1; RUN: opt < %s -S -loop-unroll -unroll-runtime=true | FileCheck %s -check-prefix=EPILOG
2; RUN: opt < %s -S -loop-unroll -unroll-runtime=true -unroll-runtime-epilog=false | FileCheck %s -check-prefix=PROLOG
3
4target 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"
5
6; Tests for unrolling loops with run-time trip counts
7
8; EPILOG: %xtraiter = and i32 %n
9; EPILOG:  %lcmp.mod = icmp ne i32 %xtraiter, 0
10; EPILOG:  br i1 %lcmp.mod, label %for.body.epil.preheader, label %for.end.loopexit
11
12; PROLOG: %xtraiter = and i32 %n
13; PROLOG:  %lcmp.mod = icmp ne i32 %xtraiter, 0
14; PROLOG:  br i1 %lcmp.mod, label %for.body.prol.preheader, label %for.body.prol.loopexit
15
16; EPILOG: for.body.epil:
17; EPILOG: %indvars.iv.epil = phi i64 [ %indvars.iv.next.epil, %for.body.epil ],  [ %indvars.iv.unr, %for.body.epil.preheader ]
18; EPILOG:  %epil.iter.sub = sub i32 %epil.iter, 1
19; EPILOG:  %epil.iter.cmp = icmp ne i32 %epil.iter.sub, 0
20; EPILOG:  br i1 %epil.iter.cmp, label %for.body.epil, label %for.end.loopexit.epilog-lcssa, !llvm.loop !0
21
22; PROLOG: for.body.prol:
23; PROLOG: %indvars.iv.prol = phi i64 [ %indvars.iv.next.prol, %for.body.prol ], [ 0, %for.body.prol.preheader ]
24; PROLOG:  %prol.iter.sub = sub i32 %prol.iter, 1
25; PROLOG:  %prol.iter.cmp = icmp ne i32 %prol.iter.sub, 0
26; PROLOG:  br i1 %prol.iter.cmp, label %for.body.prol, label %for.body.prol.loopexit, !llvm.loop !0
27
28
29define i32 @test(i32* nocapture %a, i32 %n) nounwind uwtable readonly {
30entry:
31  %cmp1 = icmp eq i32 %n, 0
32  br i1 %cmp1, label %for.end, label %for.body
33
34for.body:                                         ; preds = %for.body, %entry
35  %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ]
36  %sum.02 = phi i32 [ %add, %for.body ], [ 0, %entry ]
37  %arrayidx = getelementptr inbounds i32, i32* %a, i64 %indvars.iv
38  %0 = load i32, i32* %arrayidx, align 4
39  %add = add nsw i32 %0, %sum.02
40  %indvars.iv.next = add i64 %indvars.iv, 1
41  %lftr.wideiv = trunc i64 %indvars.iv.next to i32
42  %exitcond = icmp eq i32 %lftr.wideiv, %n
43  br i1 %exitcond, label %for.end, label %for.body
44
45for.end:                                          ; preds = %for.body, %entry
46  %sum.0.lcssa = phi i32 [ 0, %entry ], [ %add, %for.body ]
47  ret i32 %sum.0.lcssa
48}
49
50
51; Still try to completely unroll loops with compile-time trip counts
52; even if the -unroll-runtime is specified
53
54; EPILOG: for.body:
55; EPILOG-NOT: for.body.epil:
56
57; PROLOG: for.body:
58; PROLOG-NOT: for.body.prol:
59
60define i32 @test1(i32* nocapture %a) nounwind uwtable readonly {
61entry:
62  br label %for.body
63
64for.body:                                         ; preds = %for.body, %entry
65  %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
66  %sum.01 = phi i32 [ 0, %entry ], [ %add, %for.body ]
67  %arrayidx = getelementptr inbounds i32, i32* %a, i64 %indvars.iv
68  %0 = load i32, i32* %arrayidx, align 4
69  %add = add nsw i32 %0, %sum.01
70  %indvars.iv.next = add i64 %indvars.iv, 1
71  %lftr.wideiv = trunc i64 %indvars.iv.next to i32
72  %exitcond = icmp eq i32 %lftr.wideiv, 5
73  br i1 %exitcond, label %for.end, label %for.body
74
75for.end:                                          ; preds = %for.body
76  ret i32 %add
77}
78
79; This is test 2007-05-09-UnknownTripCount.ll which can be unrolled now
80; if the -unroll-runtime option is turned on
81
82; EPILOG: bb72.2:
83; PROLOG: bb72.2:
84
85define void @foo(i32 %trips) {
86entry:
87        br label %cond_true.outer
88
89cond_true.outer:
90        %indvar1.ph = phi i32 [ 0, %entry ], [ %indvar.next2, %bb72 ]
91        br label %bb72
92
93bb72:
94        %indvar.next2 = add i32 %indvar1.ph, 1
95        %exitcond3 = icmp eq i32 %indvar.next2, %trips
96        br i1 %exitcond3, label %cond_true138, label %cond_true.outer
97
98cond_true138:
99        ret void
100}
101
102
103; Test run-time unrolling for a loop that counts down by -2.
104
105; EPILOG: for.body.epil:
106; EPILOG: br i1 %epil.iter.cmp, label %for.body.epil, label %for.cond.for.end_crit_edge.epilog-lcssa
107
108; PROLOG: for.body.prol:
109; PROLOG: br i1 %prol.iter.cmp, label %for.body.prol, label %for.body.prol.loopexit
110
111define zeroext i16 @down(i16* nocapture %p, i32 %len) nounwind uwtable readonly {
112entry:
113  %cmp2 = icmp eq i32 %len, 0
114  br i1 %cmp2, label %for.end, label %for.body
115
116for.body:                                         ; preds = %for.body, %entry
117  %p.addr.05 = phi i16* [ %incdec.ptr, %for.body ], [ %p, %entry ]
118  %len.addr.04 = phi i32 [ %sub, %for.body ], [ %len, %entry ]
119  %res.03 = phi i32 [ %add, %for.body ], [ 0, %entry ]
120  %incdec.ptr = getelementptr inbounds i16, i16* %p.addr.05, i64 1
121  %0 = load i16, i16* %p.addr.05, align 2
122  %conv = zext i16 %0 to i32
123  %add = add i32 %conv, %res.03
124  %sub = add nsw i32 %len.addr.04, -2
125  %cmp = icmp eq i32 %sub, 0
126  br i1 %cmp, label %for.cond.for.end_crit_edge, label %for.body
127
128for.cond.for.end_crit_edge:                       ; preds = %for.body
129  %phitmp = trunc i32 %add to i16
130  br label %for.end
131
132for.end:                                          ; preds = %for.cond.for.end_crit_edge, %entry
133  %res.0.lcssa = phi i16 [ %phitmp, %for.cond.for.end_crit_edge ], [ 0, %entry ]
134  ret i16 %res.0.lcssa
135}
136
137; Test run-time unrolling disable metadata.
138; EPILOG: for.body:
139; EPILOG-NOT: for.body.epil:
140
141; PROLOG: for.body:
142; PROLOG-NOT: for.body.prol:
143
144define zeroext i16 @test2(i16* nocapture %p, i32 %len) nounwind uwtable readonly {
145entry:
146  %cmp2 = icmp eq i32 %len, 0
147  br i1 %cmp2, label %for.end, label %for.body
148
149for.body:                                         ; preds = %for.body, %entry
150  %p.addr.05 = phi i16* [ %incdec.ptr, %for.body ], [ %p, %entry ]
151  %len.addr.04 = phi i32 [ %sub, %for.body ], [ %len, %entry ]
152  %res.03 = phi i32 [ %add, %for.body ], [ 0, %entry ]
153  %incdec.ptr = getelementptr inbounds i16, i16* %p.addr.05, i64 1
154  %0 = load i16, i16* %p.addr.05, align 2
155  %conv = zext i16 %0 to i32
156  %add = add i32 %conv, %res.03
157  %sub = add nsw i32 %len.addr.04, -2
158  %cmp = icmp eq i32 %sub, 0
159  br i1 %cmp, label %for.cond.for.end_crit_edge, label %for.body, !llvm.loop !0
160
161for.cond.for.end_crit_edge:                       ; preds = %for.body
162  %phitmp = trunc i32 %add to i16
163  br label %for.end
164
165for.end:                                          ; preds = %for.cond.for.end_crit_edge, %entry
166  %res.0.lcssa = phi i16 [ %phitmp, %for.cond.for.end_crit_edge ], [ 0, %entry ]
167  ret i16 %res.0.lcssa
168}
169
170!0 = distinct !{!0, !1}
171!1 = !{!"llvm.loop.unroll.runtime.disable"}
172
173; EPILOG: !0 = distinct !{!0, !1}
174; EPILOG: !1 = !{!"llvm.loop.unroll.disable"}
175
176; PROLOG: !0 = distinct !{!0, !1}
177; PROLOG: !1 = !{!"llvm.loop.unroll.disable"}
178