1/* memcpy.S: Sparc optimized memcpy and memmove code
2 * Hand optimized from GNU libc's memcpy and memmove
3 * Copyright (C) 1991,1996 Free Software Foundation
4 * Copyright (C) 1995 Linus Torvalds (Linus.Torvalds@helsinki.fi)
5 * Copyright (C) 1996 David S. Miller (davem@caip.rutgers.edu)
6 * Copyright (C) 1996 Eddie C. Dost (ecd@skynet.be)
7 * Copyright (C) 1996 Jakub Jelinek (jj@sunsite.mff.cuni.cz)
8 */
9
10#define FUNC(x) 		\
11	.globl	x;		\
12	.type	x,@function;	\
13	.align	4;		\
14x:
15
16/* Both these macros have to start with exactly the same insn */
17#define MOVE_BIGCHUNK(src, dst, offset, t0, t1, t2, t3, t4, t5, t6, t7) \
18	ldd	[%src + (offset) + 0x00], %t0; \
19	ldd	[%src + (offset) + 0x08], %t2; \
20	ldd	[%src + (offset) + 0x10], %t4; \
21	ldd	[%src + (offset) + 0x18], %t6; \
22	st	%t0, [%dst + (offset) + 0x00]; \
23	st	%t1, [%dst + (offset) + 0x04]; \
24	st	%t2, [%dst + (offset) + 0x08]; \
25	st	%t3, [%dst + (offset) + 0x0c]; \
26	st	%t4, [%dst + (offset) + 0x10]; \
27	st	%t5, [%dst + (offset) + 0x14]; \
28	st	%t6, [%dst + (offset) + 0x18]; \
29	st	%t7, [%dst + (offset) + 0x1c];
30
31#define MOVE_BIGALIGNCHUNK(src, dst, offset, t0, t1, t2, t3, t4, t5, t6, t7) \
32	ldd	[%src + (offset) + 0x00], %t0; \
33	ldd	[%src + (offset) + 0x08], %t2; \
34	ldd	[%src + (offset) + 0x10], %t4; \
35	ldd	[%src + (offset) + 0x18], %t6; \
36	std	%t0, [%dst + (offset) + 0x00]; \
37	std	%t2, [%dst + (offset) + 0x08]; \
38	std	%t4, [%dst + (offset) + 0x10]; \
39	std	%t6, [%dst + (offset) + 0x18];
40
41#define MOVE_LASTCHUNK(src, dst, offset, t0, t1, t2, t3) \
42	ldd	[%src - (offset) - 0x10], %t0; \
43	ldd	[%src - (offset) - 0x08], %t2; \
44	st	%t0, [%dst - (offset) - 0x10]; \
45	st	%t1, [%dst - (offset) - 0x0c]; \
46	st	%t2, [%dst - (offset) - 0x08]; \
47	st	%t3, [%dst - (offset) - 0x04];
48
49#define MOVE_LASTALIGNCHUNK(src, dst, offset, t0, t1, t2, t3) \
50	ldd	[%src - (offset) - 0x10], %t0; \
51	ldd	[%src - (offset) - 0x08], %t2; \
52	std	%t0, [%dst - (offset) - 0x10]; \
53	std	%t2, [%dst - (offset) - 0x08];
54
55#define MOVE_SHORTCHUNK(src, dst, offset, t0, t1) \
56	ldub	[%src - (offset) - 0x02], %t0; \
57	ldub	[%src - (offset) - 0x01], %t1; \
58	stb	%t0, [%dst - (offset) - 0x02]; \
59	stb	%t1, [%dst - (offset) - 0x01];
60
61/* Both these macros have to start with exactly the same insn */
62#define RMOVE_BIGCHUNK(src, dst, offset, t0, t1, t2, t3, t4, t5, t6, t7) \
63	ldd	[%src - (offset) - 0x20], %t0; \
64	ldd	[%src - (offset) - 0x18], %t2; \
65	ldd	[%src - (offset) - 0x10], %t4; \
66	ldd	[%src - (offset) - 0x08], %t6; \
67	st	%t0, [%dst - (offset) - 0x20]; \
68	st	%t1, [%dst - (offset) - 0x1c]; \
69	st	%t2, [%dst - (offset) - 0x18]; \
70	st	%t3, [%dst - (offset) - 0x14]; \
71	st	%t4, [%dst - (offset) - 0x10]; \
72	st	%t5, [%dst - (offset) - 0x0c]; \
73	st	%t6, [%dst - (offset) - 0x08]; \
74	st	%t7, [%dst - (offset) - 0x04];
75
76#define RMOVE_BIGALIGNCHUNK(src, dst, offset, t0, t1, t2, t3, t4, t5, t6, t7) \
77	ldd	[%src - (offset) - 0x20], %t0; \
78	ldd	[%src - (offset) - 0x18], %t2; \
79	ldd	[%src - (offset) - 0x10], %t4; \
80	ldd	[%src - (offset) - 0x08], %t6; \
81	std	%t0, [%dst - (offset) - 0x20]; \
82	std	%t2, [%dst - (offset) - 0x18]; \
83	std	%t4, [%dst - (offset) - 0x10]; \
84	std	%t6, [%dst - (offset) - 0x08];
85
86#define RMOVE_LASTCHUNK(src, dst, offset, t0, t1, t2, t3) \
87	ldd	[%src + (offset) + 0x00], %t0; \
88	ldd	[%src + (offset) + 0x08], %t2; \
89	st	%t0, [%dst + (offset) + 0x00]; \
90	st	%t1, [%dst + (offset) + 0x04]; \
91	st	%t2, [%dst + (offset) + 0x08]; \
92	st	%t3, [%dst + (offset) + 0x0c];
93
94#define RMOVE_SHORTCHUNK(src, dst, offset, t0, t1) \
95	ldub	[%src + (offset) + 0x00], %t0; \
96	ldub	[%src + (offset) + 0x01], %t1; \
97	stb	%t0, [%dst + (offset) + 0x00]; \
98	stb	%t1, [%dst + (offset) + 0x01];
99
100#define SMOVE_CHUNK(src, dst, offset, t0, t1, t2, t3, t4, t5, t6, prev, shil, shir, offset2) \
101	ldd	[%src + (offset) + 0x00], %t0; \
102	ldd	[%src + (offset) + 0x08], %t2; \
103	srl	%t0, shir, %t5; \
104	srl	%t1, shir, %t6; \
105	sll	%t0, shil, %t0; \
106	or	%t5, %prev, %t5; \
107	sll	%t1, shil, %prev; \
108	or	%t6, %t0, %t0; \
109	srl	%t2, shir, %t1; \
110	srl	%t3, shir, %t6; \
111	sll	%t2, shil, %t2; \
112	or	%t1, %prev, %t1; \
113	std	%t4, [%dst + (offset) + (offset2) - 0x04]; \
114	std	%t0, [%dst + (offset) + (offset2) + 0x04]; \
115	sll	%t3, shil, %prev; \
116	or	%t6, %t2, %t4;
117
118#define SMOVE_ALIGNCHUNK(src, dst, offset, t0, t1, t2, t3, t4, t5, t6, prev, shil, shir, offset2) \
119	ldd	[%src + (offset) + 0x00], %t0; \
120	ldd	[%src + (offset) + 0x08], %t2; \
121	srl	%t0, shir, %t4;	\
122	srl	%t1, shir, %t5;	\
123	sll	%t0, shil, %t6;	\
124	or	%t4, %prev, %t0; \
125	sll	%t1, shil, %prev; \
126	or	%t5, %t6, %t1; \
127	srl	%t2, shir, %t4;	\
128	srl	%t3, shir, %t5;	\
129	sll	%t2, shil, %t6; \
130	or	%t4, %prev, %t2; \
131	sll	%t3, shil, %prev; \
132	or	%t5, %t6, %t3; \
133	std	%t0, [%dst + (offset) + (offset2) + 0x00]; \
134	std	%t2, [%dst + (offset) + (offset2) + 0x08];
135
136	.text
137	.align	4
138
1390:
140	retl
141	 nop		! Only bcopy returns here and it retuns void...
142
143#ifdef __KERNEL__
144FUNC(amemmove)
145FUNC(__memmove)
146#endif
147FUNC(memmove)
148	cmp		%o0, %o1
149	mov		%o0, %g7
150	bleu		9f
151	 sub		%o0, %o1, %o4
152
153	add		%o1, %o2, %o3
154	cmp		%o3, %o0
155	bleu		0f
156	 andcc		%o4, 3, %o5
157
158	add		%o1, %o2, %o1
159	add		%o0, %o2, %o0
160	sub		%o1, 1, %o1
161	sub		%o0, 1, %o0
162
1631:	/* reverse_bytes */
164
165	ldub		[%o1], %o4
166	subcc		%o2, 1, %o2
167	stb		%o4, [%o0]
168	sub		%o1, 1, %o1
169	bne		1b
170	 sub		%o0, 1, %o0
171
172	retl
173	 mov		%g7, %o0
174
175/* NOTE: This code is executed just for the cases,
176         where %src (=%o1) & 3 is != 0.
177	 We need to align it to 4. So, for (%src & 3)
178	 1 we need to do ldub,lduh
179	 2 lduh
180	 3 just ldub
181         so even if it looks weird, the branches
182         are correct here. -jj
183 */
18478:	/* dword_align */
185
186	andcc		%o1, 1, %g0
187	be		4f
188	 andcc		%o1, 2, %g0
189
190	ldub		[%o1], %g2
191	add		%o1, 1, %o1
192	stb		%g2, [%o0]
193	sub		%o2, 1, %o2
194	bne		3f
195	 add		%o0, 1, %o0
1964:
197	lduh		[%o1], %g2
198	add		%o1, 2, %o1
199	sth		%g2, [%o0]
200	sub		%o2, 2, %o2
201	b		3f
202	 add		%o0, 2, %o0
203
204FUNC(memcpy)	/* %o0=dst %o1=src %o2=len */
205
206	sub		%o0, %o1, %o4
207	mov		%o0, %g7
2089:
209	andcc		%o4, 3, %o5
2100:
211	bne		86f
212	 cmp		%o2, 15
213
214	bleu		90f
215	 andcc		%o1, 3, %g0
216
217	bne		78b
2183:
219	 andcc		%o1, 4, %g0
220
221	be		2f
222	 mov		%o2, %g1
223
224	ld		[%o1], %o4
225	sub		%g1, 4, %g1
226	st		%o4, [%o0]
227	add		%o1, 4, %o1
228	add		%o0, 4, %o0
2292:
230	andcc		%g1, 0xffffff80, %g0
231	be		3f
232	 andcc		%o0, 4, %g0
233
234	be		82f + 4
2355:
236	MOVE_BIGCHUNK(o1, o0, 0x00, o2, o3, o4, o5, g2, g3, g4, g5)
237	MOVE_BIGCHUNK(o1, o0, 0x20, o2, o3, o4, o5, g2, g3, g4, g5)
238	MOVE_BIGCHUNK(o1, o0, 0x40, o2, o3, o4, o5, g2, g3, g4, g5)
239	MOVE_BIGCHUNK(o1, o0, 0x60, o2, o3, o4, o5, g2, g3, g4, g5)
240	sub		%g1, 128, %g1
241	add		%o1, 128, %o1
242	cmp		%g1, 128
243	bge		5b
244	 add		%o0, 128, %o0
2453:
246	andcc		%g1, 0x70, %g4
247	be		80f
248	 andcc		%g1, 8, %g0
249
250	sethi		%hi(80f), %o5
251	srl		%g4, 1, %o4
252	add		%g4, %o4, %o4
253	add		%o1, %g4, %o1
254	sub		%o5, %o4, %o5
255	jmpl		%o5 + %lo(80f), %g0
256	 add		%o0, %g4, %o0
257
25879:	/* memcpy_table */
259
260	MOVE_LASTCHUNK(o1, o0, 0x60, g2, g3, g4, g5)
261	MOVE_LASTCHUNK(o1, o0, 0x50, g2, g3, g4, g5)
262	MOVE_LASTCHUNK(o1, o0, 0x40, g2, g3, g4, g5)
263	MOVE_LASTCHUNK(o1, o0, 0x30, g2, g3, g4, g5)
264	MOVE_LASTCHUNK(o1, o0, 0x20, g2, g3, g4, g5)
265	MOVE_LASTCHUNK(o1, o0, 0x10, g2, g3, g4, g5)
266	MOVE_LASTCHUNK(o1, o0, 0x00, g2, g3, g4, g5)
267
26880:	/* memcpy_table_end */
269	be		81f
270	 andcc		%g1, 4, %g0
271
272	ldd		[%o1], %g2
273	add		%o0, 8, %o0
274	st		%g2, [%o0 - 0x08]
275	add		%o1, 8, %o1
276	st		%g3, [%o0 - 0x04]
277
27881:	/* memcpy_last7 */
279
280	be		1f
281	 andcc		%g1, 2, %g0
282
283	ld		[%o1], %g2
284	add		%o1, 4, %o1
285	st		%g2, [%o0]
286	add		%o0, 4, %o0
2871:
288	be		1f
289	 andcc		%g1, 1, %g0
290
291	lduh		[%o1], %g2
292	add		%o1, 2, %o1
293	sth		%g2, [%o0]
294	add		%o0, 2, %o0
2951:
296	be		1f
297	 nop
298
299	ldub		[%o1], %g2
300	stb		%g2, [%o0]
3011:
302	retl
303	 mov		%g7, %o0
304
30582:	/* ldd_std */
306	MOVE_BIGALIGNCHUNK(o1, o0, 0x00, o2, o3, o4, o5, g2, g3, g4, g5)
307	MOVE_BIGALIGNCHUNK(o1, o0, 0x20, o2, o3, o4, o5, g2, g3, g4, g5)
308	MOVE_BIGALIGNCHUNK(o1, o0, 0x40, o2, o3, o4, o5, g2, g3, g4, g5)
309	MOVE_BIGALIGNCHUNK(o1, o0, 0x60, o2, o3, o4, o5, g2, g3, g4, g5)
310	subcc		%g1, 128, %g1
311	add		%o1, 128, %o1
312	cmp		%g1, 128
313	bge		82b
314	 add		%o0, 128, %o0
315
316	andcc		%g1, 0x70, %g4
317	be		84f
318	 andcc		%g1, 8, %g0
319
320	sethi		%hi(84f), %o5
321	add		%o1, %g4, %o1
322	sub		%o5, %g4, %o5
323	jmpl		%o5 + %lo(84f), %g0
324	 add		%o0, %g4, %o0
325
32683:	/* amemcpy_table */
327
328	MOVE_LASTALIGNCHUNK(o1, o0, 0x60, g2, g3, g4, g5)
329	MOVE_LASTALIGNCHUNK(o1, o0, 0x50, g2, g3, g4, g5)
330	MOVE_LASTALIGNCHUNK(o1, o0, 0x40, g2, g3, g4, g5)
331	MOVE_LASTALIGNCHUNK(o1, o0, 0x30, g2, g3, g4, g5)
332	MOVE_LASTALIGNCHUNK(o1, o0, 0x20, g2, g3, g4, g5)
333	MOVE_LASTALIGNCHUNK(o1, o0, 0x10, g2, g3, g4, g5)
334	MOVE_LASTALIGNCHUNK(o1, o0, 0x00, g2, g3, g4, g5)
335
33684:	/* amemcpy_table_end */
337	be		85f
338	 andcc		%g1, 4, %g0
339
340	ldd		[%o1], %g2
341	add		%o0, 8, %o0
342	std		%g2, [%o0 - 0x08]
343	add		%o1, 8, %o1
34485:	/* amemcpy_last7 */
345	be		1f
346	 andcc		%g1, 2, %g0
347
348	ld		[%o1], %g2
349	add		%o1, 4, %o1
350	st		%g2, [%o0]
351	add		%o0, 4, %o0
3521:
353	be		1f
354	 andcc		%g1, 1, %g0
355
356	lduh		[%o1], %g2
357	add		%o1, 2, %o1
358	sth		%g2, [%o0]
359	add		%o0, 2, %o0
3601:
361	be		1f
362	 nop
363
364	ldub		[%o1], %g2
365	stb		%g2, [%o0]
3661:
367	retl
368	 mov		%g7, %o0
369
37086:	/* non_aligned */
371	cmp		%o2, 6
372	bleu		88f
373	 nop
374
375	save		%sp, -96, %sp
376	andcc		%i0, 3, %g0
377	be		61f
378	 andcc		%i0, 1, %g0
379	be		60f
380	 andcc		%i0, 2, %g0
381
382	ldub		[%i1], %g5
383	add		%i1, 1, %i1
384	stb		%g5, [%i0]
385	sub		%i2, 1, %i2
386	bne		61f
387	 add		%i0, 1, %i0
38860:
389	ldub		[%i1], %g3
390	add		%i1, 2, %i1
391	stb		%g3, [%i0]
392	sub		%i2, 2, %i2
393	ldub		[%i1 - 1], %g3
394	add		%i0, 2, %i0
395	stb		%g3, [%i0 - 1]
39661:
397	and		%i1, 3, %g2
398	and		%i2, 0xc, %g3
399	and		%i1, -4, %i1
400	cmp		%g3, 4
401	sll		%g2, 3, %g4
402	mov		32, %g2
403	be		4f
404	 sub		%g2, %g4, %l0
405
406	blu		3f
407	 cmp		%g3, 0x8
408
409	be		2f
410	 srl		%i2, 2, %g3
411
412	ld		[%i1], %i3
413	add		%i0, -8, %i0
414	ld		[%i1 + 4], %i4
415	b		8f
416	 add		%g3, 1, %g3
4172:
418	ld		[%i1], %i4
419	add		%i0, -12, %i0
420	ld		[%i1 + 4], %i5
421	add		%g3, 2, %g3
422	b		9f
423	 add		%i1, -4, %i1
4243:
425	ld		[%i1], %g1
426	add		%i0, -4, %i0
427	ld		[%i1 + 4], %i3
428	srl		%i2, 2, %g3
429	b		7f
430	 add		%i1, 4, %i1
4314:
432	ld		[%i1], %i5
433	cmp		%i2, 7
434	ld		[%i1 + 4], %g1
435	srl		%i2, 2, %g3
436	bleu		10f
437	 add		%i1, 8, %i1
438
439	ld		[%i1], %i3
440	add		%g3, -1, %g3
4415:
442	sll		%i5, %g4, %g2
443	srl		%g1, %l0, %g5
444	or		%g2, %g5, %g2
445	st		%g2, [%i0]
4467:
447	ld		[%i1 + 4], %i4
448	sll		%g1, %g4, %g2
449	srl		%i3, %l0, %g5
450	or		%g2, %g5, %g2
451	st		%g2, [%i0 + 4]
4528:
453	ld		[%i1 + 8], %i5
454	sll		%i3, %g4, %g2
455	srl		%i4, %l0, %g5
456	or		%g2, %g5, %g2
457	st		%g2, [%i0 + 8]
4589:
459	ld		[%i1 + 12], %g1
460	sll		%i4, %g4, %g2
461	srl		%i5, %l0, %g5
462	addcc		%g3, -4, %g3
463	or		%g2, %g5, %g2
464	add		%i1, 16, %i1
465	st		%g2, [%i0 + 12]
466	add		%i0, 16, %i0
467	bne,a		5b
468	 ld		[%i1], %i3
46910:
470	sll		%i5, %g4, %g2
471	srl		%g1, %l0, %g5
472	srl		%l0, 3, %g3
473	or		%g2, %g5, %g2
474	sub		%i1, %g3, %i1
475	andcc		%i2, 2, %g0
476	st		%g2, [%i0]
477	be		1f
478	 andcc		%i2, 1, %g0
479
480	ldub		[%i1], %g2
481	add		%i1, 2, %i1
482	stb		%g2, [%i0 + 4]
483	add		%i0, 2, %i0
484	ldub		[%i1 - 1], %g2
485	stb		%g2, [%i0 + 3]
4861:
487	be		1f
488	 nop
489	ldub		[%i1], %g2
490	stb		%g2, [%i0 + 4]
4911:
492	ret
493	 restore	%g7, %g0, %o0
494
49588:	/* short_end */
496
497	and		%o2, 0xe, %o3
49820:
499	sethi		%hi(89f), %o5
500	sll		%o3, 3, %o4
501	add		%o0, %o3, %o0
502	sub		%o5, %o4, %o5
503	add		%o1, %o3, %o1
504	jmpl		%o5 + %lo(89f), %g0
505	 andcc		%o2, 1, %g0
506
507	MOVE_SHORTCHUNK(o1, o0, 0x0c, g2, g3)
508	MOVE_SHORTCHUNK(o1, o0, 0x0a, g2, g3)
509	MOVE_SHORTCHUNK(o1, o0, 0x08, g2, g3)
510	MOVE_SHORTCHUNK(o1, o0, 0x06, g2, g3)
511	MOVE_SHORTCHUNK(o1, o0, 0x04, g2, g3)
512	MOVE_SHORTCHUNK(o1, o0, 0x02, g2, g3)
513	MOVE_SHORTCHUNK(o1, o0, 0x00, g2, g3)
514
51589:	/* short_table_end */
516
517	be		1f
518	 nop
519
520	ldub		[%o1], %g2
521	stb		%g2, [%o0]
5221:
523	retl
524	 mov		%g7, %o0
525
52690:	/* short_aligned_end */
527	bne		88b
528	 andcc		%o2, 8, %g0
529
530	be		1f
531	 andcc		%o2, 4, %g0
532
533	ld		[%o1 + 0x00], %g2
534	ld		[%o1 + 0x04], %g3
535	add		%o1, 8, %o1
536	st		%g2, [%o0 + 0x00]
537	st		%g3, [%o0 + 0x04]
538	add		%o0, 8, %o0
5391:
540	b		81b
541	 mov		%o2, %g1
542