xref: /llvm-project/llvm/test/Transforms/LoopVectorize/interleave-with-i65-induction.ll (revision 4f7f71b7bccdc38f37b82981e8fa9ceb536a7016)
1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
2; RUN: opt -p loop-vectorize -force-vector-width=4 -force-vector-interleave=2 -S %s | FileCheck %s
3
4; Test case for https://github.com/llvm/llvm-project/issues/118850.
5define void @i65_induction_with_negative_step(ptr %dst) {
6; CHECK-LABEL: define void @i65_induction_with_negative_step(
7; CHECK-SAME: ptr [[DST:%.*]]) {
8; CHECK-NEXT:  [[ENTRY:.*]]:
9; CHECK-NEXT:    br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
10; CHECK:       [[VECTOR_PH]]:
11; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
12; CHECK:       [[VECTOR_BODY]]:
13; CHECK-NEXT:    [[INDEX:%.*]] = phi i65 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
14; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i64> [ <i64 poison, i64 poison, i64 poison, i64 0>, %[[VECTOR_PH]] ], [ [[STEP_ADD:%.*]], %[[VECTOR_BODY]] ]
15; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 0, i64 -1, i64 -2, i64 -3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
16; CHECK-NEXT:    [[STEP_ADD]] = add <4 x i64> [[VEC_IND]], splat (i64 -4)
17; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = sub i65 0, [[INDEX]]
18; CHECK-NEXT:    [[TMP0:%.*]] = trunc i65 [[OFFSET_IDX]] to i64
19; CHECK-NEXT:    [[TMP1:%.*]] = add i64 [[TMP0]], 0
20; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i64> [[VECTOR_RECUR]], <4 x i64> [[VEC_IND]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
21; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <4 x i64> [[VEC_IND]], <4 x i64> [[STEP_ADD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
22; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[DST]], i64 [[TMP1]]
23; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP4]], i32 0
24; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[TMP5]], i32 -3
25; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[TMP4]], i32 -4
26; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[TMP7]], i32 -3
27; CHECK-NEXT:    [[REVERSE:%.*]] = shufflevector <4 x i64> [[TMP2]], <4 x i64> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
28; CHECK-NEXT:    store <4 x i64> [[REVERSE]], ptr [[TMP6]], align 8
29; CHECK-NEXT:    [[REVERSE1:%.*]] = shufflevector <4 x i64> [[TMP3]], <4 x i64> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
30; CHECK-NEXT:    store <4 x i64> [[REVERSE1]], ptr [[TMP8]], align 8
31; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i65 [[INDEX]], 8
32; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[STEP_ADD]], splat (i64 -4)
33; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i65 [[INDEX_NEXT]], 1000
34; CHECK-NEXT:    br i1 [[TMP9]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
35; CHECK:       [[MIDDLE_BLOCK]]:
36; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i64> [[STEP_ADD]], i32 3
37; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
38; CHECK:       [[SCALAR_PH]]:
39; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
40; CHECK-NEXT:    [[BC_RESUME_VAL2:%.*]] = phi i65 [ -1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
41; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i64 [ [[VECTOR_RECUR_EXTRACT]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
42; CHECK-NEXT:    br label %[[LOOP:.*]]
43; CHECK:       [[LOOP]]:
44; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
45; CHECK-NEXT:    [[IV_I65:%.*]] = phi i65 [ [[BC_RESUME_VAL2]], %[[SCALAR_PH]] ], [ [[IV_I65_NEXT:%.*]], %[[LOOP]] ]
46; CHECK-NEXT:    [[FOR:%.*]] = phi i64 [ [[SCALAR_RECUR_INIT]], %[[SCALAR_PH]] ], [ [[TRUNC:%.*]], %[[LOOP]] ]
47; CHECK-NEXT:    [[TRUNC]] = trunc i65 [[IV_I65]] to i64
48; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds i64, ptr [[DST]], i64 [[TRUNC]]
49; CHECK-NEXT:    store i64 [[FOR]], ptr [[GEP]], align 8
50; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
51; CHECK-NEXT:    [[ICMP:%.*]] = icmp eq i64 [[IV_NEXT]], 1000
52; CHECK-NEXT:    [[IV_I65_NEXT]] = add i65 [[IV_I65]], -1
53; CHECK-NEXT:    br i1 [[ICMP]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
54; CHECK:       [[EXIT]]:
55; CHECK-NEXT:    ret void
56;
57entry:
58  br label %loop
59
60loop:
61  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
62  %iv.i65 = phi i65 [ 0, %entry ], [ %iv.i65.next, %loop ]
63  %for = phi i64 [ 0, %entry ], [ %trunc, %loop ]
64  %trunc = trunc i65 %iv.i65 to i64
65  %gep = getelementptr inbounds i64, ptr %dst, i64 %trunc
66  store i64 %for, ptr %gep, align 8
67  %iv.next = add i64 %iv, 1
68  %icmp = icmp eq i64 %iv.next, 1000
69  %iv.i65.next = add i65 %iv.i65, -1
70  br i1 %icmp, label %exit, label %loop
71
72exit:
73  ret void
74}
75;.
76; CHECK: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}
77; CHECK: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}
78; CHECK: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}
79; CHECK: [[LOOP3]] = distinct !{[[LOOP3]], [[META2]], [[META1]]}
80;.
81