1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2; RUN: opt -S -vectorize-num-stores-pred=1 -force-vector-width=1 -force-vector-interleave=2 -loop-vectorize -verify-loop-info -simplifycfg < %s | FileCheck %s --check-prefix=UNROLL
3; RUN: opt -S -vectorize-num-stores-pred=1 -force-vector-width=1 -force-vector-interleave=2 -loop-vectorize -verify-loop-info < %s | FileCheck %s --check-prefix=UNROLL-NOSIMPLIFY
4; RUN: opt -S -vectorize-num-stores-pred=1 -force-vector-width=2 -force-vector-interleave=1 -loop-vectorize -verify-loop-info -simplifycfg < %s | FileCheck %s --check-prefix=VEC
5
6target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
7
8; Test predication of stores.
9define i32 @test(i32* nocapture %f) #0 {
10; UNROLL-LABEL: @test(
11; UNROLL-NEXT:  entry:
12; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
13; UNROLL:       vector.body:
14; UNROLL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE3:%.*]] ]
15; UNROLL-NEXT:    [[INDUCTION:%.*]] = add i64 [[INDEX]], 0
16; UNROLL-NEXT:    [[INDUCTION1:%.*]] = add i64 [[INDEX]], 1
17; UNROLL-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i32, i32* [[F:%.*]], i64 [[INDUCTION]]
18; UNROLL-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, i32* [[F]], i64 [[INDUCTION1]]
19; UNROLL-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
20; UNROLL-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP1]], align 4
21; UNROLL-NEXT:    [[TMP4:%.*]] = icmp sgt i32 [[TMP2]], 100
22; UNROLL-NEXT:    [[TMP5:%.*]] = icmp sgt i32 [[TMP3]], 100
23; UNROLL-NEXT:    br i1 [[TMP4]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]]
24; UNROLL:       pred.store.if:
25; UNROLL-NEXT:    [[TMP6:%.*]] = add nsw i32 [[TMP2]], 20
26; UNROLL-NEXT:    store i32 [[TMP6]], i32* [[TMP0]], align 4
27; UNROLL-NEXT:    br label [[PRED_STORE_CONTINUE]]
28; UNROLL:       pred.store.continue:
29; UNROLL-NEXT:    br i1 [[TMP5]], label [[PRED_STORE_IF2:%.*]], label [[PRED_STORE_CONTINUE3]]
30; UNROLL:       pred.store.if2:
31; UNROLL-NEXT:    [[TMP7:%.*]] = add nsw i32 [[TMP3]], 20
32; UNROLL-NEXT:    store i32 [[TMP7]], i32* [[TMP1]], align 4
33; UNROLL-NEXT:    br label [[PRED_STORE_CONTINUE3]]
34; UNROLL:       pred.store.continue3:
35; UNROLL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], 2
36; UNROLL-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 128
37; UNROLL-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop !0
38; UNROLL:       middle.block:
39; UNROLL-NEXT:    [[CMP_N:%.*]] = icmp eq i64 128, 128
40; UNROLL-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[FOR_BODY:%.*]]
41; UNROLL:       for.body:
42; UNROLL-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_INC:%.*]] ], [ 128, [[MIDDLE_BLOCK]] ]
43; UNROLL-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[F]], i64 [[INDVARS_IV]]
44; UNROLL-NEXT:    [[TMP9:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
45; UNROLL-NEXT:    [[CMP1:%.*]] = icmp sgt i32 [[TMP9]], 100
46; UNROLL-NEXT:    br i1 [[CMP1]], label [[IF_THEN:%.*]], label [[FOR_INC]]
47; UNROLL:       if.then:
48; UNROLL-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], 20
49; UNROLL-NEXT:    store i32 [[ADD]], i32* [[ARRAYIDX]], align 4
50; UNROLL-NEXT:    br label [[FOR_INC]]
51; UNROLL:       for.inc:
52; UNROLL-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
53; UNROLL-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 128
54; UNROLL-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop !2
55; UNROLL:       for.end:
56; UNROLL-NEXT:    ret i32 0
57;
58; UNROLL-NOSIMPLIFY-LABEL: @test(
59; UNROLL-NOSIMPLIFY-NEXT:  entry:
60; UNROLL-NOSIMPLIFY-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
61; UNROLL-NOSIMPLIFY:       vector.ph:
62; UNROLL-NOSIMPLIFY-NEXT:    br label [[VECTOR_BODY:%.*]]
63; UNROLL-NOSIMPLIFY:       vector.body:
64; UNROLL-NOSIMPLIFY-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE3:%.*]] ]
65; UNROLL-NOSIMPLIFY-NEXT:    [[INDUCTION:%.*]] = add i64 [[INDEX]], 0
66; UNROLL-NOSIMPLIFY-NEXT:    [[INDUCTION1:%.*]] = add i64 [[INDEX]], 1
67; UNROLL-NOSIMPLIFY-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i32, i32* [[F:%.*]], i64 [[INDUCTION]]
68; UNROLL-NOSIMPLIFY-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, i32* [[F]], i64 [[INDUCTION1]]
69; UNROLL-NOSIMPLIFY-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
70; UNROLL-NOSIMPLIFY-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP1]], align 4
71; UNROLL-NOSIMPLIFY-NEXT:    [[TMP4:%.*]] = icmp sgt i32 [[TMP2]], 100
72; UNROLL-NOSIMPLIFY-NEXT:    [[TMP5:%.*]] = icmp sgt i32 [[TMP3]], 100
73; UNROLL-NOSIMPLIFY-NEXT:    br i1 [[TMP4]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]]
74; UNROLL-NOSIMPLIFY:       pred.store.if:
75; UNROLL-NOSIMPLIFY-NEXT:    [[TMP6:%.*]] = add nsw i32 [[TMP2]], 20
76; UNROLL-NOSIMPLIFY-NEXT:    store i32 [[TMP6]], i32* [[TMP0]], align 4
77; UNROLL-NOSIMPLIFY-NEXT:    br label [[PRED_STORE_CONTINUE]]
78; UNROLL-NOSIMPLIFY:       pred.store.continue:
79; UNROLL-NOSIMPLIFY-NEXT:    br i1 [[TMP5]], label [[PRED_STORE_IF2:%.*]], label [[PRED_STORE_CONTINUE3]]
80; UNROLL-NOSIMPLIFY:       pred.store.if2:
81; UNROLL-NOSIMPLIFY-NEXT:    [[TMP7:%.*]] = add nsw i32 [[TMP3]], 20
82; UNROLL-NOSIMPLIFY-NEXT:    store i32 [[TMP7]], i32* [[TMP1]], align 4
83; UNROLL-NOSIMPLIFY-NEXT:    br label [[PRED_STORE_CONTINUE3]]
84; UNROLL-NOSIMPLIFY:       pred.store.continue3:
85; UNROLL-NOSIMPLIFY-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], 2
86; UNROLL-NOSIMPLIFY-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 128
87; UNROLL-NOSIMPLIFY-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop !0
88; UNROLL-NOSIMPLIFY:       middle.block:
89; UNROLL-NOSIMPLIFY-NEXT:    [[CMP_N:%.*]] = icmp eq i64 128, 128
90; UNROLL-NOSIMPLIFY-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
91; UNROLL-NOSIMPLIFY:       scalar.ph:
92; UNROLL-NOSIMPLIFY-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 128, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
93; UNROLL-NOSIMPLIFY-NEXT:    br label [[FOR_BODY:%.*]]
94; UNROLL-NOSIMPLIFY:       for.body:
95; UNROLL-NOSIMPLIFY-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_INC:%.*]] ]
96; UNROLL-NOSIMPLIFY-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[F]], i64 [[INDVARS_IV]]
97; UNROLL-NOSIMPLIFY-NEXT:    [[TMP9:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
98; UNROLL-NOSIMPLIFY-NEXT:    [[CMP1:%.*]] = icmp sgt i32 [[TMP9]], 100
99; UNROLL-NOSIMPLIFY-NEXT:    br i1 [[CMP1]], label [[IF_THEN:%.*]], label [[FOR_INC]]
100; UNROLL-NOSIMPLIFY:       if.then:
101; UNROLL-NOSIMPLIFY-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], 20
102; UNROLL-NOSIMPLIFY-NEXT:    store i32 [[ADD]], i32* [[ARRAYIDX]], align 4
103; UNROLL-NOSIMPLIFY-NEXT:    br label [[FOR_INC]]
104; UNROLL-NOSIMPLIFY:       for.inc:
105; UNROLL-NOSIMPLIFY-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
106; UNROLL-NOSIMPLIFY-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 128
107; UNROLL-NOSIMPLIFY-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop !2
108; UNROLL-NOSIMPLIFY:       for.end:
109; UNROLL-NOSIMPLIFY-NEXT:    ret i32 0
110;
111; VEC-LABEL: @test(
112; VEC-NEXT:  entry:
113; VEC-NEXT:    br label [[VECTOR_BODY:%.*]]
114; VEC:       vector.body:
115; VEC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE2:%.*]] ]
116; VEC-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
117; VEC-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, i32* [[F:%.*]], i64 [[TMP0]]
118; VEC-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i32 0
119; VEC-NEXT:    [[TMP3:%.*]] = bitcast i32* [[TMP2]] to <2 x i32>*
120; VEC-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, <2 x i32>* [[TMP3]], align 4
121; VEC-NEXT:    [[TMP4:%.*]] = icmp sgt <2 x i32> [[WIDE_LOAD]], <i32 100, i32 100>
122; VEC-NEXT:    [[TMP5:%.*]] = extractelement <2 x i1> [[TMP4]], i32 0
123; VEC-NEXT:    br i1 [[TMP5]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]]
124; VEC:       pred.store.if:
125; VEC-NEXT:    [[TMP6:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i32 0
126; VEC-NEXT:    [[TMP7:%.*]] = add nsw i32 [[TMP6]], 20
127; VEC-NEXT:    store i32 [[TMP7]], i32* [[TMP1]], align 4
128; VEC-NEXT:    br label [[PRED_STORE_CONTINUE]]
129; VEC:       pred.store.continue:
130; VEC-NEXT:    [[TMP8:%.*]] = extractelement <2 x i1> [[TMP4]], i32 1
131; VEC-NEXT:    br i1 [[TMP8]], label [[PRED_STORE_IF1:%.*]], label [[PRED_STORE_CONTINUE2]]
132; VEC:       pred.store.if1:
133; VEC-NEXT:    [[TMP9:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i32 1
134; VEC-NEXT:    [[TMP10:%.*]] = add nsw i32 [[TMP9]], 20
135; VEC-NEXT:    [[TMP11:%.*]] = add i64 [[INDEX]], 1
136; VEC-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, i32* [[F]], i64 [[TMP11]]
137; VEC-NEXT:    store i32 [[TMP10]], i32* [[TMP12]], align 4
138; VEC-NEXT:    br label [[PRED_STORE_CONTINUE2]]
139; VEC:       pred.store.continue2:
140; VEC-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], 2
141; VEC-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], 128
142; VEC-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop !0
143; VEC:       middle.block:
144; VEC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 128, 128
145; VEC-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[FOR_BODY:%.*]]
146; VEC:       for.body:
147; VEC-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_INC:%.*]] ], [ 128, [[MIDDLE_BLOCK]] ]
148; VEC-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[F]], i64 [[INDVARS_IV]]
149; VEC-NEXT:    [[TMP14:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
150; VEC-NEXT:    [[CMP1:%.*]] = icmp sgt i32 [[TMP14]], 100
151; VEC-NEXT:    br i1 [[CMP1]], label [[IF_THEN:%.*]], label [[FOR_INC]]
152; VEC:       if.then:
153; VEC-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], 20
154; VEC-NEXT:    store i32 [[ADD]], i32* [[ARRAYIDX]], align 4
155; VEC-NEXT:    br label [[FOR_INC]]
156; VEC:       for.inc:
157; VEC-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
158; VEC-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 128
159; VEC-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop !2
160; VEC:       for.end:
161; VEC-NEXT:    ret i32 0
162;
163entry:
164  br label %for.body
165
166
167
168for.body:
169  %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.inc ]
170  %arrayidx = getelementptr inbounds i32, i32* %f, i64 %indvars.iv
171  %0 = load i32, i32* %arrayidx, align 4
172  %cmp1 = icmp sgt i32 %0, 100
173  br i1 %cmp1, label %if.then, label %for.inc
174
175if.then:
176  %add = add nsw i32 %0, 20
177  store i32 %add, i32* %arrayidx, align 4
178  br label %for.inc
179
180for.inc:
181  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
182  %exitcond = icmp eq i64 %indvars.iv.next, 128
183  br i1 %exitcond, label %for.end, label %for.body
184
185for.end:
186  ret i32 0
187}
188
189; Track basic blocks when unrolling conditional blocks. This code used to assert
190; because we did not update the phi nodes with the proper predecessor in the
191; vectorized loop body.
192; PR18724
193
194define void @bug18724(i1 %cond) {
195; UNROLL-LABEL: @bug18724(
196; UNROLL-NEXT:  entry:
197; UNROLL-NEXT:    [[TMP0:%.*]] = xor i1 [[COND:%.*]], true
198; UNROLL-NEXT:    call void @llvm.assume(i1 [[TMP0]])
199; UNROLL-NEXT:    br label [[FOR_BODY14:%.*]]
200; UNROLL:       for.body14:
201; UNROLL-NEXT:    [[INDVARS_IV3:%.*]] = phi i64 [ [[INDVARS_IV_NEXT4:%.*]], [[FOR_INC23:%.*]] ], [ undef, [[ENTRY:%.*]] ]
202; UNROLL-NEXT:    [[INEWCHUNKS_120:%.*]] = phi i32 [ [[INEWCHUNKS_2:%.*]], [[FOR_INC23]] ], [ undef, [[ENTRY]] ]
203; UNROLL-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds [768 x i32], [768 x i32]* undef, i64 0, i64 [[INDVARS_IV3]]
204; UNROLL-NEXT:    [[TMP:%.*]] = load i32, i32* [[ARRAYIDX16]], align 4
205; UNROLL-NEXT:    br i1 undef, label [[IF_THEN18:%.*]], label [[FOR_INC23]]
206; UNROLL:       if.then18:
207; UNROLL-NEXT:    store i32 2, i32* [[ARRAYIDX16]], align 4
208; UNROLL-NEXT:    [[INC21:%.*]] = add nsw i32 [[INEWCHUNKS_120]], 1
209; UNROLL-NEXT:    br label [[FOR_INC23]]
210; UNROLL:       for.inc23:
211; UNROLL-NEXT:    [[INEWCHUNKS_2]] = phi i32 [ [[INC21]], [[IF_THEN18]] ], [ [[INEWCHUNKS_120]], [[FOR_BODY14]] ]
212; UNROLL-NEXT:    [[INDVARS_IV_NEXT4]] = add nsw i64 [[INDVARS_IV3]], 1
213; UNROLL-NEXT:    [[TMP1:%.*]] = trunc i64 [[INDVARS_IV3]] to i32
214; UNROLL-NEXT:    [[CMP13:%.*]] = icmp slt i32 [[TMP1]], 0
215; UNROLL-NEXT:    call void @llvm.assume(i1 [[CMP13]])
216; UNROLL-NEXT:    br label [[FOR_BODY14]]
217;
218; UNROLL-NOSIMPLIFY-LABEL: @bug18724(
219; UNROLL-NOSIMPLIFY-NEXT:  entry:
220; UNROLL-NOSIMPLIFY-NEXT:    br label [[FOR_BODY9:%.*]]
221; UNROLL-NOSIMPLIFY:       for.body9:
222; UNROLL-NOSIMPLIFY-NEXT:    br i1 [[COND:%.*]], label [[FOR_INC26:%.*]], label [[FOR_BODY14_PREHEADER:%.*]]
223; UNROLL-NOSIMPLIFY:       for.body14.preheader:
224; UNROLL-NOSIMPLIFY-NEXT:    br i1 true, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
225; UNROLL-NOSIMPLIFY:       vector.ph:
226; UNROLL-NOSIMPLIFY-NEXT:    br label [[VECTOR_BODY:%.*]]
227; UNROLL-NOSIMPLIFY:       vector.body:
228; UNROLL-NOSIMPLIFY-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE4:%.*]] ]
229; UNROLL-NOSIMPLIFY-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ undef, [[VECTOR_PH]] ], [ [[PREDPHI:%.*]], [[PRED_STORE_CONTINUE4]] ]
230; UNROLL-NOSIMPLIFY-NEXT:    [[VEC_PHI2:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[PREDPHI5:%.*]], [[PRED_STORE_CONTINUE4]] ]
231; UNROLL-NOSIMPLIFY-NEXT:    [[OFFSET_IDX:%.*]] = add i64 undef, [[INDEX]]
232; UNROLL-NOSIMPLIFY-NEXT:    [[INDUCTION:%.*]] = add i64 [[OFFSET_IDX]], 0
233; UNROLL-NOSIMPLIFY-NEXT:    [[INDUCTION1:%.*]] = add i64 [[OFFSET_IDX]], 1
234; UNROLL-NOSIMPLIFY-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [768 x i32], [768 x i32]* undef, i64 0, i64 [[INDUCTION]]
235; UNROLL-NOSIMPLIFY-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [768 x i32], [768 x i32]* undef, i64 0, i64 [[INDUCTION1]]
236; UNROLL-NOSIMPLIFY-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
237; UNROLL-NOSIMPLIFY-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP1]], align 4
238; UNROLL-NOSIMPLIFY-NEXT:    br i1 undef, label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]]
239; UNROLL-NOSIMPLIFY:       pred.store.if:
240; UNROLL-NOSIMPLIFY-NEXT:    store i32 2, i32* [[TMP0]], align 4
241; UNROLL-NOSIMPLIFY-NEXT:    br label [[PRED_STORE_CONTINUE]]
242; UNROLL-NOSIMPLIFY:       pred.store.continue:
243; UNROLL-NOSIMPLIFY-NEXT:    br i1 undef, label [[PRED_STORE_IF3:%.*]], label [[PRED_STORE_CONTINUE4]]
244; UNROLL-NOSIMPLIFY:       pred.store.if3:
245; UNROLL-NOSIMPLIFY-NEXT:    store i32 2, i32* [[TMP1]], align 4
246; UNROLL-NOSIMPLIFY-NEXT:    br label [[PRED_STORE_CONTINUE4]]
247; UNROLL-NOSIMPLIFY:       pred.store.continue4:
248; UNROLL-NOSIMPLIFY-NEXT:    [[TMP4:%.*]] = add i32 [[VEC_PHI]], 1
249; UNROLL-NOSIMPLIFY-NEXT:    [[TMP5:%.*]] = add i32 [[VEC_PHI2]], 1
250; UNROLL-NOSIMPLIFY-NEXT:    [[PREDPHI]] = select i1 undef, i32 [[VEC_PHI]], i32 [[TMP4]]
251; UNROLL-NOSIMPLIFY-NEXT:    [[PREDPHI5]] = select i1 undef, i32 [[VEC_PHI2]], i32 [[TMP5]]
252; UNROLL-NOSIMPLIFY-NEXT:    [[OFFSET_IDX6:%.*]] = add i64 undef, [[INDEX]]
253; UNROLL-NOSIMPLIFY-NEXT:    [[TMP6:%.*]] = trunc i64 [[OFFSET_IDX6]] to i32
254; UNROLL-NOSIMPLIFY-NEXT:    [[INDUCTION7:%.*]] = add i32 [[TMP6]], 0
255; UNROLL-NOSIMPLIFY-NEXT:    [[INDUCTION8:%.*]] = add i32 [[TMP6]], 1
256; UNROLL-NOSIMPLIFY-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], 2
257; UNROLL-NOSIMPLIFY-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 0
258; UNROLL-NOSIMPLIFY-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop !3
259; UNROLL-NOSIMPLIFY:       middle.block:
260; UNROLL-NOSIMPLIFY-NEXT:    [[BIN_RDX:%.*]] = add i32 [[PREDPHI5]], [[PREDPHI]]
261; UNROLL-NOSIMPLIFY-NEXT:    [[CMP_N:%.*]] = icmp eq i64 1, 0
262; UNROLL-NOSIMPLIFY-NEXT:    br i1 [[CMP_N]], label [[FOR_INC26_LOOPEXIT:%.*]], label [[SCALAR_PH]]
263; UNROLL-NOSIMPLIFY:       scalar.ph:
264; UNROLL-NOSIMPLIFY-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ undef, [[MIDDLE_BLOCK]] ], [ undef, [[FOR_BODY14_PREHEADER]] ]
265; UNROLL-NOSIMPLIFY-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ undef, [[FOR_BODY14_PREHEADER]] ], [ [[BIN_RDX]], [[MIDDLE_BLOCK]] ]
266; UNROLL-NOSIMPLIFY-NEXT:    br label [[FOR_BODY14:%.*]]
267; UNROLL-NOSIMPLIFY:       for.body14:
268; UNROLL-NOSIMPLIFY-NEXT:    [[INDVARS_IV3:%.*]] = phi i64 [ [[INDVARS_IV_NEXT4:%.*]], [[FOR_INC23:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
269; UNROLL-NOSIMPLIFY-NEXT:    [[INEWCHUNKS_120:%.*]] = phi i32 [ [[INEWCHUNKS_2:%.*]], [[FOR_INC23]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
270; UNROLL-NOSIMPLIFY-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds [768 x i32], [768 x i32]* undef, i64 0, i64 [[INDVARS_IV3]]
271; UNROLL-NOSIMPLIFY-NEXT:    [[TMP:%.*]] = load i32, i32* [[ARRAYIDX16]], align 4
272; UNROLL-NOSIMPLIFY-NEXT:    br i1 undef, label [[IF_THEN18:%.*]], label [[FOR_INC23]]
273; UNROLL-NOSIMPLIFY:       if.then18:
274; UNROLL-NOSIMPLIFY-NEXT:    store i32 2, i32* [[ARRAYIDX16]], align 4
275; UNROLL-NOSIMPLIFY-NEXT:    [[INC21:%.*]] = add nsw i32 [[INEWCHUNKS_120]], 1
276; UNROLL-NOSIMPLIFY-NEXT:    br label [[FOR_INC23]]
277; UNROLL-NOSIMPLIFY:       for.inc23:
278; UNROLL-NOSIMPLIFY-NEXT:    [[INEWCHUNKS_2]] = phi i32 [ [[INC21]], [[IF_THEN18]] ], [ [[INEWCHUNKS_120]], [[FOR_BODY14]] ]
279; UNROLL-NOSIMPLIFY-NEXT:    [[INDVARS_IV_NEXT4]] = add nsw i64 [[INDVARS_IV3]], 1
280; UNROLL-NOSIMPLIFY-NEXT:    [[TMP1:%.*]] = trunc i64 [[INDVARS_IV3]] to i32
281; UNROLL-NOSIMPLIFY-NEXT:    [[CMP13:%.*]] = icmp slt i32 [[TMP1]], 0
282; UNROLL-NOSIMPLIFY-NEXT:    br i1 [[CMP13]], label [[FOR_BODY14]], label [[FOR_INC26_LOOPEXIT]], !llvm.loop !4
283; UNROLL-NOSIMPLIFY:       for.inc26.loopexit:
284; UNROLL-NOSIMPLIFY-NEXT:    [[INEWCHUNKS_2_LCSSA:%.*]] = phi i32 [ [[INEWCHUNKS_2]], [[FOR_INC23]] ], [ [[BIN_RDX]], [[MIDDLE_BLOCK]] ]
285; UNROLL-NOSIMPLIFY-NEXT:    br label [[FOR_INC26]]
286; UNROLL-NOSIMPLIFY:       for.inc26:
287; UNROLL-NOSIMPLIFY-NEXT:    [[INEWCHUNKS_1_LCSSA:%.*]] = phi i32 [ undef, [[FOR_BODY9]] ], [ [[INEWCHUNKS_2_LCSSA]], [[FOR_INC26_LOOPEXIT]] ]
288; UNROLL-NOSIMPLIFY-NEXT:    unreachable
289;
290; VEC-LABEL: @bug18724(
291; VEC-NEXT:  entry:
292; VEC-NEXT:    [[TMP0:%.*]] = xor i1 [[COND:%.*]], true
293; VEC-NEXT:    call void @llvm.assume(i1 [[TMP0]])
294; VEC-NEXT:    br label [[FOR_BODY14:%.*]]
295; VEC:       for.body14:
296; VEC-NEXT:    [[INDVARS_IV3:%.*]] = phi i64 [ [[INDVARS_IV_NEXT4:%.*]], [[FOR_INC23:%.*]] ], [ undef, [[ENTRY:%.*]] ]
297; VEC-NEXT:    [[INEWCHUNKS_120:%.*]] = phi i32 [ [[INEWCHUNKS_2:%.*]], [[FOR_INC23]] ], [ undef, [[ENTRY]] ]
298; VEC-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds [768 x i32], [768 x i32]* undef, i64 0, i64 [[INDVARS_IV3]]
299; VEC-NEXT:    [[TMP:%.*]] = load i32, i32* [[ARRAYIDX16]], align 4
300; VEC-NEXT:    br i1 undef, label [[IF_THEN18:%.*]], label [[FOR_INC23]]
301; VEC:       if.then18:
302; VEC-NEXT:    store i32 2, i32* [[ARRAYIDX16]], align 4
303; VEC-NEXT:    [[INC21:%.*]] = add nsw i32 [[INEWCHUNKS_120]], 1
304; VEC-NEXT:    br label [[FOR_INC23]]
305; VEC:       for.inc23:
306; VEC-NEXT:    [[INEWCHUNKS_2]] = phi i32 [ [[INC21]], [[IF_THEN18]] ], [ [[INEWCHUNKS_120]], [[FOR_BODY14]] ]
307; VEC-NEXT:    [[INDVARS_IV_NEXT4]] = add nsw i64 [[INDVARS_IV3]], 1
308; VEC-NEXT:    [[TMP1:%.*]] = trunc i64 [[INDVARS_IV3]] to i32
309; VEC-NEXT:    [[CMP13:%.*]] = icmp slt i32 [[TMP1]], 0
310; VEC-NEXT:    call void @llvm.assume(i1 [[CMP13]])
311; VEC-NEXT:    br label [[FOR_BODY14]]
312;
313entry:
314  br label %for.body9
315
316for.body9:
317  br i1 %cond, label %for.inc26, label %for.body14
318
319for.body14:
320  %indvars.iv3 = phi i64 [ %indvars.iv.next4, %for.inc23 ], [ undef, %for.body9 ]
321  %iNewChunks.120 = phi i32 [ %iNewChunks.2, %for.inc23 ], [ undef, %for.body9 ]
322  %arrayidx16 = getelementptr inbounds [768 x i32], [768 x i32]* undef, i64 0, i64 %indvars.iv3
323  %tmp = load i32, i32* %arrayidx16, align 4
324  br i1 undef, label %if.then18, label %for.inc23
325
326if.then18:
327  store i32 2, i32* %arrayidx16, align 4
328  %inc21 = add nsw i32 %iNewChunks.120, 1
329  br label %for.inc23
330
331for.inc23:
332  %iNewChunks.2 = phi i32 [ %inc21, %if.then18 ], [ %iNewChunks.120, %for.body14 ]
333  %indvars.iv.next4 = add nsw i64 %indvars.iv3, 1
334  %tmp1 = trunc i64 %indvars.iv3 to i32
335  %cmp13 = icmp slt i32 %tmp1, 0
336  br i1 %cmp13, label %for.body14, label %for.inc26
337
338for.inc26:
339  %iNewChunks.1.lcssa = phi i32 [ undef, %for.body9 ], [ %iNewChunks.2, %for.inc23 ]
340  unreachable
341}
342
343; In the test below, it's more profitable for the expression feeding the
344; conditional store to remain scalar. Since we can only type-shrink vector
345; types, we shouldn't try to represent the expression in a smaller type.
346;
347define void @minimal_bit_widths(i1 %c) {
348; UNROLL-LABEL: @minimal_bit_widths(
349; UNROLL-NEXT:  entry:
350; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
351; UNROLL:       vector.body:
352; UNROLL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE6:%.*]] ]
353; UNROLL-NEXT:    [[OFFSET_IDX:%.*]] = sub i64 undef, [[INDEX]]
354; UNROLL-NEXT:    [[INDUCTION3:%.*]] = add i64 [[OFFSET_IDX]], 0
355; UNROLL-NEXT:    [[INDUCTION4:%.*]] = add i64 [[OFFSET_IDX]], -1
356; UNROLL-NEXT:    br i1 [[C:%.*]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE6]]
357; UNROLL:       pred.store.if:
358; UNROLL-NEXT:    [[INDUCTION:%.*]] = add i64 [[INDEX]], 0
359; UNROLL-NEXT:    [[TMP0:%.*]] = getelementptr i8, i8* undef, i64 [[INDUCTION]]
360; UNROLL-NEXT:    [[TMP1:%.*]] = load i8, i8* [[TMP0]], align 1
361; UNROLL-NEXT:    [[TMP2:%.*]] = zext i8 [[TMP1]] to i32
362; UNROLL-NEXT:    [[TMP3:%.*]] = trunc i32 [[TMP2]] to i8
363; UNROLL-NEXT:    store i8 [[TMP3]], i8* [[TMP0]], align 1
364; UNROLL-NEXT:    [[INDUCTION2:%.*]] = add i64 [[INDEX]], 1
365; UNROLL-NEXT:    [[TMP4:%.*]] = getelementptr i8, i8* undef, i64 [[INDUCTION2]]
366; UNROLL-NEXT:    [[TMP5:%.*]] = load i8, i8* [[TMP4]], align 1
367; UNROLL-NEXT:    [[TMP6:%.*]] = zext i8 [[TMP5]] to i32
368; UNROLL-NEXT:    [[TMP7:%.*]] = trunc i32 [[TMP6]] to i8
369; UNROLL-NEXT:    store i8 [[TMP7]], i8* [[TMP4]], align 1
370; UNROLL-NEXT:    br label [[PRED_STORE_CONTINUE6]]
371; UNROLL:       pred.store.continue6:
372; UNROLL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], 2
373; UNROLL-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], undef
374; UNROLL-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop !3
375; UNROLL:       middle.block:
376; UNROLL-NEXT:    [[CMP_N:%.*]] = icmp eq i64 undef, undef
377; UNROLL-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[FOR_BODY:%.*]]
378; UNROLL:       for.body:
379; UNROLL-NEXT:    [[TMP0:%.*]] = phi i64 [ [[TMP6:%.*]], [[FOR_INC:%.*]] ], [ undef, [[MIDDLE_BLOCK]] ]
380; UNROLL-NEXT:    [[TMP1:%.*]] = phi i64 [ [[TMP7:%.*]], [[FOR_INC]] ], [ undef, [[MIDDLE_BLOCK]] ]
381; UNROLL-NEXT:    [[TMP2:%.*]] = getelementptr i8, i8* undef, i64 [[TMP0]]
382; UNROLL-NEXT:    [[TMP3:%.*]] = load i8, i8* [[TMP2]], align 1
383; UNROLL-NEXT:    br i1 [[C]], label [[IF_THEN:%.*]], label [[FOR_INC]]
384; UNROLL:       if.then:
385; UNROLL-NEXT:    [[TMP4:%.*]] = zext i8 [[TMP3]] to i32
386; UNROLL-NEXT:    [[TMP5:%.*]] = trunc i32 [[TMP4]] to i8
387; UNROLL-NEXT:    store i8 [[TMP5]], i8* [[TMP2]], align 1
388; UNROLL-NEXT:    br label [[FOR_INC]]
389; UNROLL:       for.inc:
390; UNROLL-NEXT:    [[TMP6]] = add nuw nsw i64 [[TMP0]], 1
391; UNROLL-NEXT:    [[TMP7]] = add i64 [[TMP1]], -1
392; UNROLL-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[TMP7]], 0
393; UNROLL-NEXT:    br i1 [[TMP8]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop !4
394; UNROLL:       for.end:
395; UNROLL-NEXT:    ret void
396;
397; UNROLL-NOSIMPLIFY-LABEL: @minimal_bit_widths(
398; UNROLL-NOSIMPLIFY-NEXT:  entry:
399; UNROLL-NOSIMPLIFY-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
400; UNROLL-NOSIMPLIFY:       vector.ph:
401; UNROLL-NOSIMPLIFY-NEXT:    br label [[VECTOR_BODY:%.*]]
402; UNROLL-NOSIMPLIFY:       vector.body:
403; UNROLL-NOSIMPLIFY-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE6:%.*]] ]
404; UNROLL-NOSIMPLIFY-NEXT:    [[OFFSET_IDX:%.*]] = sub i64 undef, [[INDEX]]
405; UNROLL-NOSIMPLIFY-NEXT:    [[INDUCTION3:%.*]] = add i64 [[OFFSET_IDX]], 0
406; UNROLL-NOSIMPLIFY-NEXT:    [[INDUCTION4:%.*]] = add i64 [[OFFSET_IDX]], -1
407; UNROLL-NOSIMPLIFY-NEXT:    br i1 [[C:%.*]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]]
408; UNROLL-NOSIMPLIFY:       pred.store.if:
409; UNROLL-NOSIMPLIFY-NEXT:    [[INDUCTION:%.*]] = add i64 [[INDEX]], 0
410; UNROLL-NOSIMPLIFY-NEXT:    [[TMP0:%.*]] = getelementptr i8, i8* undef, i64 [[INDUCTION]]
411; UNROLL-NOSIMPLIFY-NEXT:    [[TMP1:%.*]] = load i8, i8* [[TMP0]], align 1
412; UNROLL-NOSIMPLIFY-NEXT:    [[TMP2:%.*]] = zext i8 [[TMP1]] to i32
413; UNROLL-NOSIMPLIFY-NEXT:    [[TMP3:%.*]] = trunc i32 [[TMP2]] to i8
414; UNROLL-NOSIMPLIFY-NEXT:    store i8 [[TMP3]], i8* [[TMP0]], align 1
415; UNROLL-NOSIMPLIFY-NEXT:    br label [[PRED_STORE_CONTINUE]]
416; UNROLL-NOSIMPLIFY:       pred.store.continue:
417; UNROLL-NOSIMPLIFY-NEXT:    br i1 [[C]], label [[PRED_STORE_IF5:%.*]], label [[PRED_STORE_CONTINUE6]]
418; UNROLL-NOSIMPLIFY:       pred.store.if5:
419; UNROLL-NOSIMPLIFY-NEXT:    [[INDUCTION2:%.*]] = add i64 [[INDEX]], 1
420; UNROLL-NOSIMPLIFY-NEXT:    [[TMP4:%.*]] = getelementptr i8, i8* undef, i64 [[INDUCTION2]]
421; UNROLL-NOSIMPLIFY-NEXT:    [[TMP5:%.*]] = load i8, i8* [[TMP4]], align 1
422; UNROLL-NOSIMPLIFY-NEXT:    [[TMP6:%.*]] = zext i8 [[TMP5]] to i32
423; UNROLL-NOSIMPLIFY-NEXT:    [[TMP7:%.*]] = trunc i32 [[TMP6]] to i8
424; UNROLL-NOSIMPLIFY-NEXT:    store i8 [[TMP7]], i8* [[TMP4]], align 1
425; UNROLL-NOSIMPLIFY-NEXT:    br label [[PRED_STORE_CONTINUE6]]
426; UNROLL-NOSIMPLIFY:       pred.store.continue6:
427; UNROLL-NOSIMPLIFY-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], 2
428; UNROLL-NOSIMPLIFY-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], undef
429; UNROLL-NOSIMPLIFY-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop !5
430; UNROLL-NOSIMPLIFY:       middle.block:
431; UNROLL-NOSIMPLIFY-NEXT:    [[CMP_N:%.*]] = icmp eq i64 undef, undef
432; UNROLL-NOSIMPLIFY-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
433; UNROLL-NOSIMPLIFY:       scalar.ph:
434; UNROLL-NOSIMPLIFY-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ undef, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
435; UNROLL-NOSIMPLIFY-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i64 [ undef, [[MIDDLE_BLOCK]] ], [ undef, [[ENTRY]] ]
436; UNROLL-NOSIMPLIFY-NEXT:    br label [[FOR_BODY:%.*]]
437; UNROLL-NOSIMPLIFY:       for.body:
438; UNROLL-NOSIMPLIFY-NEXT:    [[TMP0:%.*]] = phi i64 [ [[TMP6:%.*]], [[FOR_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
439; UNROLL-NOSIMPLIFY-NEXT:    [[TMP1:%.*]] = phi i64 [ [[TMP7:%.*]], [[FOR_INC]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
440; UNROLL-NOSIMPLIFY-NEXT:    [[TMP2:%.*]] = getelementptr i8, i8* undef, i64 [[TMP0]]
441; UNROLL-NOSIMPLIFY-NEXT:    [[TMP3:%.*]] = load i8, i8* [[TMP2]], align 1
442; UNROLL-NOSIMPLIFY-NEXT:    br i1 [[C]], label [[IF_THEN:%.*]], label [[FOR_INC]]
443; UNROLL-NOSIMPLIFY:       if.then:
444; UNROLL-NOSIMPLIFY-NEXT:    [[TMP4:%.*]] = zext i8 [[TMP3]] to i32
445; UNROLL-NOSIMPLIFY-NEXT:    [[TMP5:%.*]] = trunc i32 [[TMP4]] to i8
446; UNROLL-NOSIMPLIFY-NEXT:    store i8 [[TMP5]], i8* [[TMP2]], align 1
447; UNROLL-NOSIMPLIFY-NEXT:    br label [[FOR_INC]]
448; UNROLL-NOSIMPLIFY:       for.inc:
449; UNROLL-NOSIMPLIFY-NEXT:    [[TMP6]] = add nuw nsw i64 [[TMP0]], 1
450; UNROLL-NOSIMPLIFY-NEXT:    [[TMP7]] = add i64 [[TMP1]], -1
451; UNROLL-NOSIMPLIFY-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[TMP7]], 0
452; UNROLL-NOSIMPLIFY-NEXT:    br i1 [[TMP8]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop !6
453; UNROLL-NOSIMPLIFY:       for.end:
454; UNROLL-NOSIMPLIFY-NEXT:    ret void
455;
456; VEC-LABEL: @minimal_bit_widths(
457; VEC-NEXT:  entry:
458; VEC-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i1> undef, i1 [[C:%.*]], i32 0
459; VEC-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i1> [[BROADCAST_SPLATINSERT]], <2 x i1> undef, <2 x i32> zeroinitializer
460; VEC-NEXT:    br label [[VECTOR_BODY:%.*]]
461; VEC:       vector.body:
462; VEC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE3:%.*]] ]
463; VEC-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
464; VEC-NEXT:    [[OFFSET_IDX:%.*]] = sub i64 undef, [[INDEX]]
465; VEC-NEXT:    [[TMP1:%.*]] = add i64 [[OFFSET_IDX]], 0
466; VEC-NEXT:    [[TMP2:%.*]] = getelementptr i8, i8* undef, i64 [[TMP0]]
467; VEC-NEXT:    [[TMP3:%.*]] = getelementptr i8, i8* [[TMP2]], i32 0
468; VEC-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP3]] to <2 x i8>*
469; VEC-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i8>, <2 x i8>* [[TMP4]], align 1
470; VEC-NEXT:    [[TMP5:%.*]] = extractelement <2 x i1> [[BROADCAST_SPLAT]], i32 0
471; VEC-NEXT:    br i1 [[TMP5]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]]
472; VEC:       pred.store.if:
473; VEC-NEXT:    [[TMP6:%.*]] = extractelement <2 x i8> [[WIDE_LOAD]], i32 0
474; VEC-NEXT:    [[TMP7:%.*]] = zext i8 [[TMP6]] to i32
475; VEC-NEXT:    [[TMP8:%.*]] = trunc i32 [[TMP7]] to i8
476; VEC-NEXT:    store i8 [[TMP8]], i8* [[TMP2]], align 1
477; VEC-NEXT:    br label [[PRED_STORE_CONTINUE]]
478; VEC:       pred.store.continue:
479; VEC-NEXT:    [[TMP9:%.*]] = extractelement <2 x i1> [[BROADCAST_SPLAT]], i32 1
480; VEC-NEXT:    br i1 [[TMP9]], label [[PRED_STORE_IF2:%.*]], label [[PRED_STORE_CONTINUE3]]
481; VEC:       pred.store.if2:
482; VEC-NEXT:    [[TMP10:%.*]] = extractelement <2 x i8> [[WIDE_LOAD]], i32 1
483; VEC-NEXT:    [[TMP11:%.*]] = zext i8 [[TMP10]] to i32
484; VEC-NEXT:    [[TMP12:%.*]] = trunc i32 [[TMP11]] to i8
485; VEC-NEXT:    [[TMP13:%.*]] = add i64 [[INDEX]], 1
486; VEC-NEXT:    [[TMP14:%.*]] = getelementptr i8, i8* undef, i64 [[TMP13]]
487; VEC-NEXT:    store i8 [[TMP12]], i8* [[TMP14]], align 1
488; VEC-NEXT:    br label [[PRED_STORE_CONTINUE3]]
489; VEC:       pred.store.continue3:
490; VEC-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], 2
491; VEC-NEXT:    [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], undef
492; VEC-NEXT:    br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop !4
493; VEC:       middle.block:
494; VEC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 undef, undef
495; VEC-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[FOR_BODY:%.*]]
496; VEC:       for.body:
497; VEC-NEXT:    [[TMP0:%.*]] = phi i64 [ [[TMP6:%.*]], [[FOR_INC:%.*]] ], [ undef, [[MIDDLE_BLOCK]] ]
498; VEC-NEXT:    [[TMP1:%.*]] = phi i64 [ [[TMP7:%.*]], [[FOR_INC]] ], [ undef, [[MIDDLE_BLOCK]] ]
499; VEC-NEXT:    [[TMP2:%.*]] = getelementptr i8, i8* undef, i64 [[TMP0]]
500; VEC-NEXT:    [[TMP3:%.*]] = load i8, i8* [[TMP2]], align 1
501; VEC-NEXT:    br i1 [[C]], label [[IF_THEN:%.*]], label [[FOR_INC]]
502; VEC:       if.then:
503; VEC-NEXT:    [[TMP4:%.*]] = zext i8 [[TMP3]] to i32
504; VEC-NEXT:    [[TMP5:%.*]] = trunc i32 [[TMP4]] to i8
505; VEC-NEXT:    store i8 [[TMP5]], i8* [[TMP2]], align 1
506; VEC-NEXT:    br label [[FOR_INC]]
507; VEC:       for.inc:
508; VEC-NEXT:    [[TMP6]] = add nuw nsw i64 [[TMP0]], 1
509; VEC-NEXT:    [[TMP7]] = add i64 [[TMP1]], -1
510; VEC-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[TMP7]], 0
511; VEC-NEXT:    br i1 [[TMP8]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop !5
512; VEC:       for.end:
513; VEC-NEXT:    ret void
514;
515entry:
516  br label %for.body
517
518for.body:
519  %tmp0 = phi i64 [ %tmp6, %for.inc ], [ 0, %entry ]
520  %tmp1 = phi i64 [ %tmp7, %for.inc ], [ undef, %entry ]
521  %tmp2 = getelementptr i8, i8* undef, i64 %tmp0
522  %tmp3 = load i8, i8* %tmp2, align 1
523  br i1 %c, label %if.then, label %for.inc
524
525if.then:
526  %tmp4 = zext i8 %tmp3 to i32
527  %tmp5 = trunc i32 %tmp4 to i8
528  store i8 %tmp5, i8* %tmp2, align 1
529  br label %for.inc
530
531for.inc:
532  %tmp6 = add nuw nsw i64 %tmp0, 1
533  %tmp7 = add i64 %tmp1, -1
534  %tmp8 = icmp eq i64 %tmp7, 0
535  br i1 %tmp8, label %for.end, label %for.body
536
537for.end:
538  ret void
539}
540