add_n_sub_n.c 文件参考

#include "../../../include/lammp/impl/mparam.h"
#include "../../../include/lammp/lmmpn.h"

add_n_sub_n.c 的引用(Include)关系图:

浏览源代码.

函数
mp_limb_t	lmmp_add_n_sub_n_ (mp_ptr dsta, mp_ptr dstb, mp_srcptr numa, mp_srcptr numb, mp_size_t n)
	Copyright (C) 2026 HJimmyK(Jericho Knox)

函数说明

lmmp_add_n_sub_n_()

mp_limb_t lmmp_add_n_sub_n_	(	mp_ptr	dsta,
		mp_ptr	dstb,
		mp_srcptr	numa,
		mp_srcptr	numb,
		mp_size_t	n
	)

Copyright (C) 2026 HJimmyK(Jericho Knox)

同时执行n位加法和减法 ([dsta,n],[dstb,n]) = ([numa,n]+[numb,n],[numa,n]-[numb,n])

This file is part of LAMMP.

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

This program is distributed WITHOUT ANY WARRANTY.

See https://www.gnu.org/licenses/.

在文件 add_n_sub_n.c 第 20 行定义.

                                                                                                   {
    /*
    这段代码看起来有点奇怪的原因是，对于使用x64汇编时，我们会使用带进位的加法和减法，而x64中
    只能使用同一个进位寄存器，所以我们需要将两条指令分开执行。
    而不使用汇编时，编译器通常不会使用进位寄存器。因此我们可以同时读取两路内存，以减少读写次数。
    */
#ifdef USE_ASM
    mp_limb_t acyo = 0, scyo = 0;
    mp_size_t off, this_n;
 
    if (dsta != numa && dsta != numb) {
        for (off = 0; off < n; off += PART_SIZE) {
            this_n = LMMP_MIN(n - off, PART_SIZE);
            acyo = lmmp_add_nc_(dsta + off, numa + off, numb + off, this_n, acyo);
            scyo = lmmp_sub_nc_(dstb + off, numa + off, numb + off, this_n, scyo);
        }
    } else if (dstb != numa && dstb != numb) {
        for (off = 0; off < n; off += PART_SIZE) {
            this_n = LMMP_MIN(n - off, PART_SIZE);
            scyo = lmmp_sub_nc_(dstb + off, numa + off, numb + off, this_n, scyo);
            acyo = lmmp_add_nc_(dsta + off, numa + off, numb + off, this_n, acyo);
        }
    } else {
        mp_limb_t tp[PART_SIZE];
        for (off = 0; off < n; off += PART_SIZE) {
            this_n = LMMP_MIN(n - off, PART_SIZE);
            acyo = lmmp_add_nc_(tp, numa + off, numb + off, this_n, acyo);
            scyo = lmmp_sub_nc_(dstb + off, numa + off, numb + off, this_n, scyo);
            lmmp_copy(dsta + off, tp, this_n);
        }
    }
    return 2 * acyo + scyo;
#else
    mp_size_t i;
    mp_limb_t acyo, scyo;
 
    for (i = 0, acyo = 0, scyo = 0; i < n; i++) {
        mp_limb_t a, b, r;
        a = numa[i];
        b = numb[i];
        r = a + acyo;
        acyo = (r < acyo);
        r += b;
        acyo += (r < b);
        dsta[i] = r;
 
        b += scyo;
        scyo = (b < scyo);
        scyo += (a < b);
        dstb[i] = a - b;
    }
    return 2 * acyo + scyo;
#endif
}

引用了 lmmp_add_nc_(), lmmp_copy, LMMP_MIN, lmmp_sub_nc_(), n, numb, PART_SIZE , 以及 tp.

被这些函数引用 lmmp_mul_toom32_(), lmmp_mul_toom33_(), lmmp_mul_toom42_(), lmmp_mul_toom42_cache_(), lmmp_mul_toom42_cache_init_(), lmmp_mul_toom43_(), lmmp_mul_toom52_(), lmmp_mul_toom53_(), lmmp_mul_toom62_(), lmmp_mul_toom62_cache_init_(), lmmp_sqr_toom3_(), lmmp_toom_eval_dgr3_pm1_(), lmmp_toom_eval_dgr3_pm2_(), lmmp_toom_eval_pm1_() , 以及 lmmp_toom_eval_pm2_().

函数调用图:

这是这个函数的调用关系图: