• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1; RUN: opt < %s  -loop-vectorize -force-vector-unroll=1 -force-vector-width=4 -dce -instcombine -S | 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@a = common global [2048 x i32] zeroinitializer, align 16
7
8; This is the loop.
9;  for (i=0; i<n; i++){
10;    a[i] += i;
11;  }
12;CHECK: @inc
13;CHECK: load <4 x i32>
14;CHECK: add nsw <4 x i32>
15;CHECK: store <4 x i32>
16;CHECK: ret void
17define void @inc(i32 %n) nounwind uwtable noinline ssp {
18  %1 = icmp sgt i32 %n, 0
19  br i1 %1, label %.lr.ph, label %._crit_edge
20
21.lr.ph:                                           ; preds = %0, %.lr.ph
22  %indvars.iv = phi i64 [ %indvars.iv.next, %.lr.ph ], [ 0, %0 ]
23  %2 = getelementptr inbounds [2048 x i32]* @a, i64 0, i64 %indvars.iv
24  %3 = load i32* %2, align 4
25  %4 = trunc i64 %indvars.iv to i32
26  %5 = add nsw i32 %3, %4
27  store i32 %5, i32* %2, align 4
28  %indvars.iv.next = add i64 %indvars.iv, 1
29  %lftr.wideiv = trunc i64 %indvars.iv.next to i32
30  %exitcond = icmp eq i32 %lftr.wideiv, %n
31  br i1 %exitcond, label %._crit_edge, label %.lr.ph
32
33._crit_edge:                                      ; preds = %.lr.ph, %0
34  ret void
35}
36
37; Can't vectorize this loop because the access to A[X] is non linear.
38;
39;  for (i = 0; i < n; ++i) {
40;    A[B[i]]++;
41;
42;CHECK: @histogram
43;CHECK-NOT: <4 x i32>
44;CHECK: ret i32
45define i32 @histogram(i32* nocapture noalias %A, i32* nocapture noalias %B, i32 %n) nounwind uwtable ssp {
46entry:
47  %cmp6 = icmp sgt i32 %n, 0
48  br i1 %cmp6, label %for.body, label %for.end
49
50for.body:                                         ; preds = %entry, %for.body
51  %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ]
52  %arrayidx = getelementptr inbounds i32* %B, i64 %indvars.iv
53  %0 = load i32* %arrayidx, align 4
54  %idxprom1 = sext i32 %0 to i64
55  %arrayidx2 = getelementptr inbounds i32* %A, i64 %idxprom1
56  %1 = load i32* %arrayidx2, align 4
57  %inc = add nsw i32 %1, 1
58  store i32 %inc, i32* %arrayidx2, align 4
59  %indvars.iv.next = add i64 %indvars.iv, 1
60  %lftr.wideiv = trunc i64 %indvars.iv.next to i32
61  %exitcond = icmp eq i32 %lftr.wideiv, %n
62  br i1 %exitcond, label %for.end, label %for.body
63
64for.end:                                          ; preds = %for.body, %entry
65  ret i32 0
66}
67