1 ; RUN: opt -slp-vectorizer -slp-threshold=-6 -slp-vectorize-hor -S < %s | FileCheck %s
3 ; FIXME: The threshold is changed to keep this test case a bit smaller.
4 ; The AArch64 cost model should not give such high costs to select statements.
6 target datalayout = "e-m:e-i64:64-i128:128-n32:64-S128"
7 target triple = "aarch64--linux"
9 ; CHECK-LABEL: test_select
10 ; CHECK: load <4 x i32>
11 ; CHECK: load <4 x i32>
12 ; CHECK: select <4 x i1>
13 define i32 @test_select(i32* noalias nocapture readonly %blk1, i32* noalias nocapture readonly %blk2, i32 %lx, i32 %h) {
15 %cmp.22 = icmp sgt i32 %h, 0
16 br i1 %cmp.22, label %for.body.lr.ph, label %for.end
18 for.body.lr.ph: ; preds = %entry
19 %idx.ext = sext i32 %lx to i64
22 for.body: ; preds = %for.body, %for.body.lr.ph
23 %s.026 = phi i32 [ 0, %for.body.lr.ph ], [ %add27, %for.body ]
24 %j.025 = phi i32 [ 0, %for.body.lr.ph ], [ %inc, %for.body ]
25 %p2.024 = phi i32* [ %blk2, %for.body.lr.ph ], [ %add.ptr29, %for.body ]
26 %p1.023 = phi i32* [ %blk1, %for.body.lr.ph ], [ %add.ptr, %for.body ]
27 %0 = load i32, i32* %p1.023, align 4
28 %1 = load i32, i32* %p2.024, align 4
29 %sub = sub nsw i32 %0, %1
30 %cmp2 = icmp slt i32 %sub, 0
31 %sub3 = sub nsw i32 0, %sub
32 %sub3.sub = select i1 %cmp2, i32 %sub3, i32 %sub
33 %add = add nsw i32 %sub3.sub, %s.026
34 %arrayidx4 = getelementptr inbounds i32, i32* %p1.023, i64 1
35 %2 = load i32, i32* %arrayidx4, align 4
36 %arrayidx5 = getelementptr inbounds i32, i32* %p2.024, i64 1
37 %3 = load i32, i32* %arrayidx5, align 4
38 %sub6 = sub nsw i32 %2, %3
39 %cmp7 = icmp slt i32 %sub6, 0
40 %sub9 = sub nsw i32 0, %sub6
41 %v.1 = select i1 %cmp7, i32 %sub9, i32 %sub6
42 %add11 = add nsw i32 %add, %v.1
43 %arrayidx12 = getelementptr inbounds i32, i32* %p1.023, i64 2
44 %4 = load i32, i32* %arrayidx12, align 4
45 %arrayidx13 = getelementptr inbounds i32, i32* %p2.024, i64 2
46 %5 = load i32, i32* %arrayidx13, align 4
47 %sub14 = sub nsw i32 %4, %5
48 %cmp15 = icmp slt i32 %sub14, 0
49 %sub17 = sub nsw i32 0, %sub14
50 %sub17.sub14 = select i1 %cmp15, i32 %sub17, i32 %sub14
51 %add19 = add nsw i32 %add11, %sub17.sub14
52 %arrayidx20 = getelementptr inbounds i32, i32* %p1.023, i64 3
53 %6 = load i32, i32* %arrayidx20, align 4
54 %arrayidx21 = getelementptr inbounds i32, i32* %p2.024, i64 3
55 %7 = load i32, i32* %arrayidx21, align 4
56 %sub22 = sub nsw i32 %6, %7
57 %cmp23 = icmp slt i32 %sub22, 0
58 %sub25 = sub nsw i32 0, %sub22
59 %v.3 = select i1 %cmp23, i32 %sub25, i32 %sub22
60 %add27 = add nsw i32 %add19, %v.3
61 %add.ptr = getelementptr inbounds i32, i32* %p1.023, i64 %idx.ext
62 %add.ptr29 = getelementptr inbounds i32, i32* %p2.024, i64 %idx.ext
63 %inc = add nuw nsw i32 %j.025, 1
64 %exitcond = icmp eq i32 %inc, %h
65 br i1 %exitcond, label %for.end.loopexit, label %for.body
67 for.end.loopexit: ; preds = %for.body
70 for.end: ; preds = %for.end.loopexit, %entry
71 %s.0.lcssa = phi i32 [ 0, %entry ], [ %add27, %for.end.loopexit ]