1 ; 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
3 target 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"
6 ; We want to make sure that we are vectorizeing the scalar loop only once
7 ; even if the pass manager runs the vectorizer multiple times due to inlining.
10 ; This test checks that we add metadata to vectorized loops
11 ; CHECK-LABEL: @_Z4foo1Pii(
16 ; This test comes from the loop:
18 ;int foo (int *A, int n) {
19 ; return std::accumulate(A, A + n, 0);
21 define i32 @_Z4foo1Pii(i32* %A, i32 %n) #0 {
23 %idx.ext = sext i32 %n to i64
24 %add.ptr = getelementptr inbounds i32, i32* %A, i64 %idx.ext
25 %cmp3.i = icmp eq i32 %n, 0
26 br i1 %cmp3.i, label %_ZSt10accumulateIPiiET0_T_S2_S1_.exit, label %for.body.i
28 for.body.i: ; preds = %entry, %for.body.i
29 %__init.addr.05.i = phi i32 [ %add.i, %for.body.i ], [ 0, %entry ]
30 %__first.addr.04.i = phi i32* [ %incdec.ptr.i, %for.body.i ], [ %A, %entry ]
31 %0 = load i32, i32* %__first.addr.04.i, align 4
32 %add.i = add nsw i32 %0, %__init.addr.05.i
33 %incdec.ptr.i = getelementptr inbounds i32, i32* %__first.addr.04.i, i64 1
34 %cmp.i = icmp eq i32* %incdec.ptr.i, %add.ptr
35 br i1 %cmp.i, label %_ZSt10accumulateIPiiET0_T_S2_S1_.exit, label %for.body.i
37 _ZSt10accumulateIPiiET0_T_S2_S1_.exit: ; preds = %for.body.i, %entry
38 %__init.addr.0.lcssa.i = phi i32 [ 0, %entry ], [ %add.i, %for.body.i ]
39 ret i32 %__init.addr.0.lcssa.i
42 ; This test checks that we don't vectorize loops that are marked with the "width" == 1 metadata.
43 ; CHECK-LABEL: @_Z4foo2Pii(
44 ; CHECK-NOT: <4 x i32>
47 define i32 @_Z4foo2Pii(i32* %A, i32 %n) #0 {
49 %idx.ext = sext i32 %n to i64
50 %add.ptr = getelementptr inbounds i32, i32* %A, i64 %idx.ext
51 %cmp3.i = icmp eq i32 %n, 0
52 br i1 %cmp3.i, label %_ZSt10accumulateIPiiET0_T_S2_S1_.exit, label %for.body.i
54 for.body.i: ; preds = %entry, %for.body.i
55 %__init.addr.05.i = phi i32 [ %add.i, %for.body.i ], [ 0, %entry ]
56 %__first.addr.04.i = phi i32* [ %incdec.ptr.i, %for.body.i ], [ %A, %entry ]
57 %0 = load i32, i32* %__first.addr.04.i, align 4
58 %add.i = add nsw i32 %0, %__init.addr.05.i
59 %incdec.ptr.i = getelementptr inbounds i32, i32* %__first.addr.04.i, i64 1
60 %cmp.i = icmp eq i32* %incdec.ptr.i, %add.ptr
61 br i1 %cmp.i, label %_ZSt10accumulateIPiiET0_T_S2_S1_.exit, label %for.body.i, !llvm.loop !0
63 _ZSt10accumulateIPiiET0_T_S2_S1_.exit: ; preds = %for.body.i, %entry
64 %__init.addr.0.lcssa.i = phi i32 [ 0, %entry ], [ %add.i, %for.body.i ]
65 ret i32 %__init.addr.0.lcssa.i
68 attributes #0 = { nounwind readonly ssp uwtable "fp-contract-model"="standard" "frame-pointer"="non-leaf" "realign-stack" "relocation-model"="pic" "ssp-buffers-size"="8" }
70 ; CHECK: !0 = distinct !{!0, !1}
71 ; CHECK: !1 = !{!"llvm.loop.isvectorized", i32 1}
72 ; CHECK: !2 = distinct !{!2, !3, !1}
73 ; CHECK: !3 = !{!"llvm.loop.unroll.runtime.disable"}
76 !1 = !{!"llvm.loop.vectorize.width", i32 1}