xref: /llvm-project/llvm/test/Transforms/LoopVectorize/tail-folding-alloca-in-loop.ll (revision 82821254f532c1dbdfd5d985ef7130511efaaa83)
1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
2; RUN: opt -S --passes=loop-vectorize -prefer-predicate-over-epilogue=predicate-dont-vectorize -force-vector-width=4  < %s | FileCheck %s
3
4define i32 @test(ptr %vf1, i64 %n) {
5; CHECK-LABEL: define i32 @test(
6; CHECK-SAME: ptr [[VF1:%.*]], i64 [[N:%.*]]) {
7; CHECK-NEXT:  [[ENTRY:.*]]:
8; CHECK-NEXT:    br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
9; CHECK:       [[VECTOR_PH]]:
10; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
11; CHECK:       [[VECTOR_BODY]]:
12; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_STORE_CONTINUE6:.*]] ]
13; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[PRED_STORE_CONTINUE6]] ]
14; CHECK-NEXT:    [[TMP0:%.*]] = icmp ule <4 x i64> [[VEC_IND]], splat (i64 200)
15; CHECK-NEXT:    [[TMP1:%.*]] = extractelement <4 x i1> [[TMP0]], i32 0
16; CHECK-NEXT:    br i1 [[TMP1]], label %[[PRED_STORE_IF:.*]], label %[[PRED_STORE_CONTINUE:.*]]
17; CHECK:       [[PRED_STORE_IF]]:
18; CHECK-NEXT:    [[TMP2:%.*]] = add i64 [[INDEX]], 0
19; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds ptr, ptr [[VF1]], i64 [[TMP2]]
20; CHECK-NEXT:    [[TMP4:%.*]] = alloca i8, i64 [[N]], align 16
21; CHECK-NEXT:    store ptr [[TMP4]], ptr [[TMP3]], align 8
22; CHECK-NEXT:    br label %[[PRED_STORE_CONTINUE]]
23; CHECK:       [[PRED_STORE_CONTINUE]]:
24; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <4 x i1> [[TMP0]], i32 1
25; CHECK-NEXT:    br i1 [[TMP5]], label %[[PRED_STORE_IF1:.*]], label %[[PRED_STORE_CONTINUE2:.*]]
26; CHECK:       [[PRED_STORE_IF1]]:
27; CHECK-NEXT:    [[TMP6:%.*]] = add i64 [[INDEX]], 1
28; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds ptr, ptr [[VF1]], i64 [[TMP6]]
29; CHECK-NEXT:    [[TMP8:%.*]] = alloca i8, i64 [[N]], align 16
30; CHECK-NEXT:    store ptr [[TMP8]], ptr [[TMP7]], align 8
31; CHECK-NEXT:    br label %[[PRED_STORE_CONTINUE2]]
32; CHECK:       [[PRED_STORE_CONTINUE2]]:
33; CHECK-NEXT:    [[TMP9:%.*]] = extractelement <4 x i1> [[TMP0]], i32 2
34; CHECK-NEXT:    br i1 [[TMP9]], label %[[PRED_STORE_IF3:.*]], label %[[PRED_STORE_CONTINUE4:.*]]
35; CHECK:       [[PRED_STORE_IF3]]:
36; CHECK-NEXT:    [[TMP10:%.*]] = add i64 [[INDEX]], 2
37; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds ptr, ptr [[VF1]], i64 [[TMP10]]
38; CHECK-NEXT:    [[TMP12:%.*]] = alloca i8, i64 [[N]], align 16
39; CHECK-NEXT:    store ptr [[TMP12]], ptr [[TMP11]], align 8
40; CHECK-NEXT:    br label %[[PRED_STORE_CONTINUE4]]
41; CHECK:       [[PRED_STORE_CONTINUE4]]:
42; CHECK-NEXT:    [[TMP13:%.*]] = extractelement <4 x i1> [[TMP0]], i32 3
43; CHECK-NEXT:    br i1 [[TMP13]], label %[[PRED_STORE_IF5:.*]], label %[[PRED_STORE_CONTINUE6]]
44; CHECK:       [[PRED_STORE_IF5]]:
45; CHECK-NEXT:    [[TMP14:%.*]] = add i64 [[INDEX]], 3
46; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds ptr, ptr [[VF1]], i64 [[TMP14]]
47; CHECK-NEXT:    [[TMP16:%.*]] = alloca i8, i64 [[N]], align 16
48; CHECK-NEXT:    store ptr [[TMP16]], ptr [[TMP15]], align 8
49; CHECK-NEXT:    br label %[[PRED_STORE_CONTINUE6]]
50; CHECK:       [[PRED_STORE_CONTINUE6]]:
51; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
52; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
53; CHECK-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], 204
54; CHECK-NEXT:    br i1 [[TMP17]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
55; CHECK:       [[MIDDLE_BLOCK]]:
56; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
57; CHECK:       [[SCALAR_PH]]:
58; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 204, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
59; CHECK-NEXT:    br label %[[FOR_BODY:.*]]
60; CHECK:       [[FOR_BODY]]:
61; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
62; CHECK-NEXT:    [[TMP18:%.*]] = alloca i8, i64 [[N]], align 16
63; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds ptr, ptr [[VF1]], i64 [[INDVARS_IV]]
64; CHECK-NEXT:    store ptr [[TMP18]], ptr [[ARRAYIDX]], align 8
65; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
66; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INDVARS_IV]], 200
67; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
68; CHECK:       [[EXIT]]:
69; CHECK-NEXT:    ret i32 0
70;
71entry:
72  br label %for.body
73
74for.body:
75  %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
76  %0 = alloca i8, i64 %n, align 16
77  %arrayidx = getelementptr inbounds ptr, ptr %vf1, i64 %indvars.iv
78  store ptr %0, ptr %arrayidx, align 8
79  %indvars.iv.next = add i64 %indvars.iv, 1
80  %exitcond.not = icmp eq i64 %indvars.iv, 200
81  br i1 %exitcond.not, label %exit, label %for.body
82
83exit:
84  ret i32 0
85}
86;.
87; CHECK: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}
88; CHECK: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}
89; CHECK: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}
90; CHECK: [[LOOP3]] = distinct !{[[LOOP3]], [[META2]], [[META1]]}
91;.
92