1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2; RUN: opt < %s -slp-vectorizer -S | FileCheck %s --check-prefix=DEFAULT
3; RUN: opt < %s -slp-schedule-budget=0 -slp-min-tree-size=0 -slp-threshold=-30 -slp-vectorizer -S | FileCheck %s --check-prefix=GATHER
4; RUN: opt < %s -slp-schedule-budget=0 -slp-threshold=-30 -slp-vectorizer -S | FileCheck %s --check-prefix=MAX-COST
5
6target datalayout = "e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128"
7target triple = "aarch64--linux-gnu"
8
9@a = common global [80 x i8] zeroinitializer, align 16
10
11; DEFAULT-LABEL: @PR28330(
12; DEFAULT: %tmp17 = phi i32 [ %tmp34, %for.body ], [ 0, %entry ]
13; DEFAULT: %[[S0:.+]] = select <8 x i1> %1, <8 x i32> <i32 -720, i32 -720, i32 -720, i32 -720, i32 -720, i32 -720, i32 -720, i32 -720>, <8 x i32> <i32 -80, i32 -80, i32 -80, i32 -80, i32 -80, i32 -80, i32 -80, i32 -80>
14; DEFAULT: %[[R0:.+]] = shufflevector <8 x i32> %[[S0]], <8 x i32> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef>
15; DEFAULT: %[[R1:.+]] = add <8 x i32> %[[S0]], %[[R0]]
16; DEFAULT: %[[R2:.+]] = shufflevector <8 x i32> %[[R1]], <8 x i32> undef, <8 x i32> <i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
17; DEFAULT: %[[R3:.+]] = add <8 x i32> %[[R1]], %[[R2]]
18; DEFAULT: %[[R4:.+]] = shufflevector <8 x i32> %[[R3]], <8 x i32> undef, <8 x i32> <i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
19; DEFAULT: %[[R5:.+]] = add <8 x i32> %[[R3]], %[[R4]]
20; DEFAULT: %[[R6:.+]] = extractelement <8 x i32> %[[R5]], i32 0
21; DEFAULT: %tmp34 = add i32 %[[R6]], %tmp17
22;
23; GATHER-LABEL: @PR28330(
24; GATHER: %tmp17 = phi i32 [ %tmp34, %for.body ], [ 0, %entry ]
25; GATHER: %tmp19 = select i1 %tmp1, i32 -720, i32 -80
26; GATHER: %tmp21 = select i1 %tmp3, i32 -720, i32 -80
27; GATHER: %tmp23 = select i1 %tmp5, i32 -720, i32 -80
28; GATHER: %tmp25 = select i1 %tmp7, i32 -720, i32 -80
29; GATHER: %tmp27 = select i1 %tmp9, i32 -720, i32 -80
30; GATHER: %tmp29 = select i1 %tmp11, i32 -720, i32 -80
31; GATHER: %tmp31 = select i1 %tmp13, i32 -720, i32 -80
32; GATHER: %tmp33 = select i1 %tmp15, i32 -720, i32 -80
33; GATHER: %[[I0:.+]] = insertelement <8 x i32> undef, i32 %tmp19, i32 0
34; GATHER: %[[I1:.+]] = insertelement <8 x i32> %[[I0]], i32 %tmp21, i32 1
35; GATHER: %[[I2:.+]] = insertelement <8 x i32> %[[I1]], i32 %tmp23, i32 2
36; GATHER: %[[I3:.+]] = insertelement <8 x i32> %[[I2]], i32 %tmp25, i32 3
37; GATHER: %[[I4:.+]] = insertelement <8 x i32> %[[I3]], i32 %tmp27, i32 4
38; GATHER: %[[I5:.+]] = insertelement <8 x i32> %[[I4]], i32 %tmp29, i32 5
39; GATHER: %[[I6:.+]] = insertelement <8 x i32> %[[I5]], i32 %tmp31, i32 6
40; GATHER: %[[I7:.+]] = insertelement <8 x i32> %[[I6]], i32 %tmp33, i32 7
41; GATHER: %[[R0:.+]] = shufflevector <8 x i32> %[[I7]], <8 x i32> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef>
42; GATHER: %[[R1:.+]] = add <8 x i32> %[[I7]], %[[R0]]
43; GATHER: %[[R2:.+]] = shufflevector <8 x i32> %[[R1]], <8 x i32> undef, <8 x i32> <i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
44; GATHER: %[[R3:.+]] = add <8 x i32> %[[R1]], %[[R2]]
45; GATHER: %[[R4:.+]] = shufflevector <8 x i32> %[[R3]], <8 x i32> undef, <8 x i32> <i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
46; GATHER: %[[R5:.+]] = add <8 x i32> %[[R3]], %[[R4]]
47; GATHER: %[[R6:.+]] = extractelement <8 x i32> %[[R5]], i32 0
48; GATHER: %tmp34 = add i32 %[[R6]], %tmp17
49;
50; MAX-COST-LABEL: @PR28330(
51; MAX-COST-NOT: shufflevector
52
53define void @PR28330(i32 %n) {
54entry:
55  %tmp0 = load i8, i8* getelementptr inbounds ([80 x i8], [80 x i8]* @a, i64 0, i64 1), align 1
56  %tmp1 = icmp eq i8 %tmp0, 0
57  %tmp2 = load i8, i8* getelementptr inbounds ([80 x i8], [80 x i8]* @a, i64 0, i64 2), align 2
58  %tmp3 = icmp eq i8 %tmp2, 0
59  %tmp4 = load i8, i8* getelementptr inbounds ([80 x i8], [80 x i8]* @a, i64 0, i64 3), align 1
60  %tmp5 = icmp eq i8 %tmp4, 0
61  %tmp6 = load i8, i8* getelementptr inbounds ([80 x i8], [80 x i8]* @a, i64 0, i64 4), align 4
62  %tmp7 = icmp eq i8 %tmp6, 0
63  %tmp8 = load i8, i8* getelementptr inbounds ([80 x i8], [80 x i8]* @a, i64 0, i64 5), align 1
64  %tmp9 = icmp eq i8 %tmp8, 0
65  %tmp10 = load i8, i8* getelementptr inbounds ([80 x i8], [80 x i8]* @a, i64 0, i64 6), align 2
66  %tmp11 = icmp eq i8 %tmp10, 0
67  %tmp12 = load i8, i8* getelementptr inbounds ([80 x i8], [80 x i8]* @a, i64 0, i64 7), align 1
68  %tmp13 = icmp eq i8 %tmp12, 0
69  %tmp14 = load i8, i8* getelementptr inbounds ([80 x i8], [80 x i8]* @a, i64 0, i64 8), align 8
70  %tmp15 = icmp eq i8 %tmp14, 0
71  br label %for.body
72
73for.body:
74  %tmp17 = phi i32 [ %tmp34, %for.body ], [ 0, %entry ]
75  %tmp19 = select i1 %tmp1, i32 -720, i32 -80
76  %tmp20 = add i32 %tmp17, %tmp19
77  %tmp21 = select i1 %tmp3, i32 -720, i32 -80
78  %tmp22 = add i32 %tmp20, %tmp21
79  %tmp23 = select i1 %tmp5, i32 -720, i32 -80
80  %tmp24 = add i32 %tmp22, %tmp23
81  %tmp25 = select i1 %tmp7, i32 -720, i32 -80
82  %tmp26 = add i32 %tmp24, %tmp25
83  %tmp27 = select i1 %tmp9, i32 -720, i32 -80
84  %tmp28 = add i32 %tmp26, %tmp27
85  %tmp29 = select i1 %tmp11, i32 -720, i32 -80
86  %tmp30 = add i32 %tmp28, %tmp29
87  %tmp31 = select i1 %tmp13, i32 -720, i32 -80
88  %tmp32 = add i32 %tmp30, %tmp31
89  %tmp33 = select i1 %tmp15, i32 -720, i32 -80
90  %tmp34 = add i32 %tmp32, %tmp33
91  br label %for.body
92}
93
94define void @PR32038(i32 %n) {
95; DEFAULT-LABEL: @PR32038(
96; DEFAULT-NEXT:  entry:
97; DEFAULT-NEXT:    [[TMP0:%.*]] = load <8 x i8>, <8 x i8>* bitcast (i8* getelementptr inbounds ([80 x i8], [80 x i8]* @a, i64 0, i64 1) to <8 x i8>*), align 1
98; DEFAULT-NEXT:    [[TMP1:%.*]] = icmp eq <8 x i8> [[TMP0]], zeroinitializer
99; DEFAULT-NEXT:    br label [[FOR_BODY:%.*]]
100; DEFAULT:       for.body:
101; DEFAULT-NEXT:    [[TMP17:%.*]] = phi i32 [ [[TMP34:%.*]], [[FOR_BODY]] ], [ 0, [[ENTRY:%.*]] ]
102; DEFAULT-NEXT:    [[TMP2:%.*]] = select <8 x i1> [[TMP1]], <8 x i32> <i32 -720, i32 -720, i32 -720, i32 -720, i32 -720, i32 -720, i32 -720, i32 -720>, <8 x i32> <i32 -80, i32 -80, i32 -80, i32 -80, i32 -80, i32 -80, i32 -80, i32 -80>
103; DEFAULT-NEXT:    [[TMP20:%.*]] = add i32 -5, undef
104; DEFAULT-NEXT:    [[TMP22:%.*]] = add i32 [[TMP20]], undef
105; DEFAULT-NEXT:    [[TMP24:%.*]] = add i32 [[TMP22]], undef
106; DEFAULT-NEXT:    [[TMP26:%.*]] = add i32 [[TMP24]], undef
107; DEFAULT-NEXT:    [[TMP28:%.*]] = add i32 [[TMP26]], undef
108; DEFAULT-NEXT:    [[TMP30:%.*]] = add i32 [[TMP28]], undef
109; DEFAULT-NEXT:    [[TMP32:%.*]] = add i32 [[TMP30]], undef
110; DEFAULT-NEXT:    [[RDX_SHUF:%.*]] = shufflevector <8 x i32> [[TMP2]], <8 x i32> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef>
111; DEFAULT-NEXT:    [[BIN_RDX:%.*]] = add <8 x i32> [[TMP2]], [[RDX_SHUF]]
112; DEFAULT-NEXT:    [[RDX_SHUF1:%.*]] = shufflevector <8 x i32> [[BIN_RDX]], <8 x i32> undef, <8 x i32> <i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
113; DEFAULT-NEXT:    [[BIN_RDX2:%.*]] = add <8 x i32> [[BIN_RDX]], [[RDX_SHUF1]]
114; DEFAULT-NEXT:    [[RDX_SHUF3:%.*]] = shufflevector <8 x i32> [[BIN_RDX2]], <8 x i32> undef, <8 x i32> <i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
115; DEFAULT-NEXT:    [[BIN_RDX4:%.*]] = add <8 x i32> [[BIN_RDX2]], [[RDX_SHUF3]]
116; DEFAULT-NEXT:    [[TMP3:%.*]] = extractelement <8 x i32> [[BIN_RDX4]], i32 0
117; DEFAULT-NEXT:    [[BIN_EXTRA:%.*]] = add i32 [[TMP3]], -5
118; DEFAULT-NEXT:    [[TMP34]] = add i32 [[BIN_EXTRA]], [[TMP17]]
119; DEFAULT-NEXT:    br label [[FOR_BODY]]
120;
121; GATHER-LABEL: @PR32038(
122; GATHER-NEXT:  entry:
123; GATHER-NEXT:    [[TMP0:%.*]] = load i8, i8* getelementptr inbounds ([80 x i8], [80 x i8]* @a, i64 0, i64 1), align 1
124; GATHER-NEXT:    [[TMP1:%.*]] = icmp eq i8 [[TMP0]], 0
125; GATHER-NEXT:    [[TMP2:%.*]] = load i8, i8* getelementptr inbounds ([80 x i8], [80 x i8]* @a, i64 0, i64 2), align 2
126; GATHER-NEXT:    [[TMP3:%.*]] = icmp eq i8 [[TMP2]], 0
127; GATHER-NEXT:    [[TMP4:%.*]] = load i8, i8* getelementptr inbounds ([80 x i8], [80 x i8]* @a, i64 0, i64 3), align 1
128; GATHER-NEXT:    [[TMP5:%.*]] = icmp eq i8 [[TMP4]], 0
129; GATHER-NEXT:    [[TMP6:%.*]] = load i8, i8* getelementptr inbounds ([80 x i8], [80 x i8]* @a, i64 0, i64 4), align 4
130; GATHER-NEXT:    [[TMP7:%.*]] = icmp eq i8 [[TMP6]], 0
131; GATHER-NEXT:    [[TMP8:%.*]] = load i8, i8* getelementptr inbounds ([80 x i8], [80 x i8]* @a, i64 0, i64 5), align 1
132; GATHER-NEXT:    [[TMP9:%.*]] = icmp eq i8 [[TMP8]], 0
133; GATHER-NEXT:    [[TMP10:%.*]] = load i8, i8* getelementptr inbounds ([80 x i8], [80 x i8]* @a, i64 0, i64 6), align 2
134; GATHER-NEXT:    [[TMP11:%.*]] = icmp eq i8 [[TMP10]], 0
135; GATHER-NEXT:    [[TMP12:%.*]] = load i8, i8* getelementptr inbounds ([80 x i8], [80 x i8]* @a, i64 0, i64 7), align 1
136; GATHER-NEXT:    [[TMP13:%.*]] = icmp eq i8 [[TMP12]], 0
137; GATHER-NEXT:    [[TMP14:%.*]] = load i8, i8* getelementptr inbounds ([80 x i8], [80 x i8]* @a, i64 0, i64 8), align 8
138; GATHER-NEXT:    [[TMP15:%.*]] = icmp eq i8 [[TMP14]], 0
139; GATHER-NEXT:    br label [[FOR_BODY:%.*]]
140; GATHER:       for.body:
141; GATHER-NEXT:    [[TMP17:%.*]] = phi i32 [ [[TMP34:%.*]], [[FOR_BODY]] ], [ 0, [[ENTRY:%.*]] ]
142; GATHER-NEXT:    [[TMP19:%.*]] = select i1 [[TMP1]], i32 -720, i32 -80
143; GATHER-NEXT:    [[TMP20:%.*]] = add i32 -5, [[TMP19]]
144; GATHER-NEXT:    [[TMP21:%.*]] = select i1 [[TMP3]], i32 -720, i32 -80
145; GATHER-NEXT:    [[TMP22:%.*]] = add i32 [[TMP20]], [[TMP21]]
146; GATHER-NEXT:    [[TMP23:%.*]] = select i1 [[TMP5]], i32 -720, i32 -80
147; GATHER-NEXT:    [[TMP24:%.*]] = add i32 [[TMP22]], [[TMP23]]
148; GATHER-NEXT:    [[TMP25:%.*]] = select i1 [[TMP7]], i32 -720, i32 -80
149; GATHER-NEXT:    [[TMP26:%.*]] = add i32 [[TMP24]], [[TMP25]]
150; GATHER-NEXT:    [[TMP27:%.*]] = select i1 [[TMP9]], i32 -720, i32 -80
151; GATHER-NEXT:    [[TMP28:%.*]] = add i32 [[TMP26]], [[TMP27]]
152; GATHER-NEXT:    [[TMP29:%.*]] = select i1 [[TMP11]], i32 -720, i32 -80
153; GATHER-NEXT:    [[TMP30:%.*]] = add i32 [[TMP28]], [[TMP29]]
154; GATHER-NEXT:    [[TMP31:%.*]] = select i1 [[TMP13]], i32 -720, i32 -80
155; GATHER-NEXT:    [[TMP32:%.*]] = add i32 [[TMP30]], [[TMP31]]
156; GATHER-NEXT:    [[TMP33:%.*]] = select i1 [[TMP15]], i32 -720, i32 -80
157; GATHER-NEXT:    [[TMP0:%.*]] = insertelement <8 x i32> undef, i32 [[TMP19]], i32 0
158; GATHER-NEXT:    [[TMP1:%.*]] = insertelement <8 x i32> [[TMP0]], i32 [[TMP21]], i32 1
159; GATHER-NEXT:    [[TMP2:%.*]] = insertelement <8 x i32> [[TMP1]], i32 [[TMP23]], i32 2
160; GATHER-NEXT:    [[TMP3:%.*]] = insertelement <8 x i32> [[TMP2]], i32 [[TMP25]], i32 3
161; GATHER-NEXT:    [[TMP4:%.*]] = insertelement <8 x i32> [[TMP3]], i32 [[TMP27]], i32 4
162; GATHER-NEXT:    [[TMP5:%.*]] = insertelement <8 x i32> [[TMP4]], i32 [[TMP29]], i32 5
163; GATHER-NEXT:    [[TMP6:%.*]] = insertelement <8 x i32> [[TMP5]], i32 [[TMP31]], i32 6
164; GATHER-NEXT:    [[TMP7:%.*]] = insertelement <8 x i32> [[TMP6]], i32 [[TMP33]], i32 7
165; GATHER-NEXT:    [[RDX_SHUF:%.*]] = shufflevector <8 x i32> [[TMP7]], <8 x i32> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef>
166; GATHER-NEXT:    [[BIN_RDX:%.*]] = add <8 x i32> [[TMP7]], [[RDX_SHUF]]
167; GATHER-NEXT:    [[RDX_SHUF1:%.*]] = shufflevector <8 x i32> [[BIN_RDX]], <8 x i32> undef, <8 x i32> <i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
168; GATHER-NEXT:    [[BIN_RDX2:%.*]] = add <8 x i32> [[BIN_RDX]], [[RDX_SHUF1]]
169; GATHER-NEXT:    [[RDX_SHUF3:%.*]] = shufflevector <8 x i32> [[BIN_RDX2]], <8 x i32> undef, <8 x i32> <i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
170; GATHER-NEXT:    [[BIN_RDX4:%.*]] = add <8 x i32> [[BIN_RDX2]], [[RDX_SHUF3]]
171; GATHER-NEXT:    [[TMP8:%.*]] = extractelement <8 x i32> [[BIN_RDX4]], i32 0
172; GATHER-NEXT:    [[BIN_EXTRA:%.*]] = add i32 [[TMP8]], -5
173; GATHER-NEXT:    [[TMP34]] = add i32 [[BIN_EXTRA]], [[TMP17]]
174; GATHER-NEXT:    br label [[FOR_BODY]]
175;
176; MAX-COST-LABEL: @PR32038(
177entry:
178  %tmp0 = load i8, i8* getelementptr inbounds ([80 x i8], [80 x i8]* @a, i64 0, i64 1), align 1
179  %tmp1 = icmp eq i8 %tmp0, 0
180  %tmp2 = load i8, i8* getelementptr inbounds ([80 x i8], [80 x i8]* @a, i64 0, i64 2), align 2
181  %tmp3 = icmp eq i8 %tmp2, 0
182  %tmp4 = load i8, i8* getelementptr inbounds ([80 x i8], [80 x i8]* @a, i64 0, i64 3), align 1
183  %tmp5 = icmp eq i8 %tmp4, 0
184  %tmp6 = load i8, i8* getelementptr inbounds ([80 x i8], [80 x i8]* @a, i64 0, i64 4), align 4
185  %tmp7 = icmp eq i8 %tmp6, 0
186  %tmp8 = load i8, i8* getelementptr inbounds ([80 x i8], [80 x i8]* @a, i64 0, i64 5), align 1
187  %tmp9 = icmp eq i8 %tmp8, 0
188  %tmp10 = load i8, i8* getelementptr inbounds ([80 x i8], [80 x i8]* @a, i64 0, i64 6), align 2
189  %tmp11 = icmp eq i8 %tmp10, 0
190  %tmp12 = load i8, i8* getelementptr inbounds ([80 x i8], [80 x i8]* @a, i64 0, i64 7), align 1
191  %tmp13 = icmp eq i8 %tmp12, 0
192  %tmp14 = load i8, i8* getelementptr inbounds ([80 x i8], [80 x i8]* @a, i64 0, i64 8), align 8
193  %tmp15 = icmp eq i8 %tmp14, 0
194  br label %for.body
195
196for.body:
197  %tmp17 = phi i32 [ %tmp34, %for.body ], [ 0, %entry ]
198  %tmp19 = select i1 %tmp1, i32 -720, i32 -80
199  %tmp20 = add i32 -5, %tmp19
200  %tmp21 = select i1 %tmp3, i32 -720, i32 -80
201  %tmp22 = add i32 %tmp20, %tmp21
202  %tmp23 = select i1 %tmp5, i32 -720, i32 -80
203  %tmp24 = add i32 %tmp22, %tmp23
204  %tmp25 = select i1 %tmp7, i32 -720, i32 -80
205  %tmp26 = add i32 %tmp24, %tmp25
206  %tmp27 = select i1 %tmp9, i32 -720, i32 -80
207  %tmp28 = add i32 %tmp26, %tmp27
208  %tmp29 = select i1 %tmp11, i32 -720, i32 -80
209  %tmp30 = add i32 %tmp28, %tmp29
210  %tmp31 = select i1 %tmp13, i32 -720, i32 -80
211  %tmp32 = add i32 %tmp30, %tmp31
212  %tmp33 = select i1 %tmp15, i32 -720, i32 -80
213  %tmp34 = add i32 %tmp32, %tmp33
214  br label %for.body
215}
216