aboutsummaryrefslogtreecommitdiff
path: root/sysdeps/powerpc/powerpc64/power7/memmove.S
blob: 3bd4b4bb3fc86b86a37410523a37013c5501d960 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
/* Optimized memmove implementation for PowerPC64/POWER7.
   Copyright (C) 2014-2015 Free Software Foundation, Inc.
   This file is part of the GNU C Library.

   The GNU C Library is free software; you can redistribute it and/or
   modify it under the terms of the GNU Lesser General Public
   License as published by the Free Software Foundation; either
   version 2.1 of the License, or (at your option) any later version.

   The GNU C Library is distributed in the hope that it will be useful,
   but WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
   Lesser General Public License for more details.

   You should have received a copy of the GNU Lesser General Public
   License along with the GNU C Library; if not, see
   <http://www.gnu.org/licenses/>.  */

#include <sysdep.h>


/* void* [r3] memmove (void *dest [r3], const void *src [r4], size_t len [r5])

   This optimization check if memory 'dest'  overlaps with 'src'. If it does
   not then it calls an optimized memcpy call (similar to memcpy for POWER7,
   embedded here to gain some cycles).
   If source and destiny overlaps, a optimized backwards memcpy is used
   instead.  */

	.machine power7
EALIGN (memmove, 5, 0)
	CALL_MCOUNT 3

L(_memmove):
	subf    r9,r4,r3
	cmpld   cr7,r9,r5
	blt	cr7,L(memmove_bwd)

	cmpldi	cr1,r5,31
	neg	0,3
	ble	cr1, L(copy_LT_32)  /* If move < 32 bytes use short move
				       code.  */

	andi.	10,3,15
	clrldi	11,4,60
	cmpld	cr6,10,11	/* SRC and DST alignments match?  */

	mr	r11,3
	bne	cr6,L(copy_GE_32_unaligned)
	beq	L(aligned_copy)

	mtocrf	0x01,0
	clrldi	0,0,60

/* Get the DST and SRC aligned to 8 bytes (16 for little-endian).  */
1:
	bf	31,2f
	lbz	6,0(r4)
	addi	r4,r4,1
	stb	6,0(r11)
	addi	r11,r11,1
2:
	bf	30,4f
	lhz	6,0(r4)
	addi	r4,r4,2
	sth	6,0(r11)
	addi	r11,r11,2
4:
	bf	29,8f
	lwz	6,0(r4)
	addi	r4,r4,4
	stw	6,0(r11)
	addi	r11,r11,4
8:
	bf	28,16f
	ld	6,0(r4)
	addi	r4,r4,8
	std	6,0(r11)
	addi	r11,r11,8
16:
	subf	r5,0,r5

/* Main aligned copy loop. Copies 128 bytes at a time. */
L(aligned_copy):
	li	6,16
	li	7,32
	li	8,48
	mtocrf	0x02,r5
	srdi	12,r5,7
	cmpdi	12,0
	beq	L(aligned_tail)
	lxvd2x	6,0,r4
	lxvd2x	7,r4,6
	mtctr	12
	b	L(aligned_128loop)

	.align  4
L(aligned_128head):
	/* for the 2nd + iteration of this loop. */
	lxvd2x	6,0,r4
	lxvd2x	7,r4,6
L(aligned_128loop):
	lxvd2x	8,r4,7
	lxvd2x	9,r4,8
	stxvd2x	6,0,r11
	addi	r4,r4,64
	stxvd2x	7,r11,6
	stxvd2x	8,r11,7
	stxvd2x	9,r11,8
	lxvd2x	6,0,r4
	lxvd2x	7,r4,6
	addi	r11,r11,64
	lxvd2x	8,r4,7
	lxvd2x	9,r4,8
	addi	r4,r4,64
	stxvd2x	6,0,r11
	stxvd2x	7,r11,6
	stxvd2x	8,r11,7
	stxvd2x	9,r11,8
	addi	r11,r11,64
	bdnz	L(aligned_128head)

L(aligned_tail):
	mtocrf	0x01,r5
	bf	25,32f
	lxvd2x	6,0,r4
	lxvd2x	7,r4,6
	lxvd2x	8,r4,7
	lxvd2x	9,r4,8
	addi	r4,r4,64
	stxvd2x	6,0,r11
	stxvd2x	7,r11,6
	stxvd2x	8,r11,7
	stxvd2x	9,r11,8
	addi	r11,r11,64
32:
	bf	26,16f
	lxvd2x	6,0,r4
	lxvd2x	7,r4,6
	addi	r4,r4,32
	stxvd2x	6,0,r11
	stxvd2x	7,r11,6
	addi	r11,r11,32
16:
	bf	27,8f
	lxvd2x	6,0,r4
	addi	r4,r4,16
	stxvd2x	6,0,r11
	addi	r11,r11,16
8:
	bf	28,4f
	ld	6,0(r4)
	addi	r4,r4,8
	std     6,0(r11)
	addi	r11,r11,8
4:	/* Copies 4~7 bytes.  */
	bf	29,L(tail2)
	lwz	6,0(r4)
	stw     6,0(r11)
	bf      30,L(tail5)
	lhz     7,4(r4)
	sth     7,4(r11)
	bflr	31
	lbz     8,6(r4)
	stb     8,6(r11)
	/* Return original DST pointer.  */
	blr

/* Handle copies of 0~31 bytes.  */
	.align	4
L(copy_LT_32):
	mr	r11,3
	cmpldi	cr6,r5,8
	mtocrf	0x01,r5
	ble	cr6,L(copy_LE_8)

	/* At least 9 bytes to go.  */
	neg	8,4
	andi.	0,8,3
	cmpldi	cr1,r5,16
	beq	L(copy_LT_32_aligned)

	/* Force 4-byte alignment for SRC.  */
	mtocrf	0x01,0
	subf	r5,0,r5
2:
	bf	30,1f
	lhz	6,0(r4)
	addi	r4,r4,2
	sth	6,0(r11)
	addi	r11,r11,2
1:
	bf	31,L(end_4bytes_alignment)
	lbz	6,0(r4)
	addi	r4,r4,1
	stb	6,0(r11)
	addi	r11,r11,1

	.align	4
L(end_4bytes_alignment):
	cmpldi	cr1,r5,16
	mtocrf	0x01,r5

L(copy_LT_32_aligned):
	/* At least 6 bytes to go, and SRC is word-aligned.  */
	blt	cr1,8f

	/* Copy 16 bytes.  */
	lwz	6,0(r4)
	lwz	7,4(r4)
	stw	6,0(r11)
	lwz	8,8(r4)
	stw	7,4(r11)
	lwz	6,12(r4)
	addi	r4,r4,16
	stw	8,8(r11)
	stw	6,12(r11)
	addi	r11,r11,16
8:	/* Copy 8 bytes.  */
	bf	28,L(tail4)
	lwz	6,0(r4)
	lwz	7,4(r4)
	addi	r4,r4,8
	stw	6,0(r11)
	stw	7,4(r11)
	addi	r11,r11,8

	.align	4
/* Copies 4~7 bytes.  */
L(tail4):
	bf	29,L(tail2)
	lwz	6,0(r4)
	stw	6,0(r11)
	bf	30,L(tail5)
	lhz	7,4(r4)
	sth	7,4(r11)
	bflr	31
	lbz	8,6(r4)
	stb	8,6(r11)
	/* Return original DST pointer.  */
	blr

	.align	4
/* Copies 2~3 bytes.  */
L(tail2):
	bf	30,1f
	lhz	6,0(r4)
	sth	6,0(r11)
	bflr	31
	lbz	7,2(r4)
	stb	7,2(r11)
	blr

	.align	4
L(tail5):
	bflr	31
	lbz	6,4(r4)
	stb	6,4(r11)
	blr

	.align	4
1:
	bflr	31
	lbz	6,0(r4)
	stb	6,0(r11)
	/* Return original DST pointer.  */
	blr

/* Handles copies of 0~8 bytes.  */
	.align	4
L(copy_LE_8):
	bne	cr6,L(tail4)

	/* Though we could've used ld/std here, they are still
	slow for unaligned cases.  */

	lwz	6,0(r4)
	lwz	7,4(r4)
	stw	6,0(r11)
	stw	7,4(r11)
	blr


/* Handle copies of 32+ bytes where DST is aligned (to quadword) but
   SRC is not.	Use aligned quadword loads from SRC, shifted to realign
   the data, allowing for aligned DST stores.  */
	.align	4
L(copy_GE_32_unaligned):
	clrldi	0,0,60	      /* Number of bytes until the 1st r11 quadword.  */
	srdi	9,r5,4	      /* Number of full quadwords remaining.  */

	beq	L(copy_GE_32_unaligned_cont)

	/* DST is not quadword aligned, get it aligned.  */

	mtocrf	0x01,0
	subf	r5,0,r5

	/* Vector instructions work best when proper alignment (16-bytes)
	is present.  Move 0~15 bytes as needed to get DST quadword-aligned.  */
1:
	bf	31,2f
	lbz	6,0(r4)
	addi	r4,r4,1
	stb	6,0(r11)
	addi	r11,r11,1
2:
	bf	30,4f
	lhz	6,0(r4)
	addi	r4,r4,2
	sth	6,0(r11)
	addi	r11,r11,2
4:
	bf	29,8f
	lwz	6,0(r4)
	addi	r4,r4,4
	stw	6,0(r11)
	addi	r11,r11,4
8:
	bf	28,0f
	ld	6,0(r4)
	addi	r4,r4,8
	std	6,0(r11)
	addi	r11,r11,8
0:
	srdi	9,r5,4	      /* Number of full quadwords remaining.  */

	/* The proper alignment is present, it is OK to copy the bytes now.  */
L(copy_GE_32_unaligned_cont):

	/* Setup two indexes to speed up the indexed vector operations.  */
	clrldi	10,r5,60
	li	6,16	      /* Index for 16-bytes offsets.  */
	li	7,32	      /* Index for 32-bytes offsets.  */
	cmpldi	cr1,10,0
	srdi	8,r5,5	      /* Setup the loop counter.  */
	mtocrf	0x01,9
	cmpldi	cr6,9,1
#ifdef __LITTLE_ENDIAN__
	lvsr	5,0,r4
#else
	lvsl	5,0,r4
#endif
	lvx	3,0,r4
	li	0,0
	bf	31,L(setup_unaligned_loop)

	/* Copy another 16 bytes to align to 32-bytes due to the loop.  */
	lvx	4,r4,6
#ifdef __LITTLE_ENDIAN__
	vperm	6,4,3,5
#else
	vperm	6,3,4,5
#endif
	addi	r4,r4,16
	stvx	6,0,r11
	addi	r11,r11,16
	vor	3,4,4
	clrrdi	0,r4,60

L(setup_unaligned_loop):
	mtctr	8
	ble	cr6,L(end_unaligned_loop)

	/* Copy 32 bytes at a time using vector instructions.  */
	.align	4
L(unaligned_loop):

	/* Note: vr6/vr10 may contain data that was already copied,
	but in order to get proper alignment, we may have to copy
	some portions again. This is faster than having unaligned
	vector instructions though.  */

	lvx	4,r4,6
#ifdef __LITTLE_ENDIAN__
	vperm	6,4,3,5
#else
	vperm	6,3,4,5
#endif
	lvx	3,r4,7
#ifdef __LITTLE_ENDIAN__
	vperm	10,3,4,5
#else
	vperm	10,4,3,5
#endif
	addi	r4,r4,32
	stvx	6,0,r11
	stvx	10,r11,6
	addi	r11,r11,32
	bdnz	L(unaligned_loop)

	clrrdi	0,r4,60

	.align	4
L(end_unaligned_loop):

	/* Check for tail bytes.  */
	mtocrf	0x01,r5
	beqlr	cr1

	add	r4,r4,0

	/*  We have 1~15 tail bytes to copy, and DST is quadword aligned.  */
	/* Copy 8 bytes.  */
	bf	28,4f
	lwz	6,0(r4)
	lwz	7,4(r4)
	addi	r4,r4,8
	stw	6,0(r11)
	stw	7,4(r11)
	addi	r11,r11,8
4:	/* Copy 4~7 bytes.  */
	bf	29,L(tail2)
	lwz	6,0(r4)
	stw	6,0(r11)
	bf	30,L(tail5)
	lhz	7,4(r4)
	sth	7,4(r11)
	bflr	31
	lbz	8,6(r4)
	stb	8,6(r11)
	/* Return original DST pointer.  */
	blr

	/* Start to memcpy backward implementation: the algorith first check if
	   src and dest have the same alignment and if it does align both to 16
	   bytes and copy using VSX instructions.
	   If does not, align dest to 16 bytes and use VMX (altivec) instruction
	   to read two 16 bytes at time, shift/permute the bytes read and write
	   aligned to dest.  */
L(memmove_bwd):
	cmpldi	cr1,r5,31
	/* Copy is done backwards: update the pointers and check alignment.  */
	add	r11,r3,r5
	add	r4,r4,r5
	mr	r0,r11
	ble	cr1, L(copy_LT_32_bwd)  /* If move < 32 bytes use short move
				           code.  */

	andi.	r10,r11,15	    /* Check if r11 is aligned to 16 bytes  */
	clrldi	r9,r4,60	    /* Check if r4 is aligned to 16 bytes  */
	cmpld	cr6,r10,r9	    /* SRC and DST alignments match?  */

	bne     cr6,L(copy_GE_32_unaligned_bwd)
	beq     L(aligned_copy_bwd)

	mtocrf	0x01,r0
	clrldi	r0,r0,60

/* Get the DST and SRC aligned to 16 bytes.  */
1:
	bf	31,2f
	lbz	r6,-1(r4)
	subi	r4,r4,1
	stb	r6,-1(r11)
	subi	r11,r11,1
2:
	bf	30,4f
	lhz	r6,-2(r4)
	subi	r4,r4,2
	sth	r6,-2(r11)
	subi	r11,r11,2
4:
	bf	29,8f
	lwz	r6,-4(r4)
	subi	r4,r4,4
	stw	r6,-4(r11)
	subi	r11,r11,4
8:
	bf	28,16f
	ld	r6,-8(r4)
	subi	r4,r4,8
	std	r6,-8(r11)
	subi	r11,r11,8
16:
	subf	r5,0,r5

/* Main aligned copy loop. Copies 128 bytes at a time. */
L(aligned_copy_bwd):
	li	r6,-16
	li	r7,-32
	li	r8,-48
	li	r9,-64
	mtocrf	0x02,r5
	srdi	r12,r5,7
	cmpdi	r12,0
	beq	L(aligned_tail_bwd)
	lxvd2x	v6,r4,r6
	lxvd2x	v7,r4,r7
	mtctr	12
	b	L(aligned_128loop_bwd)

	.align  4
L(aligned_128head_bwd):
	/* for the 2nd + iteration of this loop. */
	lxvd2x	v6,r4,r6
	lxvd2x	v7,r4,r7
L(aligned_128loop_bwd):
	lxvd2x	v8,r4,r8
	lxvd2x	v9,r4,r9
	stxvd2x	v6,r11,r6
	subi	r4,r4,64
	stxvd2x	v7,r11,r7
	stxvd2x	v8,r11,r8
	stxvd2x	v9,r11,r9
	lxvd2x	v6,r4,r6
	lxvd2x	v7,r4,7
	subi	r11,r11,64
	lxvd2x	v8,r4,r8
	lxvd2x	v9,r4,r9
	subi	r4,r4,64
	stxvd2x	v6,r11,r6
	stxvd2x	v7,r11,r7
	stxvd2x	v8,r11,r8
	stxvd2x	v9,r11,r9
	subi	r11,r11,64
	bdnz	L(aligned_128head_bwd)

L(aligned_tail_bwd):
	mtocrf	0x01,r5
	bf	25,32f
	lxvd2x	v6,r4,r6
	lxvd2x	v7,r4,r7
	lxvd2x	v8,r4,r8
	lxvd2x	v9,r4,r9
	subi	r4,r4,64
	stxvd2x	v6,r11,r6
	stxvd2x	v7,r11,r7
	stxvd2x	v8,r11,r8
	stxvd2x	v9,r11,r9
	subi	r11,r11,64
32:
	bf	26,16f
	lxvd2x	v6,r4,r6
	lxvd2x	v7,r4,r7
	subi	r4,r4,32
	stxvd2x	v6,r11,r6
	stxvd2x	v7,r11,r7
	subi	r11,r11,32
16:
	bf	27,8f
	lxvd2x	v6,r4,r6
	subi	r4,r4,16
	stxvd2x	v6,r11,r6
	subi	r11,r11,16
8:
	bf	28,4f
	ld	r6,-8(r4)
	subi	r4,r4,8
	std     r6,-8(r11)
	subi	r11,r11,8
4:	/* Copies 4~7 bytes.  */
	bf	29,L(tail2_bwd)
	lwz	r6,-4(r4)
	stw     r6,-4(r11)
	bf      30,L(tail5_bwd)
	lhz     r7,-6(r4)
	sth     r7,-6(r11)
	bflr	31
	lbz     r8,-7(r4)
	stb     r8,-7(r11)
	/* Return original DST pointer.  */
	blr

/* Handle copies of 0~31 bytes.  */
	.align	4
L(copy_LT_32_bwd):
	cmpldi	cr6,r5,8
	mtocrf	0x01,r5
	ble	cr6,L(copy_LE_8_bwd)

	/* At least 9 bytes to go.  */
	neg	r8,r4
	andi.	r0,r8,3
	cmpldi	cr1,r5,16
	beq	L(copy_LT_32_aligned_bwd)

	/* Force 4-byte alignment for SRC.  */
	mtocrf	0x01,0
	subf	r5,0,r5
2:
	bf	30,1f
	lhz	r6,-2(r4)
	subi	r4,r4,2
	sth	r6,-2(r11)
	subi	r11,r11,2
1:
	bf	31,L(end_4bytes_alignment_bwd)
	lbz	6,-1(r4)
	subi	r4,r4,1
	stb	6,-1(r11)
	subi	r11,r11,1

	.align	4
L(end_4bytes_alignment_bwd):
	cmpldi	cr1,r5,16
	mtocrf	0x01,r5

L(copy_LT_32_aligned_bwd):
	/* At least 6 bytes to go, and SRC is word-aligned.  */
	blt	cr1,8f

	/* Copy 16 bytes.  */
	lwz	r6,-4(r4)
	lwz	r7,-8(r4)
	stw	r6,-4(r11)
	lwz	r8,-12(r4)
	stw	r7,-8(r11)
	lwz	r6,-16(r4)
	subi	r4,r4,16
	stw	r8,-12(r11)
	stw	r6,-16(r11)
	subi	r11,r11,16
8:	/* Copy 8 bytes.  */
	bf	28,L(tail4_bwd)
	lwz	r6,-4(r4)
	lwz	r7,-8(r4)
	subi	r4,r4,8
	stw	r6,-4(r11)
	stw	r7,-8(r11)
	subi	r11,r11,8

	.align	4
/* Copies 4~7 bytes.  */
L(tail4_bwd):
	bf	29,L(tail2_bwd)
	lwz	6,-4(r4)
	stw	6,-4(r11)
	bf	30,L(tail5_bwd)
	lhz	7,-6(r4)
	sth	7,-6(r11)
	bflr	31
	lbz	8,-7(r4)
	stb	8,-7(r11)
	/* Return original DST pointer.  */
	blr

	.align	4
/* Copies 2~3 bytes.  */
L(tail2_bwd):
	bf	30,1f
	lhz	6,-2(r4)
	sth	6,-2(r11)
	bflr	31
	lbz	7,-3(r4)
	stb	7,-3(r11)
	blr

	.align	4
L(tail5_bwd):
	bflr	31
	lbz	6,-5(r4)
	stb	6,-5(r11)
	blr

	.align	4
1:
	bflr	31
	lbz	6,-1(r4)
	stb	6,-1(r11)
	/* Return original DST pointer.  */
	blr


/* Handles copies of 0~8 bytes.  */
	.align	4
L(copy_LE_8_bwd):
	bne	cr6,L(tail4_bwd)

	/* Though we could've used ld/std here, they are still
	   slow for unaligned cases.  */
	lwz	6,-8(r4)
	lwz	7,-4(r4)
	stw	6,-8(r11)
	stw	7,-4(r11)
	blr


/* Handle copies of 32+ bytes where DST is aligned (to quadword) but
   SRC is not.	Use aligned quadword loads from SRC, shifted to realign
   the data, allowing for aligned DST stores.  */
	.align	4
L(copy_GE_32_unaligned_bwd):
	andi.	r10,r11,15      /* Check alignment of DST against 16 bytes..  */
	srdi	r9,r5,4		/* Number of full quadwords remaining.  */

	beq	L(copy_GE_32_unaligned_cont_bwd)

	/* DST is not quadword aligned and r10 holds the address masked to
           compare alignments.  */
	mtocrf	0x01,r10
	subf	r5,r10,r5

	/* Vector instructions work best when proper alignment (16-bytes)
	is present.  Move 0~15 bytes as needed to get DST quadword-aligned.  */
1:
	bf	31,2f
	lbz	r6,-1(r4)
	subi	r4,r4,1
	stb	r6,-1(r11)
	subi	r11,r11,1
2:
	bf	30,4f
	lhz	r6,-2(r4)
	subi	r4,r4,2
	sth	r6,-2(r11)
	subi	r11,r11,2
4:
	bf	29,8f
	lwz	r6,-4(r4)
	subi	r4,r4,4
	stw	r6,-4(r11)
	subi	r11,r11,4
8:
	bf	28,0f
	ld	r6,-8(r4)
	subi	r4,r4,8
	std	r6,-8(r11)
	subi	r11,r11,8
0:
	srdi	r9,r5,4	      /* Number of full quadwords remaining.  */

	/* The proper alignment is present, it is OK to copy the bytes now.  */
L(copy_GE_32_unaligned_cont_bwd):

	/* Setup two indexes to speed up the indexed vector operations.  */
	clrldi	r10,r5,60
	li	r6,-16	      /* Index for 16-bytes offsets.  */
	li	r7,-32	      /* Index for 32-bytes offsets.  */
	cmpldi	cr1,10,0
	srdi	r8,r5,5	      /* Setup the loop counter.  */
	mtocrf	0x01,9
	cmpldi	cr6,r9,1
#ifdef __LITTLE_ENDIAN__
	lvsr	v5,r0,r4
#else
	lvsl	v5,r0,r4
#endif
	lvx	v3,0,r4
	li	r0,0
	bf	31,L(setup_unaligned_loop_bwd)

	/* Copy another 16 bytes to align to 32-bytes due to the loop.  */
	lvx	v4,r4,r6
#ifdef __LITTLE_ENDIAN__
	vperm	v6,v3,v4,v5
#else
	vperm	v6,v4,v3,v5
#endif
	subi	r4,r4,16
	stvx	v6,r11,r6
	subi	r11,r11,16
	vor	v3,v4,v4
	clrrdi	r0,r4,60

L(setup_unaligned_loop_bwd):
	mtctr	r8
	ble	cr6,L(end_unaligned_loop_bwd)

	/* Copy 32 bytes at a time using vector instructions.  */
	.align	4
L(unaligned_loop_bwd):

	/* Note: vr6/vr10 may contain data that was already copied,
	but in order to get proper alignment, we may have to copy
	some portions again. This is faster than having unaligned
	vector instructions though.  */

	lvx	v4,r4,r6
#ifdef __LITTLE_ENDIAN__
	vperm	v6,v3,v4,v5
#else
	vperm	v6,v4,v3,v5
#endif
	lvx	v3,r4,r7
#ifdef __LITTLE_ENDIAN__
	vperm	v10,v4,v3,v5
#else
	vperm	v10,v3,v4,v5
#endif
	subi	r4,r4,32
	stvx	v6,r11,r6
	stvx	v10,r11,r7
	subi	r11,r11,32
	bdnz	L(unaligned_loop_bwd)

	clrrdi	r0,r4,60

	.align	4
L(end_unaligned_loop_bwd):

	/* Check for tail bytes.  */
	mtocrf	0x01,r5
	beqlr	cr1

	add	r4,r4,0

	/*  We have 1~15 tail bytes to copy, and DST is quadword aligned.  */
	/* Copy 8 bytes.  */
	bf	28,4f
	lwz	r6,-4(r4)
	lwz	r7,-8(r4)
	subi	r4,r4,8
	stw	r6,-4(r11)
	stw	r7,-8(r11)
	subi	r11,r11,8
4:	/* Copy 4~7 bytes.  */
	bf	29,L(tail2_bwd)
	lwz	r6,-4(r4)
	stw	r6,-4(r11)
	bf	30,L(tail5_bwd)
	lhz	r7,-6(r4)
	sth	r7,-6(r11)
	bflr	31
	lbz	r8,-7(r4)
	stb	r8,-7(r11)
	/* Return original DST pointer.  */
	blr
END_GEN_TB (memmove, TB_TOCLESS)
libc_hidden_builtin_def (memmove)


/* void bcopy(const void *src [r3], void *dest [r4], size_t n [r5])
   Implemented in this file to avoid linker create a stub function call
   in the branch to '_memmove'.  */
ENTRY (bcopy)
	mr	r6,r3
	mr	r3,r4
	mr	r4,r6
	b	L(_memmove)
END (bcopy)