Retro68/gcc/newlib/libc/machine/hppa/strncat.S
2017-10-07 02:16:47 +02:00

239 lines
8.2 KiB
ArmAsm

/*
* (c) Copyright 1986 HEWLETT-PACKARD COMPANY
*
* To anyone who acknowledges that this file is provided "AS IS"
* without any express or implied warranty:
* permission to use, copy, modify, and distribute this file
* for any purpose is hereby granted without fee, provided that
* the above copyright notice and this notice appears in all
* copies, and that the name of Hewlett-Packard Company not be
* used in advertising or publicity pertaining to distribution
* of the software without specific, written prior permission.
* Hewlett-Packard Company makes no representations about the
* suitability of this software for any purpose.
*/
/*HPUX_ID: @(#) $Revision$ */
/* strncat(s1,s2,n) : concatonate at most n characters from s2 onto s1 */
#include "DEFS.h"
#define d_addr r26
#define s_addr r25
#define count r24
#define tmp1 r19
#define tmp2 r20
#define tmp3 r21
#define tmp4 r22
#define tmp5 arg3
#define tmp6 r31
#define save r1
#define tmp7 ret1 /* source offset-- reset to orig source addr if not aligned */
ENTRY(strncat)
comb,= r0,s_addr,quit /* quit if s2=NULL */
copy d_addr,ret0 /* The return value is the value of d_addr. DELAY SLOT*/
/* First look for end of s1 (d_addr) */
extru d_addr,31,2,tmp1 /* Extract the low two bits of the dest address. */
combt,= tmp1,r0,dont_mask
dep 0,31,2,d_addr /*set word alignment */
ldwm 4(d_addr),tmp2
sh3add tmp1,r0,save /* build mask based on tmp1 */
mtctl save,11
zvdepi -2,32,save
or save,tmp2,tmp2
uxor,nbz tmp2,r0,save
search:
b,n found_end /* nullified under uxor conditions above and below */
dont_mask:
ldwm 4(d_addr),tmp2
comib,tr r0,r0,search
uxor,nbz tmp2,r0,save
found_end: /* at this point d_addr points to word */
extru,<> save,7,8,r0 /* following word with null */
addib,tr,n -4,d_addr,begin_copy /*set d_addr to end of s1 */
extru,<> save,15,8,r0
addib,tr,n -3,d_addr,begin_copy
extru,<> save,23,8,r0
addi -1,d_addr,d_addr
addi -1,d_addr,d_addr
begin_copy:
addibt,<,n -4,count,byteloop /* If count is <= 4 don't get fancy.*/
extru s_addr,31,2,tmp4 /* Extract the low two bits of the source address.*/
extru d_addr,31,2,tmp5 /* Extract the low two bits of the destination address.*/
add count,tmp5,count /* pre increment the count by the byte address so that the count is*/
copy s_addr,tmp6 /* save original s_addr in case we find null in first word */
copy s_addr, tmp7 /* save s_addr in case we find null before first store */
comb,<> tmp5,tmp4,not_aligned /* branch if tmp5<>tmp4. */
dep 0,31,2,s_addr /* Compute the word address of the source. DELAY SLOT.*/
/* aligned*/
combt,= tmp5,r0,skip_mask
ldwm 4(0,s_addr),tmp1 /* tmp1 = *s_addr s_addr += 4 (DELAY SLOT)*/
sh3add tmp5,r0,save /* compute mask in save*/
mtctl save,11
zvdepi -2,32,save
or save,tmp1,tmp1 /* or mask with data*/
uxor,nbz tmp1,r0,save /* check for null*/
b,n null1
addibt,< -4,count,back_porch
stbys,b,m tmp1,4(0,d_addr) /* store word (delay slot)*/
chunks:
ldwm 4(0,s_addr),tmp1 /* get a word*/
skip_mask:
uxor,nbz tmp1,r0,save /* check for null*/
b,n align_null1
addibf,< -4,count,chunks
stbys,b,m tmp1,4(0,d_addr) /* store word (delay slot)*/
back_porch: /* last word to store*/
addibt,=,n 4,count,done /* if count = 0 we're, of course, done !*/
ldws 0(s_addr),tmp1 /* load up the back_porch*/
sh3add count,r0, save /* setup right mask based on count*/
mtctl save,r11
zvdepi -2,32,save /*save now has left-hand mask*/
uaddcm r0,save,save /*form right hand mask */
or tmp1,save,tmp1 /*and insert data*/
uxor,nbz tmp1,r0,save /* check for null*/
b,n null2
add d_addr,count,d_addr/* final store address is +1 too high !*/
b done
stbys,e tmp1,0(d_addr) /* done */
/* Begin non_aligned code. */
not_aligned:
sub,>= tmp5,tmp4,tmp6 /* compute the shift amt.and skip load if tmp5 > tmp4.*/
ldwm 4(0,s_addr),tmp1 /* load up the first word from the source. tmp1 = *s_addr++*/
zdep tmp6,28,29,tmp4 /* compute the number of bits to shift */
mtctl tmp4,11 /* load the shift count into cr11 = shift count register.*/
addibt,<,n -4,count,chkchnk2 /* first step in pre adjustment of count for looping.*/
ldwm 4(0,s_addr),tmp2 /* get either first or second word from source. */
combt,= tmp5,r0,skip_mask4 /* don't mask if whole word is valid*/
vshd tmp1,tmp2,tmp3 /* position data ! (delay slot)*/
sh3add tmp5,r0,save /* setup r1*/
mtctl save,r11 /* setup mask in save*/
zvdepi -2,32,save
or save, tmp3, tmp3
mtctl tmp4,11 /* re-load the shift count into cr11 */
skip_mask4:
uxor,nbz tmp3, r0, save
b,n null4 /* special case for first word */
copy r0, tmp5 /* zero out tmp5 so we don't try to mask again*/
copy r0, tmp7 /* zero out tmp7 so we don't try to use original s_addr anymore */
b continue
stbys,b,m tmp3,4(0,d_addr) /* store ! */
chunk2:
ldwm 4(0,s_addr),tmp2
vshd tmp1,tmp2,tmp3
skip_mask2:
uxor,nbz tmp3, r0, save
b,n null3
stbys,b,m tmp3,4(0,d_addr) /* store ! */
continue:
ldwm 4(0,s_addr),tmp1 /* get 2nd word ! */
vshd tmp2,tmp1,tmp3 /* position data ! */
uxor,nbz tmp3, r0, save
b,n null3
addibf,< -8,count,chunk2 /* If count is still >= 8 do another loop.*/
stbys,b,m tmp3,4(0,d_addr) /* store !*/
chkchnk2:
addibt,<,n 4,count,bp_0 /* if we don't have 4 bytes left then do the back porch (bp_0)*/
subchnk2: /* we have less than 8 chars to copy*/
ldwm 4(0,s_addr),tmp2 /* get next word !*/
combt,= tmp5,r0,skip_mask3
vshd tmp1,tmp2,tmp3 /* position data !*/
sh3add tmp5,r0,save /* setup r1*/
mtctl save,r11 /* setup mask in save*/
zvdepi -2,32,save
or save, tmp3, tmp3
mtctl tmp4,11 /* restore shift value again */
skip_mask3:
uxor,nbz tmp3,r0,save
b,n null3
copy r0,tmp5 /* zero out tmp5 so null3 does correct alignment */
copy r0,tmp7 /* zero out tmp7 so we don't use orignal s_addr since no longer valid */
b bp_1 /* we now have less than 4 bytes to move*/
stbys,b,m tmp3,4(0,d_addr) /* store !*/
bp_0:
copy tmp1,tmp2 /* switch registers for shift process */
addibt,<=,n 4,count,done /* if count = -4 this implies that count = 0 -> done */
bp_1:
ldwm 4(0,s_addr),tmp1 /* get final word ! */
vshd tmp2,tmp1,tmp3 /* position data !*/
uxor,nbz tmp3,r0,save /* if no-byte-zero */
b,n bp_null /* don't goto no_null-find which null instead */
no_null:
add d_addr,count,d_addr /* set up d_addr for stbys,e */
b done /* were done*/
stbys,e tmp3,0(0,d_addr) /* store the data !*/
/* here we do ye old byte-at-a-time moves.*/
align_null1:
b byteloop
addi -4,s_addr,s_addr
null1:
copy tmp6,s_addr /* restore orig s_addr (aligned only) */
byteloop:
addibt,= 4,count,done
null2:
ldbs,ma 1(s_addr),tmp1
encore:
combt,=,n tmp1,r0, done
stbs,ma tmp1,1(d_addr)
addibf,=,n -1,count,encore
ldbs,ma 1(s_addr),tmp1
b,n done
bp_null:
addi -4,count,count /* fudge count 'cause byteloop will re-increment */
null3: /* not_aligned case reset s_addr and finish byte-wise */
combt,=,n r0,tmp7,null3a /* if tmp7 is not valid address then branch below */
b byteloop /* otherwise reset s_addr to tmp7 and finish */
copy tmp7, s_addr
null3a: /* right shift target */
addibt,<,n 0,tmp6,null3b /* if left shifting */
sub r0,tmp6,tmp6 /* do null3b code */
addi -4,tmp6,tmp6
b byteloop
add tmp6,s_addr,s_addr /* reset s_addr by 4 + shift_amt */
null3b:
subi -8,tmp6,tmp6
add tmp5,tmp6,tmp6 /* adjust by the dest offset if this is our first store */
b byteloop
add tmp6,s_addr,s_addr /* adjust s_addr by (8-shift_amt-dest_off) */
null4:
add,> tmp6,r0,tmp6 /* if left shift */
b,n null3 /* then do null3 */
b byteloop
addi -4,s_addr,s_addr /* adj source only by 4 */
done:
bv 0(r2)
stbs r0,0(d_addr)
quit:
EXIT(strncat)