summary refs log tree commit diff
path: root/sysdeps/powerpc/powerpc64/power4/memcmp.S
blob: 65922813bd1c18148c21348751f83fc397f539a6 (plain) (blame)
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
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
/* Optimized strcmp implementation for PowerPC64.
   Copyright (C) 2003-2013 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>
#include <bp-sym.h>
#include <bp-asm.h>

/* int [r3] memcmp (const char *s1 [r3], const char *s2 [r4], size_t size [r5])  */

	.machine power4
EALIGN (BP_SYM(memcmp), 4, 0)
	CALL_MCOUNT 3

#define rTMP	r0
#define rRTN	r3
#define rSTR1	r3	/* first string arg */
#define rSTR2	r4	/* second string arg */
#define rN	r5	/* max string length */
/* Note:  The Bounded pointer support in this code is broken.  This code
   was inherited from PPC32 and that support was never completed.
   Current PPC gcc does not support -fbounds-check or -fbounded-pointers.  */
#define rWORD1	r6	/* current word in s1 */
#define rWORD2	r7	/* current word in s2 */
#define rWORD3	r8	/* next word in s1 */
#define rWORD4	r9	/* next word in s2 */
#define rWORD5	r10	/* next word in s1 */
#define rWORD6	r11	/* next word in s2 */
#define rBITDIF	r12	/* bits that differ in s1 & s2 words */
#define rWORD7	r30	/* next word in s1 */
#define rWORD8	r31	/* next word in s2 */

	xor	rTMP, rSTR2, rSTR1
	cmpldi	cr6, rN, 0
	cmpldi	cr1, rN, 12
	clrldi.	rTMP, rTMP, 61
	clrldi	rBITDIF, rSTR1, 61
	cmpldi	cr5, rBITDIF, 0
	beq-	cr6, L(zeroLength)
	dcbt	0,rSTR1
	dcbt	0,rSTR2
/* If less than 8 bytes or not aligned, use the unalligned
   byte loop.  */
	blt	cr1, L(bytealigned)
	std	rWORD8,-8(r1)	
	cfi_offset(rWORD8,-8)
	std	rWORD7,-16(r1)	
	cfi_offset(rWORD7,-16)
	bne	L(unaligned)
/* At this point we know both strings have the same alignment and the
   compare length is at least 8 bytes.  rBITDIF containes the low order
   3 bits of rSTR1 and cr5 contains the result of the logical compare
   of rBITDIF to 0.  If rBITDIF == 0 then we are already double word 
   aligned and can perform the DWaligned loop.
  
   Otherwise we know the two strings have the same alignment (but not
   yet DW).  So we can force the string addresses to the next lower DW
   boundary and special case this first DW word using shift left to
   ellimiate bits preceeding the first byte.  Since we want to join the
   normal (DWaligned) compare loop, starting at the second double word,
   we need to adjust the length (rN) and special case the loop
   versioning for the first DW. This insures that the loop count is
   correct and the first DW (shifted) is in the expected resister pair.  */
	.align 4
L(samealignment):
	clrrdi	rSTR1, rSTR1, 3
	clrrdi	rSTR2, rSTR2, 3
	beq	cr5, L(DWaligned)
	add	rN, rN, rBITDIF
	sldi	r11, rBITDIF, 3
	srdi	rTMP, rN, 5	/* Divide by 32 */
	andi.	rBITDIF, rN, 24	/* Get the DW remainder */
	ld	rWORD1, 0(rSTR1)
	ld	rWORD2, 0(rSTR2)
	cmpldi	cr1, rBITDIF, 16
	cmpldi	cr7, rN, 32
	clrldi	rN, rN, 61
	beq	L(dPs4)
	mtctr   rTMP	/* Power4 wants mtctr 1st in dispatch group */
	bgt	cr1, L(dPs3)
	beq	cr1, L(dPs2)

/* Remainder is 8 */
	.align 3
L(dsP1):
	sld	rWORD5, rWORD1, r11
	sld	rWORD6, rWORD2, r11
	cmpld	cr5, rWORD5, rWORD6
	blt	cr7, L(dP1x)
/* Do something useful in this cycle since we have to branch anyway.  */
	ld	rWORD1, 8(rSTR1)
	ld	rWORD2, 8(rSTR2)
	cmpld	cr0, rWORD1, rWORD2
	b	L(dP1e)
/* Remainder is 16 */
	.align 4
L(dPs2):
	sld	rWORD5, rWORD1, r11
	sld	rWORD6, rWORD2, r11
	cmpld	cr6, rWORD5, rWORD6
	blt	cr7, L(dP2x)
/* Do something useful in this cycle since we have to branch anyway.  */
	ld	rWORD7, 8(rSTR1)
	ld	rWORD8, 8(rSTR2)
	cmpld	cr5, rWORD7, rWORD8
	b	L(dP2e)
/* Remainder is 24 */
	.align 4
L(dPs3):
	sld	rWORD3, rWORD1, r11
	sld	rWORD4, rWORD2, r11
	cmpld	cr1, rWORD3, rWORD4
	b	L(dP3e)
/* Count is a multiple of 32, remainder is 0 */
	.align 4
L(dPs4):
	mtctr   rTMP	/* Power4 wants mtctr 1st in dispatch group */
	sld	rWORD1, rWORD1, r11
	sld	rWORD2, rWORD2, r11
	cmpld	cr0, rWORD1, rWORD2
	b	L(dP4e)

/* At this point we know both strings are double word aligned and the
   compare length is at least 8 bytes.  */
	.align 4
L(DWaligned):
	andi.	rBITDIF, rN, 24	/* Get the DW remainder */
	srdi	rTMP, rN, 5	/* Divide by 32 */
	cmpldi	cr1, rBITDIF, 16
	cmpldi	cr7, rN, 32
	clrldi	rN, rN, 61
	beq	L(dP4)
	bgt	cr1, L(dP3)
	beq	cr1, L(dP2)
		
/* Remainder is 8 */
	.align 4
L(dP1):
	mtctr   rTMP	/* Power4 wants mtctr 1st in dispatch group */
/* Normally we'd use rWORD7/rWORD8 here, but since we might exit early
   (8-15 byte compare), we want to use only volitile registers.  This
   means we can avoid restoring non-volitile registers since we did not
   change any on the early exit path.  The key here is the non-early
   exit path only cares about the condition code (cr5), not about which 
   register pair was used.  */
	ld	rWORD5, 0(rSTR1)
	ld	rWORD6, 0(rSTR2)
	cmpld	cr5, rWORD5, rWORD6
	blt	cr7, L(dP1x)
	ld	rWORD1, 8(rSTR1)
	ld	rWORD2, 8(rSTR2)
	cmpld	cr0, rWORD1, rWORD2
L(dP1e):
	ld	rWORD3, 16(rSTR1)
	ld	rWORD4, 16(rSTR2)
	cmpld	cr1, rWORD3, rWORD4
	ld	rWORD5, 24(rSTR1)
	ld	rWORD6, 24(rSTR2)
	cmpld	cr6, rWORD5, rWORD6
	bne	cr5, L(dLcr5)
	bne	cr0, L(dLcr0)
	
	ldu	rWORD7, 32(rSTR1)
	ldu	rWORD8, 32(rSTR2)
	bne	cr1, L(dLcr1)
	cmpld	cr5, rWORD7, rWORD8
	bdnz	L(dLoop)
	bne	cr6, L(dLcr6)
	ld	rWORD8,-8(r1)
	ld	rWORD7,-16(r1)
	.align 3
L(dP1x):
	sldi.	r12, rN, 3
	bne	cr5, L(dLcr5)
	subfic	rN, r12, 64	/* Shift count is 64 - (rN * 8).  */
	bne	L(d00)
	li	rRTN, 0
	blr
		
/* Remainder is 16 */
	.align 4
L(dP2):
	mtctr   rTMP	/* Power4 wants mtctr 1st in dispatch group */
	ld	rWORD5, 0(rSTR1)
	ld	rWORD6, 0(rSTR2)
	cmpld	cr6, rWORD5, rWORD6
	blt	cr7, L(dP2x)
	ld	rWORD7, 8(rSTR1)
	ld	rWORD8, 8(rSTR2)
	cmpld	cr5, rWORD7, rWORD8
L(dP2e):
	ld	rWORD1, 16(rSTR1)
	ld	rWORD2, 16(rSTR2)
	cmpld	cr0, rWORD1, rWORD2
	ld	rWORD3, 24(rSTR1)
	ld	rWORD4, 24(rSTR2)
	cmpld	cr1, rWORD3, rWORD4
	addi	rSTR1, rSTR1, 8
	addi	rSTR2, rSTR2, 8
	bne	cr6, L(dLcr6)
	bne	cr5, L(dLcr5)
	b	L(dLoop2)
/* Again we are on a early exit path (16-23 byte compare), we want to
   only use volitile registers and avoid restoring non-volitile
   registers.  */
	.align 4
L(dP2x):
	ld	rWORD3, 8(rSTR1)
	ld	rWORD4, 8(rSTR2)
	cmpld	cr5, rWORD3, rWORD4
	sldi.	r12, rN, 3
	bne	cr6, L(dLcr6)
	addi	rSTR1, rSTR1, 8
	addi	rSTR2, rSTR2, 8
	bne	cr5, L(dLcr5)
	subfic	rN, r12, 64	/* Shift count is 64 - (rN * 8).  */
	bne	L(d00)
	li	rRTN, 0
	blr
		
/* Remainder is 24 */
	.align 4
L(dP3):
	mtctr   rTMP	/* Power4 wants mtctr 1st in dispatch group */
	ld	rWORD3, 0(rSTR1)
	ld	rWORD4, 0(rSTR2)
	cmpld	cr1, rWORD3, rWORD4
L(dP3e):
	ld	rWORD5, 8(rSTR1)
	ld	rWORD6, 8(rSTR2)
	cmpld	cr6, rWORD5, rWORD6
	blt	cr7, L(dP3x)
	ld	rWORD7, 16(rSTR1)
	ld	rWORD8, 16(rSTR2)
	cmpld	cr5, rWORD7, rWORD8
	ld	rWORD1, 24(rSTR1)
	ld	rWORD2, 24(rSTR2)
	cmpld	cr0, rWORD1, rWORD2
	addi	rSTR1, rSTR1, 16
	addi	rSTR2, rSTR2, 16
	bne	cr1, L(dLcr1)
	bne	cr6, L(dLcr6)
	b	L(dLoop1)
/* Again we are on a early exit path (24-31 byte compare), we want to
   only use volitile registers and avoid restoring non-volitile
   registers.  */
	.align 4
L(dP3x):
	ld	rWORD1, 16(rSTR1)
	ld	rWORD2, 16(rSTR2)
	cmpld	cr5, rWORD1, rWORD2
	sldi.	r12, rN, 3
	bne	cr1, L(dLcr1)
	addi	rSTR1, rSTR1, 16
	addi	rSTR2, rSTR2, 16
	bne	cr6, L(dLcr6)
	subfic	rN, r12, 64	/* Shift count is 64 - (rN * 8).  */
	bne	cr5, L(dLcr5)
	bne	L(d00)
	li	rRTN, 0
	blr
	
/* Count is a multiple of 32, remainder is 0 */
	.align 4
L(dP4):
	mtctr   rTMP	/* Power4 wants mtctr 1st in dispatch group */
	ld	rWORD1, 0(rSTR1)
	ld	rWORD2, 0(rSTR2)
	cmpld	cr0, rWORD1, rWORD2
L(dP4e):
	ld	rWORD3, 8(rSTR1)
	ld	rWORD4, 8(rSTR2)
	cmpld	cr1, rWORD3, rWORD4
	ld	rWORD5, 16(rSTR1)
	ld	rWORD6, 16(rSTR2)
	cmpld	cr6, rWORD5, rWORD6
	ldu	rWORD7, 24(rSTR1)
	ldu	rWORD8, 24(rSTR2)
	cmpld	cr5, rWORD7, rWORD8
	bne	cr0, L(dLcr0)
	bne	cr1, L(dLcr1)
	bdz-	L(d24)		/* Adjust CTR as we start with +4 */
/* This is the primary loop */
	.align 4
L(dLoop):
	ld	rWORD1, 8(rSTR1)
	ld	rWORD2, 8(rSTR2)
	cmpld	cr1, rWORD3, rWORD4
	bne	cr6, L(dLcr6)
L(dLoop1):
	ld	rWORD3, 16(rSTR1)
	ld	rWORD4, 16(rSTR2)
	cmpld	cr6, rWORD5, rWORD6
	bne	cr5, L(dLcr5)
L(dLoop2):
	ld	rWORD5, 24(rSTR1)
	ld	rWORD6, 24(rSTR2)
	cmpld	cr5, rWORD7, rWORD8
	bne	cr0, L(dLcr0)
L(dLoop3):
	ldu	rWORD7, 32(rSTR1)
	ldu	rWORD8, 32(rSTR2)
	bne-	cr1, L(dLcr1)
	cmpld	cr0, rWORD1, rWORD2
	bdnz+	L(dLoop)	
	
L(dL4):
	cmpld	cr1, rWORD3, rWORD4
	bne	cr6, L(dLcr6)
	cmpld	cr6, rWORD5, rWORD6
	bne	cr5, L(dLcr5)
	cmpld	cr5, rWORD7, rWORD8
L(d44):
	bne	cr0, L(dLcr0)
L(d34):
	bne	cr1, L(dLcr1)
L(d24):
	bne	cr6, L(dLcr6)
L(d14):
	sldi.	r12, rN, 3
	bne	cr5, L(dLcr5) 
L(d04):
	ld	rWORD8,-8(r1)
	ld	rWORD7,-16(r1)
	subfic	rN, r12, 64	/* Shift count is 64 - (rN * 8).  */
	beq	L(zeroLength)
/* At this point we have a remainder of 1 to 7 bytes to compare.  Since
   we are aligned it is safe to load the whole double word, and use
   shift right double to elliminate bits beyond the compare length.  */ 
L(d00):
	ld	rWORD1, 8(rSTR1)
	ld	rWORD2, 8(rSTR2) 
	srd	rWORD1, rWORD1, rN
	srd	rWORD2, rWORD2, rN
	cmpld	cr5, rWORD1, rWORD2
 	bne	cr5, L(dLcr5x)
	li	rRTN, 0
	blr
	.align 4
L(dLcr0):
	ld	rWORD8,-8(r1)
	ld	rWORD7,-16(r1)
	li	rRTN, 1
	bgtlr	cr0
	li	rRTN, -1
	blr
	.align 4
L(dLcr1):
	ld	rWORD8,-8(r1)
	ld	rWORD7,-16(r1)
	li	rRTN, 1
	bgtlr	cr1
	li	rRTN, -1
	blr
	.align 4
L(dLcr6):
	ld	rWORD8,-8(r1)
	ld	rWORD7,-16(r1)
	li	rRTN, 1
	bgtlr	cr6
	li	rRTN, -1
	blr
	.align 4
L(dLcr5):
	ld	rWORD8,-8(r1)
	ld	rWORD7,-16(r1)
L(dLcr5x):
	li	rRTN, 1
	bgtlr	cr5
	li	rRTN, -1
	blr
	
	.align 4
L(bytealigned):
	mtctr   rN	/* Power4 wants mtctr 1st in dispatch group */
	beq-	cr6, L(zeroLength)

/* We need to prime this loop.  This loop is swing modulo scheduled
   to avoid pipe delays.  The dependent instruction latencies (load to 
   compare to conditional branch) is 2 to 3 cycles.  In this loop each
   dispatch group ends in a branch and takes 1 cycle.  Effectively
   the first iteration of the loop only serves to load operands and 
   branches based on compares are delayed until the next loop. 

   So we must precondition some registers and condition codes so that
   we don't exit the loop early on the first iteration.  */
   
	lbz	rWORD1, 0(rSTR1)
	lbz	rWORD2, 0(rSTR2)
	bdz-	L(b11)
	cmpld	cr0, rWORD1, rWORD2
	lbz	rWORD3, 1(rSTR1)
	lbz	rWORD4, 1(rSTR2)
	bdz-	L(b12)
	cmpld	cr1, rWORD3, rWORD4
	lbzu	rWORD5, 2(rSTR1)
	lbzu	rWORD6, 2(rSTR2)
	bdz-	L(b13)
	.align 4
L(bLoop):
	lbzu	rWORD1, 1(rSTR1)
	lbzu	rWORD2, 1(rSTR2)
	bne-	cr0, L(bLcr0)

	cmpld	cr6, rWORD5, rWORD6
	bdz-	L(b3i)
	
	lbzu	rWORD3, 1(rSTR1)
	lbzu	rWORD4, 1(rSTR2)
	bne-	cr1, L(bLcr1)

	cmpld	cr0, rWORD1, rWORD2
	bdz-	L(b2i)

	lbzu	rWORD5, 1(rSTR1)
	lbzu	rWORD6, 1(rSTR2)
	bne-	cr6, L(bLcr6)

	cmpld	cr1, rWORD3, rWORD4
	bdnz+	L(bLoop)
	
/* We speculatively loading bytes before we have tested the previous
   bytes.  But we must avoid overrunning the length (in the ctr) to
   prevent these speculative loads from causing a segfault.  In this 
   case the loop will exit early (before the all pending bytes are
   tested.  In this case we must complete the pending operations
   before returning.  */
L(b1i):
	bne-	cr0, L(bLcr0)
	bne-	cr1, L(bLcr1)
	b	L(bx56)
	.align 4
L(b2i):
	bne-	cr6, L(bLcr6)
	bne-	cr0, L(bLcr0)
	b	L(bx34)
	.align 4
L(b3i):
	bne-	cr1, L(bLcr1)
	bne-	cr6, L(bLcr6)
	b	L(bx12)
	.align 4
L(bLcr0):
	li	rRTN, 1
	bgtlr	cr0
	li	rRTN, -1
	blr
L(bLcr1):
	li	rRTN, 1
	bgtlr	cr1
	li	rRTN, -1
	blr
L(bLcr6):
	li	rRTN, 1
	bgtlr	cr6
	li	rRTN, -1
	blr

L(b13):
	bne-	cr0, L(bx12)
	bne-	cr1, L(bx34)
L(bx56):
	sub	rRTN, rWORD5, rWORD6
	blr
	nop
L(b12):
	bne-	cr0, L(bx12)
L(bx34):	
	sub	rRTN, rWORD3, rWORD4
	blr
L(b11):
L(bx12):
	sub	rRTN, rWORD1, rWORD2
	blr
	.align 4 
L(zeroLengthReturn):
	ld	rWORD8,-8(r1)
	ld	rWORD7,-16(r1)
L(zeroLength):
	li	rRTN, 0
	blr

	.align 4
/* At this point we know the strings have different alignment and the
   compare length is at least 8 bytes.  rBITDIF containes the low order
   3 bits of rSTR1 and cr5 contains the result of the logical compare
   of rBITDIF to 0.  If rBITDIF == 0 then rStr1 is double word 
   aligned and can perform the DWunaligned loop.
  
   Otherwise we know that rSTR1 is not aready DW aligned yet.
   So we can force the string addresses to the next lower DW
   boundary and special case this first DW word using shift left to
   ellimiate bits preceeding the first byte.  Since we want to join the
   normal (DWaligned) compare loop, starting at the second double word,
   we need to adjust the length (rN) and special case the loop
   versioning for the first DW. This insures that the loop count is
   correct and the first DW (shifted) is in the expected resister pair.  */
#define rSHL	r29	/* Unaligned shift left count.  */
#define rSHR	r28	/* Unaligned shift right count.  */
#define rB		r27	/* Left rotation temp for rWORD2.  */
#define rD		r26	/* Left rotation temp for rWORD4.  */
#define rF		r25	/* Left rotation temp for rWORD6.  */
#define rH		r24	/* Left rotation temp for rWORD8.  */
#define rA		r0	/* Right rotation temp for rWORD2.  */
#define rC		r12	/* Right rotation temp for rWORD4.  */
#define rE		r0	/* Right rotation temp for rWORD6.  */
#define rG		r12	/* Right rotation temp for rWORD8.  */
L(unaligned):
	std	r29,-24(r1)	
	cfi_offset(r29,-24)
	clrldi	rSHL, rSTR2, 61
	beq-	cr6, L(duzeroLength)
	std	r28,-32(r1)	
	cfi_offset(r28,-32)
	beq	cr5, L(DWunaligned)
	std	r27,-40(r1)	
	cfi_offset(r27,-40)
/* Adjust the logical start of rSTR2 ro compensate for the extra bits
   in the 1st rSTR1 DW.  */
	sub	r27, rSTR2, rBITDIF
/* But do not attempt to address the DW before that DW that contains
   the actual start of rSTR2.  */
	clrrdi	rSTR2, rSTR2, 3
	std	r26,-48(r1)	
	cfi_offset(r26,-48)
/* Compute the leaft/right shift counts for the unalign rSTR2,
   compensating for the logical (DW aligned) start of rSTR1.  */ 
	clrldi	rSHL, r27, 61
	clrrdi	rSTR1, rSTR1, 3	
	std	r25,-56(r1)	
	cfi_offset(r25,-56)
	sldi	rSHL, rSHL, 3
	cmpld	cr5, r27, rSTR2
	add	rN, rN, rBITDIF
	sldi	r11, rBITDIF, 3
	std	r24,-64(r1)	
	cfi_offset(r24,-64)
	subfic	rSHR, rSHL, 64
	srdi	rTMP, rN, 5	/* Divide by 32 */
	andi.	rBITDIF, rN, 24	/* Get the DW remainder */
/* We normally need to load 2 DWs to start the unaligned rSTR2, but in
   this special case those bits may be discarded anyway.  Also we
   must avoid loading a DW where none of the bits are part of rSTR2 as
   this may cross a page boundary and cause a page fault.  */
	li	rWORD8, 0
	blt	cr5, L(dus0)
	ld	rWORD8, 0(rSTR2)
	la	rSTR2, 8(rSTR2)
	sld	rWORD8, rWORD8, rSHL

L(dus0):
	ld	rWORD1, 0(rSTR1)
	ld	rWORD2, 0(rSTR2)
	cmpldi	cr1, rBITDIF, 16
	cmpldi	cr7, rN, 32
	srd	rG, rWORD2, rSHR
	clrldi	rN, rN, 61
	beq	L(duPs4)
	mtctr   rTMP	/* Power4 wants mtctr 1st in dispatch group */
	or	rWORD8, rG, rWORD8
	bgt	cr1, L(duPs3)
	beq	cr1, L(duPs2)

/* Remainder is 8 */
	.align 4
L(dusP1):
	sld	rB, rWORD2, rSHL
	sld	rWORD7, rWORD1, r11
	sld	rWORD8, rWORD8, r11
	bge	cr7, L(duP1e)
/* At this point we exit early with the first double word compare
   complete and remainder of 0 to 7 bytes.  See L(du14) for details on
   how we handle the remaining bytes.  */
	cmpld	cr5, rWORD7, rWORD8
	sldi.	rN, rN, 3
	bne	cr5, L(duLcr5)
	cmpld	cr7, rN, rSHR
	beq	L(duZeroReturn)
	li	rA, 0
	ble	cr7, L(dutrim)
	ld	rWORD2, 8(rSTR2)
	srd	rA, rWORD2, rSHR
	b	L(dutrim)
/* Remainder is 16 */
	.align 4
L(duPs2):
	sld	rH, rWORD2, rSHL
	sld	rWORD5, rWORD1, r11
	sld	rWORD6, rWORD8, r11
	b	L(duP2e)
/* Remainder is 24 */
	.align 4
L(duPs3):
	sld	rF, rWORD2, rSHL
	sld	rWORD3, rWORD1, r11
	sld	rWORD4, rWORD8, r11
	b	L(duP3e)
/* Count is a multiple of 32, remainder is 0 */
	.align 4
L(duPs4):
	mtctr   rTMP	/* Power4 wants mtctr 1st in dispatch group */
	or	rWORD8, rG, rWORD8
	sld	rD, rWORD2, rSHL
	sld	rWORD1, rWORD1, r11
	sld	rWORD2, rWORD8, r11
	b	L(duP4e)

/* At this point we know rSTR1 is double word aligned and the
   compare length is at least 8 bytes.  */
	.align 4
L(DWunaligned):
	std	r27,-40(r1)	
	cfi_offset(r27,-40)
	clrrdi	rSTR2, rSTR2, 3
	std	r26,-48(r1)	
	cfi_offset(r26,-48)
	srdi	rTMP, rN, 5	/* Divide by 32 */
	std	r25,-56(r1)	
	cfi_offset(r25,-56)
	andi.	rBITDIF, rN, 24	/* Get the DW remainder */
	std	r24,-64(r1)	
	cfi_offset(r24,-64)
	sldi	rSHL, rSHL, 3
	ld	rWORD6, 0(rSTR2)
	ldu	rWORD8, 8(rSTR2)
	cmpldi	cr1, rBITDIF, 16
	cmpldi	cr7, rN, 32
	clrldi	rN, rN, 61
	subfic	rSHR, rSHL, 64
	sld	rH, rWORD6, rSHL
	beq	L(duP4)
	mtctr   rTMP	/* Power4 wants mtctr 1st in dispatch group */
	bgt	cr1, L(duP3)
	beq	cr1, L(duP2)
		
/* Remainder is 8 */
	.align 4
L(duP1):
	srd	rG, rWORD8, rSHR
	ld	rWORD7, 0(rSTR1)
	sld	rB, rWORD8, rSHL
	or	rWORD8, rG, rH
	blt	cr7, L(duP1x)
L(duP1e):
	ld	rWORD1, 8(rSTR1)
	ld	rWORD2, 8(rSTR2)
	cmpld	cr5, rWORD7, rWORD8
	srd	rA, rWORD2, rSHR
	sld	rD, rWORD2, rSHL
	or	rWORD2, rA, rB
	ld	rWORD3, 16(rSTR1)
	ld	rWORD4, 16(rSTR2)
	cmpld	cr0, rWORD1, rWORD2
	srd	rC, rWORD4, rSHR
	sld	rF, rWORD4, rSHL
	bne	cr5, L(duLcr5)
	or	rWORD4, rC, rD
	ld	rWORD5, 24(rSTR1)
	ld	rWORD6, 24(rSTR2)
	cmpld	cr1, rWORD3, rWORD4
	srd	rE, rWORD6, rSHR
	sld	rH, rWORD6, rSHL
	bne	cr0, L(duLcr0)
	or	rWORD6, rE, rF
	cmpld	cr6, rWORD5, rWORD6
	b	L(duLoop3)	
	.align 4
/* At this point we exit early with the first double word compare
   complete and remainder of 0 to 7 bytes.  See L(du14) for details on
   how we handle the remaining bytes.  */
L(duP1x):
	cmpld	cr5, rWORD7, rWORD8
	sldi.	rN, rN, 3
	bne	cr5, L(duLcr5)
	cmpld	cr7, rN, rSHR
	beq	L(duZeroReturn)
	li	rA, 0
	ble	cr7, L(dutrim)
	ld	rWORD2, 8(rSTR2)
	srd	rA, rWORD2, rSHR
	b	L(dutrim)
/* Remainder is 16 */
	.align 4
L(duP2):
	srd	rE, rWORD8, rSHR
	ld	rWORD5, 0(rSTR1)
	or	rWORD6, rE, rH
	sld	rH, rWORD8, rSHL
L(duP2e):
	ld	rWORD7, 8(rSTR1)
	ld	rWORD8, 8(rSTR2)
	cmpld	cr6, rWORD5, rWORD6
	srd	rG, rWORD8, rSHR
	sld	rB, rWORD8, rSHL
	or	rWORD8, rG, rH
	blt	cr7, L(duP2x)
	ld	rWORD1, 16(rSTR1)
	ld	rWORD2, 16(rSTR2)
	cmpld	cr5, rWORD7, rWORD8
	bne	cr6, L(duLcr6)
	srd	rA, rWORD2, rSHR
	sld	rD, rWORD2, rSHL
	or	rWORD2, rA, rB
	ld	rWORD3, 24(rSTR1)
	ld	rWORD4, 24(rSTR2)
	cmpld	cr0, rWORD1, rWORD2
	bne	cr5, L(duLcr5)
	srd	rC, rWORD4, rSHR
	sld	rF, rWORD4, rSHL
	or	rWORD4, rC, rD
	addi	rSTR1, rSTR1, 8
	addi	rSTR2, rSTR2, 8
	cmpld	cr1, rWORD3, rWORD4
	b	L(duLoop2)
	.align 4
L(duP2x):
	cmpld	cr5, rWORD7, rWORD8
	addi	rSTR1, rSTR1, 8
	addi	rSTR2, rSTR2, 8
	bne	cr6, L(duLcr6)
	sldi.	rN, rN, 3
	bne	cr5, L(duLcr5)
	cmpld	cr7, rN, rSHR
	beq	L(duZeroReturn)
	li	rA, 0
	ble	cr7, L(dutrim)
	ld	rWORD2, 8(rSTR2)
	srd	rA, rWORD2, rSHR
	b	L(dutrim)
		
/* Remainder is 24 */
	.align 4
L(duP3):
	srd	rC, rWORD8, rSHR
	ld	rWORD3, 0(rSTR1)
	sld	rF, rWORD8, rSHL
	or	rWORD4, rC, rH
L(duP3e):
	ld	rWORD5, 8(rSTR1)
	ld	rWORD6, 8(rSTR2)
	cmpld	cr1, rWORD3, rWORD4
	srd	rE, rWORD6, rSHR
	sld	rH, rWORD6, rSHL
	or	rWORD6, rE, rF
	ld	rWORD7, 16(rSTR1)
	ld	rWORD8, 16(rSTR2)
	cmpld	cr6, rWORD5, rWORD6
	bne	cr1, L(duLcr1)
	srd	rG, rWORD8, rSHR
	sld	rB, rWORD8, rSHL
	or	rWORD8, rG, rH
	blt	cr7, L(duP3x)
	ld	rWORD1, 24(rSTR1)
	ld	rWORD2, 24(rSTR2)
	cmpld	cr5, rWORD7, rWORD8
	bne	cr6, L(duLcr6)
	srd	rA, rWORD2, rSHR
	sld	rD, rWORD2, rSHL
	or	rWORD2, rA, rB
	addi	rSTR1, rSTR1, 16
	addi	rSTR2, rSTR2, 16
	cmpld	cr0, rWORD1, rWORD2
	b	L(duLoop1)
	.align 4
L(duP3x):
	addi	rSTR1, rSTR1, 16
	addi	rSTR2, rSTR2, 16
	bne	cr1, L(duLcr1)
	cmpld	cr5, rWORD7, rWORD8
	bne	cr6, L(duLcr6)
	sldi.	rN, rN, 3
	bne	cr5, L(duLcr5)
	cmpld	cr7, rN, rSHR
	beq	L(duZeroReturn)
	li	rA, 0
	ble	cr7, L(dutrim)
	ld	rWORD2, 8(rSTR2)
	srd	rA, rWORD2, rSHR
	b	L(dutrim)
	
/* Count is a multiple of 32, remainder is 0 */
	.align 4
L(duP4):
	mtctr   rTMP	/* Power4 wants mtctr 1st in dispatch group */
	srd	rA, rWORD8, rSHR
	ld	rWORD1, 0(rSTR1)
	sld	rD, rWORD8, rSHL
	or	rWORD2, rA, rH
L(duP4e):
	ld	rWORD3, 8(rSTR1)
	ld	rWORD4, 8(rSTR2)
	cmpld	cr0, rWORD1, rWORD2
	srd	rC, rWORD4, rSHR
	sld	rF, rWORD4, rSHL
	or	rWORD4, rC, rD
	ld	rWORD5, 16(rSTR1)
	ld	rWORD6, 16(rSTR2)
	cmpld	cr1, rWORD3, rWORD4
	bne	cr0, L(duLcr0)
	srd	rE, rWORD6, rSHR
	sld	rH, rWORD6, rSHL
	or	rWORD6, rE, rF
	ldu	rWORD7, 24(rSTR1)
	ldu	rWORD8, 24(rSTR2)
	cmpld	cr6, rWORD5, rWORD6
	bne	cr1, L(duLcr1)
	srd	rG, rWORD8, rSHR
	sld	rB, rWORD8, rSHL
	or	rWORD8, rG, rH
	cmpld	cr5, rWORD7, rWORD8
	bdz-	L(du24)		/* Adjust CTR as we start with +4 */
/* This is the primary loop */
	.align 4
L(duLoop):
	ld	rWORD1, 8(rSTR1)
	ld	rWORD2, 8(rSTR2)
	cmpld	cr1, rWORD3, rWORD4
	bne	cr6, L(duLcr6)
	srd	rA, rWORD2, rSHR
	sld	rD, rWORD2, rSHL
	or	rWORD2, rA, rB
L(duLoop1):
	ld	rWORD3, 16(rSTR1)
	ld	rWORD4, 16(rSTR2)
	cmpld	cr6, rWORD5, rWORD6
	bne	cr5, L(duLcr5)
	srd	rC, rWORD4, rSHR
	sld	rF, rWORD4, rSHL
	or	rWORD4, rC, rD
L(duLoop2):
	ld	rWORD5, 24(rSTR1)
	ld	rWORD6, 24(rSTR2)
	cmpld	cr5, rWORD7, rWORD8
	bne	cr0, L(duLcr0)
	srd	rE, rWORD6, rSHR
	sld	rH, rWORD6, rSHL
	or	rWORD6, rE, rF
L(duLoop3):
	ldu	rWORD7, 32(rSTR1)
	ldu	rWORD8, 32(rSTR2)
	cmpld	cr0, rWORD1, rWORD2
	bne-	cr1, L(duLcr1)
	srd	rG, rWORD8, rSHR
	sld	rB, rWORD8, rSHL
	or	rWORD8, rG, rH
	bdnz+	L(duLoop)	
	
L(duL4):
	bne	cr1, L(duLcr1)
	cmpld	cr1, rWORD3, rWORD4
	bne	cr6, L(duLcr6)
	cmpld	cr6, rWORD5, rWORD6
	bne	cr5, L(duLcr5)
	cmpld	cr5, rWORD7, rWORD8
L(du44):
	bne	cr0, L(duLcr0)
L(du34):
	bne	cr1, L(duLcr1)
L(du24):
	bne	cr6, L(duLcr6)
L(du14):
	sldi.	rN, rN, 3
	bne	cr5, L(duLcr5)
/* At this point we have a remainder of 1 to 7 bytes to compare.  We use
   shift right double to elliminate bits beyond the compare length. 
   This allows the use of double word subtract to compute the final
   result.

   However it may not be safe to load rWORD2 which may be beyond the 
   string length. So we compare the bit length of the remainder to
   the right shift count (rSHR). If the bit count is less than or equal
   we do not need to load rWORD2 (all significant bits are already in
   rB).  */
	cmpld	cr7, rN, rSHR
	beq	L(duZeroReturn)
	li	rA, 0
	ble	cr7, L(dutrim)
	ld	rWORD2, 8(rSTR2)
	srd	rA, rWORD2, rSHR
	.align 4
L(dutrim):
	ld	rWORD1, 8(rSTR1)
	ld	rWORD8,-8(r1)
	subfic	rN, rN, 64	/* Shift count is 64 - (rN * 8).  */ 
	or	rWORD2, rA, rB
	ld	rWORD7,-16(r1)	
	ld	r29,-24(r1)
	srd	rWORD1, rWORD1, rN
	srd	rWORD2, rWORD2, rN
	ld	r28,-32(r1)	
	ld	r27,-40(r1)
	li	rRTN, 0
	cmpld	cr0, rWORD1, rWORD2	
	ld	r26,-48(r1)
	ld	r25,-56(r1)
 	beq	cr0, L(dureturn24)
	li	rRTN, 1
	ld	r24,-64(r1)
	bgtlr	cr0
	li	rRTN, -1
	blr
	.align 4
L(duLcr0):
	ld	rWORD8,-8(r1)
	ld	rWORD7,-16(r1)
	li	rRTN, 1
	bgt	cr0, L(dureturn29)	
	ld	r29,-24(r1)
	ld	r28,-32(r1)
	li	rRTN, -1
	b	L(dureturn27)
	.align 4
L(duLcr1):
	ld	rWORD8,-8(r1)
	ld	rWORD7,-16(r1)
	li	rRTN, 1
	bgt	cr1, L(dureturn29)	
	ld	r29,-24(r1)
	ld	r28,-32(r1)
	li	rRTN, -1
	b	L(dureturn27)
	.align 4
L(duLcr6):
	ld	rWORD8,-8(r1)
	ld	rWORD7,-16(r1)
	li	rRTN, 1
	bgt	cr6, L(dureturn29)	
	ld	r29,-24(r1)
	ld	r28,-32(r1)
	li	rRTN, -1
	b	L(dureturn27)
	.align 4
L(duLcr5):
	ld	rWORD8,-8(r1)
	ld	rWORD7,-16(r1)
	li	rRTN, 1
	bgt	cr5, L(dureturn29)	
	ld	r29,-24(r1)
	ld	r28,-32(r1)
	li	rRTN, -1
	b	L(dureturn27)
	.align	3
L(duZeroReturn):
	li	rRTN,0
	.align	4
L(dureturn):
	ld	rWORD8,-8(r1)
	ld	rWORD7,-16(r1)
L(dureturn29):	
	ld	r29,-24(r1)
	ld	r28,-32(r1)
L(dureturn27):	
	ld	r27,-40(r1)
L(dureturn26):	
	ld	r26,-48(r1)
L(dureturn25):	
	ld	r25,-56(r1)
L(dureturn24):
	ld	r24,-64(r1)
	blr
L(duzeroLength):
	li	rRTN,0
	blr

END (BP_SYM (memcmp))
libc_hidden_builtin_def (memcmp)
weak_alias (memcmp, bcmp)