1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2; RUN: opt < %s  -passes=loop-vectorize -force-vector-interleave=1 -force-vector-width=4 -S | FileCheck %s
3target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
4
5
6; Function Attrs: nofree norecurse nounwind
7define void @a(i8* readnone %b) {
8; CHECK-LABEL: @a(
9; CHECK-NEXT:  entry:
10; CHECK-NEXT:    [[B1:%.*]] = ptrtoint i8* [[B:%.*]] to i64
11; CHECK-NEXT:    [[CMP_NOT4:%.*]] = icmp eq i8* [[B]], null
12; CHECK-NEXT:    br i1 [[CMP_NOT4]], label [[FOR_COND_CLEANUP:%.*]], label [[FOR_BODY_PREHEADER:%.*]]
13; CHECK:       for.body.preheader:
14; CHECK-NEXT:    [[TMP0:%.*]] = sub i64 0, [[B1]]
15; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP0]], 4
16; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
17; CHECK:       vector.ph:
18; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP0]], 4
19; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP0]], [[N_MOD_VF]]
20; CHECK-NEXT:    [[TMP1:%.*]] = mul i64 [[N_VEC]], -1
21; CHECK-NEXT:    [[IND_END:%.*]] = getelementptr i8, i8* null, i64 [[TMP1]]
22; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
23; CHECK:       vector.body:
24; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE10:%.*]] ]
25; CHECK-NEXT:    [[TMP2:%.*]] = add i64 [[INDEX]], 0
26; CHECK-NEXT:    [[TMP3:%.*]] = mul i64 [[TMP2]], -1
27; CHECK-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, i8* null, i64 [[TMP3]]
28; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, i8* [[NEXT_GEP]], i64 -1
29; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, i8* [[TMP4]], i32 0
30; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8, i8* [[TMP5]], i32 -3
31; CHECK-NEXT:    [[TMP7:%.*]] = bitcast i8* [[TMP6]] to <4 x i8>*
32; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, <4 x i8>* [[TMP7]], align 1
33; CHECK-NEXT:    [[REVERSE:%.*]] = shufflevector <4 x i8> [[WIDE_LOAD]], <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
34; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq <4 x i8> [[REVERSE]], zeroinitializer
35; CHECK-NEXT:    [[TMP9:%.*]] = xor <4 x i1> [[TMP8]], <i1 true, i1 true, i1 true, i1 true>
36; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <4 x i1> [[TMP9]], i32 0
37; CHECK-NEXT:    br i1 [[TMP10]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]]
38; CHECK:       pred.store.if:
39; CHECK-NEXT:    store i8 95, i8* [[TMP4]], align 1
40; CHECK-NEXT:    br label [[PRED_STORE_CONTINUE]]
41; CHECK:       pred.store.continue:
42; CHECK-NEXT:    [[TMP12:%.*]] = extractelement <4 x i1> [[TMP9]], i32 1
43; CHECK-NEXT:    br i1 [[TMP12]], label [[PRED_STORE_IF5:%.*]], label [[PRED_STORE_CONTINUE6:%.*]]
44; CHECK:       pred.store.if5:
45; CHECK-NEXT:    [[TMP13:%.*]] = add i64 [[INDEX]], 1
46; CHECK-NEXT:    [[TMP14:%.*]] = mul i64 [[TMP13]], -1
47; CHECK-NEXT:    [[NEXT_GEP2:%.*]] = getelementptr i8, i8* null, i64 [[TMP14]]
48; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i8, i8* [[NEXT_GEP2]], i64 -1
49; CHECK-NEXT:    store i8 95, i8* [[TMP15]], align 1
50; CHECK-NEXT:    br label [[PRED_STORE_CONTINUE6]]
51; CHECK:       pred.store.continue6:
52; CHECK-NEXT:    [[TMP16:%.*]] = extractelement <4 x i1> [[TMP9]], i32 2
53; CHECK-NEXT:    br i1 [[TMP16]], label [[PRED_STORE_IF7:%.*]], label [[PRED_STORE_CONTINUE8:%.*]]
54; CHECK:       pred.store.if7:
55; CHECK-NEXT:    [[TMP17:%.*]] = add i64 [[INDEX]], 2
56; CHECK-NEXT:    [[TMP18:%.*]] = mul i64 [[TMP17]], -1
57; CHECK-NEXT:    [[NEXT_GEP3:%.*]] = getelementptr i8, i8* null, i64 [[TMP18]]
58; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i8, i8* [[NEXT_GEP3]], i64 -1
59; CHECK-NEXT:    store i8 95, i8* [[TMP19]], align 1
60; CHECK-NEXT:    br label [[PRED_STORE_CONTINUE8]]
61; CHECK:       pred.store.continue8:
62; CHECK-NEXT:    [[TMP20:%.*]] = extractelement <4 x i1> [[TMP9]], i32 3
63; CHECK-NEXT:    br i1 [[TMP20]], label [[PRED_STORE_IF9:%.*]], label [[PRED_STORE_CONTINUE10]]
64; CHECK:       pred.store.if9:
65; CHECK-NEXT:    [[TMP21:%.*]] = add i64 [[INDEX]], 3
66; CHECK-NEXT:    [[TMP22:%.*]] = mul i64 [[TMP21]], -1
67; CHECK-NEXT:    [[NEXT_GEP4:%.*]] = getelementptr i8, i8* null, i64 [[TMP22]]
68; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i8, i8* [[NEXT_GEP4]], i64 -1
69; CHECK-NEXT:    store i8 95, i8* [[TMP23]], align 1
70; CHECK-NEXT:    br label [[PRED_STORE_CONTINUE10]]
71; CHECK:       pred.store.continue10:
72; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
73; CHECK-NEXT:    [[TMP24:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
74; CHECK-NEXT:    br i1 [[TMP24]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
75; CHECK:       middle.block:
76; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC]]
77; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP_LOOPEXIT:%.*]], label [[SCALAR_PH]]
78; CHECK:       scalar.ph:
79; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8* [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ null, [[FOR_BODY_PREHEADER]] ]
80; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
81; CHECK:       for.cond.cleanup.loopexit:
82; CHECK-NEXT:    br label [[FOR_COND_CLEANUP]]
83; CHECK:       for.cond.cleanup:
84; CHECK-NEXT:    ret void
85; CHECK:       for.body:
86; CHECK-NEXT:    [[C_05:%.*]] = phi i8* [ [[INCDEC_PTR:%.*]], [[IF_END:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
87; CHECK-NEXT:    [[INCDEC_PTR]] = getelementptr inbounds i8, i8* [[C_05]], i64 -1
88; CHECK-NEXT:    [[TMP25:%.*]] = load i8, i8* [[INCDEC_PTR]], align 1
89; CHECK-NEXT:    [[TOBOOL_NOT:%.*]] = icmp eq i8 [[TMP25]], 0
90; CHECK-NEXT:    br i1 [[TOBOOL_NOT]], label [[IF_END]], label [[IF_THEN:%.*]]
91; CHECK:       if.then:
92; CHECK-NEXT:    store i8 95, i8* [[INCDEC_PTR]], align 1
93; CHECK-NEXT:    br label [[IF_END]]
94; CHECK:       if.end:
95; CHECK-NEXT:    [[CMP_NOT:%.*]] = icmp eq i8* [[INCDEC_PTR]], [[B]]
96; CHECK-NEXT:    br i1 [[CMP_NOT]], label [[FOR_COND_CLEANUP_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP2:![0-9]+]]
97;
98
99entry:
100  %cmp.not4 = icmp eq i8* %b, null
101  br i1 %cmp.not4, label %for.cond.cleanup, label %for.body.preheader
102
103for.body.preheader:                               ; preds = %entry
104  br label %for.body
105
106for.cond.cleanup.loopexit:                        ; preds = %if.end
107  br label %for.cond.cleanup
108
109for.cond.cleanup:                                 ; preds = %for.cond.cleanup.loopexit, %entry
110  ret void
111
112for.body:                                         ; preds = %for.body.preheader, %if.end
113  %c.05 = phi i8* [ %incdec.ptr, %if.end ], [ null, %for.body.preheader ]
114  %incdec.ptr = getelementptr inbounds i8, i8* %c.05, i64 -1
115  %0 = load i8, i8* %incdec.ptr, align 1
116  %tobool.not = icmp eq i8 %0, 0
117  br i1 %tobool.not, label %if.end, label %if.then
118
119if.then:                                          ; preds = %for.body
120  store i8 95, i8* %incdec.ptr, align 1
121  br label %if.end
122
123if.end:                                           ; preds = %for.body, %if.then
124  %cmp.not = icmp eq i8* %incdec.ptr, %b
125  br i1 %cmp.not, label %for.cond.cleanup.loopexit, label %for.body
126}
127
128; In the test below the pointer phi %ptr.iv.2 is used as
129;  1. As a uniform address for the load, and
130;  2. Non-uniform use by the getelementptr which is stored. This requires the
131;     vector value.
132define void @pointer_induction_used_as_vector(i8** noalias %start.1, i8* noalias %start.2, i64 %N) {
133; CHECK-LABEL: @pointer_induction_used_as_vector(
134; CHECK-NEXT:  entry:
135; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4
136; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
137; CHECK:       vector.ph:
138; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
139; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
140; CHECK-NEXT:    [[IND_END:%.*]] = getelementptr i8*, i8** [[START_1:%.*]], i64 [[N_VEC]]
141; CHECK-NEXT:    [[IND_END3:%.*]] = getelementptr i8, i8* [[START_2:%.*]], i64 [[N_VEC]]
142; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
143; CHECK:       vector.body:
144; CHECK-NEXT:    [[POINTER_PHI:%.*]] = phi i8* [ [[START_2]], [[VECTOR_PH]] ], [ [[PTR_IND:%.*]], [[VECTOR_BODY]] ]
145; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
146; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
147; CHECK-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8*, i8** [[START_1]], i64 [[TMP0]]
148; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr i8, i8* [[POINTER_PHI]], <4 x i64> <i64 0, i64 1, i64 2, i64 3>
149; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, <4 x i8*> [[TMP2]], i64 1
150; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i8*, i8** [[NEXT_GEP]], i32 0
151; CHECK-NEXT:    [[TMP5:%.*]] = bitcast i8** [[TMP4]] to <4 x i8*>*
152; CHECK-NEXT:    store <4 x i8*> [[TMP3]], <4 x i8*>* [[TMP5]], align 8
153; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <4 x i8*> [[TMP1]], i32 0
154; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr i8, i8* [[TMP6]], i32 0
155; CHECK-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to <4 x i8>*
156; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, <4 x i8>* [[TMP8]], align 1
157; CHECK-NEXT:    [[TMP9:%.*]] = add <4 x i8> [[WIDE_LOAD]], <i8 1, i8 1, i8 1, i8 1>
158; CHECK-NEXT:    [[TMP10:%.*]] = bitcast i8* [[TMP7]] to <4 x i8>*
159; CHECK-NEXT:    store <4 x i8> [[TMP9]], <4 x i8>* [[TMP10]], align 1
160; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
161; CHECK-NEXT:    [[PTR_IND]] = getelementptr i8, i8* [[POINTER_PHI]], i64 4
162; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
163; CHECK-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
164; CHECK:       middle.block:
165; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
166; CHECK-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
167; CHECK:       scalar.ph:
168; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
169; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i8** [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[START_1]], [[ENTRY]] ]
170; CHECK-NEXT:    [[BC_RESUME_VAL2:%.*]] = phi i8* [ [[IND_END3]], [[MIDDLE_BLOCK]] ], [ [[START_2]], [[ENTRY]] ]
171; CHECK-NEXT:    br label [[LOOP_BODY:%.*]]
172; CHECK:       loop.body:
173; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_BODY]] ]
174; CHECK-NEXT:    [[PTR_IV_1:%.*]] = phi i8** [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[PTR_IV_1_NEXT:%.*]], [[LOOP_BODY]] ]
175; CHECK-NEXT:    [[PTR_IV_2:%.*]] = phi i8* [ [[BC_RESUME_VAL2]], [[SCALAR_PH]] ], [ [[PTR_IV_2_NEXT:%.*]], [[LOOP_BODY]] ]
176; CHECK-NEXT:    [[PTR_IV_1_NEXT]] = getelementptr inbounds i8*, i8** [[PTR_IV_1]], i64 1
177; CHECK-NEXT:    [[PTR_IV_2_NEXT]] = getelementptr inbounds i8, i8* [[PTR_IV_2]], i64 1
178; CHECK-NEXT:    store i8* [[PTR_IV_2_NEXT]], i8** [[PTR_IV_1]], align 8
179; CHECK-NEXT:    [[LV:%.*]] = load i8, i8* [[PTR_IV_2]], align 1
180; CHECK-NEXT:    [[ADD:%.*]] = add i8 [[LV]], 1
181; CHECK-NEXT:    store i8 [[ADD]], i8* [[PTR_IV_2]], align 1
182; CHECK-NEXT:    [[IV_NEXT]] = add nuw i64 [[IV]], 1
183; CHECK-NEXT:    [[C:%.*]] = icmp ne i64 [[IV_NEXT]], [[N]]
184; CHECK-NEXT:    br i1 [[C]], label [[LOOP_BODY]], label [[EXIT]], !llvm.loop [[LOOP5:![0-9]+]]
185; CHECK:       exit:
186; CHECK-NEXT:    ret void
187;
188
189entry:
190  br label %loop.body
191
192loop.body:                                    ; preds = %loop.body, %entry
193  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.body ]
194  %ptr.iv.1 = phi i8** [ %start.1, %entry ], [ %ptr.iv.1.next, %loop.body ]
195  %ptr.iv.2 = phi i8* [ %start.2, %entry ], [ %ptr.iv.2.next, %loop.body ]
196  %ptr.iv.1.next = getelementptr inbounds i8*, i8** %ptr.iv.1, i64 1
197  %ptr.iv.2.next = getelementptr inbounds i8, i8* %ptr.iv.2, i64 1
198  store i8* %ptr.iv.2.next, i8** %ptr.iv.1, align 8
199  %lv = load i8, i8* %ptr.iv.2, align 1
200  %add = add i8 %lv, 1
201  store i8 %add, i8* %ptr.iv.2, align 1
202  %iv.next = add nuw i64 %iv, 1
203  %c = icmp ne i64 %iv.next, %N
204  br i1 %c, label %loop.body, label %exit
205
206exit:                            ; preds = %loop.body
207  ret void
208}
209