1 ; mc88100 __gmpn_sub -- Subtract two limb vectors of the same length > 0 and
2 ; store difference in a third limb vector.
4 ; Copyright (C) 1992, 1994, 1996, 2000 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., 59 Temple Place - Suite 330, Boston,
30 ; This code has been optimized to run one instruction per clock, avoiding
31 ; load stalls and writeback contention. As a result, the instruction
32 ; order is not always natural.
34 ; The speed is about 4.6 clocks/limb + 18 clocks/limb-vector on an 88100,
35 ; but on the 88110, it seems to run much slower, 6.6 clocks/limb.
41 ld r6,r3,0 ; read first limb from s1_ptr
43 ld r7,r4,0 ; read first limb from s2_ptr
48 subu.co r0,r0,r0 ; initialize carry
51 or.u r12,r12,hi16(Lbase)
52 addu r12,r12,r5 ; r12 is address for entering in loop
54 extu r5,r5,2 ; divide by 4
55 subu r2,r2,r5 ; adjust res_ptr
56 subu r3,r3,r5 ; adjust s1_ptr
57 subu r4,r4,r5 ; adjust s2_ptr
70 Lzero: subu r10,r10,1 ; subtract 0 + 8r limbs (adj loop cnt)
75 ld r6,r3,8 ; subtract 7 + 8r limbs
79 ld r8,r3,12 ; subtract 6 + 8r limbs
83 ld r6,r3,16 ; subtract 5 + 8r limbs
87 ld r8,r3,20 ; subtract 4 + 8r limbs
91 ld r6,r3,24 ; subtract 3 + 8r limbs
95 ld r8,r3,28 ; subtract 2 + 8r limbs
99 bcnd.n ne0,r10,Loop ; subtract 1 + 8r limbs
102 st r8,r2,28 ; store most significant limb
104 addu.ci r2,r0,r0 ; return carry-out from most sign. limb