===================================================================
RCS file: /home/cvs/OpenXM_contrib2/asir2000/engine/_distm.c,v
retrieving revision 1.3
retrieving revision 1.16
diff -u -p -r1.3 -r1.16
--- OpenXM_contrib2/asir2000/engine/_distm.c	2000/08/22 05:04:05	1.3
+++ OpenXM_contrib2/asir2000/engine/_distm.c	2017/08/31 02:36:21	1.16
@@ -45,54 +45,29 @@
  * DEVELOPER SHALL HAVE NO LIABILITY IN CONNECTION WITH THE USE,
  * PERFORMANCE OR NON-PERFORMANCE OF THE SOFTWARE.
  *
- * $OpenXM: OpenXM_contrib2/asir2000/engine/_distm.c,v 1.2 2000/08/21 08:31:27 noro Exp $ 
+ * $OpenXM: OpenXM_contrib2/asir2000/engine/_distm.c,v 1.15 2012/12/17 07:20:44 noro Exp $ 
 */
 #include "ca.h"
 #include "inline.h"
 
-#define NV(p) ((p)->nv)
-#define C(p) ((p)->c)
-#if 0
-#define ITOS(p) (((unsigned int)(p))>>1)
-#define STOI(i) ((P)((((unsigned int)(i))<<1)|1))
-#else
-#define ITOS(p) (((unsigned int)(p))&0x7fffffff)
-#define STOI(i) ((P)((unsigned int)(i)|0x80000000))
-#endif
-
 extern int (*cmpdl)();
 extern int do_weyl;
 
-void _dptodp();
-void adddl_dup();
-void _mulmdm_dup();
-void _free_dp();
-void _comm_mulmd_dup();
-void _weyl_mulmd_dup();
-void _weyl_mulmmm_dup();
-void _weyl_mulmdm_dup();
-
 MP _mp_free_list;
 DP _dp_free_list;
 DL _dl_free_list;
 int current_dl_length;
 
-#define _NEWDL_NOINIT(d,n) if ((n)!= current_dl_length){_dl_free_list=0; current_dl_length=(n);} if(!_dl_free_list)_DL_alloc(); (d)=_dl_free_list; _dl_free_list = *((DL *)_dl_free_list)
-#define _NEWDL(d,n) if ((n)!= current_dl_length){_dl_free_list=0; current_dl_length=(n);} if(!_dl_free_list)_DL_alloc(); (d)=_dl_free_list; _dl_free_list = *((DL *)_dl_free_list); bzero((d),(((n)+1)*sizeof(int)))
-#define _NEWMP(m) if(!_mp_free_list)_MP_alloc(); (m)=_mp_free_list; _mp_free_list = NEXT(_mp_free_list)
-#define _MKDP(n,m,d) if(!_dp_free_list)_DP_alloc(); (d)=_dp_free_list; _dp_free_list = (DP)BDY(_dp_free_list); (d)->nv=(n); BDY(d)=(m)
+void GC_gcollect();
 
-#define _NEXTMP(r,c) \
-if(!(r)){_NEWMP(r);(c)=(r);}else{_NEWMP(NEXT(c));(c)=NEXT(c);}
+void _free_private_storage()
+{
+	_mp_free_list = 0;
+	_dp_free_list = 0;
+	_dl_free_list = 0;
+	GC_gcollect();
+}
 
-#define _NEXTMP2(r,c,s) \
-if(!(r)){(c)=(r)=(s);}else{NEXT(c)=(s);(c)=(s);}
-
-#define FREEDL(m) *((DL *)m)=_dl_free_list; _dl_free_list=(m)
-#define FREEMP(m) NEXT(m)=_mp_free_list; _mp_free_list=(m)
-#define FREEDP(m) BDY(m)=(MP)_dp_free_list; _dp_free_list=(m)
-
-
 void _DL_alloc()
 {
 	int *p;
@@ -101,8 +76,12 @@ void _DL_alloc()
 
 /*	fprintf(stderr,"DL_alloc : %d \n",++DL_alloc_count); */
 	dl_len = (current_dl_length+1);
-	p = (int *)GC_malloc(128*dl_len*sizeof(int));
+#if SIZEOF_LONG == 8
+	if ( dl_len & 1 )
+		dl_len += 1;
+#endif
 	for ( i = 0; i < 128; i++, p += dl_len ) {
+		p = (int *)MALLOC(dl_len*sizeof(int));
 		*(DL *)p = _dl_free_list;
 		_dl_free_list = (DL)p;
 	}
@@ -115,9 +94,9 @@ void _MP_alloc()
 	static int MP_alloc_count;
 
 /*	fprintf(stderr,"MP_alloc : %d \n",++MP_alloc_count); */
-	p = (MP)GC_malloc(1024*sizeof(struct oMP));
 	for ( i = 0; i < 1024; i++ ) {
-		p[i].next = _mp_free_list; _mp_free_list = &p[i];
+		p = (MP)MALLOC(sizeof(struct oMP));
+		p->next = _mp_free_list; _mp_free_list = p;
 	}
 }
 		
@@ -128,17 +107,15 @@ void _DP_alloc()
 	static int DP_alloc_count;
 
 /*	fprintf(stderr,"DP_alloc : %d \n",++DP_alloc_count); */
-	p = (DP)GC_malloc(1024*sizeof(struct oDP));
 	for ( i = 0; i < 1024; i++ ) {
-		p[i].body = (MP)_dp_free_list; _dp_free_list = &p[i];
+		p = (DP)MALLOC(sizeof(struct oDP));
+		p->body = (MP)_dp_free_list; _dp_free_list = p;
 	}
 }
 
 /* merge p1 and p2 into pr */
 
-void _addmd_destructive(mod,p1,p2,pr)
-int mod;
-DP p1,p2,*pr;
+void _addmd_destructive(int mod,DP p1,DP p2,DP *pr)
 {
 	int n;
 	MP m1,m2,mr,mr0,s;
@@ -157,11 +134,11 @@ DP p1,p2,*pr;
 						t += mod;
 					s = m1; m1 = NEXT(m1);
 					if ( t ) {
-						_NEXTMP2(mr0,mr,s); C(mr) = STOI(t);
+						_NEXTMP2(mr0,mr,s); C(mr) = (Obj)STOI(t);
 					} else {
-						FREEDL(s->dl); FREEMP(s);
+						_FREEDL(s->dl); _FREEMP(s);
 					}
-					s = m2; m2 = NEXT(m2); FREEDL(s->dl); FREEMP(s);
+					s = m2; m2 = NEXT(m2); _FREEDL(s->dl); _FREEMP(s);
 					break;
 				case 1:
 					s = m1; m1 = NEXT(m1); _NEXTMP2(mr0,mr,s);
@@ -188,13 +165,11 @@ DP p1,p2,*pr;
 		_MKDP(NV(p1),mr0,*pr);
 		if ( *pr )
 			(*pr)->sugar = MAX(p1->sugar,p2->sugar);
-		FREEDP(p1); FREEDP(p2);
+		_FREEDP(p1); _FREEDP(p2);
 	}
 }
 
-void _mulmd_dup(mod,p1,p2,pr)
-int mod;
-DP p1,p2,*pr;
+void _mulmd_dup(int mod,DP p1,DP p2,DP *pr)
 {
 	if ( !do_weyl )
 		_comm_mulmd_dup(mod,p1,p2,pr);
@@ -202,9 +177,7 @@ DP p1,p2,*pr;
 		_weyl_mulmd_dup(mod,p1,p2,pr);
 }
 
-void _comm_mulmd_dup(mod,p1,p2,pr)
-int mod;
-DP p1,p2,*pr;
+void _comm_mulmd_dup(int mod,DP p1,DP p2,DP *pr)
 {
 	MP m;
 	DP s,t,u;
@@ -222,7 +195,7 @@ DP p1,p2,*pr;
 			l = l1;
 		}
 		if ( l > wlen ) {
-			if ( w ) GC_free(w);
+			if ( w ) GCFREE(w);
 			w = (MP *)MALLOC(l*sizeof(MP));
 			wlen = l;
 		}
@@ -236,44 +209,38 @@ DP p1,p2,*pr;
 	}
 }
 
-void _weyl_mulmd_dup(mod,p1,p2,pr)
-int mod;
-DP p1,p2,*pr;
+void _weyl_mulmd_dup(int mod,DP p1,DP p2,DP *pr)
 {
 	MP m;
 	DP s,t,u;
-	int i,l,l1;
+	int i,l;
 	static MP *w;
 	static int wlen;
 
 	if ( !p1 || !p2 )
 		*pr = 0;
 	else {
-		for ( m = BDY(p2), l = 0; m; m = NEXT(m), l++ );
+		for ( m = BDY(p1), l = 0; m; m = NEXT(m), l++ );
 		if ( l > wlen ) {
-			if ( w ) GC_free(w);
+			if ( w ) GCFREE(w);
 			w = (MP *)MALLOC(l*sizeof(MP));
 			wlen = l;
 		}
-		for ( m = BDY(p2), i = 0; i < l; m = NEXT(m), i++ )
+		for ( m = BDY(p1), i = 0; i < l; m = NEXT(m), i++ )
 			w[i] = m;
 		for ( s = 0, i = l-1; i >= 0; i-- ) {
-			_weyl_mulmdm_dup(mod,p1,w[i],&t); _addmd_destructive(mod,s,t,&u); s = u;
+			_weyl_mulmdm_dup(mod,w[i],p2,&t); _addmd_destructive(mod,s,t,&u); s = u;
 		}
 		bzero(w,l*sizeof(MP));
 		*pr = s;
 	}
 }
 
-void _mulmdm_dup(mod,p,m0,pr)
-int mod;
-DP p;
-MP m0;
-DP *pr;
+void _mulmdm_dup(int mod,DP p,MP m0,DP *pr)
 {
 	MP m,mr,mr0;
 	DL d,dt,dm;
-	int c,n,r,i;
+	int c,n,i,c1,c2;
 	int *pt,*p1,*p2;
 
 	if ( !p )
@@ -282,7 +249,9 @@ DP *pr;
 		for ( mr0 = 0, m = BDY(p), c = ITOS(C(m0)), d = m0->dl, n = NV(p); 
 			m; m = NEXT(m) ) {
 			_NEXTMP(mr0,mr);
-			C(mr) = STOI(dmar(ITOS(C(m)),c,0,mod));
+			c1 = ITOS(C(m));
+			DMAR(c1,c,0,mod,c2);
+			C(mr) = (Obj)STOI(c2);
 			_NEWDL_NOINIT(dt,n); mr->dl = dt;
 			dm = m->dl;
 			dt->td = d->td + dm->td;
@@ -295,171 +264,242 @@ DP *pr;
 	}
 }
 
-void _weyl_mulmdm_dup(mod,p,m0,pr)
-int mod;
-DP p;
-MP m0;
-DP *pr;
+void _weyl_mulmdm_dup(int mod,MP m0,DP p,DP *pr)
 {
 	DP r,t,t1;
 	MP m;
-	int n,l,i;
-	static MP *w;
+	DL d0;
+	int n,n2,l,i,j,tlen;
+	static MP *w,*psum;
+	static struct cdlm *tab;
 	static int wlen;
+	static int rtlen;
 
 	if ( !p )
 		*pr = 0;
 	else {
 		for ( m = BDY(p), l = 0; m; m = NEXT(m), l++ );
 		if ( l > wlen ) {
-			if ( w ) GC_free(w);
+			if ( w ) GCFREE(w);
 			w = (MP *)MALLOC(l*sizeof(MP));
 			wlen = l;
 		}
 		for ( m = BDY(p), i = 0; i < l; m = NEXT(m), i++ )
 			w[i] = m;
-		for ( r = 0, i = l-1, n = NV(p); i >= 0; i-- ) {
-			_weyl_mulmmm_dup(mod,w[i],m0,n,&t);
-			_addmd_destructive(mod,r,t,&t1); r = t1;
+		n = NV(p); n2 = n>>1;
+		d0 = m0->dl;
+
+		for ( i = 0, tlen = 1; i < n2; i++ )
+			tlen *= d0->d[n2+i]+1;
+		if ( tlen > rtlen ) {
+			if ( tab ) GCFREE(tab);
+			if ( psum ) GCFREE(psum);
+			rtlen = tlen;
+			tab = (struct cdlm *)MALLOC(rtlen*sizeof(struct cdlm));
+			psum = (MP *)MALLOC(rtlen*sizeof(MP));
 		}
-		bzero(w,l*sizeof(MP));
+		bzero(psum,tlen*sizeof(MP));
+		for ( i = l-1; i >= 0; i-- ) {
+			bzero(tab,tlen*sizeof(struct cdlm));
+			_weyl_mulmmm_dup(mod,m0,w[i],n,tab,tlen);
+			for ( j = 0; j < tlen; j++ ) {
+				if ( tab[j].c ) {
+					_NEWMP(m); m->dl = tab[j].d;
+					C(m) = (Obj)STOI(tab[j].c); NEXT(m) = psum[j];
+					psum[j] = m;
+				}
+			}
+		}
+		for ( j = tlen-1, r = 0; j >= 0; j-- ) 
+			if ( psum[j] ) {
+				_MKDP(n,psum[j],t); _addmd_destructive(mod,r,t,&t1); r = t1;
+			}
 		if ( r )
 			r->sugar = p->sugar + m0->dl->td;
 		*pr = r;
 	}
 }
+
 /* m0 = x0^d0*x1^d1*... * dx0^d(n/2)*dx1^d(n/2+1)*... */
 
-void _weyl_mulmmm_dup(mod,m0,m1,n,pr)
-int mod;
-MP m0,m1;
-int n;
-DP *pr;
+void _weyl_mulmmm_dup(int mod,MP m0,MP m1,int n,struct cdlm *rtab,int rtablen)
 {
-	MP m,mr,mr0;
-	DP r,t,t1;
-	int c,c0,c1,cc;
-	DL d,d0,d1;
-	int i,j,a,b,k,l,n2,s,min,h;
-	static int *tab;
+	int c,c0,c1;
+	DL d,d0,d1,dt;
+	int i,j,a,b,k,l,n2,s,min,curlen;
+	struct cdlm *p;
+	static int *ctab;
+	static struct cdlm *tab;
 	static int tablen;
+	static struct cdlm *tmptab;
+	static int tmptablen;
 
-	if ( !m0 || !m1 )
-		*pr = 0;
-	else {
-		c0 = ITOS(C(m0)); c1 = ITOS(C(m1));
-		c = dmar(c0,c1,0,mod);
-		d0 = m0->dl; d1 = m1->dl;
-		n2 = n>>1;
+	if ( !m0 || !m1 ) {
+		rtab[0].c = 0;
+		rtab[0].d = 0;
+		return;
+	}
+	c0 = ITOS(C(m0)); c1 = ITOS(C(m1));
+	c = dmar(c0,c1,0,mod);
+	d0 = m0->dl; d1 = m1->dl;
+	n2 = n>>1;
+	curlen = 1;
 
-		_NEWDL(d,n);
+	_NEWDL(d,n);
+	if ( n & 1 )
+		/* offset of h-degree */
+		d->td = d->d[n-1] = d0->d[n-1]+d1->d[n-1];
+	else
+		d->td = 0;
+	rtab[0].c = c;
+	rtab[0].d = d;
+
+	if ( rtablen > tmptablen ) {
+		if ( tmptab ) GCFREE(tmptab);
+		tmptab = (struct cdlm *)MALLOC(rtablen*sizeof(struct cdlm));
+		tmptablen = rtablen; 
+	}
+
+	for ( i = 0; i < n2; i++ ) {
+		a = d0->d[i]; b = d1->d[n2+i];
+		k = d0->d[n2+i]; l = d1->d[i];
+
+		/* degree of xi^a*(Di^k*xi^l)*Di^b */
+		a += l;
+		b += k;
+		s = MUL_WEIGHT(a,i)+MUL_WEIGHT(b,n2+i);
+
+		if ( !k || !l ) {
+			for ( j = 0, p = rtab; j < curlen; j++, p++ ) {
+				if ( p->c ) {
+					dt = p->d;
+					dt->d[i] = a;
+					dt->d[n2+i] = b;
+					dt->td += s;
+				}
+			}
+			curlen *= k+1;
+			continue;
+		}
+		if ( k+1 > tablen ) {
+			if ( tab ) GCFREE(tab);
+			if ( ctab ) GCFREE(ctab);
+			tablen = k+1;
+			tab = (struct cdlm *)MALLOC(tablen*sizeof(struct cdlm));
+			ctab = (int *)MALLOC(tablen*sizeof(int));
+		}
+		/* compute xi^a*(Di^k*xi^l)*Di^b */
+		min = MIN(k,l);
+		mkwcm(k,l,mod,ctab);
+		bzero(tab,(k+1)*sizeof(struct cdlm));
+		/* n&1 != 0 => homogenized computation; dx-xd=h^2 */
 		if ( n & 1 )
-			/* offset of h-degree */
-			d->td = d->d[n-1] = d0->d[n-1]+d1->d[n-1];
+			for ( j = 0; j <= min; j++ ) {
+				_NEWDL(d,n);
+				d->d[i] = a-j; d->d[n2+i] = b-j;
+				d->td = s;
+				d->d[n-1] = s-(MUL_WEIGHT(a-j,i)+MUL_WEIGHT(b-j,n2+i)); 
+				tab[j].d = d;
+				tab[j].c = ctab[j];
+			}
 		else
-			d->td = 0;
-		_NEWMP(mr); mr->c = STOI(c); mr->dl = d; NEXT(mr) = 0;
-		_MKDP(n,mr,r); r->sugar = d->td;
+			for ( j = 0; j <= min; j++ ) {
+				_NEWDL(d,n);
+				d->d[i] = a-j; d->d[n2+i] = b-j;
+				d->td = MUL_WEIGHT(a-j,i)+MUL_WEIGHT(b-j,n2+i);  /* XXX */
+				tab[j].d = d;
+				tab[j].c = ctab[j];
+			}
+#if 0
+		_comm_mulmd_tab(mod,n,rtab,curlen,tab,k+1,tmptab);
+		for ( j = 0; j < curlen; j++ )
+			if ( rtab[j].d ) { _FREEDL(rtab[j].d); }
+		for ( j = 0; j <= min; j++ )
+			if ( tab[j].d ) { _FREEDL(tab[j].d); }
+		curlen *= k+1;
+		bcopy(tmptab,rtab,curlen*sizeof(struct cdlm));
+#else
+		_comm_mulmd_tab_destructive(mod,n,rtab,curlen,tab,k+1);
+		for ( j = 0; j <= min; j++ )
+			if ( tab[j].d ) { _FREEDL(tab[j].d); }
+		curlen *= k+1;
+#endif
+	}
+}
 
-		/* homogenized computation; dx-xd=h^2 */
-		for ( i = 0; i < n2; i++ ) {
-			a = d0->d[i]; b = d1->d[n2+i];
-			k = d0->d[n2+i]; l = d1->d[i];
-			/* degree of xi^a*(Di^k*xi^l)*Di^b */
-			s = a+k+l+b;
-			/* compute xi^a*(Di^k*xi^l)*Di^b */
-			min = MIN(k,l);
+/* direct product of two cdlm tables
+  rt[] = [
+    t[0]*t1[0],...,t[n-1]*t1[0],
+    t[0]*t1[1],...,t[n-1]*t1[1],
+    ...
+    t[0]*t1[n1-1],...,t[n-1]*t1[n1-1]
+  ]
+*/
 
-			if ( min+1 > tablen ) {
-				if ( tab ) GC_free(tab);
-				tab = (int *)MALLOC((min+1)*sizeof(int));
-				tablen = min+1;
+void _comm_mulmd_tab(int mod,int nv,struct cdlm *t,int n,struct cdlm *t1,int n1,struct cdlm *rt)
+{
+	int i,j;
+	struct cdlm *p;
+	int c;
+	DL d;
+
+	bzero(rt,n*n1*sizeof(struct cdlm));
+	for ( j = 0, p = rt; j < n1; j++ ) {
+		c = t1[j].c;
+		d = t1[j].d;
+		if ( !c )
+			break;
+		for ( i = 0; i < n; i++, p++ ) {
+			if ( t[i].c ) {
+				p->c = dmar(t[i].c,c,0,mod);
+				_adddl_dup(nv,t[i].d,d,&p->d);
 			}
-			mkwcm(k,l,mod,tab);
-			if ( n & 1 )
-				for ( mr0 = 0, j = 0; j <= min; j++ ) {
-					_NEXTMP(mr0,mr); _NEWDL(d,n);
-					d->d[i] = l-j+a; d->d[n2+i] = k-j+b;
-					d->td = s;
-					d->d[n-1] = s-(d->d[i]+d->d[n2+i]); 
-					mr->c = STOI(tab[j]); mr->dl = d;
-				}
-			else
-				for ( mr0 = 0, s = 0, j = 0; j <= min; j++ ) {
-					_NEXTMP(mr0,mr); _NEWDL(d,n);
-					d->d[i] = l-j+a; d->d[n2+i] = k-j+b;
-					d->td = d->d[i]+d->d[n2+i]; /* XXX */
-					s = MAX(s,d->td); /* XXX */
-					mr->c = STOI(tab[j]); mr->dl = d;
-				}
-			bzero(tab,(min+1)*sizeof(int));
-			if ( mr0 )
-				NEXT(mr) = 0;
-			_MKDP(n,mr0,t);
-			if ( t )
-				t->sugar = s;
-			_comm_mulmd_dup(mod,r,t,&t1); 
-			_free_dp(r); _free_dp(t);
-			r = t1;
 		}
-		*pr = r;
 	}
 }
 
-void _dp_red_mod_destructive(p1,p2,mod,rp)
-DP p1,p2;
-int mod;
-DP *rp;
+void _comm_mulmd_tab_destructive(int mod,int nv,struct cdlm *t,int n,struct cdlm *t1,int n1)
 {
-	int i,n;
-	DL d1,d2,d;
-	MP m;
-	DP t,s;
-	int c,c1;
+	int i,j;
+	struct cdlm *p;
+	int c;
+	DL d;
 
-	n = p1->nv; d1 = BDY(p1)->dl; d2 = BDY(p2)->dl;
-	_NEWDL(d,n); d->td = d1->td - d2->td;
-	for ( i = 0; i < n; i++ )
-		d->d[i] = d1->d[i]-d2->d[i];
-	c = invm(ITOS(BDY(p2)->c),mod); c1 = dmar(c,ITOS(BDY(p1)->c),0,mod);
-	_NEWMP(m); m->dl = d; m->c = STOI(mod-c1); NEXT(m) = 0;
-	_MKDP(n,m,s); s->sugar = d->td;
-	_mulmd_dup(mod,s,p2,&t); _free_dp(s);
-	_addmd_destructive(mod,p1,t,rp);
+	for ( j = 1, p = t+n; j < n1; j++ ) {
+		c = t1[j].c;
+		d = t1[j].d;
+		if ( !c )
+			break;
+		for ( i = 0; i < n; i++, p++ ) {
+			if ( t[i].c ) {
+				p->c = dmar(t[i].c,c,0,mod);
+				_adddl_dup(nv,t[i].d,d,&p->d);
+			}
+		}
+	}
+	c = t1[0].c;
+	d = t1[0].d;
+	for ( i = 0, p = t; i < n; i++, p++ )
+		if ( t[i].c ) {
+			p->c = dmar(t[i].c,c,0,mod);
+			/* t[i].d += d */
+			adddl_destructive(nv,t[i].d,d);
+		}
 }
 
-void _dp_sp_mod_dup(p1,p2,mod,rp)
-DP p1,p2;
-int mod;
-DP *rp;
+void dlto_dl(DL d,DL *dr)
 {
-	int i,n,td;
-	int *w;
-	DL d1,d2,d;
-	MP m;
-	DP t,s,u;
+	int i,n;
+	DL t;
 
-	n = p1->nv; d1 = BDY(p1)->dl; d2 = BDY(p2)->dl;
-	w = (int *)ALLOCA(n*sizeof(int));
-	for ( i = 0, td = 0; i < n; i++ ) {
-		w[i] = MAX(d1->d[i],d2->d[i]); td += w[i];
-	}
-	_NEWDL(d,n); d->td = td - d1->td;
+	n = current_dl_length;
+	_NEWDL(t,n); *dr = t;
+	t->td = d->td;
 	for ( i = 0; i < n; i++ )
-		d->d[i] = w[i] - d1->d[i];
-	_NEWMP(m); m->dl = d; m->c = BDY(p2)->c; NEXT(m) = 0;
-	_MKDP(n,m,s); s->sugar = d->td; _mulmd_dup(mod,s,p1,&t); _free_dp(s);
-	_NEWDL(d,n); d->td = td - d2->td;
-	for ( i = 0; i < n; i++ )
-		d->d[i] = w[i] - d2->d[i];
-	_NEWMP(m); m->dl = d; m->c = STOI(mod - ITOS(BDY(p1)->c)); NEXT(m) = 0;
-	_MKDP(n,m,s); s->sugar = d->td; _mulmd_dup(mod,s,p2,&u); _free_dp(s);
-	_addmd_destructive(mod,t,u,rp);
+		t->d[i] = d->d[i];	
 }
 
-void _dltodl(d,dr)
-DL d;
-DL *dr;
+void _dltodl(DL d,DL *dr)
 {
 	int i,n;
 	DL t;
@@ -471,44 +511,68 @@ DL *dr;
 		t->d[i] = d->d[i];	
 }
 
-void adddl_dup(n,d1,d2,dr)
-int n;
-DL d1,d2;
-DL *dr;
+void _adddl_dup(int n,DL d1,DL d2,DL *dr)
 {
 	DL dt;
 	int i;
 
-	_NEWDL(dt,n); *dr = dt;
+	_NEWDL(dt,n);
+	*dr = dt;
 	dt->td = d1->td + d2->td;
 	for ( i = 0; i < n; i++ )
 		dt->d[i] = d1->d[i]+d2->d[i];
 }
 
-void _free_dp(f)
-DP f;
+void _free_dlarray(DL *a,int n)
 {
+	int i;
+
+	for ( i = 0; i < n; i++ ) { _FREEDL(a[i]); }
+}
+
+void _free_dp(DP f)
+{
 	MP m,s;
 
 	if ( !f )
 		return;
 	m = f->body;
 	while ( m ) {
-		s = m; m = NEXT(m); FREEDL(s->dl); FREEMP(s);
+		s = m; m = NEXT(m); _FREEDL(s->dl); _FREEMP(s);
 	}
-	FREEDP(f);
+	_FREEDP(f);
 }
 
-void _dptodp(p,r)
-DP p;
-DP *r;
+void dpto_dp(DP p,DP *r)
 {
 	MP m,mr0,mr;
+	DL t;
 
 	if ( !p )
 		*r = 0;
 	else {
+		/* XXX : dummy call to set current_dl_length */
+		_NEWDL_NOINIT(t,NV(p));
+
 		for ( m = BDY(p), mr0 = 0; m; m = NEXT(m) ) {
+			_NEXTMP(mr0,mr);
+			dlto_dl(m->dl,&mr->dl);
+			mr->c = m->c;
+		}
+		NEXT(mr) = 0;
+		_MKDP(p->nv,mr0,*r);
+		(*r)->sugar = p->sugar;
+	}
+}
+
+void _dptodp(DP p,DP *r)
+{
+	MP m,mr0,mr;
+
+	if ( !p )
+		*r = 0;
+	else {
+		for ( m = BDY(p), mr0 = 0; m; m = NEXT(m) ) {
 			NEXTMP(mr0,mr);
 			_dltodl(m->dl,&mr->dl);
 			mr->c = m->c;
@@ -519,61 +583,423 @@ DP *r;
 	}
 }
 
-void _dp_nf_mod_destructive(b,g,ps,mod,full,rp)
-NODE b;
-DP g;
-DP *ps;
-int mod,full;
-DP *rp;
+/* 
+ * destructive merge of two list
+ *
+ * p1, p2 : list of DL
+ * return : a merged list
+ */
+
+NODE _symb_merge(NODE m1,NODE m2,int n)
 {
-	DP u,p,d,s,t;
-	NODE l;
-	MP m,mr,mrd;
-	int sugar,psugar,n,h_reducible,i;
+	NODE top,prev,cur,m,t;
 
-	if ( !g ) {
-		*rp = 0; return;
+	if ( !m1 )
+		return m2;
+	else if ( !m2 )
+		return m1;
+	else {
+		switch ( (*cmpdl)(n,(DL)BDY(m1),(DL)BDY(m2)) ) {
+			case 0:
+				top = m1; _FREEDL((DL)BDY(m2)); m = NEXT(m2);
+				break;
+			case 1:
+				top = m1; m = m2;
+				break;
+			case -1:
+				top = m2; m = m1;
+				break;
+		}
+		prev = top; cur = NEXT(top);
+		/* BDY(prev) > BDY(m) always holds */
+		while ( cur && m ) {
+			switch ( (*cmpdl)(n,(DL)BDY(cur),(DL)BDY(m)) ) {
+				case 0:
+					_FREEDL(BDY(m)); m = NEXT(m);
+					prev = cur; cur = NEXT(cur);
+					break;
+				case 1:
+					t = NEXT(cur); NEXT(cur) = m; m = t;
+					prev = cur; cur = NEXT(cur);
+					break;
+				case -1:
+					NEXT(prev) = m; m = cur;
+					prev = NEXT(prev); cur = NEXT(prev);
+					break;
+			}
+		}
+		if ( !cur )
+			NEXT(prev) = m;
+		return top;
 	}
-	sugar = g->sugar;
-	n = g->nv;
-	for ( d = 0; g; ) {
-		for ( h_reducible = 0, l = b; l; l = NEXT(l) ) {
-			if ( dp_redble(g,p = ps[(int)BDY(l)]) ) {
-				h_reducible = 1;
-				psugar = (BDY(g)->dl->td - BDY(p)->dl->td) + p->sugar;
-				_dp_red_mod_destructive(g,p,mod,&u); g = u;
-				sugar = MAX(sugar,psugar);
-				if ( !g ) {
-					if ( d )
-						d->sugar = sugar;
-					_dptodp(d,rp); _free_dp(d); return;
+}
+
+/* merge p1 and p2 into pr */
+
+void _addd_destructive(VL vl,DP p1,DP p2,DP *pr)
+{
+	int n;
+	MP m1,m2,mr,mr0,s;
+	P t;
+
+	if ( !p1 )
+		*pr = p2;
+	else if ( !p2 )
+		*pr = p1;
+	else {
+		for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; )
+			switch ( (*cmpdl)(n,m1->dl,m2->dl) ) {
+				case 0:
+					addp(vl,(P)C(m1),(P)C(m2),&t);
+					s = m1; m1 = NEXT(m1);
+					if ( t ) {
+						_NEXTMP2(mr0,mr,s); C(mr) = (Obj)t;
+					} else {
+						_FREEDL(s->dl); _FREEMP(s);
+					}
+					s = m2; m2 = NEXT(m2); _FREEDL(s->dl); _FREEMP(s);
+					break;
+				case 1:
+					s = m1; m1 = NEXT(m1); _NEXTMP2(mr0,mr,s);
+					break;
+				case -1:
+					s = m2; m2 = NEXT(m2); _NEXTMP2(mr0,mr,s);
+					break;
+			}
+		if ( !mr0 )
+			if ( m1 )
+				mr0 = m1;
+			else if ( m2 )
+				mr0 = m2;
+			else {
+				*pr = 0;
+				return;
+			}
+		else if ( m1 )
+			NEXT(mr) = m1;
+		else if ( m2 )
+			NEXT(mr) = m2;
+		else
+			NEXT(mr) = 0;
+		_MKDP(NV(p1),mr0,*pr);
+		if ( *pr )
+			(*pr)->sugar = MAX(p1->sugar,p2->sugar);
+		_FREEDP(p1); _FREEDP(p2);
+	}
+}
+
+void _muld_dup(VL vl,DP p1,DP p2,DP *pr)
+{
+	if ( !do_weyl )
+		_comm_muld_dup(vl,p1,p2,pr);
+	else
+		_weyl_muld_dup(vl,p1,p2,pr);
+}
+
+void _comm_muld_dup(VL vl,DP p1,DP p2,DP *pr)
+{
+	MP m;
+	DP s,t,u;
+	int i,l,l1;
+	static MP *w;
+	static int wlen;
+
+	if ( !p1 || !p2 )
+		*pr = 0;
+	else {
+		for ( m = BDY(p1), l1 = 0; m; m = NEXT(m), l1++ );
+		for ( m = BDY(p2), l = 0; m; m = NEXT(m), l++ );
+		if ( l1 < l ) {
+			t = p1; p1 = p2; p2 = t;
+			l = l1;
+		}
+		if ( l > wlen ) {
+			if ( w ) GCFREE(w);
+			w = (MP *)MALLOC(l*sizeof(MP));
+			wlen = l;
+		}
+		for ( m = BDY(p2), i = 0; i < l; m = NEXT(m), i++ )
+			w[i] = m;
+		for ( s = 0, i = l-1; i >= 0; i-- ) {
+			_muldm_dup(vl,p1,w[i],&t); _addd_destructive(vl,s,t,&u); s = u;
+		}
+		bzero(w,l*sizeof(MP));
+		*pr = s;
+	}
+}
+
+void _weyl_muld_dup(VL vl,DP p1,DP p2,DP *pr)
+{
+	MP m;
+	DP s,t,u;
+	int i,l;
+	static MP *w;
+	static int wlen;
+
+	if ( !p1 || !p2 )
+		*pr = 0;
+	else {
+		for ( m = BDY(p1), l = 0; m; m = NEXT(m), l++ );
+		if ( l > wlen ) {
+			if ( w ) GCFREE(w);
+			w = (MP *)MALLOC(l*sizeof(MP));
+			wlen = l;
+		}
+		for ( m = BDY(p1), i = 0; i < l; m = NEXT(m), i++ )
+			w[i] = m;
+		for ( s = 0, i = l-1; i >= 0; i-- ) {
+			_weyl_muldm_dup(vl,w[i],p2,&t); _addd_destructive(vl,s,t,&u); s = u;
+		}
+		bzero(w,l*sizeof(MP));
+		*pr = s;
+	}
+}
+
+void _muldm_dup(VL vl,DP p,MP m0,DP *pr)
+{
+	MP m,mr,mr0;
+	DL d,dt,dm;
+	P c;
+	int n,i;
+	int *pt,*p1,*p2;
+
+	if ( !p )
+		*pr = 0;
+	else {
+		for ( mr0 = 0, m = BDY(p), c = (P)C(m0), d = m0->dl, n = NV(p); 
+			m; m = NEXT(m) ) {
+			_NEXTMP(mr0,mr);
+			mulp(vl,(P)C(m),c,(P *)&C(mr));
+			_NEWDL_NOINIT(dt,n); mr->dl = dt;
+			dm = m->dl;
+			dt->td = d->td + dm->td;
+			for ( i = 0, pt = dt->d, p1=d->d, p2 = dm->d; i < n; i++ )
+				*pt++ = *p1++ + *p2++;
+		}
+		NEXT(mr) = 0; _MKDP(NV(p),mr0,*pr);
+		if ( *pr )
+			(*pr)->sugar = p->sugar + m0->dl->td;
+	}
+}
+
+void _weyl_muldm_dup(VL vl,MP m0,DP p,DP *pr)
+{
+	DP r,t,t1;
+	MP m;
+	DL d0;
+	int n,n2,l,i,j,tlen;
+	static MP *w,*psum;
+	static struct cdl *tab;
+	static int wlen;
+	static int rtlen;
+
+	if ( !p )
+		*pr = 0;
+	else {
+		for ( m = BDY(p), l = 0; m; m = NEXT(m), l++ );
+		if ( l > wlen ) {
+			if ( w ) GCFREE(w);
+			w = (MP *)MALLOC(l*sizeof(MP));
+			wlen = l;
+		}
+		for ( m = BDY(p), i = 0; i < l; m = NEXT(m), i++ )
+			w[i] = m;
+		n = NV(p); n2 = n>>1;
+		d0 = m0->dl;
+
+		for ( i = 0, tlen = 1; i < n2; i++ )
+			tlen *= d0->d[n2+i]+1;
+		if ( tlen > rtlen ) {
+			if ( tab ) GCFREE(tab);
+			if ( psum ) GCFREE(psum);
+			rtlen = tlen;
+			tab = (struct cdl *)MALLOC(rtlen*sizeof(struct cdl));
+			psum = (MP *)MALLOC(rtlen*sizeof(MP));
+		}
+		bzero(psum,tlen*sizeof(MP));
+		for ( i = l-1; i >= 0; i-- ) {
+			bzero(tab,tlen*sizeof(struct cdl));
+			_weyl_mulmm_dup(vl,m0,w[i],n,tab,tlen);
+			for ( j = 0; j < tlen; j++ ) {
+				if ( tab[j].c ) {
+					_NEWMP(m); m->dl = tab[j].d;
+					C(m) = tab[j].c; NEXT(m) = psum[j];
+					psum[j] = m;
 				}
-				break;
 			}
 		}
-		if ( !h_reducible ) {
-			/* head term is not reducible */
-			if ( !full ) {
-				if ( g )
-					g->sugar = sugar;
-				_dptodp(g,rp); _free_dp(g); return;
-			} else {
-				m = BDY(g); 
-				if ( NEXT(m) ) {
-					BDY(g) = NEXT(m); NEXT(m) = 0;
-				} else {
-					FREEDP(g); g = 0;
+		for ( j = tlen-1, r = 0; j >= 0; j-- ) 
+			if ( psum[j] ) {
+				_MKDP(n,psum[j],t); _addd_destructive(vl,r,t,&t1); r = t1;
+			}
+		if ( r )
+			r->sugar = p->sugar + m0->dl->td;
+		*pr = r;
+	}
+}
+
+/* m0 = x0^d0*x1^d1*... * dx0^d(n/2)*dx1^d(n/2+1)*... */
+
+void _weyl_mulmm_dup(VL vl,MP m0,MP m1,int n,struct cdl *rtab,int rtablen)
+{
+	P c;
+	DL d,d0,d1,dt;
+	int i,j,a,b,k,l,n2,s,min,curlen;
+	struct cdl *p;
+	static Q *ctab;
+	static struct cdl *tab;
+	static int tablen;
+	static struct cdl *tmptab;
+	static int tmptablen;
+
+	if ( !m0 || !m1 ) {
+		rtab[0].c = 0;
+		rtab[0].d = 0;
+		return;
+	}
+	mulp(vl,(P)C(m0),(P)C(m1),&c);
+	d0 = m0->dl; d1 = m1->dl;
+	n2 = n>>1;
+	curlen = 1;
+
+	_NEWDL(d,n);
+	if ( n & 1 )
+		/* offset of h-degree */
+		d->td = d->d[n-1] = d0->d[n-1]+d1->d[n-1];
+	else
+		d->td = 0;
+	rtab[0].c = (Obj)c;
+	rtab[0].d = d;
+
+	if ( rtablen > tmptablen ) {
+		if ( tmptab ) GCFREE(tmptab);
+		tmptab = (struct cdl *)MALLOC(rtablen*sizeof(struct cdl));
+		tmptablen = rtablen; 
+	}
+
+	for ( i = 0; i < n2; i++ ) {
+		a = d0->d[i]; b = d1->d[n2+i];
+		k = d0->d[n2+i]; l = d1->d[i];
+
+		/* degree of xi^a*(Di^k*xi^l)*Di^b */
+		a += l;
+		b += k;
+		s = MUL_WEIGHT(a,i)+MUL_WEIGHT(b,n2+i);
+
+		if ( !k || !l ) {
+			for ( j = 0, p = rtab; j < curlen; j++, p++ ) {
+				if ( p->c ) {
+					dt = p->d;
+					dt->d[i] = a;
+					dt->d[n2+i] = b;
+					dt->td += s;
 				}
-				if ( d ) {
-					for ( mrd = BDY(d); NEXT(mrd); mrd = NEXT(mrd) );
-					NEXT(mrd) = m;
-				} else {
-					_MKDP(n,m,d);
-				}
 			}
+			curlen *= k+1;
+			continue;
 		}
+		if ( k+1 > tablen ) {
+			if ( tab ) GCFREE(tab);
+			if ( ctab ) GCFREE(ctab);
+			tablen = k+1;
+			tab = (struct cdl *)MALLOC(tablen*sizeof(struct cdl));
+			ctab = (Q *)MALLOC(tablen*sizeof(P));
+		}
+		/* compute xi^a*(Di^k*xi^l)*Di^b */
+		min = MIN(k,l);
+		mkwc(k,l,ctab);
+		bzero(tab,(k+1)*sizeof(struct cdl));
+		/* n&1 != 0 => homogenized computation; dx-xd=h^2 */
+		if ( n & 1 )
+			for ( j = 0; j <= min; j++ ) {
+				_NEWDL(d,n);
+				d->d[i] = a-j; d->d[n2+i] = b-j;
+				d->td = s;
+				d->d[n-1] = s-(MUL_WEIGHT(a-j,i)+MUL_WEIGHT(b-j,n2+i));
+				tab[j].d = d;
+				tab[j].c = (Obj)ctab[j];
+			}
+		else
+			for ( j = 0; j <= min; j++ ) {
+				_NEWDL(d,n);
+				d->d[i] = a-j; d->d[n2+i] = b-j;
+				d->td = MUL_WEIGHT(a-j,i)+MUL_WEIGHT(b-j,n2+i); /* XXX */
+				tab[j].d = d;
+				tab[j].c = (Obj)ctab[j];
+			}
+#if 0
+		_comm_muld_tab(vl,n,rtab,curlen,tab,k+1,tmptab);
+		for ( j = 0; j < curlen; j++ )
+			if ( rtab[j].d ) { _FREEDL(rtab[j].d); }
+		for ( j = 0; j <= min; j++ )
+			if ( tab[j].d ) { _FREEDL(tab[j].d); }
+		curlen *= k+1;
+		bcopy(tmptab,rtab,curlen*sizeof(struct cdl));
+#else
+		_comm_muld_tab_destructive(vl,n,rtab,curlen,tab,k+1);
+		for ( j = 0; j <= min; j++ )
+			if ( tab[j].d ) { _FREEDL(tab[j].d); }
+		curlen *= k+1;
+#endif
 	}
-	if ( d )
-		d->sugar = sugar;
-	_dptodp(d,rp); _free_dp(d);
+}
+
+/* direct product of two cdl tables
+  rt[] = [
+    t[0]*t1[0],...,t[n-1]*t1[0],
+    t[0]*t1[1],...,t[n-1]*t1[1],
+    ...
+    t[0]*t1[n1-1],...,t[n-1]*t1[n1-1]
+  ]
+*/
+
+void _comm_muld_tab(VL vl,int nv,struct cdl *t,int n,struct cdl *t1,int n1,struct cdl *rt)
+{
+	int i,j;
+	struct cdl *p;
+	P c;
+	DL d;
+
+	bzero(rt,n*n1*sizeof(struct cdl));
+	for ( j = 0, p = rt; j < n1; j++ ) {
+		c = (P)t1[j].c;
+		d = t1[j].d;
+		if ( !c )
+			break;
+		for ( i = 0; i < n; i++, p++ ) {
+			if ( t[i].c ) {
+				mulp(vl,(P)t[i].c,c,(P *)&p->c);
+				_adddl_dup(nv,t[i].d,d,&p->d);
+			}
+		}
+	}
+}
+
+void _comm_muld_tab_destructive(VL vl,int nv,struct cdl *t,int n,struct cdl *t1,int n1)
+{
+	int i,j;
+	struct cdl *p;
+	P c;
+	DL d;
+
+	for ( j = 1, p = t+n; j < n1; j++ ) {
+		c = (P)t1[j].c;
+		d = t1[j].d;
+		if ( !c )
+			break;
+		for ( i = 0; i < n; i++, p++ ) {
+			if ( t[i].c ) {
+				mulp(vl,(P)t[i].c,c,(P *)&p->c);
+				_adddl_dup(nv,t[i].d,d,&p->d);
+			}
+		}
+	}
+	c = (P)t1[0].c;
+	d = t1[0].d;
+	for ( i = 0, p = t; i < n; i++, p++ )
+		if ( t[i].c ) {
+			mulp(vl,(P)t[i].c,c,(P *)&p->c);
+			/* t[i].d += d */
+			adddl_destructive(nv,t[i].d,d);
+		}
 }