1 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2 ; RUN: opt -loop-reduce -S < %s | FileCheck %s
5 ; LSR should properly handle the post-inc offset when folding the
6 ; non-IV operand of an icmp into the IV.
8 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64"
10 %struct.Vector2 = type { i16*, [64 x i16], i32 }
12 @.str = private unnamed_addr constant [37 x i8] c"0123456789abcdefghijklmnopqrstuvwxyz\00"
14 define void @_Z15IntegerToStringjjR7Vector2(i32 %i, i32 %radix, %struct.Vector2* nocapture %result) nounwind noinline {
15 ; CHECK-LABEL: @_Z15IntegerToStringjjR7Vector2(
17 ; CHECK-NEXT: [[BUFFER:%.*]] = alloca [33 x i16], align 16
18 ; CHECK-NEXT: [[ADD_PTR:%.*]] = getelementptr inbounds [33 x i16], [33 x i16]* [[BUFFER]], i64 0, i64 33
19 ; CHECK-NEXT: [[SUB_PTR_LHS_CAST:%.*]] = ptrtoint i16* [[ADD_PTR]] to i64
20 ; CHECK-NEXT: [[SUB_PTR_RHS_CAST:%.*]] = ptrtoint i16* [[ADD_PTR]] to i64
21 ; CHECK-NEXT: [[SCEVGEP4:%.*]] = getelementptr [33 x i16], [33 x i16]* [[BUFFER]], i64 0, i64 32
22 ; CHECK-NEXT: [[SCEVGEP45:%.*]] = bitcast i16* [[SCEVGEP4]] to [33 x i16]*
23 ; CHECK-NEXT: [[SCEVGEP11:%.*]] = getelementptr [33 x i16], [33 x i16]* [[BUFFER]], i64 1, i64 0
24 ; CHECK-NEXT: [[SCEVGEP1112:%.*]] = bitcast i16* [[SCEVGEP11]] to [33 x i16]*
25 ; CHECK-NEXT: br label [[DO_BODY:%.*]]
27 ; CHECK-NEXT: [[LSR_IV15:%.*]] = phi i64 [ [[LSR_IV_NEXT16:%.*]], [[DO_BODY]] ], [ -1, [[ENTRY:%.*]] ]
28 ; CHECK-NEXT: [[LSR_IV13:%.*]] = phi [33 x i16]* [ [[TMP2:%.*]], [[DO_BODY]] ], [ [[SCEVGEP1112]], [[ENTRY]] ]
29 ; CHECK-NEXT: [[LSR_IV6:%.*]] = phi [33 x i16]* [ [[TMP1:%.*]], [[DO_BODY]] ], [ [[SCEVGEP45]], [[ENTRY]] ]
30 ; CHECK-NEXT: [[I_ADDR_0:%.*]] = phi i32 [ [[DIV:%.*]], [[DO_BODY]] ], [ [[I:%.*]], [[ENTRY]] ]
31 ; CHECK-NEXT: [[LSR_IV617:%.*]] = bitcast [33 x i16]* [[LSR_IV6]] to i16*
32 ; CHECK-NEXT: [[REM:%.*]] = urem i32 [[I_ADDR_0]], 10
33 ; CHECK-NEXT: [[DIV]] = udiv i32 [[I_ADDR_0]], 10
34 ; CHECK-NEXT: [[IDXPROM:%.*]] = zext i32 [[REM]] to i64
35 ; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [37 x i8], [37 x i8]* @.str, i64 0, i64 [[IDXPROM]]
36 ; CHECK-NEXT: [[TMP5:%.*]] = load i8, i8* [[ARRAYIDX]], align 1
37 ; CHECK-NEXT: [[CONV:%.*]] = sext i8 [[TMP5]] to i16
38 ; CHECK-NEXT: store i16 [[CONV]], i16* [[LSR_IV617]], align 2
39 ; CHECK-NEXT: [[TMP0:%.*]] = icmp ugt i32 [[I_ADDR_0]], 9
40 ; CHECK-NEXT: [[SCEVGEP7:%.*]] = getelementptr [33 x i16], [33 x i16]* [[LSR_IV6]], i64 0, i64 -1
41 ; CHECK-NEXT: [[TMP1]] = bitcast i16* [[SCEVGEP7]] to [33 x i16]*
42 ; CHECK-NEXT: [[SCEVGEP14:%.*]] = getelementptr [33 x i16], [33 x i16]* [[LSR_IV13]], i64 0, i64 -1
43 ; CHECK-NEXT: [[TMP2]] = bitcast i16* [[SCEVGEP14]] to [33 x i16]*
44 ; CHECK-NEXT: [[LSR_IV_NEXT16]] = add i64 [[LSR_IV15]], 1
45 ; CHECK-NEXT: br i1 [[TMP0]], label [[DO_BODY]], label [[DO_END:%.*]]
47 ; CHECK-NEXT: [[XAP_0:%.*]] = inttoptr i64 [[LSR_IV_NEXT16]] to i1*
48 ; CHECK-NEXT: [[CAP_0:%.*]] = ptrtoint i1* [[XAP_0]] to i64
49 ; CHECK-NEXT: [[SUB_PTR_SUB:%.*]] = sub i64 [[SUB_PTR_LHS_CAST]], [[SUB_PTR_RHS_CAST]]
50 ; CHECK-NEXT: [[SUB_PTR_DIV39:%.*]] = lshr exact i64 [[SUB_PTR_SUB]], 1
51 ; CHECK-NEXT: [[CONV11:%.*]] = trunc i64 [[SUB_PTR_DIV39]] to i32
52 ; CHECK-NEXT: [[MLENGTH:%.*]] = getelementptr inbounds [[STRUCT_VECTOR2:%.*]], %struct.Vector2* [[RESULT:%.*]], i64 0, i32 2
53 ; CHECK-NEXT: [[IDX_EXT21:%.*]] = bitcast i64 [[SUB_PTR_DIV39]] to i64
54 ; CHECK-NEXT: [[CMP2740:%.*]] = icmp eq i64 [[IDX_EXT21]], 0
55 ; CHECK-NEXT: br i1 [[CMP2740]], label [[FOR_END:%.*]], label [[FOR_BODY_LR_PH:%.*]]
56 ; CHECK: for.body.lr.ph:
57 ; CHECK-NEXT: [[TMP16:%.*]] = load i32, i32* [[MLENGTH]], align 4
58 ; CHECK-NEXT: [[MBEGIN:%.*]] = getelementptr inbounds [[STRUCT_VECTOR2]], %struct.Vector2* [[RESULT]], i64 0, i32 0
59 ; CHECK-NEXT: [[TMP14:%.*]] = load i16*, i16** [[MBEGIN]], align 8
60 ; CHECK-NEXT: [[TMP48:%.*]] = zext i32 [[TMP16]] to i64
61 ; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr i16, i16* [[TMP14]], i64 [[TMP48]]
62 ; CHECK-NEXT: [[SCEVGEP1:%.*]] = bitcast i16* [[SCEVGEP]] to i8*
63 ; CHECK-NEXT: br label [[FOR_BODY:%.*]]
65 ; CHECK-NEXT: [[LSR_IV8:%.*]] = phi [33 x i16]* [ [[TMP3:%.*]], [[FOR_BODY]] ], [ [[TMP2]], [[FOR_BODY_LR_PH]] ]
66 ; CHECK-NEXT: [[LSR_IV:%.*]] = phi i64 [ [[LSR_IV_NEXT:%.*]], [[FOR_BODY]] ], [ 0, [[FOR_BODY_LR_PH]] ]
67 ; CHECK-NEXT: [[LSR_IV810:%.*]] = bitcast [33 x i16]* [[LSR_IV8]] to i16*
68 ; CHECK-NEXT: [[UGLYGEP:%.*]] = getelementptr i8, i8* [[SCEVGEP1]], i64 [[LSR_IV]]
69 ; CHECK-NEXT: [[UGLYGEP2:%.*]] = bitcast i8* [[UGLYGEP]] to i16*
70 ; CHECK-NEXT: [[TMP29:%.*]] = load i16, i16* [[LSR_IV810]], align 2
71 ; CHECK-NEXT: store i16 [[TMP29]], i16* [[UGLYGEP2]], align 2
72 ; CHECK-NEXT: [[LSR_IV_NEXT]] = add i64 [[LSR_IV]], 2
73 ; CHECK-NEXT: [[LSR_IV_NEXT3:%.*]] = inttoptr i64 [[LSR_IV_NEXT]] to i16*
74 ; CHECK-NEXT: [[SCEVGEP9:%.*]] = getelementptr [33 x i16], [33 x i16]* [[LSR_IV8]], i64 0, i64 1
75 ; CHECK-NEXT: [[TMP3]] = bitcast i16* [[SCEVGEP9]] to [33 x i16]*
76 ; CHECK-NEXT: [[CMP27:%.*]] = icmp eq i16* [[LSR_IV_NEXT3]], null
77 ; CHECK-NEXT: br i1 [[CMP27]], label [[FOR_END_LOOPEXIT:%.*]], label [[FOR_BODY]]
78 ; CHECK: for.end.loopexit:
79 ; CHECK-NEXT: br label [[FOR_END]]
81 ; CHECK-NEXT: [[TMP38:%.*]] = load i32, i32* [[MLENGTH]], align 4
82 ; CHECK-NEXT: [[ADD:%.*]] = add i32 [[TMP38]], [[CONV11]]
83 ; CHECK-NEXT: store i32 [[ADD]], i32* [[MLENGTH]], align 4
84 ; CHECK-NEXT: ret void
87 %buffer = alloca [33 x i16], align 16
88 %add.ptr = getelementptr inbounds [33 x i16], [33 x i16]* %buffer, i64 0, i64 33
89 %sub.ptr.lhs.cast = ptrtoint i16* %add.ptr to i64
90 %sub.ptr.rhs.cast = ptrtoint i16* %add.ptr to i64
93 do.body: ; preds = %do.body, %entry
94 %0 = phi i64 [ %indvar.next44, %do.body ], [ 0, %entry ]
95 %i.addr.0 = phi i32 [ %div, %do.body ], [ %i, %entry ]
96 %tmp51 = sub i64 32, %0
97 %incdec.ptr = getelementptr [33 x i16], [33 x i16]* %buffer, i64 0, i64 %tmp51
98 %rem = urem i32 %i.addr.0, 10
99 %div = udiv i32 %i.addr.0, 10
100 %idxprom = zext i32 %rem to i64
101 %arrayidx = getelementptr inbounds [37 x i8], [37 x i8]* @.str, i64 0, i64 %idxprom
102 %tmp5 = load i8, i8* %arrayidx, align 1
103 %conv = sext i8 %tmp5 to i16
104 store i16 %conv, i16* %incdec.ptr, align 2
105 %1 = icmp ugt i32 %i.addr.0, 9
106 %indvar.next44 = add i64 %0, 1
107 br i1 %1, label %do.body, label %do.end
109 do.end: ; preds = %do.body
110 %xap.0 = inttoptr i64 %0 to i1*
111 %cap.0 = ptrtoint i1* %xap.0 to i64
112 %sub.ptr.sub = sub i64 %sub.ptr.lhs.cast, %sub.ptr.rhs.cast
113 %sub.ptr.div39 = lshr exact i64 %sub.ptr.sub, 1
114 %conv11 = trunc i64 %sub.ptr.div39 to i32
115 %mLength = getelementptr inbounds %struct.Vector2, %struct.Vector2* %result, i64 0, i32 2
116 %idx.ext21 = bitcast i64 %sub.ptr.div39 to i64
117 %incdec.ptr.sum = add i64 %idx.ext21, -1
118 %cp.0.sum = sub i64 %incdec.ptr.sum, %0
119 %add.ptr22 = getelementptr [33 x i16], [33 x i16]* %buffer, i64 1, i64 %cp.0.sum
120 %cmp2740 = icmp eq i64 %idx.ext21, 0
121 br i1 %cmp2740, label %for.end, label %for.body.lr.ph
123 for.body.lr.ph: ; preds = %do.end
124 %tmp16 = load i32, i32* %mLength, align 4
125 %mBegin = getelementptr inbounds %struct.Vector2, %struct.Vector2* %result, i64 0, i32 0
126 %tmp14 = load i16*, i16** %mBegin, align 8
127 %tmp48 = zext i32 %tmp16 to i64
130 for.body: ; preds = %for.body, %for.body.lr.ph
131 %indvar = phi i64 [ 0, %for.body.lr.ph ], [ %indvar.next, %for.body ]
132 %tmp46 = add i64 %tmp51, %indvar
133 %p.042 = getelementptr [33 x i16], [33 x i16]* %buffer, i64 0, i64 %tmp46
134 %tmp47 = sub i64 %indvar, %0
135 %incdec.ptr32 = getelementptr [33 x i16], [33 x i16]* %buffer, i64 1, i64 %tmp47
136 %tmp49 = add i64 %tmp48, %indvar
137 %dst.041 = getelementptr i16, i16* %tmp14, i64 %tmp49
138 %tmp29 = load i16, i16* %p.042, align 2
139 store i16 %tmp29, i16* %dst.041, align 2
140 %cmp27 = icmp eq i16* %incdec.ptr32, %add.ptr22
141 %indvar.next = add i64 %indvar, 1
142 br i1 %cmp27, label %for.end.loopexit, label %for.body
144 for.end.loopexit: ; preds = %for.body
147 for.end: ; preds = %for.end.loopexit, %do.end
148 %tmp38 = load i32, i32* %mLength, align 4
149 %add = add i32 %tmp38, %conv11
150 store i32 %add, i32* %mLength, align 4