mul_toom32.c 文件参考

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

mul_toom32.c 的引用(Include)关系图:

浏览源代码.

宏定义
#define	a0   numa
#define	a1   (numa + n)
#define	a2   (numa + 2 * n)
#define	am1   (dst + 3 * n + 2)
#define	ap1   (dst + 2 * n + 1)
#define	b0   numb
#define	b1   (numb + n)
#define	bm1   (dst)
#define	bp1   (dst + n)
#define	r0   (dst)
#define	r1   (dst + n)
#define	r2   (dst + 2 * n)
#define	r3   (dst + 3 * n)
#define	v1   (tp)
#define	vm1   (tp + 2 * n + 1)

函数
void	lmmp_mul_toom32_ (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

a1

#define a1   (numa + n)

a2

#define a2   (numa + 2 * n)

am1

#define am1   (dst + 3 * n + 2)

ap1

#define ap1   (dst + 2 * n + 1)

b0

#define b0   numb

b1

#define b1   (numb + n)

bm1

#define bm1   (dst)

bp1

#define bp1   (dst + n)

r0

#define r0   (dst)

r1

#define r1   (dst + n)

r2

#define r2   (dst + 2 * n)

r3

#define r3   (dst + 3 * n)

v1

#define v1   (tp)

vm1

#define vm1   (tp + 2 * n + 1)

函数说明

lmmp_mul_toom32_()

void lmmp_mul_toom32_	(	mp_ptr restrict	dst,
		mp_srcptr restrict	numa,
		mp_size_t	na,
		mp_srcptr restrict	numb,
		mp_size_t	nb
	)

在文件 mul_toom32.c 第 24 行定义.

                                                                                                                         {
    lmmp_param_assert(nb >= 12);
    lmmp_param_assert(4 * na >= 5 * nb);
    lmmp_param_assert(5 * na <= 9 * nb);
    TEMP_S_DECL;
    mp_size_t n = 1 + (2 * na >= 3 * nb ? (na - 1) / 3 : (nb - 1) >> 1), s = na - 2 * n, t = nb - n;
    int vm1_neg;
    mp_limb_t cy, hi;
    mp_limb_t* restrict tp = SALLOC_TYPE(4 * n + 2, mp_limb_t);
 
#define a0 numa
#define a1 (numa + n)
#define a2 (numa + 2 * n)
#define b0 numb
#define b1 (numb + n)
    // nb>=12, so that s+t>=n+2
#define bm1 (dst)              //[dst,n]
#define bp1 (dst + n)          //[dst+n,n+1]
#define ap1 (dst + 2 * n + 1)  //[dst+2*n+1,n+1]
#define am1 (dst + 3 * n + 2)  //[dst+3*n+2,n]:hi
#define v1 (tp)                //[tp,2*n+1]
#define vm1 (tp + 2 * n + 1)   //[tp+2*n+1,2*n+1]
#define r0 (dst)
#define r1 (dst + n)
#define r2 (dst + 2 * n)
#define r3 (dst + 3 * n)
 
    // ap1=a0+a1+a3, am1=a0-a1+a3
    ap1[n] = lmmp_add_(ap1, a0, n, a2, s);
    if (ap1[n] == 0 && lmmp_cmp_(ap1, a1, n) < 0) {
        ap1[n] = lmmp_add_n_sub_n_(ap1, am1, a1, ap1, n) >> 1;
        hi = 0;
        vm1_neg = 1;
    } else {
        cy = lmmp_add_n_sub_n_(ap1, am1, ap1, a1, n);
        hi = ap1[n] - (cy & 1);
        ap1[n] += (cy >> 1);
        vm1_neg = 0;
    }
 
    // bp1=b0+b1, bm1=b0-b1
    if (t == n) {
        if (lmmp_cmp_(b0, b1, n) < 0) {
            bp1[n] = lmmp_add_n_sub_n_(bp1, bm1, b1, b0, n) >> 1;
            vm1_neg ^= 1;
        } else {
            bp1[n] = lmmp_add_n_sub_n_(bp1, bm1, b0, b1, n) >> 1;
        }
    } else {
        if (lmmp_zero_q_(b0 + t, n - t) && lmmp_cmp_(b0, b1, t) < 0) {
            cy = lmmp_add_n_sub_n_(bp1, bm1, b1, b0, t);
            lmmp_zero(bm1 + t, n - t);
            vm1_neg ^= 1;
        } else {
            cy = lmmp_add_n_sub_n_(bp1, bm1, b0, b1, t);
            lmmp_sub_1_(bm1 + t, b0 + t, n - t, cy & 1);
        }
        bp1[n] = lmmp_add_1_(bp1 + t, b0 + t, n - t, cy >> 1);
    }
 
    // v1=ap1*bp1
    lmmp_mul_n_(v1, ap1, bp1, n + 1);
 
    // vm=am1*bm1
    lmmp_mul_n_(vm1, am1, bm1, n);
    if (hi)
        hi = lmmp_add_n_(vm1 + n, vm1 + n, bm1, n);
    vm1[2 * n] = hi;
 
    // r0=a0*b0
    // r3=a2*b1
    lmmp_mul_n_(r0, a0, b0, n);
    if (s > t)
        lmmp_mul_(r3, a2, s, b1, t);
    else
        lmmp_mul_(r3, b1, t, a2, s);
 
    // v1=(v1+vm1)/2, (=a0*b0+a2*b0+a1*b1)
    // vm1=v1-vm1, (=a1*b0+a0*b1+a2*b1)
    if (vm1_neg) {
        lmmp_shr1sub_n_(v1, v1, vm1, 2 * n + 1);
        lmmp_add_n_(vm1, v1, vm1, 2 * n + 1);
    } else {
        lmmp_shr1add_n_(v1, v1, vm1, 2 * n + 1);
        lmmp_sub_n_(vm1, v1, vm1, 2 * n + 1);
    }
 
    // vm1-=r3, (=r1)
    // v1-=r0, (=r2)
    lmmp_sub_(vm1, vm1, 2 * n + 1, r3, s + t);
    v1[2 * n] -= lmmp_sub_n_(v1, v1, r0, 2 * n);
 
    // r=r0+vm1*B+v1*B^2+r3*B^4
    cy = vm1[2 * n] + lmmp_add_(r1, vm1, 2 * n, r1, n);
    lmmp_add_(r2, r2, n + s + t, v1, 2 * n + 1);
    lmmp_inc_1(r3, cy);
    TEMP_S_FREE;
}

引用了 a0, a1, a2, am1, ap1, b0, b1, bm1, bp1, lmmp_add_(), lmmp_add_1_(), lmmp_add_n_(), lmmp_add_n_sub_n_(), lmmp_cmp_(), lmmp_inc_1, lmmp_mul_(), lmmp_mul_n_(), lmmp_param_assert, lmmp_shr1add_n_(), lmmp_shr1sub_n_(), lmmp_sub_(), lmmp_sub_1_(), lmmp_sub_n_(), lmmp_zero, lmmp_zero_q_(), r0, r1, r2, r3, SALLOC_TYPE, TEMP_S_DECL, TEMP_S_FREE, tp, v1 , 以及 vm1.

函数调用图: