1/*
2 * This program is free software; you can redistribute it and/or modify
3 * it under the terms of the GNU General Public License as published by
4 * the Free Software Foundation; either version 2 of the License, or
5 * (at your option) any later version.
6 *
7 * This program is distributed in the hope that it will be useful,
8 * but WITHOUT ANY WARRANTY; without even the implied warranty of
9 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
10 * GNU General Public License for more details.
11 *
12 * You should have received a copy of the GNU General Public License
13 * along with this program; if not, write to the Free Software
14 * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
15 *
16 * Copyright (C) IBM Corporation, 2011
17 *
18 * Author: Anton Blanchard <anton@au.ibm.com>
19 */
20#include <asm/ppc_asm.h>
21
22#ifdef __BIG_ENDIAN__
23#define LVS(VRT,RA,RB)		lvsl	VRT,RA,RB
24#define VPERM(VRT,VRA,VRB,VRC)	vperm	VRT,VRA,VRB,VRC
25#else
26#define LVS(VRT,RA,RB)		lvsr	VRT,RA,RB
27#define VPERM(VRT,VRA,VRB,VRC)	vperm	VRT,VRB,VRA,VRC
28#endif
29
30	.macro err1
31100:
32	EX_TABLE(100b,.Ldo_err1)
33	.endm
34
35	.macro err2
36200:
37	EX_TABLE(200b,.Ldo_err2)
38	.endm
39
40#ifdef CONFIG_ALTIVEC
41	.macro err3
42300:
43	EX_TABLE(300b,.Ldo_err3)
44	.endm
45
46	.macro err4
47400:
48	EX_TABLE(400b,.Ldo_err4)
49	.endm
50
51
52.Ldo_err4:
53	ld	r16,STK_REG(R16)(r1)
54	ld	r15,STK_REG(R15)(r1)
55	ld	r14,STK_REG(R14)(r1)
56.Ldo_err3:
57	bl	exit_vmx_usercopy
58	ld	r0,STACKFRAMESIZE+16(r1)
59	mtlr	r0
60	b	.Lexit
61#endif /* CONFIG_ALTIVEC */
62
63.Ldo_err2:
64	ld	r22,STK_REG(R22)(r1)
65	ld	r21,STK_REG(R21)(r1)
66	ld	r20,STK_REG(R20)(r1)
67	ld	r19,STK_REG(R19)(r1)
68	ld	r18,STK_REG(R18)(r1)
69	ld	r17,STK_REG(R17)(r1)
70	ld	r16,STK_REG(R16)(r1)
71	ld	r15,STK_REG(R15)(r1)
72	ld	r14,STK_REG(R14)(r1)
73.Lexit:
74	addi	r1,r1,STACKFRAMESIZE
75.Ldo_err1:
76	ld	r3,-STACKFRAMESIZE+STK_REG(R31)(r1)
77	ld	r4,-STACKFRAMESIZE+STK_REG(R30)(r1)
78	ld	r5,-STACKFRAMESIZE+STK_REG(R29)(r1)
79	b	__copy_tofrom_user_base
80
81
82_GLOBAL(__copy_tofrom_user_power7)
83#ifdef CONFIG_ALTIVEC
84	cmpldi	r5,16
85	cmpldi	cr1,r5,3328
86
87	std	r3,-STACKFRAMESIZE+STK_REG(R31)(r1)
88	std	r4,-STACKFRAMESIZE+STK_REG(R30)(r1)
89	std	r5,-STACKFRAMESIZE+STK_REG(R29)(r1)
90
91	blt	.Lshort_copy
92	bge	cr1,.Lvmx_copy
93#else
94	cmpldi	r5,16
95
96	std	r3,-STACKFRAMESIZE+STK_REG(R31)(r1)
97	std	r4,-STACKFRAMESIZE+STK_REG(R30)(r1)
98	std	r5,-STACKFRAMESIZE+STK_REG(R29)(r1)
99
100	blt	.Lshort_copy
101#endif
102
103.Lnonvmx_copy:
104	/* Get the source 8B aligned */
105	neg	r6,r4
106	mtocrf	0x01,r6
107	clrldi	r6,r6,(64-3)
108
109	bf	cr7*4+3,1f
110err1;	lbz	r0,0(r4)
111	addi	r4,r4,1
112err1;	stb	r0,0(r3)
113	addi	r3,r3,1
114
1151:	bf	cr7*4+2,2f
116err1;	lhz	r0,0(r4)
117	addi	r4,r4,2
118err1;	sth	r0,0(r3)
119	addi	r3,r3,2
120
1212:	bf	cr7*4+1,3f
122err1;	lwz	r0,0(r4)
123	addi	r4,r4,4
124err1;	stw	r0,0(r3)
125	addi	r3,r3,4
126
1273:	sub	r5,r5,r6
128	cmpldi	r5,128
129	blt	5f
130
131	mflr	r0
132	stdu	r1,-STACKFRAMESIZE(r1)
133	std	r14,STK_REG(R14)(r1)
134	std	r15,STK_REG(R15)(r1)
135	std	r16,STK_REG(R16)(r1)
136	std	r17,STK_REG(R17)(r1)
137	std	r18,STK_REG(R18)(r1)
138	std	r19,STK_REG(R19)(r1)
139	std	r20,STK_REG(R20)(r1)
140	std	r21,STK_REG(R21)(r1)
141	std	r22,STK_REG(R22)(r1)
142	std	r0,STACKFRAMESIZE+16(r1)
143
144	srdi	r6,r5,7
145	mtctr	r6
146
147	/* Now do cacheline (128B) sized loads and stores. */
148	.align	5
1494:
150err2;	ld	r0,0(r4)
151err2;	ld	r6,8(r4)
152err2;	ld	r7,16(r4)
153err2;	ld	r8,24(r4)
154err2;	ld	r9,32(r4)
155err2;	ld	r10,40(r4)
156err2;	ld	r11,48(r4)
157err2;	ld	r12,56(r4)
158err2;	ld	r14,64(r4)
159err2;	ld	r15,72(r4)
160err2;	ld	r16,80(r4)
161err2;	ld	r17,88(r4)
162err2;	ld	r18,96(r4)
163err2;	ld	r19,104(r4)
164err2;	ld	r20,112(r4)
165err2;	ld	r21,120(r4)
166	addi	r4,r4,128
167err2;	std	r0,0(r3)
168err2;	std	r6,8(r3)
169err2;	std	r7,16(r3)
170err2;	std	r8,24(r3)
171err2;	std	r9,32(r3)
172err2;	std	r10,40(r3)
173err2;	std	r11,48(r3)
174err2;	std	r12,56(r3)
175err2;	std	r14,64(r3)
176err2;	std	r15,72(r3)
177err2;	std	r16,80(r3)
178err2;	std	r17,88(r3)
179err2;	std	r18,96(r3)
180err2;	std	r19,104(r3)
181err2;	std	r20,112(r3)
182err2;	std	r21,120(r3)
183	addi	r3,r3,128
184	bdnz	4b
185
186	clrldi	r5,r5,(64-7)
187
188	ld	r14,STK_REG(R14)(r1)
189	ld	r15,STK_REG(R15)(r1)
190	ld	r16,STK_REG(R16)(r1)
191	ld	r17,STK_REG(R17)(r1)
192	ld	r18,STK_REG(R18)(r1)
193	ld	r19,STK_REG(R19)(r1)
194	ld	r20,STK_REG(R20)(r1)
195	ld	r21,STK_REG(R21)(r1)
196	ld	r22,STK_REG(R22)(r1)
197	addi	r1,r1,STACKFRAMESIZE
198
199	/* Up to 127B to go */
2005:	srdi	r6,r5,4
201	mtocrf	0x01,r6
202
2036:	bf	cr7*4+1,7f
204err1;	ld	r0,0(r4)
205err1;	ld	r6,8(r4)
206err1;	ld	r7,16(r4)
207err1;	ld	r8,24(r4)
208err1;	ld	r9,32(r4)
209err1;	ld	r10,40(r4)
210err1;	ld	r11,48(r4)
211err1;	ld	r12,56(r4)
212	addi	r4,r4,64
213err1;	std	r0,0(r3)
214err1;	std	r6,8(r3)
215err1;	std	r7,16(r3)
216err1;	std	r8,24(r3)
217err1;	std	r9,32(r3)
218err1;	std	r10,40(r3)
219err1;	std	r11,48(r3)
220err1;	std	r12,56(r3)
221	addi	r3,r3,64
222
223	/* Up to 63B to go */
2247:	bf	cr7*4+2,8f
225err1;	ld	r0,0(r4)
226err1;	ld	r6,8(r4)
227err1;	ld	r7,16(r4)
228err1;	ld	r8,24(r4)
229	addi	r4,r4,32
230err1;	std	r0,0(r3)
231err1;	std	r6,8(r3)
232err1;	std	r7,16(r3)
233err1;	std	r8,24(r3)
234	addi	r3,r3,32
235
236	/* Up to 31B to go */
2378:	bf	cr7*4+3,9f
238err1;	ld	r0,0(r4)
239err1;	ld	r6,8(r4)
240	addi	r4,r4,16
241err1;	std	r0,0(r3)
242err1;	std	r6,8(r3)
243	addi	r3,r3,16
244
2459:	clrldi	r5,r5,(64-4)
246
247	/* Up to 15B to go */
248.Lshort_copy:
249	mtocrf	0x01,r5
250	bf	cr7*4+0,12f
251err1;	lwz	r0,0(r4)	/* Less chance of a reject with word ops */
252err1;	lwz	r6,4(r4)
253	addi	r4,r4,8
254err1;	stw	r0,0(r3)
255err1;	stw	r6,4(r3)
256	addi	r3,r3,8
257
25812:	bf	cr7*4+1,13f
259err1;	lwz	r0,0(r4)
260	addi	r4,r4,4
261err1;	stw	r0,0(r3)
262	addi	r3,r3,4
263
26413:	bf	cr7*4+2,14f
265err1;	lhz	r0,0(r4)
266	addi	r4,r4,2
267err1;	sth	r0,0(r3)
268	addi	r3,r3,2
269
27014:	bf	cr7*4+3,15f
271err1;	lbz	r0,0(r4)
272err1;	stb	r0,0(r3)
273
27415:	li	r3,0
275	blr
276
277.Lunwind_stack_nonvmx_copy:
278	addi	r1,r1,STACKFRAMESIZE
279	b	.Lnonvmx_copy
280
281#ifdef CONFIG_ALTIVEC
282.Lvmx_copy:
283	mflr	r0
284	std	r0,16(r1)
285	stdu	r1,-STACKFRAMESIZE(r1)
286	bl	enter_vmx_usercopy
287	cmpwi	cr1,r3,0
288	ld	r0,STACKFRAMESIZE+16(r1)
289	ld	r3,STK_REG(R31)(r1)
290	ld	r4,STK_REG(R30)(r1)
291	ld	r5,STK_REG(R29)(r1)
292	mtlr	r0
293
294	/*
295	 * We prefetch both the source and destination using enhanced touch
296	 * instructions. We use a stream ID of 0 for the load side and
297	 * 1 for the store side.
298	 */
299	clrrdi	r6,r4,7
300	clrrdi	r9,r3,7
301	ori	r9,r9,1		/* stream=1 */
302
303	srdi	r7,r5,7		/* length in cachelines, capped at 0x3FF */
304	cmpldi	r7,0x3FF
305	ble	1f
306	li	r7,0x3FF
3071:	lis	r0,0x0E00	/* depth=7 */
308	sldi	r7,r7,7
309	or	r7,r7,r0
310	ori	r10,r7,1	/* stream=1 */
311
312	lis	r8,0x8000	/* GO=1 */
313	clrldi	r8,r8,32
314
315.machine push
316.machine "power4"
317	/* setup read stream 0 */
318	dcbt	0,r6,0b01000   /* addr from */
319	dcbt	0,r7,0b01010   /* length and depth from */
320	/* setup write stream 1 */
321	dcbtst	0,r9,0b01000   /* addr to */
322	dcbtst	0,r10,0b01010  /* length and depth to */
323	eieio
324	dcbt	0,r8,0b01010	/* all streams GO */
325.machine pop
326
327	beq	cr1,.Lunwind_stack_nonvmx_copy
328
329	/*
330	 * If source and destination are not relatively aligned we use a
331	 * slower permute loop.
332	 */
333	xor	r6,r4,r3
334	rldicl.	r6,r6,0,(64-4)
335	bne	.Lvmx_unaligned_copy
336
337	/* Get the destination 16B aligned */
338	neg	r6,r3
339	mtocrf	0x01,r6
340	clrldi	r6,r6,(64-4)
341
342	bf	cr7*4+3,1f
343err3;	lbz	r0,0(r4)
344	addi	r4,r4,1
345err3;	stb	r0,0(r3)
346	addi	r3,r3,1
347
3481:	bf	cr7*4+2,2f
349err3;	lhz	r0,0(r4)
350	addi	r4,r4,2
351err3;	sth	r0,0(r3)
352	addi	r3,r3,2
353
3542:	bf	cr7*4+1,3f
355err3;	lwz	r0,0(r4)
356	addi	r4,r4,4
357err3;	stw	r0,0(r3)
358	addi	r3,r3,4
359
3603:	bf	cr7*4+0,4f
361err3;	ld	r0,0(r4)
362	addi	r4,r4,8
363err3;	std	r0,0(r3)
364	addi	r3,r3,8
365
3664:	sub	r5,r5,r6
367
368	/* Get the desination 128B aligned */
369	neg	r6,r3
370	srdi	r7,r6,4
371	mtocrf	0x01,r7
372	clrldi	r6,r6,(64-7)
373
374	li	r9,16
375	li	r10,32
376	li	r11,48
377
378	bf	cr7*4+3,5f
379err3;	lvx	v1,0,r4
380	addi	r4,r4,16
381err3;	stvx	v1,0,r3
382	addi	r3,r3,16
383
3845:	bf	cr7*4+2,6f
385err3;	lvx	v1,0,r4
386err3;	lvx	v0,r4,r9
387	addi	r4,r4,32
388err3;	stvx	v1,0,r3
389err3;	stvx	v0,r3,r9
390	addi	r3,r3,32
391
3926:	bf	cr7*4+1,7f
393err3;	lvx	v3,0,r4
394err3;	lvx	v2,r4,r9
395err3;	lvx	v1,r4,r10
396err3;	lvx	v0,r4,r11
397	addi	r4,r4,64
398err3;	stvx	v3,0,r3
399err3;	stvx	v2,r3,r9
400err3;	stvx	v1,r3,r10
401err3;	stvx	v0,r3,r11
402	addi	r3,r3,64
403
4047:	sub	r5,r5,r6
405	srdi	r6,r5,7
406
407	std	r14,STK_REG(R14)(r1)
408	std	r15,STK_REG(R15)(r1)
409	std	r16,STK_REG(R16)(r1)
410
411	li	r12,64
412	li	r14,80
413	li	r15,96
414	li	r16,112
415
416	mtctr	r6
417
418	/*
419	 * Now do cacheline sized loads and stores. By this stage the
420	 * cacheline stores are also cacheline aligned.
421	 */
422	.align	5
4238:
424err4;	lvx	v7,0,r4
425err4;	lvx	v6,r4,r9
426err4;	lvx	v5,r4,r10
427err4;	lvx	v4,r4,r11
428err4;	lvx	v3,r4,r12
429err4;	lvx	v2,r4,r14
430err4;	lvx	v1,r4,r15
431err4;	lvx	v0,r4,r16
432	addi	r4,r4,128
433err4;	stvx	v7,0,r3
434err4;	stvx	v6,r3,r9
435err4;	stvx	v5,r3,r10
436err4;	stvx	v4,r3,r11
437err4;	stvx	v3,r3,r12
438err4;	stvx	v2,r3,r14
439err4;	stvx	v1,r3,r15
440err4;	stvx	v0,r3,r16
441	addi	r3,r3,128
442	bdnz	8b
443
444	ld	r14,STK_REG(R14)(r1)
445	ld	r15,STK_REG(R15)(r1)
446	ld	r16,STK_REG(R16)(r1)
447
448	/* Up to 127B to go */
449	clrldi	r5,r5,(64-7)
450	srdi	r6,r5,4
451	mtocrf	0x01,r6
452
453	bf	cr7*4+1,9f
454err3;	lvx	v3,0,r4
455err3;	lvx	v2,r4,r9
456err3;	lvx	v1,r4,r10
457err3;	lvx	v0,r4,r11
458	addi	r4,r4,64
459err3;	stvx	v3,0,r3
460err3;	stvx	v2,r3,r9
461err3;	stvx	v1,r3,r10
462err3;	stvx	v0,r3,r11
463	addi	r3,r3,64
464
4659:	bf	cr7*4+2,10f
466err3;	lvx	v1,0,r4
467err3;	lvx	v0,r4,r9
468	addi	r4,r4,32
469err3;	stvx	v1,0,r3
470err3;	stvx	v0,r3,r9
471	addi	r3,r3,32
472
47310:	bf	cr7*4+3,11f
474err3;	lvx	v1,0,r4
475	addi	r4,r4,16
476err3;	stvx	v1,0,r3
477	addi	r3,r3,16
478
479	/* Up to 15B to go */
48011:	clrldi	r5,r5,(64-4)
481	mtocrf	0x01,r5
482	bf	cr7*4+0,12f
483err3;	ld	r0,0(r4)
484	addi	r4,r4,8
485err3;	std	r0,0(r3)
486	addi	r3,r3,8
487
48812:	bf	cr7*4+1,13f
489err3;	lwz	r0,0(r4)
490	addi	r4,r4,4
491err3;	stw	r0,0(r3)
492	addi	r3,r3,4
493
49413:	bf	cr7*4+2,14f
495err3;	lhz	r0,0(r4)
496	addi	r4,r4,2
497err3;	sth	r0,0(r3)
498	addi	r3,r3,2
499
50014:	bf	cr7*4+3,15f
501err3;	lbz	r0,0(r4)
502err3;	stb	r0,0(r3)
503
50415:	addi	r1,r1,STACKFRAMESIZE
505	b	exit_vmx_usercopy	/* tail call optimise */
506
507.Lvmx_unaligned_copy:
508	/* Get the destination 16B aligned */
509	neg	r6,r3
510	mtocrf	0x01,r6
511	clrldi	r6,r6,(64-4)
512
513	bf	cr7*4+3,1f
514err3;	lbz	r0,0(r4)
515	addi	r4,r4,1
516err3;	stb	r0,0(r3)
517	addi	r3,r3,1
518
5191:	bf	cr7*4+2,2f
520err3;	lhz	r0,0(r4)
521	addi	r4,r4,2
522err3;	sth	r0,0(r3)
523	addi	r3,r3,2
524
5252:	bf	cr7*4+1,3f
526err3;	lwz	r0,0(r4)
527	addi	r4,r4,4
528err3;	stw	r0,0(r3)
529	addi	r3,r3,4
530
5313:	bf	cr7*4+0,4f
532err3;	lwz	r0,0(r4)	/* Less chance of a reject with word ops */
533err3;	lwz	r7,4(r4)
534	addi	r4,r4,8
535err3;	stw	r0,0(r3)
536err3;	stw	r7,4(r3)
537	addi	r3,r3,8
538
5394:	sub	r5,r5,r6
540
541	/* Get the desination 128B aligned */
542	neg	r6,r3
543	srdi	r7,r6,4
544	mtocrf	0x01,r7
545	clrldi	r6,r6,(64-7)
546
547	li	r9,16
548	li	r10,32
549	li	r11,48
550
551	LVS(v16,0,r4)		/* Setup permute control vector */
552err3;	lvx	v0,0,r4
553	addi	r4,r4,16
554
555	bf	cr7*4+3,5f
556err3;	lvx	v1,0,r4
557	VPERM(v8,v0,v1,v16)
558	addi	r4,r4,16
559err3;	stvx	v8,0,r3
560	addi	r3,r3,16
561	vor	v0,v1,v1
562
5635:	bf	cr7*4+2,6f
564err3;	lvx	v1,0,r4
565	VPERM(v8,v0,v1,v16)
566err3;	lvx	v0,r4,r9
567	VPERM(v9,v1,v0,v16)
568	addi	r4,r4,32
569err3;	stvx	v8,0,r3
570err3;	stvx	v9,r3,r9
571	addi	r3,r3,32
572
5736:	bf	cr7*4+1,7f
574err3;	lvx	v3,0,r4
575	VPERM(v8,v0,v3,v16)
576err3;	lvx	v2,r4,r9
577	VPERM(v9,v3,v2,v16)
578err3;	lvx	v1,r4,r10
579	VPERM(v10,v2,v1,v16)
580err3;	lvx	v0,r4,r11
581	VPERM(v11,v1,v0,v16)
582	addi	r4,r4,64
583err3;	stvx	v8,0,r3
584err3;	stvx	v9,r3,r9
585err3;	stvx	v10,r3,r10
586err3;	stvx	v11,r3,r11
587	addi	r3,r3,64
588
5897:	sub	r5,r5,r6
590	srdi	r6,r5,7
591
592	std	r14,STK_REG(R14)(r1)
593	std	r15,STK_REG(R15)(r1)
594	std	r16,STK_REG(R16)(r1)
595
596	li	r12,64
597	li	r14,80
598	li	r15,96
599	li	r16,112
600
601	mtctr	r6
602
603	/*
604	 * Now do cacheline sized loads and stores. By this stage the
605	 * cacheline stores are also cacheline aligned.
606	 */
607	.align	5
6088:
609err4;	lvx	v7,0,r4
610	VPERM(v8,v0,v7,v16)
611err4;	lvx	v6,r4,r9
612	VPERM(v9,v7,v6,v16)
613err4;	lvx	v5,r4,r10
614	VPERM(v10,v6,v5,v16)
615err4;	lvx	v4,r4,r11
616	VPERM(v11,v5,v4,v16)
617err4;	lvx	v3,r4,r12
618	VPERM(v12,v4,v3,v16)
619err4;	lvx	v2,r4,r14
620	VPERM(v13,v3,v2,v16)
621err4;	lvx	v1,r4,r15
622	VPERM(v14,v2,v1,v16)
623err4;	lvx	v0,r4,r16
624	VPERM(v15,v1,v0,v16)
625	addi	r4,r4,128
626err4;	stvx	v8,0,r3
627err4;	stvx	v9,r3,r9
628err4;	stvx	v10,r3,r10
629err4;	stvx	v11,r3,r11
630err4;	stvx	v12,r3,r12
631err4;	stvx	v13,r3,r14
632err4;	stvx	v14,r3,r15
633err4;	stvx	v15,r3,r16
634	addi	r3,r3,128
635	bdnz	8b
636
637	ld	r14,STK_REG(R14)(r1)
638	ld	r15,STK_REG(R15)(r1)
639	ld	r16,STK_REG(R16)(r1)
640
641	/* Up to 127B to go */
642	clrldi	r5,r5,(64-7)
643	srdi	r6,r5,4
644	mtocrf	0x01,r6
645
646	bf	cr7*4+1,9f
647err3;	lvx	v3,0,r4
648	VPERM(v8,v0,v3,v16)
649err3;	lvx	v2,r4,r9
650	VPERM(v9,v3,v2,v16)
651err3;	lvx	v1,r4,r10
652	VPERM(v10,v2,v1,v16)
653err3;	lvx	v0,r4,r11
654	VPERM(v11,v1,v0,v16)
655	addi	r4,r4,64
656err3;	stvx	v8,0,r3
657err3;	stvx	v9,r3,r9
658err3;	stvx	v10,r3,r10
659err3;	stvx	v11,r3,r11
660	addi	r3,r3,64
661
6629:	bf	cr7*4+2,10f
663err3;	lvx	v1,0,r4
664	VPERM(v8,v0,v1,v16)
665err3;	lvx	v0,r4,r9
666	VPERM(v9,v1,v0,v16)
667	addi	r4,r4,32
668err3;	stvx	v8,0,r3
669err3;	stvx	v9,r3,r9
670	addi	r3,r3,32
671
67210:	bf	cr7*4+3,11f
673err3;	lvx	v1,0,r4
674	VPERM(v8,v0,v1,v16)
675	addi	r4,r4,16
676err3;	stvx	v8,0,r3
677	addi	r3,r3,16
678
679	/* Up to 15B to go */
68011:	clrldi	r5,r5,(64-4)
681	addi	r4,r4,-16	/* Unwind the +16 load offset */
682	mtocrf	0x01,r5
683	bf	cr7*4+0,12f
684err3;	lwz	r0,0(r4)	/* Less chance of a reject with word ops */
685err3;	lwz	r6,4(r4)
686	addi	r4,r4,8
687err3;	stw	r0,0(r3)
688err3;	stw	r6,4(r3)
689	addi	r3,r3,8
690
69112:	bf	cr7*4+1,13f
692err3;	lwz	r0,0(r4)
693	addi	r4,r4,4
694err3;	stw	r0,0(r3)
695	addi	r3,r3,4
696
69713:	bf	cr7*4+2,14f
698err3;	lhz	r0,0(r4)
699	addi	r4,r4,2
700err3;	sth	r0,0(r3)
701	addi	r3,r3,2
702
70314:	bf	cr7*4+3,15f
704err3;	lbz	r0,0(r4)
705err3;	stb	r0,0(r3)
706
70715:	addi	r1,r1,STACKFRAMESIZE
708	b	exit_vmx_usercopy	/* tail call optimise */
709#endif /* CONFIG_ALTIVEC */
710