xref: /llvm-project/llvm/test/Transforms/LoopVectorize/pr33706.ll (revision 29441e4f5fa5f5c7709f7cf180815ba97f611297)
1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --prefix-filecheck-ir-name VAR_ --version 2
2; RUN: opt -S -passes=loop-vectorize -force-vector-interleave=1 -force-vector-width=2 < %s | FileCheck %s
3
4@global = local_unnamed_addr global i32 0, align 4
5@global.1 = local_unnamed_addr global i32 0, align 4
6@global.2 = local_unnamed_addr global float 0x3EF0000000000000, align 4
7
8define void @PR33706(ptr nocapture readonly %arg, ptr nocapture %arg1, i32 %arg2) local_unnamed_addr {
9; CHECK-LABEL: define void @PR33706
10; CHECK-SAME: (ptr readonly captures(none) [[ARG:%.*]], ptr captures(none) [[ARG1:%.*]], i32 [[ARG2:%.*]]) local_unnamed_addr {
11; CHECK-NEXT:  bb:
12; CHECK-NEXT:    [[TMP:%.*]] = load i32, ptr @global.1, align 4
13; CHECK-NEXT:    [[VAR_TMP3:%.*]] = getelementptr inbounds float, ptr [[ARG]], i64 190
14; CHECK-NEXT:    [[VAR_TMP4:%.*]] = getelementptr inbounds float, ptr [[ARG1]], i64 512
15; CHECK-NEXT:    [[VAR_TMP5:%.*]] = and i32 [[TMP]], 65535
16; CHECK-NEXT:    [[VAR_TMP6:%.*]] = icmp ugt i32 [[ARG2]], 65536
17; CHECK-NEXT:    br i1 [[VAR_TMP6]], label [[BB7:%.*]], label [[BB9:%.*]]
18; CHECK:       bb7:
19; CHECK-NEXT:    [[VAR_TMP8:%.*]] = load i32, ptr @global, align 4
20; CHECK-NEXT:    br label [[BB27:%.*]]
21; CHECK:       bb9:
22; CHECK-NEXT:    [[VAR_TMP10:%.*]] = udiv i32 65536, [[ARG2]]
23; CHECK-NEXT:    br label [[BB11:%.*]]
24; CHECK:       bb11:
25; CHECK-NEXT:    [[VAR_TMP12:%.*]] = phi i32 [ [[VAR_TMP20:%.*]], [[BB11]] ], [ [[VAR_TMP5]], [[BB9]] ]
26; CHECK-NEXT:    [[VAR_TMP13:%.*]] = phi ptr [ [[VAR_TMP18:%.*]], [[BB11]] ], [ [[VAR_TMP4]], [[BB9]] ]
27; CHECK-NEXT:    [[VAR_TMP14:%.*]] = phi i32 [ [[VAR_TMP16:%.*]], [[BB11]] ], [ [[VAR_TMP10]], [[BB9]] ]
28; CHECK-NEXT:    [[VAR_TMP15:%.*]] = phi i32 [ [[VAR_TMP19:%.*]], [[BB11]] ], [ [[TMP]], [[BB9]] ]
29; CHECK-NEXT:    [[VAR_TMP16]] = add nsw i32 [[VAR_TMP14]], -1
30; CHECK-NEXT:    [[VAR_TMP17:%.*]] = sitofp i32 [[VAR_TMP12]] to float
31; CHECK-NEXT:    store float [[VAR_TMP17]], ptr [[VAR_TMP13]], align 4
32; CHECK-NEXT:    [[VAR_TMP18]] = getelementptr inbounds float, ptr [[VAR_TMP13]], i64 1
33; CHECK-NEXT:    [[VAR_TMP19]] = add i32 [[VAR_TMP15]], [[ARG2]]
34; CHECK-NEXT:    [[VAR_TMP20]] = and i32 [[VAR_TMP19]], 65535
35; CHECK-NEXT:    [[VAR_TMP21:%.*]] = icmp eq i32 [[VAR_TMP16]], 0
36; CHECK-NEXT:    br i1 [[VAR_TMP21]], label [[BB22:%.*]], label [[BB11]]
37; CHECK:       bb22:
38; CHECK-NEXT:    [[VAR_TMP23:%.*]] = phi ptr [ [[VAR_TMP18]], [[BB11]] ]
39; CHECK-NEXT:    [[VAR_TMP24:%.*]] = phi i32 [ [[VAR_TMP19]], [[BB11]] ]
40; CHECK-NEXT:    [[VAR_TMP25:%.*]] = phi i32 [ [[VAR_TMP20]], [[BB11]] ]
41; CHECK-NEXT:    [[VAR_TMP26:%.*]] = ashr i32 [[VAR_TMP24]], 16
42; CHECK-NEXT:    store i32 [[VAR_TMP26]], ptr @global, align 4
43; CHECK-NEXT:    br label [[BB27]]
44; CHECK:       bb27:
45; CHECK-NEXT:    [[VAR_TMP28:%.*]] = phi i32 [ [[VAR_TMP26]], [[BB22]] ], [ [[VAR_TMP8]], [[BB7]] ]
46; CHECK-NEXT:    [[VAR_TMP29:%.*]] = phi ptr [ [[VAR_TMP23]], [[BB22]] ], [ [[VAR_TMP4]], [[BB7]] ]
47; CHECK-NEXT:    [[VAR_TMP30:%.*]] = phi i32 [ [[VAR_TMP25]], [[BB22]] ], [ [[VAR_TMP5]], [[BB7]] ]
48; CHECK-NEXT:    [[VAR_TMP31:%.*]] = sext i32 [[VAR_TMP28]] to i64
49; CHECK-NEXT:    [[VAR_TMP32:%.*]] = getelementptr inbounds float, ptr [[VAR_TMP3]], i64 [[VAR_TMP31]]
50; CHECK-NEXT:    [[VAR_TMP33:%.*]] = load float, ptr [[VAR_TMP32]], align 4
51; CHECK-NEXT:    [[VAR_TMP34:%.*]] = sitofp i32 [[VAR_TMP30]] to float
52; CHECK-NEXT:    [[VAR_TMP35:%.*]] = load float, ptr @global.2, align 4
53; CHECK-NEXT:    [[VAR_TMP36:%.*]] = fmul float [[VAR_TMP35]], [[VAR_TMP34]]
54; CHECK-NEXT:    [[VAR_TMP37:%.*]] = fadd float [[VAR_TMP33]], [[VAR_TMP36]]
55; CHECK-NEXT:    store float [[VAR_TMP37]], ptr [[VAR_TMP29]], align 4
56; CHECK-NEXT:    ret void
57;
58bb:
59  %tmp = load i32, ptr @global.1, align 4
60  %tmp3 = getelementptr inbounds float, ptr %arg, i64 190
61  %tmp4 = getelementptr inbounds float, ptr %arg1, i64 512
62  %tmp5 = and i32 %tmp, 65535
63  %tmp6 = icmp ugt i32 %arg2, 65536
64  br i1 %tmp6, label %bb7, label %bb9
65
66bb7:                                              ; preds = %bb
67  %tmp8 = load i32, ptr @global, align 4
68  br label %bb27
69
70bb9:                                              ; preds = %bb
71  %tmp10 = udiv i32 65536, %arg2
72  br label %bb11
73
74bb11:                                             ; preds = %bb11, %bb9
75  %tmp12 = phi i32 [ %tmp20, %bb11 ], [ %tmp5, %bb9 ]
76  %tmp13 = phi ptr [ %tmp18, %bb11 ], [ %tmp4, %bb9 ]
77  %tmp14 = phi i32 [ %tmp16, %bb11 ], [ %tmp10, %bb9 ]
78  %tmp15 = phi i32 [ %tmp19, %bb11 ], [ %tmp, %bb9 ]
79  %tmp16 = add nsw i32 %tmp14, -1
80  %tmp17 = sitofp i32 %tmp12 to float
81  store float %tmp17, ptr %tmp13, align 4
82  %tmp18 = getelementptr inbounds float, ptr %tmp13, i64 1
83  %tmp19 = add i32 %tmp15, %arg2
84  %tmp20 = and i32 %tmp19, 65535
85  %tmp21 = icmp eq i32 %tmp16, 0
86  br i1 %tmp21, label %bb22, label %bb11
87
88bb22:                                             ; preds = %bb11
89  %tmp23 = phi ptr [ %tmp18, %bb11 ]
90  %tmp24 = phi i32 [ %tmp19, %bb11 ]
91  %tmp25 = phi i32 [ %tmp20, %bb11 ]
92  %tmp26 = ashr i32 %tmp24, 16
93  store i32 %tmp26, ptr @global, align 4
94  br label %bb27
95
96bb27:                                             ; preds = %bb22, %bb7
97  %tmp28 = phi i32 [ %tmp26, %bb22 ], [ %tmp8, %bb7 ]
98  %tmp29 = phi ptr [ %tmp23, %bb22 ], [ %tmp4, %bb7 ]
99  %tmp30 = phi i32 [ %tmp25, %bb22 ], [ %tmp5, %bb7 ]
100  %tmp31 = sext i32 %tmp28 to i64
101  %tmp32 = getelementptr inbounds float, ptr %tmp3, i64 %tmp31
102  %tmp33 = load float, ptr %tmp32, align 4
103  %tmp34 = sitofp i32 %tmp30 to float
104  %tmp35 = load float, ptr @global.2, align 4
105  %tmp36 = fmul float %tmp35, %tmp34
106  %tmp37 = fadd float %tmp33, %tmp36
107  store float %tmp37, ptr %tmp29, align 4
108  ret void
109}
110