• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2; RUN: opt < %s  -loop-vectorize -force-vector-interleave=1 -force-vector-width=4 -enable-if-conversion -dce -instcombine -S | FileCheck %s
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
6define i32 @foo(i32* nocapture %A, i32* nocapture %B, i32 %n) {
7; CHECK-LABEL: @foo(
8; CHECK-NEXT:  entry:
9; CHECK-NEXT:    [[CMP26:%.*]] = icmp sgt i32 [[N:%.*]], 0
10; CHECK-NEXT:    br i1 [[CMP26]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_END:%.*]]
11; CHECK:       for.body.preheader:
12; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
13; CHECK-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
14; CHECK-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
15; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 3
16; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
17; CHECK:       vector.memcheck:
18; CHECK-NEXT:    [[TMP3:%.*]] = add i32 [[N]], -1
19; CHECK-NEXT:    [[TMP4:%.*]] = zext i32 [[TMP3]] to i64
20; CHECK-NEXT:    [[TMP5:%.*]] = add nuw nsw i64 [[TMP4]], 1
21; CHECK-NEXT:    [[SCEVGEP:%.*]] = getelementptr i32, i32* [[A:%.*]], i64 [[TMP5]]
22; CHECK-NEXT:    [[SCEVGEP4:%.*]] = getelementptr i32, i32* [[B:%.*]], i64 [[TMP5]]
23; CHECK-NEXT:    [[BOUND0:%.*]] = icmp ugt i32* [[SCEVGEP4]], [[A]]
24; CHECK-NEXT:    [[BOUND1:%.*]] = icmp ugt i32* [[SCEVGEP]], [[B]]
25; CHECK-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
26; CHECK-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
27; CHECK:       vector.ph:
28; CHECK-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934588
29; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
30; CHECK:       vector.body:
31; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
32; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDEX]]
33; CHECK-NEXT:    [[TMP7:%.*]] = bitcast i32* [[TMP6]] to <4 x i32>*
34; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, <4 x i32>* [[TMP7]], align 4, !alias.scope !0, !noalias !3
35; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDEX]]
36; CHECK-NEXT:    [[TMP9:%.*]] = bitcast i32* [[TMP8]] to <4 x i32>*
37; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i32>, <4 x i32>* [[TMP9]], align 4, !alias.scope !3
38; CHECK-NEXT:    [[TMP10:%.*]] = icmp sgt <4 x i32> [[WIDE_LOAD]], [[WIDE_LOAD6]]
39; CHECK-NEXT:    [[TMP11:%.*]] = icmp sgt <4 x i32> [[WIDE_LOAD]], <i32 19, i32 19, i32 19, i32 19>
40; CHECK-NEXT:    [[TMP12:%.*]] = icmp slt <4 x i32> [[WIDE_LOAD6]], <i32 4, i32 4, i32 4, i32 4>
41; CHECK-NEXT:    [[TMP13:%.*]] = select <4 x i1> [[TMP12]], <4 x i32> <i32 4, i32 4, i32 4, i32 4>, <4 x i32> <i32 5, i32 5, i32 5, i32 5>
42; CHECK-NEXT:    [[TMP14:%.*]] = and <4 x i1> [[TMP11]], [[TMP10]]
43; CHECK-NEXT:    [[TMP15:%.*]] = xor <4 x i1> [[TMP11]], <i1 true, i1 true, i1 true, i1 true>
44; CHECK-NEXT:    [[TMP16:%.*]] = and <4 x i1> [[TMP10]], [[TMP15]]
45; CHECK-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP14]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>, <4 x i32> <i32 9, i32 9, i32 9, i32 9>
46; CHECK-NEXT:    [[PREDPHI7:%.*]] = select <4 x i1> [[TMP16]], <4 x i32> [[TMP13]], <4 x i32> [[PREDPHI]]
47; CHECK-NEXT:    [[TMP17:%.*]] = bitcast i32* [[TMP6]] to <4 x i32>*
48; CHECK-NEXT:    store <4 x i32> [[PREDPHI7]], <4 x i32>* [[TMP17]], align 4, !alias.scope !0, !noalias !3
49; CHECK-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], 4
50; CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
51; CHECK-NEXT:    br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], [[LOOP5:!llvm.loop !.*]]
52; CHECK:       middle.block:
53; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
54; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
55; CHECK:       scalar.ph:
56; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ], [ 0, [[VECTOR_MEMCHECK]] ]
57; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
58; CHECK:       for.body:
59; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[IF_END14:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
60; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDVARS_IV]]
61; CHECK-NEXT:    [[TMP19:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
62; CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
63; CHECK-NEXT:    [[TMP20:%.*]] = load i32, i32* [[ARRAYIDX2]], align 4
64; CHECK-NEXT:    [[CMP3:%.*]] = icmp sgt i32 [[TMP19]], [[TMP20]]
65; CHECK-NEXT:    br i1 [[CMP3]], label [[IF_THEN:%.*]], label [[IF_END14]]
66; CHECK:       if.then:
67; CHECK-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[TMP19]], 19
68; CHECK-NEXT:    br i1 [[CMP6]], label [[IF_END14]], label [[IF_ELSE:%.*]]
69; CHECK:       if.else:
70; CHECK-NEXT:    [[CMP10:%.*]] = icmp slt i32 [[TMP20]], 4
71; CHECK-NEXT:    [[DOT:%.*]] = select i1 [[CMP10]], i32 4, i32 5
72; CHECK-NEXT:    br label [[IF_END14]]
73; CHECK:       if.end14:
74; CHECK-NEXT:    [[X_0:%.*]] = phi i32 [ 9, [[FOR_BODY]] ], [ 3, [[IF_THEN]] ], [ [[DOT]], [[IF_ELSE]] ]
75; CHECK-NEXT:    store i32 [[X_0]], i32* [[ARRAYIDX]], align 4
76; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
77; CHECK-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
78; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
79; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], [[LOOP7:!llvm.loop !.*]]
80; CHECK:       for.end.loopexit:
81; CHECK-NEXT:    br label [[FOR_END]]
82; CHECK:       for.end:
83; CHECK-NEXT:    ret i32 undef
84;
85entry:
86  %cmp26 = icmp sgt i32 %n, 0
87  br i1 %cmp26, label %for.body, label %for.end
88
89for.body:
90  %indvars.iv = phi i64 [ %indvars.iv.next, %if.end14 ], [ 0, %entry ]
91  %arrayidx = getelementptr inbounds i32, i32* %A, i64 %indvars.iv
92  %0 = load i32, i32* %arrayidx, align 4
93  %arrayidx2 = getelementptr inbounds i32, i32* %B, i64 %indvars.iv
94  %1 = load i32, i32* %arrayidx2, align 4
95  %cmp3 = icmp sgt i32 %0, %1
96  br i1 %cmp3, label %if.then, label %if.end14
97
98if.then:
99  %cmp6 = icmp sgt i32 %0, 19
100  br i1 %cmp6, label %if.end14, label %if.else
101
102if.else:
103  %cmp10 = icmp slt i32 %1, 4
104  %. = select i1 %cmp10, i32 4, i32 5
105  br label %if.end14
106
107if.end14:
108  %x.0 = phi i32 [ 9, %for.body ], [ 3, %if.then ], [ %., %if.else ]  ; <------------- A PHI with 3 entries that we can still vectorize.
109  store i32 %x.0, i32* %arrayidx, align 4
110  %indvars.iv.next = add i64 %indvars.iv, 1
111  %lftr.wideiv = trunc i64 %indvars.iv.next to i32
112  %exitcond = icmp eq i32 %lftr.wideiv, %n
113  br i1 %exitcond, label %for.end, label %for.body
114
115for.end:
116  ret i32 undef
117}
118
119