![[BACK]](/icons/cvsweb/back.gif){kind=icon}
Return to submul_1.S CVS log ![[TXT]](/icons/cvsweb/text.gif){kind=icon}
![[DIR]](/icons/cvsweb/dir.gif){kind=icon}
Up to [local] / OpenXM_contrib / gmp / mpn / pa64w|
Return to submul_1.S CVS log | Up to [local] / OpenXM_contrib / gmp / mpn / pa64w |
1.1 maekawa 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.
3:
4: ; Copyright (C) 1998, 1999, 2000 Free Software Foundation, Inc.
5:
6: ; This file is part of the GNU MP Library.
7:
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.
12:
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.
17:
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,
21: ; MA 02111-1307, USA.
22:
23: ; INPUT PARAMETERS
24: #define rptr %r26
25: #define sptr %r25
26: #define size %r24
27: #define s2limb %r23
28:
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
32:
33: #define t1 %r19
34: #define rlimb %r20
35: #define hi %r21
36: #define lo %r22
37: #define m0 %r28
38: #define m1 %r3
39: #define cylimb %r29
40: #define t3 %r4
41: #define t2 %r6
42: #define t5 %r23
43: #define t4 %r31
44: .level 2.0w
45: .code
46: .export __gmpn_submul_1,entry
47: __gmpn_submul_1
48: .proc
49: .callinfo frame=128,no_calls
50: .entry
51: std s2limb,-56(%r30)
52: fldd -56(%r30),%fr5
53: ldo 128(%r30),%r30
54: add %r0,%r0,cylimb ; clear cy and cylimb
55:
56: std %r3,-96(%r30)
57: std %r4,-88(%r30)
58: std %r5,-80(%r30)
59: std %r6,-72(%r30)
60: depdi,z 1,31,1,%r5
61:
62: fldd 0(sptr),%fr4
63: ldo 8(sptr),sptr
64:
65: xmpyu %fr5R,%fr4R,%fr6
66: fstd %fr6,-128(%r30)
67: xmpyu %fr5R,%fr4L,%fr7
68: fstd %fr7,-120(%r30)
69: xmpyu %fr5L,%fr4R,%fr8
70: fstd %fr8,-112(%r30)
71: xmpyu %fr5L,%fr4L,%fr9
72: fstd %fr9,-104(%r30)
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$end1
78: nop
79: fldd 0(sptr),%fr4
80: ldo 8(sptr),sptr
81: addib,= -1,%r24,L$end2
82: nop
83: L$loop
84: xmpyu %fr5R,%fr4R,%fr6
85: fstd %fr6,-128(%r30)
86: xmpyu %fr5R,%fr4L,%fr7
87: fstd %fr7,-120(%r30)
88: xmpyu %fr5L,%fr4R,%fr8
89: fstd %fr8,-112(%r30)
90: xmpyu %fr5L,%fr4L,%fr9
91: fstd %fr9,-104(%r30)
92: ldd 0(rptr),rlimb
93: extrd,u lo,31,32,t1 ; t1 = hi32(lo)
94: extrd,u lo,63,32,t4 ; t4 = lo32(lo)
95: add,l m0,t1,t1 ; t1 += m0
96: add,l,*nuv m1,t1,t1 ; t1 += m1
97: add,l %r5,hi,hi ; propagate carry
98: extrd,u t1,31,32,t2 ; t2 = hi32(t1)
99: depd,z t1,31,32,t5 ; t5 = lo32(t1)
100: add,l t5,t4,t4 ; t4 += lo32(t1)
101: ldd -128(%r30),lo ; lo = low 64 bit of product
102: add cylimb,t4,t4
103: ldd -120(%r30),m0 ; m0 = mid0 64 bit of product
104: add,dc t2,hi,cylimb
105: ldd -112(%r30),m1 ; m1 = mid1 64 bit of product
106: sub rlimb,t4,t3
107: add t4,t3,%r0
108: ldd -104(%r30),hi ; hi = high 64 bit of product
109: add,dc %r0,cylimb,cylimb
110: fldd 0(sptr),%fr4
111: ldo 8(sptr),sptr
112: std t3,0(rptr)
113: addib,<> -1,%r24,L$loop
114: ldo 8(rptr),rptr
115: L$end2
116: xmpyu %fr5R,%fr4R,%fr6
117: fstd %fr6,-128(%r30)
118: xmpyu %fr5R,%fr4L,%fr7
119: fstd %fr7,-120(%r30)
120: xmpyu %fr5L,%fr4R,%fr8
121: fstd %fr8,-112(%r30)
122: xmpyu %fr5L,%fr4L,%fr9
123: fstd %fr9,-104(%r30)
124: ldd 0(rptr),rlimb
125: extrd,u lo,31,32,t1 ; t1 = hi32(lo)
126: extrd,u lo,63,32,t4 ; t4 = lo32(lo)
127: add,l m0,t1,t1 ; t1 += m0
128: add,l,*nuv m1,t1,t1 ; t1 += m0
129: add,l %r5,hi,hi ; propagate carry
130: extrd,u t1,31,32,t2 ; t2 = hi32(t1)
131: depd,z t1,31,32,t5 ; t5 = lo32(t1)
132: add,l t5,t4,t4 ; t4 += lo32(t1)
133: ldd -128(%r30),lo ; lo = low 64 bit of product
134: add cylimb,t4,t4
135: ldd -120(%r30),m0 ; m0 = mid0 64 bit of product
136: add,dc t2,hi,cylimb
137: ldd -112(%r30),m1 ; m1 = mid1 64 bit of product
138: sub rlimb,t4,t3
139: add t4,t3,%r0
140: ldd -104(%r30),hi ; hi = high 64 bit of product
141: add,dc %r0,cylimb,cylimb
142: std t3,0(rptr)
143: ldo 8(rptr),rptr
144: L$end1
145: ldd 0(rptr),rlimb
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)
154: add cylimb,t4,t4
155: add,dc t2,hi,cylimb
156: sub rlimb,t4,t3
157: add t4,t3,%r0
158: add,dc %r0,cylimb,cylimb
159: std t3,0(rptr)
160: ldo 8(rptr),rptr
161:
162: ldd -96(%r30),%r3
163: ldd -88(%r30),%r4
164: ldd -80(%r30),%r5
165: ldd -72(%r30),%r6
166:
167: copy cylimb,%r28
168: bve (%r2)
169: .exit
170: ldo -128(%r30),%r30
171: .procend