1
/* Alpha 21064 mpih-mul1.S -- Multiply a limb vector with a limb and store
2
* the result in a second limb vector.
4
* Copyright (C) 1992, 1994, 1995, 1998,
5
2001 Free Software Foundation, Inc.
7
* This file is part of GnuPG.
9
* GnuPG is free software; you can redistribute it and/or modify
10
* it under the terms of the GNU General Public License as published by
11
* the Free Software Foundation; either version 2 of the License, or
12
* (at your option) any later version.
14
* GnuPG is distributed in the hope that it will be useful,
15
* but WITHOUT ANY WARRANTY; without even the implied warranty of
16
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17
* GNU General Public License for more details.
19
* You should have received a copy of the GNU General Public License
20
* along with this program; if not, write to the Free Software
21
* Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA
27
* mpihelp_mul_1( mpi_ptr_t res_ptr, (r16)
28
* mpi_ptr_t s1_ptr, (r17)
29
* mpi_size_t s1_size, (r18)
30
* mpi_limb_t s2_limb) (r19)
32
* This code runs at 42 cycles/limb on the EV4 and 18 cycles/limb on the EV5.
34
* To improve performance for long multiplications, we would use
35
* 'fetch' for S1 and 'fetch_m' for RES. It's not obvious how to use
36
* these instructions without slowing down the general code: 1. We can
37
* only have two prefetches in operation at any time in the Alpha
38
* architecture. 2. There will seldom be any special alignment
39
* between RES_PTR and S1_PTR. Maybe we can simply divide the current
40
* loop into an inner and outer loop, having the inner loop handle
41
* exactly one prefetch block?
53
ldq $2,0($17) # $2 = s1_limb
54
subq $18,1,$18 # size--
55
mulq $2,$19,$3 # $3 = prod_low
56
bic $31,$31,$4 # clear cy_limb
57
umulh $2,$19,$0 # $0 = prod_high
58
beq $18,Lend1 # jump if size was == 1
59
ldq $2,8($17) # $2 = s1_limb
60
subq $18,1,$18 # size--
62
beq $18,Lend2 # jump if size was == 2
65
Loop: mulq $2,$19,$3 # $3 = prod_low
66
addq $4,$0,$0 # cy_limb = cy_limb + 'cy'
67
subq $18,1,$18 # size--
68
umulh $2,$19,$4 # $4 = cy_limb
69
ldq $2,16($17) # $2 = s1_limb
70
addq $17,8,$17 # s1_ptr++
71
addq $3,$0,$3 # $3 = cy_limb + prod_low
73
cmpult $3,$0,$0 # $0 = carry from (cy_limb + prod_low)
74
addq $16,8,$16 # res_ptr++
77
Lend2: mulq $2,$19,$3 # $3 = prod_low
78
addq $4,$0,$0 # cy_limb = cy_limb + 'cy'
79
umulh $2,$19,$4 # $4 = cy_limb
80
addq $3,$0,$3 # $3 = cy_limb + prod_low
81
cmpult $3,$0,$0 # $0 = carry from (cy_limb + prod_low)
83
addq $4,$0,$0 # cy_limb = prod_high + cy