static void revol(rmptr r)
{s_monom s;
vmrec rec;
vmptr m, m1;
int i;
s = r->n;
r->n = r->d;
r->d = s;
rec.next = (r->d).v;
m = &rec;
m1 = m->next;
while (m1 != NULL)
{
if (strcmp(m1->name,"i") == 0 )
{
m->next = m1->next;
if ( (m1->deg & 1) == 0) s.c= 1 ; else s.c= -1;
s.v=m1;
m1->next=NULL;
mult_s(&(r->n),&s);
m1 = m->next;
} else
if (strcmp(m1->name,"Sqrt2") == 0 )
{
m->next = m1->next;
for(i=0; i<m1->deg; i++) (r->d).c = (r->d).c*2;
s.c=1;
s.v=m1;
m1->next=NULL;
mult_s(&(r->n),&s);
m1 = m->next;
} else
{
m = m1;
m1 = m1->next;
}
}
(r->d).v = rec.next;
reducec(&r->n.c,& r->d.c);
if (r->d.c < 0)
{ r->d.c = - r->d.c;
r->d.c = - r->d.c;
}
}
void mult_rptr(rmptr* m1,rmptr* m2)
{
reduce_s(&(*m1)->n,&(*m2)->d);
reduce_s(&(*m1)->d,&(*m2)->n);
mult_s(&(*m1)->n,&(*m2)->n);
mult_s(&(*m1)->d,&(*m2)->d);
free(*m2);
delsqrt2(&(*m1)->n);
del_i(&(*m1)->n);
reducec(&(*m1)->n.c, &(*m1)->d.c);
if ((*m1)->d.c < 0)
{ (*m1)->d.c = - (*m1)->d.c;
(*m1)->d.c = - (*m1)->d.c;
}
}
// Sequential version of Karatsuba multiplication, with recursion
// It works well when numbers are about same size
large_int* mult_kar_eqsize(large_int* a, large_int* b) {
int n;
large_int a0,a1,b0,b1;
large_int *a1_plus_a0,*b1_plus_b0, *a1b0_plus_a0b1;
large_int *a1_mult_b1,*a0_mult_b0, *a1pa0_mult_b1pb0;
large_int *temp,*result;
if (debug) printf("Calling mult_kar_eqsize\n");
// For small numbers we use standard multiplication algorithm
// This is the end of the recursion
if (a->size<=KMUX_MINLEVEL || b->size<=KMUX_MINLEVEL) {
if (debug) printf("Calling mult_gmp\n");
// return mult_gmp(a,b);
return mult_s(a,b);
}
// Here we choose first if a or b are smaller. Then out of the smallest one
// which we can call C, we find n that is <= 1/2 of it
if (a->size>b->size) {
n=b->size/2;
if (2*n<b->size) n++;
}
else {
n=a->size/2;
if (2*n<a->size) n++;
}
a0.size=n;
a1.size=a->size-n;
b0.size=n;
b1.size=b->size-n;
// allocate arrays
a0.int_array=malloc(sizeof(unsigned int)*a0.size);
a1.int_array=malloc(sizeof(unsigned int)*a1.size);
b0.int_array=malloc(sizeof(unsigned int)*b0.size);
b1.int_array=malloc(sizeof(unsigned int)*b1.size);
if (a0.int_array==NULL || a1.int_array==NULL || b0.int_array==NULL || b1.int_array==NULL)
die("memory allocation error");
// Copy data from a,b into a0,a1,b0,b1
memcpy(a0.int_array,a->int_array,a0.size*sizeof(unsigned int));
memcpy(a1.int_array,(char*)a->int_array+((size_t)a0.size*sizeof(unsigned int)),a1.size*sizeof(unsigned int));
memcpy(b0.int_array,b->int_array,b0.size*sizeof(unsigned int));
memcpy(b1.int_array,(char*)b->int_array+((size_t)b0.size*sizeof(unsigned int)),b1.size*sizeof(unsigned int));
// Calculate a0+a1, b1+b0 - these can potentially be done in parallel to each other
#pragma omp parallel sections default (shared)
{
a1_plus_a0=ADD(&a1,&a0);
#pragma omp section
b1_plus_b0=ADD(&b1,&b0);
}
// Calculate 3 intermediate products - can all be done in parallel
#pragma omp parallel sections default (shared)
{
if (debug) printf("d0 ");
a1_mult_b1=mult_kar_eqsize(&a1,&b1);
if (debug) { printf("a1*b1 : "); print_hex(a1_mult_b1); }
#pragma omp section
{
a0_mult_b0=mult_kar_eqsize(&a0,&b0);
if (debug) { printf("a0*b0 : "); print_hex(a0_mult_b0); }
}
#pragma omp section
{
a1pa0_mult_b1pb0=mult_kar_eqsize(a1_plus_a0,b1_plus_b0);
if (debug) { printf("(a1+a0)*(b1+b0) : "); print_hex(a1pa0_mult_b1pb0); }
}
}
// From here on things can not really be parallelizeable
if (debug && compare(a1_mult_b1,a1pa0_mult_b1pb0)>0) {
printf("Noticed that a1b1 > (a1+a0)(b1+b0), this should not be\n");
temp=SUB(a1_mult_b1,a1pa0_mult_b1pb0);
temp->sign=-1;
}
else {
if (debug) printf("Calling sub -a1b1 , first_size=%d, second_size=%d\n",a1pa0_mult_b1pb0->size,a1_mult_b1->size);
temp=SUB(a1pa0_mult_b1pb0,a1_mult_b1);
temp->sign=1;
if (debug) print_hex(temp);
}
if (debug && compare(temp,a0_mult_b0)<0) {
if (debug) printf("Noticed that a0b0 > reminder, should not be");
a1b0_plus_a0b1=SUB(a0_mult_b0,temp);
if (temp->sign==-1 && debug) printf("but we're good\n");
else { printf("we have a problem\n"); a1b0_plus_a0b1->sign=-1; }
}
else {
if (debug) printf("calling sub -a0b0: ");
a1b0_plus_a0b1=SUB(temp,a0_mult_b0);
if (temp->sign==-1) printf("we have a problem\n");
if (debug) print_hex(a1b0_plus_a0b1);
}
free(temp->int_array);
free(temp);
// Do some shifting around to prepare final result
shift_left(a1b0_plus_a0b1,n);
shift_left(a1_mult_b1,n*2);
if (debug) {
printf("Shifted a1b0_plus_a0b1: ");
print_hex(a1b0_plus_a0b1);
printf("Shifted a1_mult_b1: ");
print_hex(a1_mult_b1);
}
// Now we're ready to put it all together
temp=ADD(a1_mult_b1,a1b0_plus_a0b1);
result=ADD(temp,a0_mult_b0);
// Free all allocated memory
free(temp->int_array);
free(a1_mult_b1->int_array);
free(a0_mult_b0->int_array);
free(a1pa0_mult_b1pb0->int_array);
free(a1_plus_a0->int_array);
free(b1_plus_b0->int_array);
free(a1b0_plus_a0b1->int_array);
free(a0.int_array);
free(a1.int_array);
free(b0.int_array);
free(b1.int_array);
free(temp);
free(a1_mult_b1);
free(a0_mult_b0);
free(a1pa0_mult_b1pb0);
free(a1_plus_a0);
free(b1_plus_b0);
free(a1b0_plus_a0b1);
// And return
return result;
}
void diagramsrfactors(hlpcsptr gst,s_listptr* s,rmptr* totf)
{s_listptr s1, s2, ss, dl;
vcsect vcs_copy;
int i;
rmptr rcoef, rrcoef;
rmptr r;
s_monom stmp, stmp2;
int first;
s1 = NULL;
s2 = NULL;
first = 1;
vcs_copy = vcs;
while (gst != NULL)
{
ss = (s_listptr) m_alloc(sizeof(struct s_listrec));
ss->next = s1; s1 = ss;
ss = (s_listptr) m_alloc(sizeof(struct s_listrec));
ss->next = s2; s2 = ss;
coloringvcs(gst);
attachvertexes();
rcoef = (rmptr) readExpression(vertexes[0].lgrnptr->comcoef,
rd_rat,act_rat,NULL);
for (i = 2; i <= vcs.sizet; i++)
{
rrcoef = (rmptr) readExpression(vertexes[i-1].lgrnptr->comcoef,
rd_rat,act_rat,NULL);
mult_rptr(&rcoef,&rrcoef);
}
if (first)
{
r = rcoef;
s1->monom.c = 1;
s1->monom.v = NULL;
s2->monom.c = 1;
s2->monom.v = NULL;
first = 0;
}
else
{
copysmonom(r->n,&s1->monom);
s2->monom = rcoef->n;
reduce_s(&s1->monom,&s2->monom);
copysmonom(s1->monom,&stmp);
reduce_s(&r->n,&stmp);
mult_s(&r->n,&stmp); /* for case Stmp = -1 */
copysmonom(r->d,&stmp);
reduce_s(&stmp,&rcoef->d);
copysmonom(rcoef->d,&stmp2);
mult_s(&s1->monom,&rcoef->d);
mult_s(&s2->monom,&stmp);
mult_s(&r->d,&stmp2);
free(rcoef);
}
vcs = vcs_copy;
gst = gst->next;
}
vcs = vcs_copy;
ss = s2;
stmp.v = NULL;
stmp.c = 1;
while (s2 != NULL)
{
copysmonom(stmp,&stmp2);
mult_s(&s2->monom,&stmp2);
mult_s(&stmp,&s1->monom);
dl = s1;
s1 = s1->next;
free(dl);
s2 = s2->next;
}
clrvm(stmp.v);
revers((void **)&ss);
*s = ss;
*totf = r;
}