1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2; RUN: opt -force-vector-width=4 -force-vector-interleave=1 -loop-vectorize -S %s | FileCheck %s
3
4target datalayout = "e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
5
6define i32 @test(i64 %N, i32 %x) {
7; CHECK-LABEL: @test(
8; CHECK-NEXT:  entry:
9; CHECK-NEXT:    [[EXTRA_ITER:%.*]] = and i64 [[N:%.*]], 7
10; CHECK-NEXT:    br label [[CHECK:%.*]]
11; CHECK:       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:%.*]]
16; CHECK:       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]], i32 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:%.*]]
25; CHECK:       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]] ], [ [[TMP3:%.*]], [[VECTOR_BODY]] ]
28; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = sub i64 [[EXTRA_ITER]], [[INDEX]]
29; CHECK-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <4 x i64> poison, i64 [[OFFSET_IDX]], i32 0
30; CHECK-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT1]], <4 x i64> poison, <4 x i32> zeroinitializer
31; CHECK-NEXT:    [[INDUCTION:%.*]] = add <4 x i64> [[BROADCAST_SPLAT2]], <i64 0, i64 -1, i64 -2, i64 -3>
32; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[OFFSET_IDX]], 0
33; CHECK-NEXT:    [[BROADCAST_SPLATINSERT3:%.*]] = insertelement <4 x i64> poison, i64 [[INDEX]], i32 0
34; CHECK-NEXT:    [[BROADCAST_SPLAT4:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT3]], <4 x i64> poison, <4 x i32> zeroinitializer
35; CHECK-NEXT:    [[VEC_IV:%.*]] = add <4 x i64> [[BROADCAST_SPLAT4]], <i64 0, i64 1, i64 2, i64 3>
36; CHECK-NEXT:    [[TMP1:%.*]] = icmp ule <4 x i64> [[VEC_IV]], [[BROADCAST_SPLAT]]
37; CHECK-NEXT:    [[TMP2:%.*]] = icmp sgt <4 x i32> [[VEC_PHI]], <i32 10, i32 10, i32 10, i32 10>
38; CHECK-NEXT:    [[TMP3]] = select <4 x i1> [[TMP2]], <4 x i32> [[VEC_PHI]], <4 x i32> <i32 10, i32 10, i32 10, i32 10>
39; CHECK-NEXT:    [[TMP4:%.*]] = select <4 x i1> [[TMP1]], <4 x i32> [[TMP3]], <4 x i32> [[VEC_PHI]]
40; CHECK-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], 4
41; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
42; CHECK-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], [[LOOP0:!llvm.loop !.*]]
43; CHECK:       middle.block:
44; CHECK-NEXT:    [[TMP6:%.*]] = call i32 @llvm.vector.reduce.smax.v4i32(<4 x i32> [[TMP4]])
45; CHECK-NEXT:    br i1 true, label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]
46; CHECK:       scalar.ph:
47; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[EXTRA_ITER]], [[LOOP_PREHEADER]] ]
48; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[LOOP_PREHEADER]] ], [ [[TMP6]], [[MIDDLE_BLOCK]] ]
49; CHECK-NEXT:    br label [[LOOP:%.*]]
50; CHECK:       loop:
51; CHECK-NEXT:    [[NEXT:%.*]] = phi i32 [ [[SEL:%.*]], [[LOOP]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
52; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
53; CHECK-NEXT:    [[SEL_COND:%.*]] = icmp sgt i32 [[NEXT]], 10
54; CHECK-NEXT:    [[SEL]] = select i1 [[SEL_COND]], i32 [[NEXT]], i32 10
55; CHECK-NEXT:    [[IV_NEXT]] = add nsw i64 [[IV]], -1
56; CHECK-NEXT:    [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], 0
57; CHECK-NEXT:    br i1 [[EC]], label [[EXIT_LOOPEXIT]], label [[LOOP]], [[LOOP2:!llvm.loop !.*]]
58; CHECK:       exit.loopexit:
59; CHECK-NEXT:    [[SEL_LCSSA:%.*]] = phi i32 [ [[SEL]], [[LOOP]] ], [ [[TMP6]], [[MIDDLE_BLOCK]] ]
60; CHECK-NEXT:    br label [[EXIT]]
61; CHECK:       exit:
62; CHECK-NEXT:    [[RESULT:%.*]] = phi i32 [ 0, [[CHECK]] ], [ [[SEL_LCSSA]], [[EXIT_LOOPEXIT]] ]
63; CHECK-NEXT:    ret i32 [[RESULT]]
64;
65entry:
66  %extra.iter = and i64 %N, 7
67  br label %check
68
69check:
70  %extra.iter.check = icmp eq i64 %extra.iter, 0
71  br i1 %extra.iter.check, label %exit, label %loop
72
73loop:
74  %next = phi i32 [ %sel, %loop ], [ 0, %check ]
75  %iv = phi i64 [ %iv.next, %loop ], [ %extra.iter, %check ]
76  %sel.cond = icmp sgt i32 %next, 10
77  %sel = select i1 %sel.cond, i32 %next, i32 10
78  %iv.next = add nsw i64 %iv, -1
79  %ec = icmp eq i64 %iv.next, 0
80  br i1 %ec, label %exit, label %loop
81
82exit:
83  %result = phi i32 [ %sel, %loop], [ 0, %check ]
84  ret i32 %result
85}
86