Annotation of OpenXM_contrib/gmp/mpn/x86/k7/diveby3.asm, Revision 1.1.1.1
1.1 maekawa 1: dnl AMD K7 mpn_divexact_by3 -- mpn division by 3, expecting no remainder.
2: dnl
3: dnl K7: 8.0 cycles/limb
4:
5:
6: dnl Copyright (C) 2000 Free Software Foundation, Inc.
7: dnl
8: dnl This file is part of the GNU MP Library.
9: dnl
10: dnl The GNU MP Library is free software; you can redistribute it and/or
11: dnl modify it under the terms of the GNU Lesser General Public License as
12: dnl published by the Free Software Foundation; either version 2.1 of the
13: dnl License, or (at your option) any later version.
14: dnl
15: dnl The GNU MP Library is distributed in the hope that it will be useful,
16: dnl but WITHOUT ANY WARRANTY; without even the implied warranty of
17: dnl MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18: dnl Lesser General Public License for more details.
19: dnl
20: dnl You should have received a copy of the GNU Lesser General Public
21: dnl License along with the GNU MP Library; see the file COPYING.LIB. If
22: dnl not, write to the Free Software Foundation, Inc., 59 Temple Place -
23: dnl Suite 330, Boston, MA 02111-1307, USA.
24:
25:
26: include(`../config.m4')
27:
28:
29: C mp_limb_t mpn_divexact_by3c (mp_ptr dst, mp_srcptr src, mp_size_t size,
30: C mp_limb_t carry);
31:
32: defframe(PARAM_CARRY,16)
33: defframe(PARAM_SIZE, 12)
34: defframe(PARAM_SRC, 8)
35: defframe(PARAM_DST, 4)
36:
37: dnl multiplicative inverse of 3, modulo 2^32
38: deflit(INVERSE_3, 0xAAAAAAAB)
39:
40: dnl ceil(b/3) and floor(b*2/3) where b=2^32
41: deflit(ONE_THIRD_CEIL, 0x55555556)
42: deflit(TWO_THIRDS_FLOOR, 0xAAAAAAAA)
43:
44: .text
45: ALIGN(32)
46:
47: PROLOGUE(mpn_divexact_by3c)
48: deflit(`FRAME',0)
49:
50: movl PARAM_SRC, %ecx
51: pushl %ebx defframe_pushl(SAVE_EBX)
52:
53: movl PARAM_CARRY, %ebx
54: pushl %ebp defframe_pushl(SAVE_EBP)
55:
56: movl PARAM_SIZE, %ebp
57: pushl %edi defframe_pushl(SAVE_EDI)
58:
59: movl (%ecx), %eax C src low limb
60: pushl %esi defframe_pushl(SAVE_ESI)
61:
62: movl PARAM_DST, %edi
63: movl $TWO_THIRDS_FLOOR, %esi
64: leal -4(%ecx,%ebp,4), %ecx C &src[size-1]
65:
66: subl %ebx, %eax
67:
68: setc %bl
69: decl %ebp
70: jz L(last)
71:
72: leal (%edi,%ebp,4), %edi C &dst[size-1]
73: negl %ebp
74:
75:
76: ALIGN(16)
77: L(top):
78: C eax src limb, carry subtracted
79: C ebx carry limb (0 or 1)
80: C ecx &src[size-1]
81: C edx scratch
82: C esi TWO_THIRDS_FLOOR
83: C edi &dst[size-1]
84: C ebp counter, limbs, negative
85:
86: imull $INVERSE_3, %eax, %edx
87:
88: movl 4(%ecx,%ebp,4), %eax C next src limb
89: cmpl $ONE_THIRD_CEIL, %edx
90:
91: sbbl $-1, %ebx C +1 if result>=ceil(b/3)
92: cmpl %edx, %esi
93:
94: sbbl %ebx, %eax C and further 1 if result>=ceil(b*2/3)
95: movl %edx, (%edi,%ebp,4)
96: incl %ebp
97:
98: setc %bl C new carry
99: jnz L(top)
100:
101:
102:
103: L(last):
104: C eax src limb, carry subtracted
105: C ebx carry limb (0 or 1)
106: C ecx &src[size-1]
107: C edx scratch
108: C esi multiplier
109: C edi &dst[size-1]
110: C ebp
111:
112: imull $INVERSE_3, %eax
113:
114: cmpl $ONE_THIRD_CEIL, %eax
115: movl %eax, (%edi)
116: movl SAVE_EBP, %ebp
117:
118: sbbl $-1, %ebx C +1 if eax>=ceil(b/3)
119: cmpl %eax, %esi
120: movl $0, %eax
121:
122: adcl %ebx, %eax C further +1 if eax>=ceil(b*2/3)
123: movl SAVE_EDI, %edi
124: movl SAVE_ESI, %esi
125:
126: movl SAVE_EBX, %ebx
127: addl $FRAME, %esp
128:
129: ret
130:
131: EPILOGUE()
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>