#include "mltaln.h"
#include "dp.h"

#define DEBUG 0
#define XXXXXXX    0
#define USE_PENALTY_EX  0
#define STOREWM 0

#define DPTANNI 50


static reccycle = 0;

static void OpeningGapCount( float *ogcp, int clus, char **seq, double *eff )
{
	int i, j, gc, gb; 
	int len = strlen( seq[0] );
	float totaleff = 0.0;
	
	for( i=0; i<len; i++ ) ogcp[i] = 0.0;
	for( j=0; j<clus; j++ ) 
	{
		gc = 0;
		for( i=0; i<len; i++ ) 
		{
			gb = gc;
			gc = ( seq[j][i] == '-' );
			{
				if( !gb *  gc ) ogcp[i] += eff[j];
			}
		}
		totaleff+= eff[j];
	}
	for( i=0; i<len; i++ ) 
		ogcp[i] /= totaleff;
}

static void FinalGapCount( float *fgcp, int clus, char **seq, double *eff )
{
	int i, j, gc, gb; 
	int len = strlen( seq[0] );
	float totaleff = 0.0;
	
	for( i=0; i<len; i++ ) fgcp[i] = 0.0;
	for( j=0; j<clus; j++ ) 
	{
		gc = ( seq[j][0] == '-' );
		for( i=1; i<len+1; i++ ) 
		{
			gb = gc;
			gc = ( seq[j][i] == '-' );
			{
				if( gb * !gc ) fgcp[i-1] += eff[j];
			}
		}
		totaleff += eff[j];
	}
	for( i=0; i<len; i++ ) 
		fgcp[i] /= totaleff;
}

static void match_calc( float *match, float **cpmx1, float **cpmx2, int i1, int lgth2, float **floatwork, int **intwork, int initialize )
{
	int j, k, l;
	float scarr[26];
	float **cpmxpd = floatwork;
	int **cpmxpdn = intwork;
	int count = 0;

	if( initialize )
	{
		for( j=0; j<lgth2; j++ )
		{
			count = 0;
			for( l=0; l<26; l++ )
			{
				if( cpmx2[j][l] )
				{
					cpmxpd[j][count] = cpmx2[j][l];
					cpmxpdn[j][count] = l;
					count++;
				}
			}
			cpmxpdn[j][count] = -1;
		}
	}

	for( l=0; l<26; l++ )
	{
		scarr[l] = 0.0;
		for( k=0; k<26; k++ )
		{
			scarr[l] += n_dis[k][l] * cpmx1[i1][k];
		}
	}
#if 0 /* これを使うときはfloatworkのアロケートを逆にする */
	{
		float *fpt, **fptpt, *fpt2;
		int *ipt, **iptpt;
		fpt2 = match;
		iptpt = cpmxpdn;
		fptpt = cpmxpd;
		while( lgth2-- )
		{
			*fpt2 = 0.0;
			ipt=*iptpt,fpt=*fptpt;
			while( *ipt > -1 )
				*fpt2 += scarr[*ipt++] * *fpt++;
			fpt2++,iptpt++,fptpt++;
		} 
	}
#else
	for( j=0; j<lgth2; j++ )
	{
		match[j] = 0.0;
		for( k=0; cpmxpdn[j][k]>-1; k++ )
			match[j] += scarr[cpmxpdn[j][k]] * cpmxpd[j][k];
	} 
#endif
}

static float Atracking( float *lasthorizontalw, float *lastverticalw, 
						char **seq1, char **seq2, 
                        char **mseq1, char **mseq2, 
                        float **cpmx1, float **cpmx2, 
                        short **ijp, int icyc, int jcyc,
						int ist, int ien, int jst, int jen )
{
	int i, j, l, iin, jin, ifi, jfi, lgth1, lgth2, k, klim;
	char gap[] = "-";
	float wm;
	lgth1 = ien-ist+1;
	lgth2 = jen-jst+1;

#if 0
	for( i=0; i<lgth1; i++ ) 
	{
		fprintf( stderr, "lastverticalw[%d] = %f\n", i, lastverticalw[i] );
	}
#endif


//	fprintf( stderr, "in Atracking, lgth1=%d, lgth2=%d\n", lgth1, lgth2 );
 
    for( i=0; i<lgth1+1; i++ ) 
    {
        ijp[i][0] = i + 1;
    }
    for( j=0; j<lgth2+1; j++ ) 
    {
        ijp[0][j] = -( j + 1 );
    }


	for( i=0; i<icyc; i++ )
	{
		mseq1[i] += lgth1+lgth2;
		*mseq1[i] = 0;
	}
	for( j=0; j<jcyc; j++ )
	{
		mseq2[j] += lgth1+lgth2;
		*mseq2[j] = 0;
	}

//	if( lgth2 == 1 ) fprintf( stderr, "in Atracking, mseq1 = %s, mseq2 = %s\n", mseq1[0], mseq2[0] );

	iin = lgth1; jin = lgth2;
	klim = lgth1+lgth2;
	for( k=0; k<=klim; k++ ) 
	{
		if( ijp[iin][jin] < 0 ) 
		{
			ifi = iin-1; jfi = jin+ijp[iin][jin];
		}
		else if( ijp[iin][jin] > 0 )
		{
			ifi = iin-ijp[iin][jin]; jfi = jin-1;
		}
		else
		{
			ifi = iin-1; jfi = jin-1;
		}
		l = iin - ifi;
		while( --l ) 
		{
			for( i=0; i<icyc; i++ )
				*--mseq1[i] = seq1[i][ist+ifi+l];
			for( j=0; j<jcyc; j++ ) 
				*--mseq2[j] = *gap;
			k++;
		}
		l= jin - jfi;
		while( --l )
		{
			for( i=0; i<icyc; i++ ) 
				*--mseq1[i] = *gap;
			for( j=0; j<jcyc; j++ ) 
				*--mseq2[j] = seq2[j][jst+jfi+l];
			k++;
		}
		if( iin <= 0 || jin <= 0 ) break;
		for( i=0; i<icyc; i++ ) 
			*--mseq1[i] = seq1[i][ist+ifi];
		for( j=0; j<jcyc; j++ ) 
			*--mseq2[j] = seq2[j][jst+jfi];
		k++;
		iin = ifi; jin = jfi;

//		if( lgth2 == 1 ) fprintf( stderr, "in Atracking, mseq1 = %s, mseq2 = %s\n", mseq1[0], mseq2[0] );
	}
//	fprintf( stderr, "in Atracking (owari), mseq1 = %s\n", mseq1[0] );
//	fprintf( stderr, "in Atracking (owari), mseq2 = %s\n", mseq2[0] );
	return( 0.0 );
}

static float MSalignmm_tanni( int icyc, int jcyc, double *eff1, double *eff2, char **seq1, char **seq2, float **cpmx1, float **cpmx2, int ist, int ien, int jst, int jen, int alloclen, char **mseq1, char **mseq2, float *gapinfo[4]  )
/* score no keisan no sai motokaraaru gap no atukai ni mondai ga aru */
{
//	int k;
	register int i, j;
	int ll1, ll2;
	int lasti, lastj;
	int resultlen;
	float wm;   /* int ?????? */
	float g;
	float *currentw, *previousw;
	float fpenalty = (float)penalty;
	float fpenalty_ex = (float)penalty_ex;
#if 1
	float *wtmp;
	short *ijppt;
	float *mjpt, *prept, *curpt;
	int *mpjpt;
#endif
	float mi, *m;
	short **ijp;
	int mpi, *mp;
	float *w1, *w2;
	float *initverticalw;    /* kufuu sureba iranai */
	float *lastverticalw;    /* kufuu sureba iranai */
	int **intwork;
	float **floatwork;
	short **shortmtx;
	float dumfl;
	int lgth1, lgth2;
	float *ogcp1 = gapinfo[0];
	float *fgcp1 = gapinfo[1];
	float *ogcp2 = gapinfo[2];
	float *fgcp2 = gapinfo[3];
	static char **aseq1 = NULL;
	static char **aseq2 = NULL;
	static char **aseq1bk, **aseq2bk;

//	char ttt1[10000], ttt2[10000];

	if( !aseq1 )
	{
		aseq1 = AllocateCharMtx( njob, 0 );
		aseq2 = AllocateCharMtx( njob, 0 );
	}

	lgth1 = ien-ist+1;
	lgth2 = jen-jst+1;

//	strncpy( ttt1, seq1[0]+ist, lgth1 ); ttt1[lgth1] = 0;
//	strncpy( ttt2, seq2[0]+jst, lgth2 ); ttt2[lgth2] = 0;
//
//	fprintf( stderr, "in _tanni ist,ien = %d,%d, lgth1=%d\n", ist, ien, lgth1 );
//	fprintf( stderr, "in _tanni jst,jen = %d,%d, lgth2=%d\n", jst, jen, lgth2 );
//	fprintf( stderr, "ttt1 = %s\n", ttt1 );
//	fprintf( stderr, "ttt2 = %s\n", ttt2 );


	ll1 = ( (int)(1.3*lgth1) ) + 100;
	ll2 = ( (int)(1.3*lgth2) ) + 100;

	aseq1bk = AllocateCharMtx( icyc, lgth1+lgth2+100 );
	aseq2bk = AllocateCharMtx( jcyc, lgth1+lgth2+100 );
	for( i=0; i<icyc; i++ ) aseq1[i] = aseq1bk[i];
	for( i=0; i<jcyc; i++ ) aseq2[i] = aseq2bk[i];

	w1 = AllocateFloatVec( ll2+2 );
	w2 = AllocateFloatVec( ll2+2 );

	initverticalw = AllocateFloatVec( ll1+2 );
	lastverticalw = AllocateFloatVec( ll1+2 );

	m = AllocateFloatVec( ll2+2 );
	mp = AllocateIntVec( ll2+2 );

	floatwork = AllocateFloatMtx( MAX( ll1, ll2 )+2, 26 ); 
	intwork = AllocateIntMtx( MAX( ll1, ll2 )+2, 26 ); 


	shortmtx = AllocateShortMtx( ll1+1, ll2+1 );

	ijp = shortmtx;

	currentw = w1;
	previousw = w2;

	match_calc( initverticalw, cpmx2+jst, cpmx1+ist, 0, lgth1, floatwork, intwork, 1 );

	match_calc( currentw, cpmx1+ist, cpmx2+jst, 0, lgth2, floatwork, intwork, 1 );

	for( i=1; i<lgth1+1; i++ )
	{
		initverticalw[i] += fpenalty * ( ogcp1[0] + fgcp1[i-1] );
	}
	for( j=1; j<lgth2+1; j++ )
	{
		currentw[j] += fpenalty * ( ogcp2[0] + fgcp2[j-1] );
	}

	for( j=1; j<lgth2+1; ++j ) 
	{
		m[j] = currentw[j-1] + fpenalty * ogcp1[1]; mp[j] = 0;;
	}

	lastverticalw[0] = currentw[lgth2-1];


	lasti = lgth1+1;
	for( i=1; i<lasti; i++ )
	{

		wtmp = previousw; 
		previousw = currentw;
		currentw = wtmp;

		previousw[0] = initverticalw[i-1];

		match_calc( currentw, cpmx1+ist, cpmx2+jst, i, lgth2, floatwork, intwork, 0 );
		currentw[0] = initverticalw[i];

		mi = previousw[0] + fpenalty * ogcp2[1]; 
		mpi = 0;

		ijppt = ijp[i] + 1;
		mjpt = m + 1;
		prept = previousw;
		curpt = currentw + 1;
		mpjpt = mp + 1;
		lastj = lgth2+1;
		for( j=1; j<lastj; j++ )
		{
			wm = *prept;
			*ijppt = 0;

#if 0
			fprintf( stderr, "%5.0f->", wm );
#endif
			g = mi + fpenalty * fgcp2[j-1];
#if 0
			fprintf( stderr, "%5.0f?", g );
#endif
			if( g > wm )
			{
				wm = g;
				*ijppt = -( j - mpi );
			}
			g = *prept + fpenalty * ogcp2[j];
			if( g >= mi )
			{
				mi = g;
				mpi = j-1;
			}
#if USE_PENALTY_EX
			mi += fpenalty_ex;
#endif

			g = *mjpt + fpenalty * fgcp1[i-1];
#if 0 
			fprintf( stderr, "%5.0f?", g );
#endif
			if( g > wm )
			{
				wm = g;
				*ijppt = +( i - *mpjpt );
			}
			g = *prept + fpenalty * ogcp1[i];
			if( g >= *mjpt )
			{
				*mjpt = g;
				*mpjpt = i-1;
			}
#if USE_PENALTY_EX
			m[j] += fpenalty_ex;
#endif

#if 0
			fprintf( stderr, "%5.0f ", wm );
#endif
			*curpt += wm;


			ijppt++;
			mjpt++;
			prept++;
			mpjpt++;
			curpt++;
		}
		lastverticalw[i] = currentw[lgth2-1];
	}


	Atracking( currentw, lastverticalw, seq1, seq2, aseq1, aseq2, cpmx1+ist, cpmx2+jst, ijp, icyc, jcyc, ist, ien, jst, jen );

	for( i=0; i<icyc; i++ ) strcpy( mseq1[i], aseq1[i] );
	for( i=0; i<jcyc; i++ ) strcpy( mseq2[i], aseq2[i] );

//	fprintf( stderr, "in _tanni, aseq1 = %s\n", aseq1[0] );
//	fprintf( stderr, "in _tanni, mseq1 = %s\n", mseq1[0] );

	FreeFloatVec( w1 );
	FreeFloatVec( w2 );
	FreeFloatVec( initverticalw );
	FreeFloatVec( lastverticalw );

	FreeFloatVec( m );
	FreeIntVec( mp );


	FreeFloatMtx( floatwork );
	FreeIntMtx( intwork );

	FreeShortMtx( shortmtx );

	FreeCharMtx( aseq1bk );
	FreeCharMtx( aseq2bk );

	return( wm );
}

static float MSalignmm_rec( int icyc, int jcyc, double *eff1, double *eff2, char **seq1, char **seq2, float **cpmx1, float **cpmx2, int ist, int ien, int jst, int jen, int alloclen, char **mseq1, char **mseq2, int depth, float *gapinfo[4] )
/* score no keisan no sai motokaraaru gap no atukai ni mondai ga aru */
{
//	int k;
	register int i, j;
	char **aseq1, **aseq2;
	int ll1, ll2, l, len;
	int lasti, lastj, imid, jmid;
	int resultlen;
	float wm;   /* int ?????? */
	float g;
	float *currentw, *previousw;
	float fpenalty = (float)penalty;
	float fpenalty_ex = (float)penalty_ex;
	float *wtmp;
//	short *ijppt;
	int *mpjpt;
//	short **ijp;
	int *mp;
	int mpi;
	float *mjpt, *prept, *curpt;
	float mi;
	float *m;
	float *w1, *w2;
//	float *match;
	float *initverticalw;    /* kufuu sureba iranai */
	float *lastverticalw;    /* kufuu sureba iranai */
	int **intwork;
	float **floatwork;
//	short **shortmtx;
#if STOREWM
//	float **WMMTX;
//	float **WMMTX2;
#endif
	float *midw;
	float *midm;
	float *midn;
	float dumfl;
	int lgth1, lgth2;
	float maxwm;
	int *jumpforwi;
	int *jumpforwj;
	int *jumpbacki;
	int *jumpbackj;
	int *jumpdummi; //muda
	int *jumpdummj; //muda
	int jumpi, jumpj;
	char *gaps;
	int ijpi, ijpj;
//	float *ogcp1 = gapinfo[0];
//	float *fgcp1 = gapinfo[1];
//	float *ogcp2 = gapinfo[2];
//	float *fgcp2 = gapinfo[3];
//	static char ttt1[50000];
//	static char ttt2[50000];

	depth++;
	reccycle++;

	lgth1 = ien-ist+1;
	lgth2 = jen-jst+1;

//	fprintf( stderr, "==== MSalign (depth=%d, reccycle=%d), ist=%d, ien=%d, jst=%d, jen=%d\n", depth, reccycle, ist, ien, jst, jen );
//	strncpy( ttt1, seq1[0]+ist, lgth1 );
//	strncpy( ttt2, seq2[0]+jst, lgth2 );
//	ttt1[lgth1] = 0;
//	ttt2[lgth2] = 0;
//	fprintf( stderr, "seq1 = %s\n", ttt1 );
//	fprintf( stderr, "seq2 = %s\n", ttt2 );
	if( lgth2 <= 0 )
	{
		fprintf( stderr, "==== jimei\n" );
		for( i=0; i<icyc; i++ ) 
		{
			strncpy( mseq1[i], seq1[i]+ist, lgth1 );
			mseq1[i][lgth1] = 0;
		}
		for( i=0; i<jcyc; i++ ) 
		{
			mseq2[i][0] = 0;
			for( j=0; j<lgth1; j++ )
				strcat( mseq2[i], "-" );
		}

//		fprintf( stderr, "==== mseq1[0] (%d) = %s\n", depth, mseq1[0] );
//		fprintf( stderr, "==== mseq2[0] (%d) = %s\n", depth, mseq2[0] );

		return( (float)offset * lgth1 );
	}

	aseq1 = AllocateCharMtx( icyc, lgth1+lgth2+100 );
	aseq2 = AllocateCharMtx( jcyc, lgth1+lgth2+100 );

	if( lgth1 < DPTANNI || lgth2 < DPTANNI )
	{
//		fprintf( stderr, "==== Going to _tanni\n" );

		wm = MSalignmm_tanni( icyc, jcyc, eff1, eff2, seq1, seq2, cpmx1, cpmx2, ist, ien, jst, jen, alloclen, aseq1, aseq2, gapinfo );	


		for( i=0; i<icyc; i++ ) strcpy( mseq1[i], aseq1[i] );
		for( i=0; i<jcyc; i++ ) strcpy( mseq2[i], aseq2[i] );

//		fprintf( stderr, "==== mseq1[0] (%d) = %s\n", depth, mseq1[0] );
//		fprintf( stderr, "==== mseq2[0] (%d) = %s\n", depth, mseq2[0] );

//		fprintf( stderr, "freeing aseq\n" );
		FreeCharMtx( aseq1 );
		FreeCharMtx( aseq2 );

		return( wm );
	}
//	fprintf( stderr, "Trying to divide the mtx\n" );

	ll1 = ( (int)(1.3*lgth1) ) + 100;
	ll2 = ( (int)(1.3*lgth2) ) + 100;

//	fprintf( stderr, "ll1,ll2=%d,%d\n", ll1, ll2 );

	w1 = AllocateFloatVec( ll2+2 );
	w2 = AllocateFloatVec( ll2+2 );
//	match = AllocateFloatVec( ll2+2 );
	midw = AllocateFloatVec( ll2+2 );
	midn = AllocateFloatVec( ll2+2 );
	midm = AllocateFloatVec( ll2+2 );
	jumpbacki = AllocateIntVec( ll2+2 );
	jumpbackj = AllocateIntVec( ll2+2 );
	jumpforwi = AllocateIntVec( ll2+2 );
	jumpforwj = AllocateIntVec( ll2+2 );
	jumpdummi = AllocateIntVec( ll2+2 );
	jumpdummj = AllocateIntVec( ll2+2 );

	initverticalw = AllocateFloatVec( ll1+2 );
	lastverticalw = AllocateFloatVec( ll1+2 );

	m = AllocateFloatVec( ll2+2 );
	mp = AllocateIntVec( ll2+2 );
	gaps = AllocateCharVec( MAX( ll1, ll2 ) + 2 );

	floatwork = AllocateFloatMtx( MAX( ll1, ll2 )+2, 26 ); 
	intwork = AllocateIntMtx( MAX( ll1, ll2 )+2, 26 ); 

#if DEBUG
	fprintf( stderr, "succeeded\n" );
#endif

#if STOREWM
	WMMTX = AllocateFloatMtx( ll1, ll2 );
	WMMTX2 = AllocateFloatMtx( ll1, ll2 );
#endif
#if 0
	shortmtx = AllocateShortMtx( ll1, ll2 );

#if DEBUG
	fprintf( stderr, "succeeded\n\n" );
#endif

	ijp = shortmtx;
#endif

	currentw = w1;
	previousw = w2;

	match_calc( initverticalw, cpmx2+jst, cpmx1+ist, 0, lgth1, floatwork, intwork, 1 );

	match_calc( currentw, cpmx1+ist, cpmx2+jst, 0, lgth2, floatwork, intwork, 1 );

	for( i=1; i<lgth1+1; i++ )
	{
//		initverticalw[i] += fpenalty  * ( ogcp1[0] + fgcp1[i-1] );
		initverticalw[i] += fpenalty;
	}
	for( j=1; j<lgth2+1; j++ )
	{
//		currentw[j] += fpenalty * ( ogcp2[0] + fgcp2[j-1] );
		currentw[j] += fpenalty;
	}

#if STOREWM
	WMMTX[0][0] = initverticalw[0];
	for( i=1; i<lgth1+1; i++ )
	{
		WMMTX[i][0] = initverticalw[i];
	}
	for( j=1; j<lgth2+1; j++ )
	{
		WMMTX[0][j] = currentw[j];
	}
#endif

	for( j=1; j<lgth2+1; ++j ) 
	{
//		m[j] = currentw[j-1] + fpenalty * ogcp1[1];
		m[j] = currentw[j-1];
		mp[j] = 0;
	}

	lastverticalw[0] = currentw[lgth2-1];

	imid = lgth1 / 2;

	jumpi = 0; // atode kawaru.
	lasti = lgth1+1;
//	for( i=1; i<lasti; i++ )
	for( i=1; i<=imid; i++ )
	{
		wtmp = previousw; 
		previousw = currentw;
		currentw = wtmp;

		previousw[0] = initverticalw[i-1];

		match_calc( currentw, cpmx1+ist, cpmx2+jst, i, lgth2, floatwork, intwork, 0 );
		currentw[0] = initverticalw[i];


//		mi = previousw[0] + fpenalty * ogcp2[1]; 
		mi = previousw[0];
		mpi = 0;

//		ijppt = ijp[i] + 1;
		mjpt = m + 1;
		prept = previousw;
		curpt = currentw + 1;
		mpjpt = mp + 1;


		lastj = lgth2+1;
		for( j=1; j<lastj; j++ )
		{

			wm = *prept;
//			*ijppt = 0;

#if 0
			fprintf( stderr, "%5.0f->", wm );
#endif
			g = mi + fpenalty;
#if 0
			fprintf( stderr, "%5.0f?", g );
#endif
			if( g > wm )
			{
				wm = g;
//				*ijppt = -( j - mpi );
			}
//			g = *prept + fpenalty * ogcp2[j];
			g = *prept;
			if( g >= mi )
			{
				mi = g;
				mpi = j-1;
			}
#if USE_PENALTY_EX
			mi += fpenalty_ex;
#endif

//			g = *mjpt + fpenalty  * fgcp1[i-1];
			g = *mjpt + fpenalty;
#if 0 
			fprintf( stderr, "%5.0f?", g );
#endif
			if( g > wm )
			{
				wm = g;
//				*ijppt = +( i - *mpjpt );
			}


//			g = *prept + fpenalty * ogcp1[i];
			g = *prept;
			if( g >= *mjpt )
			{
				*mjpt = g;
				*mpjpt = i-1;
			}
#if USE_PENALTY_EX
			m[j] += fpenalty_ex;
#endif

#if 0
			fprintf( stderr, "%5.0f ", wm );
#endif
			*curpt += wm;

#if STOREWM
			WMMTX[i][j] = *curpt;
			WMMTX2[i][j] = *mjpt;
#endif

			if( i == imid ) //muda
			{	
				jumpbackj[j] = *mpjpt; // muda atode matomeru
				jumpbacki[j] = mpi; // muda atode matomeru
//				fprintf( stderr, "jumpbackj[%d] in forward dp is %d\n", j, *mpjpt );
//				fprintf( stderr, "jumpbacki[%d] in forward dp is %d\n", j, mpi );
				midw[j] = *curpt;
				midm[j] = *mjpt;
				midn[j] = mi;
			}

			mjpt++;
			prept++;
			mpjpt++;
			curpt++;

		}
		lastverticalw[i] = currentw[lgth2-1];
#if 0  // ue
		if( i == imid )
		{
			for( j=0; j<lgth2; j++ ) midw[j] = currentw[j];
			for( j=0; j<lgth2; j++ ) midm[j] = m[j];
		}
#endif
	}
//	for( j=0; j<lgth2; j++ ) midw[j] = WMMTX[imid][j];
//	for( j=0; j<lgth2; j++ ) midm[j] = WMMTX2[imid][j];

#if 0
    for( i=0; i<lgth1; i++ )
    {
        for( j=0; j<lgth2; j++ )
        {
            fprintf( stderr, "% 10.2f ", WMMTX[i][j] );
        }
        fprintf( stderr, "\n" );
    }
	fprintf( stderr, "\n" );
	fprintf( stderr, "WMMTX2 = \n" );
    for( i=0; i<lgth1; i++ )
    {
        for( j=0; j<lgth2; j++ )
        {
            fprintf( stderr, "% 10.2f ", WMMTX2[i][j] );
        }
        fprintf( stderr, "\n" );
    }
	fprintf( stderr, "\n" );
#endif

// gyakudp

	match_calc( initverticalw, cpmx2+jst, cpmx1+ist, lgth2-1, lgth1, floatwork, intwork, 1 );
	match_calc( currentw, cpmx1+ist, cpmx2+jst, lgth1-1, lgth2, floatwork, intwork, 1 );

	for( i=0; i<lgth1-1; i++ )
	{
//		initverticalw[i] += fpenalty * ( fgcp1[lgth1-1] + ogcp1[i+1] );
		initverticalw[i] += fpenalty;
	}
	for( j=0; j<lgth2-1; j++ )
	{
//		currentw[j] += fpenalty * ( fgcp2[lgth1-1] + ogcp2[j+1] );
		currentw[j] += fpenalty;
	}

#if STOREWM
	for( i=0; i<lgth1-1; i++ )
	{
		WMMTX[i][lgth2-1] += fpenalty;
	}
	for( j=0; j<lgth2-1; j++ )
	{
		WMMTX[lgth1-1][j] += fpenalty;
	}
#endif

	for( j=lgth2-1; j>0; --j )
	{
//		m[j-1] = currentw[j] + fpenalty * fgcp2[lgth2-2];
		m[j-1] = currentw[j];
		mp[j] = lgth1-1;
	}

//	for( j=0; j<lgth2; j++ ) m[j] = 0.0;
	// m[lgth2-1] ha irunoka?

//	for( i=lgth1-2; i>=imid; i-- )
	for( i=lgth1-2; i>-1; i-- )
	{
		wtmp = previousw;
		previousw = currentw;
		currentw = wtmp;
		previousw[lgth2-1] = initverticalw[i+1];
//		match_calc( currentw, seq1, seq2, i, lgth2 );
		match_calc( currentw, cpmx1+ist, cpmx2+jst, i, lgth2, floatwork, intwork, 0 );

		currentw[lgth2-1] = initverticalw[i];

//		mi = previousw[lgth2-1] + fpenalty * fgcp2[lgth2-2];
		mi = previousw[lgth2-1];
		mpi = lgth2 - 1;

		mjpt = m + lgth2 - 2;
		prept = previousw + lgth2 - 1;
		curpt = currentw + lgth2 - 2;
		mpjpt = mp + lgth2 - 2;


		for( j=lgth2-2; j>-1; j-- )
		{
			wm = *prept;
			ijpi = i+1;
			ijpj = j+1;

//			g = mi + fpenalty * ogcp2[j+1];
			g = mi + fpenalty;
			if( g > wm )
			{
				wm = g;
				ijpj = mpi;
			}

//			g = *prept + fpenalty + fgcp2[j];
			g = *prept;
			if( g > mi )
			{
//				fprintf( stderr, "i,j=%d,%d - renewed! mpi = %d\n", i, j, j+1 );
				mi = g;
				mpi = j + 1;
			}

#if USE_PENALTY_EX
			mi += fpenalty_ex;
#endif

//			fprintf( stderr, "i,j=%d,%d *mpjpt = %d\n", i, j, *mpjpt );
//			g = *mjpt + fpenalty * ogcp1[i+1];
			g = *mjpt + fpenalty;
			if( g > wm )
			{
				wm = g;
				ijpi = *mpjpt;
			}

//			if( i == imid )fprintf( stderr, "i,j=%d,%d \n", i, j );
//			g = *prept + fpenalty * fgcp1[i];
			g = *prept;
			if( g > *mjpt )
			{
				*mjpt = g;
				*mpjpt = i + 1;
			}

#if USE_PENALTY_EX
			m[j] += fpenalty_ex;
#endif

			if( i == jumpi || i == imid - 1 )
			{
				jumpforwi[j] = ijpi; //muda
				jumpforwj[j] = ijpj; //muda
//				fprintf( stderr, "jumpfori[%d] = %d\n", j, ijpi );
//				fprintf( stderr, "jumpforj[%d] = %d\n", j, ijpj );
			}
			if( i == imid ) // muda
			{
				midw[j] += wm;
				midm[j+1] += *mjpt + fpenalty; //??????
			}
			if( i == imid - 1 )
			{
				midn[j] += mi + fpenalty;  //????
			}
#if STOREWM
			WMMTX[i][j] += wm;
			WMMTX2[i][j+1] += *mjpt + fpenalty;
#endif
			*curpt += wm;

			mjpt--;
			prept--;
			mpjpt--;
			curpt--;
		}

		if( i == imid - 1 )	
		{
			maxwm = midw[1];
			jmid = 1;
			for( j=2; j<lgth2; j++ )
			{
				wm = midw[j];
				if( wm > maxwm )
				{
					jmid = j;
					maxwm = wm;
				}
				wm = midm[j];
				if( wm > maxwm )
				{
					jmid = j;
					maxwm = wm;
				}
			}
//			fprintf( stderr, "imid=%d, jmid=%d\n", imid, jmid );
			wm = midw[jmid];
			{
				jumpi = imid-1;
				jumpj = jmid-1;
			}
			if( midm[jmid] > wm )
			{
				jumpi = jumpbackj[jmid];
				jumpj = jmid-1;
			}
			if( midn[jmid-1] >= wm )
			{
				jumpi = imid-1;
				jumpj = jumpbacki[jmid];
			}
#if 0
			fprintf( stderr, "imid = %d\n", imid );
			fprintf( stderr, "midn = \n" );
			for( j=0; j<lgth2; j++ )
			{
				fprintf( stderr, "% 10.2f ", midn[j] );
			}
			fprintf( stderr, "\n" );
			fprintf( stderr, "midw = \n" );
			for( j=0; j<lgth2; j++ )
			{
				fprintf( stderr, "% 10.2f ", midw[j] );
			}
			fprintf( stderr, "\n" );
			fprintf( stderr, "midm = \n" );
			for( j=0; j<lgth2; j++ )
			{
				fprintf( stderr, "% 10.2f ", midm[j] );
			}
			fprintf( stderr, "\n" );
#endif
		}
		if( i == jumpi ) //saki?
		{
//			fprintf( stderr, "imid, jumpi = %d,%d\n", imid, jumpi );
//			fprintf( stderr, "jmid, jumpj = %d,%d\n", jmid, jumpj );
			imid = jumpforwi[jumpj];
			jmid = jumpforwj[jumpj];
//			fprintf( stderr, "imid -> %d\n", imid );
//			fprintf( stderr, "jmid -> %d\n", jmid );
			break;
		}
	}
//	fprintf( stderr, "imid = %d, but jumpi = %d\n", imid, jumpi );
//	fprintf( stderr, "jmid = %d, but jumpj = %d\n", jmid, jumpj );

//	for( j=0; j<lgth2; j++ ) midw[j] += currentw[j];
//	for( j=0; j<lgth2; j++ ) midm[j] += m[j+1];
//	for( j=0; j<lgth2; j++ ) midw[j] += WMMTX[imid][j];
//	for( j=0; j<lgth2; j++ ) midw[j] += WMMTX[imid][j];


#if 0
	fprintf( stderr, "WMMTX = \n" );
    for( i=0; i<lgth1; i++ )
    {
        fprintf( stderr, "%d ", i );
        for( j=0; j<lgth2; j++ )
        {
            fprintf( stderr, "% 10.2f ", WMMTX[i][j] );
        }
        fprintf( stderr, "\n" );
    }
	fprintf( stderr, "WMMTX2 = (p = %f)\n", fpenalty );
    for( i=0; i<lgth1; i++ )
    {
        fprintf( stderr, "%d ", i );
        for( j=0; j<lgth2; j++ )
        {
            fprintf( stderr, "% 10.2f ", WMMTX2[i][j] );
        }
        fprintf( stderr, "\n" );
    }

	fprintf( stderr, "jumpbacki = \n" );
	for( j=0; j<lgth2; j++ )
	{
		fprintf( stderr, "% 10d ", jumpbacki[j] );
	}
	fprintf( stderr, "\n" );
	fprintf( stderr, "jumpbackj = \n" );
	for( j=0; j<lgth2; j++ )
	{
		fprintf( stderr, "% 10d ", jumpbackj[j] );
	}
	fprintf( stderr, "\n" );
	fprintf( stderr, "jumpforwi = \n" );
	for( j=0; j<lgth2; j++ )
	{
		fprintf( stderr, "% 10d ", jumpforwi[j] );
	}
	fprintf( stderr, "\n" );
	fprintf( stderr, "jumpforwj = \n" );
	for( j=0; j<lgth2; j++ )
	{
		fprintf( stderr, "% 10d ", jumpforwj[j] );
	}
	fprintf( stderr, "\n" );


#endif

//	Atracking( currentw, lastverticalw, seq1, seq2, mseq1, mseq2, cpmx1, cpmx2, ijp, icyc, jcyc );

#if 0 // irukamo
	resultlen = strlen( mseq1[0] );
	if( alloclen < resultlen || resultlen > N )
	{
		fprintf( stderr, "alloclen=%d, resultlen=%d, N=%d\n", alloclen, resultlen, N );
		ErrorExit( "LENGTH OVER!\n" );
	}
#endif



#if 0
	fprintf( stderr, "jumpi = %d, imid = %d\n", jumpi, imid );
	fprintf( stderr, "jumpj = %d, jmid = %d\n", jumpj, jmid );

	fprintf( stderr, "imid = %d\n", imid );
	fprintf( stderr, "jmid = %d\n", jmid );
#endif


	FreeFloatVec( w1 );
	FreeFloatVec( w2 );
	FreeFloatVec( initverticalw );
	FreeFloatVec( lastverticalw );
	FreeFloatVec( midw );
	FreeFloatVec( midm );
	FreeFloatVec( midn );

	FreeIntVec( jumpbacki );
	FreeIntVec( jumpbackj );
	FreeIntVec( jumpforwi );
	FreeIntVec( jumpforwj );
	FreeIntVec( jumpdummi );
	FreeIntVec( jumpdummj );

	FreeFloatVec( m );
	FreeIntVec( mp );

	FreeFloatMtx( floatwork );
	FreeIntMtx( intwork );

#if STOREWM
	FreeFloatMtx( WMMTX );
	FreeFloatMtx( WMMTX2 );
#endif

//	fprintf( stderr, "==== calling myself (first)\n" );

	MSalignmm_rec( icyc, jcyc, eff1, eff2, seq1, seq2, cpmx1, cpmx2, ist, ist+jumpi, jst, jst+jumpj, alloclen, aseq1, aseq2, depth, gapinfo );	
	for( i=0; i<icyc; i++ ) strcpy( mseq1[i], aseq1[i] );
	for( i=0; i<jcyc; i++ ) strcpy( mseq2[i], aseq2[i] );

//	fprintf( stderr, "====(f) aseq1[0] (%d) = %s (%d-%d)\n", depth, aseq1[0], ist, ien );
//	fprintf( stderr, "====(f) aseq2[0] (%d) = %s (%d-%d)\n", depth, aseq2[0], jst, jen );

	len = strlen( mseq1[0] );
//	fprintf( stderr, "len = %d\n", len );
	l = jmid - jumpj - 1;
//	fprintf( stderr, "l=%d\n", l );
	if( l > 0 )
	{
		for( i=0; i<l; i++ ) gaps[i] = '-'; gaps[i] = 0;
		for( i=0; i<icyc; i++ ) 
		{
			strcat( mseq1[i], gaps );
			mseq1[i][len+l] = 0;
		}
		for( j=0; j<jcyc; j++ )
		{
			strncat( mseq2[j], seq2[j]+jst+jumpj+1, l );
			mseq2[j][len+l] = 0;
		}
	}
	len = strlen( mseq1[0] );
	l = imid - jumpi - 1;
//	fprintf( stderr, "l=%d\n", l );
	if( l > 0 )
	{
		for( i=0; i<l; i++ ) gaps[i] = '-'; gaps[i] = 0;
		for( i=0; i<icyc; i++ )
		{
			strncat( mseq1[i], seq1[i]+ist+jumpi+1, l );
			mseq1[i][len+l] = 0;
		}
		for( j=0; j<jcyc; j++ ) 
		{
			strcat( mseq2[j], gaps );
			mseq2[j][len+l] = 0;
		}
	}
#if DEBUG
	fprintf( stderr, "after gapfill mseq1[0]=%s\n", mseq1[0] );
	fprintf( stderr, "after gapfill mseq2[0]=%s\n", mseq2[0] );
#endif

//	fprintf( stderr, "==== calling myself (second)\n" );

	MSalignmm_rec( icyc, jcyc, eff1, eff2, seq1, seq2, cpmx1, cpmx2, ist+imid, ien, jst+jmid, jen, alloclen, aseq1, aseq2, depth, gapinfo );	

	for( i=0; i<icyc; i++ ) strcat( mseq1[i], aseq1[i] );
	for( i=0; i<jcyc; i++ ) strcat( mseq2[i], aseq2[i] );

//	fprintf( stderr, "====(s) aseq1[0] (%d) = %s (%d-%d)\n", depth, aseq1[0], ist, ien );
//	fprintf( stderr, "====(s) aseq2[0] (%d) = %s (%d-%d)\n", depth, aseq2[0], jst, jen );

//	fprintf( stderr, "==== mseq1[0] (%d) = %s\n", depth, mseq1[0] );
//	fprintf( stderr, "==== mseq2[0] (%d) = %s\n", depth, mseq2[0] );

	free( gaps );
	FreeCharMtx( aseq1 );
	FreeCharMtx( aseq2 );
	
	return( wm );
}



float MSalignmm( char **seq1, char **seq2, double *eff1, double *eff2, int icyc, int jcyc, int alloclen )
/* score no keisan no sai motokaraaru gap no atukai ni mondai ga aru */
{
//	int k;
	int i, j;
	int ll1, ll2;
	int lgth1, lgth2;
	float wm;   /* int ?????? */
	static char **mseq1 = NULL;
	static char **mseq2 = NULL;
	char **mseq;
	float *ogcp1;
	float *ogcp2;
	float *fgcp1;
	float *fgcp2;
	float **cpmx1;
	float **cpmx2;
	float *gapinfo[4];

#if 0
	fprintf( stderr, "eff in SA+++align\n" );
	for( i=0; i<icyc; i++ ) fprintf( stderr, "eff1[%d] = %f\n", i, eff1[i] );
#endif
	if( mseq1 == NULL )
	{
		mseq1 = AllocateCharMtx( njob, 0 );
		mseq2 = AllocateCharMtx( njob, 0 );
	}

	lgth1 = strlen( seq1[0] );
	lgth2 = strlen( seq2[0] );

	ll1 = ( (int)(1.3*lgth1) ) + 100;
	ll2 = ( (int)(1.3*lgth2) ) + 100;

	mseq = AllocateCharMtx( njob, ll1+ll2 );

	ogcp1 = AllocateFloatVec( ll1+2 );
	ogcp2 = AllocateFloatVec( ll2+2 );
	fgcp1 = AllocateFloatVec( ll1+2 );
	fgcp2 = AllocateFloatVec( ll2+2 );

	cpmx1 = AllocateFloatMtx( ll1+2, 26 );
	cpmx2 = AllocateFloatMtx( ll2+2, 26 );

	for( i=0; i<icyc; i++ ) mseq1[i] = mseq[i];
	for( j=0; j<jcyc; j++ ) mseq2[j] = mseq[icyc+j];

	MScpmx_calc_new( seq1, cpmx1, eff1, lgth1, icyc );
	MScpmx_calc_new( seq2, cpmx2, eff2, lgth2, jcyc );

#if 1
	OpeningGapCount( ogcp1, icyc, seq1, eff1 );
	OpeningGapCount( ogcp2, jcyc, seq2, eff2 );
	FinalGapCount( fgcp1, icyc, seq1, eff1 );
	FinalGapCount( fgcp2, jcyc, seq2, eff2 );

	for( i=0; i<lgth1; i++ ) 
	{
		ogcp1[i] = 0.5 * ( 1.0 - ogcp1[i] );
		fgcp1[i] = 0.5 * ( 1.0 - fgcp1[i] );
	}
	for( i=0; i<lgth2; i++ ) 
	{
		ogcp2[i] = 0.5 * ( 1.0 - ogcp2[i] );
		fgcp2[i] = 0.5 * ( 1.0 - fgcp2[i] );
	}

	gapinfo[0] = ogcp1;
	gapinfo[1] = fgcp1;
	gapinfo[2] = ogcp2;
	gapinfo[3] = fgcp2;
#endif

#if 0
	fprintf( stdout, "in MSalignmm.c\n" );
	for( i=0; i<icyc; i++ )
	{
		fprintf( stdout, ">%d of GROUP1\n", i );
		fprintf( stdout, "%s\n", seq1[i] );
	}
	for( i=0; i<jcyc; i++ )
	{
		fprintf( stdout, ">%d of GROUP2\n", i );
		fprintf( stdout, "%s\n", seq2[i] );
	}
	fflush( stdout );
#endif

	wm = MSalignmm_rec( icyc, jcyc, eff1, eff2, seq1, seq2, cpmx1, cpmx2, 0, lgth1-1, 0, lgth2-1, alloclen, mseq1, mseq2, 0, gapinfo );

	for( i=0; i<icyc; i++ ) strcpy( seq1[i], mseq1[i] );
	for( i=0; i<jcyc; i++ ) strcpy( seq2[i], mseq2[i] );


	FreeFloatVec( ogcp1 );
	FreeFloatVec( ogcp2 );
	FreeFloatVec( fgcp1 );
	FreeFloatVec( fgcp2 );
	FreeFloatMtx( cpmx1 );
	FreeFloatMtx( cpmx2 );

	FreeCharMtx( mseq );

	return( wm );
}