LMMP 4.3.0
Lamina High-Precision Arithmetic Library
载入中...
搜索中...
未找到
mul_toom43.c 文件参考
#include "impl/mul_hard.h"
#include "mparam.h"
#include "lmmpn/toom_interp.h"
mul_toom43.c 的引用(Include)关系图:

浏览该文件的源代码.

宏定义

#define a0   numa
#define a0a2   scratch
#define a1   (numa + n)
#define a1a3   asm1
#define a2   (numa + 2 * n)
#define a3   (numa + 3 * n)
#define as1   (dst + 4 * n + 4)
#define as2   (dst + 3 * n + 3)
#define asm1   (scratch + 3 * n + 3)
#define asm2   (scratch + 4 * n + 4)
#define b0   numb
#define b0b2   scratch
#define b1   (numb + n)
#define b1d   bsm1
#define b2   (numb + 2 * n)
#define bs1   dst
#define bs2   (dst + 2 * n + 2)
#define bsm1   (scratch + 2 * n + 2)
#define bsm2   (dst + n + 1)
#define lmmp_mul_n_(dst, numa, numb, n)
 Copyright (C) 2026 HJimmyK (Jericho Knox)
#define v0   dst
#define v1   (dst + 2 * n)
#define v2   (scratch + 4 * n + 2)
#define vinf   (dst + 5 * n)
#define vm1   (scratch)
#define vm2   (scratch + 2 * n + 1)

函数

void lmmp_mul_toom43_ (mp_ptr restrict dst, mp_srcptr restrict numa, mp_size_t na, mp_srcptr restrict numb, mp_size_t nb)

宏定义说明

◆ a0

#define a0   numa

◆ a0a2

#define a0a2   scratch

被这些函数引用 lmmp_mul_toom43_() , 以及 lmmp_mul_toom52_().

◆ a1

#define a1   (numa + n)

◆ a1a3

#define a1a3   asm1

被这些函数引用 lmmp_mul_toom43_() , 以及 lmmp_mul_toom52_().

◆ a2

#define a2   (numa + 2 * n)

◆ a3

#define a3   (numa + 3 * n)

◆ as1

◆ as2

◆ asm1

#define asm1   (scratch + 3 * n + 3)

◆ asm2

◆ b0

#define b0   numb

◆ b0b2

#define b0b2   scratch

被这些函数引用 lmmp_mul_toom43_().

◆ b1

#define b1   (numb + n)

◆ b1d

#define b1d   bsm1

被这些函数引用 lmmp_mul_toom43_().

◆ b2

#define b2   (numb + 2 * n)

◆ bs1

◆ bs2

◆ bsm1

#define bsm1   (scratch + 2 * n + 2)

◆ bsm2

◆ lmmp_mul_n_

#define lmmp_mul_n_ ( dst,
numa,
numb,
n )
值:
lmmp_mul_hard_n_((dst), (numa), (numb), (n)); \
else if ((n) < MUL_TOOM33_THRESHOLD) \
lmmp_mul_toom22_((dst), (numa), (n), (numb), (n)); \
else if ((n) < MUL_TOOM44_THRESHOLD) \
lmmp_mul_toom33_((dst), (numa), (n), (numb), (n)); \
else \
lmmp_mul_toom44_((dst), (numa), (n), (numb), (n))
#define MUL_TOOM22_THRESHOLD
#define MUL_TOOM33_THRESHOLD
#define MUL_TOOM44_THRESHOLD
#define numb
#define n

Copyright (C) 2026 HJimmyK (Jericho Knox)

This file is part of LMMP.

LMMP is free software: you can redistribute it and/or modify it under the terms of the GNU Lesser General Public License as published by the Free Software Foundation; either version 3 of the License, or (at your option) any later version.

This library is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details.

You should have received a copy of the GNU Lesser General Public License along with this library. If not, see https://www.gnu.org/licenses/.

SPDX-License-Identifier: LGPL-3.0-or-later

在文件 mul_toom43.c 第 27 行定义.

◆ v0

#define v0   dst

◆ v1

#define v1   (dst + 2 * n)

◆ v2

#define v2   (scratch + 4 * n + 2)

◆ vinf

#define vinf   (dst + 5 * n)

◆ vm1

#define vm1   (scratch)

◆ vm2

函数说明

◆ lmmp_mul_toom43_()

void lmmp_mul_toom43_ ( mp_ptr restrict dst,
mp_srcptr restrict numa,
mp_size_t na,
mp_srcptr restrict numb,
mp_size_t nb )

在文件 mul_toom43.c 第 53 行定义.

53 {
54 lmmp_param_assert(2 * nb > na);
55 lmmp_param_assert(na >= nb);
56 mp_size_t n, s, t;
57 enum toom6_flags flags;
58 mp_limb_t cy;
59
60#define a0 numa
61#define a1 (numa + n)
62#define a2 (numa + 2 * n)
63#define a3 (numa + 3 * n)
64#define b0 numb
65#define b1 (numb + n)
66#define b2 (numb + 2 * n)
67
68 n = 1 + (3 * na >= 4 * nb ? (na - 1) >> 2 : (nb - 1) / (mp_size_t)3);
71
72 s = na - 3 * n;
73 t = nb - 2 * n;
74
75 lmmp_param_assert(0 < s && s <= n);
76 lmmp_param_assert(0 < t && t <= n);
77
78 /* This is true whenever na >= 25 or nb >= 19, I think. It
79 guarantees that we can fit 5 values of size n+1 in the product
80 area. */
81 lmmp_param_assert(s + t >= 5);
82
83#define v0 dst /* 2n */
84#define vm1 (scratch) /* 2n+1 */
85#define v1 (dst + 2 * n) /* 2n+1 */
86#define vm2 (scratch + 2 * n + 1) /* 2n+1 */
87#define v2 (scratch + 4 * n + 2) /* 2n+1 */
88#define vinf (dst + 5 * n) /* s+t */
89#define bs1 dst /* n+1 */
90#define bsm1 (scratch + 2 * n + 2) /* n+1 */
91#define asm1 (scratch + 3 * n + 3) /* n+1 */
92#define asm2 (scratch + 4 * n + 4) /* n+1 */
93#define bsm2 (dst + n + 1) /* n+1 */
94#define bs2 (dst + 2 * n + 2) /* n+1 */
95#define as2 (dst + 3 * n + 3) /* n+1 */
96#define as1 (dst + 4 * n + 4) /* n+1 */
97
98 /* Total sccratch need is 6 * n + 3 + 1; we allocate one extra
99 limb, because products will overwrite 2n+2 limbs. */
100
101#define a0a2 scratch
102#define b0b2 scratch
103#define a1a3 asm1
104#define b1d bsm1
105
106 /* Compute as2 and asm2. */
107 flags = (enum toom6_flags)(toom6_vm2_neg & lmmp_toom_eval_dgr3_pm2_(as2, asm2, numa, n, s, a1a3));
108
109 /* Compute bs2 and bsm2. */
110 b1d[n] = lmmp_shl_(b1d, b1, n, 1); /* 2b1 */
111 cy = lmmp_shl_(b0b2, b2, t, 2); /* 4b2 */
112 cy += lmmp_add_n_(b0b2, b0b2, b0, t); /* 4b2 + b0 */
113 if (t != n)
114 cy = lmmp_add_1_(b0b2 + t, b0 + t, n - t, cy);
115 b0b2[n] = cy;
116
117 if (lmmp_cmp_(b0b2, b1d, n + 1) < 0) {
119 flags = (enum toom6_flags)(flags ^ toom6_vm2_neg);
120 } else {
122 }
123
124 /* Compute as1 and asm1. */
125 flags = (enum toom6_flags)(flags ^ (toom6_vm1_neg & lmmp_toom_eval_dgr3_pm1_(as1, asm1, numa, n, s, a0a2)));
126
127 /* Compute bs1 and bsm1. */
128 bsm1[n] = lmmp_add_(bsm1, b0, n, b2, t);
129 if (bsm1[n] == 0 && lmmp_cmp_(bsm1, b1, n) < 0) {
130 cy = lmmp_add_n_sub_n_(bs1, bsm1, b1, bsm1, n);
131 bs1[n] = cy >> 1;
132 flags = (enum toom6_flags)(flags ^ toom6_vm1_neg);
133 } else {
134 cy = lmmp_add_n_sub_n_(bs1, bsm1, bsm1, b1, n);
135 bs1[n] = bsm1[n] + (cy >> 1);
136 bsm1[n] -= cy & 1;
137 }
138
139 lmmp_debug_assert(as1[n] <= 3);
140 lmmp_debug_assert(bs1[n] <= 2);
141 lmmp_debug_assert(asm1[n] <= 1);
142 lmmp_debug_assert(bsm1[n] <= 1);
143 lmmp_debug_assert(as2[n] <= 14);
144 lmmp_debug_assert(bs2[n] <= 6);
145 lmmp_debug_assert(asm2[n] <= 9);
146 lmmp_debug_assert(bsm2[n] <= 4);
147
148 /* vm1, 2n+1 limbs */
149 lmmp_mul_n_(vm1, asm1, bsm1, n + 1); /* W4 */
150
151 /* vm2, 2n+1 limbs */
152 lmmp_mul_n_(vm2, asm2, bsm2, n + 1); /* W2 */
153
154 /* v2, 2n+1 limbs */
155 lmmp_mul_n_(v2, as2, bs2, n + 1); /* W1 */
156
157 /* v1, 2n+1 limbs */
158 lmmp_mul_n_(v1, as1, bs1, n + 1); /* W3 */
159
160 /* vinf, s+t limbs */ /* W0 */
161 if (s > t)
162 lmmp_mul_(vinf, a3, s, b2, t);
163 else
164 lmmp_mul_(vinf, b2, t, a3, s);
165
166 /* v0, 2n limbs */
167 lmmp_mul_n_(v0, numa, numb, n); /* W5 */
168
169 lmmp_toom_interp6_(dst, n, flags, vm1, vm2, v2, t + s);
170
172#undef v0
173#undef vm1
174#undef v1
175#undef vm2
176#undef v2
177#undef vinf
178#undef bs1
179#undef bs2
180#undef bsm1
181#undef bsm2
182#undef asm1
183#undef asm2
184#undef as1
185#undef as2
186#undef a0a2
187#undef b0b2
188#undef a1a3
189#undef b1d
190#undef a0
191#undef a1
192#undef a2
193#undef a3
194#undef b0
195#undef b1
196#undef b2
197}
#define lmmp_mul_n_
uint64_t mp_size_t
定义 lmmp.h:93
#define lmmp_debug_assert(x)
定义 lmmp.h:498
uint64_t mp_limb_t
定义 lmmp.h:92
#define lmmp_param_assert(x)
定义 lmmp.h:509
static mp_limb_t lmmp_add_(mp_ptr dst, mp_srcptr numa, mp_size_t na, mp_srcptr numb, mp_size_t nb)
加法静态内联函数 [dst,na]=[numa,na]+[numb,nb]
static mp_limb_t lmmp_add_1_(mp_ptr dst, mp_srcptr numa, mp_size_t na, mp_limb_t x)
加单精度数静态内联函数 [dst,na]=[numa,na]+x
void lmmp_mul_(mp_ptr dst, mp_srcptr numa, mp_size_t na, mp_srcptr numb, mp_size_t nb)
不等长乘法操作 [dst,na+nb] = [numa,na] * [numb,nb]
int lmmp_cmp_(mp_srcptr numa, mp_srcptr numb, mp_size_t n)
比较函数
定义 cmp.c:26
mp_limb_t lmmp_shl_(mp_ptr dst, mp_srcptr numa, mp_size_t na, mp_size_t shl)
左移操作 [dst,na] = [numa,na]<<shl,dst的低shl位填充0
定义 shl.c:26
mp_limb_t lmmp_add_n_sub_n_(mp_ptr dsta, mp_ptr dstb, mp_srcptr numa, mp_srcptr numb, mp_size_t n)
同时执行n位加法和减法 ([dsta,n],[dstb,n]) = ([numa,n]+[numb,n],[numa,n]-[numb,n])
mp_limb_t lmmp_add_n_(mp_ptr dst, mp_srcptr numa, mp_srcptr numb, mp_size_t n)
无进位的n位加法 [dst,n] = [numa,n] + [numb,n]
定义 add_n.c:76
#define b0
#define b1
#define v0
#define bsm1
#define asm1
#define vinf
#define vm1
#define v1
#define v2
#define b2
#define t
#define a3
#define s
#define bs1
#define bs2
#define as2
#define b0b2
#define asm2
#define bsm2
#define a0a2
#define as1
#define a1a3
#define vm2
#define b1d
#define scratch
LMMP_VECT_CONSERVATIVE int lmmp_toom_eval_dgr3_pm1_(mp_ptr xp1, mp_ptr xm1, mp_srcptr xp, mp_size_t n, mp_size_t x3n, mp_ptr tp)
Copyright (C) 2026 HJimmyK (Jericho Knox)
int lmmp_toom_eval_dgr3_pm2_(mp_ptr xp2, mp_ptr xm2, mp_srcptr xp, mp_size_t n, mp_size_t x3n, mp_ptr tp)
Toom-3 专用:3次多项式在 x = +2 和 x = -2 处求值 计算 P(+2) 和 P(-2),其中 P(x) 是一个3次多项式(4段系数)。
void lmmp_toom_interp6_(mp_ptr dst, mp_size_t n, enum toom6_flags flags, mp_ptr w4, mp_ptr w2, mp_ptr w1, mp_size_t w0n)
Copyright (C) 2026 HJimmyK (Jericho Knox)
#define restrict
Copyright (C) 2026 HJimmyK (Jericho Knox)
#define SALLOC_TYPE(n, type)
#define TEMP_S_DECL
#define TEMP_S_FREE
toom6_flags
Copyright (C) 2026 HJimmyK (Jericho Knox)
@ toom6_vm2_neg
@ toom6_vm1_neg

引用了 a0a2, a1a3, a3, as1, as2, asm1, asm2, b0, b0b2, b1, b1d, b2, bs1, bs2, bsm1, bsm2, lmmp_add_(), lmmp_add_1_(), lmmp_add_n_(), lmmp_add_n_sub_n_(), lmmp_cmp_(), lmmp_debug_assert, lmmp_mul_(), lmmp_mul_n_, lmmp_mul_toom43_(), lmmp_param_assert, lmmp_shl_(), lmmp_toom_eval_dgr3_pm1_(), lmmp_toom_eval_dgr3_pm2_(), lmmp_toom_interp6_(), n, numb, restrict, s, SALLOC_TYPE, scratch, t, TEMP_S_DECL, TEMP_S_FREE, toom6_vm1_neg, toom6_vm2_neg, v0, v1, v2, vinf, vm1 , 以及 vm2.

函数调用图: