1; REQUIRES: asserts
2
3; RUN: opt -loop-vectorize -debug-only=loop-vectorize -force-vector-interleave=1 -force-vector-width=4 -prefer-inloop-reductions -disable-output %s 2>&1 | FileCheck %s
4
5target 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
7; Tests for printing VPlans.
8
9define void @print_call_and_memory(i64 %n, float* noalias %y, float* noalias %x) nounwind uwtable {
10; CHECK: N0 [label =
11; CHECK-NEXT: "for.body:\n" +
12; CHECK-NEXT:       "WIDEN-INDUCTION %iv = phi %iv.next, 0\l" +
13; CHECK-NEXT:       "CLONE %arrayidx = getelementptr %y, %iv\l" +
14; CHECK-NEXT:       "WIDEN ir<%lv> = load ir<%arrayidx>\l" +
15; CHECK-NEXT:       "WIDEN-CALL ir<%call> = call @llvm.sqrt.f32(ir<%lv>)\l" +
16; CHECK-NEXT:       "CLONE %arrayidx2 = getelementptr %x, %iv\l" +
17; CHECK-NEXT:       "WIDEN store ir<%arrayidx2>, ir<%call>\l"
18; CHECK-NEXT:   ]
19
20entry:
21  %cmp6 = icmp sgt i64 %n, 0
22  br i1 %cmp6, label %for.body, label %for.end
23
24for.body:                                         ; preds = %entry, %for.body
25  %iv = phi i64 [ %iv.next, %for.body ], [ 0, %entry ]
26  %arrayidx = getelementptr inbounds float, float* %y, i64 %iv
27  %lv = load float, float* %arrayidx, align 4
28  %call = tail call float @llvm.sqrt.f32(float %lv) nounwind readnone
29  %arrayidx2 = getelementptr inbounds float, float* %x, i64 %iv
30  store float %call, float* %arrayidx2, align 4
31  %iv.next = add i64 %iv, 1
32  %exitcond = icmp eq i64 %iv.next, %n
33  br i1 %exitcond, label %for.end, label %for.body
34
35for.end:                                          ; preds = %for.body, %entry
36  ret void
37}
38
39define void @print_widen_gep_and_select(i64 %n, float* noalias %y, float* noalias %x, float* %z) nounwind uwtable {
40; CHECK: N0 [label =
41; CHECK-NEXT: "for.body:\n" +
42; CHECK-NEXT:      "WIDEN-INDUCTION %iv = phi %iv.next, 0\l" +
43; CHECK-NEXT:      "WIDEN-GEP Inv[Var] ir<%arrayidx> = getelementptr ir<%y>, ir<%iv>\l" +
44; CHECK-NEXT:      "WIDEN ir<%lv> = load ir<%arrayidx>\l" +
45; CHECK-NEXT:      "WIDEN ir<%cmp> = icmp ir<%arrayidx>, ir<%z>\l" +
46; CHECK-NEXT:      "WIDEN-SELECT ir<%sel> = select ir<%cmp>, ir<1.000000e+01>, ir<2.000000e+01>\l" +
47; CHECK-NEXT:      "WIDEN ir<%add> = fadd ir<%lv>, ir<%sel>\l" +
48; CHECK-NEXT:      "CLONE %arrayidx2 = getelementptr %x, %iv\l" +
49; CHECK-NEXT:      "WIDEN store ir<%arrayidx2>, ir<%add>\l"
50; CHECK-NEXT:   ]
51
52entry:
53  %cmp6 = icmp sgt i64 %n, 0
54  br i1 %cmp6, label %for.body, label %for.end
55
56for.body:                                         ; preds = %entry, %for.body
57  %iv = phi i64 [ %iv.next, %for.body ], [ 0, %entry ]
58  %arrayidx = getelementptr inbounds float, float* %y, i64 %iv
59  %lv = load float, float* %arrayidx, align 4
60  %cmp = icmp eq float* %arrayidx, %z
61  %sel = select i1 %cmp, float 10.0, float 20.0
62  %add = fadd float %lv, %sel
63  %arrayidx2 = getelementptr inbounds float, float* %x, i64 %iv
64  store float %add, float* %arrayidx2, align 4
65  %iv.next = add i64 %iv, 1
66  %exitcond = icmp eq i64 %iv.next, %n
67  br i1 %exitcond, label %for.end, label %for.body
68
69for.end:                                          ; preds = %for.body, %entry
70  ret void
71}
72
73define float @print_reduction(i64 %n, float* noalias %y) {
74; CHECK: N0 [label =
75; CHECK-NEXT: "for.body:\n" +
76; CHECK-NEXT:       "WIDEN-INDUCTION %iv = phi %iv.next, 0\l" +
77; CHECK-NEXT:       "WIDEN-PHI %red = phi %red.next, 0.000000e+00\l" +
78; CHECK-NEXT:       "CLONE %arrayidx = getelementptr %y, %iv\l" +
79; CHECK-NEXT:       "WIDEN ir<%lv> = load ir<%arrayidx>\l" +
80; CHECK-NEXT:       "REDUCE ir<%red.next> = ir<%red> + reduce.fadd (ir<%lv>)\l"
81; CHECK-NEXT:   ]
82
83entry:
84  br label %for.body
85
86for.body:                                         ; preds = %entry, %for.body
87  %iv = phi i64 [ %iv.next, %for.body ], [ 0, %entry ]
88  %red = phi float [ %red.next, %for.body ], [ 0.0, %entry ]
89  %arrayidx = getelementptr inbounds float, float* %y, i64 %iv
90  %lv = load float, float* %arrayidx, align 4
91  %red.next = fadd fast float %lv, %red
92  %iv.next = add i64 %iv, 1
93  %exitcond = icmp eq i64 %iv.next, %n
94  br i1 %exitcond, label %for.end, label %for.body
95
96for.end:                                          ; preds = %for.body, %entry
97  ret float %red.next
98}
99
100declare float @llvm.sqrt.f32(float) nounwind readnone
101