1*5cce4affSRoman Lebedev; RUN: opt < %s -loop-vectorize -force-vector-interleave=1 -force-vector-width=4 -dce -instcombine -S -simplifycfg -simplifycfg-require-and-preserve-domtree=1 | FileCheck %s
2cee313d2SEric Christopher
3cee313d2SEric Christophertarget 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"
4cee313d2SEric Christopher
5cee313d2SEric Christopher;
6cee313d2SEric Christopher; We want to make sure that we are vectorizeing the scalar loop only once
7cee313d2SEric Christopher; even if the pass manager runs the vectorizer multiple times due to inlining.
8cee313d2SEric Christopher
9cee313d2SEric Christopher
10cee313d2SEric Christopher; This test checks that we add metadata to vectorized loops
11cee313d2SEric Christopher; CHECK-LABEL: @_Z4foo1Pii(
12cee313d2SEric Christopher; CHECK: <4 x i32>
13cee313d2SEric Christopher; CHECK: llvm.loop
14cee313d2SEric Christopher; CHECK: ret
15cee313d2SEric Christopher
16cee313d2SEric Christopher; This test comes from the loop:
17cee313d2SEric Christopher;
18cee313d2SEric Christopher;int foo (int *A, int n) {
19cee313d2SEric Christopher;  return std::accumulate(A, A + n, 0);
20cee313d2SEric Christopher;}
21cee313d2SEric Christopherdefine i32 @_Z4foo1Pii(i32* %A, i32 %n) #0 {
22cee313d2SEric Christopherentry:
23cee313d2SEric Christopher  %idx.ext = sext i32 %n to i64
24cee313d2SEric Christopher  %add.ptr = getelementptr inbounds i32, i32* %A, i64 %idx.ext
25cee313d2SEric Christopher  %cmp3.i = icmp eq i32 %n, 0
26cee313d2SEric Christopher  br i1 %cmp3.i, label %_ZSt10accumulateIPiiET0_T_S2_S1_.exit, label %for.body.i
27cee313d2SEric Christopher
28cee313d2SEric Christopherfor.body.i:                                       ; preds = %entry, %for.body.i
29cee313d2SEric Christopher  %__init.addr.05.i = phi i32 [ %add.i, %for.body.i ], [ 0, %entry ]
30cee313d2SEric Christopher  %__first.addr.04.i = phi i32* [ %incdec.ptr.i, %for.body.i ], [ %A, %entry ]
31cee313d2SEric Christopher  %0 = load i32, i32* %__first.addr.04.i, align 4
32cee313d2SEric Christopher  %add.i = add nsw i32 %0, %__init.addr.05.i
33cee313d2SEric Christopher  %incdec.ptr.i = getelementptr inbounds i32, i32* %__first.addr.04.i, i64 1
34cee313d2SEric Christopher  %cmp.i = icmp eq i32* %incdec.ptr.i, %add.ptr
35cee313d2SEric Christopher  br i1 %cmp.i, label %_ZSt10accumulateIPiiET0_T_S2_S1_.exit, label %for.body.i
36cee313d2SEric Christopher
37cee313d2SEric Christopher_ZSt10accumulateIPiiET0_T_S2_S1_.exit:            ; preds = %for.body.i, %entry
38cee313d2SEric Christopher  %__init.addr.0.lcssa.i = phi i32 [ 0, %entry ], [ %add.i, %for.body.i ]
39cee313d2SEric Christopher  ret i32 %__init.addr.0.lcssa.i
40cee313d2SEric Christopher}
41cee313d2SEric Christopher
42cee313d2SEric Christopher; This test checks that we don't vectorize loops that are marked with the "width" == 1 metadata.
43cee313d2SEric Christopher; CHECK-LABEL: @_Z4foo2Pii(
44cee313d2SEric Christopher; CHECK-NOT: <4 x i32>
45cee313d2SEric Christopher; CHECK: llvm.loop
46cee313d2SEric Christopher; CHECK: ret
47cee313d2SEric Christopherdefine i32 @_Z4foo2Pii(i32* %A, i32 %n) #0 {
48cee313d2SEric Christopherentry:
49cee313d2SEric Christopher  %idx.ext = sext i32 %n to i64
50cee313d2SEric Christopher  %add.ptr = getelementptr inbounds i32, i32* %A, i64 %idx.ext
51cee313d2SEric Christopher  %cmp3.i = icmp eq i32 %n, 0
52cee313d2SEric Christopher  br i1 %cmp3.i, label %_ZSt10accumulateIPiiET0_T_S2_S1_.exit, label %for.body.i
53cee313d2SEric Christopher
54cee313d2SEric Christopherfor.body.i:                                       ; preds = %entry, %for.body.i
55cee313d2SEric Christopher  %__init.addr.05.i = phi i32 [ %add.i, %for.body.i ], [ 0, %entry ]
56cee313d2SEric Christopher  %__first.addr.04.i = phi i32* [ %incdec.ptr.i, %for.body.i ], [ %A, %entry ]
57cee313d2SEric Christopher  %0 = load i32, i32* %__first.addr.04.i, align 4
58cee313d2SEric Christopher  %add.i = add nsw i32 %0, %__init.addr.05.i
59cee313d2SEric Christopher  %incdec.ptr.i = getelementptr inbounds i32, i32* %__first.addr.04.i, i64 1
60cee313d2SEric Christopher  %cmp.i = icmp eq i32* %incdec.ptr.i, %add.ptr
61cee313d2SEric Christopher  br i1 %cmp.i, label %_ZSt10accumulateIPiiET0_T_S2_S1_.exit, label %for.body.i, !llvm.loop !0
62cee313d2SEric Christopher
63cee313d2SEric Christopher_ZSt10accumulateIPiiET0_T_S2_S1_.exit:            ; preds = %for.body.i, %entry
64cee313d2SEric Christopher  %__init.addr.0.lcssa.i = phi i32 [ 0, %entry ], [ %add.i, %for.body.i ]
65cee313d2SEric Christopher  ret i32 %__init.addr.0.lcssa.i
66cee313d2SEric Christopher}
67cee313d2SEric Christopher
68a36ddf0aSFangrui Songattributes #0 = { nounwind readonly ssp uwtable "fp-contract-model"="standard" "frame-pointer"="non-leaf" "realign-stack" "relocation-model"="pic" "ssp-buffers-size"="8" }
69cee313d2SEric Christopher
70cee313d2SEric Christopher; CHECK: !0 = distinct !{!0, !1}
71cee313d2SEric Christopher; CHECK: !1 = !{!"llvm.loop.isvectorized", i32 1}
72cee313d2SEric Christopher; CHECK: !2 = distinct !{!2, !3, !1}
73cee313d2SEric Christopher; CHECK: !3 = !{!"llvm.loop.unroll.runtime.disable"}
74cee313d2SEric Christopher
75cee313d2SEric Christopher!0 = !{!0, !1}
76cee313d2SEric Christopher!1 = !{!"llvm.loop.vectorize.width", i32 1}
77