2005-07-08 Carlos O'Donell <carlos@systemhalted.org>
[glibc-ports.git] / sysdeps / m88k / m88100 / add_n.s
blob0741ec91de9986f7d36a021c7b69365e6442ec20
1 ; mc88100 __mpn_add -- Add two limb vectors of the same length > 0 and store
2 ; sum in a third limb vector.
4 ; Copyright (C) 1992, 1994 Free Software Foundation, Inc.
6 ; This file is part of the GNU MP Library.
8 ; The GNU MP Library is free software; you can redistribute it and/or modify
9 ; it under the terms of the GNU Lesser General Public License as published by
10 ; the Free Software Foundation; either version 2.1 of the License, or (at your
11 ; option) any later version.
13 ; The GNU MP Library is distributed in the hope that it will be useful, but
14 ; WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
15 ; or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public
16 ; License for more details.
18 ; You should have received a copy of the GNU Lesser General Public License
19 ; along with the GNU MP Library; see the file COPYING.LIB. If not, write to
20 ; the Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
23 ; INPUT PARAMETERS
24 ; res_ptr r2
25 ; s1_ptr r3
26 ; s2_ptr r4
27 ; size r5
29 ; This code has been optimized to run one instruction per clock, avoiding
30 ; load stalls and writeback contention. As a result, the instruction
31 ; order is not always natural.
33 ; The speed is about 4.6 clocks/limb + 18 clocks/limb-vector on an 88100,
34 ; but on the 88110, it seems to run much slower, 6.6 clocks/limb.
36 text
37 align 16
38 global ___mpn_add_n
39 ___mpn_add_n:
40 ld r6,r3,0 ; read first limb from s1_ptr
41 extu r10,r5,3
42 ld r7,r4,0 ; read first limb from s2_ptr
44 subu.co r5,r0,r5 ; (clear carry as side effect)
45 mak r5,r5,3<4>
46 bcnd eq0,r5,Lzero
48 or r12,r0,lo16(Lbase)
49 or.u r12,r12,hi16(Lbase)
50 addu r12,r12,r5 ; r12 is address for entering in loop
52 extu r5,r5,2 ; divide by 4
53 subu r2,r2,r5 ; adjust res_ptr
54 subu r3,r3,r5 ; adjust s1_ptr
55 subu r4,r4,r5 ; adjust s2_ptr
57 or r8,r6,r0
59 jmp.n r12
60 or r9,r7,r0
62 Loop: addu r3,r3,32
63 st r8,r2,28
64 addu r4,r4,32
65 ld r6,r3,0
66 addu r2,r2,32
67 ld r7,r4,0
68 Lzero: subu r10,r10,1 ; add 0 + 8r limbs (adj loop cnt)
69 Lbase: ld r8,r3,4
70 addu.cio r6,r6,r7
71 ld r9,r4,4
72 st r6,r2,0
73 ld r6,r3,8 ; add 7 + 8r limbs
74 addu.cio r8,r8,r9
75 ld r7,r4,8
76 st r8,r2,4
77 ld r8,r3,12 ; add 6 + 8r limbs
78 addu.cio r6,r6,r7
79 ld r9,r4,12
80 st r6,r2,8
81 ld r6,r3,16 ; add 5 + 8r limbs
82 addu.cio r8,r8,r9
83 ld r7,r4,16
84 st r8,r2,12
85 ld r8,r3,20 ; add 4 + 8r limbs
86 addu.cio r6,r6,r7
87 ld r9,r4,20
88 st r6,r2,16
89 ld r6,r3,24 ; add 3 + 8r limbs
90 addu.cio r8,r8,r9
91 ld r7,r4,24
92 st r8,r2,20
93 ld r8,r3,28 ; add 2 + 8r limbs
94 addu.cio r6,r6,r7
95 ld r9,r4,28
96 st r6,r2,24
97 bcnd.n ne0,r10,Loop ; add 1 + 8r limbs
98 addu.cio r8,r8,r9
100 st r8,r2,28 ; store most significant limb
102 jmp.n r1
103 addu.ci r2,r0,r0 ; return carry-out from most sign. limb