1; RUN: opt %loadPolly -polly-scops -analyze < %s | FileCheck %s
2;
3; This checks that the no-wraps checks will be computed fast as some example
4; already showed huge slowdowns even though the inbounds and nsw flags were
5; all in place.
6;
7;    // Inspired by itrans8x8 in transform8x8.c from the ldecode benchmark.
8;    void fast(char *A, char N, char M) {
9;      for (char i = 0; i < 8; i++) {
10;        short index0 = (short)(i + N);
11;        #ifdef fast
12;          short index1 = (index0 *  1) + (short)M;
13;        #else
14;          short index1 = (index0 * 16) + (short)M;
15;        #endif
16;        A[index1]++;
17;      }
18;    }
19;
20; CHECK: Function: fast
21; CHECK: Function: slow
22;
23target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
24
25define void @fast(i8* %A, i8 %N, i8 %M) {
26entry:
27  br label %for.cond
28
29for.cond:                                         ; preds = %for.inc, %entry
30  %indvars.iv = phi i8 [ %indvars.iv.next, %for.inc ], [ 0, %entry ]
31  %exitcond = icmp ne i8 %indvars.iv, 8
32  br i1 %exitcond, label %for.body, label %for.end
33
34for.body:                                         ; preds = %for.cond
35  %tmp3 = add nsw i8 %indvars.iv, %N
36  %tmp3ext = sext i8 %tmp3 to i16
37  ;%mul = mul nsw i16 %tmp3ext, 16
38  %Mext = sext i8 %M to i16
39  %add2 = add nsw i16 %tmp3ext, %Mext
40  %arrayidx = getelementptr inbounds i8, i8* %A, i16 %add2
41  %tmp4 = load i8, i8* %arrayidx, align 4
42  %inc = add nsw i8 %tmp4, 1
43  store i8 %inc, i8* %arrayidx, align 4
44  br label %for.inc
45
46for.inc:                                          ; preds = %for.body
47  %indvars.iv.next = add nuw nsw i8 %indvars.iv, 1
48  br label %for.cond
49
50for.end:                                          ; preds = %for.cond
51  ret void
52}
53
54define void @slow(i8* %A, i8 %N, i8 %M) {
55entry:
56  br label %for.cond
57
58for.cond:                                         ; preds = %for.inc, %entry
59  %indvars.iv = phi i8 [ %indvars.iv.next, %for.inc ], [ 0, %entry ]
60  %exitcond = icmp ne i8 %indvars.iv, 8
61  br i1 %exitcond, label %for.body, label %for.end
62
63for.body:                                         ; preds = %for.cond
64  %tmp3 = add nsw i8 %indvars.iv, %N
65  %tmp3ext = sext i8 %tmp3 to i16
66  %mul = mul nsw i16 %tmp3ext, 16
67  %Mext = sext i8 %M to i16
68  %add2 = add nsw i16 %mul, %Mext
69  %arrayidx = getelementptr inbounds i8, i8* %A, i16 %add2
70  %tmp4 = load i8, i8* %arrayidx, align 4
71  %inc = add nsw i8 %tmp4, 1
72  store i8 %inc, i8* %arrayidx, align 4
73  br label %for.inc
74
75for.inc:                                          ; preds = %for.body
76  %indvars.iv.next = add nuw nsw i8 %indvars.iv, 1
77  br label %for.cond
78
79for.end:                                          ; preds = %for.cond
80  ret void
81}
82