Skip to content

Commit ba3fa39

Browse files
fhahnfhahn
committed
[EarlyCSE] Re-generate checks for intrinsics.ll.
1 parent 283a78a commit ba3fa39

File tree

1 file changed

+144
-15
lines changed

1 file changed

+144
-15
lines changed

‎llvm/test/Transforms/EarlyCSE/AArch64/intrinsics.ll

Lines changed: 144 additions & 15 deletions
Original file line numberDiff line numberDiff line change
@@ -1,11 +1,33 @@
1+
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
12
; RUN: opt < %s -S -mtriple=aarch64-none-linux-gnu -mattr=+neon -passes=early-cse -earlycse-debug-hash | FileCheck %s
23
; RUN: opt < %s -S -mtriple=aarch64-none-linux-gnu -mattr=+neon -aa-pipeline=basic-aa -passes='early-cse<memssa>' | FileCheck %s
34

45
define <4 x i32> @test_cse(ptr %a, [2 x <4 x i32>] %s.coerce, i32 %n) {
6+
; CHECK-LABEL: define <4 x i32> @test_cse(
7+
; CHECK-SAME: ptr [[A:%.*]], [2 x <4 x i32>] [[S_COERCE:%.*]], i32 [[N:%.*]]) #[[ATTR0:[0-9]+]] {
8+
; CHECK-NEXT: [[ENTRY:.*]]:
9+
; CHECK-NEXT: [[S_COERCE_FCA_0_EXTRACT:%.*]] = extractvalue [2 x <4 x i32>] [[S_COERCE]], 0
10+
; CHECK-NEXT: [[S_COERCE_FCA_1_EXTRACT:%.*]] = extractvalue [2 x <4 x i32>] [[S_COERCE]], 1
11+
; CHECK-NEXT: br label %[[FOR_COND:.*]]
12+
; CHECK: [[FOR_COND]]:
13+
; CHECK-NEXT: [[I_0:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ [[INC:%.*]], %[[FOR_BODY:.*]] ]
14+
; CHECK-NEXT: [[RES_0:%.*]] = phi <4 x i32> [ undef, %[[ENTRY]] ], [ [[CALL:%.*]], %[[FOR_BODY]] ]
15+
; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], [[N]]
16+
; CHECK-NEXT: br i1 [[CMP]], label %[[FOR_BODY]], label %[[FOR_END:.*]]
17+
; CHECK: [[FOR_BODY]]:
18+
; CHECK-NEXT: [[TMP0:%.*]] = bitcast <4 x i32> [[S_COERCE_FCA_0_EXTRACT]] to <16 x i8>
19+
; CHECK-NEXT: [[TMP1:%.*]] = bitcast <4 x i32> [[S_COERCE_FCA_1_EXTRACT]] to <16 x i8>
20+
; CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <4 x i32>, <4 x i32> } poison, <4 x i32> [[S_COERCE_FCA_0_EXTRACT]], 0
21+
; CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <4 x i32>, <4 x i32> } [[TMP2]], <4 x i32> [[S_COERCE_FCA_1_EXTRACT]], 1
22+
; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v4i32.p0(<4 x i32> [[S_COERCE_FCA_0_EXTRACT]], <4 x i32> [[S_COERCE_FCA_1_EXTRACT]], ptr [[A]])
23+
; CHECK-NEXT: [[CALL]] = call <4 x i32> @vaddq_s32(<4 x i32> [[S_COERCE_FCA_0_EXTRACT]], <4 x i32> [[S_COERCE_FCA_0_EXTRACT]])
24+
; CHECK-NEXT: [[INC]] = add nsw i32 [[I_0]], 1
25+
; CHECK-NEXT: br label %[[FOR_COND]]
26+
; CHECK: [[FOR_END]]:
27+
; CHECK-NEXT: ret <4 x i32> [[RES_0]]
28+
;
529
entry:
630
; Check that @llvm.aarch64.neon.ld2 is optimized away by Early CSE.
7-
; CHECK-LABEL: @test_cse
8-
; CHECK-NOT: call { <4 x i32>, <4 x i32> } @llvm.aarch64.neon.ld2.v4i32.p0
931
%s.coerce.fca.0.extract = extractvalue [2 x <4 x i32>] %s.coerce, 0
1032
%s.coerce.fca.1.extract = extractvalue [2 x <4 x i32>] %s.coerce, 1
1133
br label %for.cond
@@ -34,11 +56,32 @@ for.end: ; preds = %for.cond
3456
}
3557

3658
define <4 x i32> @test_cse2(ptr %a, [2 x <4 x i32>] %s.coerce, i32 %n) {
59+
; CHECK-LABEL: define <4 x i32> @test_cse2(
60+
; CHECK-SAME: ptr [[A:%.*]], [2 x <4 x i32>] [[S_COERCE:%.*]], i32 [[N:%.*]]) #[[ATTR0]] {
61+
; CHECK-NEXT: [[ENTRY:.*]]:
62+
; CHECK-NEXT: [[S_COERCE_FCA_0_EXTRACT:%.*]] = extractvalue [2 x <4 x i32>] [[S_COERCE]], 0
63+
; CHECK-NEXT: [[S_COERCE_FCA_1_EXTRACT:%.*]] = extractvalue [2 x <4 x i32>] [[S_COERCE]], 1
64+
; CHECK-NEXT: br label %[[FOR_COND:.*]]
65+
; CHECK: [[FOR_COND]]:
66+
; CHECK-NEXT: [[I_0:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ [[INC:%.*]], %[[FOR_BODY:.*]] ]
67+
; CHECK-NEXT: [[RES_0:%.*]] = phi <4 x i32> [ undef, %[[ENTRY]] ], [ [[CALL:%.*]], %[[FOR_BODY]] ]
68+
; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], [[N]]
69+
; CHECK-NEXT: br i1 [[CMP]], label %[[FOR_BODY]], label %[[FOR_END:.*]]
70+
; CHECK: [[FOR_BODY]]:
71+
; CHECK-NEXT: [[TMP0:%.*]] = bitcast <4 x i32> [[S_COERCE_FCA_0_EXTRACT]] to <16 x i8>
72+
; CHECK-NEXT: [[TMP1:%.*]] = bitcast <4 x i32> [[S_COERCE_FCA_1_EXTRACT]] to <16 x i8>
73+
; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v4i32.p0(<4 x i32> [[S_COERCE_FCA_0_EXTRACT]], <4 x i32> [[S_COERCE_FCA_0_EXTRACT]], ptr [[A]])
74+
; CHECK-NEXT: [[TMP2:%.*]] = insertvalue { <4 x i32>, <4 x i32> } poison, <4 x i32> [[S_COERCE_FCA_0_EXTRACT]], 0
75+
; CHECK-NEXT: [[TMP3:%.*]] = insertvalue { <4 x i32>, <4 x i32> } [[TMP2]], <4 x i32> [[S_COERCE_FCA_1_EXTRACT]], 1
76+
; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v4i32.p0(<4 x i32> [[S_COERCE_FCA_0_EXTRACT]], <4 x i32> [[S_COERCE_FCA_1_EXTRACT]], ptr [[A]])
77+
; CHECK-NEXT: [[CALL]] = call <4 x i32> @vaddq_s32(<4 x i32> [[S_COERCE_FCA_0_EXTRACT]], <4 x i32> [[S_COERCE_FCA_0_EXTRACT]])
78+
; CHECK-NEXT: [[INC]] = add nsw i32 [[I_0]], 1
79+
; CHECK-NEXT: br label %[[FOR_COND]]
80+
; CHECK: [[FOR_END]]:
81+
; CHECK-NEXT: ret <4 x i32> [[RES_0]]
82+
;
3783
entry:
3884
; Check that the first @llvm.aarch64.neon.st2 is optimized away by Early CSE.
39-
; CHECK-LABEL: @test_cse2
40-
; CHECK-NOT: call void @llvm.aarch64.neon.st2.v4i32.p0(<4 x i32> %3, <4 x i32> %3, ptr %0)
41-
; CHECK: call void @llvm.aarch64.neon.st2.v4i32.p0(<4 x i32> %s.coerce.fca.0.extract, <4 x i32> %s.coerce.fca.1.extract, ptr %a)
4285
%s.coerce.fca.0.extract = extractvalue [2 x <4 x i32>] %s.coerce, 0
4386
%s.coerce.fca.1.extract = extractvalue [2 x <4 x i32>] %s.coerce, 1
4487
br label %for.cond
@@ -68,11 +111,26 @@ for.end: ; preds = %for.cond
68111
}
69112

70113
define <4 x i32> @test_cse3(ptr %a, [2 x <4 x i32>] %s.coerce, i32 %n) #0 {
114+
; CHECK-LABEL: define <4 x i32> @test_cse3(
115+
; CHECK-SAME: ptr [[A:%.*]], [2 x <4 x i32>] [[S_COERCE:%.*]], i32 [[N:%.*]]) #[[ATTR0]] {
116+
; CHECK-NEXT: [[ENTRY:.*]]:
117+
; CHECK-NEXT: br label %[[FOR_COND:.*]]
118+
; CHECK: [[FOR_COND]]:
119+
; CHECK-NEXT: [[I_0:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ [[INC:%.*]], %[[FOR_BODY:.*]] ]
120+
; CHECK-NEXT: [[RES_0:%.*]] = phi <4 x i32> [ undef, %[[ENTRY]] ], [ [[CALL:%.*]], %[[FOR_BODY]] ]
121+
; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], [[N]]
122+
; CHECK-NEXT: br i1 [[CMP]], label %[[FOR_BODY]], label %[[FOR_END:.*]]
123+
; CHECK: [[FOR_BODY]]:
124+
; CHECK-NEXT: [[VLD2:%.*]] = call { <4 x i32>, <4 x i32> } @llvm.aarch64.neon.ld2.v4i32.p0(ptr [[A]])
125+
; CHECK-NEXT: [[VLD2_FCA_0_EXTRACT:%.*]] = extractvalue { <4 x i32>, <4 x i32> } [[VLD2]], 0
126+
; CHECK-NEXT: [[CALL]] = call <4 x i32> @vaddq_s32(<4 x i32> [[VLD2_FCA_0_EXTRACT]], <4 x i32> [[VLD2_FCA_0_EXTRACT]])
127+
; CHECK-NEXT: [[INC]] = add nsw i32 [[I_0]], 1
128+
; CHECK-NEXT: br label %[[FOR_COND]]
129+
; CHECK: [[FOR_END]]:
130+
; CHECK-NEXT: ret <4 x i32> [[RES_0]]
131+
;
71132
entry:
72133
; Check that the first @llvm.aarch64.neon.ld2 is optimized away by Early CSE.
73-
; CHECK-LABEL: @test_cse3
74-
; CHECK: call { <4 x i32>, <4 x i32> } @llvm.aarch64.neon.ld2.v4i32.p0
75-
; CHECK-NOT: call { <4 x i32>, <4 x i32> } @llvm.aarch64.neon.ld2.v4i32.p0
76134
%s.coerce.fca.0.extract = extractvalue [2 x <4 x i32>] %s.coerce, 0
77135
%s.coerce.fca.1.extract = extractvalue [2 x <4 x i32>] %s.coerce, 1
78136
br label %for.cond
@@ -100,11 +158,33 @@ for.end: ; preds = %for.cond
100158

101159

102160
define <4 x i32> @test_nocse(ptr %a, ptr %b, [2 x <4 x i32>] %s.coerce, i32 %n) {
161+
; CHECK-LABEL: define <4 x i32> @test_nocse(
162+
; CHECK-SAME: ptr [[A:%.*]], ptr [[B:%.*]], [2 x <4 x i32>] [[S_COERCE:%.*]], i32 [[N:%.*]]) #[[ATTR0]] {
163+
; CHECK-NEXT: [[ENTRY:.*]]:
164+
; CHECK-NEXT: [[S_COERCE_FCA_0_EXTRACT:%.*]] = extractvalue [2 x <4 x i32>] [[S_COERCE]], 0
165+
; CHECK-NEXT: [[S_COERCE_FCA_1_EXTRACT:%.*]] = extractvalue [2 x <4 x i32>] [[S_COERCE]], 1
166+
; CHECK-NEXT: br label %[[FOR_COND:.*]]
167+
; CHECK: [[FOR_COND]]:
168+
; CHECK-NEXT: [[I_0:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ [[INC:%.*]], %[[FOR_BODY:.*]] ]
169+
; CHECK-NEXT: [[RES_0:%.*]] = phi <4 x i32> [ undef, %[[ENTRY]] ], [ [[CALL:%.*]], %[[FOR_BODY]] ]
170+
; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], [[N]]
171+
; CHECK-NEXT: br i1 [[CMP]], label %[[FOR_BODY]], label %[[FOR_END:.*]]
172+
; CHECK: [[FOR_BODY]]:
173+
; CHECK-NEXT: [[TMP0:%.*]] = bitcast <4 x i32> [[S_COERCE_FCA_0_EXTRACT]] to <16 x i8>
174+
; CHECK-NEXT: [[TMP1:%.*]] = bitcast <4 x i32> [[S_COERCE_FCA_1_EXTRACT]] to <16 x i8>
175+
; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v4i32.p0(<4 x i32> [[S_COERCE_FCA_0_EXTRACT]], <4 x i32> [[S_COERCE_FCA_1_EXTRACT]], ptr [[A]])
176+
; CHECK-NEXT: store i32 0, ptr [[B]], align 4
177+
; CHECK-NEXT: [[VLD2:%.*]] = call { <4 x i32>, <4 x i32> } @llvm.aarch64.neon.ld2.v4i32.p0(ptr [[A]])
178+
; CHECK-NEXT: [[VLD2_FCA_0_EXTRACT:%.*]] = extractvalue { <4 x i32>, <4 x i32> } [[VLD2]], 0
179+
; CHECK-NEXT: [[CALL]] = call <4 x i32> @vaddq_s32(<4 x i32> [[VLD2_FCA_0_EXTRACT]], <4 x i32> [[VLD2_FCA_0_EXTRACT]])
180+
; CHECK-NEXT: [[INC]] = add nsw i32 [[I_0]], 1
181+
; CHECK-NEXT: br label %[[FOR_COND]]
182+
; CHECK: [[FOR_END]]:
183+
; CHECK-NEXT: ret <4 x i32> [[RES_0]]
184+
;
103185
entry:
104186
; Check that the store prevents @llvm.aarch64.neon.ld2 from being optimized
105187
; away by Early CSE.
106-
; CHECK-LABEL: @test_nocse
107-
; CHECK: call { <4 x i32>, <4 x i32> } @llvm.aarch64.neon.ld2.v4i32.p0
108188
%s.coerce.fca.0.extract = extractvalue [2 x <4 x i32>] %s.coerce, 0
109189
%s.coerce.fca.1.extract = extractvalue [2 x <4 x i32>] %s.coerce, 1
110190
br label %for.cond
@@ -134,11 +214,33 @@ for.end: ; preds = %for.cond
134214
}
135215

136216
define <4 x i32> @test_nocse2(ptr %a, [2 x <4 x i32>] %s.coerce, i32 %n) {
217+
; CHECK-LABEL: define <4 x i32> @test_nocse2(
218+
; CHECK-SAME: ptr [[A:%.*]], [2 x <4 x i32>] [[S_COERCE:%.*]], i32 [[N:%.*]]) #[[ATTR0]] {
219+
; CHECK-NEXT: [[ENTRY:.*]]:
220+
; CHECK-NEXT: [[S_COERCE_FCA_0_EXTRACT:%.*]] = extractvalue [2 x <4 x i32>] [[S_COERCE]], 0
221+
; CHECK-NEXT: [[S_COERCE_FCA_1_EXTRACT:%.*]] = extractvalue [2 x <4 x i32>] [[S_COERCE]], 1
222+
; CHECK-NEXT: br label %[[FOR_COND:.*]]
223+
; CHECK: [[FOR_COND]]:
224+
; CHECK-NEXT: [[I_0:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ [[INC:%.*]], %[[FOR_BODY:.*]] ]
225+
; CHECK-NEXT: [[RES_0:%.*]] = phi <4 x i32> [ undef, %[[ENTRY]] ], [ [[CALL:%.*]], %[[FOR_BODY]] ]
226+
; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], [[N]]
227+
; CHECK-NEXT: br i1 [[CMP]], label %[[FOR_BODY]], label %[[FOR_END:.*]]
228+
; CHECK: [[FOR_BODY]]:
229+
; CHECK-NEXT: [[TMP0:%.*]] = bitcast <4 x i32> [[S_COERCE_FCA_0_EXTRACT]] to <16 x i8>
230+
; CHECK-NEXT: [[TMP1:%.*]] = bitcast <4 x i32> [[S_COERCE_FCA_1_EXTRACT]] to <16 x i8>
231+
; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v4i32.p0(<4 x i32> [[S_COERCE_FCA_0_EXTRACT]], <4 x i32> [[S_COERCE_FCA_1_EXTRACT]], ptr [[A]])
232+
; CHECK-NEXT: [[VLD3:%.*]] = call { <4 x i32>, <4 x i32>, <4 x i32> } @llvm.aarch64.neon.ld3.v4i32.p0(ptr [[A]])
233+
; CHECK-NEXT: [[VLD3_FCA_0_EXTRACT:%.*]] = extractvalue { <4 x i32>, <4 x i32>, <4 x i32> } [[VLD3]], 0
234+
; CHECK-NEXT: [[VLD3_FCA_2_EXTRACT:%.*]] = extractvalue { <4 x i32>, <4 x i32>, <4 x i32> } [[VLD3]], 2
235+
; CHECK-NEXT: [[CALL]] = call <4 x i32> @vaddq_s32(<4 x i32> [[VLD3_FCA_0_EXTRACT]], <4 x i32> [[VLD3_FCA_2_EXTRACT]])
236+
; CHECK-NEXT: [[INC]] = add nsw i32 [[I_0]], 1
237+
; CHECK-NEXT: br label %[[FOR_COND]]
238+
; CHECK: [[FOR_END]]:
239+
; CHECK-NEXT: ret <4 x i32> [[RES_0]]
240+
;
137241
entry:
138242
; Check that @llvm.aarch64.neon.ld3 is not optimized away by Early CSE due
139243
; to mismatch between st2 and ld3.
140-
; CHECK-LABEL: @test_nocse2
141-
; CHECK: call { <4 x i32>, <4 x i32>, <4 x i32> } @llvm.aarch64.neon.ld3.v4i32.p0
142244
%s.coerce.fca.0.extract = extractvalue [2 x <4 x i32>] %s.coerce, 0
143245
%s.coerce.fca.1.extract = extractvalue [2 x <4 x i32>] %s.coerce, 1
144246
br label %for.cond
@@ -167,12 +269,33 @@ for.end: ; preds = %for.cond
167269
}
168270

169271
define <4 x i32> @test_nocse3(ptr %a, [2 x <4 x i32>] %s.coerce, i32 %n) {
272+
; CHECK-LABEL: define <4 x i32> @test_nocse3(
273+
; CHECK-SAME: ptr [[A:%.*]], [2 x <4 x i32>] [[S_COERCE:%.*]], i32 [[N:%.*]]) #[[ATTR0]] {
274+
; CHECK-NEXT: [[ENTRY:.*]]:
275+
; CHECK-NEXT: [[S_COERCE_FCA_0_EXTRACT:%.*]] = extractvalue [2 x <4 x i32>] [[S_COERCE]], 0
276+
; CHECK-NEXT: [[S_COERCE_FCA_1_EXTRACT:%.*]] = extractvalue [2 x <4 x i32>] [[S_COERCE]], 1
277+
; CHECK-NEXT: br label %[[FOR_COND:.*]]
278+
; CHECK: [[FOR_COND]]:
279+
; CHECK-NEXT: [[I_0:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ [[INC:%.*]], %[[FOR_BODY:.*]] ]
280+
; CHECK-NEXT: [[RES_0:%.*]] = phi <4 x i32> [ undef, %[[ENTRY]] ], [ [[CALL:%.*]], %[[FOR_BODY]] ]
281+
; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], [[N]]
282+
; CHECK-NEXT: br i1 [[CMP]], label %[[FOR_BODY]], label %[[FOR_END:.*]]
283+
; CHECK: [[FOR_BODY]]:
284+
; CHECK-NEXT: [[TMP0:%.*]] = bitcast <4 x i32> [[S_COERCE_FCA_0_EXTRACT]] to <16 x i8>
285+
; CHECK-NEXT: [[TMP1:%.*]] = bitcast <4 x i32> [[S_COERCE_FCA_1_EXTRACT]] to <16 x i8>
286+
; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v4i32.p0(<4 x i32> [[S_COERCE_FCA_1_EXTRACT]], <4 x i32> [[S_COERCE_FCA_0_EXTRACT]], <4 x i32> [[S_COERCE_FCA_0_EXTRACT]], ptr [[A]])
287+
; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v4i32.p0(<4 x i32> [[S_COERCE_FCA_0_EXTRACT]], <4 x i32> [[S_COERCE_FCA_0_EXTRACT]], ptr [[A]])
288+
; CHECK-NEXT: [[VLD3:%.*]] = call { <4 x i32>, <4 x i32>, <4 x i32> } @llvm.aarch64.neon.ld3.v4i32.p0(ptr [[A]])
289+
; CHECK-NEXT: [[VLD3_FCA_0_EXTRACT:%.*]] = extractvalue { <4 x i32>, <4 x i32>, <4 x i32> } [[VLD3]], 0
290+
; CHECK-NEXT: [[CALL]] = call <4 x i32> @vaddq_s32(<4 x i32> [[VLD3_FCA_0_EXTRACT]], <4 x i32> [[VLD3_FCA_0_EXTRACT]])
291+
; CHECK-NEXT: [[INC]] = add nsw i32 [[I_0]], 1
292+
; CHECK-NEXT: br label %[[FOR_COND]]
293+
; CHECK: [[FOR_END]]:
294+
; CHECK-NEXT: ret <4 x i32> [[RES_0]]
295+
;
170296
entry:
171297
; Check that @llvm.aarch64.neon.st3 is not optimized away by Early CSE due to
172298
; mismatch between st2 and st3.
173-
; CHECK-LABEL: @test_nocse3
174-
; CHECK: call void @llvm.aarch64.neon.st3.v4i32.p0
175-
; CHECK: call void @llvm.aarch64.neon.st2.v4i32.p0
176299
%s.coerce.fca.0.extract = extractvalue [2 x <4 x i32>] %s.coerce, 0
177300
%s.coerce.fca.1.extract = extractvalue [2 x <4 x i32>] %s.coerce, 1
178301
br label %for.cond
@@ -214,6 +337,12 @@ declare { <4 x i32>, <4 x i32> } @llvm.aarch64.neon.ld2.v4i32.p0(ptr)
214337
declare { <4 x i32>, <4 x i32>, <4 x i32> } @llvm.aarch64.neon.ld3.v4i32.p0(ptr)
215338

216339
define internal fastcc <4 x i32> @vaddq_s32(<4 x i32> %__p0, <4 x i32> %__p1) {
340+
; CHECK-LABEL: define internal fastcc <4 x i32> @vaddq_s32(
341+
; CHECK-SAME: <4 x i32> [[__P0:%.*]], <4 x i32> [[__P1:%.*]]) #[[ATTR0]] {
342+
; CHECK-NEXT: [[ENTRY:.*:]]
343+
; CHECK-NEXT: [[ADD:%.*]] = add <4 x i32> [[__P0]], [[__P1]]
344+
; CHECK-NEXT: ret <4 x i32> [[ADD]]
345+
;
217346
entry:
218347
%add = add <4 x i32> %__p0, %__p1
219348
ret <4 x i32> %add

0 commit comments

Comments
 (0)