1 /*******************************************************************************
3 * Copyright (c) 1993 Intel Corporation
5 * Intel hereby grants you permission to copy, modify, and distribute this
6 * software and its documentation. Intel grants this permission provided
7 * that the above copyright notice appears in all copies and that both the
8 * copyright notice and this permission notice appear in supporting
9 * documentation. In addition, Intel grants this permission provided that
10 * you prominently mark as "not part of the original" any modifications
11 * made to this software or documentation, and that the name of Intel
12 * Corporation not be used in advertising or publicity pertaining to
13 * distribution of the software or the documentation without specific,
14 * written prior permission.
16 * Intel Corporation provides this AS IS, WITHOUT ANY WARRANTY, EXPRESS OR
17 * IMPLIED, INCLUDING, WITHOUT LIMITATION, ANY WARRANTY OF MERCHANTABILITY
18 * OR FITNESS FOR A PARTICULAR PURPOSE. Intel makes no guarantee or
19 * representations regarding the use of, or the results of the use of,
20 * the software and documentation in terms of correctness, accuracy,
21 * reliability, currentness, or otherwise; and you rely on the software,
22 * documentation and results solely at your own risk.
24 * IN NO EVENT SHALL INTEL BE LIABLE FOR ANY LOSS OF USE, LOSS OF BUSINESS,
25 * LOSS OF PROFITS, INDIRECT, INCIDENTAL, SPECIAL OR CONSEQUENTIAL DAMAGES
26 * OF ANY KIND. IN NO EVENT SHALL INTEL'S TOTAL LIABILITY EXCEED THE SUM
27 * PAID TO INTEL FOR THE PRODUCT LICENSED HEREUNDER.
29 ******************************************************************************/
39 * (c) copyright 1988,1993 Intel Corp., all rights reserved
43 procedure strncpy (optimized assembler version for the CA)
45 dest_addr = strncpy (dest_addr, src_addr, max_bytes)
47 copy the null terminated string pointed to by src_addr to
48 the string space pointed to by dest_addr. Return the original
49 dest_addr. If the source string is shorter than max_bytes,
50 then null-pad the destination string.
52 This routine will fail if the source and destination string
53 overlap (in particular, if the end of the source is overlapped
54 by the beginning of the destination). The behavior is undefined.
55 This is acceptable according to the draft C standard.
57 Undefined behavior will also occur if the end of the source string
58 (i.e. the terminating null byte) is in the last word of the program's
59 allocated memory space. This is so because, in several cases, strcpy
60 will fetch ahead one word. Disallowing the fetch ahead would impose
61 a severe performance penalty.
63 This program handles five cases:
65 1) both arguments start on a word boundary
66 2) neither are word aligned, but they are offset by the same amount
67 3) source is word aligned, destination is not
68 4) destination is word aligned, source is not
69 5) neither is word aligned, and they are offset by differing amounts
71 At the time of this writing, only g0 thru g7 and g13 are available
72 for use in this leafproc; other registers would have to be saved and
73 restored. These nine registers, plus tricky use of g14 are sufficient
74 to implement the routine. The registers are used as follows:
76 g0 original dest ptr; not modified, so that it may be returned.
77 g1 src ptr; shift count
79 g3 src ptr (word aligned)
80 g4 dest ptr (word aligned)
81 g5 0xff -- byte extraction mask
83 g6 lsw of double word for extraction of 4 bytes
84 g7 msw of double word for extraction of 4 bytes
86 g6 msw of double word for extraction of 4 bytes
87 g7 lsw of double word for extraction of 4 bytes
92 #if __i960_BIG_ENDIAN__
102 .leafproc _strncpy, __strncpy
108 lda Lrett-(.+8)(ip),g14
111 notand g1,3,g3 # extract word addr of start of src
112 lda (g14),g13 # preserve return address
113 cmpibge.f 0,g2,Lexit_code # Lexit if number of bytes to move is <= zero.
114 cmpo g3,g1 # check alignment of src
115 ld (g3),LSW # fetch word containing at least first byte
116 notand g0,3,g4 # extract word addr of start of dest
117 lda 4(g3),g3 # advance src word addr
118 bne.f Lcase_245 # branch if src is NOT word aligned
121 cmpo g0,g4 # check alignment of dest
122 lda 0xff,g5 # load mask for byte extraction
123 subo 4,g4,g4 # store is pre-incrementing; back up dest addr
124 bne.f Lcase_3 # branch if dest not word aligned
126 Lcase_1: # src and dest are word aligned
127 Lcase_1_wloop: # word copying loop
128 cmpi g2,4 # check for fewer than four bytes to move
129 lda (LSW),g1 # keep a copy of the src word
130 addo 4,g4,g4 # pre-increment dest addr
131 bl.f Lcase_1_cloop.a # branch if fewer than four bytes to copy
132 scanbyte 0,g1 # check for null byte in src word
133 ld (g3),LSW # pre-fetch next word of src
134 subi 4,g2,g2 # decrease max_byte count by the 4 bytes moved
135 bo.f Lcase_1_cloop.c # branch if word contains null byte
136 addo 4,g3,g3 # post-increment src addr
137 st g1,(g4) # store word in dest string
141 Lcase_1_cloop.a: # character copying loop (max_bytes <= 3)
142 #if __i960_BIG_ENDIAN__
143 rotate 8,g1,g1 # move next byte into position for extraction
145 and g5,g1,g14 # extract next char
147 cmpdeci 0,g2,g2 # is max_bytes exhausted?
148 be.f Lexit_code # Lexit if max_bytes is exhausted
149 cmpo 0,g14 # check for null byte
150 stob g14,(g4) # store the byte in dest
151 #if ! __i960_BIG_ENDIAN__
152 shro 8,g1,g1 # move next byte into position for extraction
154 lda 1(g4),g4 # post-increment dest byte addr
155 bne.t Lcase_1_cloop.a # branch if null not reached
159 mov 0,g14 # conform to register conventions
160 bx (g13) # g0 = addr of dest; g14 = 0
166 #if __i960_BIG_ENDIAN__
167 rotate 24,g5,g5 # move mask into position for testing next byte
169 and g5,g1,g14 # extract next char
170 cmpo 0,g14 # check for null byte
171 #if ! __i960_BIG_ENDIAN__
172 lda (g5),LSW # keep a copy of the current mask
173 shlo 8,g5,g5 # move mask into position for testing next byte
175 bne.t Lcase_1_cloop.c # branch if null not reached
176 #if __i960_BIG_ENDIAN__
177 subo 1,g5,g5 # null pad.
178 andnot g5,g1,g1 # last bytes to copy, and null pad rest of word
180 subo 1,LSW,g5 # mask to get last bytes to copy, and null pad
181 and g5,g1,g1 # last bytes to copy, and null pad rest of word
185 Lcase_1_zwloop: # zero word loop
186 cmpi g2,4 # check for fewer than four bytes to move
187 addo 4,g4,g4 # pre-increment dest addr
188 bl.f Lcase_1_cloop.b # branch if fewer than four bytes to copy
189 subo 4,g2,g2 # decrease max_byte count by the 4 bytes moved
190 st g14,(g4) # store word in dest string
193 Lcase_3: # src is word aligned; dest is not
194 addo 8,g4,g4 # move dest word ptr to first word boundary
195 lda (g0),g1 # copy dest byte ptr
196 mov LSW,MSW # make copy of first word of src
197 lda 32,g14 # initialize shift count to zero (mod 32)
200 Lcase_3_cloop_at_start: # character copying loop for start of dest str
201 cmpdeci 0,g2,g2 # is max_bytes exhausted?
202 #if __i960_BIG_ENDIAN__
203 shro 24,MSW,g5 # extract next char
205 and g5,MSW,g5 # extract next char
207 be.f Lexit_code # Lexit if max_bytes is exhausted
208 cmpo 0,g5 # check for null byte
209 stob g5,(g1) # store the byte in dest
210 addo 1,g1,g1 # post-increment dest ptr
211 lda 0xff,g5 # re-initialize byte extraction mask
212 bne.t 1f # drop thru if null byte reached (to pad)
213 movl 0,g6 # blank out remainder of input buffer
215 cmpo g1,g4 # have we reached word boundary in dest yet?
216 #if __i960_BIG_ENDIAN__
217 lda -8(g14),g14 # augment the shift counter
218 rotate 8,MSW,MSW # move next byte into position for extraction
220 lda 8(g14),g14 # augment the shift counter
221 shro 8,MSW,MSW # move next byte into position for extraction
223 bne.t Lcase_3_cloop_at_start # branch if reached word boundary?
225 ld (g3),MSW # fetch msw of operand for double shift
229 #if __i960_BIG_ENDIAN__
230 cmpobne 0,g14,Lcase_3_wloop # branch if src is still unaligned.
233 cmpi g2,4 # less than four bytes to move?
234 lda (LSW),g1 # extract 4 bytes of src
235 lda 4(g3),g3 # post-increment src word addr
236 bl.f Lcase_3_cloop.a # branch if < four bytes left to move
237 scanbyte 0,g1 # check for null byte
238 lda (MSW),LSW # move msw to lsw
239 subi 4,g2,g2 # decrease max_byte count by the 4 bytes moved
240 ld (g3),MSW # pre-fetch msw of operand for double shift
241 bo.f Lcase_3_cloop.c # branch if word contains null byte
242 st g1,(g4) # store 4 bytes to dest
243 addo 4,g4,g4 # post-increment dest ptr
249 cmpi g2,4 # less than four bytes to move?
250 eshro g14,g6,g1 # extract 4 bytes of src
251 lda 4(g3),g3 # post-increment src word addr
252 bl.f Lcase_3_cloop.a # branch if < four bytes left to move
253 scanbyte 0,g1 # check for null byte
254 lda (MSW),LSW # move msw to lsw
255 subi 4,g2,g2 # decrease max_byte count by the 4 bytes moved
256 ld (g3),MSW # pre-fetch msw of operand for double shift
257 bo.f Lcase_3_cloop.c # branch if word contains null byte
258 st g1,(g4) # store 4 bytes to dest
259 addo 4,g4,g4 # post-increment dest ptr
263 cmpo g0,g4 # check alignment of dest
264 ld (g3),MSW # pre-fetch second half
265 and 3,g1,g1 # compute shift count
266 lda 0xff,g5 # load mask for byte extraction
267 #if __i960_BIG_ENDIAN__
268 subo g1,4,g14 # adjust shift count for big endian.
273 be.t Lcase_4 # branch if dest is word aligned
274 or g4,g1,g1 # is src earlier in word, later, or sync w/ dst
275 cmpo g0,g1 # < indicates first word of dest has more bytes
276 /* than first word of source. */
277 lda 4(g4),g4 # move dest word addr to first word boundary
278 eshro g14,g6,g5 # extract four bytes
282 lda 4(g3),g3 # move src word addr to second word boundary