1; RUN: opt -S -march=r600 -mcpu=cayman -loop-vectorize -force-vector-unroll=1 -force-vector-width=4 -dce -instcombine < %s | FileCheck %s
2
3; Artificial datalayout
4target datalayout = "e-p:32:32:32-p1:16:16:16-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-v16:16:16-v24:32:32-v32:32:32-v48:64:64-v64:64:64-v96:128:128-v128:128:128-v192:256:256-v256:256:256-v512:512:512-v1024:1024:1024-v2048:2048:2048-n32:64"
5
6
7define void @add_ints_1_1_1(i32 addrspace(1)* %a, i32 addrspace(1)* %b, i32 addrspace(1)* %c) #0 {
8; CHECK-LABEL: @add_ints_1_1_1(
9; CHECK: <4 x i32>
10; CHECK: ret
11entry:
12  br label %for.cond
13
14for.cond:                                         ; preds = %for.body, %entry
15  %i.0 = phi i64 [ 0, %entry ], [ %inc, %for.body ]
16  %cmp = icmp ult i64 %i.0, 200
17  br i1 %cmp, label %for.body, label %for.end
18
19for.body:                                         ; preds = %for.cond
20  %arrayidx = getelementptr inbounds i32 addrspace(1)* %b, i64 %i.0
21  %0 = load i32 addrspace(1)* %arrayidx, align 4
22  %arrayidx1 = getelementptr inbounds i32 addrspace(1)* %c, i64 %i.0
23  %1 = load i32 addrspace(1)* %arrayidx1, align 4
24  %add = add nsw i32 %0, %1
25  %arrayidx2 = getelementptr inbounds i32 addrspace(1)* %a, i64 %i.0
26  store i32 %add, i32 addrspace(1)* %arrayidx2, align 4
27  %inc = add i64 %i.0, 1
28  br label %for.cond
29
30for.end:                                          ; preds = %for.cond
31  ret void
32}
33
34define void @add_ints_as_1_0_0(i32 addrspace(1)* %a, i32* %b, i32* %c) #0 {
35; CHECK-LABEL: @add_ints_as_1_0_0(
36; CHECK-NOT: <4 x i32>
37; CHECK: ret
38entry:
39  br label %for.cond
40
41for.cond:                                         ; preds = %for.body, %entry
42  %i.0 = phi i64 [ 0, %entry ], [ %inc, %for.body ]
43  %cmp = icmp ult i64 %i.0, 200
44  br i1 %cmp, label %for.body, label %for.end
45
46for.body:                                         ; preds = %for.cond
47  %arrayidx = getelementptr inbounds i32* %b, i64 %i.0
48  %0 = load i32* %arrayidx, align 4
49  %arrayidx1 = getelementptr inbounds i32* %c, i64 %i.0
50  %1 = load i32* %arrayidx1, align 4
51  %add = add nsw i32 %0, %1
52  %arrayidx2 = getelementptr inbounds i32 addrspace(1)* %a, i64 %i.0
53  store i32 %add, i32 addrspace(1)* %arrayidx2, align 4
54  %inc = add i64 %i.0, 1
55  br label %for.cond
56
57for.end:                                          ; preds = %for.cond
58  ret void
59}
60
61define void @add_ints_as_0_1_0(i32* %a, i32 addrspace(1)* %b, i32* %c) #0 {
62; CHECK-LABEL: @add_ints_as_0_1_0(
63; CHECK-NOT: <4 x i32>
64; CHECK: ret
65entry:
66  br label %for.cond
67
68for.cond:                                         ; preds = %for.body, %entry
69  %i.0 = phi i64 [ 0, %entry ], [ %inc, %for.body ]
70  %cmp = icmp ult i64 %i.0, 200
71  br i1 %cmp, label %for.body, label %for.end
72
73for.body:                                         ; preds = %for.cond
74  %arrayidx = getelementptr inbounds i32 addrspace(1)* %b, i64 %i.0
75  %0 = load i32 addrspace(1)* %arrayidx, align 4
76  %arrayidx1 = getelementptr inbounds i32* %c, i64 %i.0
77  %1 = load i32* %arrayidx1, align 4
78  %add = add nsw i32 %0, %1
79  %arrayidx2 = getelementptr inbounds i32* %a, i64 %i.0
80  store i32 %add, i32* %arrayidx2, align 4
81  %inc = add i64 %i.0, 1
82  br label %for.cond
83
84for.end:                                          ; preds = %for.cond
85  ret void
86}
87
88define void @add_ints_as_0_1_1(i32* %a, i32 addrspace(1)* %b, i32 addrspace(1)* %c) #0 {
89; CHECK-LABEL: @add_ints_as_0_1_1(
90; CHECK-NOT: <4 x i32>
91; CHECK: ret
92entry:
93  br label %for.cond
94
95for.cond:                                         ; preds = %for.body, %entry
96  %i.0 = phi i64 [ 0, %entry ], [ %inc, %for.body ]
97  %cmp = icmp ult i64 %i.0, 200
98  br i1 %cmp, label %for.body, label %for.end
99
100for.body:                                         ; preds = %for.cond
101  %arrayidx = getelementptr inbounds i32 addrspace(1)* %b, i64 %i.0
102  %0 = load i32 addrspace(1)* %arrayidx, align 4
103  %arrayidx1 = getelementptr inbounds i32 addrspace(1)* %c, i64 %i.0
104  %1 = load i32 addrspace(1)* %arrayidx1, align 4
105  %add = add nsw i32 %0, %1
106  %arrayidx2 = getelementptr inbounds i32* %a, i64 %i.0
107  store i32 %add, i32* %arrayidx2, align 4
108  %inc = add i64 %i.0, 1
109  br label %for.cond
110
111for.end:                                          ; preds = %for.cond
112  ret void
113}
114
115define void @add_ints_as_0_1_2(i32* %a, i32 addrspace(1)* %b, i32 addrspace(2)* %c) #0 {
116; CHECK-LABEL: @add_ints_as_0_1_2(
117; CHECK-NOT: <4 x i32>
118; CHECK: ret
119entry:
120  br label %for.cond
121
122for.cond:                                         ; preds = %for.body, %entry
123  %i.0 = phi i64 [ 0, %entry ], [ %inc, %for.body ]
124  %cmp = icmp ult i64 %i.0, 200
125  br i1 %cmp, label %for.body, label %for.end
126
127for.body:                                         ; preds = %for.cond
128  %arrayidx = getelementptr inbounds i32 addrspace(1)* %b, i64 %i.0
129  %0 = load i32 addrspace(1)* %arrayidx, align 4
130  %arrayidx1 = getelementptr inbounds i32 addrspace(2)* %c, i64 %i.0
131  %1 = load i32 addrspace(2)* %arrayidx1, align 4
132  %add = add nsw i32 %0, %1
133  %arrayidx2 = getelementptr inbounds i32* %a, i64 %i.0
134  store i32 %add, i32* %arrayidx2, align 4
135  %inc = add i64 %i.0, 1
136  br label %for.cond
137
138for.end:                                          ; preds = %for.cond
139  ret void
140}
141
142attributes #0 = { nounwind uwtable "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
143