1/* U1memcpy.S: UltraSPARC-I/II/IIi/IIe optimized memcpy.
2 *
3 * Copyright (C) 1997, 2004 David S. Miller (davem@redhat.com)
4 * Copyright (C) 1996, 1997, 1998, 1999 Jakub Jelinek (jj@ultra.linux.cz)
5 */
6
7#ifdef __KERNEL__
8#include <asm/visasm.h>
9#include <asm/asi.h>
10#else
11#define ASI_BLK_P 0xf0
12#define FPRS_FEF  0x04
13#ifdef MEMCPY_DEBUG
14#define VISEntry rd %fprs, %o5; wr %g0, FPRS_FEF, %fprs; \
15		 clr %g1; clr %g2; clr %g3; subcc %g0, %g0, %g0;
16#define VISExit and %o5, FPRS_FEF, %o5; wr %o5, 0x0, %fprs
17#else
18#define VISEntry rd %fprs, %o5; wr %g0, FPRS_FEF, %fprs
19#define VISExit and %o5, FPRS_FEF, %o5; wr %o5, 0x0, %fprs
20#endif
21#endif
22
23#ifndef EX_LD
24#define EX_LD(x)	x
25#endif
26
27#ifndef EX_ST
28#define EX_ST(x)	x
29#endif
30
31#ifndef EX_RETVAL
32#define EX_RETVAL(x)	x
33#endif
34
35#ifndef LOAD
36#define LOAD(type,addr,dest)	type [addr], dest
37#endif
38
39#ifndef LOAD_BLK
40#define LOAD_BLK(addr,dest)	ldda [addr] ASI_BLK_P, dest
41#endif
42
43#ifndef STORE
44#define STORE(type,src,addr)	type src, [addr]
45#endif
46
47#ifndef STORE_BLK
48#define STORE_BLK(src,addr)	stda src, [addr] ASI_BLK_P
49#endif
50
51#ifndef FUNC_NAME
52#define FUNC_NAME	memcpy
53#endif
54
55#ifndef PREAMBLE
56#define PREAMBLE
57#endif
58
59#ifndef XCC
60#define XCC xcc
61#endif
62
63#define FREG_FROB(f1, f2, f3, f4, f5, f6, f7, f8, f9)		\
64	faligndata		%f1, %f2, %f48;			\
65	faligndata		%f2, %f3, %f50;			\
66	faligndata		%f3, %f4, %f52;			\
67	faligndata		%f4, %f5, %f54;			\
68	faligndata		%f5, %f6, %f56;			\
69	faligndata		%f6, %f7, %f58;			\
70	faligndata		%f7, %f8, %f60;			\
71	faligndata		%f8, %f9, %f62;
72
73#define MAIN_LOOP_CHUNK(src, dest, fdest, fsrc, len, jmptgt)	\
74	EX_LD(LOAD_BLK(%src, %fdest));				\
75	EX_ST(STORE_BLK(%fsrc, %dest));				\
76	add			%src, 0x40, %src;		\
77	subcc			%len, 0x40, %len;		\
78	be,pn			%xcc, jmptgt;			\
79	 add			%dest, 0x40, %dest;		\
80
81#define LOOP_CHUNK1(src, dest, len, branch_dest)		\
82	MAIN_LOOP_CHUNK(src, dest, f0,  f48, len, branch_dest)
83#define LOOP_CHUNK2(src, dest, len, branch_dest)		\
84	MAIN_LOOP_CHUNK(src, dest, f16, f48, len, branch_dest)
85#define LOOP_CHUNK3(src, dest, len, branch_dest)		\
86	MAIN_LOOP_CHUNK(src, dest, f32, f48, len, branch_dest)
87
88#define STORE_SYNC(dest, fsrc)				\
89	EX_ST(STORE_BLK(%fsrc, %dest));			\
90	add			%dest, 0x40, %dest;
91
92#define STORE_JUMP(dest, fsrc, target)			\
93	EX_ST(STORE_BLK(%fsrc, %dest));			\
94	add			%dest, 0x40, %dest;	\
95	ba,pt			%xcc, target;
96
97#define FINISH_VISCHUNK(dest, f0, f1, left)	\
98	subcc			%left, 8, %left;\
99	bl,pn			%xcc, 95f;	\
100	 faligndata		%f0, %f1, %f48;	\
101	EX_ST(STORE(std, %f48, %dest));		\
102	add			%dest, 8, %dest;
103
104#define UNEVEN_VISCHUNK_LAST(dest, f0, f1, left)	\
105	subcc			%left, 8, %left;	\
106	bl,pn			%xcc, 95f;		\
107	 fsrc1			%f0, %f1;
108
109#define UNEVEN_VISCHUNK(dest, f0, f1, left)		\
110	UNEVEN_VISCHUNK_LAST(dest, f0, f1, left)	\
111	ba,a,pt			%xcc, 93f;
112
113	.register	%g2,#scratch
114	.register	%g3,#scratch
115
116	.text
117	.align		64
118
119	.globl		FUNC_NAME
120FUNC_NAME:		/* %o0=dst, %o1=src, %o2=len */
121	PREAMBLE
122	mov		%o0, %g5
123	cmp		%o2, 0
124	be,pn		%XCC, 85f
125	 or		%o0, %o1, %o3
126	cmp		%o2, 16
127	blu,a,pn	%XCC, 80f
128	 or		%o3, %o2, %o3
129
130	cmp		%o2, (5 * 64)
131	blu,pt		%XCC, 70f
132	 andcc		%o3, 0x7, %g0
133
134	/* Clobbers o5/g1/g2/g3/g7/icc/xcc.  */
135	VISEntry
136
137	/* Is 'dst' already aligned on an 64-byte boundary? */
138	andcc		%o0, 0x3f, %g2
139	be,pt		%XCC, 2f
140
141	/* Compute abs((dst & 0x3f) - 0x40) into %g2.  This is the number
142	 * of bytes to copy to make 'dst' 64-byte aligned.  We pre-
143	 * subtract this from 'len'.
144	 */
145	 sub		%o0, %o1, %o4
146	sub		%g2, 0x40, %g2
147	sub		%g0, %g2, %g2
148	sub		%o2, %g2, %o2
149	andcc		%g2, 0x7, %g1
150	be,pt		%icc, 2f
151	 and		%g2, 0x38, %g2
152
1531:	subcc		%g1, 0x1, %g1
154	EX_LD(LOAD(ldub, %o1 + 0x00, %o3))
155	EX_ST(STORE(stb, %o3, %o1 + %o4))
156	bgu,pt		%XCC, 1b
157	 add		%o1, 0x1, %o1
158
159	add		%o1, %o4, %o0
160
1612:	cmp		%g2, 0x0
162	and		%o1, 0x7, %g1
163	be,pt		%icc, 3f
164	 alignaddr	%o1, %g0, %o1
165
166	EX_LD(LOAD(ldd, %o1, %f4))
1671:	EX_LD(LOAD(ldd, %o1 + 0x8, %f6))
168	add		%o1, 0x8, %o1
169	subcc		%g2, 0x8, %g2
170	faligndata	%f4, %f6, %f0
171	EX_ST(STORE(std, %f0, %o0))
172	be,pn		%icc, 3f
173	 add		%o0, 0x8, %o0
174
175	EX_LD(LOAD(ldd, %o1 + 0x8, %f4))
176	add		%o1, 0x8, %o1
177	subcc		%g2, 0x8, %g2
178	faligndata	%f6, %f4, %f0
179	EX_ST(STORE(std, %f0, %o0))
180	bne,pt		%icc, 1b
181	 add		%o0, 0x8, %o0
182
183	/* Destination is 64-byte aligned.  */
1843:
185	membar		  #LoadStore | #StoreStore | #StoreLoad
186
187	subcc		%o2, 0x40, %o4
188	add		%o1, %g1, %g1
189	andncc		%o4, (0x40 - 1), %o4
190	srl		%g1, 3, %g2
191	sub		%o2, %o4, %g3
192	andn		%o1, (0x40 - 1), %o1
193	and		%g2, 7, %g2
194	andncc		%g3, 0x7, %g3
195	fmovd		%f0, %f2
196	sub		%g3, 0x8, %g3
197	sub		%o2, %o4, %o2
198
199	add		%g1, %o4, %g1
200	subcc		%o2, %g3, %o2
201
202	EX_LD(LOAD_BLK(%o1, %f0))
203	add		%o1, 0x40, %o1
204	add		%g1, %g3, %g1
205	EX_LD(LOAD_BLK(%o1, %f16))
206	add		%o1, 0x40, %o1
207	sub		%o4, 0x80, %o4
208	EX_LD(LOAD_BLK(%o1, %f32))
209	add		%o1, 0x40, %o1
210
211	/* There are 8 instances of the unrolled loop,
212	 * one for each possible alignment of the
213	 * source buffer.  Each loop instance is 452
214	 * bytes.
215	 */
216	sll		%g2, 3, %o3
217	sub		%o3, %g2, %o3
218	sllx		%o3, 4, %o3
219	add		%o3, %g2, %o3
220	sllx		%o3, 2, %g2
2211:	rd		%pc, %o3
222	add		%o3, %lo(1f - 1b), %o3
223	jmpl		%o3 + %g2, %g0
224	 nop
225
226	.align		64
2271:	FREG_FROB(f0, f2, f4, f6, f8, f10,f12,f14,f16)
228	LOOP_CHUNK1(o1, o0, o4, 1f)
229	FREG_FROB(f16,f18,f20,f22,f24,f26,f28,f30,f32)
230	LOOP_CHUNK2(o1, o0, o4, 2f)
231	FREG_FROB(f32,f34,f36,f38,f40,f42,f44,f46,f0)
232	LOOP_CHUNK3(o1, o0, o4, 3f)
233	ba,pt		%xcc, 1b+4
234	 faligndata	%f0, %f2, %f48
2351:	FREG_FROB(f16,f18,f20,f22,f24,f26,f28,f30,f32)
236	STORE_SYNC(o0, f48) membar #Sync
237	FREG_FROB(f32,f34,f36,f38,f40,f42,f44,f46,f0)
238	STORE_JUMP(o0, f48, 40f) membar #Sync
2392:	FREG_FROB(f32,f34,f36,f38,f40,f42,f44,f46,f0)
240	STORE_SYNC(o0, f48) membar #Sync
241	FREG_FROB(f0, f2, f4, f6, f8, f10,f12,f14,f16)
242	STORE_JUMP(o0, f48, 48f) membar #Sync
2433:	FREG_FROB(f0, f2, f4, f6, f8, f10,f12,f14,f16)
244	STORE_SYNC(o0, f48) membar #Sync
245	FREG_FROB(f16,f18,f20,f22,f24,f26,f28,f30,f32)
246	STORE_JUMP(o0, f48, 56f) membar #Sync
247
2481:	FREG_FROB(f2, f4, f6, f8, f10,f12,f14,f16,f18)
249	LOOP_CHUNK1(o1, o0, o4, 1f)
250	FREG_FROB(f18,f20,f22,f24,f26,f28,f30,f32,f34)
251	LOOP_CHUNK2(o1, o0, o4, 2f)
252	FREG_FROB(f34,f36,f38,f40,f42,f44,f46,f0, f2)
253	LOOP_CHUNK3(o1, o0, o4, 3f)
254	ba,pt		%xcc, 1b+4
255	 faligndata	%f2, %f4, %f48
2561:	FREG_FROB(f18,f20,f22,f24,f26,f28,f30,f32,f34)
257	STORE_SYNC(o0, f48) membar #Sync
258	FREG_FROB(f34,f36,f38,f40,f42,f44,f46,f0, f2)
259	STORE_JUMP(o0, f48, 41f) membar #Sync
2602:	FREG_FROB(f34,f36,f38,f40,f42,f44,f46,f0, f2)
261	STORE_SYNC(o0, f48) membar #Sync
262	FREG_FROB(f2, f4, f6, f8, f10,f12,f14,f16,f18)
263	STORE_JUMP(o0, f48, 49f) membar #Sync
2643:	FREG_FROB(f2, f4, f6, f8, f10,f12,f14,f16,f18)
265	STORE_SYNC(o0, f48) membar #Sync
266	FREG_FROB(f18,f20,f22,f24,f26,f28,f30,f32,f34)
267	STORE_JUMP(o0, f48, 57f) membar #Sync
268
2691:	FREG_FROB(f4, f6, f8, f10,f12,f14,f16,f18,f20)
270	LOOP_CHUNK1(o1, o0, o4, 1f)
271	FREG_FROB(f20,f22,f24,f26,f28,f30,f32,f34,f36)
272	LOOP_CHUNK2(o1, o0, o4, 2f)
273	FREG_FROB(f36,f38,f40,f42,f44,f46,f0, f2, f4)
274	LOOP_CHUNK3(o1, o0, o4, 3f)
275	ba,pt		%xcc, 1b+4
276	 faligndata	%f4, %f6, %f48
2771:	FREG_FROB(f20,f22,f24,f26,f28,f30,f32,f34,f36)
278	STORE_SYNC(o0, f48) membar #Sync
279	FREG_FROB(f36,f38,f40,f42,f44,f46,f0, f2, f4)
280	STORE_JUMP(o0, f48, 42f) membar #Sync
2812:	FREG_FROB(f36,f38,f40,f42,f44,f46,f0, f2, f4)
282	STORE_SYNC(o0, f48) membar #Sync
283	FREG_FROB(f4, f6, f8, f10,f12,f14,f16,f18,f20)
284	STORE_JUMP(o0, f48, 50f) membar #Sync
2853:	FREG_FROB(f4, f6, f8, f10,f12,f14,f16,f18,f20)
286	STORE_SYNC(o0, f48) membar #Sync
287	FREG_FROB(f20,f22,f24,f26,f28,f30,f32,f34,f36)
288	STORE_JUMP(o0, f48, 58f) membar #Sync
289
2901:	FREG_FROB(f6, f8, f10,f12,f14,f16,f18,f20,f22)
291	LOOP_CHUNK1(o1, o0, o4, 1f)
292	FREG_FROB(f22,f24,f26,f28,f30,f32,f34,f36,f38)
293	LOOP_CHUNK2(o1, o0, o4, 2f)
294	FREG_FROB(f38,f40,f42,f44,f46,f0, f2, f4, f6)
295	LOOP_CHUNK3(o1, o0, o4, 3f)
296	ba,pt		%xcc, 1b+4
297	 faligndata	%f6, %f8, %f48
2981:	FREG_FROB(f22,f24,f26,f28,f30,f32,f34,f36,f38)
299	STORE_SYNC(o0, f48) membar #Sync
300	FREG_FROB(f38,f40,f42,f44,f46,f0, f2, f4, f6)
301	STORE_JUMP(o0, f48, 43f) membar #Sync
3022:	FREG_FROB(f38,f40,f42,f44,f46,f0, f2, f4, f6)
303	STORE_SYNC(o0, f48) membar #Sync
304	FREG_FROB(f6, f8, f10,f12,f14,f16,f18,f20,f22)
305	STORE_JUMP(o0, f48, 51f) membar #Sync
3063:	FREG_FROB(f6, f8, f10,f12,f14,f16,f18,f20,f22)
307	STORE_SYNC(o0, f48) membar #Sync
308	FREG_FROB(f22,f24,f26,f28,f30,f32,f34,f36,f38)
309	STORE_JUMP(o0, f48, 59f) membar #Sync
310
3111:	FREG_FROB(f8, f10,f12,f14,f16,f18,f20,f22,f24)
312	LOOP_CHUNK1(o1, o0, o4, 1f)
313	FREG_FROB(f24,f26,f28,f30,f32,f34,f36,f38,f40)
314	LOOP_CHUNK2(o1, o0, o4, 2f)
315	FREG_FROB(f40,f42,f44,f46,f0, f2, f4, f6, f8)
316	LOOP_CHUNK3(o1, o0, o4, 3f)
317	ba,pt		%xcc, 1b+4
318	 faligndata	%f8, %f10, %f48
3191:	FREG_FROB(f24,f26,f28,f30,f32,f34,f36,f38,f40)
320	STORE_SYNC(o0, f48) membar #Sync
321	FREG_FROB(f40,f42,f44,f46,f0, f2, f4, f6, f8)
322	STORE_JUMP(o0, f48, 44f) membar #Sync
3232:	FREG_FROB(f40,f42,f44,f46,f0, f2, f4, f6, f8)
324	STORE_SYNC(o0, f48) membar #Sync
325	FREG_FROB(f8, f10,f12,f14,f16,f18,f20,f22,f24)
326	STORE_JUMP(o0, f48, 52f) membar #Sync
3273:	FREG_FROB(f8, f10,f12,f14,f16,f18,f20,f22,f24)
328	STORE_SYNC(o0, f48) membar #Sync
329	FREG_FROB(f24,f26,f28,f30,f32,f34,f36,f38,f40)
330	STORE_JUMP(o0, f48, 60f) membar #Sync
331
3321:	FREG_FROB(f10,f12,f14,f16,f18,f20,f22,f24,f26)
333	LOOP_CHUNK1(o1, o0, o4, 1f)
334	FREG_FROB(f26,f28,f30,f32,f34,f36,f38,f40,f42)
335	LOOP_CHUNK2(o1, o0, o4, 2f)
336	FREG_FROB(f42,f44,f46,f0, f2, f4, f6, f8, f10)
337	LOOP_CHUNK3(o1, o0, o4, 3f)
338	ba,pt		%xcc, 1b+4
339	 faligndata	%f10, %f12, %f48
3401:	FREG_FROB(f26,f28,f30,f32,f34,f36,f38,f40,f42)
341	STORE_SYNC(o0, f48) membar #Sync
342	FREG_FROB(f42,f44,f46,f0, f2, f4, f6, f8, f10)
343	STORE_JUMP(o0, f48, 45f) membar #Sync
3442:	FREG_FROB(f42,f44,f46,f0, f2, f4, f6, f8, f10)
345	STORE_SYNC(o0, f48) membar #Sync
346	FREG_FROB(f10,f12,f14,f16,f18,f20,f22,f24,f26)
347	STORE_JUMP(o0, f48, 53f) membar #Sync
3483:	FREG_FROB(f10,f12,f14,f16,f18,f20,f22,f24,f26)
349	STORE_SYNC(o0, f48) membar #Sync
350	FREG_FROB(f26,f28,f30,f32,f34,f36,f38,f40,f42)
351	STORE_JUMP(o0, f48, 61f) membar #Sync
352
3531:	FREG_FROB(f12,f14,f16,f18,f20,f22,f24,f26,f28)
354	LOOP_CHUNK1(o1, o0, o4, 1f)
355	FREG_FROB(f28,f30,f32,f34,f36,f38,f40,f42,f44)
356	LOOP_CHUNK2(o1, o0, o4, 2f)
357	FREG_FROB(f44,f46,f0, f2, f4, f6, f8, f10,f12)
358	LOOP_CHUNK3(o1, o0, o4, 3f)
359	ba,pt		%xcc, 1b+4
360	 faligndata	%f12, %f14, %f48
3611:	FREG_FROB(f28,f30,f32,f34,f36,f38,f40,f42,f44)
362	STORE_SYNC(o0, f48) membar #Sync
363	FREG_FROB(f44,f46,f0, f2, f4, f6, f8, f10,f12)
364	STORE_JUMP(o0, f48, 46f) membar #Sync
3652:	FREG_FROB(f44,f46,f0, f2, f4, f6, f8, f10,f12)
366	STORE_SYNC(o0, f48) membar #Sync
367	FREG_FROB(f12,f14,f16,f18,f20,f22,f24,f26,f28)
368	STORE_JUMP(o0, f48, 54f) membar #Sync
3693:	FREG_FROB(f12,f14,f16,f18,f20,f22,f24,f26,f28)
370	STORE_SYNC(o0, f48) membar #Sync
371	FREG_FROB(f28,f30,f32,f34,f36,f38,f40,f42,f44)
372	STORE_JUMP(o0, f48, 62f) membar #Sync
373
3741:	FREG_FROB(f14,f16,f18,f20,f22,f24,f26,f28,f30)
375	LOOP_CHUNK1(o1, o0, o4, 1f)
376	FREG_FROB(f30,f32,f34,f36,f38,f40,f42,f44,f46)
377	LOOP_CHUNK2(o1, o0, o4, 2f)
378	FREG_FROB(f46,f0, f2, f4, f6, f8, f10,f12,f14)
379	LOOP_CHUNK3(o1, o0, o4, 3f)
380	ba,pt		%xcc, 1b+4
381	 faligndata	%f14, %f16, %f48
3821:	FREG_FROB(f30,f32,f34,f36,f38,f40,f42,f44,f46)
383	STORE_SYNC(o0, f48) membar #Sync
384	FREG_FROB(f46,f0, f2, f4, f6, f8, f10,f12,f14)
385	STORE_JUMP(o0, f48, 47f) membar #Sync
3862:	FREG_FROB(f46,f0, f2, f4, f6, f8, f10,f12,f14)
387	STORE_SYNC(o0, f48) membar #Sync
388	FREG_FROB(f14,f16,f18,f20,f22,f24,f26,f28,f30)
389	STORE_JUMP(o0, f48, 55f) membar #Sync
3903:	FREG_FROB(f14,f16,f18,f20,f22,f24,f26,f28,f30)
391	STORE_SYNC(o0, f48) membar #Sync
392	FREG_FROB(f30,f32,f34,f36,f38,f40,f42,f44,f46)
393	STORE_JUMP(o0, f48, 63f) membar #Sync
394
39540:	FINISH_VISCHUNK(o0, f0,  f2,  g3)
39641:	FINISH_VISCHUNK(o0, f2,  f4,  g3)
39742:	FINISH_VISCHUNK(o0, f4,  f6,  g3)
39843:	FINISH_VISCHUNK(o0, f6,  f8,  g3)
39944:	FINISH_VISCHUNK(o0, f8,  f10, g3)
40045:	FINISH_VISCHUNK(o0, f10, f12, g3)
40146:	FINISH_VISCHUNK(o0, f12, f14, g3)
40247:	UNEVEN_VISCHUNK(o0, f14, f0,  g3)
40348:	FINISH_VISCHUNK(o0, f16, f18, g3)
40449:	FINISH_VISCHUNK(o0, f18, f20, g3)
40550:	FINISH_VISCHUNK(o0, f20, f22, g3)
40651:	FINISH_VISCHUNK(o0, f22, f24, g3)
40752:	FINISH_VISCHUNK(o0, f24, f26, g3)
40853:	FINISH_VISCHUNK(o0, f26, f28, g3)
40954:	FINISH_VISCHUNK(o0, f28, f30, g3)
41055:	UNEVEN_VISCHUNK(o0, f30, f0,  g3)
41156:	FINISH_VISCHUNK(o0, f32, f34, g3)
41257:	FINISH_VISCHUNK(o0, f34, f36, g3)
41358:	FINISH_VISCHUNK(o0, f36, f38, g3)
41459:	FINISH_VISCHUNK(o0, f38, f40, g3)
41560:	FINISH_VISCHUNK(o0, f40, f42, g3)
41661:	FINISH_VISCHUNK(o0, f42, f44, g3)
41762:	FINISH_VISCHUNK(o0, f44, f46, g3)
41863:	UNEVEN_VISCHUNK_LAST(o0, f46, f0,  g3)
419
42093:	EX_LD(LOAD(ldd, %o1, %f2))
421	add		%o1, 8, %o1
422	subcc		%g3, 8, %g3
423	faligndata	%f0, %f2, %f8
424	EX_ST(STORE(std, %f8, %o0))
425	bl,pn		%xcc, 95f
426	 add		%o0, 8, %o0
427	EX_LD(LOAD(ldd, %o1, %f0))
428	add		%o1, 8, %o1
429	subcc		%g3, 8, %g3
430	faligndata	%f2, %f0, %f8
431	EX_ST(STORE(std, %f8, %o0))
432	bge,pt		%xcc, 93b
433	 add		%o0, 8, %o0
434
43595:	brz,pt		%o2, 2f
436	 mov		%g1, %o1
437
4381:	EX_LD(LOAD(ldub, %o1, %o3))
439	add		%o1, 1, %o1
440	subcc		%o2, 1, %o2
441	EX_ST(STORE(stb, %o3, %o0))
442	bne,pt		%xcc, 1b
443	 add		%o0, 1, %o0
444
4452:	membar		#StoreLoad | #StoreStore
446	VISExit
447	retl
448	 mov		EX_RETVAL(%g5), %o0
449
450	.align		64
45170:	/* 16 < len <= (5 * 64) */
452	bne,pn		%XCC, 75f
453	 sub		%o0, %o1, %o3
454
45572:	andn		%o2, 0xf, %o4
456	and		%o2, 0xf, %o2
4571:	EX_LD(LOAD(ldx, %o1 + 0x00, %o5))
458	EX_LD(LOAD(ldx, %o1 + 0x08, %g1))
459	subcc		%o4, 0x10, %o4
460	EX_ST(STORE(stx, %o5, %o1 + %o3))
461	add		%o1, 0x8, %o1
462	EX_ST(STORE(stx, %g1, %o1 + %o3))
463	bgu,pt		%XCC, 1b
464	 add		%o1, 0x8, %o1
46573:	andcc		%o2, 0x8, %g0
466	be,pt		%XCC, 1f
467	 nop
468	EX_LD(LOAD(ldx, %o1, %o5))
469	sub		%o2, 0x8, %o2
470	EX_ST(STORE(stx, %o5, %o1 + %o3))
471	add		%o1, 0x8, %o1
4721:	andcc		%o2, 0x4, %g0
473	be,pt		%XCC, 1f
474	 nop
475	EX_LD(LOAD(lduw, %o1, %o5))
476	sub		%o2, 0x4, %o2
477	EX_ST(STORE(stw, %o5, %o1 + %o3))
478	add		%o1, 0x4, %o1
4791:	cmp		%o2, 0
480	be,pt		%XCC, 85f
481	 nop
482	ba,pt		%xcc, 90f
483	 nop
484
48575:	andcc		%o0, 0x7, %g1
486	sub		%g1, 0x8, %g1
487	be,pn		%icc, 2f
488	 sub		%g0, %g1, %g1
489	sub		%o2, %g1, %o2
490
4911:	EX_LD(LOAD(ldub, %o1, %o5))
492	subcc		%g1, 1, %g1
493	EX_ST(STORE(stb, %o5, %o1 + %o3))
494	bgu,pt		%icc, 1b
495	 add		%o1, 1, %o1
496
4972:	add		%o1, %o3, %o0
498	andcc		%o1, 0x7, %g1
499	bne,pt		%icc, 8f
500	 sll		%g1, 3, %g1
501
502	cmp		%o2, 16
503	bgeu,pt		%icc, 72b
504	 nop
505	ba,a,pt		%xcc, 73b
506
5078:	mov		64, %o3
508	andn		%o1, 0x7, %o1
509	EX_LD(LOAD(ldx, %o1, %g2))
510	sub		%o3, %g1, %o3
511	andn		%o2, 0x7, %o4
512	sllx		%g2, %g1, %g2
5131:	EX_LD(LOAD(ldx, %o1 + 0x8, %g3))
514	subcc		%o4, 0x8, %o4
515	add		%o1, 0x8, %o1
516	srlx		%g3, %o3, %o5
517	or		%o5, %g2, %o5
518	EX_ST(STORE(stx, %o5, %o0))
519	add		%o0, 0x8, %o0
520	bgu,pt		%icc, 1b
521	 sllx		%g3, %g1, %g2
522
523	srl		%g1, 3, %g1
524	andcc		%o2, 0x7, %o2
525	be,pn		%icc, 85f
526	 add		%o1, %g1, %o1
527	ba,pt		%xcc, 90f
528	 sub		%o0, %o1, %o3
529
530	.align		64
53180:	/* 0 < len <= 16 */
532	andcc		%o3, 0x3, %g0
533	bne,pn		%XCC, 90f
534	 sub		%o0, %o1, %o3
535
5361:	EX_LD(LOAD(lduw, %o1, %g1))
537	subcc		%o2, 4, %o2
538	EX_ST(STORE(stw, %g1, %o1 + %o3))
539	bgu,pt		%XCC, 1b
540	 add		%o1, 4, %o1
541
54285:	retl
543	 mov		EX_RETVAL(%g5), %o0
544
545	.align		32
54690:	EX_LD(LOAD(ldub, %o1, %g1))
547	subcc		%o2, 1, %o2
548	EX_ST(STORE(stb, %g1, %o1 + %o3))
549	bgu,pt		%XCC, 90b
550	 add		%o1, 1, %o1
551	retl
552	 mov		EX_RETVAL(%g5), %o0
553