GFX9:       ; %bb.0:
4779; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4780; GFX9-NEXT:    ;;#ASMSTART
4781; GFX9-NEXT:    ; use s[8:13]
4782; GFX9-NEXT:    ;;#ASMEND
4783; GFX9-NEXT:    s_setpc_b64 s[30:31]
4784  %vec0 = call <3 x i64> asm "; def $0", "=s"()
4785  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 3, i32 poison, i32 poison>
4786  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
4787  ret void
4788}
4789
4790define void @s_shuffle_v3i64_v3i64__4_u_u() {
4791; GFX900-LABEL: s_shuffle_v3i64_v3i64__4_u_u:
4792; GFX900:       ; %bb.0:
4793; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4794; GFX900-NEXT:    ;;#ASMSTART
4795; GFX900-NEXT:    ; def s[4:9]
4796; GFX900-NEXT:    ;;#ASMEND
4797; GFX900-NEXT:    s_mov_b32 s8, s6
4798; GFX900-NEXT:    s_mov_b32 s9, s7
4799; GFX900-NEXT:    ;;#ASMSTART
4800; GFX900-NEXT:    ; use s[8:13]
4801; GFX900-NEXT:    ;;#ASMEND
4802; GFX900-NEXT:    s_setpc_b64 s[30:31]
4803;
4804; GFX90A-LABEL: s_shuffle_v3i64_v3i64__4_u_u:
4805; GFX90A:       ; %bb.0:
4806; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4807; GFX90A-NEXT:    ;;#ASMSTART
4808; GFX90A-NEXT:    ; def s[4:9]
4809; GFX90A-NEXT:    ;;#ASMEND
4810; GFX90A-NEXT:    s_mov_b32 s8, s6
4811; GFX90A-NEXT:    s_mov_b32 s9, s7
4812; GFX90A-NEXT:    ;;#ASMSTART
4813; GFX90A-NEXT:    ; use s[8:13]
4814; GFX90A-NEXT:    ;;#ASMEND
4815; GFX90A-NEXT:    s_setpc_b64 s[30:31]
4816;
4817; GFX940-LABEL: s_shuffle_v3i64_v3i64__4_u_u:
4818; GFX940:       ; %bb.0:
4819; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4820; GFX940-NEXT:    ;;#ASMSTART
4821; GFX940-NEXT:    ; def s[0:5]
4822; GFX940-NEXT:    ;;#ASMEND
4823; GFX940-NEXT:    s_mov_b32 s8, s2
4824; GFX940-NEXT:    s_mov_b32 s9, s3
4825; GFX940-NEXT:    ;;#ASMSTART
4826; GFX940-NEXT:    ; use s[8:13]
4827; GFX940-NEXT:    ;;#ASMEND
4828; GFX940-NEXT:    s_setpc_b64 s[30:31]
4829  %vec0 = call <3 x i64> asm "; def $0", "=s"()
4830  %vec1 = call <3 x i64> asm "; def $0", "=s"()
4831  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 4, i32 poison, i32 poison>
4832  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
4833  ret void
4834}
4835
4836define void @s_shuffle_v3i64_v3i64__5_u_u() {
4837; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_u_u:
4838; GFX900:       ; %bb.0:
4839; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4840; GFX900-NEXT:    ;;#ASMSTART
4841; GFX900-NEXT:    ; def s[4:9]
4842; GFX900-NEXT:    ;;#ASMEND
4843; GFX900-NEXT:    ;;#ASMSTART
4844; GFX900-NEXT:    ; use s[8:13]
4845; GFX900-NEXT:    ;;#ASMEND
4846; GFX900-NEXT:    s_setpc_b64 s[30:31]
4847;
4848; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_u_u:
4849; GFX90A:       ; %bb.0:
4850; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4851; GFX90A-NEXT:    ;;#ASMSTART
4852; GFX90A-NEXT:    ; def s[4:9]
4853; GFX90A-NEXT:    ;;#ASMEND
4854; GFX90A-NEXT:    ;;#ASMSTART
4855; GFX90A-NEXT:    ; use s[8:13]
4856; GFX90A-NEXT:    ;;#ASMEND
4857; GFX90A-NEXT:    s_setpc_b64 s[30:31]
4858;
4859; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_u_u:
4860; GFX940:       ; %bb.0:
4861; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4862; GFX940-NEXT:    ;;#ASMSTART
4863; GFX940-NEXT:    ; def s[0:5]
4864; GFX940-NEXT:    ;;#ASMEND
4865; GFX940-NEXT:    s_mov_b32 s8, s4
4866; GFX940-NEXT:    s_mov_b32 s9, s5
4867; GFX940-NEXT:    ;;#ASMSTART
4868; GFX940-NEXT:    ; use s[8:13]
4869; GFX940-NEXT:    ;;#ASMEND
4870; GFX940-NEXT:    s_setpc_b64 s[30:31]
4871  %vec0 = call <3 x i64> asm "; def $0", "=s"()
4872  %vec1 = call <3 x i64> asm "; def $0", "=s"()
4873  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 poison, i32 poison>
4874  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
4875  ret void
4876}
4877
4878define void @s_shuffle_v3i64_v3i64__5_0_u() {
4879; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_0_u:
4880; GFX900:       ; %bb.0:
4881; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4882; GFX900-NEXT:    ;;#ASMSTART
4883; GFX900-NEXT:    ; def s[4:9]
4884; GFX900-NEXT:    ;;#ASMEND
4885; GFX900-NEXT:    ;;#ASMSTART
4886; GFX900-NEXT:    ; def s[8:13]
4887; GFX900-NEXT:    ;;#ASMEND
4888; GFX900-NEXT:    s_mov_b32 s8, s12
4889; GFX900-NEXT:    s_mov_b32 s9, s13
4890; GFX900-NEXT:    s_mov_b32 s10, s4
4891; GFX900-NEXT:    s_mov_b32 s11, s5
4892; GFX900-NEXT:    ;;#ASMSTART
4893; GFX900-NEXT:    ; use s[8:13]
4894; GFX900-NEXT:    ;;#ASMEND
4895; GFX900-NEXT:    s_setpc_b64 s[30:31]
4896;
4897; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_0_u:
4898; GFX90A:       ; %bb.0:
4899; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4900; GFX90A-NEXT:    ;;#ASMSTART
4901; GFX90A-NEXT:    ; def s[4:9]
4902; GFX90A-NEXT:    ;;#ASMEND
4903; GFX90A-NEXT:    ;;#ASMSTART
4904; GFX90A-NEXT:    ; def s[8:13]
4905; GFX90A-NEXT:    ;;#ASMEND
4906; GFX90A-NEXT:    s_mov_b32 s8, s12
4907; GFX90A-NEXT:    s_mov_b32 s9, s13
4908; GFX90A-NEXT:    s_mov_b32 s10, s4
4909; GFX90A-NEXT:    s_mov_b32 s11, s5
4910; GFX90A-NEXT:    ;;#ASMSTART
4911; GFX90A-NEXT:    ; use s[8:13]
4912; GFX90A-NEXT:    ;;#ASMEND
4913; GFX90A-NEXT:    s_setpc_b64 s[30:31]
4914;
4915; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_0_u:
4916; GFX940:       ; %bb.0:
4917; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4918; GFX940-NEXT:    ;;#ASMSTART
4919; GFX940-NEXT:    ; def s[0:5]
4920; GFX940-NEXT:    ;;#ASMEND
4921; GFX940-NEXT:    s_mov_b32 s10, s0
4922; GFX940-NEXT:    ;;#ASMSTART
4923; GFX940-NEXT:    ; def s[4:9]
4924; GFX940-NEXT:    ;;#ASMEND
4925; GFX940-NEXT:    s_mov_b32 s11, s1
4926; GFX940-NEXT:    ;;#ASMSTART
4927; GFX940-NEXT:    ; use s[8:13]
4928; GFX940-NEXT:    ;;#ASMEND
4929; GFX940-NEXT:    s_setpc_b64 s[30:31]
4930  %vec0 = call <3 x i64> asm "; def $0", "=s"()
4931  %vec1 = call <3 x i64> asm "; def $0", "=s"()
4932  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 0, i32 poison>
4933  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
4934  ret void
4935}
4936
4937define void @s_shuffle_v3i64_v3i64__5_1_u() {
4938; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_1_u:
4939; GFX900:       ; %bb.0:
4940; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4941; GFX900-NEXT:    ;;#ASMSTART
4942; GFX900-NEXT:    ; def s[8:13]
4943; GFX900-NEXT:    ;;#ASMEND
4944; GFX900-NEXT:    ;;#ASMSTART
4945; GFX900-NEXT:    ; def s[4:9]
4946; GFX900-NEXT:    ;;#ASMEND
4947; GFX900-NEXT:    ;;#ASMSTART
4948; GFX900-NEXT:    ; use s[8:13]
4949; GFX900-NEXT:    ;;#ASMEND
4950; GFX900-NEXT:    s_setpc_b64 s[30:31]
4951;
4952; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_1_u:
4953; GFX90A:       ; %bb.0:
4954; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4955; GFX90A-NEXT:    ;;#ASMSTART
4956; GFX90A-NEXT:    ; def s[8:13]
4957; GFX90A-NEXT:    ;;#ASMEND
4958; GFX90A-NEXT:    ;;#ASMSTART
4959; GFX90A-NEXT:    ; def s[4:9]
4960; GFX90A-NEXT:    ;;#ASMEND
4961; GFX90A-NEXT:    ;;#ASMSTART
4962; GFX90A-NEXT:    ; use s[8:13]
4963; GFX90A-NEXT:    ;;#ASMEND
4964; GFX90A-NEXT:    s_setpc_b64 s[30:31]
4965;
4966; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_1_u:
4967; GFX940:       ; %bb.0:
4968; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4969; GFX940-NEXT:    ;;#ASMSTART
4970; GFX940-NEXT:    ; def s[8:13]
4971; GFX940-NEXT:    ;;#ASMEND
4972; GFX940-NEXT:    ;;#ASMSTART
4973; GFX940-NEXT:    ; def s[0:5]
4974; GFX940-NEXT:    ;;#ASMEND
4975; GFX940-NEXT:    s_mov_b32 s8, s4
4976; GFX940-NEXT:    s_mov_b32 s9, s5
4977; GFX940-NEXT:    ;;#ASMSTART
4978; GFX940-NEXT:    ; use s[8:13]
4979; GFX940-NEXT:    ;;#ASMEND
4980; GFX940-NEXT:    s_setpc_b64 s[30:31]
4981  %vec0 = call <3 x i64> asm "; def $0", "=s"()
4982  %vec1 = call <3 x i64> asm "; def $0", "=s"()
4983  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 1, i32 poison>
4984  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
4985  ret void
4986}
4987
4988define void @s_shuffle_v3i64_v3i64__5_2_u() {
4989; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_2_u:
4990; GFX900:       ; %bb.0:
4991; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
4992; GFX900-NEXT:    ;;#ASMSTART
4993; GFX900-NEXT:    ; def s[8:13]
4994; GFX900-NEXT:    ;;#ASMEND
4995; GFX900-NEXT:    ;;#ASMSTART
4996; GFX900-NEXT:    ; def s[4:9]
4997; GFX900-NEXT:    ;;#ASMEND
4998; GFX900-NEXT:    s_mov_b32 s10, s12
4999; GFX900-NEXT:    s_mov_b32 s11, s13
5000; GFX900-NEXT:    ;;#ASMSTART
5001; GFX900-NEXT:    ; use s[8:13]
5002; GFX900-NEXT:    ;;#ASMEND
5003; GFX900-NEXT:    s_setpc_b64 s[30:31]
5004;
5005; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_2_u:
5006; GFX90A:       ; %bb.0:
5007; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5008; GFX90A-NEXT:    ;;#ASMSTART
5009; GFX90A-NEXT:    ; def s[8:13]
5010; GFX90A-NEXT:    ;;#ASMEND
5011; GFX90A-NEXT:    ;;#ASMSTART
5012; GFX90A-NEXT:    ; def s[4:9]
5013; GFX90A-NEXT:    ;;#ASMEND
5014; GFX90A-NEXT:    s_mov_b32 s10, s12
5015; GFX90A-NEXT:    s_mov_b32 s11, s13
5016; GFX90A-NEXT:    ;;#ASMSTART
5017; GFX90A-NEXT:    ; use s[8:13]
5018; GFX90A-NEXT:    ;;#ASMEND
5019; GFX90A-NEXT:    s_setpc_b64 s[30:31]
5020;
5021; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_2_u:
5022; GFX940:       ; %bb.0:
5023; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5024; GFX940-NEXT:    ;;#ASMSTART
5025; GFX940-NEXT:    ; def s[8:13]
5026; GFX940-NEXT:    ;;#ASMEND
5027; GFX940-NEXT:    ;;#ASMSTART
5028; GFX940-NEXT:    ; def s[0:5]
5029; GFX940-NEXT:    ;;#ASMEND
5030; GFX940-NEXT:    s_mov_b32 s8, s12
5031; GFX940-NEXT:    s_mov_b32 s9, s13
5032; GFX940-NEXT:    s_mov_b32 s10, s4
5033; GFX940-NEXT:    s_mov_b32 s11, s5
5034; GFX940-NEXT:    ;;#ASMSTART
5035; GFX940-NEXT:    ; use s[8:13]
5036; GFX940-NEXT:    ;;#ASMEND
5037; GFX940-NEXT:    s_setpc_b64 s[30:31]
5038  %vec0 = call <3 x i64> asm "; def $0", "=s"()
5039  %vec1 = call <3 x i64> asm "; def $0", "=s"()
5040  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 2, i32 poison>
5041  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
5042  ret void
5043}
5044
5045define void @s_shuffle_v3i64_v3i64__5_3_u() {
5046; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_3_u:
5047; GFX900:       ; %bb.0:
5048; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5049; GFX900-NEXT:    ;;#ASMSTART
5050; GFX900-NEXT:    ; def s[4:9]
5051; GFX900-NEXT:    ;;#ASMEND
5052; GFX900-NEXT:    s_mov_b32 s10, s4
5053; GFX900-NEXT:    s_mov_b32 s11, s5
5054; GFX900-NEXT:    ;;#ASMSTART
5055; GFX900-NEXT:    ; use s[8:13]
5056; GFX900-NEXT:    ;;#ASMEND
5057; GFX900-NEXT:    s_setpc_b64 s[30:31]
5058;
5059; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_3_u:
5060; GFX90A:       ; %bb.0:
5061; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5062; GFX90A-NEXT:    ;;#ASMSTART
5063; GFX90A-NEXT:    ; def s[4:9]
5064; GFX90A-NEXT:    ;;#ASMEND
5065; GFX90A-NEXT:    s_mov_b32 s10, s4
5066; GFX90A-NEXT:    s_mov_b32 s11, s5
5067; GFX90A-NEXT:    ;;#ASMSTART
5068; GFX90A-NEXT:    ; use s[8:13]
5069; GFX90A-NEXT:    ;;#ASMEND
5070; GFX90A-NEXT:    s_setpc_b64 s[30:31]
5071;
5072; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_3_u:
5073; GFX940:       ; %bb.0:
5074; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5075; GFX940-NEXT:    ;;#ASMSTART
5076; GFX940-NEXT:    ; def s[0:5]
5077; GFX940-NEXT:    ;;#ASMEND
5078; GFX940-NEXT:    s_mov_b32 s8, s4
5079; GFX940-NEXT:    s_mov_b32 s9, s5
5080; GFX940-NEXT:    s_mov_b32 s10, s0
5081; GFX940-NEXT:    s_mov_b32 s11, s1
5082; GFX940-NEXT:    ;;#ASMSTART
5083; GFX940-NEXT:    ; use s[8:13]
5084; GFX940-NEXT:    ;;#ASMEND
5085; GFX940-NEXT:    s_setpc_b64 s[30:31]
5086  %vec0 = call <3 x i64> asm "; def $0", "=s"()
5087  %vec1 = call <3 x i64> asm "; def $0", "=s"()
5088  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 3, i32 poison>
5089  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
5090  ret void
5091}
5092
5093define void @s_shuffle_v3i64_v3i64__5_4_u() {
5094; GFX9-LABEL: s_shuffle_v3i64_v3i64__5_4_u:
5095; GFX9:       ; %bb.0:
5096; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5097; GFX9-NEXT:    ;;#ASMSTART
5098; GFX9-NEXT:    ; def s[8:13]
5099; GFX9-NEXT:    ;;#ASMEND
5100; GFX9-NEXT:    s_mov_b32 s8, s12
5101; GFX9-NEXT:    s_mov_b32 s9, s13
5102; GFX9-NEXT:    ;;#ASMSTART
5103; GFX9-NEXT:    ; use s[8:13]
5104; GFX9-NEXT:    ;;#ASMEND
5105; GFX9-NEXT:    s_setpc_b64 s[30:31]
5106  %vec0 = call <3 x i64> asm "; def $0", "=s"()
5107  %vec1 = call <3 x i64> asm "; def $0", "=s"()
5108  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 4, i32 poison>
5109  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
5110  ret void
5111}
5112
5113define void @s_shuffle_v3i64_v3i64__5_5_u() {
5114; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_5_u:
5115; GFX900:       ; %bb.0:
5116; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5117; GFX900-NEXT:    ;;#ASMSTART
5118; GFX900-NEXT:    ; def s[8:13]
5119; GFX900-NEXT:    ;;#ASMEND
5120; GFX900-NEXT:    s_mov_b32 s8, s12
5121; GFX900-NEXT:    s_mov_b32 s9, s13
5122; GFX900-NEXT:    s_mov_b32 s10, s12
5123; GFX900-NEXT:    s_mov_b32 s11, s13
5124; GFX900-NEXT:    ;;#ASMSTART
5125; GFX900-NEXT:    ; use s[8:13]
5126; GFX900-NEXT:    ;;#ASMEND
5127; GFX900-NEXT:    s_setpc_b64 s[30:31]
5128;
5129; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_5_u:
5130; GFX90A:       ; %bb.0:
5131; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5132; GFX90A-NEXT:    ;;#ASMSTART
5133; GFX90A-NEXT:    ; def s[8:13]
5134; GFX90A-NEXT:    ;;#ASMEND
5135; GFX90A-NEXT:    s_mov_b32 s8, s12
5136; GFX90A-NEXT:    s_mov_b32 s9, s13
5137; GFX90A-NEXT:    s_mov_b32 s10, s12
5138; GFX90A-NEXT:    s_mov_b32 s11, s13
5139; GFX90A-NEXT:    ;;#ASMSTART
5140; GFX90A-NEXT:    ; use s[8:13]
5141; GFX90A-NEXT:    ;;#ASMEND
5142; GFX90A-NEXT:    s_setpc_b64 s[30:31]
5143;
5144; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_5_u:
5145; GFX940:       ; %bb.0:
5146; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5147; GFX940-NEXT:    ;;#ASMSTART
5148; GFX940-NEXT:    ; def s[0:5]
5149; GFX940-NEXT:    ;;#ASMEND
5150; GFX940-NEXT:    s_mov_b32 s8, s4
5151; GFX940-NEXT:    s_mov_b32 s9, s5
5152; GFX940-NEXT:    s_mov_b32 s10, s4
5153; GFX940-NEXT:    s_mov_b32 s11, s5
5154; GFX940-NEXT:    ;;#ASMSTART
5155; GFX940-NEXT:    ; use s[8:13]
5156; GFX940-NEXT:    ;;#ASMEND
5157; GFX940-NEXT:    s_setpc_b64 s[30:31]
5158  %vec0 = call <3 x i64> asm "; def $0", "=s"()
5159  %vec1 = call <3 x i64> asm "; def $0", "=s"()
5160  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 5, i32 poison>
5161  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
5162  ret void
5163}
5164
5165define void @s_shuffle_v3i64_v3i64__5_5_0() {
5166; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_5_0:
5167; GFX900:       ; %bb.0:
5168; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5169; GFX900-NEXT:    ;;#ASMSTART
5170; GFX900-NEXT:    ; def s[4:9]
5171; GFX900-NEXT:    ;;#ASMEND
5172; GFX900-NEXT:    ;;#ASMSTART
5173; GFX900-NEXT:    ; def s[8:13]
5174; GFX900-NEXT:    ;;#ASMEND
5175; GFX900-NEXT:    s_mov_b32 s8, s12
5176; GFX900-NEXT:    s_mov_b32 s9, s13
5177; GFX900-NEXT:    s_mov_b32 s10, s12
5178; GFX900-NEXT:    s_mov_b32 s11, s13
5179; GFX900-NEXT:    s_mov_b32 s12, s4
5180; GFX900-NEXT:    s_mov_b32 s13, s5
5181; GFX900-NEXT:    ;;#ASMSTART
5182; GFX900-NEXT:    ; use s[8:13]
5183; GFX900-NEXT:    ;;#ASMEND
5184; GFX900-NEXT:    s_setpc_b64 s[30:31]
5185;
5186; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_5_0:
5187; GFX90A:       ; %bb.0:
5188; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5189; GFX90A-NEXT:    ;;#ASMSTART
5190; GFX90A-NEXT:    ; def s[4:9]
5191; GFX90A-NEXT:    ;;#ASMEND
5192; GFX90A-NEXT:    ;;#ASMSTART
5193; GFX90A-NEXT:    ; def s[8:13]
5194; GFX90A-NEXT:    ;;#ASMEND
5195; GFX90A-NEXT:    s_mov_b32 s8, s12
5196; GFX90A-NEXT:    s_mov_b32 s9, s13
5197; GFX90A-NEXT:    s_mov_b32 s10, s12
5198; GFX90A-NEXT:    s_mov_b32 s11, s13
5199; GFX90A-NEXT:    s_mov_b32 s12, s4
5200; GFX90A-NEXT:    s_mov_b32 s13, s5
5201; GFX90A-NEXT:    ;;#ASMSTART
5202; GFX90A-NEXT:    ; use s[8:13]
5203; GFX90A-NEXT:    ;;#ASMEND
5204; GFX90A-NEXT:    s_setpc_b64 s[30:31]
5205;
5206; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_5_0:
5207; GFX940:       ; %bb.0:
5208; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5209; GFX940-NEXT:    ;;#ASMSTART
5210; GFX940-NEXT:    ; def s[8:13]
5211; GFX940-NEXT:    ;;#ASMEND
5212; GFX940-NEXT:    ;;#ASMSTART
5213; GFX940-NEXT:    ; def s[0:5]
5214; GFX940-NEXT:    ;;#ASMEND
5215; GFX940-NEXT:    s_mov_b32 s8, s12
5216; GFX940-NEXT:    s_mov_b32 s9, s13
5217; GFX940-NEXT:    s_mov_b32 s10, s12
5218; GFX940-NEXT:    s_mov_b32 s11, s13
5219; GFX940-NEXT:    s_mov_b32 s12, s0
5220; GFX940-NEXT:    s_mov_b32 s13, s1
5221; GFX940-NEXT:    ;;#ASMSTART
5222; GFX940-NEXT:    ; use s[8:13]
5223; GFX940-NEXT:    ;;#ASMEND
5224; GFX940-NEXT:    s_setpc_b64 s[30:31]
5225  %vec0 = call <3 x i64> asm "; def $0", "=s"()
5226  %vec1 = call <3 x i64> asm "; def $0", "=s"()
5227  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 5, i32 0>
5228  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
5229  ret void
5230}
5231
5232define void @s_shuffle_v3i64_v3i64__5_5_1() {
5233; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_5_1:
5234; GFX900:       ; %bb.0:
5235; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5236; GFX900-NEXT:    ;;#ASMSTART
5237; GFX900-NEXT:    ; def s[4:9]
5238; GFX900-NEXT:    ;;#ASMEND
5239; GFX900-NEXT:    ;;#ASMSTART
5240; GFX900-NEXT:    ; def s[8:13]
5241; GFX900-NEXT:    ;;#ASMEND
5242; GFX900-NEXT:    s_mov_b32 s8, s12
5243; GFX900-NEXT:    s_mov_b32 s9, s13
5244; GFX900-NEXT:    s_mov_b32 s10, s12
5245; GFX900-NEXT:    s_mov_b32 s11, s13
5246; GFX900-NEXT:    s_mov_b32 s12, s6
5247; GFX900-NEXT:    s_mov_b32 s13, s7
5248; GFX900-NEXT:    ;;#ASMSTART
5249; GFX900-NEXT:    ; use s[8:13]
5250; GFX900-NEXT:    ;;#ASMEND
5251; GFX900-NEXT:    s_setpc_b64 s[30:31]
5252;
5253; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_5_1:
5254; GFX90A:       ; %bb.0:
5255; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5256; GFX90A-NEXT:    ;;#ASMSTART
5257; GFX90A-NEXT:    ; def s[4:9]
5258; GFX90A-NEXT:    ;;#ASMEND
5259; GFX90A-NEXT:    ;;#ASMSTART
5260; GFX90A-NEXT:    ; def s[8:13]
5261; GFX90A-NEXT:    ;;#ASMEND
5262; GFX90A-NEXT:    s_mov_b32 s8, s12
5263; GFX90A-NEXT:    s_mov_b32 s9, s13
5264; GFX90A-NEXT:    s_mov_b32 s10, s12
5265; GFX90A-NEXT:    s_mov_b32 s11, s13
5266; GFX90A-NEXT:    s_mov_b32 s12, s6
5267; GFX90A-NEXT:    s_mov_b32 s13, s7
5268; GFX90A-NEXT:    ;;#ASMSTART
5269; GFX90A-NEXT:    ; use s[8:13]
5270; GFX90A-NEXT:    ;;#ASMEND
5271; GFX90A-NEXT:    s_setpc_b64 s[30:31]
5272;
5273; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_5_1:
5274; GFX940:       ; %bb.0:
5275; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5276; GFX940-NEXT:    ;;#ASMSTART
5277; GFX940-NEXT:    ; def s[8:13]
5278; GFX940-NEXT:    ;;#ASMEND
5279; GFX940-NEXT:    ;;#ASMSTART
5280; GFX940-NEXT:    ; def s[0:5]
5281; GFX940-NEXT:    ;;#ASMEND
5282; GFX940-NEXT:    s_mov_b32 s8, s12
5283; GFX940-NEXT:    s_mov_b32 s9, s13
5284; GFX940-NEXT:    s_mov_b32 s10, s12
5285; GFX940-NEXT:    s_mov_b32 s11, s13
5286; GFX940-NEXT:    s_mov_b32 s12, s2
5287; GFX940-NEXT:    s_mov_b32 s13, s3
5288; GFX940-NEXT:    ;;#ASMSTART
5289; GFX940-NEXT:    ; use s[8:13]
5290; GFX940-NEXT:    ;;#ASMEND
5291; GFX940-NEXT:    s_setpc_b64 s[30:31]
5292  %vec0 = call <3 x i64> asm "; def $0", "=s"()
5293  %vec1 = call <3 x i64> asm "; def $0", "=s"()
5294  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 5, i32 1>
5295  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
5296  ret void
5297}
5298
5299define void @s_shuffle_v3i64_v3i64__5_5_2() {
5300; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_5_2:
5301; GFX900:       ; %bb.0:
5302; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5303; GFX900-NEXT:    ;;#ASMSTART
5304; GFX900-NEXT:    ; def s[8:13]
5305; GFX900-NEXT:    ;;#ASMEND
5306; GFX900-NEXT:    ;;#ASMSTART
5307; GFX900-NEXT:    ; def s[16:21]
5308; GFX900-NEXT:    ;;#ASMEND
5309; GFX900-NEXT:    s_mov_b32 s8, s20
5310; GFX900-NEXT:    s_mov_b32 s9, s21
5311; GFX900-NEXT:    s_mov_b32 s10, s20
5312; GFX900-NEXT:    s_mov_b32 s11, s21
5313; GFX900-NEXT:    ;;#ASMSTART
5314; GFX900-NEXT:    ; use s[8:13]
5315; GFX900-NEXT:    ;;#ASMEND
5316; GFX900-NEXT:    s_setpc_b64 s[30:31]
5317;
5318; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_5_2:
5319; GFX90A:       ; %bb.0:
5320; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5321; GFX90A-NEXT:    ;;#ASMSTART
5322; GFX90A-NEXT:    ; def s[8:13]
5323; GFX90A-NEXT:    ;;#ASMEND
5324; GFX90A-NEXT:    ;;#ASMSTART
5325; GFX90A-NEXT:    ; def s[16:21]
5326; GFX90A-NEXT:    ;;#ASMEND
5327; GFX90A-NEXT:    s_mov_b32 s8, s20
5328; GFX90A-NEXT:    s_mov_b32 s9, s21
5329; GFX90A-NEXT:    s_mov_b32 s10, s20
5330; GFX90A-NEXT:    s_mov_b32 s11, s21
5331; GFX90A-NEXT:    ;;#ASMSTART
5332; GFX90A-NEXT:    ; use s[8:13]
5333; GFX90A-NEXT:    ;;#ASMEND
5334; GFX90A-NEXT:    s_setpc_b64 s[30:31]
5335;
5336; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_5_2:
5337; GFX940:       ; %bb.0:
5338; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5339; GFX940-NEXT:    ;;#ASMSTART
5340; GFX940-NEXT:    ; def s[8:13]
5341; GFX940-NEXT:    ;;#ASMEND
5342; GFX940-NEXT:    ;;#ASMSTART
5343; GFX940-NEXT:    ; def s[0:5]
5344; GFX940-NEXT:    ;;#ASMEND
5345; GFX940-NEXT:    s_mov_b32 s8, s4
5346; GFX940-NEXT:    s_mov_b32 s9, s5
5347; GFX940-NEXT:    s_mov_b32 s10, s4
5348; GFX940-NEXT:    s_mov_b32 s11, s5
5349; GFX940-NEXT:    ;;#ASMSTART
5350; GFX940-NEXT:    ; use s[8:13]
5351; GFX940-NEXT:    ;;#ASMEND
5352; GFX940-NEXT:    s_setpc_b64 s[30:31]
5353  %vec0 = call <3 x i64> asm "; def $0", "=s"()
5354  %vec1 = call <3 x i64> asm "; def $0", "=s"()
5355  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 5, i32 2>
5356  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
5357  ret void
5358}
5359
5360define void @s_shuffle_v3i64_v3i64__5_5_3() {
5361; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_5_3:
5362; GFX900:       ; %bb.0:
5363; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5364; GFX900-NEXT:    ;;#ASMSTART
5365; GFX900-NEXT:    ; def s[12:17]
5366; GFX900-NEXT:    ;;#ASMEND
5367; GFX900-NEXT:    s_mov_b32 s8, s16
5368; GFX900-NEXT:    s_mov_b32 s9, s17
5369; GFX900-NEXT:    s_mov_b32 s10, s16
5370; GFX900-NEXT:    s_mov_b32 s11, s17
5371; GFX900-NEXT:    ;;#ASMSTART
5372; GFX900-NEXT:    ; use s[8:13]
5373; GFX900-NEXT:    ;;#ASMEND
5374; GFX900-NEXT:    s_setpc_b64 s[30:31]
5375;
5376; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_5_3:
5377; GFX90A:       ; %bb.0:
5378; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5379; GFX90A-NEXT:    ;;#ASMSTART
5380; GFX90A-NEXT:    ; def s[12:17]
5381; GFX90A-NEXT:    ;;#ASMEND
5382; GFX90A-NEXT:    s_mov_b32 s8, s16
5383; GFX90A-NEXT:    s_mov_b32 s9, s17
5384; GFX90A-NEXT:    s_mov_b32 s10, s16
5385; GFX90A-NEXT:    s_mov_b32 s11, s17
5386; GFX90A-NEXT:    ;;#ASMSTART
5387; GFX90A-NEXT:    ; use s[8:13]
5388; GFX90A-NEXT:    ;;#ASMEND
5389; GFX90A-NEXT:    s_setpc_b64 s[30:31]
5390;
5391; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_5_3:
5392; GFX940:       ; %bb.0:
5393; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5394; GFX940-NEXT:    ;;#ASMSTART
5395; GFX940-NEXT:    ; def s[0:5]
5396; GFX940-NEXT:    ;;#ASMEND
5397; GFX940-NEXT:    s_mov_b32 s8, s4
5398; GFX940-NEXT:    s_mov_b32 s9, s5
5399; GFX940-NEXT:    s_mov_b32 s10, s4
5400; GFX940-NEXT:    s_mov_b32 s11, s5
5401; GFX940-NEXT:    s_mov_b32 s12, s0
5402; GFX940-NEXT:    s_mov_b32 s13, s1
5403; GFX940-NEXT:    ;;#ASMSTART
5404; GFX940-NEXT:    ; use s[8:13]
5405; GFX940-NEXT:    ;;#ASMEND
5406; GFX940-NEXT:    s_setpc_b64 s[30:31]
5407  %vec0 = call <3 x i64> asm "; def $0", "=s"()
5408  %vec1 = call <3 x i64> asm "; def $0", "=s"()
5409  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 5, i32 3>
5410  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
5411  ret void
5412}
5413
5414define void @s_shuffle_v3i64_v3i64__5_5_4() {
5415; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_5_4:
5416; GFX900:       ; %bb.0:
5417; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5418; GFX900-NEXT:    ;;#ASMSTART
5419; GFX900-NEXT:    ; def s[12:17]
5420; GFX900-NEXT:    ;;#ASMEND
5421; GFX900-NEXT:    s_mov_b32 s8, s16
5422; GFX900-NEXT:    s_mov_b32 s9, s17
5423; GFX900-NEXT:    s_mov_b32 s10, s16
5424; GFX900-NEXT:    s_mov_b32 s11, s17
5425; GFX900-NEXT:    s_mov_b32 s12, s14
5426; GFX900-NEXT:    s_mov_b32 s13, s15
5427; GFX900-NEXT:    ;;#ASMSTART
5428; GFX900-NEXT:    ; use s[8:13]
5429; GFX900-NEXT:    ;;#ASMEND
5430; GFX900-NEXT:    s_setpc_b64 s[30:31]
5431;
5432; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_5_4:
5433; GFX90A:       ; %bb.0:
5434; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5435; GFX90A-NEXT:    ;;#ASMSTART
5436; GFX90A-NEXT:    ; def s[12:17]
5437; GFX90A-NEXT:    ;;#ASMEND
5438; GFX90A-NEXT:    s_mov_b32 s8, s16
5439; GFX90A-NEXT:    s_mov_b32 s9, s17
5440; GFX90A-NEXT:    s_mov_b32 s10, s16
5441; GFX90A-NEXT:    s_mov_b32 s11, s17
5442; GFX90A-NEXT:    s_mov_b32 s12, s14
5443; GFX90A-NEXT:    s_mov_b32 s13, s15
5444; GFX90A-NEXT:    ;;#ASMSTART
5445; GFX90A-NEXT:    ; use s[8:13]
5446; GFX90A-NEXT:    ;;#ASMEND
5447; GFX90A-NEXT:    s_setpc_b64 s[30:31]
5448;
5449; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_5_4:
5450; GFX940:       ; %bb.0:
5451; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5452; GFX940-NEXT:    ;;#ASMSTART
5453; GFX940-NEXT:    ; def s[0:5]
5454; GFX940-NEXT:    ;;#ASMEND
5455; GFX940-NEXT:    s_mov_b32 s8, s4
5456; GFX940-NEXT:    s_mov_b32 s9, s5
5457; GFX940-NEXT:    s_mov_b32 s10, s4
5458; GFX940-NEXT:    s_mov_b32 s11, s5
5459; GFX940-NEXT:    s_mov_b32 s12, s2
5460; GFX940-NEXT:    s_mov_b32 s13, s3
5461; GFX940-NEXT:    ;;#ASMSTART
5462; GFX940-NEXT:    ; use s[8:13]
5463; GFX940-NEXT:    ;;#ASMEND
5464; GFX940-NEXT:    s_setpc_b64 s[30:31]
5465  %vec0 = call <3 x i64> asm "; def $0", "=s"()
5466  %vec1 = call <3 x i64> asm "; def $0", "=s"()
5467  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 5, i32 4>
5468  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
5469  ret void
5470}
5471
5472define void @s_shuffle_v3i64_v3i64__5_5_5() {
5473; GFX9-LABEL: s_shuffle_v3i64_v3i64__5_5_5:
5474; GFX9:       ; %bb.0:
5475; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5476; GFX9-NEXT:    ;;#ASMSTART
5477; GFX9-NEXT:    ; def s[8:13]
5478; GFX9-NEXT:    ;;#ASMEND
5479; GFX9-NEXT:    s_mov_b32 s8, s12
5480; GFX9-NEXT:    s_mov_b32 s9, s13
5481; GFX9-NEXT:    s_mov_b32 s10, s12
5482; GFX9-NEXT:    s_mov_b32 s11, s13
5483; GFX9-NEXT:    ;;#ASMSTART
5484; GFX9-NEXT:    ; use s[8:13]
5485; GFX9-NEXT:    ;;#ASMEND
5486; GFX9-NEXT:    s_setpc_b64 s[30:31]
5487  %vec0 = call <3 x i64> asm "; def $0", "=s"()
5488  %vec1 = call <3 x i64> asm "; def $0", "=s"()
5489  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 5, i32 5>
5490  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
5491  ret void
5492}
5493
5494define void @s_shuffle_v3i64_v3i64__u_0_0() {
5495; GFX900-LABEL: s_shuffle_v3i64_v3i64__u_0_0:
5496; GFX900:       ; %bb.0:
5497; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5498; GFX900-NEXT:    ;;#ASMSTART
5499; GFX900-NEXT:    ; def s[4:9]
5500; GFX900-NEXT:    ;;#ASMEND
5501; GFX900-NEXT:    s_mov_b32 s10, s4
5502; GFX900-NEXT:    s_mov_b32 s11, s5
5503; GFX900-NEXT:    s_mov_b32 s12, s4
5504; GFX900-NEXT:    s_mov_b32 s13, s5
5505; GFX900-NEXT:    ;;#ASMSTART
5506; GFX900-NEXT:    ; use s[8:13]
5507; GFX900-NEXT:    ;;#ASMEND
5508; GFX900-NEXT:    s_setpc_b64 s[30:31]
5509;
5510; GFX90A-LABEL: s_shuffle_v3i64_v3i64__u_0_0:
5511; GFX90A:       ; %bb.0:
5512; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5513; GFX90A-NEXT:    ;;#ASMSTART
5514; GFX90A-NEXT:    ; def s[4:9]
5515; GFX90A-NEXT:    ;;#ASMEND
5516; GFX90A-NEXT:    s_mov_b32 s10, s4
5517; GFX90A-NEXT:    s_mov_b32 s11, s5
5518; GFX90A-NEXT:    s_mov_b32 s12, s4
5519; GFX90A-NEXT:    s_mov_b32 s13, s5
5520; GFX90A-NEXT:    ;;#ASMSTART
5521; GFX90A-NEXT:    ; use s[8:13]
5522; GFX90A-NEXT:    ;;#ASMEND
5523; GFX90A-NEXT:    s_setpc_b64 s[30:31]
5524;
5525; GFX940-LABEL: s_shuffle_v3i64_v3i64__u_0_0:
5526; GFX940:       ; %bb.0:
5527; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5528; GFX940-NEXT:    ;;#ASMSTART
5529; GFX940-NEXT:    ; def s[0:5]
5530; GFX940-NEXT:    ;;#ASMEND
5531; GFX940-NEXT:    s_mov_b32 s10, s0
5532; GFX940-NEXT:    s_mov_b32 s11, s1
5533; GFX940-NEXT:    s_mov_b32 s12, s0
5534; GFX940-NEXT:    s_mov_b32 s13, s1
5535; GFX940-NEXT:    ;;#ASMSTART
5536; GFX940-NEXT:    ; use s[8:13]
5537; GFX940-NEXT:    ;;#ASMEND
5538; GFX940-NEXT:    s_setpc_b64 s[30:31]
5539  %vec0 = call <3 x i64> asm "; def $0", "=s"()
5540  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 poison, i32 0, i32 0>
5541  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
5542  ret void
5543}
5544
5545define void @s_shuffle_v3i64_v3i64__0_0_0() {
5546; GFX9-LABEL: s_shuffle_v3i64_v3i64__0_0_0:
5547; GFX9:       ; %bb.0:
5548; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5549; GFX9-NEXT:    ;;#ASMSTART
5550; GFX9-NEXT:    ; def s[8:13]
5551; GFX9-NEXT:    ;;#ASMEND
5552; GFX9-NEXT:    s_mov_b32 s10, s8
5553; GFX9-NEXT:    s_mov_b32 s11, s9
5554; GFX9-NEXT:    s_mov_b32 s12, s8
5555; GFX9-NEXT:    s_mov_b32 s13, s9
5556; GFX9-NEXT:    ;;#ASMSTART
5557; GFX9-NEXT:    ; use s[8:13]
5558; GFX9-NEXT:    ;;#ASMEND
5559; GFX9-NEXT:    s_setpc_b64 s[30:31]
5560  %vec0 = call <3 x i64> asm "; def $0", "=s"()
5561  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> zeroinitializer
5562  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
5563  ret void
5564}
5565
5566define void @s_shuffle_v3i64_v3i64__1_0_0() {
5567; GFX900-LABEL: s_shuffle_v3i64_v3i64__1_0_0:
5568; GFX900:       ; %bb.0:
5569; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5570; GFX900-NEXT:    ;;#ASMSTART
5571; GFX900-NEXT:    ; def s[4:9]
5572; GFX900-NEXT:    ;;#ASMEND
5573; GFX900-NEXT:    s_mov_b32 s8, s6
5574; GFX900-NEXT:    s_mov_b32 s9, s7
5575; GFX900-NEXT:    s_mov_b32 s10, s4
5576; GFX900-NEXT:    s_mov_b32 s11, s5
5577; GFX900-NEXT:    s_mov_b32 s12, s4
5578; GFX900-NEXT:    s_mov_b32 s13, s5
5579; GFX900-NEXT:    ;;#ASMSTART
5580; GFX900-NEXT:    ; use s[8:13]
5581; GFX900-NEXT:    ;;#ASMEND
5582; GFX900-NEXT:    s_setpc_b64 s[30:31]
5583;
5584; GFX90A-LABEL: s_shuffle_v3i64_v3i64__1_0_0:
5585; GFX90A:       ; %bb.0:
5586; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5587; GFX90A-NEXT:    ;;#ASMSTART
5588; GFX90A-NEXT:    ; def s[4:9]
5589; GFX90A-NEXT:    ;;#ASMEND
5590; GFX90A-NEXT:    s_mov_b32 s8, s6
5591; GFX90A-NEXT:    s_mov_b32 s9, s7
5592; GFX90A-NEXT:    s_mov_b32 s10, s4
5593; GFX90A-NEXT:    s_mov_b32 s11, s5
5594; GFX90A-NEXT:    s_mov_b32 s12, s4
5595; GFX90A-NEXT:    s_mov_b32 s13, s5
5596; GFX90A-NEXT:    ;;#ASMSTART
5597; GFX90A-NEXT:    ; use s[8:13]
5598; GFX90A-NEXT:    ;;#ASMEND
5599; GFX90A-NEXT:    s_setpc_b64 s[30:31]
5600;
5601; GFX940-LABEL: s_shuffle_v3i64_v3i64__1_0_0:
5602; GFX940:       ; %bb.0:
5603; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5604; GFX940-NEXT:    ;;#ASMSTART
5605; GFX940-NEXT:    ; def s[0:5]
5606; GFX940-NEXT:    ;;#ASMEND
5607; GFX940-NEXT:    s_mov_b32 s8, s2
5608; GFX940-NEXT:    s_mov_b32 s9, s3
5609; GFX940-NEXT:    s_mov_b32 s10, s0
5610; GFX940-NEXT:    s_mov_b32 s11, s1
5611; GFX940-NEXT:    s_mov_b32 s12, s0
5612; GFX940-NEXT:    s_mov_b32 s13, s1
5613; GFX940-NEXT:    ;;#ASMSTART
5614; GFX940-NEXT:    ; use s[8:13]
5615; GFX940-NEXT:    ;;#ASMEND
5616; GFX940-NEXT:    s_setpc_b64 s[30:31]
5617  %vec0 = call <3 x i64> asm "; def $0", "=s"()
5618  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 1, i32 0, i32 0>
5619  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
5620  ret void
5621}
5622
5623define void @s_shuffle_v3i64_v3i64__2_0_0() {
5624; GFX900-LABEL: s_shuffle_v3i64_v3i64__2_0_0:
5625; GFX900:       ; %bb.0:
5626; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5627; GFX900-NEXT:    ;;#ASMSTART
5628; GFX900-NEXT:    ; def s[4:9]
5629; GFX900-NEXT:    ;;#ASMEND
5630; GFX900-NEXT:    s_mov_b32 s10, s4
5631; GFX900-NEXT:    s_mov_b32 s11, s5
5632; GFX900-NEXT:    s_mov_b32 s12, s4
5633; GFX900-NEXT:    s_mov_b32 s13, s5
5634; GFX900-NEXT:    ;;#ASMSTART
5635; GFX900-NEXT:    ; use s[8:13]
5636; GFX900-NEXT:    ;;#ASMEND
5637; GFX900-NEXT:    s_setpc_b64 s[30:31]
5638;
5639; GFX90A-LABEL: s_shuffle_v3i64_v3i64__2_0_0:
5640; GFX90A:       ; %bb.0:
5641; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5642; GFX90A-NEXT:    ;;#ASMSTART
5643; GFX90A-NEXT:    ; def s[4:9]
5644; GFX90A-NEXT:    ;;#ASMEND
5645; GFX90A-NEXT:    s_mov_b32 s10, s4
5646; GFX90A-NEXT:    s_mov_b32 s11, s5
5647; GFX90A-NEXT:    s_mov_b32 s12, s4
5648; GFX90A-NEXT:    s_mov_b32 s13, s5
5649; GFX90A-NEXT:    ;;#ASMSTART
5650; GFX90A-NEXT:    ; use s[8:13]
5651; GFX90A-NEXT:    ;;#ASMEND
5652; GFX90A-NEXT:    s_setpc_b64 s[30:31]
5653;
5654; GFX940-LABEL: s_shuffle_v3i64_v3i64__2_0_0:
5655; GFX940:       ; %bb.0:
5656; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5657; GFX940-NEXT:    ;;#ASMSTART
5658; GFX940-NEXT:    ; def s[0:5]
5659; GFX940-NEXT:    ;;#ASMEND
5660; GFX940-NEXT:    s_mov_b32 s8, s4
5661; GFX940-NEXT:    s_mov_b32 s9, s5
5662; GFX940-NEXT:    s_mov_b32 s10, s0
5663; GFX940-NEXT:    s_mov_b32 s11, s1
5664; GFX940-NEXT:    s_mov_b32 s12, s0
5665; GFX940-NEXT:    s_mov_b32 s13, s1
5666; GFX940-NEXT:    ;;#ASMSTART
5667; GFX940-NEXT:    ; use s[8:13]
5668; GFX940-NEXT:    ;;#ASMEND
5669; GFX940-NEXT:    s_setpc_b64 s[30:31]
5670  %vec0 = call <3 x i64> asm "; def $0", "=s"()
5671  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 2, i32 0, i32 0>
5672  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
5673  ret void
5674}
5675
5676define void @s_shuffle_v3i64_v3i64__3_0_0() {
5677; GFX900-LABEL: s_shuffle_v3i64_v3i64__3_0_0:
5678; GFX900:       ; %bb.0:
5679; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5680; GFX900-NEXT:    ;;#ASMSTART
5681; GFX900-NEXT:    ; def s[4:9]
5682; GFX900-NEXT:    ;;#ASMEND
5683; GFX900-NEXT:    s_mov_b32 s10, s4
5684; GFX900-NEXT:    s_mov_b32 s11, s5
5685; GFX900-NEXT:    s_mov_b32 s12, s4
5686; GFX900-NEXT:    s_mov_b32 s13, s5
5687; GFX900-NEXT:    ;;#ASMSTART
5688; GFX900-NEXT:    ; use s[8:13]
5689; GFX900-NEXT:    ;;#ASMEND
5690; GFX900-NEXT:    s_setpc_b64 s[30:31]
5691;
5692; GFX90A-LABEL: s_shuffle_v3i64_v3i64__3_0_0:
5693; GFX90A:       ; %bb.0:
5694; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5695; GFX90A-NEXT:    ;;#ASMSTART
5696; GFX90A-NEXT:    ; def s[4:9]
5697; GFX90A-NEXT:    ;;#ASMEND
5698; GFX90A-NEXT:    s_mov_b32 s10, s4
5699; GFX90A-NEXT:    s_mov_b32 s11, s5
5700; GFX90A-NEXT:    s_mov_b32 s12, s4
5701; GFX90A-NEXT:    s_mov_b32 s13, s5
5702; GFX90A-NEXT:    ;;#ASMSTART
5703; GFX90A-NEXT:    ; use s[8:13]
5704; GFX90A-NEXT:    ;;#ASMEND
5705; GFX90A-NEXT:    s_setpc_b64 s[30:31]
5706;
5707; GFX940-LABEL: s_shuffle_v3i64_v3i64__3_0_0:
5708; GFX940:       ; %bb.0:
5709; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5710; GFX940-NEXT:    ;;#ASMSTART
5711; GFX940-NEXT:    ; def s[0:5]
5712; GFX940-NEXT:    ;;#ASMEND
5713; GFX940-NEXT:    s_mov_b32 s10, s0
5714; GFX940-NEXT:    s_mov_b32 s11, s1
5715; GFX940-NEXT:    s_mov_b32 s12, s0
5716; GFX940-NEXT:    s_mov_b32 s13, s1
5717; GFX940-NEXT:    ;;#ASMSTART
5718; GFX940-NEXT:    ; use s[8:13]
5719; GFX940-NEXT:    ;;#ASMEND
5720; GFX940-NEXT:    s_setpc_b64 s[30:31]
5721  %vec0 = call <3 x i64> asm "; def $0", "=s"()
5722  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 3, i32 0, i32 0>
5723  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
5724  ret void
5725}
5726
5727define void @s_shuffle_v3i64_v3i64__4_0_0() {
5728; GFX900-LABEL: s_shuffle_v3i64_v3i64__4_0_0:
5729; GFX900:       ; %bb.0:
5730; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5731; GFX900-NEXT:    ;;#ASMSTART
5732; GFX900-NEXT:    ; def s[4:9]
5733; GFX900-NEXT:    ;;#ASMEND
5734; GFX900-NEXT:    ;;#ASMSTART
5735; GFX900-NEXT:    ; def s[8:13]
5736; GFX900-NEXT:    ;;#ASMEND
5737; GFX900-NEXT:    s_mov_b32 s8, s10
5738; GFX900-NEXT:    s_mov_b32 s9, s11
5739; GFX900-NEXT:    s_mov_b32 s10, s4
5740; GFX900-NEXT:    s_mov_b32 s11, s5
5741; GFX900-NEXT:    s_mov_b32 s12, s4
5742; GFX900-NEXT:    s_mov_b32 s13, s5
5743; GFX900-NEXT:    ;;#ASMSTART
5744; GFX900-NEXT:    ; use s[8:13]
5745; GFX900-NEXT:    ;;#ASMEND
5746; GFX900-NEXT:    s_setpc_b64 s[30:31]
5747;
5748; GFX90A-LABEL: s_shuffle_v3i64_v3i64__4_0_0:
5749; GFX90A:       ; %bb.0:
5750; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5751; GFX90A-NEXT:    ;;#ASMSTART
5752; GFX90A-NEXT:    ; def s[4:9]
5753; GFX90A-NEXT:    ;;#ASMEND
5754; GFX90A-NEXT:    ;;#ASMSTART
5755; GFX90A-NEXT:    ; def s[8:13]
5756; GFX90A-NEXT:    ;;#ASMEND
5757; GFX90A-NEXT:    s_mov_b32 s8, s10
5758; GFX90A-NEXT:    s_mov_b32 s9, s11
5759; GFX90A-NEXT:    s_mov_b32 s10, s4
5760; GFX90A-NEXT:    s_mov_b32 s11, s5
5761; GFX90A-NEXT:    s_mov_b32 s12, s4
5762; GFX90A-NEXT:    s_mov_b32 s13, s5
5763; GFX90A-NEXT:    ;;#ASMSTART
5764; GFX90A-NEXT:    ; use s[8:13]
5765; GFX90A-NEXT:    ;;#ASMEND
5766; GFX90A-NEXT:    s_setpc_b64 s[30:31]
5767;
5768; GFX940-LABEL: s_shuffle_v3i64_v3i64__4_0_0:
5769; GFX940:       ; %bb.0:
5770; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5771; GFX940-NEXT:    ;;#ASMSTART
5772; GFX940-NEXT:    ; def s[0:5]
5773; GFX940-NEXT:    ;;#ASMEND
5774; GFX940-NEXT:    s_mov_b32 s10, s0
5775; GFX940-NEXT:    ;;#ASMSTART
5776; GFX940-NEXT:    ; def s[4:9]
5777; GFX940-NEXT:    ;;#ASMEND
5778; GFX940-NEXT:    s_mov_b32 s8, s6
5779; GFX940-NEXT:    s_mov_b32 s9, s7
5780; GFX940-NEXT:    s_mov_b32 s11, s1
5781; GFX940-NEXT:    s_mov_b32 s12, s0
5782; GFX940-NEXT:    s_mov_b32 s13, s1
5783; GFX940-NEXT:    ;;#ASMSTART
5784; GFX940-NEXT:    ; use s[8:13]
5785; GFX940-NEXT:    ;;#ASMEND
5786; GFX940-NEXT:    s_setpc_b64 s[30:31]
5787  %vec0 = call <3 x i64> asm "; def $0", "=s"()
5788  %vec1 = call <3 x i64> asm "; def $0", "=s"()
5789  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 4, i32 0, i32 0>
5790  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
5791  ret void
5792}
5793
5794define void @s_shuffle_v3i64_v3i64__5_0_0() {
5795; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_0_0:
5796; GFX900:       ; %bb.0:
5797; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5798; GFX900-NEXT:    ;;#ASMSTART
5799; GFX900-NEXT:    ; def s[4:9]
5800; GFX900-NEXT:    ;;#ASMEND
5801; GFX900-NEXT:    ;;#ASMSTART
5802; GFX900-NEXT:    ; def s[8:13]
5803; GFX900-NEXT:    ;;#ASMEND
5804; GFX900-NEXT:    s_mov_b32 s8, s12
5805; GFX900-NEXT:    s_mov_b32 s9, s13
5806; GFX900-NEXT:    s_mov_b32 s10, s4
5807; GFX900-NEXT:    s_mov_b32 s11, s5
5808; GFX900-NEXT:    s_mov_b32 s12, s4
5809; GFX900-NEXT:    s_mov_b32 s13, s5
5810; GFX900-NEXT:    ;;#ASMSTART
5811; GFX900-NEXT:    ; use s[8:13]
5812; GFX900-NEXT:    ;;#ASMEND
5813; GFX900-NEXT:    s_setpc_b64 s[30:31]
5814;
5815; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_0_0:
5816; GFX90A:       ; %bb.0:
5817; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5818; GFX90A-NEXT:    ;;#ASMSTART
5819; GFX90A-NEXT:    ; def s[4:9]
5820; GFX90A-NEXT:    ;;#ASMEND
5821; GFX90A-NEXT:    ;;#ASMSTART
5822; GFX90A-NEXT:    ; def s[8:13]
5823; GFX90A-NEXT:    ;;#ASMEND
5824; GFX90A-NEXT:    s_mov_b32 s8, s12
5825; GFX90A-NEXT:    s_mov_b32 s9, s13
5826; GFX90A-NEXT:    s_mov_b32 s10, s4
5827; GFX90A-NEXT:    s_mov_b32 s11, s5
5828; GFX90A-NEXT:    s_mov_b32 s12, s4
5829; GFX90A-NEXT:    s_mov_b32 s13, s5
5830; GFX90A-NEXT:    ;;#ASMSTART
5831; GFX90A-NEXT:    ; use s[8:13]
5832; GFX90A-NEXT:    ;;#ASMEND
5833; GFX90A-NEXT:    s_setpc_b64 s[30:31]
5834;
5835; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_0_0:
5836; GFX940:       ; %bb.0:
5837; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5838; GFX940-NEXT:    ;;#ASMSTART
5839; GFX940-NEXT:    ; def s[0:5]
5840; GFX940-NEXT:    ;;#ASMEND
5841; GFX940-NEXT:    s_mov_b32 s10, s0
5842; GFX940-NEXT:    ;;#ASMSTART
5843; GFX940-NEXT:    ; def s[4:9]
5844; GFX940-NEXT:    ;;#ASMEND
5845; GFX940-NEXT:    s_mov_b32 s11, s1
5846; GFX940-NEXT:    s_mov_b32 s12, s0
5847; GFX940-NEXT:    s_mov_b32 s13, s1
5848; GFX940-NEXT:    ;;#ASMSTART
5849; GFX940-NEXT:    ; use s[8:13]
5850; GFX940-NEXT:    ;;#ASMEND
5851; GFX940-NEXT:    s_setpc_b64 s[30:31]
5852  %vec0 = call <3 x i64> asm "; def $0", "=s"()
5853  %vec1 = call <3 x i64> asm "; def $0", "=s"()
5854  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 0, i32 0>
5855  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
5856  ret void
5857}
5858
5859define void @s_shuffle_v3i64_v3i64__5_u_0() {
5860; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_u_0:
5861; GFX900:       ; %bb.0:
5862; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5863; GFX900-NEXT:    ;;#ASMSTART
5864; GFX900-NEXT:    ; def s[4:9]
5865; GFX900-NEXT:    ;;#ASMEND
5866; GFX900-NEXT:    ;;#ASMSTART
5867; GFX900-NEXT:    ; def s[8:13]
5868; GFX900-NEXT:    ;;#ASMEND
5869; GFX900-NEXT:    s_mov_b32 s8, s12
5870; GFX900-NEXT:    s_mov_b32 s9, s13
5871; GFX900-NEXT:    s_mov_b32 s12, s4
5872; GFX900-NEXT:    s_mov_b32 s13, s5
5873; GFX900-NEXT:    ;;#ASMSTART
5874; GFX900-NEXT:    ; use s[8:13]
5875; GFX900-NEXT:    ;;#ASMEND
5876; GFX900-NEXT:    s_setpc_b64 s[30:31]
5877;
5878; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_u_0:
5879; GFX90A:       ; %bb.0:
5880; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5881; GFX90A-NEXT:    ;;#ASMSTART
5882; GFX90A-NEXT:    ; def s[4:9]
5883; GFX90A-NEXT:    ;;#ASMEND
5884; GFX90A-NEXT:    ;;#ASMSTART
5885; GFX90A-NEXT:    ; def s[8:13]
5886; GFX90A-NEXT:    ;;#ASMEND
5887; GFX90A-NEXT:    s_mov_b32 s8, s12
5888; GFX90A-NEXT:    s_mov_b32 s9, s13
5889; GFX90A-NEXT:    s_mov_b32 s12, s4
5890; GFX90A-NEXT:    s_mov_b32 s13, s5
5891; GFX90A-NEXT:    ;;#ASMSTART
5892; GFX90A-NEXT:    ; use s[8:13]
5893; GFX90A-NEXT:    ;;#ASMEND
5894; GFX90A-NEXT:    s_setpc_b64 s[30:31]
5895;
5896; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_u_0:
5897; GFX940:       ; %bb.0:
5898; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5899; GFX940-NEXT:    ;;#ASMSTART
5900; GFX940-NEXT:    ; def s[0:5]
5901; GFX940-NEXT:    ;;#ASMEND
5902; GFX940-NEXT:    s_mov_b32 s12, s0
5903; GFX940-NEXT:    ;;#ASMSTART
5904; GFX940-NEXT:    ; def s[4:9]
5905; GFX940-NEXT:    ;;#ASMEND
5906; GFX940-NEXT:    s_mov_b32 s13, s1
5907; GFX940-NEXT:    ;;#ASMSTART
5908; GFX940-NEXT:    ; use s[8:13]
5909; GFX940-NEXT:    ;;#ASMEND
5910; GFX940-NEXT:    s_setpc_b64 s[30:31]
5911  %vec0 = call <3 x i64> asm "; def $0", "=s"()
5912  %vec1 = call <3 x i64> asm "; def $0", "=s"()
5913  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 poison, i32 0>
5914  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
5915  ret void
5916}
5917
5918define void @s_shuffle_v3i64_v3i64__5_1_0() {
5919; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_1_0:
5920; GFX900:       ; %bb.0:
5921; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5922; GFX900-NEXT:    ;;#ASMSTART
5923; GFX900-NEXT:    ; def s[4:9]
5924; GFX900-NEXT:    ;;#ASMEND
5925; GFX900-NEXT:    ;;#ASMSTART
5926; GFX900-NEXT:    ; def s[8:13]
5927; GFX900-NEXT:    ;;#ASMEND
5928; GFX900-NEXT:    s_mov_b32 s8, s12
5929; GFX900-NEXT:    s_mov_b32 s9, s13
5930; GFX900-NEXT:    s_mov_b32 s10, s6
5931; GFX900-NEXT:    s_mov_b32 s11, s7
5932; GFX900-NEXT:    s_mov_b32 s12, s4
5933; GFX900-NEXT:    s_mov_b32 s13, s5
5934; GFX900-NEXT:    ;;#ASMSTART
5935; GFX900-NEXT:    ; use s[8:13]
5936; GFX900-NEXT:    ;;#ASMEND
5937; GFX900-NEXT:    s_setpc_b64 s[30:31]
5938;
5939; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_1_0:
5940; GFX90A:       ; %bb.0:
5941; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5942; GFX90A-NEXT:    ;;#ASMSTART
5943; GFX90A-NEXT:    ; def s[4:9]
5944; GFX90A-NEXT:    ;;#ASMEND
5945; GFX90A-NEXT:    ;;#ASMSTART
5946; GFX90A-NEXT:    ; def s[8:13]
5947; GFX90A-NEXT:    ;;#ASMEND
5948; GFX90A-NEXT:    s_mov_b32 s8, s12
5949; GFX90A-NEXT:    s_mov_b32 s9, s13
5950; GFX90A-NEXT:    s_mov_b32 s10, s6
5951; GFX90A-NEXT:    s_mov_b32 s11, s7
5952; GFX90A-NEXT:    s_mov_b32 s12, s4
5953; GFX90A-NEXT:    s_mov_b32 s13, s5
5954; GFX90A-NEXT:    ;;#ASMSTART
5955; GFX90A-NEXT:    ; use s[8:13]
5956; GFX90A-NEXT:    ;;#ASMEND
5957; GFX90A-NEXT:    s_setpc_b64 s[30:31]
5958;
5959; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_1_0:
5960; GFX940:       ; %bb.0:
5961; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5962; GFX940-NEXT:    ;;#ASMSTART
5963; GFX940-NEXT:    ; def s[0:5]
5964; GFX940-NEXT:    ;;#ASMEND
5965; GFX940-NEXT:    s_mov_b32 s10, s2
5966; GFX940-NEXT:    ;;#ASMSTART
5967; GFX940-NEXT:    ; def s[4:9]
5968; GFX940-NEXT:    ;;#ASMEND
5969; GFX940-NEXT:    s_mov_b32 s11, s3
5970; GFX940-NEXT:    s_mov_b32 s12, s0
5971; GFX940-NEXT:    s_mov_b32 s13, s1
5972; GFX940-NEXT:    ;;#ASMSTART
5973; GFX940-NEXT:    ; use s[8:13]
5974; GFX940-NEXT:    ;;#ASMEND
5975; GFX940-NEXT:    s_setpc_b64 s[30:31]
5976  %vec0 = call <3 x i64> asm "; def $0", "=s"()
5977  %vec1 = call <3 x i64> asm "; def $0", "=s"()
5978  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 1, i32 0>
5979  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
5980  ret void
5981}
5982
5983define void @s_shuffle_v3i64_v3i64__5_2_0() {
5984; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_2_0:
5985; GFX900:       ; %bb.0:
5986; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5987; GFX900-NEXT:    ;;#ASMSTART
5988; GFX900-NEXT:    ; def s[12:17]
5989; GFX900-NEXT:    ;;#ASMEND
5990; GFX900-NEXT:    ;;#ASMSTART
5991; GFX900-NEXT:    ; def s[4:9]
5992; GFX900-NEXT:    ;;#ASMEND
5993; GFX900-NEXT:    s_mov_b32 s10, s16
5994; GFX900-NEXT:    s_mov_b32 s11, s17
5995; GFX900-NEXT:    ;;#ASMSTART
5996; GFX900-NEXT:    ; use s[8:13]
5997; GFX900-NEXT:    ;;#ASMEND
5998; GFX900-NEXT:    s_setpc_b64 s[30:31]
5999;
6000; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_2_0:
6001; GFX90A:       ; %bb.0:
6002; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6003; GFX90A-NEXT:    ;;#ASMSTART
6004; GFX90A-NEXT:    ; def s[12:17]
6005; GFX90A-NEXT:    ;;#ASMEND
6006; GFX90A-NEXT:    ;;#ASMSTART
6007; GFX90A-NEXT:    ; def s[4:9]
6008; GFX90A-NEXT:    ;;#ASMEND
6009; GFX90A-NEXT:    s_mov_b32 s10, s16
6010; GFX90A-NEXT:    s_mov_b32 s11, s17
6011; GFX90A-NEXT:    ;;#ASMSTART
6012; GFX90A-NEXT:    ; use s[8:13]
6013; GFX90A-NEXT:    ;;#ASMEND
6014; GFX90A-NEXT:    s_setpc_b64 s[30:31]
6015;
6016; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_2_0:
6017; GFX940:       ; %bb.0:
6018; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6019; GFX940-NEXT:    ;;#ASMSTART
6020; GFX940-NEXT:    ; def s[8:13]
6021; GFX940-NEXT:    ;;#ASMEND
6022; GFX940-NEXT:    ;;#ASMSTART
6023; GFX940-NEXT:    ; def s[0:5]
6024; GFX940-NEXT:    ;;#ASMEND
6025; GFX940-NEXT:    s_mov_b32 s8, s12
6026; GFX940-NEXT:    s_mov_b32 s9, s13
6027; GFX940-NEXT:    s_mov_b32 s10, s4
6028; GFX940-NEXT:    s_mov_b32 s11, s5
6029; GFX940-NEXT:    s_mov_b32 s12, s0
6030; GFX940-NEXT:    s_mov_b32 s13, s1
6031; GFX940-NEXT:    ;;#ASMSTART
6032; GFX940-NEXT:    ; use s[8:13]
6033; GFX940-NEXT:    ;;#ASMEND
6034; GFX940-NEXT:    s_setpc_b64 s[30:31]
6035  %vec0 = call <3 x i64> asm "; def $0", "=s"()
6036  %vec1 = call <3 x i64> asm "; def $0", "=s"()
6037  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 2, i32 0>
6038  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
6039  ret void
6040}
6041
6042define void @s_shuffle_v3i64_v3i64__5_3_0() {
6043; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_3_0:
6044; GFX900:       ; %bb.0:
6045; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6046; GFX900-NEXT:    ;;#ASMSTART
6047; GFX900-NEXT:    ; def s[4:9]
6048; GFX900-NEXT:    ;;#ASMEND
6049; GFX900-NEXT:    ;;#ASMSTART
6050; GFX900-NEXT:    ; def s[12:17]
6051; GFX900-NEXT:    ;;#ASMEND
6052; GFX900-NEXT:    s_mov_b32 s8, s16
6053; GFX900-NEXT:    s_mov_b32 s9, s17
6054; GFX900-NEXT:    s_mov_b32 s10, s12
6055; GFX900-NEXT:    s_mov_b32 s11, s13
6056; GFX900-NEXT:    s_mov_b32 s12, s4
6057; GFX900-NEXT:    s_mov_b32 s13, s5
6058; GFX900-NEXT:    ;;#ASMSTART
6059; GFX900-NEXT:    ; use s[8:13]
6060; GFX900-NEXT:    ;;#ASMEND
6061; GFX900-NEXT:    s_setpc_b64 s[30:31]
6062;
6063; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_3_0:
6064; GFX90A:       ; %bb.0:
6065; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6066; GFX90A-NEXT:    ;;#ASMSTART
6067; GFX90A-NEXT:    ; def s[4:9]
6068; GFX90A-NEXT:    ;;#ASMEND
6069; GFX90A-NEXT:    ;;#ASMSTART
6070; GFX90A-NEXT:    ; def s[12:17]
6071; GFX90A-NEXT:    ;;#ASMEND
6072; GFX90A-NEXT:    s_mov_b32 s8, s16
6073; GFX90A-NEXT:    s_mov_b32 s9, s17
6074; GFX90A-NEXT:    s_mov_b32 s10, s12
6075; GFX90A-NEXT:    s_mov_b32 s11, s13
6076; GFX90A-NEXT:    s_mov_b32 s12, s4
6077; GFX90A-NEXT:    s_mov_b32 s13, s5
6078; GFX90A-NEXT:    ;;#ASMSTART
6079; GFX90A-NEXT:    ; use s[8:13]
6080; GFX90A-NEXT:    ;;#ASMEND
6081; GFX90A-NEXT:    s_setpc_b64 s[30:31]
6082;
6083; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_3_0:
6084; GFX940:       ; %bb.0:
6085; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6086; GFX940-NEXT:    ;;#ASMSTART
6087; GFX940-NEXT:    ; def s[0:5]
6088; GFX940-NEXT:    ;;#ASMEND
6089; GFX940-NEXT:    s_mov_b32 s12, s0
6090; GFX940-NEXT:    ;;#ASMSTART
6091; GFX940-NEXT:    ; def s[4:9]
6092; GFX940-NEXT:    ;;#ASMEND
6093; GFX940-NEXT:    s_mov_b32 s10, s4
6094; GFX940-NEXT:    s_mov_b32 s11, s5
6095; GFX940-NEXT:    s_mov_b32 s13, s1
6096; GFX940-NEXT:    ;;#ASMSTART
6097; GFX940-NEXT:    ; use s[8:13]
6098; GFX940-NEXT:    ;;#ASMEND
6099; GFX940-NEXT:    s_setpc_b64 s[30:31]
6100  %vec0 = call <3 x i64> asm "; def $0", "=s"()
6101  %vec1 = call <3 x i64> asm "; def $0", "=s"()
6102  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 3, i32 0>
6103  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
6104  ret void
6105}
6106
6107define void @s_shuffle_v3i64_v3i64__5_4_0() {
6108; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_4_0:
6109; GFX900:       ; %bb.0:
6110; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6111; GFX900-NEXT:    ;;#ASMSTART
6112; GFX900-NEXT:    ; def s[4:9]
6113; GFX900-NEXT:    ;;#ASMEND
6114; GFX900-NEXT:    ;;#ASMSTART
6115; GFX900-NEXT:    ; def s[8:13]
6116; GFX900-NEXT:    ;;#ASMEND
6117; GFX900-NEXT:    s_mov_b32 s8, s12
6118; GFX900-NEXT:    s_mov_b32 s9, s13
6119; GFX900-NEXT:    s_mov_b32 s12, s4
6120; GFX900-NEXT:    s_mov_b32 s13, s5
6121; GFX900-NEXT:    ;;#ASMSTART
6122; GFX900-NEXT:    ; use s[8:13]
6123; GFX900-NEXT:    ;;#ASMEND
6124; GFX900-NEXT:    s_setpc_b64 s[30:31]
6125;
6126; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_4_0:
6127; GFX90A:       ; %bb.0:
6128; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6129; GFX90A-NEXT:    ;;#ASMSTART
6130; GFX90A-NEXT:    ; def s[4:9]
6131; GFX90A-NEXT:    ;;#ASMEND
6132; GFX90A-NEXT:    ;;#ASMSTART
6133; GFX90A-NEXT:    ; def s[8:13]
6134; GFX90A-NEXT:    ;;#ASMEND
6135; GFX90A-NEXT:    s_mov_b32 s8, s12
6136; GFX90A-NEXT:    s_mov_b32 s9, s13
6137; GFX90A-NEXT:    s_mov_b32 s12, s4
6138; GFX90A-NEXT:    s_mov_b32 s13, s5
6139; GFX90A-NEXT:    ;;#ASMSTART
6140; GFX90A-NEXT:    ; use s[8:13]
6141; GFX90A-NEXT:    ;;#ASMEND
6142; GFX90A-NEXT:    s_setpc_b64 s[30:31]
6143;
6144; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_4_0:
6145; GFX940:       ; %bb.0:
6146; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6147; GFX940-NEXT:    ;;#ASMSTART
6148; GFX940-NEXT:    ; def s[8:13]
6149; GFX940-NEXT:    ;;#ASMEND
6150; GFX940-NEXT:    ;;#ASMSTART
6151; GFX940-NEXT:    ; def s[0:5]
6152; GFX940-NEXT:    ;;#ASMEND
6153; GFX940-NEXT:    s_mov_b32 s8, s12
6154; GFX940-NEXT:    s_mov_b32 s9, s13
6155; GFX940-NEXT:    s_mov_b32 s12, s0
6156; GFX940-NEXT:    s_mov_b32 s13, s1
6157; GFX940-NEXT:    ;;#ASMSTART
6158; GFX940-NEXT:    ; use s[8:13]
6159; GFX940-NEXT:    ;;#ASMEND
6160; GFX940-NEXT:    s_setpc_b64 s[30:31]
6161  %vec0 = call <3 x i64> asm "; def $0", "=s"()
6162  %vec1 = call <3 x i64> asm "; def $0", "=s"()
6163  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 4, i32 0>
6164  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
6165  ret void
6166}
6167
6168define void @s_shuffle_v3i64_v3i64__u_1_1() {
6169; GFX9-LABEL: s_shuffle_v3i64_v3i64__u_1_1:
6170; GFX9:       ; %bb.0:
6171; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6172; GFX9-NEXT:    ;;#ASMSTART
6173; GFX9-NEXT:    ; def s[8:13]
6174; GFX9-NEXT:    ;;#ASMEND
6175; GFX9-NEXT:    s_mov_b32 s12, s10
6176; GFX9-NEXT:    s_mov_b32 s13, s11
6177; GFX9-NEXT:    ;;#ASMSTART
6178; GFX9-NEXT:    ; use s[8:13]
6179; GFX9-NEXT:    ;;#ASMEND
6180; GFX9-NEXT:    s_setpc_b64 s[30:31]
6181  %vec0 = call <3 x i64> asm "; def $0", "=s"()
6182  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 poison, i32 1, i32 1>
6183  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
6184  ret void
6185}
6186
6187define void @s_shuffle_v3i64_v3i64__0_1_1() {
6188; GFX9-LABEL: s_shuffle_v3i64_v3i64__0_1_1:
6189; GFX9:       ; %bb.0:
6190; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6191; GFX9-NEXT:    ;;#ASMSTART
6192; GFX9-NEXT:    ; def s[8:13]
6193; GFX9-NEXT:    ;;#ASMEND
6194; GFX9-NEXT:    s_mov_b32 s12, s10
6195; GFX9-NEXT:    s_mov_b32 s13, s11
6196; GFX9-NEXT:    ;;#ASMSTART
6197; GFX9-NEXT:    ; use s[8:13]
6198; GFX9-NEXT:    ;;#ASMEND
6199; GFX9-NEXT:    s_setpc_b64 s[30:31]
6200  %vec0 = call <3 x i64> asm "; def $0", "=s"()
6201  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 0, i32 1, i32 1>
6202  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
6203  ret void
6204}
6205
6206define void @s_shuffle_v3i64_v3i64__1_1_1() {
6207; GFX9-LABEL: s_shuffle_v3i64_v3i64__1_1_1:
6208; GFX9:       ; %bb.0:
6209; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6210; GFX9-NEXT:    ;;#ASMSTART
6211; GFX9-NEXT:    ; def s[8:13]
6212; GFX9-NEXT:    ;;#ASMEND
6213; GFX9-NEXT:    s_mov_b32 s8, s10
6214; GFX9-NEXT:    s_mov_b32 s9, s11
6215; GFX9-NEXT:    s_mov_b32 s12, s10
6216; GFX9-NEXT:    s_mov_b32 s13, s11
6217; GFX9-NEXT:    ;;#ASMSTART
6218; GFX9-NEXT:    ; use s[8:13]
6219; GFX9-NEXT:    ;;#ASMEND
6220; GFX9-NEXT:    s_setpc_b64 s[30:31]
6221  %vec0 = call <3 x i64> asm "; def $0", "=s"()
6222  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 1, i32 1, i32 1>
6223  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
6224  ret void
6225}
6226
6227define void @s_shuffle_v3i64_v3i64__2_1_1() {
6228; GFX9-LABEL: s_shuffle_v3i64_v3i64__2_1_1:
6229; GFX9:       ; %bb.0:
6230; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6231; GFX9-NEXT:    ;;#ASMSTART
6232; GFX9-NEXT:    ; def s[8:13]
6233; GFX9-NEXT:    ;;#ASMEND
6234; GFX9-NEXT:    s_mov_b32 s8, s12
6235; GFX9-NEXT:    s_mov_b32 s9, s13
6236; GFX9-NEXT:    s_mov_b32 s12, s10
6237; GFX9-NEXT:    s_mov_b32 s13, s11
6238; GFX9-NEXT:    ;;#ASMSTART
6239; GFX9-NEXT:    ; use s[8:13]
6240; GFX9-NEXT:    ;;#ASMEND
6241; GFX9-NEXT:    s_setpc_b64 s[30:31]
6242  %vec0 = call <3 x i64> asm "; def $0", "=s"()
6243  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 2, i32 1, i32 1>
6244  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
6245  ret void
6246}
6247
6248define void @s_shuffle_v3i64_v3i64__3_1_1() {
6249; GFX9-LABEL: s_shuffle_v3i64_v3i64__3_1_1:
6250; GFX9:       ; %bb.0:
6251; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6252; GFX9-NEXT:    ;;#ASMSTART
6253; GFX9-NEXT:    ; def s[8:13]
6254; GFX9-NEXT:    ;;#ASMEND
6255; GFX9-NEXT:    s_mov_b32 s12, s10
6256; GFX9-NEXT:    s_mov_b32 s13, s11
6257; GFX9-NEXT:    ;;#ASMSTART
6258; GFX9-NEXT:    ; use s[8:13]
6259; GFX9-NEXT:    ;;#ASMEND
6260; GFX9-NEXT:    s_setpc_b64 s[30:31]
6261  %vec0 = call <3 x i64> asm "; def $0", "=s"()
6262  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 3, i32 1, i32 1>
6263  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
6264  ret void
6265}
6266
6267define void @s_shuffle_v3i64_v3i64__4_1_1() {
6268; GFX900-LABEL: s_shuffle_v3i64_v3i64__4_1_1:
6269; GFX900:       ; %bb.0:
6270; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6271; GFX900-NEXT:    ;;#ASMSTART
6272; GFX900-NEXT:    ; def s[8:13]
6273; GFX900-NEXT:    ;;#ASMEND
6274; GFX900-NEXT:    ;;#ASMSTART
6275; GFX900-NEXT:    ; def s[4:9]
6276; GFX900-NEXT:    ;;#ASMEND
6277; GFX900-NEXT:    s_mov_b32 s8, s6
6278; GFX900-NEXT:    s_mov_b32 s9, s7
6279; GFX900-NEXT:    s_mov_b32 s12, s10
6280; GFX900-NEXT:    s_mov_b32 s13, s11
6281; GFX900-NEXT:    ;;#ASMSTART
6282; GFX900-NEXT:    ; use s[8:13]
6283; GFX900-NEXT:    ;;#ASMEND
6284; GFX900-NEXT:    s_setpc_b64 s[30:31]
6285;
6286; GFX90A-LABEL: s_shuffle_v3i64_v3i64__4_1_1:
6287; GFX90A:       ; %bb.0:
6288; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6289; GFX90A-NEXT:    ;;#ASMSTART
6290; GFX90A-NEXT:    ; def s[8:13]
6291; GFX90A-NEXT:    ;;#ASMEND
6292; GFX90A-NEXT:    ;;#ASMSTART
6293; GFX90A-NEXT:    ; def s[4:9]
6294; GFX90A-NEXT:    ;;#ASMEND
6295; GFX90A-NEXT:    s_mov_b32 s8, s6
6296; GFX90A-NEXT:    s_mov_b32 s9, s7
6297; GFX90A-NEXT:    s_mov_b32 s12, s10
6298; GFX90A-NEXT:    s_mov_b32 s13, s11
6299; GFX90A-NEXT:    ;;#ASMSTART
6300; GFX90A-NEXT:    ; use s[8:13]
6301; GFX90A-NEXT:    ;;#ASMEND
6302; GFX90A-NEXT:    s_setpc_b64 s[30:31]
6303;
6304; GFX940-LABEL: s_shuffle_v3i64_v3i64__4_1_1:
6305; GFX940:       ; %bb.0:
6306; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6307; GFX940-NEXT:    ;;#ASMSTART
6308; GFX940-NEXT:    ; def s[8:13]
6309; GFX940-NEXT:    ;;#ASMEND
6310; GFX940-NEXT:    ;;#ASMSTART
6311; GFX940-NEXT:    ; def s[0:5]
6312; GFX940-NEXT:    ;;#ASMEND
6313; GFX940-NEXT:    s_mov_b32 s8, s2
6314; GFX940-NEXT:    s_mov_b32 s9, s3
6315; GFX940-NEXT:    s_mov_b32 s12, s10
6316; GFX940-NEXT:    s_mov_b32 s13, s11
6317; GFX940-NEXT:    ;;#ASMSTART
6318; GFX940-NEXT:    ; use s[8:13]
6319; GFX940-NEXT:    ;;#ASMEND
6320; GFX940-NEXT:    s_setpc_b64 s[30:31]
6321  %vec0 = call <3 x i64> asm "; def $0", "=s"()
6322  %vec1 = call <3 x i64> asm "; def $0", "=s"()
6323  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 4, i32 1, i32 1>
6324  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
6325  ret void
6326}
6327
6328define void @s_shuffle_v3i64_v3i64__5_1_1() {
6329; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_1_1:
6330; GFX900:       ; %bb.0:
6331; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6332; GFX900-NEXT:    ;;#ASMSTART
6333; GFX900-NEXT:    ; def s[8:13]
6334; GFX900-NEXT:    ;;#ASMEND
6335; GFX900-NEXT:    ;;#ASMSTART
6336; GFX900-NEXT:    ; def s[4:9]
6337; GFX900-NEXT:    ;;#ASMEND
6338; GFX900-NEXT:    s_mov_b32 s12, s10
6339; GFX900-NEXT:    s_mov_b32 s13, s11
6340; GFX900-NEXT:    ;;#ASMSTART
6341; GFX900-NEXT:    ; use s[8:13]
6342; GFX900-NEXT:    ;;#ASMEND
6343; GFX900-NEXT:    s_setpc_b64 s[30:31]
6344;
6345; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_1_1:
6346; GFX90A:       ; %bb.0:
6347; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6348; GFX90A-NEXT:    ;;#ASMSTART
6349; GFX90A-NEXT:    ; def s[8:13]
6350; GFX90A-NEXT:    ;;#ASMEND
6351; GFX90A-NEXT:    ;;#ASMSTART
6352; GFX90A-NEXT:    ; def s[4:9]
6353; GFX90A-NEXT:    ;;#ASMEND
6354; GFX90A-NEXT:    s_mov_b32 s12, s10
6355; GFX90A-NEXT:    s_mov_b32 s13, s11
6356; GFX90A-NEXT:    ;;#ASMSTART
6357; GFX90A-NEXT:    ; use s[8:13]
6358; GFX90A-NEXT:    ;;#ASMEND
6359; GFX90A-NEXT:    s_setpc_b64 s[30:31]
6360;
6361; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_1_1:
6362; GFX940:       ; %bb.0:
6363; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6364; GFX940-NEXT:    ;;#ASMSTART
6365; GFX940-NEXT:    ; def s[8:13]
6366; GFX940-NEXT:    ;;#ASMEND
6367; GFX940-NEXT:    ;;#ASMSTART
6368; GFX940-NEXT:    ; def s[0:5]
6369; GFX940-NEXT:    ;;#ASMEND
6370; GFX940-NEXT:    s_mov_b32 s8, s4
6371; GFX940-NEXT:    s_mov_b32 s9, s5
6372; GFX940-NEXT:    s_mov_b32 s12, s10
6373; GFX940-NEXT:    s_mov_b32 s13, s11
6374; GFX940-NEXT:    ;;#ASMSTART
6375; GFX940-NEXT:    ; use s[8:13]
6376; GFX940-NEXT:    ;;#ASMEND
6377; GFX940-NEXT:    s_setpc_b64 s[30:31]
6378  %vec0 = call <3 x i64> asm "; def $0", "=s"()
6379  %vec1 = call <3 x i64> asm "; def $0", "=s"()
6380  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 1, i32 1>
6381  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
6382  ret void
6383}
6384
6385define void @s_shuffle_v3i64_v3i64__5_u_1() {
6386; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_u_1:
6387; GFX900:       ; %bb.0:
6388; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6389; GFX900-NEXT:    ;;#ASMSTART
6390; GFX900-NEXT:    ; def s[4:9]
6391; GFX900-NEXT:    ;;#ASMEND
6392; GFX900-NEXT:    ;;#ASMSTART
6393; GFX900-NEXT:    ; def s[8:13]
6394; GFX900-NEXT:    ;;#ASMEND
6395; GFX900-NEXT:    s_mov_b32 s8, s12
6396; GFX900-NEXT:    s_mov_b32 s9, s13
6397; GFX900-NEXT:    s_mov_b32 s12, s6
6398; GFX900-NEXT:    s_mov_b32 s13, s7
6399; GFX900-NEXT:    ;;#ASMSTART
6400; GFX900-NEXT:    ; use s[8:13]
6401; GFX900-NEXT:    ;;#ASMEND
6402; GFX900-NEXT:    s_setpc_b64 s[30:31]
6403;
6404; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_u_1:
6405; GFX90A:       ; %bb.0:
6406; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6407; GFX90A-NEXT:    ;;#ASMSTART
6408; GFX90A-NEXT:    ; def s[4:9]
6409; GFX90A-NEXT:    ;;#ASMEND
6410; GFX90A-NEXT:    ;;#ASMSTART
6411; GFX90A-NEXT:    ; def s[8:13]
6412; GFX90A-NEXT:    ;;#ASMEND
6413; GFX90A-NEXT:    s_mov_b32 s8, s12
6414; GFX90A-NEXT:    s_mov_b32 s9, s13
6415; GFX90A-NEXT:    s_mov_b32 s12, s6
6416; GFX90A-NEXT:    s_mov_b32 s13, s7
6417; GFX90A-NEXT:    ;;#ASMSTART
6418; GFX90A-NEXT:    ; use s[8:13]
6419; GFX90A-NEXT:    ;;#ASMEND
6420; GFX90A-NEXT:    s_setpc_b64 s[30:31]
6421;
6422; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_u_1:
6423; GFX940:       ; %bb.0:
6424; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6425; GFX940-NEXT:    ;;#ASMSTART
6426; GFX940-NEXT:    ; def s[0:5]
6427; GFX940-NEXT:    ;;#ASMEND
6428; GFX940-NEXT:    s_mov_b32 s12, s2
6429; GFX940-NEXT:    ;;#ASMSTART
6430; GFX940-NEXT:    ; def s[4:9]
6431; GFX940-NEXT:    ;;#ASMEND
6432; GFX940-NEXT:    s_mov_b32 s13, s3
6433; GFX940-NEXT:    ;;#ASMSTART
6434; GFX940-NEXT:    ; use s[8:13]
6435; GFX940-NEXT:    ;;#ASMEND
6436; GFX940-NEXT:    s_setpc_b64 s[30:31]
6437  %vec0 = call <3 x i64> asm "; def $0", "=s"()
6438  %vec1 = call <3 x i64> asm "; def $0", "=s"()
6439  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 poison, i32 1>
6440  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
6441  ret void
6442}
6443
6444define void @s_shuffle_v3i64_v3i64__5_0_1() {
6445; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_0_1:
6446; GFX900:       ; %bb.0:
6447; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6448; GFX900-NEXT:    ;;#ASMSTART
6449; GFX900-NEXT:    ; def s[4:9]
6450; GFX900-NEXT:    ;;#ASMEND
6451; GFX900-NEXT:    ;;#ASMSTART
6452; GFX900-NEXT:    ; def s[8:13]
6453; GFX900-NEXT:    ;;#ASMEND
6454; GFX900-NEXT:    s_mov_b32 s8, s12
6455; GFX900-NEXT:    s_mov_b32 s9, s13
6456; GFX900-NEXT:    s_mov_b32 s10, s4
6457; GFX900-NEXT:    s_mov_b32 s11, s5
6458; GFX900-NEXT:    s_mov_b32 s12, s6
6459; GFX900-NEXT:    s_mov_b32 s13, s7
6460; GFX900-NEXT:    ;;#ASMSTART
6461; GFX900-NEXT:    ; use s[8:13]
6462; GFX900-NEXT:    ;;#ASMEND
6463; GFX900-NEXT:    s_setpc_b64 s[30:31]
6464;
6465; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_0_1:
6466; GFX90A:       ; %bb.0:
6467; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6468; GFX90A-NEXT:    ;;#ASMSTART
6469; GFX90A-NEXT:    ; def s[4:9]
6470; GFX90A-NEXT:    ;;#ASMEND
6471; GFX90A-NEXT:    ;;#ASMSTART
6472; GFX90A-NEXT:    ; def s[8:13]
6473; GFX90A-NEXT:    ;;#ASMEND
6474; GFX90A-NEXT:    s_mov_b32 s8, s12
6475; GFX90A-NEXT:    s_mov_b32 s9, s13
6476; GFX90A-NEXT:    s_mov_b32 s10, s4
6477; GFX90A-NEXT:    s_mov_b32 s11, s5
6478; GFX90A-NEXT:    s_mov_b32 s12, s6
6479; GFX90A-NEXT:    s_mov_b32 s13, s7
6480; GFX90A-NEXT:    ;;#ASMSTART
6481; GFX90A-NEXT:    ; use s[8:13]
6482; GFX90A-NEXT:    ;;#ASMEND
6483; GFX90A-NEXT:    s_setpc_b64 s[30:31]
6484;
6485; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_0_1:
6486; GFX940:       ; %bb.0:
6487; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6488; GFX940-NEXT:    ;;#ASMSTART
6489; GFX940-NEXT:    ; def s[0:5]
6490; GFX940-NEXT:    ;;#ASMEND
6491; GFX940-NEXT:    s_mov_b32 s10, s0
6492; GFX940-NEXT:    ;;#ASMSTART
6493; GFX940-NEXT:    ; def s[4:9]
6494; GFX940-NEXT:    ;;#ASMEND
6495; GFX940-NEXT:    s_mov_b32 s11, s1
6496; GFX940-NEXT:    s_mov_b32 s12, s2
6497; GFX940-NEXT:    s_mov_b32 s13, s3
6498; GFX940-NEXT:    ;;#ASMSTART
6499; GFX940-NEXT:    ; use s[8:13]
6500; GFX940-NEXT:    ;;#ASMEND
6501; GFX940-NEXT:    s_setpc_b64 s[30:31]
6502  %vec0 = call <3 x i64> asm "; def $0", "=s"()
6503  %vec1 = call <3 x i64> asm "; def $0", "=s"()
6504  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 0, i32 1>
6505  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
6506  ret void
6507}
6508
6509define void @s_shuffle_v3i64_v3i64__5_2_1() {
6510; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_2_1:
6511; GFX900:       ; %bb.0:
6512; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6513; GFX900-NEXT:    ;;#ASMSTART
6514; GFX900-NEXT:    ; def s[12:17]
6515; GFX900-NEXT:    ;;#ASMEND
6516; GFX900-NEXT:    ;;#ASMSTART
6517; GFX900-NEXT:    ; def s[4:9]
6518; GFX900-NEXT:    ;;#ASMEND
6519; GFX900-NEXT:    s_mov_b32 s10, s16
6520; GFX900-NEXT:    s_mov_b32 s11, s17
6521; GFX900-NEXT:    s_mov_b32 s12, s14
6522; GFX900-NEXT:    s_mov_b32 s13, s15
6523; GFX900-NEXT:    ;;#ASMSTART
6524; GFX900-NEXT:    ; use s[8:13]
6525; GFX900-NEXT:    ;;#ASMEND
6526; GFX900-NEXT:    s_setpc_b64 s[30:31]
6527;
6528; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_2_1:
6529; GFX90A:       ; %bb.0:
6530; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6531; GFX90A-NEXT:    ;;#ASMSTART
6532; GFX90A-NEXT:    ; def s[12:17]
6533; GFX90A-NEXT:    ;;#ASMEND
6534; GFX90A-NEXT:    ;;#ASMSTART
6535; GFX90A-NEXT:    ; def s[4:9]
6536; GFX90A-NEXT:    ;;#ASMEND
6537; GFX90A-NEXT:    s_mov_b32 s10, s16
6538; GFX90A-NEXT:    s_mov_b32 s11, s17
6539; GFX90A-NEXT:    s_mov_b32 s12, s14
6540; GFX90A-NEXT:    s_mov_b32 s13, s15
6541; GFX90A-NEXT:    ;;#ASMSTART
6542; GFX90A-NEXT:    ; use s[8:13]
6543; GFX90A-NEXT:    ;;#ASMEND
6544; GFX90A-NEXT:    s_setpc_b64 s[30:31]
6545;
6546; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_2_1:
6547; GFX940:       ; %bb.0:
6548; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6549; GFX940-NEXT:    ;;#ASMSTART
6550; GFX940-NEXT:    ; def s[8:13]
6551; GFX940-NEXT:    ;;#ASMEND
6552; GFX940-NEXT:    ;;#ASMSTART
6553; GFX940-NEXT:    ; def s[0:5]
6554; GFX940-NEXT:    ;;#ASMEND
6555; GFX940-NEXT:    s_mov_b32 s8, s12
6556; GFX940-NEXT:    s_mov_b32 s9, s13
6557; GFX940-NEXT:    s_mov_b32 s10, s4
6558; GFX940-NEXT:    s_mov_b32 s11, s5
6559; GFX940-NEXT:    s_mov_b32 s12, s2
6560; GFX940-NEXT:    s_mov_b32 s13, s3
6561; GFX940-NEXT:    ;;#ASMSTART
6562; GFX940-NEXT:    ; use s[8:13]
6563; GFX940-NEXT:    ;;#ASMEND
6564; GFX940-NEXT:    s_setpc_b64 s[30:31]
6565  %vec0 = call <3 x i64> asm "; def $0", "=s"()
6566  %vec1 = call <3 x i64> asm "; def $0", "=s"()
6567  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 2, i32 1>
6568  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
6569  ret void
6570}
6571
6572define void @s_shuffle_v3i64_v3i64__5_3_1() {
6573; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_3_1:
6574; GFX900:       ; %bb.0:
6575; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6576; GFX900-NEXT:    ;;#ASMSTART
6577; GFX900-NEXT:    ; def s[4:9]
6578; GFX900-NEXT:    ;;#ASMEND
6579; GFX900-NEXT:    ;;#ASMSTART
6580; GFX900-NEXT:    ; def s[12:17]
6581; GFX900-NEXT:    ;;#ASMEND
6582; GFX900-NEXT:    s_mov_b32 s8, s16
6583; GFX900-NEXT:    s_mov_b32 s9, s17
6584; GFX900-NEXT:    s_mov_b32 s10, s12
6585; GFX900-NEXT:    s_mov_b32 s11, s13
6586; GFX900-NEXT:    s_mov_b32 s12, s6
6587; GFX900-NEXT:    s_mov_b32 s13, s7
6588; GFX900-NEXT:    ;;#ASMSTART
6589; GFX900-NEXT:    ; use s[8:13]
6590; GFX900-NEXT:    ;;#ASMEND
6591; GFX900-NEXT:    s_setpc_b64 s[30:31]
6592;
6593; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_3_1:
6594; GFX90A:       ; %bb.0:
6595; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6596; GFX90A-NEXT:    ;;#ASMSTART
6597; GFX90A-NEXT:    ; def s[4:9]
6598; GFX90A-NEXT:    ;;#ASMEND
6599; GFX90A-NEXT:    ;;#ASMSTART
6600; GFX90A-NEXT:    ; def s[12:17]
6601; GFX90A-NEXT:    ;;#ASMEND
6602; GFX90A-NEXT:    s_mov_b32 s8, s16
6603; GFX90A-NEXT:    s_mov_b32 s9, s17
6604; GFX90A-NEXT:    s_mov_b32 s10, s12
6605; GFX90A-NEXT:    s_mov_b32 s11, s13
6606; GFX90A-NEXT:    s_mov_b32 s12, s6
6607; GFX90A-NEXT:    s_mov_b32 s13, s7
6608; GFX90A-NEXT:    ;;#ASMSTART
6609; GFX90A-NEXT:    ; use s[8:13]
6610; GFX90A-NEXT:    ;;#ASMEND
6611; GFX90A-NEXT:    s_setpc_b64 s[30:31]
6612;
6613; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_3_1:
6614; GFX940:       ; %bb.0:
6615; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6616; GFX940-NEXT:    ;;#ASMSTART
6617; GFX940-NEXT:    ; def s[0:5]
6618; GFX940-NEXT:    ;;#ASMEND
6619; GFX940-NEXT:    s_mov_b32 s12, s2
6620; GFX940-NEXT:    ;;#ASMSTART
6621; GFX940-NEXT:    ; def s[4:9]
6622; GFX940-NEXT:    ;;#ASMEND
6623; GFX940-NEXT:    s_mov_b32 s10, s4
6624; GFX940-NEXT:    s_mov_b32 s11, s5
6625; GFX940-NEXT:    s_mov_b32 s13, s3
6626; GFX940-NEXT:    ;;#ASMSTART
6627; GFX940-NEXT:    ; use s[8:13]
6628; GFX940-NEXT:    ;;#ASMEND
6629; GFX940-NEXT:    s_setpc_b64 s[30:31]
6630  %vec0 = call <3 x i64> asm "; def $0", "=s"()
6631  %vec1 = call <3 x i64> asm "; def $0", "=s"()
6632  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 3, i32 1>
6633  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
6634  ret void
6635}
6636
6637define void @s_shuffle_v3i64_v3i64__5_4_1() {
6638; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_4_1:
6639; GFX900:       ; %bb.0:
6640; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6641; GFX900-NEXT:    ;;#ASMSTART
6642; GFX900-NEXT:    ; def s[4:9]
6643; GFX900-NEXT:    ;;#ASMEND
6644; GFX900-NEXT:    ;;#ASMSTART
6645; GFX900-NEXT:    ; def s[8:13]
6646; GFX900-NEXT:    ;;#ASMEND
6647; GFX900-NEXT:    s_mov_b32 s8, s12
6648; GFX900-NEXT:    s_mov_b32 s9, s13
6649; GFX900-NEXT:    s_mov_b32 s12, s6
6650; GFX900-NEXT:    s_mov_b32 s13, s7
6651; GFX900-NEXT:    ;;#ASMSTART
6652; GFX900-NEXT:    ; use s[8:13]
6653; GFX900-NEXT:    ;;#ASMEND
6654; GFX900-NEXT:    s_setpc_b64 s[30:31]
6655;
6656; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_4_1:
6657; GFX90A:       ; %bb.0:
6658; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6659; GFX90A-NEXT:    ;;#ASMSTART
6660; GFX90A-NEXT:    ; def s[4:9]
6661; GFX90A-NEXT:    ;;#ASMEND
6662; GFX90A-NEXT:    ;;#ASMSTART
6663; GFX90A-NEXT:    ; def s[8:13]
6664; GFX90A-NEXT:    ;;#ASMEND
6665; GFX90A-NEXT:    s_mov_b32 s8, s12
6666; GFX90A-NEXT:    s_mov_b32 s9, s13
6667; GFX90A-NEXT:    s_mov_b32 s12, s6
6668; GFX90A-NEXT:    s_mov_b32 s13, s7
6669; GFX90A-NEXT:    ;;#ASMSTART
6670; GFX90A-NEXT:    ; use s[8:13]
6671; GFX90A-NEXT:    ;;#ASMEND
6672; GFX90A-NEXT:    s_setpc_b64 s[30:31]
6673;
6674; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_4_1:
6675; GFX940:       ; %bb.0:
6676; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6677; GFX940-NEXT:    ;;#ASMSTART
6678; GFX940-NEXT:    ; def s[8:13]
6679; GFX940-NEXT:    ;;#ASMEND
6680; GFX940-NEXT:    ;;#ASMSTART
6681; GFX940-NEXT:    ; def s[0:5]
6682; GFX940-NEXT:    ;;#ASMEND
6683; GFX940-NEXT:    s_mov_b32 s8, s12
6684; GFX940-NEXT:    s_mov_b32 s9, s13
6685; GFX940-NEXT:    s_mov_b32 s12, s2
6686; GFX940-NEXT:    s_mov_b32 s13, s3
6687; GFX940-NEXT:    ;;#ASMSTART
6688; GFX940-NEXT:    ; use s[8:13]
6689; GFX940-NEXT:    ;;#ASMEND
6690; GFX940-NEXT:    s_setpc_b64 s[30:31]
6691  %vec0 = call <3 x i64> asm "; def $0", "=s"()
6692  %vec1 = call <3 x i64> asm "; def $0", "=s"()
6693  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 4, i32 1>
6694  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
6695  ret void
6696}
6697
6698define void @s_shuffle_v3i64_v3i64__u_2_2() {
6699; GFX9-LABEL: s_shuffle_v3i64_v3i64__u_2_2:
6700; GFX9:       ; %bb.0:
6701; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6702; GFX9-NEXT:    ;;#ASMSTART
6703; GFX9-NEXT:    ; def s[8:13]
6704; GFX9-NEXT:    ;;#ASMEND
6705; GFX9-NEXT:    s_mov_b32 s10, s12
6706; GFX9-NEXT:    s_mov_b32 s11, s13
6707; GFX9-NEXT:    ;;#ASMSTART
6708; GFX9-NEXT:    ; use s[8:13]
6709; GFX9-NEXT:    ;;#ASMEND
6710; GFX9-NEXT:    s_setpc_b64 s[30:31]
6711  %vec0 = call <3 x i64> asm "; def $0", "=s"()
6712  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 poison, i32 2, i32 2>
6713  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
6714  ret void
6715}
6716
6717define void @s_shuffle_v3i64_v3i64__0_2_2() {
6718; GFX9-LABEL: s_shuffle_v3i64_v3i64__0_2_2:
6719; GFX9:       ; %bb.0:
6720; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6721; GFX9-NEXT:    ;;#ASMSTART
6722; GFX9-NEXT:    ; def s[8:13]
6723; GFX9-NEXT:    ;;#ASMEND
6724; GFX9-NEXT:    s_mov_b32 s10, s12
6725; GFX9-NEXT:    s_mov_b32 s11, s13
6726; GFX9-NEXT:    ;;#ASMSTART
6727; GFX9-NEXT:    ; use s[8:13]
6728; GFX9-NEXT:    ;;#ASMEND
6729; GFX9-NEXT:    s_setpc_b64 s[30:31]
6730  %vec0 = call <3 x i64> asm "; def $0", "=s"()
6731  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 0, i32 2, i32 2>
6732  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
6733  ret void
6734}
6735
6736define void @s_shuffle_v3i64_v3i64__1_2_2() {
6737; GFX9-LABEL: s_shuffle_v3i64_v3i64__1_2_2:
6738; GFX9:       ; %bb.0:
6739; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6740; GFX9-NEXT:    ;;#ASMSTART
6741; GFX9-NEXT:    ; def s[8:13]
6742; GFX9-NEXT:    ;;#ASMEND
6743; GFX9-NEXT:    s_mov_b32 s8, s10
6744; GFX9-NEXT:    s_mov_b32 s9, s11
6745; GFX9-NEXT:    s_mov_b32 s10, s12
6746; GFX9-NEXT:    s_mov_b32 s11, s13
6747; GFX9-NEXT:    ;;#ASMSTART
6748; GFX9-NEXT:    ; use s[8:13]
6749; GFX9-NEXT:    ;;#ASMEND
6750; GFX9-NEXT:    s_setpc_b64 s[30:31]
6751  %vec0 = call <3 x i64> asm "; def $0", "=s"()
6752  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 1, i32 2, i32 2>
6753  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
6754  ret void
6755}
6756
6757define void @s_shuffle_v3i64_v3i64__2_2_2() {
6758; GFX9-LABEL: s_shuffle_v3i64_v3i64__2_2_2:
6759; GFX9:       ; %bb.0:
6760; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6761; GFX9-NEXT:    ;;#ASMSTART
6762; GFX9-NEXT:    ; def s[8:13]
6763; GFX9-NEXT:    ;;#ASMEND
6764; GFX9-NEXT:    s_mov_b32 s8, s12
6765; GFX9-NEXT:    s_mov_b32 s9, s13
6766; GFX9-NEXT:    s_mov_b32 s10, s12
6767; GFX9-NEXT:    s_mov_b32 s11, s13
6768; GFX9-NEXT:    ;;#ASMSTART
6769; GFX9-NEXT:    ; use s[8:13]
6770; GFX9-NEXT:    ;;#ASMEND
6771; GFX9-NEXT:    s_setpc_b64 s[30:31]
6772  %vec0 = call <3 x i64> asm "; def $0", "=s"()
6773  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 2, i32 2, i32 2>
6774  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
6775  ret void
6776}
6777
6778define void @s_shuffle_v3i64_v3i64__3_2_2() {
6779; GFX9-LABEL: s_shuffle_v3i64_v3i64__3_2_2:
6780; GFX9:       ; %bb.0:
6781; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6782; GFX9-NEXT:    ;;#ASMSTART
6783; GFX9-NEXT:    ; def s[8:13]
6784; GFX9-NEXT:    ;;#ASMEND
6785; GFX9-NEXT:    s_mov_b32 s10, s12
6786; GFX9-NEXT:    s_mov_b32 s11, s13
6787; GFX9-NEXT:    ;;#ASMSTART
6788; GFX9-NEXT:    ; use s[8:13]
6789; GFX9-NEXT:    ;;#ASMEND
6790; GFX9-NEXT:    s_setpc_b64 s[30:31]
6791  %vec0 = call <3 x i64> asm "; def $0", "=s"()
6792  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 3, i32 2, i32 2>
6793  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
6794  ret void
6795}
6796
6797define void @s_shuffle_v3i64_v3i64__4_2_2() {
6798; GFX900-LABEL: s_shuffle_v3i64_v3i64__4_2_2:
6799; GFX900:       ; %bb.0:
6800; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6801; GFX900-NEXT:    ;;#ASMSTART
6802; GFX900-NEXT:    ; def s[8:13]
6803; GFX900-NEXT:    ;;#ASMEND
6804; GFX900-NEXT:    ;;#ASMSTART
6805; GFX900-NEXT:    ; def s[4:9]
6806; GFX900-NEXT:    ;;#ASMEND
6807; GFX900-NEXT:    s_mov_b32 s8, s6
6808; GFX900-NEXT:    s_mov_b32 s9, s7
6809; GFX900-NEXT:    s_mov_b32 s10, s12
6810; GFX900-NEXT:    s_mov_b32 s11, s13
6811; GFX900-NEXT:    ;;#ASMSTART
6812; GFX900-NEXT:    ; use s[8:13]
6813; GFX900-NEXT:    ;;#ASMEND
6814; GFX900-NEXT:    s_setpc_b64 s[30:31]
6815;
6816; GFX90A-LABEL: s_shuffle_v3i64_v3i64__4_2_2:
6817; GFX90A:       ; %bb.0:
6818; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6819; GFX90A-NEXT:    ;;#ASMSTART
6820; GFX90A-NEXT:    ; def s[8:13]
6821; GFX90A-NEXT:    ;;#ASMEND
6822; GFX90A-NEXT:    ;;#ASMSTART
6823; GFX90A-NEXT:    ; def s[4:9]
6824; GFX90A-NEXT:    ;;#ASMEND
6825; GFX90A-NEXT:    s_mov_b32 s8, s6
6826; GFX90A-NEXT:    s_mov_b32 s9, s7
6827; GFX90A-NEXT:    s_mov_b32 s10, s12
6828; GFX90A-NEXT:    s_mov_b32 s11, s13
6829; GFX90A-NEXT:    ;;#ASMSTART
6830; GFX90A-NEXT:    ; use s[8:13]
6831; GFX90A-NEXT:    ;;#ASMEND
6832; GFX90A-NEXT:    s_setpc_b64 s[30:31]
6833;
6834; GFX940-LABEL: s_shuffle_v3i64_v3i64__4_2_2:
6835; GFX940:       ; %bb.0:
6836; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6837; GFX940-NEXT:    ;;#ASMSTART
6838; GFX940-NEXT:    ; def s[8:13]
6839; GFX940-NEXT:    ;;#ASMEND
6840; GFX940-NEXT:    ;;#ASMSTART
6841; GFX940-NEXT:    ; def s[0:5]
6842; GFX940-NEXT:    ;;#ASMEND
6843; GFX940-NEXT:    s_mov_b32 s8, s2
6844; GFX940-NEXT:    s_mov_b32 s9, s3
6845; GFX940-NEXT:    s_mov_b32 s10, s12
6846; GFX940-NEXT:    s_mov_b32 s11, s13
6847; GFX940-NEXT:    ;;#ASMSTART
6848; GFX940-NEXT:    ; use s[8:13]
6849; GFX940-NEXT:    ;;#ASMEND
6850; GFX940-NEXT:    s_setpc_b64 s[30:31]
6851  %vec0 = call <3 x i64> asm "; def $0", "=s"()
6852  %vec1 = call <3 x i64> asm "; def $0", "=s"()
6853  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 4, i32 2, i32 2>
6854  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
6855  ret void
6856}
6857
6858define void @s_shuffle_v3i64_v3i64__5_2_2() {
6859; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_2_2:
6860; GFX900:       ; %bb.0:
6861; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6862; GFX900-NEXT:    ;;#ASMSTART
6863; GFX900-NEXT:    ; def s[8:13]
6864; GFX900-NEXT:    ;;#ASMEND
6865; GFX900-NEXT:    ;;#ASMSTART
6866; GFX900-NEXT:    ; def s[4:9]
6867; GFX900-NEXT:    ;;#ASMEND
6868; GFX900-NEXT:    s_mov_b32 s10, s12
6869; GFX900-NEXT:    s_mov_b32 s11, s13
6870; GFX900-NEXT:    ;;#ASMSTART
6871; GFX900-NEXT:    ; use s[8:13]
6872; GFX900-NEXT:    ;;#ASMEND
6873; GFX900-NEXT:    s_setpc_b64 s[30:31]
6874;
6875; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_2_2:
6876; GFX90A:       ; %bb.0:
6877; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6878; GFX90A-NEXT:    ;;#ASMSTART
6879; GFX90A-NEXT:    ; def s[8:13]
6880; GFX90A-NEXT:    ;;#ASMEND
6881; GFX90A-NEXT:    ;;#ASMSTART
6882; GFX90A-NEXT:    ; def s[4:9]
6883; GFX90A-NEXT:    ;;#ASMEND
6884; GFX90A-NEXT:    s_mov_b32 s10, s12
6885; GFX90A-NEXT:    s_mov_b32 s11, s13
6886; GFX90A-NEXT:    ;;#ASMSTART
6887; GFX90A-NEXT:    ; use s[8:13]
6888; GFX90A-NEXT:    ;;#ASMEND
6889; GFX90A-NEXT:    s_setpc_b64 s[30:31]
6890;
6891; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_2_2:
6892; GFX940:       ; %bb.0:
6893; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6894; GFX940-NEXT:    ;;#ASMSTART
6895; GFX940-NEXT:    ; def s[8:13]
6896; GFX940-NEXT:    ;;#ASMEND
6897; GFX940-NEXT:    ;;#ASMSTART
6898; GFX940-NEXT:    ; def s[0:5]
6899; GFX940-NEXT:    ;;#ASMEND
6900; GFX940-NEXT:    s_mov_b32 s8, s4
6901; GFX940-NEXT:    s_mov_b32 s9, s5
6902; GFX940-NEXT:    s_mov_b32 s10, s12
6903; GFX940-NEXT:    s_mov_b32 s11, s13
6904; GFX940-NEXT:    ;;#ASMSTART
6905; GFX940-NEXT:    ; use s[8:13]
6906; GFX940-NEXT:    ;;#ASMEND
6907; GFX940-NEXT:    s_setpc_b64 s[30:31]
6908  %vec0 = call <3 x i64> asm "; def $0", "=s"()
6909  %vec1 = call <3 x i64> asm "; def $0", "=s"()
6910  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 2, i32 2>
6911  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
6912  ret void
6913}
6914
6915define void @s_shuffle_v3i64_v3i64__5_u_2() {
6916; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_u_2:
6917; GFX900:       ; %bb.0:
6918; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6919; GFX900-NEXT:    ;;#ASMSTART
6920; GFX900-NEXT:    ; def s[8:13]
6921; GFX900-NEXT:    ;;#ASMEND
6922; GFX900-NEXT:    ;;#ASMSTART
6923; GFX900-NEXT:    ; def s[4:9]
6924; GFX900-NEXT:    ;;#ASMEND
6925; GFX900-NEXT:    ;;#ASMSTART
6926; GFX900-NEXT:    ; use s[8:13]
6927; GFX900-NEXT:    ;;#ASMEND
6928; GFX900-NEXT:    s_setpc_b64 s[30:31]
6929;
6930; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_u_2:
6931; GFX90A:       ; %bb.0:
6932; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6933; GFX90A-NEXT:    ;;#ASMSTART
6934; GFX90A-NEXT:    ; def s[8:13]
6935; GFX90A-NEXT:    ;;#ASMEND
6936; GFX90A-NEXT:    ;;#ASMSTART
6937; GFX90A-NEXT:    ; def s[4:9]
6938; GFX90A-NEXT:    ;;#ASMEND
6939; GFX90A-NEXT:    ;;#ASMSTART
6940; GFX90A-NEXT:    ; use s[8:13]
6941; GFX90A-NEXT:    ;;#ASMEND
6942; GFX90A-NEXT:    s_setpc_b64 s[30:31]
6943;
6944; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_u_2:
6945; GFX940:       ; %bb.0:
6946; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6947; GFX940-NEXT:    ;;#ASMSTART
6948; GFX940-NEXT:    ; def s[8:13]
6949; GFX940-NEXT:    ;;#ASMEND
6950; GFX940-NEXT:    ;;#ASMSTART
6951; GFX940-NEXT:    ; def s[0:5]
6952; GFX940-NEXT:    ;;#ASMEND
6953; GFX940-NEXT:    s_mov_b32 s8, s4
6954; GFX940-NEXT:    s_mov_b32 s9, s5
6955; GFX940-NEXT:    ;;#ASMSTART
6956; GFX940-NEXT:    ; use s[8:13]
6957; GFX940-NEXT:    ;;#ASMEND
6958; GFX940-NEXT:    s_setpc_b64 s[30:31]
6959  %vec0 = call <3 x i64> asm "; def $0", "=s"()
6960  %vec1 = call <3 x i64> asm "; def $0", "=s"()
6961  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 poison, i32 2>
6962  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
6963  ret void
6964}
6965
6966define void @s_shuffle_v3i64_v3i64__5_0_2() {
6967; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_0_2:
6968; GFX900:       ; %bb.0:
6969; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6970; GFX900-NEXT:    ;;#ASMSTART
6971; GFX900-NEXT:    ; def s[12:17]
6972; GFX900-NEXT:    ;;#ASMEND
6973; GFX900-NEXT:    ;;#ASMSTART
6974; GFX900-NEXT:    ; def s[4:9]
6975; GFX900-NEXT:    ;;#ASMEND
6976; GFX900-NEXT:    s_mov_b32 s10, s12
6977; GFX900-NEXT:    s_mov_b32 s11, s13
6978; GFX900-NEXT:    s_mov_b32 s12, s16
6979; GFX900-NEXT:    s_mov_b32 s13, s17
6980; GFX900-NEXT:    ;;#ASMSTART
6981; GFX900-NEXT:    ; use s[8:13]
6982; GFX900-NEXT:    ;;#ASMEND
6983; GFX900-NEXT:    s_setpc_b64 s[30:31]
6984;
6985; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_0_2:
6986; GFX90A:       ; %bb.0:
6987; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
6988; GFX90A-NEXT:    ;;#ASMSTART
6989; GFX90A-NEXT:    ; def s[12:17]
6990; GFX90A-NEXT:    ;;#ASMEND
6991; GFX90A-NEXT:    ;;#ASMSTART
6992; GFX90A-NEXT:    ; def s[4:9]
6993; GFX90A-NEXT:    ;;#ASMEND
6994; GFX90A-NEXT:    s_mov_b32 s10, s12
6995; GFX90A-NEXT:    s_mov_b32 s11, s13
6996; GFX90A-NEXT:    s_mov_b32 s12, s16
6997; GFX90A-NEXT:    s_mov_b32 s13, s17
6998; GFX90A-NEXT:    ;;#ASMSTART
6999; GFX90A-NEXT:    ; use s[8:13]
7000; GFX90A-NEXT:    ;;#ASMEND
7001; GFX90A-NEXT:    s_setpc_b64 s[30:31]
7002;
7003; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_0_2:
7004; GFX940:       ; %bb.0:
7005; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7006; GFX940-NEXT:    ;;#ASMSTART
7007; GFX940-NEXT:    ; def s[8:13]
7008; GFX940-NEXT:    ;;#ASMEND
7009; GFX940-NEXT:    ;;#ASMSTART
7010; GFX940-NEXT:    ; def s[0:5]
7011; GFX940-NEXT:    ;;#ASMEND
7012; GFX940-NEXT:    s_mov_b32 s8, s12
7013; GFX940-NEXT:    s_mov_b32 s9, s13
7014; GFX940-NEXT:    s_mov_b32 s10, s0
7015; GFX940-NEXT:    s_mov_b32 s11, s1
7016; GFX940-NEXT:    s_mov_b32 s12, s4
7017; GFX940-NEXT:    s_mov_b32 s13, s5
7018; GFX940-NEXT:    ;;#ASMSTART
7019; GFX940-NEXT:    ; use s[8:13]
7020; GFX940-NEXT:    ;;#ASMEND
7021; GFX940-NEXT:    s_setpc_b64 s[30:31]
7022  %vec0 = call <3 x i64> asm "; def $0", "=s"()
7023  %vec1 = call <3 x i64> asm "; def $0", "=s"()
7024  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 0, i32 2>
7025  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
7026  ret void
7027}
7028
7029define void @s_shuffle_v3i64_v3i64__5_1_2() {
7030; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_1_2:
7031; GFX900:       ; %bb.0:
7032; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7033; GFX900-NEXT:    ;;#ASMSTART
7034; GFX900-NEXT:    ; def s[8:13]
7035; GFX900-NEXT:    ;;#ASMEND
7036; GFX900-NEXT:    ;;#ASMSTART
7037; GFX900-NEXT:    ; def s[4:9]
7038; GFX900-NEXT:    ;;#ASMEND
7039; GFX900-NEXT:    ;;#ASMSTART
7040; GFX900-NEXT:    ; use s[8:13]
7041; GFX900-NEXT:    ;;#ASMEND
7042; GFX900-NEXT:    s_setpc_b64 s[30:31]
7043;
7044; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_1_2:
7045; GFX90A:       ; %bb.0:
7046; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7047; GFX90A-NEXT:    ;;#ASMSTART
7048; GFX90A-NEXT:    ; def s[8:13]
7049; GFX90A-NEXT:    ;;#ASMEND
7050; GFX90A-NEXT:    ;;#ASMSTART
7051; GFX90A-NEXT:    ; def s[4:9]
7052; GFX90A-NEXT:    ;;#ASMEND
7053; GFX90A-NEXT:    ;;#ASMSTART
7054; GFX90A-NEXT:    ; use s[8:13]
7055; GFX90A-NEXT:    ;;#ASMEND
7056; GFX90A-NEXT:    s_setpc_b64 s[30:31]
7057;
7058; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_1_2:
7059; GFX940:       ; %bb.0:
7060; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7061; GFX940-NEXT:    ;;#ASMSTART
7062; GFX940-NEXT:    ; def s[8:13]
7063; GFX940-NEXT:    ;;#ASMEND
7064; GFX940-NEXT:    ;;#ASMSTART
7065; GFX940-NEXT:    ; def s[0:5]
7066; GFX940-NEXT:    ;;#ASMEND
7067; GFX940-NEXT:    s_mov_b32 s8, s4
7068; GFX940-NEXT:    s_mov_b32 s9, s5
7069; GFX940-NEXT:    ;;#ASMSTART
7070; GFX940-NEXT:    ; use s[8:13]
7071; GFX940-NEXT:    ;;#ASMEND
7072; GFX940-NEXT:    s_setpc_b64 s[30:31]
7073  %vec0 = call <3 x i64> asm "; def $0", "=s"()
7074  %vec1 = call <3 x i64> asm "; def $0", "=s"()
7075  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 1, i32 2>
7076  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
7077  ret void
7078}
7079
7080define void @s_shuffle_v3i64_v3i64__5_3_2() {
7081; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_3_2:
7082; GFX900:       ; %bb.0:
7083; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7084; GFX900-NEXT:    ;;#ASMSTART
7085; GFX900-NEXT:    ; def s[8:13]
7086; GFX900-NEXT:    ;;#ASMEND
7087; GFX900-NEXT:    ;;#ASMSTART
7088; GFX900-NEXT:    ; def s[4:9]
7089; GFX900-NEXT:    ;;#ASMEND
7090; GFX900-NEXT:    s_mov_b32 s10, s4
7091; GFX900-NEXT:    s_mov_b32 s11, s5
7092; GFX900-NEXT:    ;;#ASMSTART
7093; GFX900-NEXT:    ; use s[8:13]
7094; GFX900-NEXT:    ;;#ASMEND
7095; GFX900-NEXT:    s_setpc_b64 s[30:31]
7096;
7097; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_3_2:
7098; GFX90A:       ; %bb.0:
7099; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7100; GFX90A-NEXT:    ;;#ASMSTART
7101; GFX90A-NEXT:    ; def s[8:13]
7102; GFX90A-NEXT:    ;;#ASMEND
7103; GFX90A-NEXT:    ;;#ASMSTART
7104; GFX90A-NEXT:    ; def s[4:9]
7105; GFX90A-NEXT:    ;;#ASMEND
7106; GFX90A-NEXT:    s_mov_b32 s10, s4
7107; GFX90A-NEXT:    s_mov_b32 s11, s5
7108; GFX90A-NEXT:    ;;#ASMSTART
7109; GFX90A-NEXT:    ; use s[8:13]
7110; GFX90A-NEXT:    ;;#ASMEND
7111; GFX90A-NEXT:    s_setpc_b64 s[30:31]
7112;
7113; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_3_2:
7114; GFX940:       ; %bb.0:
7115; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7116; GFX940-NEXT:    ;;#ASMSTART
7117; GFX940-NEXT:    ; def s[8:13]
7118; GFX940-NEXT:    ;;#ASMEND
7119; GFX940-NEXT:    ;;#ASMSTART
7120; GFX940-NEXT:    ; def s[0:5]
7121; GFX940-NEXT:    ;;#ASMEND
7122; GFX940-NEXT:    s_mov_b32 s8, s4
7123; GFX940-NEXT:    s_mov_b32 s9, s5
7124; GFX940-NEXT:    s_mov_b32 s10, s0
7125; GFX940-NEXT:    s_mov_b32 s11, s1
7126; GFX940-NEXT:    ;;#ASMSTART
7127; GFX940-NEXT:    ; use s[8:13]
7128; GFX940-NEXT:    ;;#ASMEND
7129; GFX940-NEXT:    s_setpc_b64 s[30:31]
7130  %vec0 = call <3 x i64> asm "; def $0", "=s"()
7131  %vec1 = call <3 x i64> asm "; def $0", "=s"()
7132  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 3, i32 2>
7133  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
7134  ret void
7135}
7136
7137define void @s_shuffle_v3i64_v3i64__5_4_2() {
7138; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_4_2:
7139; GFX900:       ; %bb.0:
7140; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7141; GFX900-NEXT:    ;;#ASMSTART
7142; GFX900-NEXT:    ; def s[12:17]
7143; GFX900-NEXT:    ;;#ASMEND
7144; GFX900-NEXT:    ;;#ASMSTART
7145; GFX900-NEXT:    ; def s[8:13]
7146; GFX900-NEXT:    ;;#ASMEND
7147; GFX900-NEXT:    s_mov_b32 s8, s12
7148; GFX900-NEXT:    s_mov_b32 s9, s13
7149; GFX900-NEXT:    s_mov_b32 s12, s16
7150; GFX900-NEXT:    s_mov_b32 s13, s17
7151; GFX900-NEXT:    ;;#ASMSTART
7152; GFX900-NEXT:    ; use s[8:13]
7153; GFX900-NEXT:    ;;#ASMEND
7154; GFX900-NEXT:    s_setpc_b64 s[30:31]
7155;
7156; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_4_2:
7157; GFX90A:       ; %bb.0:
7158; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7159; GFX90A-NEXT:    ;;#ASMSTART
7160; GFX90A-NEXT:    ; def s[12:17]
7161; GFX90A-NEXT:    ;;#ASMEND
7162; GFX90A-NEXT:    ;;#ASMSTART
7163; GFX90A-NEXT:    ; def s[8:13]
7164; GFX90A-NEXT:    ;;#ASMEND
7165; GFX90A-NEXT:    s_mov_b32 s8, s12
7166; GFX90A-NEXT:    s_mov_b32 s9, s13
7167; GFX90A-NEXT:    s_mov_b32 s12, s16
7168; GFX90A-NEXT:    s_mov_b32 s13, s17
7169; GFX90A-NEXT:    ;;#ASMSTART
7170; GFX90A-NEXT:    ; use s[8:13]
7171; GFX90A-NEXT:    ;;#ASMEND
7172; GFX90A-NEXT:    s_setpc_b64 s[30:31]
7173;
7174; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_4_2:
7175; GFX940:       ; %bb.0:
7176; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7177; GFX940-NEXT:    ;;#ASMSTART
7178; GFX940-NEXT:    ; def s[8:13]
7179; GFX940-NEXT:    ;;#ASMEND
7180; GFX940-NEXT:    ;;#ASMSTART
7181; GFX940-NEXT:    ; def s[0:5]
7182; GFX940-NEXT:    ;;#ASMEND
7183; GFX940-NEXT:    s_mov_b32 s8, s12
7184; GFX940-NEXT:    s_mov_b32 s9, s13
7185; GFX940-NEXT:    s_mov_b32 s12, s4
7186; GFX940-NEXT:    s_mov_b32 s13, s5
7187; GFX940-NEXT:    ;;#ASMSTART
7188; GFX940-NEXT:    ; use s[8:13]
7189; GFX940-NEXT:    ;;#ASMEND
7190; GFX940-NEXT:    s_setpc_b64 s[30:31]
7191  %vec0 = call <3 x i64> asm "; def $0", "=s"()
7192  %vec1 = call <3 x i64> asm "; def $0", "=s"()
7193  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 4, i32 2>
7194  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
7195  ret void
7196}
7197
7198define void @s_shuffle_v3i64_v3i64__u_3_3() {
7199; GFX9-LABEL: s_shuffle_v3i64_v3i64__u_3_3:
7200; GFX9:       ; %bb.0:
7201; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7202; GFX9-NEXT:    ;;#ASMSTART
7203; GFX9-NEXT:    ; use s[8:13]
7204; GFX9-NEXT:    ;;#ASMEND
7205; GFX9-NEXT:    s_setpc_b64 s[30:31]
7206  %vec0 = call <3 x i64> asm "; def $0", "=s"()
7207  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 poison, i32 3, i32 3>
7208  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
7209  ret void
7210}
7211
7212define void @s_shuffle_v3i64_v3i64__0_3_3() {
7213; GFX900-LABEL: s_shuffle_v3i64_v3i64__0_3_3:
7214; GFX900:       ; %bb.0:
7215; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7216; GFX900-NEXT:    ;;#ASMSTART
7217; GFX900-NEXT:    ; def s[8:13]
7218; GFX900-NEXT:    ;;#ASMEND
7219; GFX900-NEXT:    ;;#ASMSTART
7220; GFX900-NEXT:    ; use s[8:13]
7221; GFX900-NEXT:    ;;#ASMEND
7222; GFX900-NEXT:    s_setpc_b64 s[30:31]
7223;
7224; GFX90A-LABEL: s_shuffle_v3i64_v3i64__0_3_3:
7225; GFX90A:       ; %bb.0:
7226; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7227; GFX90A-NEXT:    ;;#ASMSTART
7228; GFX90A-NEXT:    ; def s[8:13]
7229; GFX90A-NEXT:    ;;#ASMEND
7230; GFX90A-NEXT:    ;;#ASMSTART
7231; GFX90A-NEXT:    ; use s[8:13]
7232; GFX90A-NEXT:    ;;#ASMEND
7233; GFX90A-NEXT:    s_setpc_b64 s[30:31]
7234;
7235; GFX940-LABEL: s_shuffle_v3i64_v3i64__0_3_3:
7236; GFX940:       ; %bb.0:
7237; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7238; GFX940-NEXT:    ;;#ASMSTART
7239; GFX940-NEXT:    ; def s[8:13]
7240; GFX940-NEXT:    ;;#ASMEND
7241; GFX940-NEXT:    s_nop 0
7242; GFX940-NEXT:    ;;#ASMSTART
7243; GFX940-NEXT:    ; use s[8:13]
7244; GFX940-NEXT:    ;;#ASMEND
7245; GFX940-NEXT:    s_setpc_b64 s[30:31]
7246  %vec0 = call <3 x i64> asm "; def $0", "=s"()
7247  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 0, i32 3, i32 3>
7248  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
7249  ret void
7250}
7251
7252define void @s_shuffle_v3i64_v3i64__1_3_3() {
7253; GFX900-LABEL: s_shuffle_v3i64_v3i64__1_3_3:
7254; GFX900:       ; %bb.0:
7255; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7256; GFX900-NEXT:    ;;#ASMSTART
7257; GFX900-NEXT:    ; def s[4:9]
7258; GFX900-NEXT:    ;;#ASMEND
7259; GFX900-NEXT:    s_mov_b32 s8, s6
7260; GFX900-NEXT:    s_mov_b32 s9, s7
7261; GFX900-NEXT:    ;;#ASMSTART
7262; GFX900-NEXT:    ; use s[8:13]
7263; GFX900-NEXT:    ;;#ASMEND
7264; GFX900-NEXT:    s_setpc_b64 s[30:31]
7265;
7266; GFX90A-LABEL: s_shuffle_v3i64_v3i64__1_3_3:
7267; GFX90A:       ; %bb.0:
7268; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7269; GFX90A-NEXT:    ;;#ASMSTART
7270; GFX90A-NEXT:    ; def s[4:9]
7271; GFX90A-NEXT:    ;;#ASMEND
7272; GFX90A-NEXT:    s_mov_b32 s8, s6
7273; GFX90A-NEXT:    s_mov_b32 s9, s7
7274; GFX90A-NEXT:    ;;#ASMSTART
7275; GFX90A-NEXT:    ; use s[8:13]
7276; GFX90A-NEXT:    ;;#ASMEND
7277; GFX90A-NEXT:    s_setpc_b64 s[30:31]
7278;
7279; GFX940-LABEL: s_shuffle_v3i64_v3i64__1_3_3:
7280; GFX940:       ; %bb.0:
7281; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7282; GFX940-NEXT:    ;;#ASMSTART
7283; GFX940-NEXT:    ; def s[0:5]
7284; GFX940-NEXT:    ;;#ASMEND
7285; GFX940-NEXT:    s_mov_b32 s8, s2
7286; GFX940-NEXT:    s_mov_b32 s9, s3
7287; GFX940-NEXT:    ;;#ASMSTART
7288; GFX940-NEXT:    ; use s[8:13]
7289; GFX940-NEXT:    ;;#ASMEND
7290; GFX940-NEXT:    s_setpc_b64 s[30:31]
7291  %vec0 = call <3 x i64> asm "; def $0", "=s"()
7292  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 1, i32 3, i32 3>
7293  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
7294  ret void
7295}
7296
7297define void @s_shuffle_v3i64_v3i64__2_3_3() {
7298; GFX900-LABEL: s_shuffle_v3i64_v3i64__2_3_3:
7299; GFX900:       ; %bb.0:
7300; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7301; GFX900-NEXT:    ;;#ASMSTART
7302; GFX900-NEXT:    ; def s[4:9]
7303; GFX900-NEXT:    ;;#ASMEND
7304; GFX900-NEXT:    ;;#ASMSTART
7305; GFX900-NEXT:    ; use s[8:13]
7306; GFX900-NEXT:    ;;#ASMEND
7307; GFX900-NEXT:    s_setpc_b64 s[30:31]
7308;
7309; GFX90A-LABEL: s_shuffle_v3i64_v3i64__2_3_3:
7310; GFX90A:       ; %bb.0:
7311; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7312; GFX90A-NEXT:    ;;#ASMSTART
7313; GFX90A-NEXT:    ; def s[4:9]
7314; GFX90A-NEXT:    ;;#ASMEND
7315; GFX90A-NEXT:    ;;#ASMSTART
7316; GFX90A-NEXT:    ; use s[8:13]
7317; GFX90A-NEXT:    ;;#ASMEND
7318; GFX90A-NEXT:    s_setpc_b64 s[30:31]
7319;
7320; GFX940-LABEL: s_shuffle_v3i64_v3i64__2_3_3:
7321; GFX940:       ; %bb.0:
7322; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7323; GFX940-NEXT:    ;;#ASMSTART
7324; GFX940-NEXT:    ; def s[0:5]
7325; GFX940-NEXT:    ;;#ASMEND
7326; GFX940-NEXT:    s_mov_b32 s8, s4
7327; GFX940-NEXT:    s_mov_b32 s9, s5
7328; GFX940-NEXT:    ;;#ASMSTART
7329; GFX940-NEXT:    ; use s[8:13]
7330; GFX940-NEXT:    ;;#ASMEND
7331; GFX940-NEXT:    s_setpc_b64 s[30:31]
7332  %vec0 = call <3 x i64> asm "; def $0", "=s"()
7333  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 2, i32 3, i32 3>
7334  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
7335  ret void
7336}
7337
7338define void @s_shuffle_v3i64_v3i64__3_3_3() {
7339; GFX9-LABEL: s_shuffle_v3i64_v3i64__3_3_3:
7340; GFX9:       ; %bb.0:
7341; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7342; GFX9-NEXT:    ;;#ASMSTART
7343; GFX9-NEXT:    ; use s[8:13]
7344; GFX9-NEXT:    ;;#ASMEND
7345; GFX9-NEXT:    s_setpc_b64 s[30:31]
7346  %vec0 = call <3 x i64> asm "; def $0", "=s"()
7347  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> poison, <3 x i32> <i32 3, i32 3, i32 3>
7348  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
7349  ret void
7350}
7351
7352define void @s_shuffle_v3i64_v3i64__4_3_3() {
7353; GFX900-LABEL: s_shuffle_v3i64_v3i64__4_3_3:
7354; GFX900:       ; %bb.0:
7355; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7356; GFX900-NEXT:    ;;#ASMSTART
7357; GFX900-NEXT:    ; def s[4:9]
7358; GFX900-NEXT:    ;;#ASMEND
7359; GFX900-NEXT:    s_mov_b32 s8, s6
7360; GFX900-NEXT:    s_mov_b32 s9, s7
7361; GFX900-NEXT:    s_mov_b32 s10, s4
7362; GFX900-NEXT:    s_mov_b32 s11, s5
7363; GFX900-NEXT:    s_mov_b32 s12, s4
7364; GFX900-NEXT:    s_mov_b32 s13, s5
7365; GFX900-NEXT:    ;;#ASMSTART
7366; GFX900-NEXT:    ; use s[8:13]
7367; GFX900-NEXT:    ;;#ASMEND
7368; GFX900-NEXT:    s_setpc_b64 s[30:31]
7369;
7370; GFX90A-LABEL: s_shuffle_v3i64_v3i64__4_3_3:
7371; GFX90A:       ; %bb.0:
7372; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7373; GFX90A-NEXT:    ;;#ASMSTART
7374; GFX90A-NEXT:    ; def s[4:9]
7375; GFX90A-NEXT:    ;;#ASMEND
7376; GFX90A-NEXT:    s_mov_b32 s8, s6
7377; GFX90A-NEXT:    s_mov_b32 s9, s7
7378; GFX90A-NEXT:    s_mov_b32 s10, s4
7379; GFX90A-NEXT:    s_mov_b32 s11, s5
7380; GFX90A-NEXT:    s_mov_b32 s12, s4
7381; GFX90A-NEXT:    s_mov_b32 s13, s5
7382; GFX90A-NEXT:    ;;#ASMSTART
7383; GFX90A-NEXT:    ; use s[8:13]
7384; GFX90A-NEXT:    ;;#ASMEND
7385; GFX90A-NEXT:    s_setpc_b64 s[30:31]
7386;
7387; GFX940-LABEL: s_shuffle_v3i64_v3i64__4_3_3:
7388; GFX940:       ; %bb.0:
7389; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7390; GFX940-NEXT:    ;;#ASMSTART
7391; GFX940-NEXT:    ; def s[0:5]
7392; GFX940-NEXT:    ;;#ASMEND
7393; GFX940-NEXT:    s_mov_b32 s8, s2
7394; GFX940-NEXT:    s_mov_b32 s9, s3
7395; GFX940-NEXT:    s_mov_b32 s10, s0
7396; GFX940-NEXT:    s_mov_b32 s11, s1
7397; GFX940-NEXT:    s_mov_b32 s12, s0
7398; GFX940-NEXT:    s_mov_b32 s13, s1
7399; GFX940-NEXT:    ;;#ASMSTART
7400; GFX940-NEXT:    ; use s[8:13]
7401; GFX940-NEXT:    ;;#ASMEND
7402; GFX940-NEXT:    s_setpc_b64 s[30:31]
7403  %vec0 = call <3 x i64> asm "; def $0", "=s"()
7404  %vec1 = call <3 x i64> asm "; def $0", "=s"()
7405  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 4, i32 3, i32 3>
7406  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
7407  ret void
7408}
7409
7410define void @s_shuffle_v3i64_v3i64__5_3_3() {
7411; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_3_3:
7412; GFX900:       ; %bb.0:
7413; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7414; GFX900-NEXT:    ;;#ASMSTART
7415; GFX900-NEXT:    ; def s[4:9]
7416; GFX900-NEXT:    ;;#ASMEND
7417; GFX900-NEXT:    s_mov_b32 s10, s4
7418; GFX900-NEXT:    s_mov_b32 s11, s5
7419; GFX900-NEXT:    s_mov_b32 s12, s4
7420; GFX900-NEXT:    s_mov_b32 s13, s5
7421; GFX900-NEXT:    ;;#ASMSTART
7422; GFX900-NEXT:    ; use s[8:13]
7423; GFX900-NEXT:    ;;#ASMEND
7424; GFX900-NEXT:    s_setpc_b64 s[30:31]
7425;
7426; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_3_3:
7427; GFX90A:       ; %bb.0:
7428; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7429; GFX90A-NEXT:    ;;#ASMSTART
7430; GFX90A-NEXT:    ; def s[4:9]
7431; GFX90A-NEXT:    ;;#ASMEND
7432; GFX90A-NEXT:    s_mov_b32 s10, s4
7433; GFX90A-NEXT:    s_mov_b32 s11, s5
7434; GFX90A-NEXT:    s_mov_b32 s12, s4
7435; GFX90A-NEXT:    s_mov_b32 s13, s5
7436; GFX90A-NEXT:    ;;#ASMSTART
7437; GFX90A-NEXT:    ; use s[8:13]
7438; GFX90A-NEXT:    ;;#ASMEND
7439; GFX90A-NEXT:    s_setpc_b64 s[30:31]
7440;
7441; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_3_3:
7442; GFX940:       ; %bb.0:
7443; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7444; GFX940-NEXT:    ;;#ASMSTART
7445; GFX940-NEXT:    ; def s[0:5]
7446; GFX940-NEXT:    ;;#ASMEND
7447; GFX940-NEXT:    s_mov_b32 s8, s4
7448; GFX940-NEXT:    s_mov_b32 s9, s5
7449; GFX940-NEXT:    s_mov_b32 s10, s0
7450; GFX940-NEXT:    s_mov_b32 s11, s1
7451; GFX940-NEXT:    s_mov_b32 s12, s0
7452; GFX940-NEXT:    s_mov_b32 s13, s1
7453; GFX940-NEXT:    ;;#ASMSTART
7454; GFX940-NEXT:    ; use s[8:13]
7455; GFX940-NEXT:    ;;#ASMEND
7456; GFX940-NEXT:    s_setpc_b64 s[30:31]
7457  %vec0 = call <3 x i64> asm "; def $0", "=s"()
7458  %vec1 = call <3 x i64> asm "; def $0", "=s"()
7459  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 3, i32 3>
7460  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
7461  ret void
7462}
7463
7464define void @s_shuffle_v3i64_v3i64__5_u_3() {
7465; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_u_3:
7466; GFX900:       ; %bb.0:
7467; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7468; GFX900-NEXT:    ;;#ASMSTART
7469; GFX900-NEXT:    ; def s[4:9]
7470; GFX900-NEXT:    ;;#ASMEND
7471; GFX900-NEXT:    s_mov_b32 s12, s4
7472; GFX900-NEXT:    s_mov_b32 s13, s5
7473; GFX900-NEXT:    ;;#ASMSTART
7474; GFX900-NEXT:    ; use s[8:13]
7475; GFX900-NEXT:    ;;#ASMEND
7476; GFX900-NEXT:    s_setpc_b64 s[30:31]
7477;
7478; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_u_3:
7479; GFX90A:       ; %bb.0:
7480; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7481; GFX90A-NEXT:    ;;#ASMSTART
7482; GFX90A-NEXT:    ; def s[4:9]
7483; GFX90A-NEXT:    ;;#ASMEND
7484; GFX90A-NEXT:    s_mov_b32 s12, s4
7485; GFX90A-NEXT:    s_mov_b32 s13, s5
7486; GFX90A-NEXT:    ;;#ASMSTART
7487; GFX90A-NEXT:    ; use s[8:13]
7488; GFX90A-NEXT:    ;;#ASMEND
7489; GFX90A-NEXT:    s_setpc_b64 s[30:31]
7490;
7491; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_u_3:
7492; GFX940:       ; %bb.0:
7493; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7494; GFX940-NEXT:    ;;#ASMSTART
7495; GFX940-NEXT:    ; def s[0:5]
7496; GFX940-NEXT:    ;;#ASMEND
7497; GFX940-NEXT:    s_mov_b32 s8, s4
7498; GFX940-NEXT:    s_mov_b32 s9, s5
7499; GFX940-NEXT:    s_mov_b32 s12, s0
7500; GFX940-NEXT:    s_mov_b32 s13, s1
7501; GFX940-NEXT:    ;;#ASMSTART
7502; GFX940-NEXT:    ; use s[8:13]
7503; GFX940-NEXT:    ;;#ASMEND
7504; GFX940-NEXT:    s_setpc_b64 s[30:31]
7505  %vec0 = call <3 x i64> asm "; def $0", "=s"()
7506  %vec1 = call <3 x i64> asm "; def $0", "=s"()
7507  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 poison, i32 3>
7508  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
7509  ret void
7510}
7511
7512define void @s_shuffle_v3i64_v3i64__5_0_3() {
7513; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_0_3:
7514; GFX900:       ; %bb.0:
7515; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7516; GFX900-NEXT:    ;;#ASMSTART
7517; GFX900-NEXT:    ; def s[4:9]
7518; GFX900-NEXT:    ;;#ASMEND
7519; GFX900-NEXT:    ;;#ASMSTART
7520; GFX900-NEXT:    ; def s[12:17]
7521; GFX900-NEXT:    ;;#ASMEND
7522; GFX900-NEXT:    s_mov_b32 s8, s16
7523; GFX900-NEXT:    s_mov_b32 s9, s17
7524; GFX900-NEXT:    s_mov_b32 s10, s4
7525; GFX900-NEXT:    s_mov_b32 s11, s5
7526; GFX900-NEXT:    ;;#ASMSTART
7527; GFX900-NEXT:    ; use s[8:13]
7528; GFX900-NEXT:    ;;#ASMEND
7529; GFX900-NEXT:    s_setpc_b64 s[30:31]
7530;
7531; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_0_3:
7532; GFX90A:       ; %bb.0:
7533; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7534; GFX90A-NEXT:    ;;#ASMSTART
7535; GFX90A-NEXT:    ; def s[4:9]
7536; GFX90A-NEXT:    ;;#ASMEND
7537; GFX90A-NEXT:    ;;#ASMSTART
7538; GFX90A-NEXT:    ; def s[12:17]
7539; GFX90A-NEXT:    ;;#ASMEND
7540; GFX90A-NEXT:    s_mov_b32 s8, s16
7541; GFX90A-NEXT:    s_mov_b32 s9, s17
7542; GFX90A-NEXT:    s_mov_b32 s10, s4
7543; GFX90A-NEXT:    s_mov_b32 s11, s5
7544; GFX90A-NEXT:    ;;#ASMSTART
7545; GFX90A-NEXT:    ; use s[8:13]
7546; GFX90A-NEXT:    ;;#ASMEND
7547; GFX90A-NEXT:    s_setpc_b64 s[30:31]
7548;
7549; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_0_3:
7550; GFX940:       ; %bb.0:
7551; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7552; GFX940-NEXT:    ;;#ASMSTART
7553; GFX940-NEXT:    ; def s[0:5]
7554; GFX940-NEXT:    ;;#ASMEND
7555; GFX940-NEXT:    s_mov_b32 s10, s0
7556; GFX940-NEXT:    ;;#ASMSTART
7557; GFX940-NEXT:    ; def s[4:9]
7558; GFX940-NEXT:    ;;#ASMEND
7559; GFX940-NEXT:    s_mov_b32 s11, s1
7560; GFX940-NEXT:    s_mov_b32 s12, s4
7561; GFX940-NEXT:    s_mov_b32 s13, s5
7562; GFX940-NEXT:    ;;#ASMSTART
7563; GFX940-NEXT:    ; use s[8:13]
7564; GFX940-NEXT:    ;;#ASMEND
7565; GFX940-NEXT:    s_setpc_b64 s[30:31]
7566  %vec0 = call <3 x i64> asm "; def $0", "=s"()
7567  %vec1 = call <3 x i64> asm "; def $0", "=s"()
7568  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 0, i32 3>
7569  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
7570  ret void
7571}
7572
7573define void @s_shuffle_v3i64_v3i64__5_1_3() {
7574; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_1_3:
7575; GFX900:       ; %bb.0:
7576; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7577; GFX900-NEXT:    ;;#ASMSTART
7578; GFX900-NEXT:    ; def s[8:13]
7579; GFX900-NEXT:    ;;#ASMEND
7580; GFX900-NEXT:    ;;#ASMSTART
7581; GFX900-NEXT:    ; def s[4:9]
7582; GFX900-NEXT:    ;;#ASMEND
7583; GFX900-NEXT:    s_mov_b32 s12, s4
7584; GFX900-NEXT:    s_mov_b32 s13, s5
7585; GFX900-NEXT:    ;;#ASMSTART
7586; GFX900-NEXT:    ; use s[8:13]
7587; GFX900-NEXT:    ;;#ASMEND
7588; GFX900-NEXT:    s_setpc_b64 s[30:31]
7589;
7590; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_1_3:
7591; GFX90A:       ; %bb.0:
7592; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7593; GFX90A-NEXT:    ;;#ASMSTART
7594; GFX90A-NEXT:    ; def s[8:13]
7595; GFX90A-NEXT:    ;;#ASMEND
7596; GFX90A-NEXT:    ;;#ASMSTART
7597; GFX90A-NEXT:    ; def s[4:9]
7598; GFX90A-NEXT:    ;;#ASMEND
7599; GFX90A-NEXT:    s_mov_b32 s12, s4
7600; GFX90A-NEXT:    s_mov_b32 s13, s5
7601; GFX90A-NEXT:    ;;#ASMSTART
7602; GFX90A-NEXT:    ; use s[8:13]
7603; GFX90A-NEXT:    ;;#ASMEND
7604; GFX90A-NEXT:    s_setpc_b64 s[30:31]
7605;
7606; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_1_3:
7607; GFX940:       ; %bb.0:
7608; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7609; GFX940-NEXT:    ;;#ASMSTART
7610; GFX940-NEXT:    ; def s[8:13]
7611; GFX940-NEXT:    ;;#ASMEND
7612; GFX940-NEXT:    ;;#ASMSTART
7613; GFX940-NEXT:    ; def s[0:5]
7614; GFX940-NEXT:    ;;#ASMEND
7615; GFX940-NEXT:    s_mov_b32 s8, s4
7616; GFX940-NEXT:    s_mov_b32 s9, s5
7617; GFX940-NEXT:    s_mov_b32 s12, s0
7618; GFX940-NEXT:    s_mov_b32 s13, s1
7619; GFX940-NEXT:    ;;#ASMSTART
7620; GFX940-NEXT:    ; use s[8:13]
7621; GFX940-NEXT:    ;;#ASMEND
7622; GFX940-NEXT:    s_setpc_b64 s[30:31]
7623  %vec0 = call <3 x i64> asm "; def $0", "=s"()
7624  %vec1 = call <3 x i64> asm "; def $0", "=s"()
7625  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 1, i32 3>
7626  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
7627  ret void
7628}
7629
7630define void @s_shuffle_v3i64_v3i64__5_2_3() {
7631; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_2_3:
7632; GFX900:       ; %bb.0:
7633; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7634; GFX900-NEXT:    ;;#ASMSTART
7635; GFX900-NEXT:    ; def s[8:13]
7636; GFX900-NEXT:    ;;#ASMEND
7637; GFX900-NEXT:    ;;#ASMSTART
7638; GFX900-NEXT:    ; def s[4:9]
7639; GFX900-NEXT:    ;;#ASMEND
7640; GFX900-NEXT:    s_mov_b32 s10, s12
7641; GFX900-NEXT:    s_mov_b32 s11, s13
7642; GFX900-NEXT:    s_mov_b32 s12, s4
7643; GFX900-NEXT:    s_mov_b32 s13, s5
7644; GFX900-NEXT:    ;;#ASMSTART
7645; GFX900-NEXT:    ; use s[8:13]
7646; GFX900-NEXT:    ;;#ASMEND
7647; GFX900-NEXT:    s_setpc_b64 s[30:31]
7648;
7649; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_2_3:
7650; GFX90A:       ; %bb.0:
7651; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7652; GFX90A-NEXT:    ;;#ASMSTART
7653; GFX90A-NEXT:    ; def s[8:13]
7654; GFX90A-NEXT:    ;;#ASMEND
7655; GFX90A-NEXT:    ;;#ASMSTART
7656; GFX90A-NEXT:    ; def s[4:9]
7657; GFX90A-NEXT:    ;;#ASMEND
7658; GFX90A-NEXT:    s_mov_b32 s10, s12
7659; GFX90A-NEXT:    s_mov_b32 s11, s13
7660; GFX90A-NEXT:    s_mov_b32 s12, s4
7661; GFX90A-NEXT:    s_mov_b32 s13, s5
7662; GFX90A-NEXT:    ;;#ASMSTART
7663; GFX90A-NEXT:    ; use s[8:13]
7664; GFX90A-NEXT:    ;;#ASMEND
7665; GFX90A-NEXT:    s_setpc_b64 s[30:31]
7666;
7667; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_2_3:
7668; GFX940:       ; %bb.0:
7669; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7670; GFX940-NEXT:    ;;#ASMSTART
7671; GFX940-NEXT:    ; def s[0:5]
7672; GFX940-NEXT:    ;;#ASMEND
7673; GFX940-NEXT:    ;;#ASMSTART
7674; GFX940-NEXT:    ; def s[12:17]
7675; GFX940-NEXT:    ;;#ASMEND
7676; GFX940-NEXT:    s_mov_b32 s8, s16
7677; GFX940-NEXT:    s_mov_b32 s9, s17
7678; GFX940-NEXT:    s_mov_b32 s10, s4
7679; GFX940-NEXT:    s_mov_b32 s11, s5
7680; GFX940-NEXT:    ;;#ASMSTART
7681; GFX940-NEXT:    ; use s[8:13]
7682; GFX940-NEXT:    ;;#ASMEND
7683; GFX940-NEXT:    s_setpc_b64 s[30:31]
7684  %vec0 = call <3 x i64> asm "; def $0", "=s"()
7685  %vec1 = call <3 x i64> asm "; def $0", "=s"()
7686  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 2, i32 3>
7687  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
7688  ret void
7689}
7690
7691define void @s_shuffle_v3i64_v3i64__5_4_3() {
7692; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_4_3:
7693; GFX900:       ; %bb.0:
7694; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7695; GFX900-NEXT:    ;;#ASMSTART
7696; GFX900-NEXT:    ; def s[4:9]
7697; GFX900-NEXT:    ;;#ASMEND
7698; GFX900-NEXT:    s_mov_b32 s10, s6
7699; GFX900-NEXT:    s_mov_b32 s11, s7
7700; GFX900-NEXT:    s_mov_b32 s12, s4
7701; GFX900-NEXT:    s_mov_b32 s13, s5
7702; GFX900-NEXT:    ;;#ASMSTART
7703; GFX900-NEXT:    ; use s[8:13]
7704; GFX900-NEXT:    ;;#ASMEND
7705; GFX900-NEXT:    s_setpc_b64 s[30:31]
7706;
7707; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_4_3:
7708; GFX90A:       ; %bb.0:
7709; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7710; GFX90A-NEXT:    ;;#ASMSTART
7711; GFX90A-NEXT:    ; def s[4:9]
7712; GFX90A-NEXT:    ;;#ASMEND
7713; GFX90A-NEXT:    s_mov_b32 s10, s6
7714; GFX90A-NEXT:    s_mov_b32 s11, s7
7715; GFX90A-NEXT:    s_mov_b32 s12, s4
7716; GFX90A-NEXT:    s_mov_b32 s13, s5
7717; GFX90A-NEXT:    ;;#ASMSTART
7718; GFX90A-NEXT:    ; use s[8:13]
7719; GFX90A-NEXT:    ;;#ASMEND
7720; GFX90A-NEXT:    s_setpc_b64 s[30:31]
7721;
7722; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_4_3:
7723; GFX940:       ; %bb.0:
7724; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7725; GFX940-NEXT:    ;;#ASMSTART
7726; GFX940-NEXT:    ; def s[0:5]
7727; GFX940-NEXT:    ;;#ASMEND
7728; GFX940-NEXT:    s_mov_b32 s8, s4
7729; GFX940-NEXT:    s_mov_b32 s9, s5
7730; GFX940-NEXT:    s_mov_b32 s10, s2
7731; GFX940-NEXT:    s_mov_b32 s11, s3
7732; GFX940-NEXT:    s_mov_b32 s12, s0
7733; GFX940-NEXT:    s_mov_b32 s13, s1
7734; GFX940-NEXT:    ;;#ASMSTART
7735; GFX940-NEXT:    ; use s[8:13]
7736; GFX940-NEXT:    ;;#ASMEND
7737; GFX940-NEXT:    s_setpc_b64 s[30:31]
7738  %vec0 = call <3 x i64> asm "; def $0", "=s"()
7739  %vec1 = call <3 x i64> asm "; def $0", "=s"()
7740  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 4, i32 3>
7741  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
7742  ret void
7743}
7744
7745define void @s_shuffle_v3i64_v3i64__u_4_4() {
7746; GFX9-LABEL: s_shuffle_v3i64_v3i64__u_4_4:
7747; GFX9:       ; %bb.0:
7748; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7749; GFX9-NEXT:    ;;#ASMSTART
7750; GFX9-NEXT:    ; def s[8:13]
7751; GFX9-NEXT:    ;;#ASMEND
7752; GFX9-NEXT:    s_mov_b32 s12, s10
7753; GFX9-NEXT:    s_mov_b32 s13, s11
7754; GFX9-NEXT:    ;;#ASMSTART
7755; GFX9-NEXT:    ; use s[8:13]
7756; GFX9-NEXT:    ;;#ASMEND
7757; GFX9-NEXT:    s_setpc_b64 s[30:31]
7758  %vec0 = call <3 x i64> asm "; def $0", "=s"()
7759  %vec1 = call <3 x i64> asm "; def $0", "=s"()
7760  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 poison, i32 4, i32 4>
7761  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
7762  ret void
7763}
7764
7765define void @s_shuffle_v3i64_v3i64__0_4_4() {
7766; GFX900-LABEL: s_shuffle_v3i64_v3i64__0_4_4:
7767; GFX900:       ; %bb.0:
7768; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7769; GFX900-NEXT:    ;;#ASMSTART
7770; GFX900-NEXT:    ; def s[8:13]
7771; GFX900-NEXT:    ;;#ASMEND
7772; GFX900-NEXT:    ;;#ASMSTART
7773; GFX900-NEXT:    ; def s[12:17]
7774; GFX900-NEXT:    ;;#ASMEND
7775; GFX900-NEXT:    s_mov_b32 s10, s14
7776; GFX900-NEXT:    s_mov_b32 s11, s15
7777; GFX900-NEXT:    s_mov_b32 s12, s14
7778; GFX900-NEXT:    s_mov_b32 s13, s15
7779; GFX900-NEXT:    ;;#ASMSTART
7780; GFX900-NEXT:    ; use s[8:13]
7781; GFX900-NEXT:    ;;#ASMEND
7782; GFX900-NEXT:    s_setpc_b64 s[30:31]
7783;
7784; GFX90A-LABEL: s_shuffle_v3i64_v3i64__0_4_4:
7785; GFX90A:       ; %bb.0:
7786; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7787; GFX90A-NEXT:    ;;#ASMSTART
7788; GFX90A-NEXT:    ; def s[8:13]
7789; GFX90A-NEXT:    ;;#ASMEND
7790; GFX90A-NEXT:    ;;#ASMSTART
7791; GFX90A-NEXT:    ; def s[12:17]
7792; GFX90A-NEXT:    ;;#ASMEND
7793; GFX90A-NEXT:    s_mov_b32 s10, s14
7794; GFX90A-NEXT:    s_mov_b32 s11, s15
7795; GFX90A-NEXT:    s_mov_b32 s12, s14
7796; GFX90A-NEXT:    s_mov_b32 s13, s15
7797; GFX90A-NEXT:    ;;#ASMSTART
7798; GFX90A-NEXT:    ; use s[8:13]
7799; GFX90A-NEXT:    ;;#ASMEND
7800; GFX90A-NEXT:    s_setpc_b64 s[30:31]
7801;
7802; GFX940-LABEL: s_shuffle_v3i64_v3i64__0_4_4:
7803; GFX940:       ; %bb.0:
7804; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7805; GFX940-NEXT:    ;;#ASMSTART
7806; GFX940-NEXT:    ; def s[8:13]
7807; GFX940-NEXT:    ;;#ASMEND
7808; GFX940-NEXT:    ;;#ASMSTART
7809; GFX940-NEXT:    ; def s[0:5]
7810; GFX940-NEXT:    ;;#ASMEND
7811; GFX940-NEXT:    s_mov_b32 s10, s2
7812; GFX940-NEXT:    s_mov_b32 s11, s3
7813; GFX940-NEXT:    s_mov_b32 s12, s2
7814; GFX940-NEXT:    s_mov_b32 s13, s3
7815; GFX940-NEXT:    ;;#ASMSTART
7816; GFX940-NEXT:    ; use s[8:13]
7817; GFX940-NEXT:    ;;#ASMEND
7818; GFX940-NEXT:    s_setpc_b64 s[30:31]
7819  %vec0 = call <3 x i64> asm "; def $0", "=s"()
7820  %vec1 = call <3 x i64> asm "; def $0", "=s"()
7821  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 0, i32 4, i32 4>
7822  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
7823  ret void
7824}
7825
7826define void @s_shuffle_v3i64_v3i64__1_4_4() {
7827; GFX900-LABEL: s_shuffle_v3i64_v3i64__1_4_4:
7828; GFX900:       ; %bb.0:
7829; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7830; GFX900-NEXT:    ;;#ASMSTART
7831; GFX900-NEXT:    ; def s[4:9]
7832; GFX900-NEXT:    ;;#ASMEND
7833; GFX900-NEXT:    ;;#ASMSTART
7834; GFX900-NEXT:    ; def s[8:13]
7835; GFX900-NEXT:    ;;#ASMEND
7836; GFX900-NEXT:    s_mov_b32 s8, s6
7837; GFX900-NEXT:    s_mov_b32 s9, s7
7838; GFX900-NEXT:    s_mov_b32 s12, s10
7839; GFX900-NEXT:    s_mov_b32 s13, s11
7840; GFX900-NEXT:    ;;#ASMSTART
7841; GFX900-NEXT:    ; use s[8:13]
7842; GFX900-NEXT:    ;;#ASMEND
7843; GFX900-NEXT:    s_setpc_b64 s[30:31]
7844;
7845; GFX90A-LABEL: s_shuffle_v3i64_v3i64__1_4_4:
7846; GFX90A:       ; %bb.0:
7847; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7848; GFX90A-NEXT:    ;;#ASMSTART
7849; GFX90A-NEXT:    ; def s[4:9]
7850; GFX90A-NEXT:    ;;#ASMEND
7851; GFX90A-NEXT:    ;;#ASMSTART
7852; GFX90A-NEXT:    ; def s[8:13]
7853; GFX90A-NEXT:    ;;#ASMEND
7854; GFX90A-NEXT:    s_mov_b32 s8, s6
7855; GFX90A-NEXT:    s_mov_b32 s9, s7
7856; GFX90A-NEXT:    s_mov_b32 s12, s10
7857; GFX90A-NEXT:    s_mov_b32 s13, s11
7858; GFX90A-NEXT:    ;;#ASMSTART
7859; GFX90A-NEXT:    ; use s[8:13]
7860; GFX90A-NEXT:    ;;#ASMEND
7861; GFX90A-NEXT:    s_setpc_b64 s[30:31]
7862;
7863; GFX940-LABEL: s_shuffle_v3i64_v3i64__1_4_4:
7864; GFX940:       ; %bb.0:
7865; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7866; GFX940-NEXT:    ;;#ASMSTART
7867; GFX940-NEXT:    ; def s[8:13]
7868; GFX940-NEXT:    ;;#ASMEND
7869; GFX940-NEXT:    ;;#ASMSTART
7870; GFX940-NEXT:    ; def s[0:5]
7871; GFX940-NEXT:    ;;#ASMEND
7872; GFX940-NEXT:    s_mov_b32 s8, s2
7873; GFX940-NEXT:    s_mov_b32 s9, s3
7874; GFX940-NEXT:    s_mov_b32 s12, s10
7875; GFX940-NEXT:    s_mov_b32 s13, s11
7876; GFX940-NEXT:    ;;#ASMSTART
7877; GFX940-NEXT:    ; use s[8:13]
7878; GFX940-NEXT:    ;;#ASMEND
7879; GFX940-NEXT:    s_setpc_b64 s[30:31]
7880  %vec0 = call <3 x i64> asm "; def $0", "=s"()
7881  %vec1 = call <3 x i64> asm "; def $0", "=s"()
7882  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 1, i32 4, i32 4>
7883  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
7884  ret void
7885}
7886
7887define void @s_shuffle_v3i64_v3i64__2_4_4() {
7888; GFX900-LABEL: s_shuffle_v3i64_v3i64__2_4_4:
7889; GFX900:       ; %bb.0:
7890; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7891; GFX900-NEXT:    ;;#ASMSTART
7892; GFX900-NEXT:    ; def s[12:17]
7893; GFX900-NEXT:    ;;#ASMEND
7894; GFX900-NEXT:    ;;#ASMSTART
7895; GFX900-NEXT:    ; def s[8:13]
7896; GFX900-NEXT:    ;;#ASMEND
7897; GFX900-NEXT:    s_mov_b32 s8, s16
7898; GFX900-NEXT:    s_mov_b32 s9, s17
7899; GFX900-NEXT:    s_mov_b32 s12, s10
7900; GFX900-NEXT:    s_mov_b32 s13, s11
7901; GFX900-NEXT:    ;;#ASMSTART
7902; GFX900-NEXT:    ; use s[8:13]
7903; GFX900-NEXT:    ;;#ASMEND
7904; GFX900-NEXT:    s_setpc_b64 s[30:31]
7905;
7906; GFX90A-LABEL: s_shuffle_v3i64_v3i64__2_4_4:
7907; GFX90A:       ; %bb.0:
7908; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7909; GFX90A-NEXT:    ;;#ASMSTART
7910; GFX90A-NEXT:    ; def s[12:17]
7911; GFX90A-NEXT:    ;;#ASMEND
7912; GFX90A-NEXT:    ;;#ASMSTART
7913; GFX90A-NEXT:    ; def s[8:13]
7914; GFX90A-NEXT:    ;;#ASMEND
7915; GFX90A-NEXT:    s_mov_b32 s8, s16
7916; GFX90A-NEXT:    s_mov_b32 s9, s17
7917; GFX90A-NEXT:    s_mov_b32 s12, s10
7918; GFX90A-NEXT:    s_mov_b32 s13, s11
7919; GFX90A-NEXT:    ;;#ASMSTART
7920; GFX90A-NEXT:    ; use s[8:13]
7921; GFX90A-NEXT:    ;;#ASMEND
7922; GFX90A-NEXT:    s_setpc_b64 s[30:31]
7923;
7924; GFX940-LABEL: s_shuffle_v3i64_v3i64__2_4_4:
7925; GFX940:       ; %bb.0:
7926; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7927; GFX940-NEXT:    ;;#ASMSTART
7928; GFX940-NEXT:    ; def s[8:13]
7929; GFX940-NEXT:    ;;#ASMEND
7930; GFX940-NEXT:    ;;#ASMSTART
7931; GFX940-NEXT:    ; def s[0:5]
7932; GFX940-NEXT:    ;;#ASMEND
7933; GFX940-NEXT:    s_mov_b32 s8, s4
7934; GFX940-NEXT:    s_mov_b32 s9, s5
7935; GFX940-NEXT:    s_mov_b32 s12, s10
7936; GFX940-NEXT:    s_mov_b32 s13, s11
7937; GFX940-NEXT:    ;;#ASMSTART
7938; GFX940-NEXT:    ; use s[8:13]
7939; GFX940-NEXT:    ;;#ASMEND
7940; GFX940-NEXT:    s_setpc_b64 s[30:31]
7941  %vec0 = call <3 x i64> asm "; def $0", "=s"()
7942  %vec1 = call <3 x i64> asm "; def $0", "=s"()
7943  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 2, i32 4, i32 4>
7944  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
7945  ret void
7946}
7947
7948define void @s_shuffle_v3i64_v3i64__3_4_4() {
7949; GFX9-LABEL: s_shuffle_v3i64_v3i64__3_4_4:
7950; GFX9:       ; %bb.0:
7951; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7952; GFX9-NEXT:    ;;#ASMSTART
7953; GFX9-NEXT:    ; def s[8:13]
7954; GFX9-NEXT:    ;;#ASMEND
7955; GFX9-NEXT:    s_mov_b32 s12, s10
7956; GFX9-NEXT:    s_mov_b32 s13, s11
7957; GFX9-NEXT:    ;;#ASMSTART
7958; GFX9-NEXT:    ; use s[8:13]
7959; GFX9-NEXT:    ;;#ASMEND
7960; GFX9-NEXT:    s_setpc_b64 s[30:31]
7961  %vec0 = call <3 x i64> asm "; def $0", "=s"()
7962  %vec1 = call <3 x i64> asm "; def $0", "=s"()
7963  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 3, i32 4, i32 4>
7964  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
7965  ret void
7966}
7967
7968define void @s_shuffle_v3i64_v3i64__4_4_4() {
7969; GFX9-LABEL: s_shuffle_v3i64_v3i64__4_4_4:
7970; GFX9:       ; %bb.0:
7971; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7972; GFX9-NEXT:    ;;#ASMSTART
7973; GFX9-NEXT:    ; def s[8:13]
7974; GFX9-NEXT:    ;;#ASMEND
7975; GFX9-NEXT:    s_mov_b32 s8, s10
7976; GFX9-NEXT:    s_mov_b32 s9, s11
7977; GFX9-NEXT:    s_mov_b32 s12, s10
7978; GFX9-NEXT:    s_mov_b32 s13, s11
7979; GFX9-NEXT:    ;;#ASMSTART
7980; GFX9-NEXT:    ; use s[8:13]
7981; GFX9-NEXT:    ;;#ASMEND
7982; GFX9-NEXT:    s_setpc_b64 s[30:31]
7983  %vec0 = call <3 x i64> asm "; def $0", "=s"()
7984  %vec1 = call <3 x i64> asm "; def $0", "=s"()
7985  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 4, i32 4, i32 4>
7986  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
7987  ret void
7988}
7989
7990define void @s_shuffle_v3i64_v3i64__5_4_4() {
7991; GFX9-LABEL: s_shuffle_v3i64_v3i64__5_4_4:
7992; GFX9:       ; %bb.0:
7993; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
7994; GFX9-NEXT:    ;;#ASMSTART
7995; GFX9-NEXT:    ; def s[8:13]
7996; GFX9-NEXT:    ;;#ASMEND
7997; GFX9-NEXT:    s_mov_b32 s8, s12
7998; GFX9-NEXT:    s_mov_b32 s9, s13
7999; GFX9-NEXT:    s_mov_b32 s12, s10
8000; GFX9-NEXT:    s_mov_b32 s13, s11
8001; GFX9-NEXT:    ;;#ASMSTART
8002; GFX9-NEXT:    ; use s[8:13]
8003; GFX9-NEXT:    ;;#ASMEND
8004; GFX9-NEXT:    s_setpc_b64 s[30:31]
8005  %vec0 = call <3 x i64> asm "; def $0", "=s"()
8006  %vec1 = call <3 x i64> asm "; def $0", "=s"()
8007  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 4, i32 4>
8008  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
8009  ret void
8010}
8011
8012define void @s_shuffle_v3i64_v3i64__5_u_4() {
8013; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_u_4:
8014; GFX900:       ; %bb.0:
8015; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8016; GFX900-NEXT:    ;;#ASMSTART
8017; GFX900-NEXT:    ; def s[4:9]
8018; GFX900-NEXT:    ;;#ASMEND
8019; GFX900-NEXT:    s_mov_b32 s12, s6
8020; GFX900-NEXT:    s_mov_b32 s13, s7
8021; GFX900-NEXT:    ;;#ASMSTART
8022; GFX900-NEXT:    ; use s[8:13]
8023; GFX900-NEXT:    ;;#ASMEND
8024; GFX900-NEXT:    s_setpc_b64 s[30:31]
8025;
8026; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_u_4:
8027; GFX90A:       ; %bb.0:
8028; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8029; GFX90A-NEXT:    ;;#ASMSTART
8030; GFX90A-NEXT:    ; def s[4:9]
8031; GFX90A-NEXT:    ;;#ASMEND
8032; GFX90A-NEXT:    s_mov_b32 s12, s6
8033; GFX90A-NEXT:    s_mov_b32 s13, s7
8034; GFX90A-NEXT:    ;;#ASMSTART
8035; GFX90A-NEXT:    ; use s[8:13]
8036; GFX90A-NEXT:    ;;#ASMEND
8037; GFX90A-NEXT:    s_setpc_b64 s[30:31]
8038;
8039; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_u_4:
8040; GFX940:       ; %bb.0:
8041; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8042; GFX940-NEXT:    ;;#ASMSTART
8043; GFX940-NEXT:    ; def s[0:5]
8044; GFX940-NEXT:    ;;#ASMEND
8045; GFX940-NEXT:    s_mov_b32 s8, s4
8046; GFX940-NEXT:    s_mov_b32 s9, s5
8047; GFX940-NEXT:    s_mov_b32 s12, s2
8048; GFX940-NEXT:    s_mov_b32 s13, s3
8049; GFX940-NEXT:    ;;#ASMSTART
8050; GFX940-NEXT:    ; use s[8:13]
8051; GFX940-NEXT:    ;;#ASMEND
8052; GFX940-NEXT:    s_setpc_b64 s[30:31]
8053  %vec0 = call <3 x i64> asm "; def $0", "=s"()
8054  %vec1 = call <3 x i64> asm "; def $0", "=s"()
8055  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 poison, i32 4>
8056  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
8057  ret void
8058}
8059
8060define void @s_shuffle_v3i64_v3i64__5_0_4() {
8061; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_0_4:
8062; GFX900:       ; %bb.0:
8063; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8064; GFX900-NEXT:    ;;#ASMSTART
8065; GFX900-NEXT:    ; def s[4:9]
8066; GFX900-NEXT:    ;;#ASMEND
8067; GFX900-NEXT:    ;;#ASMSTART
8068; GFX900-NEXT:    ; def s[12:17]
8069; GFX900-NEXT:    ;;#ASMEND
8070; GFX900-NEXT:    s_mov_b32 s8, s16
8071; GFX900-NEXT:    s_mov_b32 s9, s17
8072; GFX900-NEXT:    s_mov_b32 s10, s4
8073; GFX900-NEXT:    s_mov_b32 s11, s5
8074; GFX900-NEXT:    s_mov_b32 s12, s14
8075; GFX900-NEXT:    s_mov_b32 s13, s15
8076; GFX900-NEXT:    ;;#ASMSTART
8077; GFX900-NEXT:    ; use s[8:13]
8078; GFX900-NEXT:    ;;#ASMEND
8079; GFX900-NEXT:    s_setpc_b64 s[30:31]
8080;
8081; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_0_4:
8082; GFX90A:       ; %bb.0:
8083; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8084; GFX90A-NEXT:    ;;#ASMSTART
8085; GFX90A-NEXT:    ; def s[4:9]
8086; GFX90A-NEXT:    ;;#ASMEND
8087; GFX90A-NEXT:    ;;#ASMSTART
8088; GFX90A-NEXT:    ; def s[12:17]
8089; GFX90A-NEXT:    ;;#ASMEND
8090; GFX90A-NEXT:    s_mov_b32 s8, s16
8091; GFX90A-NEXT:    s_mov_b32 s9, s17
8092; GFX90A-NEXT:    s_mov_b32 s10, s4
8093; GFX90A-NEXT:    s_mov_b32 s11, s5
8094; GFX90A-NEXT:    s_mov_b32 s12, s14
8095; GFX90A-NEXT:    s_mov_b32 s13, s15
8096; GFX90A-NEXT:    ;;#ASMSTART
8097; GFX90A-NEXT:    ; use s[8:13]
8098; GFX90A-NEXT:    ;;#ASMEND
8099; GFX90A-NEXT:    s_setpc_b64 s[30:31]
8100;
8101; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_0_4:
8102; GFX940:       ; %bb.0:
8103; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8104; GFX940-NEXT:    ;;#ASMSTART
8105; GFX940-NEXT:    ; def s[0:5]
8106; GFX940-NEXT:    ;;#ASMEND
8107; GFX940-NEXT:    s_mov_b32 s10, s0
8108; GFX940-NEXT:    ;;#ASMSTART
8109; GFX940-NEXT:    ; def s[4:9]
8110; GFX940-NEXT:    ;;#ASMEND
8111; GFX940-NEXT:    s_mov_b32 s11, s1
8112; GFX940-NEXT:    s_mov_b32 s12, s6
8113; GFX940-NEXT:    s_mov_b32 s13, s7
8114; GFX940-NEXT:    ;;#ASMSTART
8115; GFX940-NEXT:    ; use s[8:13]
8116; GFX940-NEXT:    ;;#ASMEND
8117; GFX940-NEXT:    s_setpc_b64 s[30:31]
8118  %vec0 = call <3 x i64> asm "; def $0", "=s"()
8119  %vec1 = call <3 x i64> asm "; def $0", "=s"()
8120  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 0, i32 4>
8121  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
8122  ret void
8123}
8124
8125define void @s_shuffle_v3i64_v3i64__5_1_4() {
8126; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_1_4:
8127; GFX900:       ; %bb.0:
8128; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8129; GFX900-NEXT:    ;;#ASMSTART
8130; GFX900-NEXT:    ; def s[8:13]
8131; GFX900-NEXT:    ;;#ASMEND
8132; GFX900-NEXT:    ;;#ASMSTART
8133; GFX900-NEXT:    ; def s[4:9]
8134; GFX900-NEXT:    ;;#ASMEND
8135; GFX900-NEXT:    s_mov_b32 s12, s6
8136; GFX900-NEXT:    s_mov_b32 s13, s7
8137; GFX900-NEXT:    ;;#ASMSTART
8138; GFX900-NEXT:    ; use s[8:13]
8139; GFX900-NEXT:    ;;#ASMEND
8140; GFX900-NEXT:    s_setpc_b64 s[30:31]
8141;
8142; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_1_4:
8143; GFX90A:       ; %bb.0:
8144; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8145; GFX90A-NEXT:    ;;#ASMSTART
8146; GFX90A-NEXT:    ; def s[8:13]
8147; GFX90A-NEXT:    ;;#ASMEND
8148; GFX90A-NEXT:    ;;#ASMSTART
8149; GFX90A-NEXT:    ; def s[4:9]
8150; GFX90A-NEXT:    ;;#ASMEND
8151; GFX90A-NEXT:    s_mov_b32 s12, s6
8152; GFX90A-NEXT:    s_mov_b32 s13, s7
8153; GFX90A-NEXT:    ;;#ASMSTART
8154; GFX90A-NEXT:    ; use s[8:13]
8155; GFX90A-NEXT:    ;;#ASMEND
8156; GFX90A-NEXT:    s_setpc_b64 s[30:31]
8157;
8158; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_1_4:
8159; GFX940:       ; %bb.0:
8160; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8161; GFX940-NEXT:    ;;#ASMSTART
8162; GFX940-NEXT:    ; def s[8:13]
8163; GFX940-NEXT:    ;;#ASMEND
8164; GFX940-NEXT:    ;;#ASMSTART
8165; GFX940-NEXT:    ; def s[0:5]
8166; GFX940-NEXT:    ;;#ASMEND
8167; GFX940-NEXT:    s_mov_b32 s8, s4
8168; GFX940-NEXT:    s_mov_b32 s9, s5
8169; GFX940-NEXT:    s_mov_b32 s12, s2
8170; GFX940-NEXT:    s_mov_b32 s13, s3
8171; GFX940-NEXT:    ;;#ASMSTART
8172; GFX940-NEXT:    ; use s[8:13]
8173; GFX940-NEXT:    ;;#ASMEND
8174; GFX940-NEXT:    s_setpc_b64 s[30:31]
8175  %vec0 = call <3 x i64> asm "; def $0", "=s"()
8176  %vec1 = call <3 x i64> asm "; def $0", "=s"()
8177  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 1, i32 4>
8178  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
8179  ret void
8180}
8181
8182define void @s_shuffle_v3i64_v3i64__5_2_4() {
8183; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_2_4:
8184; GFX900:       ; %bb.0:
8185; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8186; GFX900-NEXT:    ;;#ASMSTART
8187; GFX900-NEXT:    ; def s[8:13]
8188; GFX900-NEXT:    ;;#ASMEND
8189; GFX900-NEXT:    ;;#ASMSTART
8190; GFX900-NEXT:    ; def s[4:9]
8191; GFX900-NEXT:    ;;#ASMEND
8192; GFX900-NEXT:    s_mov_b32 s10, s12
8193; GFX900-NEXT:    s_mov_b32 s11, s13
8194; GFX900-NEXT:    s_mov_b32 s12, s6
8195; GFX900-NEXT:    s_mov_b32 s13, s7
8196; GFX900-NEXT:    ;;#ASMSTART
8197; GFX900-NEXT:    ; use s[8:13]
8198; GFX900-NEXT:    ;;#ASMEND
8199; GFX900-NEXT:    s_setpc_b64 s[30:31]
8200;
8201; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_2_4:
8202; GFX90A:       ; %bb.0:
8203; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8204; GFX90A-NEXT:    ;;#ASMSTART
8205; GFX90A-NEXT:    ; def s[8:13]
8206; GFX90A-NEXT:    ;;#ASMEND
8207; GFX90A-NEXT:    ;;#ASMSTART
8208; GFX90A-NEXT:    ; def s[4:9]
8209; GFX90A-NEXT:    ;;#ASMEND
8210; GFX90A-NEXT:    s_mov_b32 s10, s12
8211; GFX90A-NEXT:    s_mov_b32 s11, s13
8212; GFX90A-NEXT:    s_mov_b32 s12, s6
8213; GFX90A-NEXT:    s_mov_b32 s13, s7
8214; GFX90A-NEXT:    ;;#ASMSTART
8215; GFX90A-NEXT:    ; use s[8:13]
8216; GFX90A-NEXT:    ;;#ASMEND
8217; GFX90A-NEXT:    s_setpc_b64 s[30:31]
8218;
8219; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_2_4:
8220; GFX940:       ; %bb.0:
8221; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8222; GFX940-NEXT:    ;;#ASMSTART
8223; GFX940-NEXT:    ; def s[12:17]
8224; GFX940-NEXT:    ;;#ASMEND
8225; GFX940-NEXT:    ;;#ASMSTART
8226; GFX940-NEXT:    ; def s[0:5]
8227; GFX940-NEXT:    ;;#ASMEND
8228; GFX940-NEXT:    s_mov_b32 s8, s16
8229; GFX940-NEXT:    s_mov_b32 s9, s17
8230; GFX940-NEXT:    s_mov_b32 s10, s4
8231; GFX940-NEXT:    s_mov_b32 s11, s5
8232; GFX940-NEXT:    s_mov_b32 s12, s14
8233; GFX940-NEXT:    s_mov_b32 s13, s15
8234; GFX940-NEXT:    ;;#ASMSTART
8235; GFX940-NEXT:    ; use s[8:13]
8236; GFX940-NEXT:    ;;#ASMEND
8237; GFX940-NEXT:    s_setpc_b64 s[30:31]
8238  %vec0 = call <3 x i64> asm "; def $0", "=s"()
8239  %vec1 = call <3 x i64> asm "; def $0", "=s"()
8240  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 2, i32 4>
8241  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
8242  ret void
8243}
8244
8245define void @s_shuffle_v3i64_v3i64__5_3_4() {
8246; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_3_4:
8247; GFX900:       ; %bb.0:
8248; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8249; GFX900-NEXT:    ;;#ASMSTART
8250; GFX900-NEXT:    ; def s[4:9]
8251; GFX900-NEXT:    ;;#ASMEND
8252; GFX900-NEXT:    s_mov_b32 s10, s4
8253; GFX900-NEXT:    s_mov_b32 s11, s5
8254; GFX900-NEXT:    s_mov_b32 s12, s6
8255; GFX900-NEXT:    s_mov_b32 s13, s7
8256; GFX900-NEXT:    ;;#ASMSTART
8257; GFX900-NEXT:    ; use s[8:13]
8258; GFX900-NEXT:    ;;#ASMEND
8259; GFX900-NEXT:    s_setpc_b64 s[30:31]
8260;
8261; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_3_4:
8262; GFX90A:       ; %bb.0:
8263; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8264; GFX90A-NEXT:    ;;#ASMSTART
8265; GFX90A-NEXT:    ; def s[4:9]
8266; GFX90A-NEXT:    ;;#ASMEND
8267; GFX90A-NEXT:    s_mov_b32 s10, s4
8268; GFX90A-NEXT:    s_mov_b32 s11, s5
8269; GFX90A-NEXT:    s_mov_b32 s12, s6
8270; GFX90A-NEXT:    s_mov_b32 s13, s7
8271; GFX90A-NEXT:    ;;#ASMSTART
8272; GFX90A-NEXT:    ; use s[8:13]
8273; GFX90A-NEXT:    ;;#ASMEND
8274; GFX90A-NEXT:    s_setpc_b64 s[30:31]
8275;
8276; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_3_4:
8277; GFX940:       ; %bb.0:
8278; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8279; GFX940-NEXT:    ;;#ASMSTART
8280; GFX940-NEXT:    ; def s[0:5]
8281; GFX940-NEXT:    ;;#ASMEND
8282; GFX940-NEXT:    s_mov_b32 s8, s4
8283; GFX940-NEXT:    s_mov_b32 s9, s5
8284; GFX940-NEXT:    s_mov_b32 s10, s0
8285; GFX940-NEXT:    s_mov_b32 s11, s1
8286; GFX940-NEXT:    s_mov_b32 s12, s2
8287; GFX940-NEXT:    s_mov_b32 s13, s3
8288; GFX940-NEXT:    ;;#ASMSTART
8289; GFX940-NEXT:    ; use s[8:13]
8290; GFX940-NEXT:    ;;#ASMEND
8291; GFX940-NEXT:    s_setpc_b64 s[30:31]
8292  %vec0 = call <3 x i64> asm "; def $0", "=s"()
8293  %vec1 = call <3 x i64> asm "; def $0", "=s"()
8294  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 3, i32 4>
8295  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
8296  ret void
8297}
8298
8299define void @s_shuffle_v3i64_v3i64__u_5_5() {
8300; GFX9-LABEL: s_shuffle_v3i64_v3i64__u_5_5:
8301; GFX9:       ; %bb.0:
8302; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8303; GFX9-NEXT:    ;;#ASMSTART
8304; GFX9-NEXT:    ; def s[8:13]
8305; GFX9-NEXT:    ;;#ASMEND
8306; GFX9-NEXT:    s_mov_b32 s10, s12
8307; GFX9-NEXT:    s_mov_b32 s11, s13
8308; GFX9-NEXT:    ;;#ASMSTART
8309; GFX9-NEXT:    ; use s[8:13]
8310; GFX9-NEXT:    ;;#ASMEND
8311; GFX9-NEXT:    s_setpc_b64 s[30:31]
8312  %vec0 = call <3 x i64> asm "; def $0", "=s"()
8313  %vec1 = call <3 x i64> asm "; def $0", "=s"()
8314  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 poison, i32 5, i32 5>
8315  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
8316  ret void
8317}
8318
8319define void @s_shuffle_v3i64_v3i64__0_5_5() {
8320; GFX900-LABEL: s_shuffle_v3i64_v3i64__0_5_5:
8321; GFX900:       ; %bb.0:
8322; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8323; GFX900-NEXT:    ;;#ASMSTART
8324; GFX900-NEXT:    ; def s[8:13]
8325; GFX900-NEXT:    ;;#ASMEND
8326; GFX900-NEXT:    ;;#ASMSTART
8327; GFX900-NEXT:    ; def s[12:17]
8328; GFX900-NEXT:    ;;#ASMEND
8329; GFX900-NEXT:    s_mov_b32 s10, s16
8330; GFX900-NEXT:    s_mov_b32 s11, s17
8331; GFX900-NEXT:    s_mov_b32 s12, s16
8332; GFX900-NEXT:    s_mov_b32 s13, s17
8333; GFX900-NEXT:    ;;#ASMSTART
8334; GFX900-NEXT:    ; use s[8:13]
8335; GFX900-NEXT:    ;;#ASMEND
8336; GFX900-NEXT:    s_setpc_b64 s[30:31]
8337;
8338; GFX90A-LABEL: s_shuffle_v3i64_v3i64__0_5_5:
8339; GFX90A:       ; %bb.0:
8340; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8341; GFX90A-NEXT:    ;;#ASMSTART
8342; GFX90A-NEXT:    ; def s[8:13]
8343; GFX90A-NEXT:    ;;#ASMEND
8344; GFX90A-NEXT:    ;;#ASMSTART
8345; GFX90A-NEXT:    ; def s[12:17]
8346; GFX90A-NEXT:    ;;#ASMEND
8347; GFX90A-NEXT:    s_mov_b32 s10, s16
8348; GFX90A-NEXT:    s_mov_b32 s11, s17
8349; GFX90A-NEXT:    s_mov_b32 s12, s16
8350; GFX90A-NEXT:    s_mov_b32 s13, s17
8351; GFX90A-NEXT:    ;;#ASMSTART
8352; GFX90A-NEXT:    ; use s[8:13]
8353; GFX90A-NEXT:    ;;#ASMEND
8354; GFX90A-NEXT:    s_setpc_b64 s[30:31]
8355;
8356; GFX940-LABEL: s_shuffle_v3i64_v3i64__0_5_5:
8357; GFX940:       ; %bb.0:
8358; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8359; GFX940-NEXT:    ;;#ASMSTART
8360; GFX940-NEXT:    ; def s[8:13]
8361; GFX940-NEXT:    ;;#ASMEND
8362; GFX940-NEXT:    ;;#ASMSTART
8363; GFX940-NEXT:    ; def s[0:5]
8364; GFX940-NEXT:    ;;#ASMEND
8365; GFX940-NEXT:    s_mov_b32 s10, s4
8366; GFX940-NEXT:    s_mov_b32 s11, s5
8367; GFX940-NEXT:    s_mov_b32 s12, s4
8368; GFX940-NEXT:    s_mov_b32 s13, s5
8369; GFX940-NEXT:    ;;#ASMSTART
8370; GFX940-NEXT:    ; use s[8:13]
8371; GFX940-NEXT:    ;;#ASMEND
8372; GFX940-NEXT:    s_setpc_b64 s[30:31]
8373  %vec0 = call <3 x i64> asm "; def $0", "=s"()
8374  %vec1 = call <3 x i64> asm "; def $0", "=s"()
8375  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 0, i32 5, i32 5>
8376  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
8377  ret void
8378}
8379
8380define void @s_shuffle_v3i64_v3i64__1_5_5() {
8381; GFX900-LABEL: s_shuffle_v3i64_v3i64__1_5_5:
8382; GFX900:       ; %bb.0:
8383; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8384; GFX900-NEXT:    ;;#ASMSTART
8385; GFX900-NEXT:    ; def s[4:9]
8386; GFX900-NEXT:    ;;#ASMEND
8387; GFX900-NEXT:    ;;#ASMSTART
8388; GFX900-NEXT:    ; def s[8:13]
8389; GFX900-NEXT:    ;;#ASMEND
8390; GFX900-NEXT:    s_mov_b32 s8, s6
8391; GFX900-NEXT:    s_mov_b32 s9, s7
8392; GFX900-NEXT:    s_mov_b32 s10, s12
8393; GFX900-NEXT:    s_mov_b32 s11, s13
8394; GFX900-NEXT:    ;;#ASMSTART
8395; GFX900-NEXT:    ; use s[8:13]
8396; GFX900-NEXT:    ;;#ASMEND
8397; GFX900-NEXT:    s_setpc_b64 s[30:31]
8398;
8399; GFX90A-LABEL: s_shuffle_v3i64_v3i64__1_5_5:
8400; GFX90A:       ; %bb.0:
8401; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8402; GFX90A-NEXT:    ;;#ASMSTART
8403; GFX90A-NEXT:    ; def s[4:9]
8404; GFX90A-NEXT:    ;;#ASMEND
8405; GFX90A-NEXT:    ;;#ASMSTART
8406; GFX90A-NEXT:    ; def s[8:13]
8407; GFX90A-NEXT:    ;;#ASMEND
8408; GFX90A-NEXT:    s_mov_b32 s8, s6
8409; GFX90A-NEXT:    s_mov_b32 s9, s7
8410; GFX90A-NEXT:    s_mov_b32 s10, s12
8411; GFX90A-NEXT:    s_mov_b32 s11, s13
8412; GFX90A-NEXT:    ;;#ASMSTART
8413; GFX90A-NEXT:    ; use s[8:13]
8414; GFX90A-NEXT:    ;;#ASMEND
8415; GFX90A-NEXT:    s_setpc_b64 s[30:31]
8416;
8417; GFX940-LABEL: s_shuffle_v3i64_v3i64__1_5_5:
8418; GFX940:       ; %bb.0:
8419; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8420; GFX940-NEXT:    ;;#ASMSTART
8421; GFX940-NEXT:    ; def s[8:13]
8422; GFX940-NEXT:    ;;#ASMEND
8423; GFX940-NEXT:    ;;#ASMSTART
8424; GFX940-NEXT:    ; def s[0:5]
8425; GFX940-NEXT:    ;;#ASMEND
8426; GFX940-NEXT:    s_mov_b32 s8, s2
8427; GFX940-NEXT:    s_mov_b32 s9, s3
8428; GFX940-NEXT:    s_mov_b32 s10, s12
8429; GFX940-NEXT:    s_mov_b32 s11, s13
8430; GFX940-NEXT:    ;;#ASMSTART
8431; GFX940-NEXT:    ; use s[8:13]
8432; GFX940-NEXT:    ;;#ASMEND
8433; GFX940-NEXT:    s_setpc_b64 s[30:31]
8434  %vec0 = call <3 x i64> asm "; def $0", "=s"()
8435  %vec1 = call <3 x i64> asm "; def $0", "=s"()
8436  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 1, i32 5, i32 5>
8437  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
8438  ret void
8439}
8440
8441define void @s_shuffle_v3i64_v3i64__2_5_5() {
8442; GFX900-LABEL: s_shuffle_v3i64_v3i64__2_5_5:
8443; GFX900:       ; %bb.0:
8444; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8445; GFX900-NEXT:    ;;#ASMSTART
8446; GFX900-NEXT:    ; def s[12:17]
8447; GFX900-NEXT:    ;;#ASMEND
8448; GFX900-NEXT:    ;;#ASMSTART
8449; GFX900-NEXT:    ; def s[8:13]
8450; GFX900-NEXT:    ;;#ASMEND
8451; GFX900-NEXT:    s_mov_b32 s8, s16
8452; GFX900-NEXT:    s_mov_b32 s9, s17
8453; GFX900-NEXT:    s_mov_b32 s10, s12
8454; GFX900-NEXT:    s_mov_b32 s11, s13
8455; GFX900-NEXT:    ;;#ASMSTART
8456; GFX900-NEXT:    ; use s[8:13]
8457; GFX900-NEXT:    ;;#ASMEND
8458; GFX900-NEXT:    s_setpc_b64 s[30:31]
8459;
8460; GFX90A-LABEL: s_shuffle_v3i64_v3i64__2_5_5:
8461; GFX90A:       ; %bb.0:
8462; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8463; GFX90A-NEXT:    ;;#ASMSTART
8464; GFX90A-NEXT:    ; def s[12:17]
8465; GFX90A-NEXT:    ;;#ASMEND
8466; GFX90A-NEXT:    ;;#ASMSTART
8467; GFX90A-NEXT:    ; def s[8:13]
8468; GFX90A-NEXT:    ;;#ASMEND
8469; GFX90A-NEXT:    s_mov_b32 s8, s16
8470; GFX90A-NEXT:    s_mov_b32 s9, s17
8471; GFX90A-NEXT:    s_mov_b32 s10, s12
8472; GFX90A-NEXT:    s_mov_b32 s11, s13
8473; GFX90A-NEXT:    ;;#ASMSTART
8474; GFX90A-NEXT:    ; use s[8:13]
8475; GFX90A-NEXT:    ;;#ASMEND
8476; GFX90A-NEXT:    s_setpc_b64 s[30:31]
8477;
8478; GFX940-LABEL: s_shuffle_v3i64_v3i64__2_5_5:
8479; GFX940:       ; %bb.0:
8480; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8481; GFX940-NEXT:    ;;#ASMSTART
8482; GFX940-NEXT:    ; def s[8:13]
8483; GFX940-NEXT:    ;;#ASMEND
8484; GFX940-NEXT:    ;;#ASMSTART
8485; GFX940-NEXT:    ; def s[0:5]
8486; GFX940-NEXT:    ;;#ASMEND
8487; GFX940-NEXT:    s_mov_b32 s8, s4
8488; GFX940-NEXT:    s_mov_b32 s9, s5
8489; GFX940-NEXT:    s_mov_b32 s10, s12
8490; GFX940-NEXT:    s_mov_b32 s11, s13
8491; GFX940-NEXT:    ;;#ASMSTART
8492; GFX940-NEXT:    ; use s[8:13]
8493; GFX940-NEXT:    ;;#ASMEND
8494; GFX940-NEXT:    s_setpc_b64 s[30:31]
8495  %vec0 = call <3 x i64> asm "; def $0", "=s"()
8496  %vec1 = call <3 x i64> asm "; def $0", "=s"()
8497  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 2, i32 5, i32 5>
8498  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
8499  ret void
8500}
8501
8502define void @s_shuffle_v3i64_v3i64__3_5_5() {
8503; GFX9-LABEL: s_shuffle_v3i64_v3i64__3_5_5:
8504; GFX9:       ; %bb.0:
8505; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8506; GFX9-NEXT:    ;;#ASMSTART
8507; GFX9-NEXT:    ; def s[8:13]
8508; GFX9-NEXT:    ;;#ASMEND
8509; GFX9-NEXT:    s_mov_b32 s10, s12
8510; GFX9-NEXT:    s_mov_b32 s11, s13
8511; GFX9-NEXT:    ;;#ASMSTART
8512; GFX9-NEXT:    ; use s[8:13]
8513; GFX9-NEXT:    ;;#ASMEND
8514; GFX9-NEXT:    s_setpc_b64 s[30:31]
8515  %vec0 = call <3 x i64> asm "; def $0", "=s"()
8516  %vec1 = call <3 x i64> asm "; def $0", "=s"()
8517  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 3, i32 5, i32 5>
8518  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
8519  ret void
8520}
8521
8522define void @s_shuffle_v3i64_v3i64__4_5_5() {
8523; GFX9-LABEL: s_shuffle_v3i64_v3i64__4_5_5:
8524; GFX9:       ; %bb.0:
8525; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8526; GFX9-NEXT:    ;;#ASMSTART
8527; GFX9-NEXT:    ; def s[8:13]
8528; GFX9-NEXT:    ;;#ASMEND
8529; GFX9-NEXT:    s_mov_b32 s8, s10
8530; GFX9-NEXT:    s_mov_b32 s9, s11
8531; GFX9-NEXT:    s_mov_b32 s10, s12
8532; GFX9-NEXT:    s_mov_b32 s11, s13
8533; GFX9-NEXT:    ;;#ASMSTART
8534; GFX9-NEXT:    ; use s[8:13]
8535; GFX9-NEXT:    ;;#ASMEND
8536; GFX9-NEXT:    s_setpc_b64 s[30:31]
8537  %vec0 = call <3 x i64> asm "; def $0", "=s"()
8538  %vec1 = call <3 x i64> asm "; def $0", "=s"()
8539  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 4, i32 5, i32 5>
8540  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
8541  ret void
8542}
8543
8544define void @s_shuffle_v3i64_v3i64__5_u_5() {
8545; GFX9-LABEL: s_shuffle_v3i64_v3i64__5_u_5:
8546; GFX9:       ; %bb.0:
8547; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8548; GFX9-NEXT:    ;;#ASMSTART
8549; GFX9-NEXT:    ; def s[8:13]
8550; GFX9-NEXT:    ;;#ASMEND
8551; GFX9-NEXT:    s_mov_b32 s8, s12
8552; GFX9-NEXT:    s_mov_b32 s9, s13
8553; GFX9-NEXT:    ;;#ASMSTART
8554; GFX9-NEXT:    ; use s[8:13]
8555; GFX9-NEXT:    ;;#ASMEND
8556; GFX9-NEXT:    s_setpc_b64 s[30:31]
8557  %vec0 = call <3 x i64> asm "; def $0", "=s"()
8558  %vec1 = call <3 x i64> asm "; def $0", "=s"()
8559  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 poison, i32 5>
8560  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
8561  ret void
8562}
8563
8564define void @s_shuffle_v3i64_v3i64__5_0_5() {
8565; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_0_5:
8566; GFX900:       ; %bb.0:
8567; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8568; GFX900-NEXT:    ;;#ASMSTART
8569; GFX900-NEXT:    ; def s[4:9]
8570; GFX900-NEXT:    ;;#ASMEND
8571; GFX900-NEXT:    ;;#ASMSTART
8572; GFX900-NEXT:    ; def s[8:13]
8573; GFX900-NEXT:    ;;#ASMEND
8574; GFX900-NEXT:    s_mov_b32 s8, s12
8575; GFX900-NEXT:    s_mov_b32 s9, s13
8576; GFX900-NEXT:    s_mov_b32 s10, s4
8577; GFX900-NEXT:    s_mov_b32 s11, s5
8578; GFX900-NEXT:    ;;#ASMSTART
8579; GFX900-NEXT:    ; use s[8:13]
8580; GFX900-NEXT:    ;;#ASMEND
8581; GFX900-NEXT:    s_setpc_b64 s[30:31]
8582;
8583; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_0_5:
8584; GFX90A:       ; %bb.0:
8585; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8586; GFX90A-NEXT:    ;;#ASMSTART
8587; GFX90A-NEXT:    ; def s[4:9]
8588; GFX90A-NEXT:    ;;#ASMEND
8589; GFX90A-NEXT:    ;;#ASMSTART
8590; GFX90A-NEXT:    ; def s[8:13]
8591; GFX90A-NEXT:    ;;#ASMEND
8592; GFX90A-NEXT:    s_mov_b32 s8, s12
8593; GFX90A-NEXT:    s_mov_b32 s9, s13
8594; GFX90A-NEXT:    s_mov_b32 s10, s4
8595; GFX90A-NEXT:    s_mov_b32 s11, s5
8596; GFX90A-NEXT:    ;;#ASMSTART
8597; GFX90A-NEXT:    ; use s[8:13]
8598; GFX90A-NEXT:    ;;#ASMEND
8599; GFX90A-NEXT:    s_setpc_b64 s[30:31]
8600;
8601; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_0_5:
8602; GFX940:       ; %bb.0:
8603; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8604; GFX940-NEXT:    ;;#ASMSTART
8605; GFX940-NEXT:    ; def s[8:13]
8606; GFX940-NEXT:    ;;#ASMEND
8607; GFX940-NEXT:    ;;#ASMSTART
8608; GFX940-NEXT:    ; def s[0:5]
8609; GFX940-NEXT:    ;;#ASMEND
8610; GFX940-NEXT:    s_mov_b32 s8, s12
8611; GFX940-NEXT:    s_mov_b32 s9, s13
8612; GFX940-NEXT:    s_mov_b32 s10, s0
8613; GFX940-NEXT:    s_mov_b32 s11, s1
8614; GFX940-NEXT:    ;;#ASMSTART
8615; GFX940-NEXT:    ; use s[8:13]
8616; GFX940-NEXT:    ;;#ASMEND
8617; GFX940-NEXT:    s_setpc_b64 s[30:31]
8618  %vec0 = call <3 x i64> asm "; def $0", "=s"()
8619  %vec1 = call <3 x i64> asm "; def $0", "=s"()
8620  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 0, i32 5>
8621  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
8622  ret void
8623}
8624
8625define void @s_shuffle_v3i64_v3i64__5_1_5() {
8626; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_1_5:
8627; GFX900:       ; %bb.0:
8628; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8629; GFX900-NEXT:    ;;#ASMSTART
8630; GFX900-NEXT:    ; def s[8:13]
8631; GFX900-NEXT:    ;;#ASMEND
8632; GFX900-NEXT:    ;;#ASMSTART
8633; GFX900-NEXT:    ; def s[12:17]
8634; GFX900-NEXT:    ;;#ASMEND
8635; GFX900-NEXT:    s_mov_b32 s8, s16
8636; GFX900-NEXT:    s_mov_b32 s9, s17
8637; GFX900-NEXT:    s_mov_b32 s12, s16
8638; GFX900-NEXT:    s_mov_b32 s13, s17
8639; GFX900-NEXT:    ;;#ASMSTART
8640; GFX900-NEXT:    ; use s[8:13]
8641; GFX900-NEXT:    ;;#ASMEND
8642; GFX900-NEXT:    s_setpc_b64 s[30:31]
8643;
8644; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_1_5:
8645; GFX90A:       ; %bb.0:
8646; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8647; GFX90A-NEXT:    ;;#ASMSTART
8648; GFX90A-NEXT:    ; def s[8:13]
8649; GFX90A-NEXT:    ;;#ASMEND
8650; GFX90A-NEXT:    ;;#ASMSTART
8651; GFX90A-NEXT:    ; def s[12:17]
8652; GFX90A-NEXT:    ;;#ASMEND
8653; GFX90A-NEXT:    s_mov_b32 s8, s16
8654; GFX90A-NEXT:    s_mov_b32 s9, s17
8655; GFX90A-NEXT:    s_mov_b32 s12, s16
8656; GFX90A-NEXT:    s_mov_b32 s13, s17
8657; GFX90A-NEXT:    ;;#ASMSTART
8658; GFX90A-NEXT:    ; use s[8:13]
8659; GFX90A-NEXT:    ;;#ASMEND
8660; GFX90A-NEXT:    s_setpc_b64 s[30:31]
8661;
8662; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_1_5:
8663; GFX940:       ; %bb.0:
8664; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8665; GFX940-NEXT:    ;;#ASMSTART
8666; GFX940-NEXT:    ; def s[8:13]
8667; GFX940-NEXT:    ;;#ASMEND
8668; GFX940-NEXT:    ;;#ASMSTART
8669; GFX940-NEXT:    ; def s[0:5]
8670; GFX940-NEXT:    ;;#ASMEND
8671; GFX940-NEXT:    s_mov_b32 s8, s4
8672; GFX940-NEXT:    s_mov_b32 s9, s5
8673; GFX940-NEXT:    s_mov_b32 s12, s4
8674; GFX940-NEXT:    s_mov_b32 s13, s5
8675; GFX940-NEXT:    ;;#ASMSTART
8676; GFX940-NEXT:    ; use s[8:13]
8677; GFX940-NEXT:    ;;#ASMEND
8678; GFX940-NEXT:    s_setpc_b64 s[30:31]
8679  %vec0 = call <3 x i64> asm "; def $0", "=s"()
8680  %vec1 = call <3 x i64> asm "; def $0", "=s"()
8681  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 1, i32 5>
8682  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
8683  ret void
8684}
8685
8686define void @s_shuffle_v3i64_v3i64__5_2_5() {
8687; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_2_5:
8688; GFX900:       ; %bb.0:
8689; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8690; GFX900-NEXT:    ;;#ASMSTART
8691; GFX900-NEXT:    ; def s[12:17]
8692; GFX900-NEXT:    ;;#ASMEND
8693; GFX900-NEXT:    ;;#ASMSTART
8694; GFX900-NEXT:    ; def s[8:13]
8695; GFX900-NEXT:    ;;#ASMEND
8696; GFX900-NEXT:    s_mov_b32 s8, s12
8697; GFX900-NEXT:    s_mov_b32 s9, s13
8698; GFX900-NEXT:    s_mov_b32 s10, s16
8699; GFX900-NEXT:    s_mov_b32 s11, s17
8700; GFX900-NEXT:    ;;#ASMSTART
8701; GFX900-NEXT:    ; use s[8:13]
8702; GFX900-NEXT:    ;;#ASMEND
8703; GFX900-NEXT:    s_setpc_b64 s[30:31]
8704;
8705; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_2_5:
8706; GFX90A:       ; %bb.0:
8707; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8708; GFX90A-NEXT:    ;;#ASMSTART
8709; GFX90A-NEXT:    ; def s[12:17]
8710; GFX90A-NEXT:    ;;#ASMEND
8711; GFX90A-NEXT:    ;;#ASMSTART
8712; GFX90A-NEXT:    ; def s[8:13]
8713; GFX90A-NEXT:    ;;#ASMEND
8714; GFX90A-NEXT:    s_mov_b32 s8, s12
8715; GFX90A-NEXT:    s_mov_b32 s9, s13
8716; GFX90A-NEXT:    s_mov_b32 s10, s16
8717; GFX90A-NEXT:    s_mov_b32 s11, s17
8718; GFX90A-NEXT:    ;;#ASMSTART
8719; GFX90A-NEXT:    ; use s[8:13]
8720; GFX90A-NEXT:    ;;#ASMEND
8721; GFX90A-NEXT:    s_setpc_b64 s[30:31]
8722;
8723; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_2_5:
8724; GFX940:       ; %bb.0:
8725; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8726; GFX940-NEXT:    ;;#ASMSTART
8727; GFX940-NEXT:    ; def s[8:13]
8728; GFX940-NEXT:    ;;#ASMEND
8729; GFX940-NEXT:    ;;#ASMSTART
8730; GFX940-NEXT:    ; def s[0:5]
8731; GFX940-NEXT:    ;;#ASMEND
8732; GFX940-NEXT:    s_mov_b32 s8, s12
8733; GFX940-NEXT:    s_mov_b32 s9, s13
8734; GFX940-NEXT:    s_mov_b32 s10, s4
8735; GFX940-NEXT:    s_mov_b32 s11, s5
8736; GFX940-NEXT:    ;;#ASMSTART
8737; GFX940-NEXT:    ; use s[8:13]
8738; GFX940-NEXT:    ;;#ASMEND
8739; GFX940-NEXT:    s_setpc_b64 s[30:31]
8740  %vec0 = call <3 x i64> asm "; def $0", "=s"()
8741  %vec1 = call <3 x i64> asm "; def $0", "=s"()
8742  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 2, i32 5>
8743  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
8744  ret void
8745}
8746
8747define void @s_shuffle_v3i64_v3i64__5_3_5() {
8748; GFX900-LABEL: s_shuffle_v3i64_v3i64__5_3_5:
8749; GFX900:       ; %bb.0:
8750; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8751; GFX900-NEXT:    ;;#ASMSTART
8752; GFX900-NEXT:    ; def s[12:17]
8753; GFX900-NEXT:    ;;#ASMEND
8754; GFX900-NEXT:    s_mov_b32 s8, s16
8755; GFX900-NEXT:    s_mov_b32 s9, s17
8756; GFX900-NEXT:    s_mov_b32 s10, s12
8757; GFX900-NEXT:    s_mov_b32 s11, s13
8758; GFX900-NEXT:    s_mov_b32 s12, s16
8759; GFX900-NEXT:    s_mov_b32 s13, s17
8760; GFX900-NEXT:    ;;#ASMSTART
8761; GFX900-NEXT:    ; use s[8:13]
8762; GFX900-NEXT:    ;;#ASMEND
8763; GFX900-NEXT:    s_setpc_b64 s[30:31]
8764;
8765; GFX90A-LABEL: s_shuffle_v3i64_v3i64__5_3_5:
8766; GFX90A:       ; %bb.0:
8767; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8768; GFX90A-NEXT:    ;;#ASMSTART
8769; GFX90A-NEXT:    ; def s[12:17]
8770; GFX90A-NEXT:    ;;#ASMEND
8771; GFX90A-NEXT:    s_mov_b32 s8, s16
8772; GFX90A-NEXT:    s_mov_b32 s9, s17
8773; GFX90A-NEXT:    s_mov_b32 s10, s12
8774; GFX90A-NEXT:    s_mov_b32 s11, s13
8775; GFX90A-NEXT:    s_mov_b32 s12, s16
8776; GFX90A-NEXT:    s_mov_b32 s13, s17
8777; GFX90A-NEXT:    ;;#ASMSTART
8778; GFX90A-NEXT:    ; use s[8:13]
8779; GFX90A-NEXT:    ;;#ASMEND
8780; GFX90A-NEXT:    s_setpc_b64 s[30:31]
8781;
8782; GFX940-LABEL: s_shuffle_v3i64_v3i64__5_3_5:
8783; GFX940:       ; %bb.0:
8784; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8785; GFX940-NEXT:    ;;#ASMSTART
8786; GFX940-NEXT:    ; def s[0:5]
8787; GFX940-NEXT:    ;;#ASMEND
8788; GFX940-NEXT:    s_mov_b32 s8, s4
8789; GFX940-NEXT:    s_mov_b32 s9, s5
8790; GFX940-NEXT:    s_mov_b32 s10, s0
8791; GFX940-NEXT:    s_mov_b32 s11, s1
8792; GFX940-NEXT:    s_mov_b32 s12, s4
8793; GFX940-NEXT:    s_mov_b32 s13, s5
8794; GFX940-NEXT:    ;;#ASMSTART
8795; GFX940-NEXT:    ; use s[8:13]
8796; GFX940-NEXT:    ;;#ASMEND
8797; GFX940-NEXT:    s_setpc_b64 s[30:31]
8798  %vec0 = call <3 x i64> asm "; def $0", "=s"()
8799  %vec1 = call <3 x i64> asm "; def $0", "=s"()
8800  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 3, i32 5>
8801  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
8802  ret void
8803}
8804
8805define void @s_shuffle_v3i64_v3i64__5_4_5() {
8806; GFX9-LABEL: s_shuffle_v3i64_v3i64__5_4_5:
8807; GFX9:       ; %bb.0:
8808; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
8809; GFX9-NEXT:    ;;#ASMSTART
8810; GFX9-NEXT:    ; def s[8:13]
8811; GFX9-NEXT:    ;;#ASMEND
8812; GFX9-NEXT:    s_mov_b32 s8, s12
8813; GFX9-NEXT:    s_mov_b32 s9, s13
8814; GFX9-NEXT:    ;;#ASMSTART
8815; GFX9-NEXT:    ; use s[8:13]
8816; GFX9-NEXT:    ;;#ASMEND
8817; GFX9-NEXT:    s_setpc_b64 s[30:31]
8818  %vec0 = call <3 x i64> asm "; def $0", "=s"()
8819  %vec1 = call <3 x i64> asm "; def $0", "=s"()
8820  %shuf = shufflevector <3 x i64> %vec0, <3 x i64> %vec1, <3 x i32> <i32 5, i32 4, i32 5>
8821  call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
8822  ret void
8823}
8824;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
8825; GFX90APLUS: {{.*}}
8826