xref: /llvm-project/llvm/test/CodeGen/SystemZ/atomicrmw-ops-i128.ll (revision ddb87e0f96c7d23b9b824fa9e2710c0fe11bb86b) 
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
2; Test i128 atomicrmw operations.
3;
4; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z15 | FileCheck %s
5;
6; Test expansion of AtomicRMW instructions, which assume a natural alignment.
7; Note that the multiple regmoves inside the CDSG loops hopefully will go away
8; when the new i128 support is added.
9
10; Check register exchange.
11define i128 @atomicrmw_xchg(ptr %src, i128 %b) {
12; CHECK-LABEL: atomicrmw_xchg:
13; CHECK:       # %bb.0:
14; CHECK-NEXT:    vl %v1, 0(%r4), 3
15; CHECK-NEXT:    vl %v0, 0(%r3), 4
16; CHECK-NEXT:    vlgvg %r1, %v1, 1
17; CHECK-NEXT:    vlgvg %r0, %v1, 0
18; CHECK-NEXT:  .LBB0_1: # %atomicrmw.start
19; CHECK-NEXT:    # =>This Inner Loop Header: Depth=1
20; CHECK-NEXT:    vlgvg %r5, %v0, 1
21; CHECK-NEXT:    vlgvg %r4, %v0, 0
22; CHECK-NEXT:    cdsg %r4, %r0, 0(%r3)
23; CHECK-NEXT:    vlvgp %v0, %r4, %r5
24; CHECK-NEXT:    jl .LBB0_1
25; CHECK-NEXT:  # %bb.2: # %atomicrmw.end
26; CHECK-NEXT:    vst %v0, 0(%r2), 3
27; CHECK-NEXT:    br %r14
28  %res = atomicrmw xchg ptr %src, i128 %b seq_cst
29  ret i128 %res
30}
31
32; Check addition of a variable.
33define i128 @atomicrmw_add(ptr %src, i128 %b) {
34; CHECK-LABEL: atomicrmw_add:
35; CHECK:       # %bb.0:
36; CHECK-NEXT:    vl %v0, 0(%r4), 3
37; CHECK-NEXT:    vl %v1, 0(%r3), 4
38; CHECK-NEXT:  .LBB1_1: # %atomicrmw.start
39; CHECK-NEXT:    # =>This Inner Loop Header: Depth=1
40; CHECK-NEXT:    vaq %v2, %v1, %v0
41; CHECK-NEXT:    vlgvg %r1, %v2, 1
42; CHECK-NEXT:    vlgvg %r0, %v2, 0
43; CHECK-NEXT:    vlgvg %r5, %v1, 1
44; CHECK-NEXT:    vlgvg %r4, %v1, 0
45; CHECK-NEXT:    cdsg %r4, %r0, 0(%r3)
46; CHECK-NEXT:    vlvgp %v1, %r4, %r5
47; CHECK-NEXT:    jl .LBB1_1
48; CHECK-NEXT:  # %bb.2: # %atomicrmw.end
49; CHECK-NEXT:    vst %v1, 0(%r2), 3
50; CHECK-NEXT:    br %r14
51  %res = atomicrmw add ptr %src, i128 %b seq_cst
52  ret i128 %res
53}
54
55; Check subtraction of a variable.
56define i128 @atomicrmw_sub(ptr %src, i128 %b) {
57; CHECK-LABEL: atomicrmw_sub:
58; CHECK:       # %bb.0:
59; CHECK-NEXT:    vl %v0, 0(%r4), 3
60; CHECK-NEXT:    vl %v1, 0(%r3), 4
61; CHECK-NEXT:  .LBB2_1: # %atomicrmw.start
62; CHECK-NEXT:    # =>This Inner Loop Header: Depth=1
63; CHECK-NEXT:    vsq %v2, %v1, %v0
64; CHECK-NEXT:    vlgvg %r1, %v2, 1
65; CHECK-NEXT:    vlgvg %r0, %v2, 0
66; CHECK-NEXT:    vlgvg %r5, %v1, 1
67; CHECK-NEXT:    vlgvg %r4, %v1, 0
68; CHECK-NEXT:    cdsg %r4, %r0, 0(%r3)
69; CHECK-NEXT:    vlvgp %v1, %r4, %r5
70; CHECK-NEXT:    jl .LBB2_1
71; CHECK-NEXT:  # %bb.2: # %atomicrmw.end
72; CHECK-NEXT:    vst %v1, 0(%r2), 3
73; CHECK-NEXT:    br %r14
74  %res = atomicrmw sub ptr %src, i128 %b seq_cst
75  ret i128 %res
76}
77
78; Check AND of a variable.
79define i128 @atomicrmw_and(ptr %src, i128 %b) {
80; CHECK-LABEL: atomicrmw_and:
81; CHECK:       # %bb.0:
82; CHECK-NEXT:    vl %v0, 0(%r4), 3
83; CHECK-NEXT:    vl %v1, 0(%r3), 4
84; CHECK-NEXT:  .LBB3_1: # %atomicrmw.start
85; CHECK-NEXT:    # =>This Inner Loop Header: Depth=1
86; CHECK-NEXT:    vn %v2, %v1, %v0
87; CHECK-NEXT:    vlgvg %r1, %v2, 1
88; CHECK-NEXT:    vlgvg %r0, %v2, 0
89; CHECK-NEXT:    vlgvg %r5, %v1, 1
90; CHECK-NEXT:    vlgvg %r4, %v1, 0
91; CHECK-NEXT:    cdsg %r4, %r0, 0(%r3)
92; CHECK-NEXT:    vlvgp %v1, %r4, %r5
93; CHECK-NEXT:    jl .LBB3_1
94; CHECK-NEXT:  # %bb.2: # %atomicrmw.end
95; CHECK-NEXT:    vst %v1, 0(%r2), 3
96; CHECK-NEXT:    br %r14
97  %res = atomicrmw and ptr %src, i128 %b seq_cst
98  ret i128 %res
99}
100
101; Check NAND of a variable.
102define i128 @atomicrmw_nand(ptr %src, i128 %b) {
103; CHECK-LABEL: atomicrmw_nand:
104; CHECK:       # %bb.0:
105; CHECK-NEXT:    vl %v0, 0(%r4), 3
106; CHECK-NEXT:    vl %v1, 0(%r3), 4
107; CHECK-NEXT:  .LBB4_1: # %atomicrmw.start
108; CHECK-NEXT:    # =>This Inner Loop Header: Depth=1
109; CHECK-NEXT:    vlgvg %r1, %v1, 1
110; CHECK-NEXT:    vlgvg %r0, %v1, 0
111; CHECK-NEXT:    vnn %v1, %v1, %v0
112; CHECK-NEXT:    vlgvg %r5, %v1, 1
113; CHECK-NEXT:    vlgvg %r4, %v1, 0
114; CHECK-NEXT:    cdsg %r0, %r4, 0(%r3)
115; CHECK-NEXT:    vlvgp %v1, %r0, %r1
116; CHECK-NEXT:    jl .LBB4_1
117; CHECK-NEXT:  # %bb.2: # %atomicrmw.end
118; CHECK-NEXT:    vst %v1, 0(%r2), 3
119; CHECK-NEXT:    br %r14
120  %res = atomicrmw nand ptr %src, i128 %b seq_cst
121  ret i128 %res
122}
123
124; Check OR of a variable.
125define i128 @atomicrmw_or(ptr %src, i128 %b) {
126; CHECK-LABEL: atomicrmw_or:
127; CHECK:       # %bb.0:
128; CHECK-NEXT:    vl %v0, 0(%r4), 3
129; CHECK-NEXT:    vl %v1, 0(%r3), 4
130; CHECK-NEXT:  .LBB5_1: # %atomicrmw.start
131; CHECK-NEXT:    # =>This Inner Loop Header: Depth=1
132; CHECK-NEXT:    vo %v2, %v1, %v0
133; CHECK-NEXT:    vlgvg %r1, %v2, 1
134; CHECK-NEXT:    vlgvg %r0, %v2, 0
135; CHECK-NEXT:    vlgvg %r5, %v1, 1
136; CHECK-NEXT:    vlgvg %r4, %v1, 0
137; CHECK-NEXT:    cdsg %r4, %r0, 0(%r3)
138; CHECK-NEXT:    vlvgp %v1, %r4, %r5
139; CHECK-NEXT:    jl .LBB5_1
140; CHECK-NEXT:  # %bb.2: # %atomicrmw.end
141; CHECK-NEXT:    vst %v1, 0(%r2), 3
142; CHECK-NEXT:    br %r14
143  %res = atomicrmw or ptr %src, i128 %b seq_cst
144  ret i128 %res
145}
146
147; Check XOR of a variable.
148define i128 @atomicrmw_xor(ptr %src, i128 %b) {
149; CHECK-LABEL: atomicrmw_xor:
150; CHECK:       # %bb.0:
151; CHECK-NEXT:    vl %v0, 0(%r4), 3
152; CHECK-NEXT:    vl %v1, 0(%r3), 4
153; CHECK-NEXT:  .LBB6_1: # %atomicrmw.start
154; CHECK-NEXT:    # =>This Inner Loop Header: Depth=1
155; CHECK-NEXT:    vx %v2, %v1, %v0
156; CHECK-NEXT:    vlgvg %r1, %v2, 1
157; CHECK-NEXT:    vlgvg %r0, %v2, 0
158; CHECK-NEXT:    vlgvg %r5, %v1, 1
159; CHECK-NEXT:    vlgvg %r4, %v1, 0
160; CHECK-NEXT:    cdsg %r4, %r0, 0(%r3)
161; CHECK-NEXT:    vlvgp %v1, %r4, %r5
162; CHECK-NEXT:    jl .LBB6_1
163; CHECK-NEXT:  # %bb.2: # %atomicrmw.end
164; CHECK-NEXT:    vst %v1, 0(%r2), 3
165; CHECK-NEXT:    br %r14
166  %res = atomicrmw xor ptr %src, i128 %b seq_cst
167  ret i128 %res
168}
169
170; Check signed minimum.
171define i128 @atomicrmw_min(ptr %src, i128 %b) {
172; CHECK-LABEL: atomicrmw_min:
173; CHECK:       # %bb.0:
174; CHECK-NEXT:    vl %v0, 0(%r4), 3
175; CHECK-NEXT:    vl %v1, 0(%r3), 4
176; CHECK-NEXT:    j .LBB7_2
177; CHECK-NEXT:  .LBB7_1: # %atomicrmw.start
178; CHECK-NEXT:    # in Loop: Header=BB7_2 Depth=1
179; CHECK-NEXT:    vlgvg %r1, %v1, 1
180; CHECK-NEXT:    vlgvg %r0, %v1, 0
181; CHECK-NEXT:    vlgvg %r5, %v2, 1
182; CHECK-NEXT:    vlgvg %r4, %v2, 0
183; CHECK-NEXT:    cdsg %r0, %r4, 0(%r3)
184; CHECK-NEXT:    vlvgp %v1, %r0, %r1
185; CHECK-NEXT:    je .LBB7_6
186; CHECK-NEXT:  .LBB7_2: # %atomicrmw.start
187; CHECK-NEXT:    # =>This Inner Loop Header: Depth=1
188; CHECK-NEXT:    vecg %v0, %v1
189; CHECK-NEXT:    jlh .LBB7_4
190; CHECK-NEXT:  # %bb.3: # %atomicrmw.start
191; CHECK-NEXT:    # in Loop: Header=BB7_2 Depth=1
192; CHECK-NEXT:    vchlgs %v2, %v1, %v0
193; CHECK-NEXT:  .LBB7_4: # %atomicrmw.start
194; CHECK-NEXT:    # in Loop: Header=BB7_2 Depth=1
195; CHECK-NEXT:    vlr %v2, %v1
196; CHECK-NEXT:    jnl .LBB7_1
197; CHECK-NEXT:  # %bb.5: # %atomicrmw.start
198; CHECK-NEXT:    # in Loop: Header=BB7_2 Depth=1
199; CHECK-NEXT:    vlr %v2, %v0
200; CHECK-NEXT:    j .LBB7_1
201; CHECK-NEXT:  .LBB7_6: # %atomicrmw.end
202; CHECK-NEXT:    vst %v1, 0(%r2), 3
203; CHECK-NEXT:    br %r14
204  %res = atomicrmw min ptr %src, i128 %b seq_cst
205  ret i128 %res
206}
207
208; Check signed maximum.
209define i128 @atomicrmw_max(ptr %src, i128 %b) {
210; CHECK-LABEL: atomicrmw_max:
211; CHECK:       # %bb.0:
212; CHECK-NEXT:    vl %v0, 0(%r4), 3
213; CHECK-NEXT:    vl %v1, 0(%r3), 4
214; CHECK-NEXT:    j .LBB8_2
215; CHECK-NEXT:  .LBB8_1: # %atomicrmw.start
216; CHECK-NEXT:    # in Loop: Header=BB8_2 Depth=1
217; CHECK-NEXT:    vlgvg %r1, %v1, 1
218; CHECK-NEXT:    vlgvg %r0, %v1, 0
219; CHECK-NEXT:    vlgvg %r5, %v2, 1
220; CHECK-NEXT:    vlgvg %r4, %v2, 0
221; CHECK-NEXT:    cdsg %r0, %r4, 0(%r3)
222; CHECK-NEXT:    vlvgp %v1, %r0, %r1
223; CHECK-NEXT:    je .LBB8_6
224; CHECK-NEXT:  .LBB8_2: # %atomicrmw.start
225; CHECK-NEXT:    # =>This Inner Loop Header: Depth=1
226; CHECK-NEXT:    vecg %v0, %v1
227; CHECK-NEXT:    jlh .LBB8_4
228; CHECK-NEXT:  # %bb.3: # %atomicrmw.start
229; CHECK-NEXT:    # in Loop: Header=BB8_2 Depth=1
230; CHECK-NEXT:    vchlgs %v2, %v1, %v0
231; CHECK-NEXT:  .LBB8_4: # %atomicrmw.start
232; CHECK-NEXT:    # in Loop: Header=BB8_2 Depth=1
233; CHECK-NEXT:    vlr %v2, %v1
234; CHECK-NEXT:    jl .LBB8_1
235; CHECK-NEXT:  # %bb.5: # %atomicrmw.start
236; CHECK-NEXT:    # in Loop: Header=BB8_2 Depth=1
237; CHECK-NEXT:    vlr %v2, %v0
238; CHECK-NEXT:    j .LBB8_1
239; CHECK-NEXT:  .LBB8_6: # %atomicrmw.end
240; CHECK-NEXT:    vst %v1, 0(%r2), 3
241; CHECK-NEXT:    br %r14
242  %res = atomicrmw max ptr %src, i128 %b seq_cst
243  ret i128 %res
244}
245
246; Check unsigned minimum.
247define i128 @atomicrmw_umin(ptr %src, i128 %b) {
248; CHECK-LABEL: atomicrmw_umin:
249; CHECK:       # %bb.0:
250; CHECK-NEXT:    vl %v0, 0(%r4), 3
251; CHECK-NEXT:    vl %v1, 0(%r3), 4
252; CHECK-NEXT:    j .LBB9_2
253; CHECK-NEXT:  .LBB9_1: # %atomicrmw.start
254; CHECK-NEXT:    # in Loop: Header=BB9_2 Depth=1
255; CHECK-NEXT:    vlgvg %r1, %v1, 1
256; CHECK-NEXT:    vlgvg %r0, %v1, 0
257; CHECK-NEXT:    vlgvg %r5, %v2, 1
258; CHECK-NEXT:    vlgvg %r4, %v2, 0
259; CHECK-NEXT:    cdsg %r0, %r4, 0(%r3)
260; CHECK-NEXT:    vlvgp %v1, %r0, %r1
261; CHECK-NEXT:    je .LBB9_6
262; CHECK-NEXT:  .LBB9_2: # %atomicrmw.start
263; CHECK-NEXT:    # =>This Inner Loop Header: Depth=1
264; CHECK-NEXT:    veclg %v0, %v1
265; CHECK-NEXT:    jlh .LBB9_4
266; CHECK-NEXT:  # %bb.3: # %atomicrmw.start
267; CHECK-NEXT:    # in Loop: Header=BB9_2 Depth=1
268; CHECK-NEXT:    vchlgs %v2, %v1, %v0
269; CHECK-NEXT:  .LBB9_4: # %atomicrmw.start
270; CHECK-NEXT:    # in Loop: Header=BB9_2 Depth=1
271; CHECK-NEXT:    vlr %v2, %v1
272; CHECK-NEXT:    jnl .LBB9_1
273; CHECK-NEXT:  # %bb.5: # %atomicrmw.start
274; CHECK-NEXT:    # in Loop: Header=BB9_2 Depth=1
275; CHECK-NEXT:    vlr %v2, %v0
276; CHECK-NEXT:    j .LBB9_1
277; CHECK-NEXT:  .LBB9_6: # %atomicrmw.end
278; CHECK-NEXT:    vst %v1, 0(%r2), 3
279; CHECK-NEXT:    br %r14
280  %res = atomicrmw umin ptr %src, i128 %b seq_cst
281  ret i128 %res
282}
283
284; Check unsigned maximum.
285define i128 @atomicrmw_umax(ptr %src, i128 %b) {
286; CHECK-LABEL: atomicrmw_umax:
287; CHECK:       # %bb.0:
288; CHECK-NEXT:    vl %v0, 0(%r4), 3
289; CHECK-NEXT:    vl %v1, 0(%r3), 4
290; CHECK-NEXT:    j .LBB10_2
291; CHECK-NEXT:  .LBB10_1: # %atomicrmw.start
292; CHECK-NEXT:    # in Loop: Header=BB10_2 Depth=1
293; CHECK-NEXT:    vlgvg %r1, %v1, 1
294; CHECK-NEXT:    vlgvg %r0, %v1, 0
295; CHECK-NEXT:    vlgvg %r5, %v2, 1
296; CHECK-NEXT:    vlgvg %r4, %v2, 0
297; CHECK-NEXT:    cdsg %r0, %r4, 0(%r3)
298; CHECK-NEXT:    vlvgp %v1, %r0, %r1
299; CHECK-NEXT:    je .LBB10_6
300; CHECK-NEXT:  .LBB10_2: # %atomicrmw.start
301; CHECK-NEXT:    # =>This Inner Loop Header: Depth=1
302; CHECK-NEXT:    veclg %v0, %v1
303; CHECK-NEXT:    jlh .LBB10_4
304; CHECK-NEXT:  # %bb.3: # %atomicrmw.start
305; CHECK-NEXT:    # in Loop: Header=BB10_2 Depth=1
306; CHECK-NEXT:    vchlgs %v2, %v1, %v0
307; CHECK-NEXT:  .LBB10_4: # %atomicrmw.start
308; CHECK-NEXT:    # in Loop: Header=BB10_2 Depth=1
309; CHECK-NEXT:    vlr %v2, %v1
310; CHECK-NEXT:    jl .LBB10_1
311; CHECK-NEXT:  # %bb.5: # %atomicrmw.start
312; CHECK-NEXT:    # in Loop: Header=BB10_2 Depth=1
313; CHECK-NEXT:    vlr %v2, %v0
314; CHECK-NEXT:    j .LBB10_1
315; CHECK-NEXT:  .LBB10_6: # %atomicrmw.end
316; CHECK-NEXT:    vst %v1, 0(%r2), 3
317; CHECK-NEXT:    br %r14
318  %res = atomicrmw umax ptr %src, i128 %b seq_cst
319  ret i128 %res
320}
321
322; Check increment with wraparound.
323define i128 @atomicrmw_uinc_wrap(ptr %src, i128 %b) {
324; CHECK-LABEL: atomicrmw_uinc_wrap:
325; CHECK:       # %bb.0:
326; CHECK-NEXT:    larl %r1, .LCPI11_0
327; CHECK-NEXT:    vl %v0, 0(%r4), 3
328; CHECK-NEXT:    vl %v2, 0(%r3), 4
329; CHECK-NEXT:    vl %v1, 0(%r1), 3
330; CHECK-NEXT:    j .LBB11_2
331; CHECK-NEXT:  .LBB11_1: # %atomicrmw.start
332; CHECK-NEXT:    # in Loop: Header=BB11_2 Depth=1
333; CHECK-NEXT:    vlgvg %r1, %v2, 1
334; CHECK-NEXT:    vlgvg %r0, %v2, 0
335; CHECK-NEXT:    vlgvg %r5, %v3, 1
336; CHECK-NEXT:    vlgvg %r4, %v3, 0
337; CHECK-NEXT:    cdsg %r0, %r4, 0(%r3)
338; CHECK-NEXT:    vlvgp %v2, %r0, %r1
339; CHECK-NEXT:    je .LBB11_6
340; CHECK-NEXT:  .LBB11_2: # %atomicrmw.start
341; CHECK-NEXT:    # =>This Inner Loop Header: Depth=1
342; CHECK-NEXT:    veclg %v2, %v0
343; CHECK-NEXT:    jlh .LBB11_4
344; CHECK-NEXT:  # %bb.3: # %atomicrmw.start
345; CHECK-NEXT:    # in Loop: Header=BB11_2 Depth=1
346; CHECK-NEXT:    vchlgs %v3, %v0, %v2
347; CHECK-NEXT:  .LBB11_4: # %atomicrmw.start
348; CHECK-NEXT:    # in Loop: Header=BB11_2 Depth=1
349; CHECK-NEXT:    vgbm %v3, 0
350; CHECK-NEXT:    jnl .LBB11_1
351; CHECK-NEXT:  # %bb.5: # %atomicrmw.start
352; CHECK-NEXT:    # in Loop: Header=BB11_2 Depth=1
353; CHECK-NEXT:    vaq %v3, %v2, %v1
354; CHECK-NEXT:    j .LBB11_1
355; CHECK-NEXT:  .LBB11_6: # %atomicrmw.end
356; CHECK-NEXT:    vst %v2, 0(%r2), 3
357; CHECK-NEXT:    br %r14
358  %res = atomicrmw uinc_wrap ptr %src, i128 %b seq_cst
359  ret i128 %res
360}
361
362; Check decrement with wraparound.
363define i128 @atomicrmw_udec_wrap(ptr %src, i128 %b) {
364; CHECK-LABEL: atomicrmw_udec_wrap:
365; CHECK:       # %bb.0:
366; CHECK-NEXT:    vl %v0, 0(%r4), 3
367; CHECK-NEXT:    vl %v3, 0(%r3), 4
368; CHECK-NEXT:    vgbm %v1, 65535
369; CHECK-NEXT:    vgbm %v2, 0
370; CHECK-NEXT:    j .LBB12_2
371; CHECK-NEXT:  .LBB12_1: # %atomicrmw.start
372; CHECK-NEXT:    # in Loop: Header=BB12_2 Depth=1
373; CHECK-NEXT:    vlgvg %r1, %v3, 1
374; CHECK-NEXT:    vlgvg %r0, %v3, 0
375; CHECK-NEXT:    vlgvg %r5, %v5, 1
376; CHECK-NEXT:    vlgvg %r4, %v5, 0
377; CHECK-NEXT:    cdsg %r0, %r4, 0(%r3)
378; CHECK-NEXT:    vlvgp %v3, %r0, %r1
379; CHECK-NEXT:    je .LBB12_8
380; CHECK-NEXT:  .LBB12_2: # %atomicrmw.start
381; CHECK-NEXT:    # =>This Inner Loop Header: Depth=1
382; CHECK-NEXT:    veclg %v0, %v3
383; CHECK-NEXT:    jlh .LBB12_4
384; CHECK-NEXT:  # %bb.3: # %atomicrmw.start
385; CHECK-NEXT:    # in Loop: Header=BB12_2 Depth=1
386; CHECK-NEXT:    vchlgs %v4, %v3, %v0
387; CHECK-NEXT:  .LBB12_4: # %atomicrmw.start
388; CHECK-NEXT:    # in Loop: Header=BB12_2 Depth=1
389; CHECK-NEXT:    vlr %v4, %v0
390; CHECK-NEXT:    jl .LBB12_6
391; CHECK-NEXT:  # %bb.5: # %atomicrmw.start
392; CHECK-NEXT:    # in Loop: Header=BB12_2 Depth=1
393; CHECK-NEXT:    vaq %v4, %v3, %v1
394; CHECK-NEXT:  .LBB12_6: # %atomicrmw.start
395; CHECK-NEXT:    # in Loop: Header=BB12_2 Depth=1
396; CHECK-NEXT:    vceqgs %v5, %v3, %v2
397; CHECK-NEXT:    vlr %v5, %v0
398; CHECK-NEXT:    je .LBB12_1
399; CHECK-NEXT:  # %bb.7: # %atomicrmw.start
400; CHECK-NEXT:    # in Loop: Header=BB12_2 Depth=1
401; CHECK-NEXT:    vlr %v5, %v4
402; CHECK-NEXT:    j .LBB12_1
403; CHECK-NEXT:  .LBB12_8: # %atomicrmw.end
404; CHECK-NEXT:    vst %v3, 0(%r2), 3
405; CHECK-NEXT:    br %r14
406  %res = atomicrmw udec_wrap ptr %src, i128 %b seq_cst
407  ret i128 %res
408}
409