| blob | 599b6e544efe0edc1243de6be90eadd5910521c8 |
1 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2 ; RUN: opt -loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -S < %s | FileCheck %s
4 define void @m(i32* nocapture %p, i32* nocapture %p2, i32 %q) {
5 ; CHECK-LABEL: @m(
6 ; CHECK-NEXT: entry:
7 ; CHECK-NEXT: [[P1:%.*]] = bitcast i32* [[P:%.*]] to i8*
8 ; CHECK-NEXT: [[P23:%.*]] = bitcast i32* [[P2:%.*]] to i8*
9 ; CHECK-NEXT: [[ARRAYIDX9_1:%.*]] = getelementptr inbounds i32, i32* [[P2]], i64 1
10 ; CHECK-NEXT: [[ARRAYIDX9_2:%.*]] = getelementptr inbounds i32, i32* [[P2]], i64 2
11 ; CHECK-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
12 ; CHECK: vector.memcheck:
13 ; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr i32, i32* [[P]], i64 63
14 ; CHECK-NEXT: [[SCEVGEP2:%.*]] = bitcast i32* [[SCEVGEP]] to i8*
15 ; CHECK-NEXT: [[SCEVGEP4:%.*]] = getelementptr i32, i32* [[P2]], i64 3
16 ; CHECK-NEXT: [[SCEVGEP45:%.*]] = bitcast i32* [[SCEVGEP4]] to i8*
17 ; CHECK-NEXT: [[BOUND0:%.*]] = icmp ult i8* [[P1]], [[SCEVGEP45]]
18 ; CHECK-NEXT: [[BOUND1:%.*]] = icmp ult i8* [[P23]], [[SCEVGEP2]]
19 ; CHECK-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
20 ; CHECK-NEXT: [[MEMCHECK_CONFLICT:%.*]] = and i1 [[FOUND_CONFLICT]], true
21 ; CHECK-NEXT: br i1 [[MEMCHECK_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
22 ; CHECK: vector.ph:
23 ; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
24 ; CHECK: vector.body:
25 ; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
26 ; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0
27 ; CHECK-NEXT: [[TMP1:%.*]] = load i32, i32* [[P2]], align 4, !alias.scope !0
28 ; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[TMP1]], i32 0
29 ; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
30 ; CHECK-NEXT: [[TMP2:%.*]] = sub nsw <4 x i32> zeroinitializer, [[BROADCAST_SPLAT]]
31 ; CHECK-NEXT: [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX9_1]], align 4, !alias.scope !0
32 ; CHECK-NEXT: [[BROADCAST_SPLATINSERT6:%.*]] = insertelement <4 x i32> poison, i32 [[TMP3]], i32 0
33 ; CHECK-NEXT: [[BROADCAST_SPLAT7:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT6]], <4 x i32> poison, <4 x i32> zeroinitializer
34 ; CHECK-NEXT: [[TMP4:%.*]] = sub nsw <4 x i32> [[TMP2]], [[BROADCAST_SPLAT7]]
35 ; CHECK-NEXT: [[TMP5:%.*]] = load i32, i32* [[ARRAYIDX9_2]], align 4, !alias.scope !0
36 ; CHECK-NEXT: [[BROADCAST_SPLATINSERT8:%.*]] = insertelement <4 x i32> poison, i32 [[TMP5]], i32 0
37 ; CHECK-NEXT: [[BROADCAST_SPLAT9:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT8]], <4 x i32> poison, <4 x i32> zeroinitializer
38 ; CHECK-NEXT: [[TMP6:%.*]] = sub nsw <4 x i32> [[TMP4]], [[BROADCAST_SPLAT9]]
39 ; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds i32, i32* [[P]], i64 [[TMP0]]
40 ; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds i32, i32* [[TMP7]], i32 0
41 ; CHECK-NEXT: [[TMP9:%.*]] = bitcast i32* [[TMP8]] to <4 x i32>*
42 ; CHECK-NEXT: store <4 x i32> [[TMP6]], <4 x i32>* [[TMP9]], align 4, !alias.scope !3, !noalias !0
43 ; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
44 ; CHECK-NEXT: [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 60
45 ; CHECK-NEXT: br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
46 ; CHECK: middle.block:
47 ; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 63, 60
48 ; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END17:%.*]], label [[SCALAR_PH]]
49 ; CHECK: scalar.ph:
50 ; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 60, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
51 ; CHECK-NEXT: br label [[FOR_COND5:%.*]]
52 ; CHECK: for.cond5:
53 ; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_COND5]] ]
54 ; CHECK-NEXT: [[I3:%.*]] = load i32, i32* [[P2]], align 4
55 ; CHECK-NEXT: [[SUB:%.*]] = sub nsw i32 0, [[I3]]
56 ; CHECK-NEXT: [[I4:%.*]] = load i32, i32* [[ARRAYIDX9_1]], align 4
57 ; CHECK-NEXT: [[SUB_1:%.*]] = sub nsw i32 [[SUB]], [[I4]]
58 ; CHECK-NEXT: [[I5:%.*]] = load i32, i32* [[ARRAYIDX9_2]], align 4
59 ; CHECK-NEXT: [[SUB_2:%.*]] = sub nsw i32 [[SUB_1]], [[I5]]
60 ; CHECK-NEXT: [[ARRAYIDX14:%.*]] = getelementptr inbounds i32, i32* [[P]], i64 [[INDVARS_IV]]
61 ; CHECK-NEXT: store i32 [[SUB_2]], i32* [[ARRAYIDX14]], align 4
62 ; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
63 ; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 63
64 ; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_END17]], label [[FOR_COND5]], !llvm.loop [[LOOP7:![0-9]+]]
65 ; CHECK: for.end17:
66 ; CHECK-NEXT: ret void
67 ;
68 entry:
69 %arrayidx9.1 = getelementptr inbounds i32, i32* %p2, i64 1
70 %arrayidx9.2 = getelementptr inbounds i32, i32* %p2, i64 2
71 br label %for.cond5
73 for.cond5: ; preds = %entry, %for.cond5
74 %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.cond5 ]
75 %i3 = load i32, i32* %p2, align 4
76 %sub = sub nsw i32 0, %i3
77 %i4 = load i32, i32* %arrayidx9.1, align 4
78 %sub.1 = sub nsw i32 %sub, %i4
79 %i5 = load i32, i32* %arrayidx9.2, align 4
80 %sub.2 = sub nsw i32 %sub.1, %i5
81 %arrayidx14 = getelementptr inbounds i32, i32* %p, i64 %indvars.iv
82 store i32 %sub.2, i32* %arrayidx14, align 4
83 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
84 %exitcond = icmp eq i64 %indvars.iv.next, 63
85 br i1 %exitcond, label %for.end17, label %for.cond5
87 for.end17: ; preds = %for.cond5
88 ret void
89 }