1 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2 ; RUN: opt -passes=loop-vectorize -force-vector-width=16 -force-vector-interleave=2 < %s -S -o - | FileCheck %s
4 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
5 target triple = "x86_64-pc_linux"
7 define void @firstorderrec(ptr nocapture noundef readonly %x, ptr noalias nocapture noundef writeonly %y, i32 noundef %n) {
8 ; CHECK-LABEL: @firstorderrec(
10 ; CHECK-NEXT: [[CMP18:%.*]] = icmp sgt i32 [[N:%.*]], 1
11 ; CHECK-NEXT: br i1 [[CMP18]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_COND_CLEANUP:%.*]]
12 ; CHECK: for.body.preheader:
13 ; CHECK-NEXT: [[WIDE_TRIP_COUNT:%.*]] = zext i32 [[N]] to i64
14 ; CHECK-NEXT: [[DOTPRE:%.*]] = load i8, ptr [[X:%.*]], align 1
15 ; CHECK-NEXT: [[TMP0:%.*]] = add nsw i64 [[WIDE_TRIP_COUNT]], -1
16 ; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP0]], 32
17 ; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
19 ; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP0]], 32
20 ; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP0]], [[N_MOD_VF]]
21 ; CHECK-NEXT: [[IND_END:%.*]] = add i64 1, [[N_VEC]]
22 ; CHECK-NEXT: [[VECTOR_RECUR_INIT:%.*]] = insertelement <16 x i8> poison, i8 [[DOTPRE]], i32 15
23 ; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
25 ; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
26 ; CHECK-NEXT: [[VECTOR_RECUR:%.*]] = phi <16 x i8> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD1:%.*]], [[VECTOR_BODY]] ]
27 ; CHECK-NEXT: [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
28 ; CHECK-NEXT: [[TMP1:%.*]] = add i64 [[OFFSET_IDX]], 0
29 ; CHECK-NEXT: [[TMP2:%.*]] = add i64 [[OFFSET_IDX]], 16
30 ; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[X]], i64 [[TMP1]]
31 ; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[X]], i64 [[TMP2]]
32 ; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
33 ; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 16
34 ; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
35 ; CHECK-NEXT: [[WIDE_LOAD1]] = load <16 x i8>, ptr [[TMP6]], align 1
36 ; CHECK-NEXT: [[TMP7:%.*]] = shufflevector <16 x i8> [[VECTOR_RECUR]], <16 x i8> [[WIDE_LOAD]], <16 x i32> <i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30>
37 ; CHECK-NEXT: [[TMP8:%.*]] = shufflevector <16 x i8> [[WIDE_LOAD]], <16 x i8> [[WIDE_LOAD1]], <16 x i32> <i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30>
38 ; CHECK-NEXT: [[TMP9:%.*]] = add <16 x i8> [[WIDE_LOAD]], [[TMP7]]
39 ; CHECK-NEXT: [[TMP10:%.*]] = add <16 x i8> [[WIDE_LOAD1]], [[TMP8]]
40 ; CHECK-NEXT: [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[Y:%.*]], i64 [[TMP1]]
41 ; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[Y]], i64 [[TMP2]]
42 ; CHECK-NEXT: [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i32 0
43 ; CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i32 16
44 ; CHECK-NEXT: store <16 x i8> [[TMP9]], ptr [[TMP13]], align 1
45 ; CHECK-NEXT: store <16 x i8> [[TMP10]], ptr [[TMP14]], align 1
46 ; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 32
47 ; CHECK-NEXT: [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
48 ; CHECK-NEXT: br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
49 ; CHECK: middle.block:
50 ; CHECK-NEXT: [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <16 x i8> [[WIDE_LOAD1]], i32 15
51 ; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC]]
52 ; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_COND_CLEANUP_LOOPEXIT:%.*]], label [[SCALAR_PH]]
54 ; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 1, [[FOR_BODY_PREHEADER]] ]
55 ; CHECK-NEXT: [[SCALAR_RECUR_INIT:%.*]] = phi i8 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ [[DOTPRE]], [[FOR_BODY_PREHEADER]] ]
56 ; CHECK-NEXT: br label [[FOR_BODY:%.*]]
57 ; CHECK: for.cond.cleanup.loopexit:
58 ; CHECK-NEXT: br label [[FOR_COND_CLEANUP]]
59 ; CHECK: for.cond.cleanup:
60 ; CHECK-NEXT: ret void
62 ; CHECK-NEXT: [[SCALAR_RECUR:%.*]] = phi i8 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP16:%.*]], [[FOR_BODY]] ]
63 ; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
64 ; CHECK-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds i8, ptr [[X]], i64 [[INDVARS_IV]]
65 ; CHECK-NEXT: [[TMP16]] = load i8, ptr [[ARRAYIDX4]], align 1
66 ; CHECK-NEXT: [[ADD7:%.*]] = add i8 [[TMP16]], [[SCALAR_RECUR]]
67 ; CHECK-NEXT: [[ARRAYIDX10:%.*]] = getelementptr inbounds i8, ptr [[Y]], i64 [[INDVARS_IV]]
68 ; CHECK-NEXT: store i8 [[ADD7]], ptr [[ARRAYIDX10]], align 1
69 ; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
70 ; CHECK-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[WIDE_TRIP_COUNT]]
71 ; CHECK-NEXT: br i1 [[EXITCOND_NOT]], label [[FOR_COND_CLEANUP_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
74 %cmp18 = icmp sgt i32 %n, 1
75 br i1 %cmp18, label %for.body.preheader, label %for.cond.cleanup
77 for.body.preheader: ; preds = %entry
78 %wide.trip.count = zext i32 %n to i64
79 %.pre = load i8, ptr %x, align 1
82 for.cond.cleanup: ; preds = %for.body, %entry
85 for.body: ; preds = %for.body.preheader, %for.body
86 %0 = phi i8 [ %.pre, %for.body.preheader ], [ %1, %for.body ]
87 %indvars.iv = phi i64 [ 1, %for.body.preheader ], [ %indvars.iv.next, %for.body ]
88 %arrayidx4 = getelementptr inbounds i8, ptr %x, i64 %indvars.iv
89 %1 = load i8, ptr %arrayidx4, align 1
91 %arrayidx10 = getelementptr inbounds i8, ptr %y, i64 %indvars.iv
92 store i8 %add7, ptr %arrayidx10, align 1
93 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
94 %exitcond.not = icmp eq i64 %indvars.iv.next, %wide.trip.count
95 br i1 %exitcond.not, label %for.cond.cleanup, label %for.body
98 define void @thirdorderrec(ptr nocapture noundef readonly %x, ptr noalias nocapture noundef writeonly %y, i32 noundef %n) #0 {
99 ; CHECK-LABEL: @thirdorderrec(
101 ; CHECK-NEXT: [[CMP38:%.*]] = icmp sgt i32 [[N:%.*]], 3
102 ; CHECK-NEXT: br i1 [[CMP38]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_COND_CLEANUP:%.*]]
103 ; CHECK: for.body.preheader:
104 ; CHECK-NEXT: [[WIDE_TRIP_COUNT:%.*]] = zext i32 [[N]] to i64
105 ; CHECK-NEXT: [[DOTPRE:%.*]] = load i8, ptr [[X:%.*]], align 1
106 ; CHECK-NEXT: [[ARRAYIDX5_PHI_TRANS_INSERT:%.*]] = getelementptr inbounds i8, ptr [[X]], i64 1
107 ; CHECK-NEXT: [[DOTPRE44:%.*]] = load i8, ptr [[ARRAYIDX5_PHI_TRANS_INSERT]], align 1
108 ; CHECK-NEXT: [[ARRAYIDX12_PHI_TRANS_INSERT:%.*]] = getelementptr inbounds i8, ptr [[X]], i64 2
109 ; CHECK-NEXT: [[DOTPRE45:%.*]] = load i8, ptr [[ARRAYIDX12_PHI_TRANS_INSERT]], align 1
110 ; CHECK-NEXT: [[TMP0:%.*]] = add nsw i64 [[WIDE_TRIP_COUNT]], -3
111 ; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP0]], 32
112 ; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
114 ; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP0]], 32
115 ; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP0]], [[N_MOD_VF]]
116 ; CHECK-NEXT: [[IND_END:%.*]] = add i64 3, [[N_VEC]]
117 ; CHECK-NEXT: [[VECTOR_RECUR_INIT:%.*]] = insertelement <16 x i8> poison, i8 [[DOTPRE45]], i32 15
118 ; CHECK-NEXT: [[VECTOR_RECUR_INIT1:%.*]] = insertelement <16 x i8> poison, i8 [[DOTPRE44]], i32 15
119 ; CHECK-NEXT: [[VECTOR_RECUR_INIT3:%.*]] = insertelement <16 x i8> poison, i8 [[DOTPRE]], i32 15
120 ; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
121 ; CHECK: vector.body:
122 ; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
123 ; CHECK-NEXT: [[VECTOR_RECUR:%.*]] = phi <16 x i8> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD5:%.*]], [[VECTOR_BODY]] ]
124 ; CHECK-NEXT: [[VECTOR_RECUR2:%.*]] = phi <16 x i8> [ [[VECTOR_RECUR_INIT1]], [[VECTOR_PH]] ], [ [[TMP8:%.*]], [[VECTOR_BODY]] ]
125 ; CHECK-NEXT: [[VECTOR_RECUR4:%.*]] = phi <16 x i8> [ [[VECTOR_RECUR_INIT3]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
126 ; CHECK-NEXT: [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX]]
127 ; CHECK-NEXT: [[TMP1:%.*]] = add i64 [[OFFSET_IDX]], 0
128 ; CHECK-NEXT: [[TMP2:%.*]] = add i64 [[OFFSET_IDX]], 16
129 ; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[X]], i64 [[TMP1]]
130 ; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[X]], i64 [[TMP2]]
131 ; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
132 ; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 16
133 ; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
134 ; CHECK-NEXT: [[WIDE_LOAD5]] = load <16 x i8>, ptr [[TMP6]], align 1
135 ; CHECK-NEXT: [[TMP7:%.*]] = shufflevector <16 x i8> [[VECTOR_RECUR]], <16 x i8> [[WIDE_LOAD]], <16 x i32> <i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30>
136 ; CHECK-NEXT: [[TMP8]] = shufflevector <16 x i8> [[WIDE_LOAD]], <16 x i8> [[WIDE_LOAD5]], <16 x i32> <i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30>
137 ; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <16 x i8> [[VECTOR_RECUR2]], <16 x i8> [[TMP7]], <16 x i32> <i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30>
138 ; CHECK-NEXT: [[TMP10]] = shufflevector <16 x i8> [[TMP7]], <16 x i8> [[TMP8]], <16 x i32> <i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30>
139 ; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <16 x i8> [[VECTOR_RECUR4]], <16 x i8> [[TMP9]], <16 x i32> <i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30>
140 ; CHECK-NEXT: [[TMP12:%.*]] = shufflevector <16 x i8> [[TMP9]], <16 x i8> [[TMP10]], <16 x i32> <i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30>
141 ; CHECK-NEXT: [[TMP13:%.*]] = add <16 x i8> [[TMP9]], [[TMP11]]
142 ; CHECK-NEXT: [[TMP14:%.*]] = add <16 x i8> [[TMP10]], [[TMP12]]
143 ; CHECK-NEXT: [[TMP15:%.*]] = add <16 x i8> [[TMP13]], [[TMP7]]
144 ; CHECK-NEXT: [[TMP16:%.*]] = add <16 x i8> [[TMP14]], [[TMP8]]
145 ; CHECK-NEXT: [[TMP17:%.*]] = add <16 x i8> [[TMP15]], [[WIDE_LOAD]]
146 ; CHECK-NEXT: [[TMP18:%.*]] = add <16 x i8> [[TMP16]], [[WIDE_LOAD5]]
147 ; CHECK-NEXT: [[TMP19:%.*]] = getelementptr inbounds i8, ptr [[Y:%.*]], i64 [[TMP1]]
148 ; CHECK-NEXT: [[TMP20:%.*]] = getelementptr inbounds i8, ptr [[Y]], i64 [[TMP2]]
149 ; CHECK-NEXT: [[TMP21:%.*]] = getelementptr inbounds i8, ptr [[TMP19]], i32 0
150 ; CHECK-NEXT: [[TMP22:%.*]] = getelementptr inbounds i8, ptr [[TMP19]], i32 16
151 ; CHECK-NEXT: store <16 x i8> [[TMP17]], ptr [[TMP21]], align 1
152 ; CHECK-NEXT: store <16 x i8> [[TMP18]], ptr [[TMP22]], align 1
153 ; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 32
154 ; CHECK-NEXT: [[TMP23:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
155 ; CHECK-NEXT: br i1 [[TMP23]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
156 ; CHECK: middle.block:
157 ; CHECK-NEXT: [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <16 x i8> [[WIDE_LOAD5]], i32 15
158 ; CHECK-NEXT: [[VECTOR_RECUR_EXTRACT6:%.*]] = extractelement <16 x i8> [[TMP8]], i32 15
159 ; CHECK-NEXT: [[VECTOR_RECUR_EXTRACT7:%.*]] = extractelement <16 x i8> [[TMP10]], i32 15
160 ; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC]]
161 ; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_COND_CLEANUP_LOOPEXIT:%.*]], label [[SCALAR_PH]]
163 ; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 3, [[FOR_BODY_PREHEADER]] ]
164 ; CHECK-NEXT: [[SCALAR_RECUR_INIT:%.*]] = phi i8 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ [[DOTPRE45]], [[FOR_BODY_PREHEADER]] ]
165 ; CHECK-NEXT: [[SCALAR_RECUR_INIT8:%.*]] = phi i8 [ [[VECTOR_RECUR_EXTRACT6]], [[MIDDLE_BLOCK]] ], [ [[DOTPRE44]], [[FOR_BODY_PREHEADER]] ]
166 ; CHECK-NEXT: [[SCALAR_RECUR_INIT10:%.*]] = phi i8 [ [[VECTOR_RECUR_EXTRACT7]], [[MIDDLE_BLOCK]] ], [ [[DOTPRE]], [[FOR_BODY_PREHEADER]] ]
167 ; CHECK-NEXT: br label [[FOR_BODY:%.*]]
168 ; CHECK: for.cond.cleanup.loopexit:
169 ; CHECK-NEXT: br label [[FOR_COND_CLEANUP]]
170 ; CHECK: for.cond.cleanup:
171 ; CHECK-NEXT: ret void
173 ; CHECK-NEXT: [[SCALAR_RECUR:%.*]] = phi i8 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP24:%.*]], [[FOR_BODY]] ]
174 ; CHECK-NEXT: [[SCALAR_RECUR9:%.*]] = phi i8 [ [[SCALAR_RECUR_INIT8]], [[SCALAR_PH]] ], [ [[SCALAR_RECUR]], [[FOR_BODY]] ]
175 ; CHECK-NEXT: [[SCALAR_RECUR11:%.*]] = phi i8 [ [[SCALAR_RECUR_INIT10]], [[SCALAR_PH]] ], [ [[SCALAR_RECUR9]], [[FOR_BODY]] ]
176 ; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
177 ; CHECK-NEXT: [[ADD8:%.*]] = add i8 [[SCALAR_RECUR9]], [[SCALAR_RECUR11]]
178 ; CHECK-NEXT: [[ADD15:%.*]] = add i8 [[ADD8]], [[SCALAR_RECUR]]
179 ; CHECK-NEXT: [[ARRAYIDX18:%.*]] = getelementptr inbounds i8, ptr [[X]], i64 [[INDVARS_IV]]
180 ; CHECK-NEXT: [[TMP24]] = load i8, ptr [[ARRAYIDX18]], align 1
181 ; CHECK-NEXT: [[ADD21:%.*]] = add i8 [[ADD15]], [[TMP24]]
182 ; CHECK-NEXT: [[ARRAYIDX24:%.*]] = getelementptr inbounds i8, ptr [[Y]], i64 [[INDVARS_IV]]
183 ; CHECK-NEXT: store i8 [[ADD21]], ptr [[ARRAYIDX24]], align 1
184 ; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
185 ; CHECK-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[WIDE_TRIP_COUNT]]
186 ; CHECK-NEXT: br i1 [[EXITCOND_NOT]], label [[FOR_COND_CLEANUP_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
189 %cmp38 = icmp sgt i32 %n, 3
190 br i1 %cmp38, label %for.body.preheader, label %for.cond.cleanup
192 for.body.preheader: ; preds = %entry
193 %wide.trip.count = zext i32 %n to i64
194 %.pre = load i8, ptr %x, align 1
195 %arrayidx5.phi.trans.insert = getelementptr inbounds i8, ptr %x, i64 1
196 %.pre44 = load i8, ptr %arrayidx5.phi.trans.insert, align 1
197 %arrayidx12.phi.trans.insert = getelementptr inbounds i8, ptr %x, i64 2
198 %.pre45 = load i8, ptr %arrayidx12.phi.trans.insert, align 1
201 for.cond.cleanup: ; preds = %for.body, %entry
204 for.body: ; preds = %for.body.preheader, %for.body
205 %0 = phi i8 [ %.pre45, %for.body.preheader ], [ %3, %for.body ]
206 %1 = phi i8 [ %.pre44, %for.body.preheader ], [ %0, %for.body ]
207 %2 = phi i8 [ %.pre, %for.body.preheader ], [ %1, %for.body ]
208 %indvars.iv = phi i64 [ 3, %for.body.preheader ], [ %indvars.iv.next, %for.body ]
209 %add8 = add i8 %1, %2
210 %add15 = add i8 %add8, %0
211 %arrayidx18 = getelementptr inbounds i8, ptr %x, i64 %indvars.iv
212 %3 = load i8, ptr %arrayidx18, align 1
213 %add21 = add i8 %add15, %3
214 %arrayidx24 = getelementptr inbounds i8, ptr %y, i64 %indvars.iv
215 store i8 %add21, ptr %arrayidx24, align 1
216 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
217 %exitcond.not = icmp eq i64 %indvars.iv.next, %wide.trip.count
218 br i1 %exitcond.not, label %for.cond.cleanup, label %for.body
221 define i64 @test_pr62954_scalar_epilogue_required(ptr %A, ptr noalias %B, ptr %C) {
222 ; CHECK-LABEL: @test_pr62954_scalar_epilogue_required(
224 ; CHECK-NEXT: [[GEP:%.*]] = getelementptr i8, ptr [[A:%.*]], i64 872
225 ; CHECK-NEXT: [[REC_START:%.*]] = load i64, ptr [[GEP]], align 8
226 ; CHECK-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
228 ; CHECK-NEXT: [[VECTOR_RECUR_INIT:%.*]] = insertelement <16 x i64> poison, i64 [[REC_START]], i32 15
229 ; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
230 ; CHECK: vector.body:
231 ; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
232 ; CHECK-NEXT: [[VEC_IND:%.*]] = phi <16 x i64> [ <i64 1, i64 3, i64 5, i64 7, i64 9, i64 11, i64 13, i64 15, i64 17, i64 19, i64 21, i64 23, i64 25, i64 27, i64 29, i64 31>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
233 ; CHECK-NEXT: [[VECTOR_RECUR:%.*]] = phi <16 x i64> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[TMP1:%.*]], [[VECTOR_BODY]] ]
234 ; CHECK-NEXT: [[STEP_ADD:%.*]] = add <16 x i64> [[VEC_IND]], <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
235 ; CHECK-NEXT: [[TMP0:%.*]] = sub nsw <16 x i64> zeroinitializer, [[VEC_IND]]
236 ; CHECK-NEXT: [[TMP1]] = sub nsw <16 x i64> zeroinitializer, [[STEP_ADD]]
237 ; CHECK-NEXT: [[TMP4:%.*]] = extractelement <16 x i64> [[TMP1]], i32 15
238 ; CHECK-NEXT: store i64 [[TMP4]], ptr [[GEP]], align 8
239 ; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 32
240 ; CHECK-NEXT: [[VEC_IND_NEXT]] = add <16 x i64> [[STEP_ADD]], <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
241 ; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 32
242 ; CHECK-NEXT: br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
243 ; CHECK: middle.block:
244 ; CHECK-NEXT: [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <16 x i64> [[TMP1]], i32 15
245 ; CHECK-NEXT: br label [[SCALAR_PH]]
247 ; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 65, [[MIDDLE_BLOCK]] ], [ 1, [[ENTRY:%.*]] ]
248 ; CHECK-NEXT: [[SCALAR_RECUR_INIT:%.*]] = phi i64 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ [[REC_START]], [[ENTRY]] ]
249 ; CHECK-NEXT: br label [[LOOP:%.*]]
251 ; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
252 ; CHECK-NEXT: [[SCALAR_RECUR:%.*]] = phi i64 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[NEG_IV:%.*]], [[LOOP]] ]
253 ; CHECK-NEXT: [[GEP_B:%.*]] = getelementptr double, ptr [[B:%.*]], i64 [[IV]]
254 ; CHECK-NEXT: [[L_B:%.*]] = load double, ptr [[GEP_B]], align 8
255 ; CHECK-NEXT: [[NEG_IV]] = sub nsw i64 0, [[IV]]
256 ; CHECK-NEXT: store i64 [[NEG_IV]], ptr [[GEP]], align 8
257 ; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 2
258 ; CHECK-NEXT: [[EC:%.*]] = icmp ugt i64 [[IV]], 74
259 ; CHECK-NEXT: br i1 [[EC]], label [[EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP7:![0-9]+]]
261 ; CHECK-NEXT: [[DOTIN_LCSSA:%.*]] = phi i64 [ [[SCALAR_RECUR]], [[LOOP]] ]
262 ; CHECK-NEXT: [[DOTLCSSA:%.*]] = phi double [ [[L_B]], [[LOOP]] ]
263 ; CHECK-NEXT: store double [[DOTLCSSA]], ptr [[C:%.*]], align 8
264 ; CHECK-NEXT: ret i64 [[DOTIN_LCSSA]]
267 %gep = getelementptr i8, ptr %A, i64 872
268 %rec.start = load i64, ptr %gep, align 8
272 %iv = phi i64 [ 1, %entry ], [ %iv.next, %loop ]
273 %for = phi i64 [ %rec.start, %entry ], [ %neg.iv, %loop ]
274 %gep.B = getelementptr double, ptr %B, i64 %iv
275 %l.B = load double, ptr %gep.B, align 8
276 %neg.iv = sub nsw i64 0, %iv
277 store i64 %neg.iv, ptr %gep, align 8
278 %iv.next = add nuw nsw i64 %iv, 2
279 %ec = icmp ugt i64 %iv, 74
280 br i1 %ec, label %exit, label %loop
283 %.in.lcssa = phi i64 [ %for, %loop ]
284 %.lcssa = phi double [ %l.B, %loop ]
285 store double %.lcssa, ptr %C