1; REQUIRES: asserts
2; RUN: opt < %s -loop-vectorize -force-vector-width=2 -force-vector-interleave=1 -instcombine -debug-only=loop-vectorize -disable-output -print-after=instcombine -enable-new-pm=0 2>&1 | FileCheck %s
3; RUN: opt < %s -passes='function(loop-vectorize,instcombine)' -force-vector-width=2 -force-vector-interleave=1 -debug-only=loop-vectorize -disable-output -print-after=instcombine 2>&1 | FileCheck %s
4
5target datalayout = "e-m:e-i64:64-i128:128-n32:64-S128"
6
7; CHECK-LABEL: phi_two_incoming_values
8; CHECK:       LV: Found an estimated cost of 1 for VF 2 For instruction: %i = phi i64 [ %i.next, %if.end ], [ 0, %entry ]
9; CHECK:       LV: Found an estimated cost of 1 for VF 2 For instruction: %tmp5 = phi i32 [ %tmp1, %for.body ], [ %tmp4, %if.then ]
10; CHECK:       define void @phi_two_incoming_values(
11; CHECK:       vector.body:
12; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %vector.ph ], [ [[INDEX_NEXT:%.*]], %vector.body ]
13; CHECK:         [[WIDE_LOAD:%.*]] = load <2 x i32>, <2 x i32>* {{.*}}
14; CHECK:         [[TMP5:%.*]] = icmp sgt <2 x i32> [[WIDE_LOAD]], zeroinitializer
15; CHECK-NEXT:    [[TMP6:%.*]] = zext <2 x i1> [[TMP5]] to <2 x i32>
16; CHECK-NEXT:    [[PREDPHI:%.*]] = add <2 x i32> [[WIDE_LOAD]], [[TMP6]]
17; CHECK:         store <2 x i32> [[PREDPHI]], <2 x i32>* {{.*}}
18; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
19;
20define void @phi_two_incoming_values(i32* %a, i32* %b, i64 %n) {
21entry:
22  br label %for.body
23
24for.body:
25  %i = phi i64 [ %i.next, %if.end ], [ 0, %entry ]
26  %tmp0 = getelementptr inbounds i32, i32* %a, i64 %i
27  %tmp1 = load i32, i32* %tmp0, align 4
28  %tmp2 = getelementptr inbounds i32, i32* %b, i64 %i
29  %tmp3 = icmp sgt i32 %tmp1, 0
30  br i1 %tmp3, label %if.then, label %if.end
31
32if.then:
33  %tmp4 = add i32 %tmp1, 1
34  br label %if.end
35
36if.end:
37  %tmp5 = phi i32 [ %tmp1, %for.body ], [ %tmp4, %if.then ]
38  store i32 %tmp5, i32* %tmp2, align 4
39  %i.next = add i64 %i, 1
40  %cond = icmp eq i64 %i, %n
41  br i1 %cond, label %for.end, label %for.body
42
43for.end:
44  ret void
45}
46
47; CHECK-LABEL: phi_three_incoming_values
48; CHECK:       LV: Found an estimated cost of 1 for VF 2 For instruction: %i = phi i64 [ %i.next, %if.end ], [ 0, %entry ]
49; CHECK:       LV: Found an estimated cost of 2 for VF 2 For instruction: %tmp8 = phi i32 [ 9, %for.body ], [ 3, %if.then ], [ %tmp7, %if.else ]
50; CHECK:       define void @phi_three_incoming_values(
51; CHECK:       vector.body:
52; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %vector.ph ], [ [[INDEX_NEXT:%.*]], %vector.body ]
53; CHECK:         [[PREDPHI:%.*]] = select <2 x i1> {{.*}}, <2 x i32> <i32 3, i32 3>, <2 x i32> <i32 9, i32 9>
54; CHECK:         [[PREDPHI7:%.*]] = select <2 x i1> {{.*}}, <2 x i32> {{.*}}, <2 x i32> [[PREDPHI]]
55; CHECK:         store <2 x i32> [[PREDPHI7]], <2 x i32>* {{.*}}
56; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
57;
58define void @phi_three_incoming_values(i32* %a, i32* %b, i64 %n) {
59entry:
60  br label %for.body
61
62for.body:
63  %i = phi i64 [ %i.next, %if.end ], [ 0, %entry ]
64  %tmp0 = getelementptr inbounds i32, i32* %a, i64 %i
65  %tmp1 = load i32, i32* %tmp0, align 4
66  %tmp2 = getelementptr inbounds i32, i32* %b, i64 %i
67  %tmp3 = load i32, i32* %tmp2, align 4
68  %tmp4 = icmp sgt i32 %tmp1, %tmp3
69  br i1 %tmp4, label %if.then, label %if.end
70
71if.then:
72  %tmp5 = icmp sgt i32 %tmp1, 19
73  br i1 %tmp5, label %if.end, label %if.else
74
75if.else:
76  %tmp6 = icmp slt i32 %tmp3, 4
77  %tmp7 = select i1 %tmp6, i32 4, i32 5
78  br label %if.end
79
80if.end:
81  %tmp8 = phi i32 [ 9, %for.body ], [ 3, %if.then ], [ %tmp7, %if.else ]
82  store i32 %tmp8, i32* %tmp0, align 4
83  %i.next = add i64 %i, 1
84  %cond = icmp eq i64 %i, %n
85  br i1 %cond, label %for.end, label %for.body
86
87for.end:
88  ret void
89}
90