/* $OpenXM: OpenXM_contrib2/asir2000/lib/dmul,v 1.1 1999/12/27 04:16:32 noro Exp $ */
#define MAX(a,b) ((a)>(b)?(a):(b))
#define MIN(a,b) ((a)>(b)?(b):(a))

/* CAUTION: functions in this file are experimental. */

/*
	return: F1*F2
	if option 'proc' is supplied as a list of server id's, 
	F1*F2 is calculated by distributed computation.
*/

def d_mul(F1,F2)
{
	Procs = getopt(proc);
	if ( type(Procs) == -1 )
		Procs = [];
	NP = length(Procs)+1;
	V =var(F1);
	D1 = deg(F1,V);
	D2 = deg(F2,V);
	Dmin = MIN(D1,D2);
	Dfft = p_mag(D1+D2+1)+1;
	Bound = maxblen(F1)+maxblen(F2)+p_mag(Dmin)+1;

	Marray = newvect(NP);
	MIarray = newvect(NP);
	for ( I = 0; I < NP; I++ ) {
		Marray[I] = 1;
		MIarray[I] = [];
	}

	for ( M = 1, I = 0, J = 0; p_mag(M) <= Bound; J = (J+1)%NP ) {
		T = get_next_fft_prime(I,Dfft);
		if ( !T )
			error("fft_mul_d : fft_prime exhausted.");
		Marray[J] *= T[1];
		MIarray[J] = cons(T[0],MIarray[J]);
		M *= T[1];
		I = T[0]+1;
	}
	/* Now, 
		Marray[J] = FFTprime[Marray[J][0]]*...*FFTprime[Marray[J][...]]
		M = Marray[0]*...*Marray[NP-1]
	*/
	C = newvect(NP);
	T0 = time();
	for ( J = 0; J < NP-1; J++ )
		ox_cmo_rpc(Procs[J],"call_umul",F1,F2,MIarray[J],Marray[J],M);
	T1 = time();
	R = call_umul(F1,F2,MIarray[NP-1],Marray[NP-1],M);
	T2 = time();
	for ( J = 0; J < NP-1; J++ )
		R += ox_pop_cmo(Procs[J]);
	T3 = time();
	print(["send",T1[3]-T0[3],"self",T2[3]-T1[3],"recv",T3[3]-T2[3]]);
	return R%M;
}

/*
	return: F1^2
	if option 'proc' is supplied as a list of server id's, 
	F1^2 is calculated by distributed computation.
*/

def d_square(F1)
{
	Procs = getopt(proc);
	if ( type(Procs) == -1 )
		Procs = [];
	NP = length(Procs)+1;
	V =var(F1);
	D1 = deg(F1,V);
	Dfft = p_mag(2*D1+1)+1;
	Bound = 2*maxblen(F1)+p_mag(D1)+1;

	Marray = newvect(NP);
	MIarray = newvect(NP);
	for ( I = 0; I < NP; I++ ) {
		Marray[I] = 1;
		MIarray[I] = [];
	}

	for ( M = 1, I = 0, J = 0; p_mag(M) <= Bound; J = (J+1)%NP ) {
		T = get_next_fft_prime(I,Dfft);
		if ( !T )
			error("fft_mul_d : fft_prime exhausted.");
		Marray[J] *= T[1];
		MIarray[J] = cons(T[0],MIarray[J]);
		M *= T[1];
		I = T[0]+1;
	}
	/* Now, 
		Marray[J] = FFTprime[Marray[J][0]]*...*FFTprime[Marray[J][...]]
		M = Marray[0]*...*Marray[NP-1]
	*/
	C = newvect(NP);
	T0 = time();
	for ( J = 0; J < NP-1; J++ )
		ox_cmo_rpc(Procs[J],"call_usquare",F1,MIarray[J],Marray[J],M);
	T1 = time();
	R = call_usquare(F1,MIarray[NP-1],Marray[NP-1],M);
	T2 = time();
	for ( J = 0; J < NP-1; J++ )
		R += ox_pop_cmo(Procs[J]);
	T3 = time();
	print(["send",T1[3]-T0[3],"self",T2[3]-T1[3],"recv",T3[3]-T2[3]]);
	return R%M;
}

/*
	return: F1^2 mod V^(D+1)
	if option 'proc' is supplied as a list of server id's, 
	F1*F2 mod V^(D+1) is calculated by distributed computation.
*/

def d_tmul(F1,F2,D)
{
	Procs = getopt(proc);
	if ( type(Procs) == -1 )
		Procs = [];
	NP = length(Procs)+1;
	V =var(F1);
	D1 = deg(F1,V);
	D2 = deg(F2,V);
	Dmin = MIN(D1,D2);
	Dfft = p_mag(D1+D2+1)+1;
	Bound = maxblen(F1)+maxblen(F2)+p_mag(Dmin)+1;

	Marray = newvect(NP);
	MIarray = newvect(NP);
	for ( I = 0; I < NP; I++ ) {
		Marray[I] = 1;
		MIarray[I] = [];
	}

	for ( M = 1, I = 0, J = 0; p_mag(M) <= Bound; J = (J+1)%NP ) {
		T = get_next_fft_prime(I,Dfft);
		if ( !T )
			error("fft_mul_d : fft_prime exhausted.");
		Marray[J] *= T[1];
		MIarray[J] = cons(T[0],MIarray[J]);
		M *= T[1];
		I = T[0]+1;
	}
	/* Now, 
		Marray[J] = FFTprime[Marray[J][0]]*...*FFTprime[Marray[J][...]]
		M = Marray[0]*...*Marray[NP-1]
	*/
	C = newvect(NP);
	T0 = time();
	for ( J = 0; J < NP-1; J++ )
		ox_cmo_rpc(Procs[J],"call_utmul",F1,F2,D,MIarray[J],Marray[J],M);
	T1 = time();
	R = call_utmul(F1,F2,D,MIarray[NP-1],Marray[NP-1],M);
	T2 = time();
	for ( J = 0; J < NP-1; J++ )
		R += ox_pop_cmo(Procs[J]);
	T3 = time();
	print(["send",T1[3]-T0[3],"self",T2[3]-T1[3],"recv",T3[3]-T2[3]]);
	return R%M;
}

def call_umul(F1,F2,Ind,M1,M)
{
	C = umul_specialmod(F1,F2,Ind);
	Mhat = idiv(M,M1);
	MhatInv = inv(Mhat,M1);
	return (MhatInv*Mhat)*C;
}

def call_usquare(F1,Ind,M1,M)
{
	C = usquare_specialmod(F1,Ind);
	Mhat = idiv(M,M1);
	MhatInv = inv(Mhat,M1);
	return (MhatInv*Mhat)*C;
}

def call_utmul(F1,F2,D,Ind,M1,M)
{
	C = utmul_specialmod(F1,F2,D,Ind);
	Mhat = idiv(M,M1);
	MhatInv = inv(Mhat,M1);
	return (MhatInv*Mhat)*C;
}
end$