(CFLAGS-tst-align.c): Add -mpreferred-stack-boundary=4.
[glibc.git] / sysdeps / rs6000 / sub_n.s
blobd9318709353d3b76fb8501367c8992760eaa6812
1 # IBM POWER __mpn_sub_n -- Subtract two limb vectors of equal, non-zero length.
3 # Copyright (C) 1992, 1994, 1995, 1996 Free Software Foundation, Inc.
5 # This file is part of the GNU MP Library.
7 # The GNU MP Library is free software; you can redistribute it and/or modify
8 # it under the terms of the GNU Lesser General Public License as published by
9 # the Free Software Foundation; either version 2.1 of the License, or (at your
10 # option) any later version.
12 # The GNU MP Library is distributed in the hope that it will be useful, but
13 # WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
14 # or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public
15 # License for more details.
17 # You should have received a copy of the GNU Lesser General Public License
18 # along with the GNU MP Library; see the file COPYING.LIB. If not, write to
19 # the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
20 # MA 02111-1307, USA.
23 # INPUT PARAMETERS
24 # res_ptr r3
25 # s1_ptr r4
26 # s2_ptr r5
27 # size r6
29 .toc
30 .extern __mpn_sub_n[DS]
31 .extern .__mpn_sub_n
32 .csect [PR]
33 .align 2
34 .globl __mpn_sub_n
35 .globl .__mpn_sub_n
36 .csect __mpn_sub_n[DS]
37 __mpn_sub_n:
38 .long .__mpn_sub_n, TOC[tc0], 0
39 .csect [PR]
40 .__mpn_sub_n:
41 andil. 10,6,1 # odd or even number of limbs?
42 l 8,0(4) # load least significant s1 limb
43 l 0,0(5) # load least significant s2 limb
44 cal 3,-4(3) # offset res_ptr, it's updated before it's used
45 sri 10,6,1 # count for unrolled loop
46 sf 7,0,8 # subtract least significant limbs, set cy
47 mtctr 10 # copy count into CTR
48 beq 0,Leven # branch if even # of limbs (# of limbs >= 2)
50 # We have an odd # of limbs. Add the first limbs separately.
51 cmpi 1,10,0 # is count for unrolled loop zero?
52 bne 1,L1 # branch if not
53 st 7,4(3)
54 sfe 3,0,0 # load !cy into ...
55 sfi 3,3,0 # ... return value register
56 br # return
58 # We added least significant limbs. Now reload the next limbs to enter loop.
59 L1: lu 8,4(4) # load s1 limb and update s1_ptr
60 lu 0,4(5) # load s2 limb and update s2_ptr
61 stu 7,4(3)
62 sfe 7,0,8 # subtract limbs, set cy
63 Leven: lu 9,4(4) # load s1 limb and update s1_ptr
64 lu 10,4(5) # load s2 limb and update s2_ptr
65 bdz Lend # If done, skip loop
67 Loop: lu 8,4(4) # load s1 limb and update s1_ptr
68 lu 0,4(5) # load s2 limb and update s2_ptr
69 sfe 11,10,9 # subtract previous limbs with cy, set cy
70 stu 7,4(3) #
71 lu 9,4(4) # load s1 limb and update s1_ptr
72 lu 10,4(5) # load s2 limb and update s2_ptr
73 sfe 7,0,8 # subtract previous limbs with cy, set cy
74 stu 11,4(3) #
75 bdn Loop # decrement CTR and loop back
77 Lend: sfe 11,10,9 # subtract limbs with cy, set cy
78 st 7,4(3) #
79 st 11,8(3) #
80 sfe 3,0,0 # load !cy into ...
81 sfi 3,3,0 # ... return value register