1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2; RUN: opt -slp-vectorizer  -mtriple=arm64-apple-ios -S %s | FileCheck %s
3
4define void @loop1(i32* %A, i32* %B, i64 %N) {
5; CHECK-LABEL: @loop1(
6; CHECK-NEXT:  entry:
7; CHECK-NEXT:    br label [[LOOP:%.*]]
8; CHECK:       loop:
9; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
10; CHECK-NEXT:    [[B_GEP_0:%.*]] = getelementptr inbounds i32, i32* [[B:%.*]], i64 [[IV]]
11; CHECK-NEXT:    [[B_0:%.*]] = load i32, i32* [[B_GEP_0]], align 4
12; CHECK-NEXT:    [[A_GEP_0:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[IV]]
13; CHECK-NEXT:    [[A_0:%.*]] = load i32, i32* [[A_GEP_0]], align 4
14; CHECK-NEXT:    [[ADD_0:%.*]] = add i32 [[A_0]], 20
15; CHECK-NEXT:    [[XOR_0:%.*]] = xor i32 [[ADD_0]], [[B_0]]
16; CHECK-NEXT:    store i32 [[XOR_0]], i32* [[A_GEP_0]], align 4
17; CHECK-NEXT:    [[IV_1:%.*]] = or i64 [[IV]], 1
18; CHECK-NEXT:    [[B_GEP_1:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[IV_1]]
19; CHECK-NEXT:    [[B_1:%.*]] = load i32, i32* [[B_GEP_1]], align 4
20; CHECK-NEXT:    [[A_GEP_1:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[IV_1]]
21; CHECK-NEXT:    [[A_1:%.*]] = load i32, i32* [[A_GEP_1]], align 4
22; CHECK-NEXT:    [[ADD_1:%.*]] = add i32 [[A_1]], 20
23; CHECK-NEXT:    [[XOR_1:%.*]] = xor i32 [[ADD_1]], [[B_1]]
24; CHECK-NEXT:    store i32 [[XOR_1]], i32* [[A_GEP_1]], align 4
25; CHECK-NEXT:    [[IV_2:%.*]] = or i64 [[IV]], 2
26; CHECK-NEXT:    [[B_GEP_2:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[IV_2]]
27; CHECK-NEXT:    [[B_2:%.*]] = load i32, i32* [[B_GEP_2]], align 4
28; CHECK-NEXT:    [[A_GEP_2:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[IV_2]]
29; CHECK-NEXT:    [[A_2:%.*]] = load i32, i32* [[A_GEP_2]], align 4
30; CHECK-NEXT:    [[ADD_2:%.*]] = add i32 [[A_2]], 20
31; CHECK-NEXT:    [[XOR_2:%.*]] = xor i32 [[ADD_2]], [[B_2]]
32; CHECK-NEXT:    store i32 [[XOR_2]], i32* [[A_GEP_2]], align 4
33; CHECK-NEXT:    [[IV_3:%.*]] = or i64 [[IV]], 3
34; CHECK-NEXT:    [[B_GEP_3:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[IV_3]]
35; CHECK-NEXT:    [[B_3:%.*]] = load i32, i32* [[B_GEP_3]], align 4
36; CHECK-NEXT:    [[A_GEP_3:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[IV_3]]
37; CHECK-NEXT:    [[A_3:%.*]] = load i32, i32* [[A_GEP_3]], align 4
38; CHECK-NEXT:    [[ADD_3:%.*]] = add i32 [[A_3]], 20
39; CHECK-NEXT:    [[XOR_3:%.*]] = xor i32 [[ADD_3]], [[B_3]]
40; CHECK-NEXT:    store i32 [[XOR_3]], i32* [[A_GEP_3]], align 4
41; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 16
42; CHECK-NEXT:    [[COND:%.*]] = icmp ult i64 [[IV_NEXT]], [[N:%.*]]
43; CHECK-NEXT:    br i1 [[COND]], label [[LOOP]], label [[EXIT:%.*]]
44; CHECK:       exit:
45; CHECK-NEXT:    ret void
46;
47entry:
48  br label %loop
49
50loop:
51  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
52  %B.gep.0 = getelementptr inbounds i32, i32* %B, i64 %iv
53  %B.0 = load i32, i32* %B.gep.0, align 4
54  %A.gep.0 = getelementptr inbounds i32, i32* %A, i64 %iv
55  %A.0 = load i32, i32* %A.gep.0, align 4
56  %add.0 = add i32  %A.0, 20
57  %xor.0 = xor i32 %add.0, %B.0
58  store i32 %xor.0, i32* %A.gep.0, align 4
59  %iv.1 = or i64 %iv, 1
60  %B.gep.1 = getelementptr inbounds i32, i32* %B, i64 %iv.1
61  %B.1 = load i32, i32* %B.gep.1, align 4
62  %A.gep.1 = getelementptr inbounds i32, i32* %A, i64 %iv.1
63  %A.1 = load i32, i32* %A.gep.1, align 4
64  %add.1 = add i32  %A.1, 20
65  %xor.1 = xor i32 %add.1, %B.1
66  store i32 %xor.1, i32* %A.gep.1, align 4
67  %iv.2 = or i64 %iv, 2
68  %B.gep.2 = getelementptr inbounds i32, i32* %B, i64 %iv.2
69  %B.2 = load i32, i32* %B.gep.2, align 4
70  %A.gep.2 = getelementptr inbounds i32, i32* %A, i64 %iv.2
71  %A.2 = load i32, i32* %A.gep.2, align 4
72  %add.2 = add i32  %A.2, 20
73  %xor.2 = xor i32 %add.2, %B.2
74  store i32 %xor.2, i32* %A.gep.2, align 4
75  %iv.3 = or i64 %iv, 3
76  %B.gep.3 = getelementptr inbounds i32, i32* %B, i64 %iv.3
77  %B.3 = load i32, i32* %B.gep.3, align 4
78  %A.gep.3 = getelementptr inbounds i32, i32* %A, i64 %iv.3
79  %A.3 = load i32, i32* %A.gep.3, align 4
80  %add.3 = add i32  %A.3, 20
81  %xor.3 = xor i32 %add.3, %B.3
82  store i32 %xor.3, i32* %A.gep.3, align 4
83  %iv.next = add nuw nsw i64 %iv, 16
84  %cond = icmp ult i64 %iv.next, %N
85  br i1 %cond, label %loop, label %exit
86
87exit:
88  ret void
89}
90
91define void @loop_iv_update_at_start(float* %src, float* %dst) #0 {
92; CHECK-LABEL: @loop_iv_update_at_start(
93; CHECK-NEXT:  entry:
94; CHECK-NEXT:    br label [[LOOP:%.*]]
95; CHECK:       loop:
96; CHECK-NEXT:    [[IV:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
97; CHECK-NEXT:    [[IV_NEXT]] = add i32 [[IV]], 1
98; CHECK-NEXT:    [[COND:%.*]] = icmp ult i32 [[IV]], 2000
99; CHECK-NEXT:    [[SRC_GEP_0:%.*]] = getelementptr inbounds float, float* [[SRC:%.*]], i64 0
100; CHECK-NEXT:    [[SRC_0:%.*]] = load float, float* [[SRC_GEP_0]], align 8
101; CHECK-NEXT:    [[ADD_0:%.*]] = fadd float [[SRC_0]], 1.000000e+00
102; CHECK-NEXT:    [[MUL_0:%.*]] = fmul float [[ADD_0]], [[SRC_0]]
103; CHECK-NEXT:    [[DST_GEP_0:%.*]] = getelementptr inbounds float, float* [[DST:%.*]], i64 0
104; CHECK-NEXT:    store float [[MUL_0]], float* [[DST_GEP_0]], align 8
105; CHECK-NEXT:    [[SRC_GEP_1:%.*]] = getelementptr inbounds float, float* [[SRC]], i64 1
106; CHECK-NEXT:    [[SRC_1:%.*]] = load float, float* [[SRC_GEP_1]], align 8
107; CHECK-NEXT:    [[ADD_1:%.*]] = fadd float [[SRC_1]], 1.000000e+00
108; CHECK-NEXT:    [[MUL_1:%.*]] = fmul float [[ADD_1]], [[SRC_1]]
109; CHECK-NEXT:    [[DST_GEP_1:%.*]] = getelementptr inbounds float, float* [[DST]], i64 1
110; CHECK-NEXT:    store float [[MUL_1]], float* [[DST_GEP_1]], align 8
111; CHECK-NEXT:    [[SRC_GEP_2:%.*]] = getelementptr inbounds float, float* [[SRC]], i64 2
112; CHECK-NEXT:    [[SRC_2:%.*]] = load float, float* [[SRC_GEP_2]], align 8
113; CHECK-NEXT:    [[ADD_2:%.*]] = fadd float [[SRC_2]], 1.000000e+00
114; CHECK-NEXT:    [[MUL_2:%.*]] = fmul float [[ADD_2]], [[SRC_2]]
115; CHECK-NEXT:    [[DST_GEP_2:%.*]] = getelementptr inbounds float, float* [[DST]], i64 2
116; CHECK-NEXT:    store float [[MUL_2]], float* [[DST_GEP_2]], align 8
117; CHECK-NEXT:    [[SRC_GEP_3:%.*]] = getelementptr inbounds float, float* [[SRC]], i64 3
118; CHECK-NEXT:    [[SRC_3:%.*]] = load float, float* [[SRC_GEP_3]], align 8
119; CHECK-NEXT:    [[ADD_3:%.*]] = fadd float [[SRC_3]], 1.000000e+00
120; CHECK-NEXT:    [[MUL_3:%.*]] = fmul float [[ADD_3]], [[SRC_3]]
121; CHECK-NEXT:    [[DST_GEP_3:%.*]] = getelementptr inbounds float, float* [[DST]], i64 3
122; CHECK-NEXT:    store float [[MUL_3]], float* [[DST_GEP_3]], align 8
123; CHECK-NEXT:    [[SRC_GEP_4:%.*]] = getelementptr inbounds float, float* [[SRC]], i64 4
124; CHECK-NEXT:    [[SRC_4:%.*]] = load float, float* [[SRC_GEP_4]], align 8
125; CHECK-NEXT:    [[ADD_4:%.*]] = fadd float [[SRC_4]], 1.000000e+00
126; CHECK-NEXT:    [[MUL_4:%.*]] = fmul float [[ADD_4]], [[SRC_4]]
127; CHECK-NEXT:    [[DST_GEP_4:%.*]] = getelementptr inbounds float, float* [[DST]], i64 4
128; CHECK-NEXT:    store float [[MUL_4]], float* [[DST_GEP_4]], align 8
129; CHECK-NEXT:    br i1 [[COND]], label [[LOOP]], label [[EXIT:%.*]]
130; CHECK:       exit:
131; CHECK-NEXT:    ret void
132;
133entry:
134  br label %loop
135
136loop:
137  %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop ]
138  %iv.next = add i32 %iv, 1
139  %cond = icmp ult i32 %iv, 2000
140
141  %src.gep.0 = getelementptr inbounds float, float* %src, i64 0
142  %src.0 = load float, float* %src.gep.0, align 8
143  %add.0 = fadd float %src.0, 1.0
144  %mul.0 = fmul float %add.0, %src.0
145  %dst.gep.0 = getelementptr inbounds float, float* %dst, i64 0
146  store float %mul.0, float* %dst.gep.0, align 8
147
148  %src.gep.1 = getelementptr inbounds float, float* %src, i64 1
149  %src.1 = load float, float* %src.gep.1, align 8
150  %add.1 = fadd float %src.1, 1.0
151  %mul.1 = fmul float %add.1, %src.1
152  %dst.gep.1 = getelementptr inbounds float, float* %dst, i64 1
153  store float %mul.1, float* %dst.gep.1, align 8
154  %src.gep.2 = getelementptr inbounds float, float* %src, i64 2
155  %src.2 = load float, float* %src.gep.2, align 8
156  %add.2 = fadd float %src.2, 1.0
157  %mul.2 = fmul float %add.2, %src.2
158  %dst.gep.2 = getelementptr inbounds float, float* %dst, i64 2
159  store float %mul.2, float* %dst.gep.2, align 8
160  %src.gep.3 = getelementptr inbounds float, float* %src, i64 3
161  %src.3 = load float, float* %src.gep.3, align 8
162  %add.3 = fadd float %src.3, 1.0
163  %mul.3 = fmul float %add.3, %src.3
164  %dst.gep.3 = getelementptr inbounds float, float* %dst, i64 3
165  store float %mul.3, float* %dst.gep.3, align 8
166  %src.gep.4 = getelementptr inbounds float, float* %src, i64 4
167  %src.4 = load float, float* %src.gep.4, align 8
168  %add.4 = fadd float %src.4, 1.0
169  %mul.4 = fmul float %add.4, %src.4
170  %dst.gep.4 = getelementptr inbounds float, float* %dst, i64 4
171  store float %mul.4, float* %dst.gep.4, align 8
172  br i1 %cond, label %loop, label %exit
173
174exit:
175  ret void
176}
177