1 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2 ; RUN: opt -force-vector-width=4 -force-vector-interleave=1 -passes=loop-vectorize -S %s | FileCheck %s
4 target datalayout = "e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
6 define i32 @test(i64 %N, i32 %x) {
9 ; CHECK-NEXT: [[EXTRA_ITER:%.*]] = and i64 [[N:%.*]], 7
10 ; CHECK-NEXT: br label [[CHECK:%.*]]
12 ; CHECK-NEXT: [[EXTRA_ITER_CHECK:%.*]] = icmp eq i64 [[EXTRA_ITER]], 0
13 ; CHECK-NEXT: br i1 [[EXTRA_ITER_CHECK]], label [[EXIT:%.*]], label [[LOOP_PREHEADER:%.*]]
14 ; CHECK: loop.preheader:
15 ; CHECK-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
17 ; CHECK-NEXT: [[N_RND_UP:%.*]] = add i64 [[EXTRA_ITER]], 3
18 ; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 4
19 ; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
20 ; CHECK-NEXT: [[IND_END:%.*]] = sub i64 [[EXTRA_ITER]], [[N_VEC]]
21 ; CHECK-NEXT: [[TRIP_COUNT_MINUS_1:%.*]] = sub i64 [[EXTRA_ITER]], 1
22 ; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i64> poison, i64 [[TRIP_COUNT_MINUS_1]], i64 0
23 ; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT]], <4 x i64> poison, <4 x i32> zeroinitializer
24 ; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
26 ; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
27 ; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP2:%.*]], [[VECTOR_BODY]] ]
28 ; CHECK-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <4 x i64> poison, i64 [[INDEX]], i64 0
29 ; CHECK-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT1]], <4 x i64> poison, <4 x i32> zeroinitializer
30 ; CHECK-NEXT: [[VEC_IV:%.*]] = add <4 x i64> [[BROADCAST_SPLAT2]], <i64 0, i64 1, i64 2, i64 3>
31 ; CHECK-NEXT: [[TMP0:%.*]] = icmp ule <4 x i64> [[VEC_IV]], [[BROADCAST_SPLAT]]
32 ; CHECK-NEXT: [[TMP1:%.*]] = icmp sgt <4 x i32> [[VEC_PHI]], <i32 10, i32 10, i32 10, i32 10>
33 ; CHECK-NEXT: [[TMP2]] = select <4 x i1> [[TMP1]], <4 x i32> [[VEC_PHI]], <4 x i32> <i32 10, i32 10, i32 10, i32 10>
34 ; CHECK-NEXT: [[TMP3:%.*]] = select <4 x i1> [[TMP0]], <4 x i32> [[TMP2]], <4 x i32> [[VEC_PHI]]
35 ; CHECK-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], 4
36 ; CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
37 ; CHECK-NEXT: br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
38 ; CHECK: middle.block:
39 ; CHECK-NEXT: [[TMP5:%.*]] = call i32 @llvm.vector.reduce.smax.v4i32(<4 x i32> [[TMP3]])
40 ; CHECK-NEXT: br i1 true, label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]
42 ; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[EXTRA_ITER]], [[LOOP_PREHEADER]] ]
43 ; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP5]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ]
44 ; CHECK-NEXT: br label [[LOOP:%.*]]
46 ; CHECK-NEXT: [[NEXT:%.*]] = phi i32 [ [[SEL:%.*]], [[LOOP]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
47 ; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
48 ; CHECK-NEXT: [[SEL_COND:%.*]] = icmp sgt i32 [[NEXT]], 10
49 ; CHECK-NEXT: [[SEL]] = select i1 [[SEL_COND]], i32 [[NEXT]], i32 10
50 ; CHECK-NEXT: [[IV_NEXT]] = add nsw i64 [[IV]], -1
51 ; CHECK-NEXT: [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], 0
52 ; CHECK-NEXT: br i1 [[EC]], label [[EXIT_LOOPEXIT]], label [[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
53 ; CHECK: exit.loopexit:
54 ; CHECK-NEXT: [[SEL_LCSSA:%.*]] = phi i32 [ [[SEL]], [[LOOP]] ], [ [[TMP5]], [[MIDDLE_BLOCK]] ]
55 ; CHECK-NEXT: br label [[EXIT]]
57 ; CHECK-NEXT: [[RESULT:%.*]] = phi i32 [ 0, [[CHECK]] ], [ [[SEL_LCSSA]], [[EXIT_LOOPEXIT]] ]
58 ; CHECK-NEXT: ret i32 [[RESULT]]
61 %extra.iter = and i64 %N, 7
65 %extra.iter.check = icmp eq i64 %extra.iter, 0
66 br i1 %extra.iter.check, label %exit, label %loop
69 %next = phi i32 [ %sel, %loop ], [ 0, %check ]
70 %iv = phi i64 [ %iv.next, %loop ], [ %extra.iter, %check ]
71 %sel.cond = icmp sgt i32 %next, 10
72 %sel = select i1 %sel.cond, i32 %next, i32 10
73 %iv.next = add nsw i64 %iv, -1
74 %ec = icmp eq i64 %iv.next, 0
75 br i1 %ec, label %exit, label %loop
78 %result = phi i32 [ %sel, %loop], [ 0, %check ]
82 define i32 @pr66895_tail_fold_reduction_exit_inst_gets_simplified(i32 %n) {
83 ; CHECK-LABEL: @pr66895_tail_fold_reduction_exit_inst_gets_simplified(
85 ; CHECK-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
87 ; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
89 ; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
90 ; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 1, i32 1>, [[VECTOR_PH]] ], [ [[VEC_PHI]], [[VECTOR_BODY]] ]
91 ; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[INDEX]], i64 0
92 ; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
93 ; CHECK-NEXT: [[VEC_IV:%.*]] = add <4 x i32> [[BROADCAST_SPLAT]], <i32 0, i32 1, i32 2, i32 3>
94 ; CHECK-NEXT: [[TMP0:%.*]] = icmp ule <4 x i32> [[VEC_IV]], <i32 12, i32 12, i32 12, i32 12>
95 ; CHECK-NEXT: [[TMP1:%.*]] = select <4 x i1> [[TMP0]], <4 x i32> [[VEC_PHI]], <4 x i32> [[VEC_PHI]]
96 ; CHECK-NEXT: [[INDEX_NEXT]] = add i32 [[INDEX]], 4
97 ; CHECK-NEXT: [[TMP2:%.*]] = icmp eq i32 [[INDEX_NEXT]], 16
98 ; CHECK-NEXT: br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
99 ; CHECK: middle.block:
100 ; CHECK-NEXT: [[TMP3:%.*]] = call i32 @llvm.vector.reduce.mul.v4i32(<4 x i32> [[TMP1]])
101 ; CHECK-NEXT: br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
103 ; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ -4, [[MIDDLE_BLOCK]] ], [ 12, [[ENTRY:%.*]] ]
104 ; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP3]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
105 ; CHECK-NEXT: br label [[LOOP:%.*]]
107 ; CHECK-NEXT: [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
108 ; CHECK-NEXT: [[RED:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[RED_NEXT:%.*]], [[LOOP]] ]
109 ; CHECK-NEXT: [[IV_NEXT]] = add i32 [[IV]], -1
110 ; CHECK-NEXT: [[RED_NEXT]] = mul i32 [[RED]], 1
111 ; CHECK-NEXT: [[EC:%.*]] = icmp eq i32 [[IV]], 0
112 ; CHECK-NEXT: br i1 [[EC]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP5:![0-9]+]]
114 ; CHECK-NEXT: [[RED_LCSSA:%.*]] = phi i32 [ [[RED_NEXT]], [[LOOP]] ], [ [[TMP3]], [[MIDDLE_BLOCK]] ]
115 ; CHECK-NEXT: ret i32 [[RED_LCSSA]]
121 %iv = phi i32 [ 12, %entry ], [ %iv.next, %loop ]
122 %red = phi i32 [ 0, %entry ], [ %red.next, %loop ]
123 %iv.next = add i32 %iv, -1
124 %red.next = mul i32 %red, 1
125 %ec = icmp eq i32 %iv, 0
126 br i1 %ec, label %exit, label %loop
129 %red.lcssa = phi i32 [ %red.next, %loop ]