Lines Matching +full:src +full:- +full:2
1 /* SPDX-License-Identifier: GPL-2.0-or-later */
5 * Copyright (C) 2000-2002 Hewlett-Packard (John Marvin)
6 * Copyright (C) 2000 Richard Hirst <rhirst with parisc-linux.org>
8 * Copyright (C) 2003 Randolph Chung <tausq with parisc-linux.org>
40 addib,<> -1,%r25,$lclu_loop
47 2: b $lclu_done
50 ASM_EXCEPTIONTABLE_ENTRY(1b,2b)
58 * - sr1 already contains space of source region
59 * - sr2 already contains space of destination region
62 * - number of bytes that could not be copied.
65 * This code is based on a C-implementation of a copy routine written by
69 * conditions. In the optimal case, we copy by loops that copy 32- or 16-bytes
71 * aligning the destination and then using shift-and-write method, or in a few
72 * cases by falling back to a byte-at-a-time copy.
75 * often >10x faster than a simple byte-at-a-time copy, even for strangely
78 * it by 30-40% for aligned copies because of the loop unrolling, but in some
83 * - add cache prefetching
84 * - try not to use the post-increment address modifiers; they may create
90 src = arg1 define
118 xor src,dst,t0
119 extru t0,31,2,t1
123 /* only do 64-bit copies if we can get aligned. */
127 /* loop until we are 64-bit aligned */
131 20: ldb,ma 1(srcspc,src),t1
134 ldo -1(len),len
144 10: ldd 0(srcspc,src),t1
145 11: ldd 8(srcspc,src),t2
146 ldo 16(src),src
149 14: ldd 0(srcspc,src),t1
150 15: ldd 8(srcspc,src),t2
151 ldo 16(src),src
165 ldo -32(len),len
169 20: ldw,ma 4(srcspc,src),t1
172 ldo -4(len),len
179 /* loop until we are 32-bit aligned */
181 extru dst,31,2,t1
183 20: ldb,ma 1(srcspc,src),t1
186 ldo -1(len),len
195 10: ldw 0(srcspc,src),t1
196 11: ldw 4(srcspc,src),t2
199 14: ldw 8(srcspc,src),t1
200 15: ldw 12(srcspc,src),t2
201 ldo 16(src),src
215 ldo -16(len),len
220 20: ldb 0(srcspc,src),t1
221 ldo 1(src),src
224 ldo -1(len),len
234 /* src and dst are not aligned the same way. */
237 /* align until dst is 32bit-word-aligned */
238 extru dst,31,2,t1
240 20: ldb 0(srcspc,src),t1
241 ldo 1(src),src
244 ldo -1(len),len
251 /* store src, dst and len in safe place */
252 copy src,save_src
257 SHRREG len,2,len
260 * Copy from a not-aligned src to an aligned dst using shifts.
264 depw,z src,28,2,t0
267 extru len,31,2,t0
268 cmpib,= 2,t0,.Lcase2
269 /* Make src aligned by rounding it down. */
270 depi 0,31,2,src
280 1: ldw,ma 4(srcspc,src), a3
282 1: ldw,ma 4(srcspc,src), a0
286 1: ldw,ma 4(srcspc,src), a2
288 1: ldw,ma 4(srcspc,src), a3
290 ldo -1(len),len
293 1: ldw,ma 4(srcspc,src), a0
299 1: ldw,ma 4(srcspc,src), a1
305 1: ldw,ma 4(srcspc,src), a2
311 1: ldw,ma 4(srcspc,src), a3
316 ldo -4(len),len
326 /* calculate new src, dst and len and jump to byte-copy loop */
328 add save_src,t0,src
333 1: ldw,ma 4(srcspc,src), a0
335 1: ldw,ma 4(srcspc,src), a1
340 1: ldw,ma 4(srcspc,src), a1
342 1: ldw,ma 4(srcspc,src), a2
345 ldo 2(len),len