1 ; HP-PA 2.0 64-bit __gmpn_submul_1 -- Multiply a limb vector with a limb and
2 ; subtract the result from a second limb vector.
4 ; Copyright (C) 1998, 1999, 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,
29 ; This runs at 11 cycles/limb on a PA8000. It might be possible to make
30 ; it faster, but the PA8000 pipeline is not publically documented and it
31 ; is very complex to reverse engineer
46 .export __gmpn_submul_1,entry
49 .callinfo frame=128,no_calls
54 add %r0,%r0,cylimb ; clear cy and cylimb
65 xmpyu %fr5R,%fr4R,%fr6
67 xmpyu %fr5R,%fr4L,%fr7
69 xmpyu %fr5L,%fr4R,%fr8
71 xmpyu %fr5L,%fr4L,%fr9
73 ldd -128(%r30),lo ; lo = low 64 bit of product
74 ldd -120(%r30),m0 ; m0 = mid0 64 bit of product
75 ldd -112(%r30),m1 ; m1 = mid1 64 bit of product
76 ldd -104(%r30),hi ; hi = high 64 bit of product
77 addib,= -1,%r24,L$end1
81 addib,= -1,%r24,L$end2
84 xmpyu %fr5R,%fr4R,%fr6
86 xmpyu %fr5R,%fr4L,%fr7
88 xmpyu %fr5L,%fr4R,%fr8
90 xmpyu %fr5L,%fr4L,%fr9
93 extrd,u lo,31,32,t1 ; t1 = hi32(lo)
94 extrd,u lo,63,32,t4 ; t4 = lo32(lo)
95 add,l m0,t1,t1 ; t1 += m0
96 add,l,*nuv m1,t1,t1 ; t1 += m1
97 add,l %r5,hi,hi ; propagate carry
98 extrd,u t1,31,32,t2 ; t2 = hi32(t1)
99 depd,z t1,31,32,t5 ; t5 = lo32(t1)
100 add,l t5,t4,t4 ; t4 += lo32(t1)
101 ldd -128(%r30),lo ; lo = low 64 bit of product
103 ldd -120(%r30),m0 ; m0 = mid0 64 bit of product
105 ldd -112(%r30),m1 ; m1 = mid1 64 bit of product
108 ldd -104(%r30),hi ; hi = high 64 bit of product
109 add,dc %r0,cylimb,cylimb
113 addib,<> -1,%r24,L$loop
116 xmpyu %fr5R,%fr4R,%fr6
118 xmpyu %fr5R,%fr4L,%fr7
120 xmpyu %fr5L,%fr4R,%fr8
122 xmpyu %fr5L,%fr4L,%fr9
125 extrd,u lo,31,32,t1 ; t1 = hi32(lo)
126 extrd,u lo,63,32,t4 ; t4 = lo32(lo)
127 add,l m0,t1,t1 ; t1 += m0
128 add,l,*nuv m1,t1,t1 ; t1 += m0
129 add,l %r5,hi,hi ; propagate carry
130 extrd,u t1,31,32,t2 ; t2 = hi32(t1)
131 depd,z t1,31,32,t5 ; t5 = lo32(t1)
132 add,l t5,t4,t4 ; t4 += lo32(t1)
133 ldd -128(%r30),lo ; lo = low 64 bit of product
135 ldd -120(%r30),m0 ; m0 = mid0 64 bit of product
137 ldd -112(%r30),m1 ; m1 = mid1 64 bit of product
140 ldd -104(%r30),hi ; hi = high 64 bit of product
141 add,dc %r0,cylimb,cylimb
146 extrd,u lo,31,32,t1 ; t1 = hi32(lo)
147 extrd,u lo,63,32,t4 ; t4 = lo32(lo)
148 add,l m0,t1,t1 ; t1 += m0
149 add,l,*nuv m1,t1,t1 ; t1 += m0
150 add,l %r5,hi,hi ; propagate carry
151 extrd,u t1,31,32,t2 ; t2 = hi32(t1)
152 depd,z t1,31,32,t5 ; t5 = lo32(t1)
153 add,l t5,t4,t4 ; t4 += lo32(t1)
158 add,dc %r0,cylimb,cylimb