xref: /llvm-project/llvm/test/Transforms/LoopVectorize/vplan-sink-scalars-and-merge-vf1.ll (revision 1de3dc7d23dd6b856efad3a3a04f2396328726d7)
1; REQUIRES: asserts
2
3; RUN: opt -passes=loop-vectorize -force-vector-interleave=2 -force-vector-width=1 -debug -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; Make sure recipes with side-effects are not sunk.
8define void @sink_with_sideeffects(i1 %c, ptr %ptr) {
9; CHECK-LABEL: sink_with_sideeffects
10; CHECK:      VPlan 'Initial VPlan for VF={1},UF>=1' {
11; CHECK-NEXT: Live-in vp<[[VFxUF:%.+]]> = VF * UF
12; CHECK-NEXT: Live-in vp<[[VEC_TC:%.+]]> = vector-trip-count
13; CHECK-NEXT: ir<1024> = original trip-count
14; CHECK-EMPTY:
15; CHECK-NEXT: ir-bb<entry>:
16; CHECK-NEXT: Successor(s): vector.ph
17; CHECK-EMPTY:
18; CHECK-NEXT: vector.ph:
19; CHECK-NEXT:   vp<[[END:%.+]]> = DERIVED-IV ir<1024> + vp<[[VEC_TC]]> * ir<-1>
20; CHECK-NEXT: Successor(s): vector loop
21; CHECK-EMPTY:
22; CHECK-NEXT: <x1> vector loop: {
23; CHECK-NEXT: vector.body:
24; CHECK-NEXT:   EMIT vp<[[CAN_IV:%.+]]> = CANONICAL-INDUCTION
25; CHECK-NEXT:   vp<[[STEPS:%.+]]> = SCALAR-STEPS vp<[[CAN_IV]]>, ir<1>
26; CHECK-NEXT:   CLONE ir<%tmp2> = getelementptr ir<%ptr>, vp<[[STEPS]]>
27; CHECK-NEXT:   CLONE ir<%tmp3> = load ir<%tmp2>
28; CHECK-NEXT:   CLONE store ir<0>, ir<%tmp2>
29; CHECK-NEXT: Successor(s): pred.store
30
31; CHECK:      <xVFxUF> pred.store: {
32; CHECK-NEXT:   pred.store.entry:
33; CHECK-NEXT:     BRANCH-ON-MASK ir<%c>
34; CHECK-NEXT:   Successor(s): pred.store.if, pred.store.continue
35
36; CHECK:      pred.store.if:
37; CHECK-NEXT:   CLONE store ir<%tmp3>, ir<%tmp2>
38; CHECK-NEXT:   Successor(s): pred.store.continue
39
40; CHECK:      pred.store.continue:
41; CHECK-NEXT:   No successors
42; CHECK-NEXT: }
43
44; CHECK:      if.then.0:
45; CHECK-NEXT:  EMIT vp<[[CAN_IV_NEXT:%.+]]> = add nuw vp<[[CAN_IV]]>, vp<[[VFxUF]]>
46; CHECK-NEXT:  EMIT branch-on-count vp<[[CAN_IV_NEXT]]>, vp<[[VEC_TC]]>
47; CHECK-NEXT: No successors
48; CHECK-NEXT: }
49; CHECK-NEXT: Successor(s): middle.block
50; CHECK-EMPTY:
51; CHECK-NEXT: middle.block:
52; CHECK-NEXT:    EMIT vp<[[CMP:%.+]]> = icmp eq ir<1024>, vp<[[VEC_TC]]>
53; CHECK-NEXT:    EMIT branch-on-cond vp<[[CMP]]>
54; CHECK-NEXT:  Successor(s): ir-bb<for.end>, scalar.ph
55; CHECK-EMPTY:
56; CHECK-NEXT:  scalar.ph:
57; CHECK-NEXT:    EMIT vp<[[RESUME1:%.+]]> = resume-phi vp<[[VEC_TC]]>, ir<0>
58; CHECK-NEXT:    EMIT vp<[[RESUME2:%.+]]>.1 = resume-phi vp<[[END]]>, ir<1024>
59; CHECK-NEXT:  Successor(s): ir-bb<for.body>
60; CHECK-EMPTY:
61; CHECK-NEXT:  ir-bb<for.body>:
62; CHECK-NEXT:    IR   %tmp0 = phi i64 [ %tmp6, %for.inc ], [ 0, %entry ] (extra operand: vp<[[RESUME1]]> from scalar.ph)
63; CHECK-NEXT:    IR   %tmp1 = phi i64 [ %tmp7, %for.inc ], [ 1024, %entry ] (extra operand: vp<[[RESUME2]]>.1 from scalar.ph)
64; CHECK:         IR   %tmp5 = trunc i32 %tmp4 to i8
65; CHECK-NEXT:  No successors
66; CHECK-EMPTY:
67; CHECK-NEXT:  ir-bb<for.end>:
68; CHECK-NEXT:  No successors
69; CHECK-NEXT: }
70;
71entry:
72  br label %for.body
73
74for.body:
75  %tmp0 = phi i64 [ %tmp6, %for.inc ], [ 0, %entry ]
76  %tmp1 = phi i64 [ %tmp7, %for.inc ], [ 1024, %entry ]
77  %tmp2 = getelementptr i8, ptr %ptr, i64 %tmp0
78  %tmp3 = load i8, ptr %tmp2, align 1
79  store i8 0, ptr %tmp2
80  %tmp4 = zext i8 %tmp3 to i32
81  %tmp5 = trunc i32 %tmp4 to i8
82  br i1 %c, label %if.then, label %for.inc
83
84if.then:
85  store i8 %tmp5, ptr %tmp2, align 1
86  br label %for.inc
87
88for.inc:
89  %tmp6 = add nuw nsw i64 %tmp0, 1
90  %tmp7 = add i64 %tmp1, -1
91  %tmp8 = icmp eq i64 %tmp7, 0
92  br i1 %tmp8, label %for.end, label %for.body
93
94for.end:
95  ret void
96}
97