1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc --arm-memtransfer-tploop=allow -mtriple=thumbv8.1m.main-none-eabi -mattr=+mve --verify-machineinstrs %s -o - | FileCheck %s
4 ; Check that WLSTP loop is not generated for alignment < 4
5 ; void test1(char* dest, char* src, int n){
6 ; memcpy(dest, src, n);
9 declare void @llvm.memcpy.p0.p0.i32(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i32, i1 immarg)
10 declare void @llvm.memcpy.p0.p0.i64(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i64, i1 immarg)
11 declare void @llvm.memset.p0.i32(ptr nocapture writeonly, i8, i32, i1 immarg)
13 define void @test1(ptr noalias nocapture %X, ptr noalias nocapture readonly %Y, i32 %n){
15 ; CHECK: @ %bb.0: @ %entry
16 ; CHECK-NEXT: .save {r7, lr}
17 ; CHECK-NEXT: push {r7, lr}
18 ; CHECK-NEXT: bl __aeabi_memcpy
19 ; CHECK-NEXT: pop {r7, pc}
21 call void @llvm.memcpy.p0.p0.i32(ptr align 1 %X, ptr align 1 %Y, i32 %n, i1 false)
26 ; Check that WLSTP loop is generated for alignment >= 4
27 ; void test2(int* restrict X, int* restrict Y, int n){
31 define void @test2(ptr noalias %X, ptr noalias readonly %Y, i32 %n){
33 ; CHECK: @ %bb.0: @ %entry
34 ; CHECK-NEXT: .save {r7, lr}
35 ; CHECK-NEXT: push {r7, lr}
36 ; CHECK-NEXT: wlstp.8 lr, r2, .LBB1_2
37 ; CHECK-NEXT: .LBB1_1: @ =>This Inner Loop Header: Depth=1
38 ; CHECK-NEXT: vldrb.u8 q0, [r1], #16
39 ; CHECK-NEXT: vstrb.8 q0, [r0], #16
40 ; CHECK-NEXT: letp lr, .LBB1_1
41 ; CHECK-NEXT: .LBB1_2: @ %entry
42 ; CHECK-NEXT: pop {r7, pc}
44 call void @llvm.memcpy.p0.p0.i32(ptr align 4 %X, ptr align 4 %Y, i32 %n, i1 false)
49 ; Checks that transform handles some arithmetic on the input arguments.
50 ; void test3(int* restrict X, int* restrict Y, int n)
52 ; memcpy(X+2, Y+3, (n*2)+10);
55 define void @test3(ptr noalias nocapture %X, ptr noalias nocapture readonly %Y, i32 %n) {
57 ; CHECK: @ %bb.0: @ %entry
58 ; CHECK-NEXT: .save {r7, lr}
59 ; CHECK-NEXT: push {r7, lr}
60 ; CHECK-NEXT: movs r3, #10
61 ; CHECK-NEXT: add.w r2, r3, r2, lsl #1
62 ; CHECK-NEXT: adds r1, #12
63 ; CHECK-NEXT: adds r0, #8
64 ; CHECK-NEXT: wlstp.8 lr, r2, .LBB2_2
65 ; CHECK-NEXT: .LBB2_1: @ =>This Inner Loop Header: Depth=1
66 ; CHECK-NEXT: vldrb.u8 q0, [r1], #16
67 ; CHECK-NEXT: vstrb.8 q0, [r0], #16
68 ; CHECK-NEXT: letp lr, .LBB2_1
69 ; CHECK-NEXT: .LBB2_2: @ %entry
70 ; CHECK-NEXT: pop {r7, pc}
72 %add.ptr = getelementptr inbounds i32, ptr %X, i32 2
73 %add.ptr1 = getelementptr inbounds i32, ptr %Y, i32 3
74 %mul = shl nsw i32 %n, 1
75 %add = add nsw i32 %mul, 10
76 call void @llvm.memcpy.p0.p0.i32(ptr nonnull align 4 %add.ptr, ptr nonnull align 4 %add.ptr1, i32 %add, i1 false)
81 ; Checks that transform handles for loops that are implicitly converted to mempcy
82 ; void test4(int* restrict X, int* restrict Y, int n){
83 ; for(int i = 0; i < n; ++i){
88 define void @test4(ptr noalias %X, ptr noalias readonly %Y, i32 %n) {
90 ; CHECK: @ %bb.0: @ %entry
91 ; CHECK-NEXT: cmp r2, #1
94 ; CHECK-NEXT: .LBB3_1: @ %for.body.preheader
95 ; CHECK-NEXT: .save {r7, lr}
96 ; CHECK-NEXT: push {r7, lr}
97 ; CHECK-NEXT: wlstp.8 lr, r2, .LBB3_3
98 ; CHECK-NEXT: .LBB3_2: @ =>This Inner Loop Header: Depth=1
99 ; CHECK-NEXT: vldrb.u8 q0, [r1], #16
100 ; CHECK-NEXT: vstrb.8 q0, [r0], #16
101 ; CHECK-NEXT: letp lr, .LBB3_2
102 ; CHECK-NEXT: .LBB3_3: @ %for.body.preheader
103 ; CHECK-NEXT: pop.w {r7, lr}
106 %cmp6 = icmp sgt i32 %n, 0
107 br i1 %cmp6, label %for.body.preheader, label %for.cond.cleanup
109 for.body.preheader: ; preds = %entry
110 call void @llvm.memcpy.p0.p0.i32(ptr align 4 %X, ptr align 4 %Y, i32 %n, i1 false)
111 br label %for.cond.cleanup
113 for.cond.cleanup: ; preds = %for.body.preheader, %entry
117 ; Checks that transform can handle > i32 size inputs
118 define void @test5(ptr noalias %X, ptr noalias %Y, i64 %n){
119 ; CHECK-LABEL: test5:
121 ; CHECK-NEXT: .save {r7, lr}
122 ; CHECK-NEXT: push {r7, lr}
123 ; CHECK-NEXT: wlstp.8 lr, r2, .LBB4_2
124 ; CHECK-NEXT: .LBB4_1: @ =>This Inner Loop Header: Depth=1
125 ; CHECK-NEXT: vldrb.u8 q0, [r1], #16
126 ; CHECK-NEXT: vstrb.8 q0, [r0], #16
127 ; CHECK-NEXT: letp lr, .LBB4_1
128 ; CHECK-NEXT: .LBB4_2:
129 ; CHECK-NEXT: pop {r7, pc}
130 call void @llvm.memcpy.p0.p0.i64(ptr align 4 %X, ptr align 4 %Y, i64 %n, i1 false)
134 ; Checks the transform is applied for constant size inputs below a certain threshold (128 in this case)
135 define void @test6(ptr noalias nocapture %X, ptr noalias nocapture readonly %Y, i32 %n) {
136 ; CHECK-LABEL: test6:
137 ; CHECK: @ %bb.0: @ %entry
138 ; CHECK-NEXT: .save {r7, lr}
139 ; CHECK-NEXT: push {r7, lr}
140 ; CHECK-NEXT: movs r2, #127
141 ; CHECK-NEXT: wlstp.8 lr, r2, .LBB5_2
142 ; CHECK-NEXT: .LBB5_1: @ =>This Inner Loop Header: Depth=1
143 ; CHECK-NEXT: vldrb.u8 q0, [r1], #16
144 ; CHECK-NEXT: vstrb.8 q0, [r0], #16
145 ; CHECK-NEXT: letp lr, .LBB5_1
146 ; CHECK-NEXT: .LBB5_2: @ %entry
147 ; CHECK-NEXT: pop {r7, pc}
149 call void @llvm.memcpy.p0.p0.i32(ptr noundef nonnull align 4 dereferenceable(127) %X, ptr noundef nonnull align 4 dereferenceable(127) %Y, i32 127, i1 false)
153 ; Checks the transform is NOT applied for constant size inputs above a certain threshold (128 in this case)
154 define void @test7(ptr noalias nocapture %X, ptr noalias nocapture readonly %Y, i32 %n) {
155 ; CHECK-LABEL: test7:
156 ; CHECK: @ %bb.0: @ %entry
157 ; CHECK-NEXT: .save {r7, lr}
158 ; CHECK-NEXT: push {r7, lr}
159 ; CHECK-NEXT: movs r2, #128
160 ; CHECK-NEXT: bl __aeabi_memcpy4
161 ; CHECK-NEXT: pop {r7, pc}
163 call void @llvm.memcpy.p0.p0.i32(ptr align 4 %X, ptr align 4 %Y, i32 128, i1 false)
167 ; Checks the transform is NOT applied for constant size inputs below a certain threshold (64 in this case)
168 define void @test8(ptr noalias nocapture %X, ptr noalias nocapture readonly %Y, i32 %n) {
169 ; CHECK-LABEL: test8:
170 ; CHECK: @ %bb.0: @ %entry
171 ; CHECK-NEXT: .save {r4, lr}
172 ; CHECK-NEXT: push {r4, lr}
173 ; CHECK-NEXT: ldm.w r1!, {r2, r3, r4, r12, lr}
174 ; CHECK-NEXT: stm.w r0!, {r2, r3, r4, r12, lr}
175 ; CHECK-NEXT: ldm.w r1!, {r2, r3, r4, r12, lr}
176 ; CHECK-NEXT: stm.w r0!, {r2, r3, r4, r12, lr}
177 ; CHECK-NEXT: ldm.w r1, {r2, r3, r4, r12, lr}
178 ; CHECK-NEXT: stm.w r0, {r2, r3, r4, r12, lr}
179 ; CHECK-NEXT: pop {r4, pc}
181 call void @llvm.memcpy.p0.p0.i32(ptr align 4 %X, ptr align 4 %Y, i32 60, i1 false)
185 ; Checks the transform is NOT applied (regardless of alignment) when optimizations are disabled
186 define void @test9(ptr noalias nocapture %X, ptr noalias nocapture readonly %Y, i32 %n) #0 {
187 ; CHECK-LABEL: test9:
188 ; CHECK: @ %bb.0: @ %entry
189 ; CHECK-NEXT: .save {r7, lr}
190 ; CHECK-NEXT: push {r7, lr}
191 ; CHECK-NEXT: bl __aeabi_memcpy4
192 ; CHECK-NEXT: pop {r7, pc}
194 call void @llvm.memcpy.p0.p0.i32(ptr align 4 %X, ptr align 4 %Y, i32 %n, i1 false)
198 ; Checks the transform is NOT applied (regardless of alignment) when optimization for size is on (-Os or -Oz)
199 define void @test10(ptr noalias nocapture %X, ptr noalias nocapture readonly %Y, i32 %n) #1 {
200 ; CHECK-LABEL: test10:
201 ; CHECK: @ %bb.0: @ %entry
202 ; CHECK-NEXT: .save {r7, lr}
203 ; CHECK-NEXT: push {r7, lr}
204 ; CHECK-NEXT: bl __aeabi_memcpy4
205 ; CHECK-NEXT: pop {r7, pc}
207 call void @llvm.memcpy.p0.p0.i32(ptr align 4 %X, ptr align 4 %Y, i32 %n, i1 false)
211 define void @test11(ptr nocapture %x, ptr nocapture %y, i32 %n) {
212 ; CHECK-LABEL: test11:
213 ; CHECK: @ %bb.0: @ %entry
214 ; CHECK-NEXT: cmp.w r2, #-1
216 ; CHECK-NEXT: bxgt lr
217 ; CHECK-NEXT: .LBB10_1: @ %prehead
218 ; CHECK-NEXT: .save {r4, lr}
219 ; CHECK-NEXT: push {r4, lr}
220 ; CHECK-NEXT: mov r12, r1
221 ; CHECK-NEXT: mov r4, r0
222 ; CHECK-NEXT: wlstp.8 lr, r2, .LBB10_3
223 ; CHECK-NEXT: .LBB10_2: @ =>This Inner Loop Header: Depth=1
224 ; CHECK-NEXT: vldrb.u8 q0, [r12], #16
225 ; CHECK-NEXT: vstrb.8 q0, [r4], #16
226 ; CHECK-NEXT: letp lr, .LBB10_2
227 ; CHECK-NEXT: .LBB10_3: @ %for.body
228 ; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
229 ; CHECK-NEXT: ldrb r3, [r0], #1
230 ; CHECK-NEXT: subs r2, #2
231 ; CHECK-NEXT: strb r3, [r1], #1
232 ; CHECK-NEXT: bne .LBB10_3
233 ; CHECK-NEXT: @ %bb.4:
234 ; CHECK-NEXT: pop.w {r4, lr}
237 %cmp6 = icmp slt i32 %n, 0
238 br i1 %cmp6, label %prehead, label %for.cond.cleanup
240 prehead: ; preds = %entry
241 call void @llvm.memcpy.p0.p0.i32(ptr align 4 %x, ptr align 4 %y, i32 %n, i1 false)
244 for.body: ; preds = %for.body, %prehead
245 %i.09 = phi i32 [ %inc, %for.body ], [ 0, %prehead ]
246 %x.addr.08 = phi ptr [ %add.ptr, %for.body ], [ %x, %prehead ]
247 %y.addr.07 = phi ptr [ %add.ptr1, %for.body ], [ %y, %prehead ]
248 %add.ptr = getelementptr inbounds i8, ptr %x.addr.08, i32 1
249 %add.ptr1 = getelementptr inbounds i8, ptr %y.addr.07, i32 1
250 %l = load i8, ptr %x.addr.08, align 1
251 store i8 %l, ptr %y.addr.07, align 1
252 %inc = add nuw nsw i32 %i.09, 2
253 %exitcond.not = icmp eq i32 %inc, %n
254 br i1 %exitcond.not, label %for.cond.cleanup, label %for.body
256 for.cond.cleanup: ; preds = %entry
260 ; Check that WLSTP loop is generated for simplest case of align = 1
261 define void @test12(ptr %X, i8 zeroext %c, i32 %n) {
262 ; CHECK-LABEL: test12:
263 ; CHECK: @ %bb.0: @ %entry
264 ; CHECK-NEXT: .save {r7, lr}
265 ; CHECK-NEXT: push {r7, lr}
266 ; CHECK-NEXT: vdup.8 q0, r1
267 ; CHECK-NEXT: wlstp.8 lr, r2, .LBB11_2
268 ; CHECK-NEXT: .LBB11_1: @ =>This Inner Loop Header: Depth=1
269 ; CHECK-NEXT: vstrb.8 q0, [r0], #16
270 ; CHECK-NEXT: letp lr, .LBB11_1
271 ; CHECK-NEXT: .LBB11_2: @ %entry
272 ; CHECK-NEXT: pop {r7, pc}
274 call void @llvm.memset.p0.i32(ptr align 1 %X, i8 %c, i32 %n, i1 false)
279 ; Check that WLSTP loop is generated for alignment >= 4
280 define void @test13(ptr %X, i8 zeroext %c, i32 %n) {
281 ; CHECK-LABEL: test13:
282 ; CHECK: @ %bb.0: @ %entry
283 ; CHECK-NEXT: .save {r7, lr}
284 ; CHECK-NEXT: push {r7, lr}
285 ; CHECK-NEXT: vdup.8 q0, r1
286 ; CHECK-NEXT: wlstp.8 lr, r2, .LBB12_2
287 ; CHECK-NEXT: .LBB12_1: @ =>This Inner Loop Header: Depth=1
288 ; CHECK-NEXT: vstrb.8 q0, [r0], #16
289 ; CHECK-NEXT: letp lr, .LBB12_1
290 ; CHECK-NEXT: .LBB12_2: @ %entry
291 ; CHECK-NEXT: pop {r7, pc}
293 call void @llvm.memset.p0.i32(ptr align 4 %X, i8 %c, i32 %n, i1 false)
297 define void @twoloops(ptr %X, i32 %n, i32 %m) {
298 ; CHECK-LABEL: twoloops:
299 ; CHECK: @ %bb.0: @ %entry
300 ; CHECK-NEXT: .save {r7, lr}
301 ; CHECK-NEXT: push {r7, lr}
302 ; CHECK-NEXT: vmov.i32 q0, #0x0
303 ; CHECK-NEXT: mov r3, r0
304 ; CHECK-NEXT: wlstp.8 lr, r2, .LBB13_2
305 ; CHECK-NEXT: .LBB13_1: @ =>This Inner Loop Header: Depth=1
306 ; CHECK-NEXT: vstrb.8 q0, [r3], #16
307 ; CHECK-NEXT: letp lr, .LBB13_1
308 ; CHECK-NEXT: .LBB13_2: @ %entry
309 ; CHECK-NEXT: wlstp.8 lr, r2, .LBB13_4
310 ; CHECK-NEXT: .LBB13_3: @ =>This Inner Loop Header: Depth=1
311 ; CHECK-NEXT: vstrb.8 q0, [r0], #16
312 ; CHECK-NEXT: letp lr, .LBB13_3
313 ; CHECK-NEXT: .LBB13_4: @ %entry
314 ; CHECK-NEXT: pop {r7, pc}
316 call void @llvm.memset.p0.i32(ptr align 4 %X, i8 0, i32 %m, i1 false)
317 call void @llvm.memset.p0.i32(ptr align 4 %X, i8 0, i32 %m, i1 false)
322 ; Checks that transform correctly handles input with some arithmetic on input arguments.
323 ; void test14(int* X, char c, int n)
325 ; memset(X+2, c, (n*2)+10);
328 define void @test14(ptr %X, i8 zeroext %c, i32 %n) {
329 ; CHECK-LABEL: test14:
330 ; CHECK: @ %bb.0: @ %entry
331 ; CHECK-NEXT: .save {r7, lr}
332 ; CHECK-NEXT: push {r7, lr}
333 ; CHECK-NEXT: movs r3, #10
334 ; CHECK-NEXT: add.w r2, r3, r2, lsl #1
335 ; CHECK-NEXT: vdup.8 q0, r1
336 ; CHECK-NEXT: adds r0, #8
337 ; CHECK-NEXT: wlstp.8 lr, r2, .LBB14_2
338 ; CHECK-NEXT: .LBB14_1: @ =>This Inner Loop Header: Depth=1
339 ; CHECK-NEXT: vstrb.8 q0, [r0], #16
340 ; CHECK-NEXT: letp lr, .LBB14_1
341 ; CHECK-NEXT: .LBB14_2: @ %entry
342 ; CHECK-NEXT: pop {r7, pc}
344 %add.ptr = getelementptr inbounds i32, ptr %X, i32 2
345 %mul = shl nsw i32 %n, 1
346 %add = add nsw i32 %mul, 10
347 call void @llvm.memset.p0.i32(ptr nonnull align 4 %add.ptr, i8 %c, i32 %add, i1 false)
354 ; Checks that transform handles for-loops (that get implicitly converted to memset)
355 ; void test15(int* X, char Y, int n){
356 ; for(int i = 0; i < n; ++i){
361 define void @test15(ptr nocapture %X, i8 zeroext %c, i32 %n) {
362 ; CHECK-LABEL: test15:
363 ; CHECK: @ %bb.0: @ %entry
364 ; CHECK-NEXT: cmp r2, #1
366 ; CHECK-NEXT: bxlt lr
367 ; CHECK-NEXT: .LBB15_1: @ %for.body.preheader
368 ; CHECK-NEXT: .save {r7, lr}
369 ; CHECK-NEXT: push {r7, lr}
370 ; CHECK-NEXT: vdup.8 q0, r1
371 ; CHECK-NEXT: wlstp.8 lr, r2, .LBB15_3
372 ; CHECK-NEXT: .LBB15_2: @ =>This Inner Loop Header: Depth=1
373 ; CHECK-NEXT: vstrb.8 q0, [r0], #16
374 ; CHECK-NEXT: letp lr, .LBB15_2
375 ; CHECK-NEXT: .LBB15_3: @ %for.body.preheader
376 ; CHECK-NEXT: pop.w {r7, lr}
379 %cmp4 = icmp sgt i32 %n, 0
380 br i1 %cmp4, label %for.body.preheader, label %for.cond.cleanup
382 for.body.preheader: ; preds = %entry
383 call void @llvm.memset.p0.i32(ptr align 4 %X, i8 %c, i32 %n, i1 false)
384 br label %for.cond.cleanup
386 for.cond.cleanup: ; preds = %for.body.preheader, %entry
390 ; Checks that transform handles case with 0 as src value. No difference is expected.
391 define void @test16(ptr %X, i8 zeroext %c, i32 %n) {
392 ; CHECK-LABEL: test16:
393 ; CHECK: @ %bb.0: @ %entry
394 ; CHECK-NEXT: .save {r7, lr}
395 ; CHECK-NEXT: push {r7, lr}
396 ; CHECK-NEXT: vmov.i32 q0, #0x0
397 ; CHECK-NEXT: wlstp.8 lr, r2, .LBB16_2
398 ; CHECK-NEXT: .LBB16_1: @ =>This Inner Loop Header: Depth=1
399 ; CHECK-NEXT: vstrb.8 q0, [r0], #16
400 ; CHECK-NEXT: letp lr, .LBB16_1
401 ; CHECK-NEXT: .LBB16_2: @ %entry
402 ; CHECK-NEXT: pop {r7, pc}
404 call void @llvm.memset.p0.i32(ptr align 4 %X, i8 0, i32 %n, i1 false)
408 define void @csprlive(ptr noalias %X, ptr noalias readonly %Y, i32 %n) {
409 ; CHECK-LABEL: csprlive:
410 ; CHECK: @ %bb.0: @ %entry
411 ; CHECK-NEXT: .save {r7, lr}
412 ; CHECK-NEXT: push {r7, lr}
413 ; CHECK-NEXT: wlstp.8 lr, r2, .LBB17_2
414 ; CHECK-NEXT: .LBB17_1: @ =>This Inner Loop Header: Depth=1
415 ; CHECK-NEXT: vldrb.u8 q0, [r1], #16
416 ; CHECK-NEXT: vstrb.8 q0, [r0], #16
417 ; CHECK-NEXT: letp lr, .LBB17_1
418 ; CHECK-NEXT: .LBB17_2: @ %entry
419 ; CHECK-NEXT: bl other
420 ; CHECK-NEXT: pop {r7, pc}
422 %cmp6 = icmp sgt i32 %n, 0
423 call void @llvm.memcpy.p0.p0.i32(ptr align 4 %X, ptr align 4 %Y, i32 %n, i1 false)
424 br i1 %cmp6, label %if, label %else
438 declare void @other()
440 @arr_56 = external dso_local local_unnamed_addr global [21 x [16 x [11 x i8]]], align 1
441 define void @multilooped_exit(i32 %b) {
442 ; CHECK-LABEL: multilooped_exit:
443 ; CHECK: @ %bb.0: @ %entry
444 ; CHECK-NEXT: cmp r0, #1
446 ; CHECK-NEXT: bxlt lr
447 ; CHECK-NEXT: .LBB18_1: @ %loop.preheader
448 ; CHECK-NEXT: .save {r4, lr}
449 ; CHECK-NEXT: push {r4, lr}
450 ; CHECK-NEXT: mov.w r4, #-1
451 ; CHECK-NEXT: vmov.i32 q0, #0x0
452 ; CHECK-NEXT: b .LBB18_3
453 ; CHECK-NEXT: .LBB18_2: @ %loop
454 ; CHECK-NEXT: @ in Loop: Header=BB18_3 Depth=1
455 ; CHECK-NEXT: adds r4, #1
456 ; CHECK-NEXT: cmp.w r4, #1024
457 ; CHECK-NEXT: bge .LBB18_12
458 ; CHECK-NEXT: .LBB18_3: @ %loop
459 ; CHECK-NEXT: @ =>This Loop Header: Depth=1
460 ; CHECK-NEXT: @ Child Loop BB18_4 Depth 2
461 ; CHECK-NEXT: @ Child Loop BB18_6 Depth 2
462 ; CHECK-NEXT: @ Child Loop BB18_8 Depth 2
463 ; CHECK-NEXT: @ Child Loop BB18_11 Depth 2
464 ; CHECK-NEXT: movw r3, :lower16:arr_56
465 ; CHECK-NEXT: add.w r1, r0, #15
466 ; CHECK-NEXT: movt r3, :upper16:arr_56
467 ; CHECK-NEXT: lsr.w r12, r1, #4
468 ; CHECK-NEXT: mov r2, r3
469 ; CHECK-NEXT: wlstp.8 lr, r0, .LBB18_5
470 ; CHECK-NEXT: .LBB18_4: @ Parent Loop BB18_3 Depth=1
471 ; CHECK-NEXT: @ => This Inner Loop Header: Depth=2
472 ; CHECK-NEXT: vstrb.8 q0, [r2], #16
473 ; CHECK-NEXT: letp lr, .LBB18_4
474 ; CHECK-NEXT: .LBB18_5: @ %loop
475 ; CHECK-NEXT: @ in Loop: Header=BB18_3 Depth=1
476 ; CHECK-NEXT: mov r2, r3
477 ; CHECK-NEXT: wlstp.8 lr, r0, .LBB18_7
478 ; CHECK-NEXT: .LBB18_6: @ Parent Loop BB18_3 Depth=1
479 ; CHECK-NEXT: @ => This Inner Loop Header: Depth=2
480 ; CHECK-NEXT: vstrb.8 q0, [r2], #16
481 ; CHECK-NEXT: letp lr, .LBB18_6
482 ; CHECK-NEXT: .LBB18_7: @ %loop
483 ; CHECK-NEXT: @ in Loop: Header=BB18_3 Depth=1
484 ; CHECK-NEXT: mov r2, r3
485 ; CHECK-NEXT: wlstp.8 lr, r0, .LBB18_9
486 ; CHECK-NEXT: .LBB18_8: @ Parent Loop BB18_3 Depth=1
487 ; CHECK-NEXT: @ => This Inner Loop Header: Depth=2
488 ; CHECK-NEXT: vstrb.8 q0, [r2], #16
489 ; CHECK-NEXT: letp lr, .LBB18_8
490 ; CHECK-NEXT: .LBB18_9: @ %loop
491 ; CHECK-NEXT: @ in Loop: Header=BB18_3 Depth=1
492 ; CHECK-NEXT: cmp.w r12, #0
493 ; CHECK-NEXT: beq .LBB18_2
494 ; CHECK-NEXT: @ %bb.10: @ %loop
495 ; CHECK-NEXT: @ in Loop: Header=BB18_3 Depth=1
496 ; CHECK-NEXT: dlstp.8 lr, r0
497 ; CHECK-NEXT: .LBB18_11: @ Parent Loop BB18_3 Depth=1
498 ; CHECK-NEXT: @ => This Inner Loop Header: Depth=2
499 ; CHECK-NEXT: vstrb.8 q0, [r3], #16
500 ; CHECK-NEXT: letp lr, .LBB18_11
501 ; CHECK-NEXT: b .LBB18_2
502 ; CHECK-NEXT: .LBB18_12:
503 ; CHECK-NEXT: pop.w {r4, lr}
506 %cmp8 = icmp sgt i32 %b, 0
507 br i1 %cmp8, label %loop, label %exit
510 %p = phi i32 [ 0, %entry ], [ %inc, %loop ]
511 call void @llvm.memset.p0.i32(ptr align 1 getelementptr ([21 x [16 x [11 x i8]]], ptr @arr_56, i32 0, i32 0, i32 undef, i32 0), i8 0, i32 %b, i1 false)
512 call void @llvm.memset.p0.i32(ptr align 1 getelementptr ([21 x [16 x [11 x i8]]], ptr @arr_56, i32 0, i32 0, i32 undef, i32 0), i8 0, i32 %b, i1 false)
513 call void @llvm.memset.p0.i32(ptr align 1 getelementptr ([21 x [16 x [11 x i8]]], ptr @arr_56, i32 0, i32 0, i32 undef, i32 0), i8 0, i32 %b, i1 false)
514 call void @llvm.memset.p0.i32(ptr align 1 getelementptr ([21 x [16 x [11 x i8]]], ptr @arr_56, i32 0, i32 0, i32 undef, i32 0), i8 0, i32 %b, i1 false)
516 %c = icmp slt i32 %p, 1024
517 br i1 %c, label %loop, label %exit
523 @arr_21 = external dso_local local_unnamed_addr global [17 x [12 x [19 x i16]]], align 2
524 @arr_20 = external dso_local local_unnamed_addr global [17 x [12 x [19 x i64]]], align 8
525 @arr_22 = external dso_local local_unnamed_addr global [17 x [12 x [19 x i16]]], align 2
526 define i32 @reverted(i1 zeroext %b) {
527 ; CHECK-LABEL: reverted:
528 ; CHECK: @ %bb.0: @ %entry
529 ; CHECK-NEXT: .save {r4, r5, r6, r7, r8, r9, r10, r11, lr}
530 ; CHECK-NEXT: push.w {r4, r5, r6, r7, r8, r9, r10, r11, lr}
531 ; CHECK-NEXT: .pad #12
532 ; CHECK-NEXT: sub sp, #12
533 ; CHECK-NEXT: cmp r0, #0
534 ; CHECK-NEXT: mov.w r1, #11
535 ; CHECK-NEXT: cinc r1, r1, ne
536 ; CHECK-NEXT: movs r0, #38
537 ; CHECK-NEXT: mul r2, r1, r0
538 ; CHECK-NEXT: str r1, [sp, #8] @ 4-byte Spill
539 ; CHECK-NEXT: movw r0, :lower16:arr_22
540 ; CHECK-NEXT: vmov.i32 q0, #0x0
541 ; CHECK-NEXT: movt r0, :upper16:arr_22
542 ; CHECK-NEXT: add.w r1, r2, #15
543 ; CHECK-NEXT: lsrs r3, r1, #4
544 ; CHECK-NEXT: strd r3, r2, [sp] @ 8-byte Folded Spill
545 ; CHECK-NEXT: wlstp.8 lr, r2, .LBB19_2
546 ; CHECK-NEXT: .LBB19_1: @ =>This Inner Loop Header: Depth=1
547 ; CHECK-NEXT: vstrb.8 q0, [r0], #16
548 ; CHECK-NEXT: letp lr, .LBB19_1
549 ; CHECK-NEXT: .LBB19_2: @ %entry
550 ; CHECK-NEXT: ldr r0, [sp, #8] @ 4-byte Reload
551 ; CHECK-NEXT: movw r6, :lower16:arr_20
552 ; CHECK-NEXT: movt r6, :upper16:arr_20
553 ; CHECK-NEXT: add.w r3, r6, #80
554 ; CHECK-NEXT: dls lr, r0
555 ; CHECK-NEXT: movw r0, :lower16:arr_21
556 ; CHECK-NEXT: movt r0, :upper16:arr_21
557 ; CHECK-NEXT: add.w r5, r0, #36
558 ; CHECK-NEXT: add.w r11, r6, #128
559 ; CHECK-NEXT: add.w r7, r6, #112
560 ; CHECK-NEXT: add.w r2, r6, #96
561 ; CHECK-NEXT: add.w r4, r6, #64
562 ; CHECK-NEXT: add.w r0, r6, #48
563 ; CHECK-NEXT: add.w r1, r6, #32
564 ; CHECK-NEXT: add.w r12, r6, #16
565 ; CHECK-NEXT: adr r6, .LCPI19_0
566 ; CHECK-NEXT: vldrw.u32 q0, [r6]
567 ; CHECK-NEXT: movw r6, :lower16:arr_20
568 ; CHECK-NEXT: mov.w r8, #327685
569 ; CHECK-NEXT: mov.w r9, #5
570 ; CHECK-NEXT: vmov.i16 q1, #0x5
571 ; CHECK-NEXT: mov.w r10, #0
572 ; CHECK-NEXT: movt r6, :upper16:arr_20
573 ; CHECK-NEXT: .LBB19_3: @ %for.cond8.preheader
574 ; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
575 ; CHECK-NEXT: str r8, [r5, #-4]
576 ; CHECK-NEXT: vstrh.16 q1, [r5, #-36]
577 ; CHECK-NEXT: strh.w r9, [r5]
578 ; CHECK-NEXT: vstrh.16 q1, [r5, #-20]
579 ; CHECK-NEXT: vstrw.32 q0, [r3]
580 ; CHECK-NEXT: vstrh.16 q0, [r12], #152
581 ; CHECK-NEXT: vstrh.16 q0, [r6], #152
582 ; CHECK-NEXT: vstrh.16 q0, [r1], #152
583 ; CHECK-NEXT: vstrh.16 q0, [r0], #152
584 ; CHECK-NEXT: vstrh.16 q0, [r4], #152
585 ; CHECK-NEXT: vstrh.16 q0, [r2], #152
586 ; CHECK-NEXT: vstrh.16 q0, [r7], #152
587 ; CHECK-NEXT: vstrh.16 q0, [r11], #152
588 ; CHECK-NEXT: strd r9, r10, [r3, #64]
589 ; CHECK-NEXT: adds r5, #38
590 ; CHECK-NEXT: adds r3, #152
591 ; CHECK-NEXT: le lr, .LBB19_3
592 ; CHECK-NEXT: @ %bb.4: @ %for.cond.cleanup6
593 ; CHECK-NEXT: movw r0, :lower16:arr_22
594 ; CHECK-NEXT: ldr r2, [sp, #4] @ 4-byte Reload
595 ; CHECK-NEXT: movt r0, :upper16:arr_22
596 ; CHECK-NEXT: ldr r3, [sp] @ 4-byte Reload
597 ; CHECK-NEXT: add.w r0, r0, #1824
598 ; CHECK-NEXT: vmov.i32 q1, #0x0
599 ; CHECK-NEXT: wlstp.8 lr, r2, .LBB19_6
600 ; CHECK-NEXT: .LBB19_5: @ =>This Inner Loop Header: Depth=1
601 ; CHECK-NEXT: vstrb.8 q1, [r0], #16
602 ; CHECK-NEXT: letp lr, .LBB19_5
603 ; CHECK-NEXT: .LBB19_6: @ %for.cond.cleanup6
604 ; CHECK-NEXT: movw r6, :lower16:arr_20
605 ; CHECK-NEXT: movw r0, #7376
606 ; CHECK-NEXT: movt r6, :upper16:arr_20
607 ; CHECK-NEXT: adds r3, r6, r0
608 ; CHECK-NEXT: movw r0, #7408
609 ; CHECK-NEXT: add.w r12, r6, r0
610 ; CHECK-NEXT: movw r0, #7344
611 ; CHECK-NEXT: add.w r9, r6, r0
612 ; CHECK-NEXT: movw r0, #7312
613 ; CHECK-NEXT: adds r2, r6, r0
614 ; CHECK-NEXT: movw r0, :lower16:arr_21
615 ; CHECK-NEXT: add.w r1, r6, #7424
616 ; CHECK-NEXT: add.w r7, r6, #7392
617 ; CHECK-NEXT: add.w r4, r6, #7360
618 ; CHECK-NEXT: add.w r5, r6, #7328
619 ; CHECK-NEXT: add.w r8, r6, #7296
620 ; CHECK-NEXT: ldr r6, [sp, #8] @ 4-byte Reload
621 ; CHECK-NEXT: movt r0, :upper16:arr_21
622 ; CHECK-NEXT: addw r0, r0, #1860
623 ; CHECK-NEXT: mov.w r10, #5
624 ; CHECK-NEXT: dls lr, r6
625 ; CHECK-NEXT: mov.w r6, #327685
626 ; CHECK-NEXT: vmov.i16 q1, #0x5
627 ; CHECK-NEXT: mov.w r11, #0
628 ; CHECK-NEXT: .LBB19_7: @ %for.cond8.preheader.1
629 ; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
630 ; CHECK-NEXT: str r6, [r0, #-4]
631 ; CHECK-NEXT: vstrh.16 q1, [r0, #-36]
632 ; CHECK-NEXT: strh.w r10, [r0]
633 ; CHECK-NEXT: vstrh.16 q1, [r0, #-20]
634 ; CHECK-NEXT: vstrw.32 q0, [r3]
635 ; CHECK-NEXT: vstrh.16 q0, [r2], #152
636 ; CHECK-NEXT: vstrh.16 q0, [r8], #152
637 ; CHECK-NEXT: vstrh.16 q0, [r5], #152
638 ; CHECK-NEXT: vstrh.16 q0, [r9], #152
639 ; CHECK-NEXT: vstrh.16 q0, [r4], #152
640 ; CHECK-NEXT: vstrh.16 q0, [r7], #152
641 ; CHECK-NEXT: vstrh.16 q0, [r12], #152
642 ; CHECK-NEXT: vstrh.16 q0, [r1], #152
643 ; CHECK-NEXT: strd r10, r11, [r3, #64]
644 ; CHECK-NEXT: adds r0, #38
645 ; CHECK-NEXT: adds r3, #152
646 ; CHECK-NEXT: le lr, .LBB19_7
647 ; CHECK-NEXT: @ %bb.8: @ %for.cond.cleanup6.1
648 ; CHECK-NEXT: movw r0, :lower16:arr_22
649 ; CHECK-NEXT: ldr r2, [sp, #4] @ 4-byte Reload
650 ; CHECK-NEXT: movt r0, :upper16:arr_22
651 ; CHECK-NEXT: ldr r3, [sp] @ 4-byte Reload
652 ; CHECK-NEXT: add.w r0, r0, #3648
653 ; CHECK-NEXT: vmov.i32 q1, #0x0
654 ; CHECK-NEXT: wlstp.8 lr, r2, .LBB19_10
655 ; CHECK-NEXT: .LBB19_9: @ =>This Inner Loop Header: Depth=1
656 ; CHECK-NEXT: vstrb.8 q1, [r0], #16
657 ; CHECK-NEXT: letp lr, .LBB19_9
658 ; CHECK-NEXT: .LBB19_10: @ %for.cond.cleanup6.1
659 ; CHECK-NEXT: movw r7, :lower16:arr_20
660 ; CHECK-NEXT: movw r0, #14672
661 ; CHECK-NEXT: movt r7, :upper16:arr_20
662 ; CHECK-NEXT: adds r3, r7, r0
663 ; CHECK-NEXT: movw r0, #14704
664 ; CHECK-NEXT: add.w r12, r7, r0
665 ; CHECK-NEXT: movw r0, #14688
666 ; CHECK-NEXT: add.w r8, r7, r0
667 ; CHECK-NEXT: movw r0, #14640
668 ; CHECK-NEXT: add.w r9, r7, r0
669 ; CHECK-NEXT: movw r0, #14624
670 ; CHECK-NEXT: adds r2, r7, r0
671 ; CHECK-NEXT: movw r0, #14608
672 ; CHECK-NEXT: movw r1, :lower16:arr_21
673 ; CHECK-NEXT: add r0, r7
674 ; CHECK-NEXT: add.w r4, r7, #14720
675 ; CHECK-NEXT: add.w r5, r7, #14656
676 ; CHECK-NEXT: add.w r6, r7, #14592
677 ; CHECK-NEXT: ldr r7, [sp, #8] @ 4-byte Reload
678 ; CHECK-NEXT: movt r1, :upper16:arr_21
679 ; CHECK-NEXT: addw r1, r1, #3684
680 ; CHECK-NEXT: mov.w r10, #5
681 ; CHECK-NEXT: dls lr, r7
682 ; CHECK-NEXT: mov.w r7, #327685
683 ; CHECK-NEXT: vmov.i16 q1, #0x5
684 ; CHECK-NEXT: mov.w r11, #0
685 ; CHECK-NEXT: .LBB19_11: @ %for.cond8.preheader.2
686 ; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
687 ; CHECK-NEXT: str r7, [r1, #-4]
688 ; CHECK-NEXT: vstrh.16 q1, [r1, #-36]
689 ; CHECK-NEXT: strh.w r10, [r1]
690 ; CHECK-NEXT: vstrh.16 q1, [r1, #-20]
691 ; CHECK-NEXT: vstrw.32 q0, [r3]
692 ; CHECK-NEXT: vstrh.16 q0, [r0], #152
693 ; CHECK-NEXT: vstrh.16 q0, [r6], #152
694 ; CHECK-NEXT: vstrh.16 q0, [r2], #152
695 ; CHECK-NEXT: vstrh.16 q0, [r9], #152
696 ; CHECK-NEXT: vstrh.16 q0, [r5], #152
697 ; CHECK-NEXT: vstrh.16 q0, [r8], #152
698 ; CHECK-NEXT: vstrh.16 q0, [r12], #152
699 ; CHECK-NEXT: vstrh.16 q0, [r4], #152
700 ; CHECK-NEXT: strd r10, r11, [r3, #64]
701 ; CHECK-NEXT: adds r1, #38
702 ; CHECK-NEXT: adds r3, #152
703 ; CHECK-NEXT: le lr, .LBB19_11
704 ; CHECK-NEXT: @ %bb.12: @ %for.cond.cleanup6.2
705 ; CHECK-NEXT: movw r0, :lower16:arr_22
706 ; CHECK-NEXT: ldrd r2, r1, [sp] @ 8-byte Folded Reload
707 ; CHECK-NEXT: movt r0, :upper16:arr_22
708 ; CHECK-NEXT: vmov.i32 q1, #0x0
709 ; CHECK-NEXT: add.w r0, r0, #5472
710 ; CHECK-NEXT: wlstp.8 lr, r1, .LBB19_14
711 ; CHECK-NEXT: .LBB19_13: @ =>This Inner Loop Header: Depth=1
712 ; CHECK-NEXT: vstrb.8 q1, [r0], #16
713 ; CHECK-NEXT: letp lr, .LBB19_13
714 ; CHECK-NEXT: .LBB19_14: @ %for.cond.cleanup6.2
715 ; CHECK-NEXT: movw r2, :lower16:arr_21
716 ; CHECK-NEXT: movw r1, #5508
717 ; CHECK-NEXT: movt r2, :upper16:arr_21
718 ; CHECK-NEXT: movw r7, :lower16:arr_20
719 ; CHECK-NEXT: add r2, r1
720 ; CHECK-NEXT: movw r1, #22000
721 ; CHECK-NEXT: movt r7, :upper16:arr_20
722 ; CHECK-NEXT: add.w r12, r7, r1
723 ; CHECK-NEXT: movw r1, #21984
724 ; CHECK-NEXT: add.w r8, r7, r1
725 ; CHECK-NEXT: movw r1, #21952
726 ; CHECK-NEXT: add.w r9, r7, r1
727 ; CHECK-NEXT: movw r1, #21936
728 ; CHECK-NEXT: movw r0, #21968
729 ; CHECK-NEXT: adds r5, r7, r1
730 ; CHECK-NEXT: movw r1, #21920
731 ; CHECK-NEXT: movw r3, #21904
732 ; CHECK-NEXT: adds r4, r7, r3
733 ; CHECK-NEXT: add r0, r7
734 ; CHECK-NEXT: add r1, r7
735 ; CHECK-NEXT: add.w r3, r7, #22016
736 ; CHECK-NEXT: add.w r6, r7, #21888
737 ; CHECK-NEXT: ldr r7, [sp, #8] @ 4-byte Reload
738 ; CHECK-NEXT: mov.w r10, #5
739 ; CHECK-NEXT: vmov.i16 q1, #0x5
740 ; CHECK-NEXT: mov.w r11, #0
741 ; CHECK-NEXT: dls lr, r7
742 ; CHECK-NEXT: mov.w r7, #327685
743 ; CHECK-NEXT: .LBB19_15: @ %for.cond8.preheader.3
744 ; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
745 ; CHECK-NEXT: str r7, [r2, #-4]
746 ; CHECK-NEXT: vstrh.16 q1, [r2, #-36]
747 ; CHECK-NEXT: strh.w r10, [r2]
748 ; CHECK-NEXT: vstrh.16 q1, [r2, #-20]
749 ; CHECK-NEXT: vstrw.32 q0, [r0]
750 ; CHECK-NEXT: vstrh.16 q0, [r4], #152
751 ; CHECK-NEXT: vstrh.16 q0, [r6], #152
752 ; CHECK-NEXT: vstrh.16 q0, [r1], #152
753 ; CHECK-NEXT: vstrh.16 q0, [r5], #152
754 ; CHECK-NEXT: vstrh.16 q0, [r9], #152
755 ; CHECK-NEXT: vstrh.16 q0, [r8], #152
756 ; CHECK-NEXT: vstrh.16 q0, [r12], #152
757 ; CHECK-NEXT: vstrh.16 q0, [r3], #152
758 ; CHECK-NEXT: strd r10, r11, [r0, #64]
759 ; CHECK-NEXT: adds r2, #38
760 ; CHECK-NEXT: adds r0, #152
761 ; CHECK-NEXT: le lr, .LBB19_15
762 ; CHECK-NEXT: @ %bb.16: @ %for.cond.cleanup6.3
763 ; CHECK-NEXT: add sp, #12
764 ; CHECK-NEXT: pop.w {r4, r5, r6, r7, r8, r9, r10, r11, pc}
765 ; CHECK-NEXT: .p2align 4
766 ; CHECK-NEXT: @ %bb.17:
767 ; CHECK-NEXT: .LCPI19_0:
768 ; CHECK-NEXT: .long 5 @ 0x5
769 ; CHECK-NEXT: .long 0 @ 0x0
770 ; CHECK-NEXT: .long 5 @ 0x5
771 ; CHECK-NEXT: .long 0 @ 0x0
773 %add = select i1 %b, i32 12, i32 11
774 %0 = mul nuw nsw i32 %add, 38
775 call void @llvm.memset.p0.i32(ptr noundef nonnull align 2 dereferenceable(1) @arr_22, i8 0, i32 %0, i1 false)
776 br label %for.cond8.preheader
778 for.cond8.preheader: ; preds = %entry, %for.cond8.preheader
779 %d.051 = phi i32 [ 0, %entry ], [ %inc, %for.cond8.preheader ]
780 %arrayidx16 = getelementptr inbounds [17 x [12 x [19 x i16]]], ptr @arr_21, i32 0, i32 0, i32 %d.051, i32 0
781 %arrayidx21 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 0, i32 %d.051, i32 0
782 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21, align 8
783 %arrayidx21.2 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 0, i32 %d.051, i32 2
784 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.2, align 8
785 %arrayidx21.4 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 0, i32 %d.051, i32 4
786 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.4, align 8
787 %arrayidx21.6 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 0, i32 %d.051, i32 6
788 store <8 x i16> <i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5>, ptr %arrayidx16, align 2
789 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.6, align 8
790 %arrayidx16.8 = getelementptr inbounds [17 x [12 x [19 x i16]]], ptr @arr_21, i32 0, i32 0, i32 %d.051, i32 8
791 %arrayidx21.8 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 0, i32 %d.051, i32 8
792 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.8, align 8
793 %arrayidx21.10 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 0, i32 %d.051, i32 10
794 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.10, align 8
795 %arrayidx21.12 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 0, i32 %d.051, i32 12
796 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.12, align 8
797 %arrayidx21.14 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 0, i32 %d.051, i32 14
798 store <8 x i16> <i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5>, ptr %arrayidx16.8, align 2
799 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.14, align 8
800 %arrayidx16.16 = getelementptr inbounds [17 x [12 x [19 x i16]]], ptr @arr_21, i32 0, i32 0, i32 %d.051, i32 16
801 store i16 5, ptr %arrayidx16.16, align 2
802 %arrayidx21.16 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 0, i32 %d.051, i32 16
803 %arrayidx16.17 = getelementptr inbounds [17 x [12 x [19 x i16]]], ptr @arr_21, i32 0, i32 0, i32 %d.051, i32 17
804 store i16 5, ptr %arrayidx16.17, align 2
805 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.16, align 8
806 %arrayidx16.18 = getelementptr inbounds [17 x [12 x [19 x i16]]], ptr @arr_21, i32 0, i32 0, i32 %d.051, i32 18
807 store i16 5, ptr %arrayidx16.18, align 2
808 %arrayidx21.18 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 0, i32 %d.051, i32 18
809 store i64 5, ptr %arrayidx21.18, align 8
810 %inc = add nuw nsw i32 %d.051, 1
811 %exitcond.not = icmp eq i32 %inc, %add
812 br i1 %exitcond.not, label %for.cond.cleanup6, label %for.cond8.preheader
814 for.cond.cleanup6: ; preds = %for.cond8.preheader
815 call void @llvm.memset.p0.i32(ptr noundef nonnull align 2 dereferenceable(1) getelementptr inbounds ([17 x [12 x [19 x i16]]], ptr @arr_22, i32 0, i32 4, i32 0, i32 0), i8 0, i32 %0, i1 false)
816 br label %for.cond8.preheader.1
818 for.cond8.preheader.1: ; preds = %for.cond8.preheader.1, %for.cond.cleanup6
819 %d.051.1 = phi i32 [ 0, %for.cond.cleanup6 ], [ %inc.1, %for.cond8.preheader.1 ]
820 %arrayidx16.1 = getelementptr inbounds [17 x [12 x [19 x i16]]], ptr @arr_21, i32 0, i32 4, i32 %d.051.1, i32 0
821 %arrayidx21.1 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 4, i32 %d.051.1, i32 0
822 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.1, align 8
823 %arrayidx21.2.1 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 4, i32 %d.051.1, i32 2
824 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.2.1, align 8
825 %arrayidx21.4.1 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 4, i32 %d.051.1, i32 4
826 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.4.1, align 8
827 %arrayidx21.6.1 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 4, i32 %d.051.1, i32 6
828 store <8 x i16> <i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5>, ptr %arrayidx16.1, align 2
829 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.6.1, align 8
830 %arrayidx16.8.1 = getelementptr inbounds [17 x [12 x [19 x i16]]], ptr @arr_21, i32 0, i32 4, i32 %d.051.1, i32 8
831 %arrayidx21.8.1 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 4, i32 %d.051.1, i32 8
832 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.8.1, align 8
833 %arrayidx21.10.1 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 4, i32 %d.051.1, i32 10
834 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.10.1, align 8
835 %arrayidx21.12.1 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 4, i32 %d.051.1, i32 12
836 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.12.1, align 8
837 %arrayidx21.14.1 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 4, i32 %d.051.1, i32 14
838 store <8 x i16> <i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5>, ptr %arrayidx16.8.1, align 2
839 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.14.1, align 8
840 %arrayidx16.16.1 = getelementptr inbounds [17 x [12 x [19 x i16]]], ptr @arr_21, i32 0, i32 4, i32 %d.051.1, i32 16
841 store i16 5, ptr %arrayidx16.16.1, align 2
842 %arrayidx21.16.1 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 4, i32 %d.051.1, i32 16
843 %arrayidx16.17.1 = getelementptr inbounds [17 x [12 x [19 x i16]]], ptr @arr_21, i32 0, i32 4, i32 %d.051.1, i32 17
844 store i16 5, ptr %arrayidx16.17.1, align 2
845 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.16.1, align 8
846 %arrayidx16.18.1 = getelementptr inbounds [17 x [12 x [19 x i16]]], ptr @arr_21, i32 0, i32 4, i32 %d.051.1, i32 18
847 store i16 5, ptr %arrayidx16.18.1, align 2
848 %arrayidx21.18.1 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 4, i32 %d.051.1, i32 18
849 store i64 5, ptr %arrayidx21.18.1, align 8
850 %inc.1 = add nuw nsw i32 %d.051.1, 1
851 %exitcond.not.1 = icmp eq i32 %inc.1, %add
852 br i1 %exitcond.not.1, label %for.cond.cleanup6.1, label %for.cond8.preheader.1
854 for.cond.cleanup6.1: ; preds = %for.cond8.preheader.1
855 call void @llvm.memset.p0.i32(ptr noundef nonnull align 2 dereferenceable(1) getelementptr inbounds ([17 x [12 x [19 x i16]]], ptr @arr_22, i32 0, i32 8, i32 0, i32 0), i8 0, i32 %0, i1 false)
856 br label %for.cond8.preheader.2
858 for.cond8.preheader.2: ; preds = %for.cond8.preheader.2, %for.cond.cleanup6.1
859 %d.051.2 = phi i32 [ 0, %for.cond.cleanup6.1 ], [ %inc.2, %for.cond8.preheader.2 ]
860 %arrayidx16.2 = getelementptr inbounds [17 x [12 x [19 x i16]]], ptr @arr_21, i32 0, i32 8, i32 %d.051.2, i32 0
861 %arrayidx21.254 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 8, i32 %d.051.2, i32 0
862 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.254, align 8
863 %arrayidx21.2.2 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 8, i32 %d.051.2, i32 2
864 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.2.2, align 8
865 %arrayidx21.4.2 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 8, i32 %d.051.2, i32 4
866 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.4.2, align 8
867 %arrayidx21.6.2 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 8, i32 %d.051.2, i32 6
868 store <8 x i16> <i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5>, ptr %arrayidx16.2, align 2
869 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.6.2, align 8
870 %arrayidx16.8.2 = getelementptr inbounds [17 x [12 x [19 x i16]]], ptr @arr_21, i32 0, i32 8, i32 %d.051.2, i32 8
871 %arrayidx21.8.2 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 8, i32 %d.051.2, i32 8
872 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.8.2, align 8
873 %arrayidx21.10.2 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 8, i32 %d.051.2, i32 10
874 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.10.2, align 8
875 %arrayidx21.12.2 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 8, i32 %d.051.2, i32 12
876 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.12.2, align 8
877 %arrayidx21.14.2 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 8, i32 %d.051.2, i32 14
878 store <8 x i16> <i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5>, ptr %arrayidx16.8.2, align 2
879 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.14.2, align 8
880 %arrayidx16.16.2 = getelementptr inbounds [17 x [12 x [19 x i16]]], ptr @arr_21, i32 0, i32 8, i32 %d.051.2, i32 16
881 store i16 5, ptr %arrayidx16.16.2, align 2
882 %arrayidx21.16.2 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 8, i32 %d.051.2, i32 16
883 %arrayidx16.17.2 = getelementptr inbounds [17 x [12 x [19 x i16]]], ptr @arr_21, i32 0, i32 8, i32 %d.051.2, i32 17
884 store i16 5, ptr %arrayidx16.17.2, align 2
885 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.16.2, align 8
886 %arrayidx16.18.2 = getelementptr inbounds [17 x [12 x [19 x i16]]], ptr @arr_21, i32 0, i32 8, i32 %d.051.2, i32 18
887 store i16 5, ptr %arrayidx16.18.2, align 2
888 %arrayidx21.18.2 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 8, i32 %d.051.2, i32 18
889 store i64 5, ptr %arrayidx21.18.2, align 8
890 %inc.2 = add nuw nsw i32 %d.051.2, 1
891 %exitcond.not.2 = icmp eq i32 %inc.2, %add
892 br i1 %exitcond.not.2, label %for.cond.cleanup6.2, label %for.cond8.preheader.2
894 for.cond.cleanup6.2: ; preds = %for.cond8.preheader.2
895 call void @llvm.memset.p0.i32(ptr noundef nonnull align 2 dereferenceable(1) getelementptr inbounds ([17 x [12 x [19 x i16]]], ptr @arr_22, i32 0, i32 12, i32 0, i32 0), i8 0, i32 %0, i1 false)
896 br label %for.cond8.preheader.3
898 for.cond8.preheader.3: ; preds = %for.cond8.preheader.3, %for.cond.cleanup6.2
899 %d.051.3 = phi i32 [ 0, %for.cond.cleanup6.2 ], [ %inc.3, %for.cond8.preheader.3 ]
900 %arrayidx16.3 = getelementptr inbounds [17 x [12 x [19 x i16]]], ptr @arr_21, i32 0, i32 12, i32 %d.051.3, i32 0
901 %arrayidx21.3 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 12, i32 %d.051.3, i32 0
902 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.3, align 8
903 %arrayidx21.2.3 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 12, i32 %d.051.3, i32 2
904 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.2.3, align 8
905 %arrayidx21.4.3 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 12, i32 %d.051.3, i32 4
906 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.4.3, align 8
907 %arrayidx21.6.3 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 12, i32 %d.051.3, i32 6
908 store <8 x i16> <i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5>, ptr %arrayidx16.3, align 2
909 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.6.3, align 8
910 %arrayidx16.8.3 = getelementptr inbounds [17 x [12 x [19 x i16]]], ptr @arr_21, i32 0, i32 12, i32 %d.051.3, i32 8
911 %arrayidx21.8.3 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 12, i32 %d.051.3, i32 8
912 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.8.3, align 8
913 %arrayidx21.10.3 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 12, i32 %d.051.3, i32 10
914 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.10.3, align 8
915 %arrayidx21.12.3 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 12, i32 %d.051.3, i32 12
916 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.12.3, align 8
917 %arrayidx21.14.3 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 12, i32 %d.051.3, i32 14
918 store <8 x i16> <i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5>, ptr %arrayidx16.8.3, align 2
919 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.14.3, align 8
920 %arrayidx16.16.3 = getelementptr inbounds [17 x [12 x [19 x i16]]], ptr @arr_21, i32 0, i32 12, i32 %d.051.3, i32 16
921 store i16 5, ptr %arrayidx16.16.3, align 2
922 %arrayidx21.16.3 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 12, i32 %d.051.3, i32 16
923 %arrayidx16.17.3 = getelementptr inbounds [17 x [12 x [19 x i16]]], ptr @arr_21, i32 0, i32 12, i32 %d.051.3, i32 17
924 store i16 5, ptr %arrayidx16.17.3, align 2
925 store <2 x i64> <i64 5, i64 5>, ptr %arrayidx21.16.3, align 8
926 %arrayidx16.18.3 = getelementptr inbounds [17 x [12 x [19 x i16]]], ptr @arr_21, i32 0, i32 12, i32 %d.051.3, i32 18
927 store i16 5, ptr %arrayidx16.18.3, align 2
928 %arrayidx21.18.3 = getelementptr inbounds [17 x [12 x [19 x i64]]], ptr @arr_20, i32 0, i32 12, i32 %d.051.3, i32 18
929 store i64 5, ptr %arrayidx21.18.3, align 8
930 %inc.3 = add nuw nsw i32 %d.051.3, 1
931 %exitcond.not.3 = icmp eq i32 %inc.3, %add
932 br i1 %exitcond.not.3, label %for.cond.cleanup6.3, label %for.cond8.preheader.3
934 for.cond.cleanup6.3: ; preds = %for.cond8.preheader.3
938 attributes #0 = { noinline optnone }
939 attributes #1 = { optsize }