xref: /llvm-project/llvm/test/CodeGen/Thumb2/mve-blockplacement.ll (revision eeac0ffaf46cf9f9b0f680b9940cc4b68a0286d8) 
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc -mtriple=thumbv8.1m.main-none-none-eabi -verify-machineinstrs -mattr=+mve %s -o - | FileCheck %s
3
4@var_36 = hidden local_unnamed_addr global i8 0, align 1
5@arr_61 = hidden local_unnamed_addr global [1 x i32] zeroinitializer, align 4
6
7define i32 @test(i8 zeroext %var_2, i16 signext %var_15, ptr %arr_60) {
8; CHECK-LABEL: test:
9; CHECK:       @ %bb.0: @ %entry
10; CHECK-NEXT:    .save {r4, r5, r6, r7, r8, r9, r10, r11, lr}
11; CHECK-NEXT:    push.w {r4, r5, r6, r7, r8, r9, r10, r11, lr}
12; CHECK-NEXT:    .pad #16
13; CHECK-NEXT:    sub sp, #16
14; CHECK-NEXT:    cmp r0, #0
15; CHECK-NEXT:    beq.w .LBB0_11
16; CHECK-NEXT:  @ %bb.1: @ %for.cond1.preheader
17; CHECK-NEXT:    cmp r2, #0
18; CHECK-NEXT:    beq.w .LBB0_12
19; CHECK-NEXT:  @ %bb.2: @ %for.cond1.preheader1
20; CHECK-NEXT:    addw r0, r2, #441
21; CHECK-NEXT:    str r0, [sp, #8] @ 4-byte Spill
22; CHECK-NEXT:    addw r0, r2, #419
23; CHECK-NEXT:    str r0, [sp, #4] @ 4-byte Spill
24; CHECK-NEXT:    addw r0, r2, #397
25; CHECK-NEXT:    movw r5, :lower16:var_36
26; CHECK-NEXT:    str r0, [sp] @ 4-byte Spill
27; CHECK-NEXT:    movw r0, #27476
28; CHECK-NEXT:    movt r5, :upper16:var_36
29; CHECK-NEXT:    sdiv r1, r0, r1
30; CHECK-NEXT:    str r2, [sp, #12] @ 4-byte Spill
31; CHECK-NEXT:  .LBB0_3: @ %for.cond1
32; CHECK-NEXT:    @ =>This Loop Header: Depth=1
33; CHECK-NEXT:    @ Child Loop BB0_4 Depth 2
34; CHECK-NEXT:    @ Child Loop BB0_5 Depth 3
35; CHECK-NEXT:    @ Child Loop BB0_7 Depth 3
36; CHECK-NEXT:    @ Child Loop BB0_9 Depth 3
37; CHECK-NEXT:    ldr.w r12, [sp, #8] @ 4-byte Reload
38; CHECK-NEXT:    mov.w r10, #0
39; CHECK-NEXT:    ldrd r8, r0, [sp] @ 8-byte Folded Reload
40; CHECK-NEXT:  .LBB0_4: @ %for.cond6.preheader
41; CHECK-NEXT:    @ Parent Loop BB0_3 Depth=1
42; CHECK-NEXT:    @ => This Loop Header: Depth=2
43; CHECK-NEXT:    @ Child Loop BB0_5 Depth 3
44; CHECK-NEXT:    @ Child Loop BB0_7 Depth 3
45; CHECK-NEXT:    @ Child Loop BB0_9 Depth 3
46; CHECK-NEXT:    movw r3, :lower16:arr_61
47; CHECK-NEXT:    movt r3, :upper16:arr_61
48; CHECK-NEXT:    add.w r11, r3, #4
49; CHECK-NEXT:    movs r3, #11
50; CHECK-NEXT:    dls lr, r3
51; CHECK-NEXT:    mov r4, r11
52; CHECK-NEXT:    mov r3, r8
53; CHECK-NEXT:  .LBB0_5: @ %for.body10
54; CHECK-NEXT:    @ Parent Loop BB0_3 Depth=1
55; CHECK-NEXT:    @ Parent Loop BB0_4 Depth=2
56; CHECK-NEXT:    @ => This Inner Loop Header: Depth=3
57; CHECK-NEXT:    str r2, [r4, #-4]
58; CHECK-NEXT:    ldrb r6, [r3, #-1]
59; CHECK-NEXT:    cmp r6, #0
60; CHECK-NEXT:    ite ne
61; CHECK-NEXT:    sxthne.w r9, r1
62; CHECK-NEXT:    moveq.w r9, #0
63; CHECK-NEXT:    add.w r6, r2, #396
64; CHECK-NEXT:    cmp.w r9, #0
65; CHECK-NEXT:    str r6, [r4]
66; CHECK-NEXT:    cset r6, ne
67; CHECK-NEXT:    strb r6, [r5]
68; CHECK-NEXT:    add.w r2, r2, #792
69; CHECK-NEXT:    ldrb r6, [r3]
70; CHECK-NEXT:    adds r4, #8
71; CHECK-NEXT:    adds r3, #2
72; CHECK-NEXT:    cmp r6, #0
73; CHECK-NEXT:    ite ne
74; CHECK-NEXT:    sxthne r6, r1
75; CHECK-NEXT:    moveq r6, #0
76; CHECK-NEXT:    cmp r6, #0
77; CHECK-NEXT:    cset r6, ne
78; CHECK-NEXT:    strb r6, [r5]
79; CHECK-NEXT:    le lr, .LBB0_5
80; CHECK-NEXT:  @ %bb.6: @ %for.cond.cleanup9
81; CHECK-NEXT:    @ in Loop: Header=BB0_4 Depth=2
82; CHECK-NEXT:    movs r2, #11
83; CHECK-NEXT:    mov r4, r11
84; CHECK-NEXT:    dls lr, r2
85; CHECK-NEXT:    mov r3, r0
86; CHECK-NEXT:    ldr r2, [sp, #12] @ 4-byte Reload
87; CHECK-NEXT:  .LBB0_7: @ %for.body10.1
88; CHECK-NEXT:    @ Parent Loop BB0_3 Depth=1
89; CHECK-NEXT:    @ Parent Loop BB0_4 Depth=2
90; CHECK-NEXT:    @ => This Inner Loop Header: Depth=3
91; CHECK-NEXT:    str r2, [r4, #-4]
92; CHECK-NEXT:    add.w r7, r2, #396
93; CHECK-NEXT:    ldrb r6, [r3, #-1]
94; CHECK-NEXT:    add.w r2, r2, #792
95; CHECK-NEXT:    cmp r6, #0
96; CHECK-NEXT:    ite ne
97; CHECK-NEXT:    sxthne r6, r1
98; CHECK-NEXT:    moveq r6, #0
99; CHECK-NEXT:    cmp r6, #0
100; CHECK-NEXT:    str r7, [r4]
101; CHECK-NEXT:    cset r6, ne
102; CHECK-NEXT:    adds r4, #8
103; CHECK-NEXT:    strb r6, [r5]
104; CHECK-NEXT:    ldrb r6, [r3]
105; CHECK-NEXT:    adds r3, #2
106; CHECK-NEXT:    cmp r6, #0
107; CHECK-NEXT:    ite ne
108; CHECK-NEXT:    sxthne r6, r1
109; CHECK-NEXT:    moveq r6, #0
110; CHECK-NEXT:    cmp r6, #0
111; CHECK-NEXT:    cset r6, ne
112; CHECK-NEXT:    strb r6, [r5]
113; CHECK-NEXT:    le lr, .LBB0_7
114; CHECK-NEXT:  @ %bb.8: @ %for.cond.cleanup9.1
115; CHECK-NEXT:    @ in Loop: Header=BB0_4 Depth=2
116; CHECK-NEXT:    movs r2, #11
117; CHECK-NEXT:    mov r3, r12
118; CHECK-NEXT:    dls lr, r2
119; CHECK-NEXT:    ldr r2, [sp, #12] @ 4-byte Reload
120; CHECK-NEXT:  .LBB0_9: @ %for.body10.2
121; CHECK-NEXT:    @ Parent Loop BB0_3 Depth=1
122; CHECK-NEXT:    @ Parent Loop BB0_4 Depth=2
123; CHECK-NEXT:    @ => This Inner Loop Header: Depth=3
124; CHECK-NEXT:    str r2, [r11, #-4]
125; CHECK-NEXT:    add.w r6, r2, #396
126; CHECK-NEXT:    ldrb r4, [r3, #-1]
127; CHECK-NEXT:    add.w r2, r2, #792
128; CHECK-NEXT:    cmp r4, #0
129; CHECK-NEXT:    ite ne
130; CHECK-NEXT:    sxthne r4, r1
131; CHECK-NEXT:    moveq r4, #0
132; CHECK-NEXT:    cmp r4, #0
133; CHECK-NEXT:    str.w r6, [r11]
134; CHECK-NEXT:    cset r4, ne
135; CHECK-NEXT:    add.w r11, r11, #8
136; CHECK-NEXT:    strb r4, [r5]
137; CHECK-NEXT:    ldrb r4, [r3]
138; CHECK-NEXT:    adds r3, #2
139; CHECK-NEXT:    cmp r4, #0
140; CHECK-NEXT:    ite ne
141; CHECK-NEXT:    sxthne r4, r1
142; CHECK-NEXT:    moveq r4, #0
143; CHECK-NEXT:    cmp r4, #0
144; CHECK-NEXT:    cset r4, ne
145; CHECK-NEXT:    strb r4, [r5]
146; CHECK-NEXT:    le lr, .LBB0_9
147; CHECK-NEXT:  @ %bb.10: @ %for.cond.cleanup9.2
148; CHECK-NEXT:    @ in Loop: Header=BB0_4 Depth=2
149; CHECK-NEXT:    add.w r2, r10, #3
150; CHECK-NEXT:    add.w r12, r12, #66
151; CHECK-NEXT:    adds r0, #66
152; CHECK-NEXT:    add.w r8, r8, #66
153; CHECK-NEXT:    uxtb.w r10, r2
154; CHECK-NEXT:    ldr r2, [sp, #12] @ 4-byte Reload
155; CHECK-NEXT:    cmp.w r10, #18
156; CHECK-NEXT:    blo.w .LBB0_4
157; CHECK-NEXT:    b .LBB0_3
158; CHECK-NEXT:  .LBB0_11: @ %for.cond.cleanup
159; CHECK-NEXT:    add sp, #16
160; CHECK-NEXT:    pop.w {r4, r5, r6, r7, r8, r9, r10, r11, pc}
161; CHECK-NEXT:  .LBB0_12: @ %for.cond1.us.preheader
162; CHECK-NEXT:    movw r0, :lower16:arr_61
163; CHECK-NEXT:    movs r1, #0
164; CHECK-NEXT:    movt r0, :upper16:arr_61
165; CHECK-NEXT:    str r1, [r0, #84]
166; CHECK-NEXT:    .inst.n 0xdefe
167entry:
168  %tobool.not = icmp eq i8 %var_2, 0
169  br i1 %tobool.not, label %for.cond.cleanup, label %for.cond1.preheader
170
171for.cond1.preheader:                              ; preds = %entry
172  %cmp11.not = icmp eq ptr %arr_60, null
173  br i1 %cmp11.not, label %for.cond1.us.preheader, label %for.cond1
174
175for.cond1.us.preheader:                           ; preds = %for.cond1.preheader
176  store i32 0, ptr getelementptr ([1 x i32], ptr @arr_61, i32 21, i32 0), align 4
177  call void @llvm.trap()
178  unreachable
179
180for.cond.cleanup:                                 ; preds = %entry
181  ret i32 undef
182
183for.cond1:                                        ; preds = %for.cond.cleanup9.2, %for.cond1.preheader
184  br label %for.cond6.preheader
185
186for.cond6.preheader:                              ; preds = %for.cond.cleanup9.2, %for.cond1
187  %conv45 = phi i32 [ 0, %for.cond1 ], [ %conv.2, %for.cond.cleanup9.2 ]
188  br label %for.body10
189
190for.cond.cleanup9:                                ; preds = %cond.end22.1
191  %add27 = add nuw nsw i32 %conv45, 1
192  %conv = and i32 %add27, 255
193  br label %for.body10.1
194
195for.body10:                                       ; preds = %cond.end22.1, %for.cond6.preheader
196  %i_15.044 = phi i32 [ 0, %for.cond6.preheader ], [ %add.1, %cond.end22.1 ]
197  %arraydecay = getelementptr inbounds [18 x [22 x i8]], ptr %arr_60, i32 %i_15.044, i32 0
198  %0 = ptrtoint ptr %arraydecay to i32
199  %arrayidx13 = getelementptr inbounds [1 x i32], ptr @arr_61, i32 0, i32 %i_15.044
200  store i32 %0, ptr %arrayidx13, align 4
201  %arrayidx16 = getelementptr inbounds [18 x [22 x i8]], ptr %arr_60, i32 1, i32 %conv45, i32 %i_15.044
202  %1 = load i8, ptr %arrayidx16, align 1
203  %tobool18.not = icmp eq i8 %1, 0
204  br i1 %tobool18.not, label %cond.end22, label %cond.true19
205
206cond.true19:                                      ; preds = %for.body10
207  %div43 = sdiv i16 27476, %var_15
208  %div.sext = sext i16 %div43 to i32
209  br label %cond.end22
210
211cond.end22:                                       ; preds = %for.body10, %cond.true19
212  %cond23 = phi i32 [ %div.sext, %cond.true19 ], [ 0, %for.body10 ]
213  %tobool24 = icmp ne i32 %cond23, 0
214  %frombool = zext i1 %tobool24 to i8
215  store i8 %frombool, ptr @var_36, align 1
216  %add = or disjoint i32 %i_15.044, 1
217  %arraydecay.1 = getelementptr inbounds [18 x [22 x i8]], ptr %arr_60, i32 %add, i32 0
218  %2 = ptrtoint ptr %arraydecay.1 to i32
219  %arrayidx13.1 = getelementptr inbounds [1 x i32], ptr @arr_61, i32 0, i32 %add
220  store i32 %2, ptr %arrayidx13.1, align 4
221  %arrayidx16.1 = getelementptr inbounds [18 x [22 x i8]], ptr %arr_60, i32 1, i32 %conv45, i32 %add
222  %3 = load i8, ptr %arrayidx16.1, align 1
223  %tobool18.not.1 = icmp eq i8 %3, 0
224  br i1 %tobool18.not.1, label %cond.end22.1, label %cond.true19.1
225
226cond.true19.1:                                    ; preds = %cond.end22
227  %div43.1 = sdiv i16 27476, %var_15
228  %div.sext.1 = sext i16 %div43.1 to i32
229  br label %cond.end22.1
230
231cond.end22.1:                                     ; preds = %cond.true19.1, %cond.end22
232  %cond23.1 = phi i32 [ %div.sext.1, %cond.true19.1 ], [ 0, %cond.end22 ]
233  %tobool24.1 = icmp ne i32 %cond23.1, 0
234  %frombool.1 = zext i1 %tobool24.1 to i8
235  store i8 %frombool.1, ptr @var_36, align 1
236  %add.1 = add nuw nsw i32 %i_15.044, 2
237  %exitcond105.not.1 = icmp eq i32 %add.1, 22
238  br i1 %exitcond105.not.1, label %for.cond.cleanup9, label %for.body10
239
240for.body10.1:                                     ; preds = %cond.end22.1.1, %for.cond.cleanup9
241  %i_15.044.1 = phi i32 [ 0, %for.cond.cleanup9 ], [ %add.1.1, %cond.end22.1.1 ]
242  %arraydecay.1108 = getelementptr inbounds [18 x [22 x i8]], ptr %arr_60, i32 %i_15.044.1, i32 0
243  %4 = ptrtoint ptr %arraydecay.1108 to i32
244  %arrayidx13.1109 = getelementptr inbounds [1 x i32], ptr @arr_61, i32 0, i32 %i_15.044.1
245  store i32 %4, ptr %arrayidx13.1109, align 4
246  %arrayidx16.1110 = getelementptr inbounds [18 x [22 x i8]], ptr %arr_60, i32 1, i32 %conv, i32 %i_15.044.1
247  %5 = load i8, ptr %arrayidx16.1110, align 1
248  %tobool18.not.1111 = icmp eq i8 %5, 0
249  br i1 %tobool18.not.1111, label %cond.end22.1119, label %cond.true19.1114
250
251cond.true19.1114:                                 ; preds = %for.body10.1
252  %div43.1112 = sdiv i16 27476, %var_15
253  %div.sext.1113 = sext i16 %div43.1112 to i32
254  br label %cond.end22.1119
255
256cond.end22.1119:                                  ; preds = %cond.true19.1114, %for.body10.1
257  %cond23.1115 = phi i32 [ %div.sext.1113, %cond.true19.1114 ], [ 0, %for.body10.1 ]
258  %tobool24.1116 = icmp ne i32 %cond23.1115, 0
259  %frombool.1117 = zext i1 %tobool24.1116 to i8
260  store i8 %frombool.1117, ptr @var_36, align 1
261  %add.1118 = or disjoint i32 %i_15.044.1, 1
262  %arraydecay.1.1 = getelementptr inbounds [18 x [22 x i8]], ptr %arr_60, i32 %add.1118, i32 0
263  %6 = ptrtoint ptr %arraydecay.1.1 to i32
264  %arrayidx13.1.1 = getelementptr inbounds [1 x i32], ptr @arr_61, i32 0, i32 %add.1118
265  store i32 %6, ptr %arrayidx13.1.1, align 4
266  %arrayidx16.1.1 = getelementptr inbounds [18 x [22 x i8]], ptr %arr_60, i32 1, i32 %conv, i32 %add.1118
267  %7 = load i8, ptr %arrayidx16.1.1, align 1
268  %tobool18.not.1.1 = icmp eq i8 %7, 0
269  br i1 %tobool18.not.1.1, label %cond.end22.1.1, label %cond.true19.1.1
270
271cond.true19.1.1:                                  ; preds = %cond.end22.1119
272  %div43.1.1 = sdiv i16 27476, %var_15
273  %div.sext.1.1 = sext i16 %div43.1.1 to i32
274  br label %cond.end22.1.1
275
276cond.end22.1.1:                                   ; preds = %cond.true19.1.1, %cond.end22.1119
277  %cond23.1.1 = phi i32 [ %div.sext.1.1, %cond.true19.1.1 ], [ 0, %cond.end22.1119 ]
278  %tobool24.1.1 = icmp ne i32 %cond23.1.1, 0
279  %frombool.1.1 = zext i1 %tobool24.1.1 to i8
280  store i8 %frombool.1.1, ptr @var_36, align 1
281  %add.1.1 = add nuw nsw i32 %i_15.044.1, 2
282  %exitcond105.not.1.1 = icmp eq i32 %add.1.1, 22
283  br i1 %exitcond105.not.1.1, label %for.cond.cleanup9.1, label %for.body10.1
284
285for.cond.cleanup9.1:                              ; preds = %cond.end22.1.1
286  %add27.1 = add nuw nsw i32 %conv45, 2
287  %conv.1 = and i32 %add27.1, 255
288  br label %for.body10.2
289
290for.body10.2:                                     ; preds = %cond.end22.1.2, %for.cond.cleanup9.1
291  %i_15.044.2 = phi i32 [ 0, %for.cond.cleanup9.1 ], [ %add.1.2, %cond.end22.1.2 ]
292  %arraydecay.2 = getelementptr inbounds [18 x [22 x i8]], ptr %arr_60, i32 %i_15.044.2, i32 0
293  %8 = ptrtoint ptr %arraydecay.2 to i32
294  %arrayidx13.2 = getelementptr inbounds [1 x i32], ptr @arr_61, i32 0, i32 %i_15.044.2
295  store i32 %8, ptr %arrayidx13.2, align 4
296  %arrayidx16.2 = getelementptr inbounds [18 x [22 x i8]], ptr %arr_60, i32 1, i32 %conv.1, i32 %i_15.044.2
297  %9 = load i8, ptr %arrayidx16.2, align 1
298  %tobool18.not.2 = icmp eq i8 %9, 0
299  br i1 %tobool18.not.2, label %cond.end22.2, label %cond.true19.2
300
301cond.true19.2:                                    ; preds = %for.body10.2
302  %div43.2 = sdiv i16 27476, %var_15
303  %div.sext.2 = sext i16 %div43.2 to i32
304  br label %cond.end22.2
305
306cond.end22.2:                                     ; preds = %cond.true19.2, %for.body10.2
307  %cond23.2 = phi i32 [ %div.sext.2, %cond.true19.2 ], [ 0, %for.body10.2 ]
308  %tobool24.2 = icmp ne i32 %cond23.2, 0
309  %frombool.2 = zext i1 %tobool24.2 to i8
310  store i8 %frombool.2, ptr @var_36, align 1
311  %add.2 = or disjoint i32 %i_15.044.2, 1
312  %arraydecay.1.2 = getelementptr inbounds [18 x [22 x i8]], ptr %arr_60, i32 %add.2, i32 0
313  %10 = ptrtoint ptr %arraydecay.1.2 to i32
314  %arrayidx13.1.2 = getelementptr inbounds [1 x i32], ptr @arr_61, i32 0, i32 %add.2
315  store i32 %10, ptr %arrayidx13.1.2, align 4
316  %arrayidx16.1.2 = getelementptr inbounds [18 x [22 x i8]], ptr %arr_60, i32 1, i32 %conv.1, i32 %add.2
317  %11 = load i8, ptr %arrayidx16.1.2, align 1
318  %tobool18.not.1.2 = icmp eq i8 %11, 0
319  br i1 %tobool18.not.1.2, label %cond.end22.1.2, label %cond.true19.1.2
320
321cond.true19.1.2:                                  ; preds = %cond.end22.2
322  %div43.1.2 = sdiv i16 27476, %var_15
323  %div.sext.1.2 = sext i16 %div43.1.2 to i32
324  br label %cond.end22.1.2
325
326cond.end22.1.2:                                   ; preds = %cond.true19.1.2, %cond.end22.2
327  %cond23.1.2 = phi i32 [ %div.sext.1.2, %cond.true19.1.2 ], [ 0, %cond.end22.2 ]
328  %tobool24.1.2 = icmp ne i32 %cond23.1.2, 0
329  %frombool.1.2 = zext i1 %tobool24.1.2 to i8
330  store i8 %frombool.1.2, ptr @var_36, align 1
331  %add.1.2 = add nuw nsw i32 %i_15.044.2, 2
332  %exitcond105.not.1.2 = icmp eq i32 %add.1.2, 22
333  br i1 %exitcond105.not.1.2, label %for.cond.cleanup9.2, label %for.body10.2
334
335for.cond.cleanup9.2:                              ; preds = %cond.end22.1.2
336  %add27.2 = add nuw nsw i32 %conv45, 3
337  %conv.2 = and i32 %add27.2, 255
338  %cmp.2 = icmp ult i32 %conv.2, 18
339  br i1 %cmp.2, label %for.cond6.preheader, label %for.cond1
340}
341
342declare void @llvm.trap() #1
343
344
345@b = hidden local_unnamed_addr global i32 0, align 4
346@a = hidden local_unnamed_addr global i32 0, align 4
347@c = hidden local_unnamed_addr global [1 x i32] zeroinitializer, align 4
348
349define i32 @d(i64 %e, i32 %f, i64 %g, i32 %h) {
350; CHECK-LABEL: d:
351; CHECK:       @ %bb.0: @ %entry
352; CHECK-NEXT:    .save {r4, r5, r6, r7, r8, r9, r10, r11, lr}
353; CHECK-NEXT:    push.w {r4, r5, r6, r7, r8, r9, r10, r11, lr}
354; CHECK-NEXT:    .pad #4
355; CHECK-NEXT:    sub sp, #4
356; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
357; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
358; CHECK-NEXT:    .pad #16
359; CHECK-NEXT:    sub sp, #16
360; CHECK-NEXT:    mov lr, r0
361; CHECK-NEXT:    subs r0, #1
362; CHECK-NEXT:    sbcs r0, r1, #0
363; CHECK-NEXT:    blt.w .LBB1_28
364; CHECK-NEXT:  @ %bb.1: @ %for.cond2.preheader.lr.ph
365; CHECK-NEXT:    movs r0, #1
366; CHECK-NEXT:    cmp r2, #1
367; CHECK-NEXT:    csel r7, r2, r0, lt
368; CHECK-NEXT:    mov r12, r1
369; CHECK-NEXT:    mov r1, r7
370; CHECK-NEXT:    cmp r7, #3
371; CHECK-NEXT:    it ls
372; CHECK-NEXT:    movls r1, #3
373; CHECK-NEXT:    mov r4, r2
374; CHECK-NEXT:    subs r1, r1, r7
375; CHECK-NEXT:    movw r2, #43691
376; CHECK-NEXT:    adds r1, #2
377; CHECK-NEXT:    movt r2, #43690
378; CHECK-NEXT:    ldr r6, [sp, #128]
379; CHECK-NEXT:    movw r8, :lower16:c
380; CHECK-NEXT:    umull r1, r2, r1, r2
381; CHECK-NEXT:    movt r8, :upper16:c
382; CHECK-NEXT:    movs r1, #4
383; CHECK-NEXT:    @ implicit-def: $r10
384; CHECK-NEXT:    @ implicit-def: $r5
385; CHECK-NEXT:    @ implicit-def: $r11
386; CHECK-NEXT:    mov.w r9, #12
387; CHECK-NEXT:    str r4, [sp, #12] @ 4-byte Spill
388; CHECK-NEXT:    add.w r1, r1, r2, lsr #1
389; CHECK-NEXT:    add.w r0, r0, r2, lsr #1
390; CHECK-NEXT:    bic r3, r1, #3
391; CHECK-NEXT:    adr r1, .LCPI1_0
392; CHECK-NEXT:    vldrw.u32 q0, [r1]
393; CHECK-NEXT:    adr r1, .LCPI1_1
394; CHECK-NEXT:    vldrw.u32 q5, [r1]
395; CHECK-NEXT:    vdup.32 q6, r0
396; CHECK-NEXT:    vadd.i32 q4, q0, r7
397; CHECK-NEXT:    vdup.32 q7, r0
398; CHECK-NEXT:    strd r3, r7, [sp, #4] @ 8-byte Folded Spill
399; CHECK-NEXT:    b .LBB1_6
400; CHECK-NEXT:  .LBB1_2: @ %for.body6.preheader
401; CHECK-NEXT:    @ in Loop: Header=BB1_6 Depth=1
402; CHECK-NEXT:    mov r0, r11
403; CHECK-NEXT:    cmn.w r11, #4
404; CHECK-NEXT:    it le
405; CHECK-NEXT:    mvnle r0, #3
406; CHECK-NEXT:    movw r2, #18725
407; CHECK-NEXT:    adds r0, #6
408; CHECK-NEXT:    movt r2, #9362
409; CHECK-NEXT:    sub.w r1, r0, r11
410; CHECK-NEXT:    mov r10, r3
411; CHECK-NEXT:    umull r2, r3, r1, r2
412; CHECK-NEXT:    subs r2, r1, r3
413; CHECK-NEXT:    add.w r2, r3, r2, lsr #1
414; CHECK-NEXT:    lsrs r3, r2, #2
415; CHECK-NEXT:    lsls r3, r3, #3
416; CHECK-NEXT:    sub.w r2, r3, r2, lsr #2
417; CHECK-NEXT:    subs r1, r2, r1
418; CHECK-NEXT:    mov r3, r10
419; CHECK-NEXT:    add r0, r1
420; CHECK-NEXT:  .LBB1_3: @ %for.cond.cleanup5.loopexit134.split.loop.exit139
421; CHECK-NEXT:    @ in Loop: Header=BB1_6 Depth=1
422; CHECK-NEXT:    add.w r11, r0, #7
423; CHECK-NEXT:  .LBB1_4: @ %for.cond.cleanup5
424; CHECK-NEXT:    @ in Loop: Header=BB1_6 Depth=1
425; CHECK-NEXT:    mov.w r10, #0
426; CHECK-NEXT:  .LBB1_5: @ %for.cond.cleanup5
427; CHECK-NEXT:    @ in Loop: Header=BB1_6 Depth=1
428; CHECK-NEXT:    adds r5, #2
429; CHECK-NEXT:    subs.w r1, r5, lr
430; CHECK-NEXT:    asr.w r0, r5, #31
431; CHECK-NEXT:    sbcs.w r0, r0, r12
432; CHECK-NEXT:    bge.w .LBB1_28
433; CHECK-NEXT:  .LBB1_6: @ %for.cond2.preheader
434; CHECK-NEXT:    @ =>This Loop Header: Depth=1
435; CHECK-NEXT:    @ Child Loop BB1_19 Depth 2
436; CHECK-NEXT:    @ Child Loop BB1_10 Depth 2
437; CHECK-NEXT:    @ Child Loop BB1_12 Depth 3
438; CHECK-NEXT:    @ Child Loop BB1_14 Depth 3
439; CHECK-NEXT:    cmp.w r11, #2
440; CHECK-NEXT:    bgt .LBB1_5
441; CHECK-NEXT:  @ %bb.7: @ %for.body6.lr.ph
442; CHECK-NEXT:    @ in Loop: Header=BB1_6 Depth=1
443; CHECK-NEXT:    cmp r7, #5
444; CHECK-NEXT:    bhi .LBB1_17
445; CHECK-NEXT:  @ %bb.8: @ %for.body6.us.preheader
446; CHECK-NEXT:    @ in Loop: Header=BB1_6 Depth=1
447; CHECK-NEXT:    ldrd r2, r3, [sp, #120]
448; CHECK-NEXT:    movs r0, #32
449; CHECK-NEXT:    movs r1, #0
450; CHECK-NEXT:    mov r4, r6
451; CHECK-NEXT:    mov r7, r12
452; CHECK-NEXT:    mov r6, lr
453; CHECK-NEXT:    bl __aeabi_ldivmod
454; CHECK-NEXT:    mov lr, r6
455; CHECK-NEXT:    mov r6, r4
456; CHECK-NEXT:    mov r12, r7
457; CHECK-NEXT:    ldr r3, [sp, #4] @ 4-byte Reload
458; CHECK-NEXT:    ldr r4, [sp, #12] @ 4-byte Reload
459; CHECK-NEXT:    vdup.32 q0, r2
460; CHECK-NEXT:    ldr r7, [sp, #8] @ 4-byte Reload
461; CHECK-NEXT:    mov r0, r11
462; CHECK-NEXT:    b .LBB1_10
463; CHECK-NEXT:  .LBB1_9: @ %for.cond.cleanup17.us
464; CHECK-NEXT:    @ in Loop: Header=BB1_10 Depth=2
465; CHECK-NEXT:    add.w r11, r0, #7
466; CHECK-NEXT:    cmn.w r0, #4
467; CHECK-NEXT:    mov.w r10, #0
468; CHECK-NEXT:    mov r0, r11
469; CHECK-NEXT:    bge .LBB1_5
470; CHECK-NEXT:  .LBB1_10: @ %for.body6.us
471; CHECK-NEXT:    @ Parent Loop BB1_6 Depth=1
472; CHECK-NEXT:    @ => This Loop Header: Depth=2
473; CHECK-NEXT:    @ Child Loop BB1_12 Depth 3
474; CHECK-NEXT:    @ Child Loop BB1_14 Depth 3
475; CHECK-NEXT:    movs r1, #0
476; CHECK-NEXT:    cbz r4, .LBB1_13
477; CHECK-NEXT:  @ %bb.11: @ %for.body13.us51.preheader
478; CHECK-NEXT:    @ in Loop: Header=BB1_10 Depth=2
479; CHECK-NEXT:    movw r2, :lower16:a
480; CHECK-NEXT:    vmov q1, q4
481; CHECK-NEXT:    movt r2, :upper16:a
482; CHECK-NEXT:    str r1, [r2]
483; CHECK-NEXT:    movw r2, :lower16:b
484; CHECK-NEXT:    movt r2, :upper16:b
485; CHECK-NEXT:    str r1, [r2]
486; CHECK-NEXT:    mov r2, r3
487; CHECK-NEXT:  .LBB1_12: @ %vector.body111
488; CHECK-NEXT:    @ Parent Loop BB1_6 Depth=1
489; CHECK-NEXT:    @ Parent Loop BB1_10 Depth=2
490; CHECK-NEXT:    @ => This Inner Loop Header: Depth=3
491; CHECK-NEXT:    vqadd.u32 q2, q5, r1
492; CHECK-NEXT:    adds r1, #4
493; CHECK-NEXT:    vcmp.u32 hi, q7, q2
494; CHECK-NEXT:    vshl.i32 q2, q1, #2
495; CHECK-NEXT:    subs r2, #4
496; CHECK-NEXT:    vadd.i32 q2, q2, r8
497; CHECK-NEXT:    vadd.i32 q1, q1, r9
498; CHECK-NEXT:    vpst
499; CHECK-NEXT:    vstrwt.32 q0, [q2]
500; CHECK-NEXT:    bne .LBB1_12
501; CHECK-NEXT:    b .LBB1_15
502; CHECK-NEXT:  .LBB1_13: @ %vector.body.preheader
503; CHECK-NEXT:    @ in Loop: Header=BB1_10 Depth=2
504; CHECK-NEXT:    mov r2, r3
505; CHECK-NEXT:    vmov q1, q4
506; CHECK-NEXT:  .LBB1_14: @ %vector.body
507; CHECK-NEXT:    @ Parent Loop BB1_6 Depth=1
508; CHECK-NEXT:    @ Parent Loop BB1_10 Depth=2
509; CHECK-NEXT:    @ => This Inner Loop Header: Depth=3
510; CHECK-NEXT:    vqadd.u32 q2, q5, r1
511; CHECK-NEXT:    adds r1, #4
512; CHECK-NEXT:    vcmp.u32 hi, q6, q2
513; CHECK-NEXT:    vshl.i32 q2, q1, #2
514; CHECK-NEXT:    subs r2, #4
515; CHECK-NEXT:    vadd.i32 q2, q2, r8
516; CHECK-NEXT:    vadd.i32 q1, q1, r9
517; CHECK-NEXT:    vpst
518; CHECK-NEXT:    vstrwt.32 q0, [q2]
519; CHECK-NEXT:    bne .LBB1_14
520; CHECK-NEXT:  .LBB1_15: @ %for.cond9.for.cond15.preheader_crit_edge.us
521; CHECK-NEXT:    @ in Loop: Header=BB1_10 Depth=2
522; CHECK-NEXT:    cmp r6, #0
523; CHECK-NEXT:    beq .LBB1_9
524; CHECK-NEXT:  @ %bb.16: @ %for.cond9.for.cond15.preheader_crit_edge.us
525; CHECK-NEXT:    @ in Loop: Header=BB1_10 Depth=2
526; CHECK-NEXT:    eor r1, r10, #1
527; CHECK-NEXT:    lsls r1, r1, #31
528; CHECK-NEXT:    bne .LBB1_9
529; CHECK-NEXT:    b .LBB1_26
530; CHECK-NEXT:  .LBB1_17: @ %for.body6.lr.ph.split
531; CHECK-NEXT:    @ in Loop: Header=BB1_6 Depth=1
532; CHECK-NEXT:    cmp r6, #0
533; CHECK-NEXT:    beq.w .LBB1_2
534; CHECK-NEXT:  @ %bb.18: @ in Loop: Header=BB1_6 Depth=1
535; CHECK-NEXT:    mov r0, r11
536; CHECK-NEXT:  .LBB1_19: @ %for.body6.us60
537; CHECK-NEXT:    @ Parent Loop BB1_6 Depth=1
538; CHECK-NEXT:    @ => This Inner Loop Header: Depth=2
539; CHECK-NEXT:    lsls.w r1, r10, #31
540; CHECK-NEXT:    bne .LBB1_27
541; CHECK-NEXT:  @ %bb.20: @ %for.cond.cleanup17.us63
542; CHECK-NEXT:    @ in Loop: Header=BB1_19 Depth=2
543; CHECK-NEXT:    cmn.w r0, #4
544; CHECK-NEXT:    bge.w .LBB1_3
545; CHECK-NEXT:  @ %bb.21: @ %for.cond.cleanup17.us63.1
546; CHECK-NEXT:    @ in Loop: Header=BB1_19 Depth=2
547; CHECK-NEXT:    cmn.w r0, #12
548; CHECK-NEXT:    bgt .LBB1_24
549; CHECK-NEXT:  @ %bb.22: @ %for.cond.cleanup17.us63.2
550; CHECK-NEXT:    @ in Loop: Header=BB1_19 Depth=2
551; CHECK-NEXT:    cmn.w r0, #19
552; CHECK-NEXT:    bgt .LBB1_25
553; CHECK-NEXT:  @ %bb.23: @ %for.cond.cleanup17.us63.3
554; CHECK-NEXT:    @ in Loop: Header=BB1_19 Depth=2
555; CHECK-NEXT:    add.w r11, r0, #28
556; CHECK-NEXT:    cmn.w r0, #25
557; CHECK-NEXT:    mov.w r10, #0
558; CHECK-NEXT:    mov r0, r11
559; CHECK-NEXT:    blt .LBB1_19
560; CHECK-NEXT:    b .LBB1_5
561; CHECK-NEXT:  .LBB1_24: @ %for.cond.cleanup5.loopexit134.split.loop.exit137
562; CHECK-NEXT:    @ in Loop: Header=BB1_6 Depth=1
563; CHECK-NEXT:    add.w r11, r0, #14
564; CHECK-NEXT:    b .LBB1_4
565; CHECK-NEXT:  .LBB1_25: @ %for.cond.cleanup5.loopexit134.split.loop.exit135
566; CHECK-NEXT:    @ in Loop: Header=BB1_6 Depth=1
567; CHECK-NEXT:    add.w r11, r0, #21
568; CHECK-NEXT:    b .LBB1_4
569; CHECK-NEXT:  .LBB1_26: @ %for.inc19.us
570; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
571; CHECK-NEXT:    b .LBB1_26
572; CHECK-NEXT:  .LBB1_27: @ %for.inc19.us66
573; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
574; CHECK-NEXT:    b .LBB1_27
575; CHECK-NEXT:  .LBB1_28: @ %for.cond.cleanup
576; CHECK-NEXT:    add sp, #16
577; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
578; CHECK-NEXT:    add sp, #4
579; CHECK-NEXT:    pop.w {r4, r5, r6, r7, r8, r9, r10, r11, pc}
580; CHECK-NEXT:    .p2align 4
581; CHECK-NEXT:  @ %bb.29:
582; CHECK-NEXT:  .LCPI1_0:
583; CHECK-NEXT:    .long 0 @ 0x0
584; CHECK-NEXT:    .long 3 @ 0x3
585; CHECK-NEXT:    .long 6 @ 0x6
586; CHECK-NEXT:    .long 9 @ 0x9
587; CHECK-NEXT:  .LCPI1_1:
588; CHECK-NEXT:    .long 0 @ 0x0
589; CHECK-NEXT:    .long 1 @ 0x1
590; CHECK-NEXT:    .long 2 @ 0x2
591; CHECK-NEXT:    .long 3 @ 0x3
592entry:
593  %cmp47 = icmp sgt i64 %e, 0
594  br i1 %cmp47, label %for.cond2.preheader.lr.ph, label %for.cond.cleanup
595
596for.cond2.preheader.lr.ph:                        ; preds = %entry
597  %cmp7.inv = icmp slt i32 %f, 1
598  %spec.select = select i1 %cmp7.inv, i32 %f, i32 1
599  %cmp1041 = icmp ult i32 %spec.select, 6
600  %tobool.not = icmp eq i32 %f, 0
601  %tobool20.not97 = icmp eq i32 %h, 0
602  %0 = icmp ugt i32 %spec.select, 3
603  %umax = select i1 %0, i32 %spec.select, i32 3
604  %1 = add i32 %umax, 2
605  %2 = sub i32 %1, %spec.select
606  %3 = udiv i32 %2, 3
607  %4 = add nuw nsw i32 %3, 1
608  %5 = icmp ugt i32 %spec.select, 3
609  %umax112 = select i1 %5, i32 %spec.select, i32 3
610  %6 = add i32 %umax112, 2
611  %7 = sub i32 %6, %spec.select
612  %8 = udiv i32 %7, 3
613  %9 = add nuw nsw i32 %8, 1
614  %n.rnd.up114 = add nuw nsw i32 %8, 4
615  %n.vec116 = and i32 %n.rnd.up114, 2147483644
616  %.splatinsert121 = insertelement <4 x i32> poison, i32 %spec.select, i32 0
617  %.splat122 = shufflevector <4 x i32> %.splatinsert121, <4 x i32> poison, <4 x i32> zeroinitializer
618  %induction123 = add <4 x i32> %.splat122, <i32 0, i32 3, i32 6, i32 9>
619  %n.rnd.up = add nuw nsw i32 %3, 4
620  %n.vec = and i32 %n.rnd.up, 2147483644
621  %.splatinsert = insertelement <4 x i32> poison, i32 %spec.select, i32 0
622  %.splat = shufflevector <4 x i32> %.splatinsert, <4 x i32> poison, <4 x i32> zeroinitializer
623  %induction = add <4 x i32> %.splat, <i32 0, i32 3, i32 6, i32 9>
624  br label %for.cond2.preheader
625
626for.cond2.preheader:                              ; preds = %for.cond2.preheader.lr.ph, %for.cond.cleanup5
627  %l.0.off050 = phi i1 [ undef, %for.cond2.preheader.lr.ph ], [ %l.1.off0.lcssa, %for.cond.cleanup5 ]
628  %i.049 = phi i32 [ undef, %for.cond2.preheader.lr.ph ], [ %add26, %for.cond.cleanup5 ]
629  %j.048 = phi i32 [ undef, %for.cond2.preheader.lr.ph ], [ %j.1.lcssa, %for.cond.cleanup5 ]
630  %cmp343 = icmp slt i32 %j.048, 3
631  br i1 %cmp343, label %for.body6.lr.ph, label %for.cond.cleanup5
632
633for.body6.lr.ph:                                  ; preds = %for.cond2.preheader
634  br i1 %cmp1041, label %for.body6.us.preheader, label %for.body6.lr.ph.split
635
636for.body6.us.preheader:                           ; preds = %for.body6.lr.ph
637  %rem.us = srem i64 32, %g
638  %conv14.us = trunc i64 %rem.us to i32
639  %broadcast.splatinsert131 = insertelement <4 x i32> poison, i32 %conv14.us, i32 0
640  %broadcast.splat132 = shufflevector <4 x i32> %broadcast.splatinsert131, <4 x i32> poison, <4 x i32> zeroinitializer
641  %broadcast.splatinsert107 = insertelement <4 x i32> poison, i32 %conv14.us, i32 0
642  %broadcast.splat108 = shufflevector <4 x i32> %broadcast.splatinsert107, <4 x i32> poison, <4 x i32> zeroinitializer
643  br label %for.body6.us
644
645for.body6.us:                                     ; preds = %for.body6.us.preheader, %for.cond.cleanup17.us
646  %l.1.off045.us = phi i1 [ false, %for.cond.cleanup17.us ], [ %l.0.off050, %for.body6.us.preheader ]
647  %j.144.us = phi i32 [ %add23.us, %for.cond.cleanup17.us ], [ %j.048, %for.body6.us.preheader ]
648  br i1 %tobool.not, label %vector.body, label %for.body13.us51.preheader
649
650vector.body:                                      ; preds = %for.body6.us, %vector.body
651  %index = phi i32 [ %index.next, %vector.body ], [ 0, %for.body6.us ]
652  %vec.ind = phi <4 x i32> [ %vec.ind.next, %vector.body ], [ %induction, %for.body6.us ]
653  %active.lane.mask = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32 %index, i32 %4)
654  %10 = getelementptr inbounds [1 x i32], ptr @c, i32 0, <4 x i32> %vec.ind
655  call void @llvm.masked.scatter.v4i32.v4p0(<4 x i32> %broadcast.splat108, <4 x ptr> %10, i32 4, <4 x i1> %active.lane.mask)
656  %index.next = add i32 %index, 4
657  %vec.ind.next = add <4 x i32> %vec.ind, <i32 12, i32 12, i32 12, i32 12>
658  %11 = icmp eq i32 %index.next, %n.vec
659  br i1 %11, label %for.cond9.for.cond15.preheader_crit_edge.us, label %vector.body
660
661for.body13.us51.preheader:                        ; preds = %for.body6.us
662  store i32 0, ptr @b, align 4
663  store i32 0, ptr @a, align 4
664  br label %vector.body111
665
666vector.body111:                                   ; preds = %vector.body111, %for.body13.us51.preheader
667  %index117 = phi i32 [ 0, %for.body13.us51.preheader ], [ %index.next118, %vector.body111 ]
668  %vec.ind124 = phi <4 x i32> [ %induction123, %for.body13.us51.preheader ], [ %vec.ind.next125, %vector.body111 ]
669  %active.lane.mask130 = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32 %index117, i32 %9)
670  %12 = getelementptr inbounds [1 x i32], ptr @c, i32 0, <4 x i32> %vec.ind124
671  call void @llvm.masked.scatter.v4i32.v4p0(<4 x i32> %broadcast.splat132, <4 x ptr> %12, i32 4, <4 x i1> %active.lane.mask130)
672  %index.next118 = add i32 %index117, 4
673  %vec.ind.next125 = add <4 x i32> %vec.ind124, <i32 12, i32 12, i32 12, i32 12>
674  %13 = icmp eq i32 %index.next118, %n.vec116
675  br i1 %13, label %for.cond9.for.cond15.preheader_crit_edge.us, label %vector.body111
676
677for.cond.cleanup17.us:                            ; preds = %for.cond9.for.cond15.preheader_crit_edge.us
678  %add23.us = add nsw i32 %j.144.us, 7
679  %cmp3.us = icmp slt i32 %j.144.us, -4
680  br i1 %cmp3.us, label %for.body6.us, label %for.cond.cleanup5
681
682for.inc19.us:                                     ; preds = %for.cond9.for.cond15.preheader_crit_edge.us, %for.inc19.us
683  br label %for.inc19.us
684
685for.cond9.for.cond15.preheader_crit_edge.us:      ; preds = %vector.body111, %vector.body
686  %l.1.off045.us.not = xor i1 %l.1.off045.us, true
687  %brmerge = or i1 %tobool20.not97, %l.1.off045.us.not
688  br i1 %brmerge, label %for.cond.cleanup17.us, label %for.inc19.us
689
690for.body6.lr.ph.split:                            ; preds = %for.body6.lr.ph
691  br i1 %tobool20.not97, label %for.body6.preheader, label %for.body6.us60
692
693for.body6.preheader:                              ; preds = %for.body6.lr.ph.split
694  %14 = icmp sgt i32 %j.048, -4
695  %smax = select i1 %14, i32 %j.048, i32 -4
696  %15 = add nsw i32 %smax, 6
697  %16 = sub i32 %15, %j.048
698  %17 = urem i32 %16, 7
699  %18 = sub i32 %16, %17
700  %19 = add nsw i32 %j.048, 7
701  %20 = add i32 %19, %18
702  br label %for.cond.cleanup5
703
704for.body6.us60:                                   ; preds = %for.body6.lr.ph.split, %for.cond.cleanup17.us63.3
705  %l.1.off045.us61 = phi i1 [ false, %for.cond.cleanup17.us63.3 ], [ %l.0.off050, %for.body6.lr.ph.split ]
706  %j.144.us62 = phi i32 [ %add23.us64.3, %for.cond.cleanup17.us63.3 ], [ %j.048, %for.body6.lr.ph.split ]
707  br i1 %l.1.off045.us61, label %for.inc19.us66, label %for.cond.cleanup17.us63
708
709for.cond.cleanup17.us63:                          ; preds = %for.body6.us60
710  %cmp3.us65 = icmp slt i32 %j.144.us62, -4
711  br i1 %cmp3.us65, label %for.cond.cleanup17.us63.1, label %for.cond.cleanup5.loopexit134.split.loop.exit139
712
713for.inc19.us66:                                   ; preds = %for.body6.us60, %for.inc19.us66
714  br label %for.inc19.us66
715
716for.cond.cleanup:                                 ; preds = %for.cond.cleanup5, %entry
717  ret i32 undef
718
719for.cond.cleanup5.loopexit134.split.loop.exit135: ; preds = %for.cond.cleanup17.us63.2
720  %add23.us64.2.le = add nsw i32 %j.144.us62, 21
721  br label %for.cond.cleanup5
722
723for.cond.cleanup5.loopexit134.split.loop.exit137: ; preds = %for.cond.cleanup17.us63.1
724  %add23.us64.1.le = add nsw i32 %j.144.us62, 14
725  br label %for.cond.cleanup5
726
727for.cond.cleanup5.loopexit134.split.loop.exit139: ; preds = %for.cond.cleanup17.us63
728  %add23.us64.le = add nsw i32 %j.144.us62, 7
729  br label %for.cond.cleanup5
730
731for.cond.cleanup5:                                ; preds = %for.cond.cleanup5.loopexit134.split.loop.exit135, %for.cond.cleanup5.loopexit134.split.loop.exit137, %for.cond.cleanup5.loopexit134.split.loop.exit139, %for.cond.cleanup17.us63.3, %for.cond.cleanup17.us, %for.body6.preheader, %for.cond2.preheader
732  %j.1.lcssa = phi i32 [ %j.048, %for.cond2.preheader ], [ %20, %for.body6.preheader ], [ %add23.us, %for.cond.cleanup17.us ], [ %add23.us64.2.le, %for.cond.cleanup5.loopexit134.split.loop.exit135 ], [ %add23.us64.1.le, %for.cond.cleanup5.loopexit134.split.loop.exit137 ], [ %add23.us64.le, %for.cond.cleanup5.loopexit134.split.loop.exit139 ], [ %add23.us64.3, %for.cond.cleanup17.us63.3 ]
733  %l.1.off0.lcssa = phi i1 [ %l.0.off050, %for.cond2.preheader ], [ false, %for.body6.preheader ], [ false, %for.cond.cleanup17.us ], [ false, %for.cond.cleanup17.us63.3 ], [ false, %for.cond.cleanup5.loopexit134.split.loop.exit139 ], [ false, %for.cond.cleanup5.loopexit134.split.loop.exit137 ], [ false, %for.cond.cleanup5.loopexit134.split.loop.exit135 ]
734  %add26 = add nsw i32 %i.049, 2
735  %conv = sext i32 %add26 to i64
736  %cmp = icmp slt i64 %conv, %e
737  br i1 %cmp, label %for.cond2.preheader, label %for.cond.cleanup
738
739for.cond.cleanup17.us63.1:                        ; preds = %for.cond.cleanup17.us63
740  %cmp3.us65.1 = icmp slt i32 %j.144.us62, -11
741  br i1 %cmp3.us65.1, label %for.cond.cleanup17.us63.2, label %for.cond.cleanup5.loopexit134.split.loop.exit137
742
743for.cond.cleanup17.us63.2:                        ; preds = %for.cond.cleanup17.us63.1
744  %cmp3.us65.2 = icmp slt i32 %j.144.us62, -18
745  br i1 %cmp3.us65.2, label %for.cond.cleanup17.us63.3, label %for.cond.cleanup5.loopexit134.split.loop.exit135
746
747for.cond.cleanup17.us63.3:                        ; preds = %for.cond.cleanup17.us63.2
748  %add23.us64.3 = add nsw i32 %j.144.us62, 28
749  %cmp3.us65.3 = icmp slt i32 %j.144.us62, -25
750  br i1 %cmp3.us65.3, label %for.body6.us60, label %for.cond.cleanup5
751}
752
753declare <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32, i32) #1
754declare void @llvm.masked.scatter.v4i32.v4p0(<4 x i32>, <4 x ptr>, i32 immarg, <4 x i1>) #2
755