2 # Alpha 21064 __mpn_mul_1 -- Multiply a limb vector with a limb and store
3 # the result in a second limb vector.
5 # Copyright (C) 1992, 1994, 1995 Free Software Foundation, Inc.
7 # This file is part of the GNU MP Library.
9 # The GNU MP Library is free software; you can redistribute it and/or modify
10 # it under the terms of the GNU Library General Public License as published by
11 # the Free Software Foundation; either version 2 of the License, or (at your
12 # option) any later version.
14 # The GNU MP Library is distributed in the hope that it will be useful, but
15 # WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
16 # or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public
17 # License for more details.
19 # You should have received a copy of the GNU Library General Public License
20 # along with the GNU MP Library; see the file COPYING.LIB. If not, write to
21 # the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
31 # This code runs at 42 cycles/limb on the EV4 and 18 cycles/limb on the EV5.
33 # To improve performance for long multiplications, we would use
34 # 'fetch' for S1 and 'fetch_m' for RES. It's not obvious how to use
35 # these instructions without slowing down the general code: 1. We can
36 # only have two prefetches in operation at any time in the Alpha
37 # architecture. 2. There will seldom be any special alignment
38 # between RES_PTR and S1_PTR. Maybe we can simply divide the current
39 # loop into an inner and outer loop, having the inner loop handle
40 # exactly one prefetch block?
52 ldq $2,0($17) # $2 = s1_limb
53 subq $18,1,$18 # size--
54 mulq $2,$19,$3 # $3 = prod_low
55 bic $31,$31,$4 # clear cy_limb
56 umulh $2,$19,$0 # $0 = prod_high
57 beq $18,Lend1 # jump if size was == 1
58 ldq $2,8($17) # $2 = s1_limb
59 subq $18,1,$18 # size--
61 beq $18,Lend2 # jump if size was == 2
64 Loop: mulq $2,$19,$3 # $3 = prod_low
65 addq $4,$0,$0 # cy_limb = cy_limb + 'cy'
66 subq $18,1,$18 # size--
67 umulh $2,$19,$4 # $4 = cy_limb
68 ldq $2,16($17) # $2 = s1_limb
69 addq $17,8,$17 # s1_ptr++
70 addq $3,$0,$3 # $3 = cy_limb + prod_low
72 cmpult $3,$0,$0 # $0 = carry from (cy_limb + prod_low)
73 addq $16,8,$16 # res_ptr++
76 Lend2: mulq $2,$19,$3 # $3 = prod_low
77 addq $4,$0,$0 # cy_limb = cy_limb + 'cy'
78 umulh $2,$19,$4 # $4 = cy_limb
79 addq $3,$0,$3 # $3 = cy_limb + prod_low
80 cmpult $3,$0,$0 # $0 = carry from (cy_limb + prod_low)
82 addq $4,$0,$0 # cy_limb = prod_high + cy