1; RUN: opt < %s -loop-vectorize -force-vector-interleave=1 -force-vector-width=4 -dce -instcombine -S -simplifycfg | FileCheck %s
2
3target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
4target triple = "x86_64-apple-macosx10.8.0"
5
6;
7; We want to make sure that we are vectorizeing the scalar loop only once
8; even if the pass manager runs the vectorizer multiple times due to inlining.
9
10
11; This test checks that we add metadata to vectorized loops
12; CHECK-LABEL: @_Z4foo1Pii(
13; CHECK: <4 x i32>
14; CHECK: llvm.loop
15; CHECK: ret
16
17; This test comes from the loop:
18;
19;int foo (int *A, int n) {
20;  return std::accumulate(A, A + n, 0);
21;}
22define i32 @_Z4foo1Pii(i32* %A, i32 %n) #0 {
23entry:
24  %idx.ext = sext i32 %n to i64
25  %add.ptr = getelementptr inbounds i32, i32* %A, i64 %idx.ext
26  %cmp3.i = icmp eq i32 %n, 0
27  br i1 %cmp3.i, label %_ZSt10accumulateIPiiET0_T_S2_S1_.exit, label %for.body.i
28
29for.body.i:                                       ; preds = %entry, %for.body.i
30  %__init.addr.05.i = phi i32 [ %add.i, %for.body.i ], [ 0, %entry ]
31  %__first.addr.04.i = phi i32* [ %incdec.ptr.i, %for.body.i ], [ %A, %entry ]
32  %0 = load i32, i32* %__first.addr.04.i, align 4
33  %add.i = add nsw i32 %0, %__init.addr.05.i
34  %incdec.ptr.i = getelementptr inbounds i32, i32* %__first.addr.04.i, i64 1
35  %cmp.i = icmp eq i32* %incdec.ptr.i, %add.ptr
36  br i1 %cmp.i, label %_ZSt10accumulateIPiiET0_T_S2_S1_.exit, label %for.body.i
37
38_ZSt10accumulateIPiiET0_T_S2_S1_.exit:            ; preds = %for.body.i, %entry
39  %__init.addr.0.lcssa.i = phi i32 [ 0, %entry ], [ %add.i, %for.body.i ]
40  ret i32 %__init.addr.0.lcssa.i
41}
42
43; This test checks that we don't vectorize loops that are marked with the "width" == 1 metadata.
44; CHECK-LABEL: @_Z4foo2Pii(
45; CHECK-NOT: <4 x i32>
46; CHECK: llvm.loop
47; CHECK: ret
48define i32 @_Z4foo2Pii(i32* %A, i32 %n) #0 {
49entry:
50  %idx.ext = sext i32 %n to i64
51  %add.ptr = getelementptr inbounds i32, i32* %A, i64 %idx.ext
52  %cmp3.i = icmp eq i32 %n, 0
53  br i1 %cmp3.i, label %_ZSt10accumulateIPiiET0_T_S2_S1_.exit, label %for.body.i
54
55for.body.i:                                       ; preds = %entry, %for.body.i
56  %__init.addr.05.i = phi i32 [ %add.i, %for.body.i ], [ 0, %entry ]
57  %__first.addr.04.i = phi i32* [ %incdec.ptr.i, %for.body.i ], [ %A, %entry ]
58  %0 = load i32, i32* %__first.addr.04.i, align 4
59  %add.i = add nsw i32 %0, %__init.addr.05.i
60  %incdec.ptr.i = getelementptr inbounds i32, i32* %__first.addr.04.i, i64 1
61  %cmp.i = icmp eq i32* %incdec.ptr.i, %add.ptr
62  br i1 %cmp.i, label %_ZSt10accumulateIPiiET0_T_S2_S1_.exit, label %for.body.i, !llvm.loop !0
63
64_ZSt10accumulateIPiiET0_T_S2_S1_.exit:            ; preds = %for.body.i, %entry
65  %__init.addr.0.lcssa.i = phi i32 [ 0, %entry ], [ %add.i, %for.body.i ]
66  ret i32 %__init.addr.0.lcssa.i
67}
68
69attributes #0 = { nounwind readonly ssp uwtable "fp-contract-model"="standard" "no-frame-pointer-elim" "no-frame-pointer-elim-non-leaf" "realign-stack" "relocation-model"="pic" "ssp-buffers-size"="8" }
70
71; CHECK: !0 = distinct !{!0, !1, !2}
72; CHECK: !1 = !{!"llvm.loop.vectorize.width", i32 1}
73; CHECK: !2 = !{!"llvm.loop.interleave.count", i32 1}
74; CHECK: !3 = distinct !{!3, !4, !1, !2}
75; CHECK: !4 = !{!"llvm.loop.unroll.runtime.disable"}
76
77!0 = !{!0, !1}
78!1 = !{!"llvm.loop.vectorize.width", i32 1}
79