=================================================================== RCS file: /home/cvs/OpenXM_contrib2/asir2000/engine/nd.c,v retrieving revision 1.25 retrieving revision 1.65 diff -u -p -r1.25 -r1.65 --- OpenXM_contrib2/asir2000/engine/nd.c 2003/08/07 08:46:50 1.25 +++ OpenXM_contrib2/asir2000/engine/nd.c 2003/09/12 08:01:51 1.65 @@ -1,7 +1,8 @@ -/* $OpenXM: OpenXM_contrib2/asir2000/engine/nd.c,v 1.24 2003/08/05 08:06:21 noro Exp $ */ +/* $OpenXM: OpenXM_contrib2/asir2000/engine/nd.c,v 1.64 2003/09/12 01:12:40 noro Exp $ */ #include "ca.h" #include "inline.h" +#include #if defined(__GNUC__) #define INLINE inline @@ -11,200 +12,345 @@ #define INLINE #endif +typedef unsigned int UINT; + +#define USE_GEOBUCKET 1 +#define USE_UNROLL 1 + #define REDTAB_LEN 32003 +/* GeoBucket for polynomial addition */ + typedef struct oPGeoBucket { int m; struct oND *body[32]; } *PGeoBucket; +/* distributed polynomial; linked list rep. */ typedef struct oND { struct oNM *body; int nv; + int len; int sugar; } *ND; +/* distributed polynomial; array rep. */ typedef struct oNDV { struct oNMV *body; int nv; - int sugar; int len; + int sugar; } *NDV; +/* monomial; linked list rep. */ typedef struct oNM { struct oNM *next; union { int m; Q z; + P p; } c; - int td; - unsigned int dl[1]; + UINT dl[1]; } *NM; +/* monomial; array rep. */ typedef struct oNMV { union { int m; Q z; + P p; } c; - int td; - unsigned int dl[1]; + UINT dl[1]; } *NMV; +/* history of reducer */ typedef struct oRHist { struct oRHist *next; int index; - int td,sugar; - unsigned int dl[1]; + int sugar; + UINT dl[1]; } *RHist; +/* S-pair list */ typedef struct oND_pairs { struct oND_pairs *next; int i1,i2; - int td,sugar; - unsigned int lcm[1]; + int sugar; + UINT lcm[1]; } *ND_pairs; -double nd_scale=2; -static unsigned int **nd_bound; -int nd_nvar; -int is_rlex; -int nd_epw,nd_bpe,nd_wpd; -unsigned int nd_mask[32]; -unsigned int nd_mask0,nd_mask1; +/* index and shift count for each exponent */ +typedef struct oEPOS { + int i; /* index */ + int s; /* shift */ +} *EPOS; -NM _nm_free_list; -ND _nd_free_list; -ND_pairs _ndp_free_list; +typedef struct oBlockMask { + int n; + struct order_pair *order_pair; + UINT **mask; +} *BlockMask; +typedef struct oBaseSet { + int len; + NDV *ps; + UINT **bound; +} *BaseSet; + +typedef struct oNM_ind_pair +{ + NM mul; + int index; +} *NM_ind_pair; + + +int (*ndl_compare_function)(UINT *a1,UINT *a2); + +static double nd_scale=2; +static UINT **nd_bound; +static struct order_spec *nd_ord; +static EPOS nd_epos; +static BlockMask nd_blockmask; +static int nd_nvar; +static int nd_isrlex; +static int nd_epw,nd_bpe,nd_wpd,nd_exporigin; +static UINT nd_mask[32]; +static UINT nd_mask0,nd_mask1; + +static NM _nm_free_list; +static ND _nd_free_list; +static ND_pairs _ndp_free_list; + static NDV *nd_ps; -static NDV *nd_psq; -int *nd_psl; -RHist *nd_psh; -int nd_psn,nd_pslen; +static NDV *nd_ps_trace; +static RHist *nd_psh; +static int nd_psn,nd_pslen; -RHist *nd_red; -int nd_red_len; +static RHist *nd_red; -int nd_found,nd_create,nd_notfirst; -int nm_adv; -int nmv_adv; +static int nd_found,nd_create,nd_notfirst; +static int nm_adv; +static int nmv_adv; +static int nd_dcomp; -extern int Top,Reverse; +extern VL CO; +extern int Top,Reverse,dp_nelim,do_weyl; +extern int *current_weyl_weight_vector; -#define HTD(d) ((d)->body->td) +/* fundamental macros */ +#define TD(d) (d[0]) #define HDL(d) ((d)->body->dl) +#define HTD(d) (TD(HDL(d))) #define HCM(d) ((d)->body->c.m) #define HCQ(d) ((d)->body->c.z) +#define HCP(d) ((d)->body->c.p) #define CM(a) ((a)->c.m) #define CQ(a) ((a)->c.z) +#define CP(a) ((a)->c.p) #define DL(a) ((a)->dl) -#define TD(a) ((a)->td) #define SG(a) ((a)->sugar) #define LEN(a) ((a)->len) +#define LCM(a) ((a)->lcm) +#define GET_EXP(d,a) (((d)[nd_epos[a].i]>>nd_epos[a].s)&nd_mask0) +#define GET_EXP_MASK(d,a,m) ((((d)[nd_epos[a].i]&(m)[nd_epos[a].i])>>nd_epos[a].s)&nd_mask0) +#define PUT_EXP(r,a,e) ((r)[nd_epos[a].i] |= ((e)<>oepos[a].s)&omask0) +#define PUT_EXP_OLD(r,a,e) ((r)[oepos[a].i] |= ((e)<TD(d2)?1:(TD(d1)0?TD_DL_COMPARE(d1,d2)\ + :(nd_dcomp==0?ndl_lex_compare(d1,d2)\ + :(nd_blockmask?ndl_block_compare(d1,d2)\ + :(*ndl_compare_function)(d1,d2)))) +#else +#define DL_COMPARE(d1,d2)\ +(nd_dcomp>0?TD_DL_COMPARE(d1,d2):(*ndl_compare_function)(d1,d2)) +#endif + +/* allocators */ #define NEWRHist(r) \ -((r)=(RHist)MALLOC(sizeof(struct oRHist)+(nd_wpd-1)*sizeof(unsigned int))) -#define NEWND_pairs(m) if(!_ndp_free_list)_NDP_alloc(); (m)=_ndp_free_list; _ndp_free_list = NEXT(_ndp_free_list) -#define NEWNM(m) if(!_nm_free_list)_NM_alloc(); (m)=_nm_free_list; _nm_free_list = NEXT(_nm_free_list) -#define MKND(n,m,d) if(!_nd_free_list)_ND_alloc(); (d)=_nd_free_list; _nd_free_list = (ND)BDY(_nd_free_list); (d)->nv=(n); BDY(d)=(m) +((r)=(RHist)MALLOC(sizeof(struct oRHist)+(nd_wpd-1)*sizeof(UINT))) +#define NEWND_pairs(m) \ +if(!_ndp_free_list)_NDP_alloc();\ +(m)=_ndp_free_list; _ndp_free_list = NEXT(_ndp_free_list) +#define NEWNM(m)\ +if(!_nm_free_list)_NM_alloc();\ +(m)=_nm_free_list; _nm_free_list = NEXT(_nm_free_list) +#define MKND(n,m,len,d)\ +if(!_nd_free_list)_ND_alloc();\ +(d)=_nd_free_list; _nd_free_list = (ND)BDY(_nd_free_list);\ +NV(d)=(n); LEN(d)=(len); BDY(d)=(m) +#define NEWNDV(d) ((d)=(NDV)MALLOC(sizeof(struct oNDV))) +#define MKNDV(n,m,l,d) NEWNDV(d); NV(d)=(n); BDY(d)=(m); LEN(d) = l; +#define NEWNM_ind_pair(p)\ +((p)=(NM_ind_pair)MALLOC(sizeof(struct oNM_ind_pair))) +/* allocate and link a new object */ #define NEXTRHist(r,c) \ if(!(r)){NEWRHist(r);(c)=(r);}else{NEWRHist(NEXT(c));(c)=NEXT(c);} #define NEXTNM(r,c) \ if(!(r)){NEWNM(r);(c)=(r);}else{NEWNM(NEXT(c));(c)=NEXT(c);} #define NEXTNM2(r,c,s) \ if(!(r)){(c)=(r)=(s);}else{NEXT(c)=(s);(c)=(s);} +#define NEXTND_pairs(r,c) \ +if(!(r)){NEWND_pairs(r);(c)=(r);}else{NEWND_pairs(NEXT(c));(c)=NEXT(c);} +#define MKNM_ind_pair(p,m,i) (NEWNM_ind_pair(p),(p)->mul=(m),(p)->index=(i)) + +/* deallocators */ #define FREENM(m) NEXT(m)=_nm_free_list; _nm_free_list=(m) #define FREENDP(m) NEXT(m)=_ndp_free_list; _ndp_free_list=(m) #define FREEND(m) BDY(m)=(NM)_nd_free_list; _nd_free_list=(m) -#define NEXTND_pairs(r,c) \ -if(!(r)){NEWND_pairs(r);(c)=(r);}else{NEWND_pairs(NEXT(c));(c)=NEXT(c);} +/* macro for increasing pointer to NMV */ +#define NMV_ADV(m) (m = (NMV)(((char *)m)+nmv_adv)) +#define NMV_OADV(m) (m = (NMV)(((char *)m)+oadv)) +#define NDV_NADV(m) (m = (NMV)(((char *)m)+newadv)) +#define NMV_PREV(m) (m = (NMV)(((char *)m)-nmv_adv)) +#define NMV_OPREV(m) (m = (NMV)(((char *)m)-oadv)) +/* external functions */ +void GC_gcollect(); +NODE append_one(NODE,int); + +/* manipulation of coefficients */ void nd_removecont(int mod,ND p); void nd_removecont2(ND p1,ND p2); -void ndv_removecont(int mod,NDV p); -void ndv_dehomogenize(NDV p); -void ndv_mul_c_q(NDV p,Q mul); -void nd_mul_c_q(ND p,Q mul); +void removecont_array(Q *c,int n); + +/* GeoBucket functions */ ND normalize_pbucket(int mod,PGeoBucket g); int head_pbucket(int mod,PGeoBucket g); -void add_pbucket(int mod,PGeoBucket g,ND d,int l); +int head_pbucket_q(PGeoBucket g); +void add_pbucket_symbolic(PGeoBucket g,ND d); +void add_pbucket(int mod,PGeoBucket g,ND d); void free_pbucket(PGeoBucket b); +void mulq_pbucket(PGeoBucket g,Q c); +NM remove_head_pbucket_symbolic(PGeoBucket g); PGeoBucket create_pbucket(); -ND nd_remove_head(ND p); -void GC_gcollect(); -NODE append_one(NODE,int); - -void removecont_array(Q *c,int n); -ND_pairs crit_B( ND_pairs d, int s ); -void nd_gr(LIST f,LIST v,int m,struct order_spec *ord,LIST *rp); -void nd_gr_trace(LIST f,LIST v,int m,int homo,struct order_spec *ord,LIST *rp); -void nd_setup(int mod,NODE f); -void nd_setup_trace(int mod,NODE f); -int nd_newps(int mod,ND a); -int nd_newps_trace(int mod,ND nf,ND nfq); +/* manipulation of pairs and bases */ +int nd_newps(int mod,ND a,ND aq); +ND_pairs nd_newpairs( NODE g, int t ); ND_pairs nd_minp( ND_pairs d, ND_pairs *prest ); +ND_pairs nd_minsugarp( ND_pairs d, ND_pairs *prest ); NODE update_base(NODE nd,int ndp); -static ND_pairs equivalent_pairs( ND_pairs d1, ND_pairs *prest ); -int crit_2( int dp1, int dp2 ); -ND_pairs crit_F( ND_pairs d1 ); -ND_pairs crit_M( ND_pairs d1 ); -ND_pairs nd_newpairs( NODE g, int t ); ND_pairs update_pairs( ND_pairs d, NODE /* of index */ g, int t); -NODE nd_gb(int m); +ND_pairs equivalent_pairs( ND_pairs d1, ND_pairs *prest ); +ND_pairs crit_B( ND_pairs d, int s ); +ND_pairs crit_M( ND_pairs d1 ); +ND_pairs crit_F( ND_pairs d1 ); +int crit_2( int dp1, int dp2 ); +int ndv_newps(NDV a,NDV aq); + +/* top level functions */ +void nd_gr(LIST f,LIST v,int m,int f4,struct order_spec *ord,LIST *rp); +void nd_gr_trace(LIST f,LIST v,int trace,int homo,struct order_spec *ord,LIST *rp); +NODE nd_f4(int m); +NODE nd_gb(int m,int checkonly); NODE nd_gb_trace(int m); + +/* ndl functions */ +int ndl_weight(UINT *d); +void ndl_weight_mask(UINT *d); +void ndl_homogenize(UINT *d,UINT *r,int obpe,EPOS oepos,int weight); +void ndl_dehomogenize(UINT *p); +void ndl_reconstruct(UINT *d,UINT *r,int obpe,EPOS oepos); +INLINE int ndl_reducible(UINT *d1,UINT *d2); +INLINE int ndl_lex_compare(UINT *d1,UINT *d2); +INLINE int ndl_block_compare(UINT *d1,UINT *d2); +INLINE int ndl_equal(UINT *d1,UINT *d2); +INLINE void ndl_copy(UINT *d1,UINT *d2); +INLINE void ndl_zero(UINT *d); +INLINE void ndl_add(UINT *d1,UINT *d2,UINT *d); +INLINE void ndl_addto(UINT *d1,UINT *d2); +INLINE void ndl_sub(UINT *d1,UINT *d2,UINT *d); +INLINE int ndl_hash_value(UINT *d); + +/* normal forms */ +INLINE int ndl_find_reducer(UINT *g); +INLINE int ndl_find_reducer_direct(UINT *g,NDV *ps,int len); +int nd_sp(int mod,int trace,ND_pairs p,ND *nf); +int nd_nf(int mod,ND g,NDV *ps,int full,ND *nf); +int nd_nf_pbucket(int mod,ND g,NDV *ps,int full,ND *nf); +int nd_nf_direct(int mod,ND g,BaseSet base,int full,ND *rp); + +/* finalizers */ +NODE ndv_reducebase(NODE x); +NODE ndv_reduceall(int m,NODE f); + +/* allocators */ void nd_free_private_storage(); void _NM_alloc(); void _ND_alloc(); -int ndl_td(unsigned int *d); -int ndl_dehomogenize(unsigned int *p); -ND nd_add(int mod,ND p1,ND p2); -ND nd_add_q(ND p1,ND p2); -ND nd_mul_nm(int mod,ND p,NM m0); -ND nd_mul_ind_nm(int mod,int index,NM m0); -int nd_sp(int mod,ND_pairs p,ND *nf); -int nd_find_reducer(ND g); -int nd_find_reducer_direct(ND g,NDV *ps,int len); -int nd_nf(int mod,ND g,int full,ND *nf); -ND nd_reduce(ND p1,ND p2); -ND nd_reduce_special(ND p1,ND p2); -NODE nd_reduceall(int m,NODE f); void nd_free(ND p); -void ndv_free(NDV p); -void ndl_print(unsigned int *dl); +void nd_free_redlist(); + +/* printing */ +void ndl_print(UINT *dl); void nd_print(ND p); void nd_print_q(ND p); -void ndv_print(NDV p); -void ndv_print_q(NDV p); void ndp_print(ND_pairs d); -int nd_length(ND p); -void nd_mul_c(int mod,ND p,int mul); -void nd_free_redlist(); -void nd_append_red(unsigned int *d,int td,int i); -unsigned int *nd_compute_bound(ND p); -unsigned int *dp_compute_bound(DP p); + + +/* setup, reconstruct */ +void nd_init_ord(struct order_spec *spec); ND_pairs nd_reconstruct(int mod,int trace,ND_pairs ndp); void nd_reconstruct_direct(int mod,NDV *ps,int len); -void nd_setup_parameters(); -void nd_realloc(ND p,int obpe); +void ndv_setup(int mod,int trace,NODE f); +void nd_setup_parameters(int nvar,int max); +BlockMask nd_create_blockmask(struct order_spec *ord); +EPOS nd_create_epos(struct order_spec *ord); +int nd_get_exporigin(struct order_spec *ord); +void ndv_mod(int mod,NDV p); +NDV ndv_dup(int mod,NDV p); +ND nd_dup(ND p); + +/* ND functions */ +int ndv_check_candidate(NODE input,int obpe,int oadv,EPOS oepos,NODE cand); +void nd_mul_c(int mod,ND p,int mul); +void nd_mul_c_q(ND p,Q mul); +void nd_mul_c_p(VL vl,ND p,P mul); +ND nd_remove_head(ND p); +int nd_length(ND p); +void nd_append_red(UINT *d,int i); +UINT *ndv_compute_bound(NDV p); ND nd_copy(ND p); -void ndl_dup(int obpe,unsigned int *d,unsigned int *r); +ND nd_merge(ND p1,ND p2); +ND nd_add(int mod,ND p1,ND p2); +ND nd_add_q(ND p1,ND p2); +INLINE int nd_length(ND p); -#define NMV_ADV(m) (m = (NMV)(((char *)m)+nmv_adv)) -#define NEWNDV(d) ((d)=(NDV)MALLOC(sizeof(struct oNDV))) -#define MKNDV(n,m,l,d) NEWNDV(d); NV(d)=(n); BDY(d)=(m); LEN(d) = l; +/* NDV functions */ +ND weyl_ndv_mul_nm(int mod,NM m0,NDV p); +void weyl_mul_nm_nmv(int n,int mod,NM m0,NMV m1,NM *tab,int tlen); void ndv_mul_c(int mod,NDV p,int mul); +void ndv_mul_c_q(NDV p,Q mul); +ND ndv_mul_nm_symbolic(NM m0,NDV p); +ND ndv_mul_nm(int mod,NM m0,NDV p); +void ndv_realloc(NDV p,int obpe,int oadv,EPOS oepos); +NDV ndv_dup_realloc(NDV p,int obpe,int oadv,EPOS oepos); +void ndv_homogenize(NDV p,int obpe,int oadv,EPOS eops); +void ndv_dehomogenize(NDV p,struct order_spec *spec); +void ndv_removecont(int mod,NDV p); +void ndv_print(NDV p); +void ndv_print_q(NDV p); +void ndv_free(NDV p); + +/* converters */ +ND ptond(VL vl,VL dvl,P p); +NDV ptondv(VL vl,VL dvl,P p); +P ndvtop(int mod,VL vl,VL dvl,NDV p); NDV ndtondv(int mod,ND p); ND ndvtond(int mod,NDV p); -ND ndv_mul_nm(int mod,NDV p,NM m0); -void ndv_realloc(NDV p,int obpe,int oadv); -NDV dptondv(int,DP); -DP ndvtodp(int,NDV); -ND dptond(int,DP); -DP ndtodp(int,ND); +int nm_ind_pair_to_vect(int m,UINT *s0,int n,NM_ind_pair pair,UINT *r); +void nm_ind_pair_to_vect_compress(int m,UINT *s0,int n,NM_ind_pair pair,int *ind); +int nd_to_vect(int mod,UINT *s0,int n,ND d,UINT *r); void nd_free_private_storage() { @@ -221,7 +367,7 @@ void _NM_alloc() int i; for ( i = 0; i < 1024; i++ ) { - p = (NM)GC_malloc(sizeof(struct oNM)+(nd_wpd-1)*sizeof(unsigned int)); + p = (NM)GC_malloc(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT)); p->next = _nm_free_list; _nm_free_list = p; } } @@ -244,12 +390,12 @@ void _NDP_alloc() for ( i = 0; i < 1024; i++ ) { p = (ND_pairs)GC_malloc(sizeof(struct oND_pairs) - +(nd_wpd-1)*sizeof(unsigned int)); + +(nd_wpd-1)*sizeof(UINT)); p->next = _ndp_free_list; _ndp_free_list = p; } } -INLINE nd_length(ND p) +INLINE int nd_length(ND p) { NM m; int i; @@ -262,156 +408,230 @@ INLINE nd_length(ND p) } } -INLINE int ndl_reducible(unsigned int *d1,unsigned int *d2) +INLINE int ndl_reducible(UINT *d1,UINT *d2) { - unsigned int u1,u2; + UINT u1,u2; int i,j; + if ( TD(d1) < TD(d2) ) return 0; +#if USE_UNROLL switch ( nd_bpe ) { + case 3: + for ( i = nd_exporigin; i < nd_wpd; i++ ) { + u1 = d1[i]; u2 = d2[i]; + if ( (u1&0x38000000) < (u2&0x38000000) ) return 0; + if ( (u1& 0x7000000) < (u2& 0x7000000) ) return 0; + if ( (u1& 0xe00000) < (u2& 0xe00000) ) return 0; + if ( (u1& 0x1c0000) < (u2& 0x1c0000) ) return 0; + if ( (u1& 0x38000) < (u2& 0x38000) ) return 0; + if ( (u1& 0x7000) < (u2& 0x7000) ) return 0; + if ( (u1& 0xe00) < (u2& 0xe00) ) return 0; + if ( (u1& 0x1c0) < (u2& 0x1c0) ) return 0; + if ( (u1& 0x38) < (u2& 0x38) ) return 0; + if ( (u1& 0x7) < (u2& 0x7) ) return 0; + } + return 1; + break; case 4: - for ( i = 0; i < nd_wpd; i++ ) { + for ( i = nd_exporigin; i < nd_wpd; i++ ) { u1 = d1[i]; u2 = d2[i]; if ( (u1&0xf0000000) < (u2&0xf0000000) ) return 0; - if ( (u1&0xf000000) < (u2&0xf000000) ) return 0; - if ( (u1&0xf00000) < (u2&0xf00000) ) return 0; - if ( (u1&0xf0000) < (u2&0xf0000) ) return 0; - if ( (u1&0xf000) < (u2&0xf000) ) return 0; - if ( (u1&0xf00) < (u2&0xf00) ) return 0; - if ( (u1&0xf0) < (u2&0xf0) ) return 0; - if ( (u1&0xf) < (u2&0xf) ) return 0; + if ( (u1& 0xf000000) < (u2& 0xf000000) ) return 0; + if ( (u1& 0xf00000) < (u2& 0xf00000) ) return 0; + if ( (u1& 0xf0000) < (u2& 0xf0000) ) return 0; + if ( (u1& 0xf000) < (u2& 0xf000) ) return 0; + if ( (u1& 0xf00) < (u2& 0xf00) ) return 0; + if ( (u1& 0xf0) < (u2& 0xf0) ) return 0; + if ( (u1& 0xf) < (u2& 0xf) ) return 0; } return 1; break; case 6: - for ( i = 0; i < nd_wpd; i++ ) { + for ( i = nd_exporigin; i < nd_wpd; i++ ) { u1 = d1[i]; u2 = d2[i]; if ( (u1&0x3f000000) < (u2&0x3f000000) ) return 0; - if ( (u1&0xfc0000) < (u2&0xfc0000) ) return 0; - if ( (u1&0x3f000) < (u2&0x3f000) ) return 0; - if ( (u1&0xfc0) < (u2&0xfc0) ) return 0; - if ( (u1&0x3f) < (u2&0x3f) ) return 0; + if ( (u1& 0xfc0000) < (u2& 0xfc0000) ) return 0; + if ( (u1& 0x3f000) < (u2& 0x3f000) ) return 0; + if ( (u1& 0xfc0) < (u2& 0xfc0) ) return 0; + if ( (u1& 0x3f) < (u2& 0x3f) ) return 0; } return 1; break; case 8: - for ( i = 0; i < nd_wpd; i++ ) { + for ( i = nd_exporigin; i < nd_wpd; i++ ) { u1 = d1[i]; u2 = d2[i]; if ( (u1&0xff000000) < (u2&0xff000000) ) return 0; - if ( (u1&0xff0000) < (u2&0xff0000) ) return 0; - if ( (u1&0xff00) < (u2&0xff00) ) return 0; - if ( (u1&0xff) < (u2&0xff) ) return 0; + if ( (u1& 0xff0000) < (u2& 0xff0000) ) return 0; + if ( (u1& 0xff00) < (u2& 0xff00) ) return 0; + if ( (u1& 0xff) < (u2& 0xff) ) return 0; } return 1; break; case 16: - for ( i = 0; i < nd_wpd; i++ ) { + for ( i = nd_exporigin; i < nd_wpd; i++ ) { u1 = d1[i]; u2 = d2[i]; if ( (u1&0xffff0000) < (u2&0xffff0000) ) return 0; - if ( (u1&0xffff) < (u2&0xffff) ) return 0; + if ( (u1& 0xffff) < (u2& 0xffff) ) return 0; } return 1; break; case 32: - for ( i = 0; i < nd_wpd; i++ ) + for ( i = nd_exporigin; i < nd_wpd; i++ ) if ( d1[i] < d2[i] ) return 0; return 1; break; default: - for ( i = 0; i < nd_wpd; i++ ) { + for ( i = nd_exporigin; i < nd_wpd; i++ ) { u1 = d1[i]; u2 = d2[i]; for ( j = 0; j < nd_epw; j++ ) if ( (u1&nd_mask[j]) < (u2&nd_mask[j]) ) return 0; } return 1; } +#else + for ( i = nd_exporigin; i < nd_wpd; i++ ) { + u1 = d1[i]; u2 = d2[i]; + for ( j = 0; j < nd_epw; j++ ) + if ( (u1&nd_mask[j]) < (u2&nd_mask[j]) ) return 0; + } + return 1; +#endif } -/* returns the exponent of homo variable */ +/* + * If the current order is a block order, + * then the last block is length 1 and contains + * the homo variable. Otherwise, the original + * order is either 0 or 2. + */ -int ndl_dehomogenize(unsigned int *d) +void ndl_homogenize(UINT *d,UINT *r,int obpe,EPOS oepos,int weight) { - unsigned int mask; - unsigned int h; - int i; + int w,i,e,n,omask0; - if ( is_rlex ) { - if ( nd_bpe == 32 ) { - h = d[0]; - for ( i = 1; i < nd_wpd; i++ ) - d[i-1] = d[i]; - d[i-1] = 0; - return h; + omask0 = (1<>((nd_epw-1)*nd_bpe))&nd_mask0; + for ( i = nd_exporigin; i < nd_wpd; i++ ) + d[i] = ((d[i]<>((nd_epw-1)*nd_bpe))&nd_mask0):0); + TD(d) -= h; + } } else { - mask = (1<<(nd_epw*nd_bpe))-1; - h = (d[0]>>((nd_epw-1)*nd_bpe))&nd_mask0; - for ( i = 0; i < nd_wpd; i++ ) - d[i] = ((d[i]<>((nd_epw-1)*nd_bpe))&nd_mask0):0); - return h; + h = GET_EXP(d,nd_nvar-1); + XOR_EXP(d,nd_nvar-1,h); + TD(d) -= h; } - } else { - /* do nothing */ - return d[nd_wpd-1]; } } -void ndl_lcm(unsigned int *d1,unsigned *d2,unsigned int *d) +void ndl_lcm(UINT *d1,unsigned *d2,UINT *d) { - unsigned int t1,t2,u,u1,u2; - int i,j; + UINT t1,t2,u,u1,u2; + int i,j,l; +#if USE_UNROLL switch ( nd_bpe ) { + case 3: + for ( i = nd_exporigin; i < nd_wpd; i++ ) { + u1 = d1[i]; u2 = d2[i]; + t1 = (u1&0x38000000); t2 = (u2&0x38000000); u = t1>t2?t1:t2; + t1 = (u1& 0x7000000); t2 = (u2& 0x7000000); u |= t1>t2?t1:t2; + t1 = (u1& 0xe00000); t2 = (u2& 0xe00000); u |= t1>t2?t1:t2; + t1 = (u1& 0x1c0000); t2 = (u2& 0x1c0000); u |= t1>t2?t1:t2; + t1 = (u1& 0x38000); t2 = (u2& 0x38000); u |= t1>t2?t1:t2; + t1 = (u1& 0x7000); t2 = (u2& 0x7000); u |= t1>t2?t1:t2; + t1 = (u1& 0xe00); t2 = (u2& 0xe00); u |= t1>t2?t1:t2; + t1 = (u1& 0x1c0); t2 = (u2& 0x1c0); u |= t1>t2?t1:t2; + t1 = (u1& 0x38); t2 = (u2& 0x38); u |= t1>t2?t1:t2; + t1 = (u1& 0x7); t2 = (u2& 0x7); u |= t1>t2?t1:t2; + d[i] = u; + } + break; case 4: - for ( i = 0; i < nd_wpd; i++ ) { + for ( i = nd_exporigin; i < nd_wpd; i++ ) { u1 = d1[i]; u2 = d2[i]; t1 = (u1&0xf0000000); t2 = (u2&0xf0000000); u = t1>t2?t1:t2; - t1 = (u1&0xf000000); t2 = (u2&0xf000000); u |= t1>t2?t1:t2; - t1 = (u1&0xf00000); t2 = (u2&0xf00000); u |= t1>t2?t1:t2; - t1 = (u1&0xf0000); t2 = (u2&0xf0000); u |= t1>t2?t1:t2; - t1 = (u1&0xf000); t2 = (u2&0xf000); u |= t1>t2?t1:t2; - t1 = (u1&0xf00); t2 = (u2&0xf00); u |= t1>t2?t1:t2; - t1 = (u1&0xf0); t2 = (u2&0xf0); u |= t1>t2?t1:t2; - t1 = (u1&0xf); t2 = (u2&0xf); u |= t1>t2?t1:t2; + t1 = (u1& 0xf000000); t2 = (u2& 0xf000000); u |= t1>t2?t1:t2; + t1 = (u1& 0xf00000); t2 = (u2& 0xf00000); u |= t1>t2?t1:t2; + t1 = (u1& 0xf0000); t2 = (u2& 0xf0000); u |= t1>t2?t1:t2; + t1 = (u1& 0xf000); t2 = (u2& 0xf000); u |= t1>t2?t1:t2; + t1 = (u1& 0xf00); t2 = (u2& 0xf00); u |= t1>t2?t1:t2; + t1 = (u1& 0xf0); t2 = (u2& 0xf0); u |= t1>t2?t1:t2; + t1 = (u1& 0xf); t2 = (u2& 0xf); u |= t1>t2?t1:t2; d[i] = u; } break; case 6: - for ( i = 0; i < nd_wpd; i++ ) { + for ( i = nd_exporigin; i < nd_wpd; i++ ) { u1 = d1[i]; u2 = d2[i]; t1 = (u1&0x3f000000); t2 = (u2&0x3f000000); u = t1>t2?t1:t2; - t1 = (u1&0xfc0000); t2 = (u2&0xfc0000); u |= t1>t2?t1:t2; - t1 = (u1&0x3f000); t2 = (u2&0x3f000); u |= t1>t2?t1:t2; - t1 = (u1&0xfc0); t2 = (u2&0xfc0); u |= t1>t2?t1:t2; - t1 = (u1&0x3f); t2 = (u2&0x3f); u |= t1>t2?t1:t2; + t1 = (u1& 0xfc0000); t2 = (u2& 0xfc0000); u |= t1>t2?t1:t2; + t1 = (u1& 0x3f000); t2 = (u2& 0x3f000); u |= t1>t2?t1:t2; + t1 = (u1& 0xfc0); t2 = (u2& 0xfc0); u |= t1>t2?t1:t2; + t1 = (u1& 0x3f); t2 = (u2& 0x3f); u |= t1>t2?t1:t2; d[i] = u; } break; case 8: - for ( i = 0; i < nd_wpd; i++ ) { + for ( i = nd_exporigin; i < nd_wpd; i++ ) { u1 = d1[i]; u2 = d2[i]; t1 = (u1&0xff000000); t2 = (u2&0xff000000); u = t1>t2?t1:t2; - t1 = (u1&0xff0000); t2 = (u2&0xff0000); u |= t1>t2?t1:t2; - t1 = (u1&0xff00); t2 = (u2&0xff00); u |= t1>t2?t1:t2; - t1 = (u1&0xff); t2 = (u2&0xff); u |= t1>t2?t1:t2; + t1 = (u1& 0xff0000); t2 = (u2& 0xff0000); u |= t1>t2?t1:t2; + t1 = (u1& 0xff00); t2 = (u2& 0xff00); u |= t1>t2?t1:t2; + t1 = (u1& 0xff); t2 = (u2& 0xff); u |= t1>t2?t1:t2; d[i] = u; } break; case 16: - for ( i = 0; i < nd_wpd; i++ ) { + for ( i = nd_exporigin; i < nd_wpd; i++ ) { u1 = d1[i]; u2 = d2[i]; t1 = (u1&0xffff0000); t2 = (u2&0xffff0000); u = t1>t2?t1:t2; - t1 = (u1&0xffff); t2 = (u2&0xffff); u |= t1>t2?t1:t2; + t1 = (u1& 0xffff); t2 = (u2& 0xffff); u |= t1>t2?t1:t2; d[i] = u; } break; case 32: - for ( i = 0; i < nd_wpd; i++ ) { + for ( i = nd_exporigin; i < nd_wpd; i++ ) { u1 = d1[i]; u2 = d2[i]; d[i] = u1>u2?u1:u2; } break; default: - for ( i = 0; i < nd_wpd; i++ ) { + for ( i = nd_exporigin; i < nd_wpd; i++ ) { u1 = d1[i]; u2 = d2[i]; for ( j = 0, u = 0; j < nd_epw; j++ ) { t1 = (u1&nd_mask[j]); t2 = (u2&nd_mask[j]); u |= t1>t2?t1:t2; @@ -420,55 +640,146 @@ void ndl_lcm(unsigned int *d1,unsigned *d2,unsigned in } break; } +#else + for ( i = nd_exporigin; i < nd_wpd; i++ ) { + u1 = d1[i]; u2 = d2[i]; + for ( j = 0, u = 0; j < nd_epw; j++ ) { + t1 = (u1&nd_mask[j]); t2 = (u2&nd_mask[j]); u |= t1>t2?t1:t2; + } + d[i] = u; + } +#endif + TD(d) = ndl_weight(d); + if ( nd_blockmask ) ndl_weight_mask(d); } -int ndl_td(unsigned int *d) +int ndl_weight(UINT *d) { - unsigned int t,u; + UINT t,u; int i,j; - for ( t = 0, i = 0; i < nd_wpd; i++ ) { - u = d[i]; - for ( j = 0; j < nd_epw; j++, u>>=nd_bpe ) - t += (u&nd_mask0); - } + if ( current_dl_weight_vector ) + for ( i = 0, t = 0; i < nd_nvar; i++ ) { + u = GET_EXP(d,i); + t += MUL_WEIGHT(u,i); + } + else + for ( t = 0, i = nd_exporigin; i < nd_wpd; i++ ) { + u = d[i]; + for ( j = 0; j < nd_epw; j++, u>>=nd_bpe ) + t += (u&nd_mask0); + } return t; } -INLINE int ndl_compare(unsigned int *d1,unsigned int *d2) +void ndl_weight_mask(UINT *d) { + UINT t,u; + UINT *mask; + int i,j,k,l; + + l = nd_blockmask->n; + for ( k = 0; k < l; k++ ) { + mask = nd_blockmask->mask[k]; + if ( current_dl_weight_vector ) + for ( i = 0, t = 0; i < nd_nvar; i++ ) { + u = GET_EXP_MASK(d,i,mask); + t += MUL_WEIGHT(u,i); + } + else + for ( t = 0, i = nd_exporigin; i < nd_wpd; i++ ) { + u = d[i]&mask[i]; + for ( j = 0; j < nd_epw; j++, u>>=nd_bpe ) + t += (u&nd_mask0); + } + d[k+1] = t; + } +} + +int ndl_lex_compare(UINT *d1,UINT *d2) +{ int i; - for ( i = 0; i < nd_wpd; i++, d1++, d2++ ) + d1 += nd_exporigin; + d2 += nd_exporigin; + for ( i = nd_exporigin; i < nd_wpd; i++, d1++, d2++ ) if ( *d1 > *d2 ) - return is_rlex ? -1 : 1; + return nd_isrlex ? -1 : 1; else if ( *d1 < *d2 ) - return is_rlex ? 1 : -1; + return nd_isrlex ? 1 : -1; return 0; } -INLINE int ndl_equal(unsigned int *d1,unsigned int *d2) +int ndl_block_compare(UINT *d1,UINT *d2) { + int i,l,j,ord_o,ord_l; + struct order_pair *op; + UINT t1,t2,m; + UINT *mask; + + l = nd_blockmask->n; + op = nd_blockmask->order_pair; + for ( j = 0; j < l; j++ ) { + mask = nd_blockmask->mask[j]; + ord_o = op[j].order; + if ( ord_o < 2 ) + if ( (t1=d1[j+1]) > (t2=d2[j+1]) ) return 1; + else if ( t1 < t2 ) return -1; + for ( i = nd_exporigin; i < nd_wpd; i++ ) { + m = mask[i]; + t1 = d1[i]&m; + t2 = d2[i]&m; + if ( t1 > t2 ) + return !ord_o ? -1 : 1; + else if ( t1 < t2 ) + return !ord_o ? 1 : -1; + } + } + return 0; +} + +/* TDH -> WW -> TD-> RL */ + +int ndl_ww_lex_compare(UINT *d1,UINT *d2) +{ + int i,m,e1,e2; + + if ( TD(d1) > TD(d2) ) return 1; + else if ( TD(d1) < TD(d2) ) return -1; + m = nd_nvar>>1; + for ( i = 0, e1 = e2 = 0; i < m; i++ ) { + e1 += current_weyl_weight_vector[i]*(GET_EXP(d1,m+i)-GET_EXP(d1,i)); + e2 += current_weyl_weight_vector[i]*(GET_EXP(d2,m+i)-GET_EXP(d2,i)); + } + if ( e1 > e2 ) return 1; + else if ( e1 < e2 ) return -1; + return ndl_lex_compare(d1,d2); +} + +INLINE int ndl_equal(UINT *d1,UINT *d2) +{ int i; for ( i = 0; i < nd_wpd; i++ ) - if ( d1[i] != d2[i] ) + if ( *d1++ != *d2++ ) return 0; return 1; } -INLINE void ndl_copy(unsigned int *d1,unsigned int *d2) +INLINE void ndl_copy(UINT *d1,UINT *d2) { int i; switch ( nd_wpd ) { - case 1: - d2[0] = d1[0]; - break; case 2: - d2[0] = d1[0]; + TD(d2) = TD(d1); d2[1] = d1[1]; break; + case 3: + TD(d2) = TD(d1); + d2[1] = d1[1]; + d2[2] = d1[2]; + break; default: for ( i = 0; i < nd_wpd; i++ ) d2[i] = d1[i]; @@ -476,108 +787,141 @@ INLINE void ndl_copy(unsigned int *d1,unsigned int *d2 } } -INLINE void ndl_add(unsigned int *d1,unsigned int *d2,unsigned int *d) +INLINE void ndl_zero(UINT *d) { int i; + for ( i = 0; i < nd_wpd; i++ ) d[i] = 0; +} +INLINE void ndl_add(UINT *d1,UINT *d2,UINT *d) +{ + int i; + +#if 1 switch ( nd_wpd ) { - case 1: - d[0] = d1[0]+d2[0]; - break; case 2: - d[0] = d1[0]+d2[0]; + TD(d) = TD(d1)+TD(d2); d[1] = d1[1]+d2[1]; break; + case 3: + TD(d) = TD(d1)+TD(d2); + d[1] = d1[1]+d2[1]; + d[2] = d1[2]+d2[2]; + break; default: - for ( i = 0; i < nd_wpd; i++ ) - d[i] = d1[i]+d2[i]; + for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]+d2[i]; break; } +#else + for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]+d2[i]; +#endif } -INLINE void ndl_add2(unsigned int *d1,unsigned int *d2) +/* d1 += d2 */ +INLINE void ndl_addto(UINT *d1,UINT *d2) { int i; +#if 1 switch ( nd_wpd ) { - case 1: - d2[0] += d1[0]; - break; case 2: - d2[0] += d1[0]; - d2[1] += d1[1]; + TD(d1) += TD(d2); + d1[1] += d2[1]; break; + case 3: + TD(d1) += TD(d2); + d1[1] += d2[1]; + d1[2] += d2[2]; + break; default: - for ( i = 0; i < nd_wpd; i++ ) - d2[i] += d1[i]; + for ( i = 0; i < nd_wpd; i++ ) d1[i] += d2[i]; break; } +#else + for ( i = 0; i < nd_wpd; i++ ) d1[i] += d2[i]; +#endif } -void ndl_sub(unsigned int *d1,unsigned int *d2,unsigned int *d) +INLINE void ndl_sub(UINT *d1,UINT *d2,UINT *d) { int i; - for ( i = 0; i < nd_wpd; i++ ) - d[i] = d1[i]-d2[i]; + for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]-d2[i]; } -int ndl_disjoint(unsigned int *d1,unsigned int *d2) +int ndl_disjoint(UINT *d1,UINT *d2) { - unsigned int t1,t2,u,u1,u2; + UINT t1,t2,u,u1,u2; int i,j; +#if USE_UNROLL switch ( nd_bpe ) { + case 3: + for ( i = nd_exporigin; i < nd_wpd; i++ ) { + u1 = d1[i]; u2 = d2[i]; + t1 = u1&0x38000000; t2 = u2&0x38000000; if ( t1&&t2 ) return 0; + t1 = u1& 0x7000000; t2 = u2& 0x7000000; if ( t1&&t2 ) return 0; + t1 = u1& 0xe00000; t2 = u2& 0xe00000; if ( t1&&t2 ) return 0; + t1 = u1& 0x1c0000; t2 = u2& 0x1c0000; if ( t1&&t2 ) return 0; + t1 = u1& 0x38000; t2 = u2& 0x38000; if ( t1&&t2 ) return 0; + t1 = u1& 0x7000; t2 = u2& 0x7000; if ( t1&&t2 ) return 0; + t1 = u1& 0xe00; t2 = u2& 0xe00; if ( t1&&t2 ) return 0; + t1 = u1& 0x1c0; t2 = u2& 0x1c0; if ( t1&&t2 ) return 0; + t1 = u1& 0x38; t2 = u2& 0x38; if ( t1&&t2 ) return 0; + t1 = u1& 0x7; t2 = u2& 0x7; if ( t1&&t2 ) return 0; + } + return 1; + break; case 4: - for ( i = 0; i < nd_wpd; i++ ) { + for ( i = nd_exporigin; i < nd_wpd; i++ ) { u1 = d1[i]; u2 = d2[i]; t1 = u1&0xf0000000; t2 = u2&0xf0000000; if ( t1&&t2 ) return 0; - t1 = u1&0xf000000; t2 = u2&0xf000000; if ( t1&&t2 ) return 0; - t1 = u1&0xf00000; t2 = u2&0xf00000; if ( t1&&t2 ) return 0; - t1 = u1&0xf0000; t2 = u2&0xf0000; if ( t1&&t2 ) return 0; - t1 = u1&0xf000; t2 = u2&0xf000; if ( t1&&t2 ) return 0; - t1 = u1&0xf00; t2 = u2&0xf00; if ( t1&&t2 ) return 0; - t1 = u1&0xf0; t2 = u2&0xf0; if ( t1&&t2 ) return 0; - t1 = u1&0xf; t2 = u2&0xf; if ( t1&&t2 ) return 0; + t1 = u1& 0xf000000; t2 = u2& 0xf000000; if ( t1&&t2 ) return 0; + t1 = u1& 0xf00000; t2 = u2& 0xf00000; if ( t1&&t2 ) return 0; + t1 = u1& 0xf0000; t2 = u2& 0xf0000; if ( t1&&t2 ) return 0; + t1 = u1& 0xf000; t2 = u2& 0xf000; if ( t1&&t2 ) return 0; + t1 = u1& 0xf00; t2 = u2& 0xf00; if ( t1&&t2 ) return 0; + t1 = u1& 0xf0; t2 = u2& 0xf0; if ( t1&&t2 ) return 0; + t1 = u1& 0xf; t2 = u2& 0xf; if ( t1&&t2 ) return 0; } return 1; break; case 6: - for ( i = 0; i < nd_wpd; i++ ) { + for ( i = nd_exporigin; i < nd_wpd; i++ ) { u1 = d1[i]; u2 = d2[i]; t1 = u1&0x3f000000; t2 = u2&0x3f000000; if ( t1&&t2 ) return 0; - t1 = u1&0xfc0000; t2 = u2&0xfc0000; if ( t1&&t2 ) return 0; - t1 = u1&0x3f000; t2 = u2&0x3f000; if ( t1&&t2 ) return 0; - t1 = u1&0xfc0; t2 = u2&0xfc0; if ( t1&&t2 ) return 0; - t1 = u1&0x3f; t2 = u2&0x3f; if ( t1&&t2 ) return 0; + t1 = u1& 0xfc0000; t2 = u2& 0xfc0000; if ( t1&&t2 ) return 0; + t1 = u1& 0x3f000; t2 = u2& 0x3f000; if ( t1&&t2 ) return 0; + t1 = u1& 0xfc0; t2 = u2& 0xfc0; if ( t1&&t2 ) return 0; + t1 = u1& 0x3f; t2 = u2& 0x3f; if ( t1&&t2 ) return 0; } return 1; break; case 8: - for ( i = 0; i < nd_wpd; i++ ) { + for ( i = nd_exporigin; i < nd_wpd; i++ ) { u1 = d1[i]; u2 = d2[i]; t1 = u1&0xff000000; t2 = u2&0xff000000; if ( t1&&t2 ) return 0; - t1 = u1&0xff0000; t2 = u2&0xff0000; if ( t1&&t2 ) return 0; - t1 = u1&0xff00; t2 = u2&0xff00; if ( t1&&t2 ) return 0; - t1 = u1&0xff; t2 = u2&0xff; if ( t1&&t2 ) return 0; + t1 = u1& 0xff0000; t2 = u2& 0xff0000; if ( t1&&t2 ) return 0; + t1 = u1& 0xff00; t2 = u2& 0xff00; if ( t1&&t2 ) return 0; + t1 = u1& 0xff; t2 = u2& 0xff; if ( t1&&t2 ) return 0; } return 1; break; case 16: - for ( i = 0; i < nd_wpd; i++ ) { + for ( i = nd_exporigin; i < nd_wpd; i++ ) { u1 = d1[i]; u2 = d2[i]; t1 = u1&0xffff0000; t2 = u2&0xffff0000; if ( t1&&t2 ) return 0; - t1 = u1&0xffff; t2 = u2&0xffff; if ( t1&&t2 ) return 0; + t1 = u1& 0xffff; t2 = u2& 0xffff; if ( t1&&t2 ) return 0; } return 1; break; case 32: - for ( i = 0; i < nd_wpd; i++ ) + for ( i = nd_exporigin; i < nd_wpd; i++ ) if ( d1[i] && d2[i] ) return 0; return 1; break; default: - for ( i = 0; i < nd_wpd; i++ ) { + for ( i = nd_exporigin; i < nd_wpd; i++ ) { u1 = d1[i]; u2 = d2[i]; for ( j = 0; j < nd_epw; j++ ) { if ( (u1&nd_mask0) && (u2&nd_mask0) ) return 0; @@ -587,19 +931,46 @@ int ndl_disjoint(unsigned int *d1,unsigned int *d2) return 1; break; } +#else + for ( i = nd_exporigin; i < nd_wpd; i++ ) { + u1 = d1[i]; u2 = d2[i]; + for ( j = 0; j < nd_epw; j++ ) { + if ( (u1&nd_mask0) && (u2&nd_mask0) ) return 0; + u1 >>= nd_bpe; u2 >>= nd_bpe; + } + } + return 1; +#endif } -int ndl_check_bound2(int index,unsigned int *d2) +int ndl_check_bound2(int index,UINT *d2) { - unsigned int u2; - unsigned int *d1; + UINT u2; + UINT *d1; int i,j,ind,k; d1 = nd_bound[index]; ind = 0; +#if USE_UNROLL switch ( nd_bpe ) { + case 3: + for ( i = nd_exporigin; i < nd_wpd; i++ ) { + u2 = d2[i]; + if ( d1[ind++]+((u2>>27)&0x7) >= 0x8 ) return 1; + if ( d1[ind++]+((u2>>24)&0x7) >= 0x8 ) return 1; + if ( d1[ind++]+((u2>>21)&0x7) >= 0x8 ) return 1; + if ( d1[ind++]+((u2>>18)&0x7) >= 0x8 ) return 1; + if ( d1[ind++]+((u2>>15)&0x7) >= 0x8 ) return 1; + if ( d1[ind++]+((u2>>12)&0x7) >= 0x8 ) return 1; + if ( d1[ind++]+((u2>>9)&0x7) >= 0x8 ) return 1; + if ( d1[ind++]+((u2>>6)&0x7) >= 0x8 ) return 1; + if ( d1[ind++]+((u2>>3)&0x7) >= 0x8 ) return 1; + if ( d1[ind++]+(u2&0x7) >= 0x8 ) return 1; + } + return 0; + break; case 4: - for ( i = 0; i < nd_wpd; i++ ) { + for ( i = nd_exporigin; i < nd_wpd; i++ ) { u2 = d2[i]; if ( d1[ind++]+((u2>>28)&0xf) >= 0x10 ) return 1; if ( d1[ind++]+((u2>>24)&0xf) >= 0x10 ) return 1; @@ -613,7 +984,7 @@ int ndl_check_bound2(int index,unsigned int *d2) return 0; break; case 6: - for ( i = 0; i < nd_wpd; i++ ) { + for ( i = nd_exporigin; i < nd_wpd; i++ ) { u2 = d2[i]; if ( d1[ind++]+((u2>>24)&0x3f) >= 0x40 ) return 1; if ( d1[ind++]+((u2>>18)&0x3f) >= 0x40 ) return 1; @@ -624,7 +995,7 @@ int ndl_check_bound2(int index,unsigned int *d2) return 0; break; case 8: - for ( i = 0; i < nd_wpd; i++ ) { + for ( i = nd_exporigin; i < nd_wpd; i++ ) { u2 = d2[i]; if ( d1[ind++]+((u2>>24)&0xff) >= 0x100 ) return 1; if ( d1[ind++]+((u2>>16)&0xff) >= 0x100 ) return 1; @@ -634,7 +1005,7 @@ int ndl_check_bound2(int index,unsigned int *d2) return 0; break; case 16: - for ( i = 0; i < nd_wpd; i++ ) { + for ( i = nd_exporigin; i < nd_wpd; i++ ) { u2 = d2[i]; if ( d1[ind++]+((u2>>16)&0xffff) > 0x10000 ) return 1; if ( d1[ind++]+(u2&0xffff) > 0x10000 ) return 1; @@ -642,12 +1013,12 @@ int ndl_check_bound2(int index,unsigned int *d2) return 0; break; case 32: - for ( i = 0; i < nd_wpd; i++ ) + for ( i = nd_exporigin; i < nd_wpd; i++ ) if ( d1[i]+d2[i]>k)&nd_mask0) > nd_mask0 ) return 1; + } + return 0; +#endif } -int ndl_check_bound2_direct(unsigned int *d1,unsigned int *d2) +int ndl_check_bound2_direct(UINT *d1,UINT *d2) { - unsigned int u1,u2; - int i,j,k; + UINT u2; + int i,j,ind,k; + ind = 0; +#if USE_UNROLL switch ( nd_bpe ) { + case 3: + for ( i = nd_exporigin; i < nd_wpd; i++ ) { + u2 = d2[i]; + if ( d1[ind++]+((u2>>27)&0x7) >= 0x8 ) return 1; + if ( d1[ind++]+((u2>>24)&0x7) >= 0x8 ) return 1; + if ( d1[ind++]+((u2>>21)&0x7) >= 0x8 ) return 1; + if ( d1[ind++]+((u2>>18)&0x7) >= 0x8 ) return 1; + if ( d1[ind++]+((u2>>15)&0x7) >= 0x8 ) return 1; + if ( d1[ind++]+((u2>>12)&0x7) >= 0x8 ) return 1; + if ( d1[ind++]+((u2>>9)&0x7) >= 0x8 ) return 1; + if ( d1[ind++]+((u2>>6)&0x7) >= 0x8 ) return 1; + if ( d1[ind++]+((u2>>3)&0x7) >= 0x8 ) return 1; + if ( d1[ind++]+(u2&0x7) >= 0x8 ) return 1; + } + return 0; + break; case 4: - for ( i = 0; i < nd_wpd; i++ ) { - u1 = d1[i]; u2 = d2[i]; - if ( ((u1>>28)&0xf)+((u2>>28)&0xf) >= 0x10 ) return 1; - if ( ((u1>>24)&0xf)+((u2>>24)&0xf) >= 0x10 ) return 1; - if ( ((u1>>20)&0xf)+((u2>>20)&0xf) >= 0x10 ) return 1; - if ( ((u1>>16)&0xf)+((u2>>16)&0xf) >= 0x10 ) return 1; - if ( ((u1>>12)&0xf)+((u2>>12)&0xf) >= 0x10 ) return 1; - if ( ((u1>>8)&0xf)+((u2>>8)&0xf) >= 0x10 ) return 1; - if ( ((u1>>4)&0xf)+((u2>>4)&0xf) >= 0x10 ) return 1; - if ( (u1&0xf)+(u2&0xf) >= 0x10 ) return 1; + for ( i = nd_exporigin; i < nd_wpd; i++ ) { + u2 = d2[i]; + if ( d1[ind++]+((u2>>28)&0xf) >= 0x10 ) return 1; + if ( d1[ind++]+((u2>>24)&0xf) >= 0x10 ) return 1; + if ( d1[ind++]+((u2>>20)&0xf) >= 0x10 ) return 1; + if ( d1[ind++]+((u2>>16)&0xf) >= 0x10 ) return 1; + if ( d1[ind++]+((u2>>12)&0xf) >= 0x10 ) return 1; + if ( d1[ind++]+((u2>>8)&0xf) >= 0x10 ) return 1; + if ( d1[ind++]+((u2>>4)&0xf) >= 0x10 ) return 1; + if ( d1[ind++]+(u2&0xf) >= 0x10 ) return 1; } return 0; break; case 6: - for ( i = 0; i < nd_wpd; i++ ) { - u1 = d1[i]; u2 = d2[i]; - if ( ((u1>>24)&0x3f)+((u2>>24)&0x3f) >= 0x40 ) return 1; - if ( ((u1>>18)&0x3f)+((u2>>18)&0x3f) >= 0x40 ) return 1; - if ( ((u1>>12)&0x3f)+((u2>>12)&0x3f) >= 0x40 ) return 1; - if ( ((u1>>6)&0x3f)+((u2>>6)&0x3f) >= 0x40 ) return 1; - if ( (u1&0x3f)+(u2&0x3f) >= 0x40 ) return 1; + for ( i = nd_exporigin; i < nd_wpd; i++ ) { + u2 = d2[i]; + if ( d1[ind++]+((u2>>24)&0x3f) >= 0x40 ) return 1; + if ( d1[ind++]+((u2>>18)&0x3f) >= 0x40 ) return 1; + if ( d1[ind++]+((u2>>12)&0x3f) >= 0x40 ) return 1; + if ( d1[ind++]+((u2>>6)&0x3f) >= 0x40 ) return 1; + if ( d1[ind++]+(u2&0x3f) >= 0x40 ) return 1; } return 0; break; case 8: - for ( i = 0; i < nd_wpd; i++ ) { - u1 = d1[i]; u2 = d2[i]; - if ( ((u1>>24)&0xff)+((u2>>24)&0xff) >= 0x100 ) return 1; - if ( ((u1>>16)&0xff)+((u2>>16)&0xff) >= 0x100 ) return 1; - if ( ((u1>>8)&0xff)+((u2>>8)&0xff) >= 0x100 ) return 1; - if ( (u1&0xff)+(u2&0xff) >= 0x100 ) return 1; + for ( i = nd_exporigin; i < nd_wpd; i++ ) { + u2 = d2[i]; + if ( d1[ind++]+((u2>>24)&0xff) >= 0x100 ) return 1; + if ( d1[ind++]+((u2>>16)&0xff) >= 0x100 ) return 1; + if ( d1[ind++]+((u2>>8)&0xff) >= 0x100 ) return 1; + if ( d1[ind++]+(u2&0xff) >= 0x100 ) return 1; } return 0; break; case 16: - for ( i = 0; i < nd_wpd; i++ ) { - u1 = d1[i]; u2 = d2[i]; - if ( ((u1>>16)&0xffff)+((u2>>16)&0xffff) > 0x10000 ) return 1; - if ( (u2&0xffff)+(u2&0xffff) > 0x10000 ) return 1; + for ( i = nd_exporigin; i < nd_wpd; i++ ) { + u2 = d2[i]; + if ( d1[ind++]+((u2>>16)&0xffff) > 0x10000 ) return 1; + if ( d1[ind++]+(u2&0xffff) > 0x10000 ) return 1; } return 0; break; case 32: - for ( i = 0; i < nd_wpd; i++ ) + for ( i = nd_exporigin; i < nd_wpd; i++ ) if ( d1[i]+d2[i]>k)&nd_mask0)+((u2>>k)&nd_mask0) > nd_mask0 ) - return 1; + if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1; } return 0; break; } +#else + for ( i = nd_exporigin; i < nd_wpd; i++ ) { + u2 = d2[i]; + k = (nd_epw-1)*nd_bpe; + for ( j = 0; j < nd_epw; j++, k -= nd_bpe ) + if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1; + } + return 0; +#endif } -INLINE int ndl_hash_value(int td,unsigned int *d) +INLINE int ndl_hash_value(UINT *d) { int i; int r; - r = td; + r = 0; for ( i = 0; i < nd_wpd; i++ ) r = ((r<<16)+d[i])%REDTAB_LEN; return r; } -INLINE int nd_find_reducer(ND g) +INLINE int ndl_find_reducer(UINT *dg) { RHist r; int d,k,i; - d = ndl_hash_value(HTD(g),HDL(g)); + d = ndl_hash_value(dg); for ( r = nd_red[d], k = 0; r; r = NEXT(r), k++ ) { - if ( HTD(g) == TD(r) && ndl_equal(HDL(g),DL(r)) ) { + if ( ndl_equal(dg,DL(r)) ) { if ( k > 0 ) nd_notfirst++; nd_found++; return r->index; } } - if ( Reverse ) for ( i = nd_psn-1; i >= 0; i-- ) { r = nd_psh[i]; - if ( HTD(g) >= TD(r) && ndl_reducible(HDL(g),DL(r)) ) { + if ( ndl_reducible(dg,DL(r)) ) { nd_create++; - nd_append_red(HDL(g),HTD(g),i); + nd_append_red(dg,i); return i; } } else for ( i = 0; i < nd_psn; i++ ) { r = nd_psh[i]; - if ( HTD(g) >= TD(r) && ndl_reducible(HDL(g),DL(r)) ) { + if ( ndl_reducible(dg,DL(r)) ) { nd_create++; - nd_append_red(HDL(g),HTD(g),i); + nd_append_red(dg,i); return i; } } return -1; } -INLINE int nd_find_reducer_direct(ND g,NDV *ps,int len) +INLINE int ndl_find_reducer_direct(UINT *dg,NDV *ps,int len) { - int i; NDV r; + RHist s; + int d,k,i; if ( Reverse ) for ( i = len-1; i >= 0; i-- ) { r = ps[i]; - if ( HTD(g) >= HTD(r) && ndl_reducible(HDL(g),HDL(r)) ) { - nd_append_red(HDL(g),HTD(g),i); + if ( ndl_reducible(dg,HDL(r)) ) return i; - } } else for ( i = 0; i < len; i++ ) { r = ps[i]; - if ( HTD(g) >= HTD(r) && ndl_reducible(HDL(g),HDL(r)) ) { - nd_append_red(HDL(g),HTD(g),i); + if ( ndl_reducible(dg,HDL(r)) ) return i; - } } return -1; } -ND nd_add(int mod,ND p1,ND p2) +ND nd_merge(ND p1,ND p2) { int n,c; - int t; + int t,can,td1,td2; ND r; NM m1,m2,mr0,mr,s; - if ( !p1 ) - return p2; - else if ( !p2 ) + if ( !p1 ) return p2; + else if ( !p2 ) return p1; + else { + can = 0; + for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) { + c = DL_COMPARE(DL(m1),DL(m2)); + switch ( c ) { + case 0: + s = m1; m1 = NEXT(m1); + can++; NEXTNM2(mr0,mr,s); + s = m2; m2 = NEXT(m2); FREENM(s); + break; + case 1: + s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s); + break; + case -1: + s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s); + break; + } + } + if ( !mr0 ) + if ( m1 ) mr0 = m1; + else if ( m2 ) mr0 = m2; + else return 0; + else if ( m1 ) NEXT(mr) = m1; + else if ( m2 ) NEXT(mr) = m2; + else NEXT(mr) = 0; + BDY(p1) = mr0; + SG(p1) = MAX(SG(p1),SG(p2)); + LEN(p1) = LEN(p1)+LEN(p2)-can; + FREEND(p2); return p1; - else if ( !mod ) - return nd_add_q(p1,p2); + } +} + +ND nd_add(int mod,ND p1,ND p2) +{ + int n,c; + int t,can,td1,td2; + ND r; + NM m1,m2,mr0,mr,s; + + if ( !p1 ) return p2; + else if ( !p2 ) return p1; + else if ( !mod ) return nd_add_q(p1,p2); else { + can = 0; for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) { - if ( TD(m1) > TD(m2) ) - c = 1; - else if ( TD(m1) < TD(m2) ) - c = -1; - else - c = ndl_compare(DL(m1),DL(m2)); + c = DL_COMPARE(DL(m1),DL(m2)); switch ( c ) { case 0: t = ((CM(m1))+(CM(m2))) - mod; - if ( t < 0 ) - t += mod; + if ( t < 0 ) t += mod; s = m1; m1 = NEXT(m1); if ( t ) { - NEXTNM2(mr0,mr,s); CM(mr) = (t); + can++; NEXTNM2(mr0,mr,s); CM(mr) = (t); } else { - FREENM(s); + can += 2; FREENM(s); } s = m2; m2 = NEXT(m2); FREENM(s); break; @@ -838,20 +1274,15 @@ ND nd_add(int mod,ND p1,ND p2) } } if ( !mr0 ) - if ( m1 ) - mr0 = m1; - else if ( m2 ) - mr0 = m2; - else - return 0; - else if ( m1 ) - NEXT(mr) = m1; - else if ( m2 ) - NEXT(mr) = m2; - else - NEXT(mr) = 0; + if ( m1 ) mr0 = m1; + else if ( m2 ) mr0 = m2; + else return 0; + else if ( m1 ) NEXT(mr) = m1; + else if ( m2 ) NEXT(mr) = m2; + else NEXT(mr) = 0; BDY(p1) = mr0; SG(p1) = MAX(SG(p1),SG(p2)); + LEN(p1) = LEN(p1)+LEN(p2)-can; FREEND(p2); return p1; } @@ -859,31 +1290,25 @@ ND nd_add(int mod,ND p1,ND p2) ND nd_add_q(ND p1,ND p2) { - int n,c; + int n,c,can; ND r; NM m1,m2,mr0,mr,s; Q t; - if ( !p1 ) - return p2; - else if ( !p2 ) - return p1; + if ( !p1 ) return p2; + else if ( !p2 ) return p1; else { + can = 0; for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) { - if ( TD(m1) > TD(m2) ) - c = 1; - else if ( TD(m1) < TD(m2) ) - c = -1; - else - c = ndl_compare(DL(m1),DL(m2)); + c = DL_COMPARE(DL(m1),DL(m2)); switch ( c ) { case 0: addq(CQ(m1),CQ(m2),&t); s = m1; m1 = NEXT(m1); if ( t ) { - NEXTNM2(mr0,mr,s); CQ(mr) = (t); + can++; NEXTNM2(mr0,mr,s); CQ(mr) = (t); } else { - FREENM(s); + can += 2; FREENM(s); } s = m2; m2 = NEXT(m2); FREENM(s); break; @@ -896,34 +1321,28 @@ ND nd_add_q(ND p1,ND p2) } } if ( !mr0 ) - if ( m1 ) - mr0 = m1; - else if ( m2 ) - mr0 = m2; - else - return 0; - else if ( m1 ) - NEXT(mr) = m1; - else if ( m2 ) - NEXT(mr) = m2; - else - NEXT(mr) = 0; + if ( m1 ) mr0 = m1; + else if ( m2 ) mr0 = m2; + else return 0; + else if ( m1 ) NEXT(mr) = m1; + else if ( m2 ) NEXT(mr) = m2; + else NEXT(mr) = 0; BDY(p1) = mr0; SG(p1) = MAX(SG(p1),SG(p2)); + LEN(p1) = LEN(p1)+LEN(p2)-can; FREEND(p2); return p1; } } -#if 1 /* ret=1 : success, ret=0 : overflow */ -int nd_nf(int mod,ND g,int full,ND *rp) +int nd_nf(int mod,ND g,NDV *ps,int full,ND *rp) { ND d; NM m,mrd,tail; NM mul; int n,sugar,psugar,sugar0,stat,index; - int c,c1,c2; + int c,c1,c2,dummy; RHist h; NDV p,red; Q cg,cred,gcd; @@ -933,39 +1352,31 @@ int nd_nf(int mod,ND g,int full,ND *rp) *rp = 0; return 1; } - if ( !mod ) - hmag = ((double)p_mag((P)HCQ(g)))*nd_scale; + if ( !mod ) hmag = ((double)p_mag((P)HCQ(g)))*nd_scale; sugar0 = sugar = SG(g); n = NV(g); - mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(unsigned int)); + mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT)); for ( d = 0; g; ) { - index = nd_find_reducer(g); + index = ndl_find_reducer(HDL(g)); if ( index >= 0 ) { h = nd_psh[index]; ndl_sub(HDL(g),DL(h),DL(mul)); - TD(mul) = HTD(g)-TD(h); -#if 0 - if ( d && (SG(p)+TD(mul)) > sugar ) { - goto afo; - } -#endif if ( ndl_check_bound2(index,DL(mul)) ) { nd_free(g); nd_free(d); return 0; } + p = ps[index]; if ( mod ) { - p = nd_ps[index]; c1 = invm(HCM(p),mod); c2 = mod-HCM(g); DMAR(c1,c2,0,mod,c); CM(mul) = c; } else { - p = nd_psq[index]; igcd_cofactor(HCQ(g),HCQ(p),&gcd,&cg,&cred); chsgnq(cg,&CQ(mul)); nd_mul_c_q(d,cred); nd_mul_c_q(g,cred); } - g = nd_add(mod,g,ndv_mul_nm(mod,p,mul)); - sugar = MAX(sugar,SG(p)+TD(mul)); + g = nd_add(mod,g,ndv_mul_nm(mod,mul,p)); + sugar = MAX(sugar,SG(p)+TD(DL(mul))); if ( !mod && hmag && g && ((double)(p_mag((P)HCQ(g))) > hmag) ) { nd_removecont2(d,g); hmag = ((double)p_mag((P)HCQ(g)))*nd_scale; @@ -974,40 +1385,37 @@ int nd_nf(int mod,ND g,int full,ND *rp) *rp = g; return 1; } else { -afo: m = BDY(g); if ( NEXT(m) ) { - BDY(g) = NEXT(m); NEXT(m) = 0; + BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--; } else { FREEND(g); g = 0; } if ( d ) { - NEXT(tail)=m; - tail=m; + NEXT(tail)=m; tail=m; LEN(d)++; } else { - MKND(n,m,d); - tail = BDY(d); + MKND(n,m,1,d); tail = BDY(d); } } } - if ( d ) - SG(d) = sugar; + if ( d ) SG(d) = sugar; *rp = d; return 1; } -#else -int nd_nf(int mod,ND g,int full,ND *rp) + +int nd_nf_pbucket(int mod,ND g,NDV *ps,int full,ND *rp) { int hindex,index; NDV p; ND u,d,red; NODE l; - NM mul,m,mrd; + NM mul,m,mrd,tail; int sugar,psugar,n,h_reducible; PGeoBucket bucket; int c,c1,c2; - Q cg,cred,gcd; + Q cg,cred,gcd,zzz; RHist h; + double hmag,gmag; if ( !g ) { *rp = 0; @@ -1015,78 +1423,90 @@ int nd_nf(int mod,ND g,int full,ND *rp) } sugar = SG(g); n = NV(g); + if ( !mod ) hmag = ((double)p_mag((P)HCQ(g)))*nd_scale; bucket = create_pbucket(); - add_pbucket(mod,bucket,g,nd_length(g)); + add_pbucket(mod,bucket,g); d = 0; - mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(unsigned int)); + mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT)); while ( 1 ) { - hindex = head_pbucket(mod,bucket); + hindex = mod?head_pbucket(mod,bucket):head_pbucket_q(bucket); if ( hindex < 0 ) { - if ( d ) - SG(d) = sugar; + if ( d ) SG(d) = sugar; *rp = d; return 1; } g = bucket->body[hindex]; - index = nd_find_reducer(g); + index = ndl_find_reducer(HDL(g)); if ( index >= 0 ) { h = nd_psh[index]; ndl_sub(HDL(g),DL(h),DL(mul)); - TD(mul) = HTD(g)-TD(h); if ( ndl_check_bound2(index,DL(mul)) ) { - free_pbucket(bucket); nd_free(d); + free_pbucket(bucket); *rp = 0; return 0; } + p = ps[index]; if ( mod ) { - p = nd_ps[index]; c1 = invm(HCM(p),mod); c2 = mod-HCM(g); DMAR(c1,c2,0,mod,c); CM(mul) = c; } else { - p = nd_psq[index]; igcd_cofactor(HCQ(g),HCQ(p),&gcd,&cg,&cred); chsgnq(cg,&CQ(mul)); - nd_mul_c_q(d,cred); nd_mul_c_q(g,cred); + nd_mul_c_q(d,cred); + mulq_pbucket(bucket,cred); + g = bucket->body[hindex]; + gmag = (double)p_mag((P)HCQ(g)); } - red = ndv_mul_nm(mod,p,mul); + red = ndv_mul_nm(mod,mul,p); bucket->body[hindex] = nd_remove_head(g); red = nd_remove_head(red); - add_pbucket(mod,bucket,red,LEN(p)); - sugar = MAX(sugar,SG(p)+TD(mul)); + add_pbucket(mod,bucket,red); + psugar = SG(p)+TD(DL(mul)); + sugar = MAX(sugar,psugar); + if ( !mod && hmag && (gmag > hmag) ) { + g = normalize_pbucket(mod,bucket); + if ( !g ) { + if ( d ) SG(d) = sugar; + *rp = d; + return 1; + } + nd_removecont2(d,g); + hmag = ((double)p_mag((P)HCQ(g)))*nd_scale; + add_pbucket(mod,bucket,g); + } } else if ( !full ) { g = normalize_pbucket(mod,bucket); - if ( g ) - SG(g) = sugar; + if ( g ) SG(g) = sugar; *rp = g; return 1; } else { m = BDY(g); if ( NEXT(m) ) { - BDY(g) = NEXT(m); NEXT(m) = 0; + BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--; } else { FREEND(g); g = 0; } bucket->body[hindex] = g; NEXT(m) = 0; if ( d ) { - for ( mrd = BDY(d); NEXT(mrd); mrd = NEXT(mrd) ); - NEXT(mrd) = m; + NEXT(tail)=m; tail=m; LEN(d)++; } else { - MKND(n,m,d); + MKND(n,m,1,d); tail = BDY(d); } } } } -#endif -int nd_nf_direct(int mod,ND g,NDV *ps,int len,int full,ND *rp) +int nd_nf_direct(int mod,ND g,BaseSet base,int full,ND *rp) { ND d; NM m,mrd,tail; NM mul; - int n,sugar,psugar,sugar0,stat,index; + NDV *ps; + int n,sugar,psugar,sugar0,stat,index,len; int c,c1,c2; + UINT **bound; RHist h; NDV p,red; Q cg,cred,gcd; @@ -1104,16 +1524,18 @@ int nd_nf_direct(int mod,ND g,NDV *ps,int len,int full hmag = 0; #endif + ps = base->ps; + bound = base->bound; + len = base->len; sugar0 = sugar = SG(g); n = NV(g); - mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(unsigned int)); + mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT)); for ( d = 0; g; ) { - index = nd_find_reducer_direct(g,ps,len); + index = ndl_find_reducer_direct(HDL(g),ps,len); if ( index >= 0 ) { p = ps[index]; ndl_sub(HDL(g),HDL(p),DL(mul)); - TD(mul) = HTD(g)-HTD(p); - if ( ndl_check_bound2_direct(HDL(p),DL(mul)) ) { + if ( ndl_check_bound2_direct(bound[index],DL(mul)) ) { nd_free(g); nd_free(d); return 0; } @@ -1125,8 +1547,8 @@ int nd_nf_direct(int mod,ND g,NDV *ps,int len,int full chsgnq(cg,&CQ(mul)); nd_mul_c_q(d,cred); nd_mul_c_q(g,cred); } - g = nd_add(mod,g,ndv_mul_nm(mod,p,mul)); - sugar = MAX(sugar,SG(p)+TD(mul)); + g = nd_add(mod,g,ndv_mul_nm(mod,mul,p)); + sugar = MAX(sugar,SG(p)+TD(DL(mul))); if ( !mod && hmag && g && ((double)(p_mag((P)HCQ(g))) > hmag) ) { nd_removecont2(d,g); hmag = ((double)p_mag((P)HCQ(g)))*nd_scale; @@ -1137,35 +1559,66 @@ int nd_nf_direct(int mod,ND g,NDV *ps,int len,int full } else { m = BDY(g); if ( NEXT(m) ) { - BDY(g) = NEXT(m); NEXT(m) = 0; + BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--; } else { FREEND(g); g = 0; } if ( d ) { - NEXT(tail)=m; - tail=m; + NEXT(tail)=m; tail=m; LEN(d)++; } else { - MKND(n,m,d); - tail = BDY(d); + MKND(n,m,1,d); tail = BDY(d); } } } - if ( d ) - SG(d) = sugar; + if ( d ) SG(d) = sugar; *rp = d; return 1; } +/* input : list of NDV, cand : list of NDV */ + +int ndv_check_candidate(NODE input,int obpe,int oadv,EPOS oepos,NODE cand) +{ + int n,i,stat; + ND nf,d; + NDV r; + NODE t,s; + + ndv_setup(0,0,cand); + n = length(cand); + + /* membercheck : list is a subset of Id(cand) ? */ + for ( t = input; t; t = NEXT(t) ) { +again: + if ( nd_bpe > obpe ) + r = ndv_dup_realloc((NDV)BDY(t),obpe,oadv,oepos); + else + r = (NDV)BDY(t); + d = ndvtond(0,r); + stat = nd_nf(0,d,nd_ps,0,&nf); + if ( !stat ) { + nd_reconstruct(0,0,0); + goto again; + } else if ( nf ) return 0; + printf("."); fflush(stdout); + } + printf("\n"); + /* gbcheck : cand is a GB of Id(cand) ? */ + if ( !nd_gb(0,1) ) return 0; + /* XXX */ + return 1; +} + ND nd_remove_head(ND p) { NM m; m = BDY(p); if ( !NEXT(m) ) { - FREEND(p); - p = 0; - } else - BDY(p) = NEXT(m); + FREEND(p); p = 0; + } else { + BDY(p) = NEXT(m); LEN(p)--; + } FREENM(m); return p; } @@ -1182,7 +1635,7 @@ PGeoBucket create_pbucket() void free_pbucket(PGeoBucket b) { int i; - for ( i = 0; i < b->m; i++ ) + for ( i = 0; i <= b->m; i++ ) if ( b->body[i] ) { nd_free(b->body[i]); b->body[i] = 0; @@ -1190,14 +1643,35 @@ void free_pbucket(PGeoBucket b) { GC_free(b); } -void add_pbucket(int mod,PGeoBucket g,ND d,int l) +void add_pbucket_symbolic(PGeoBucket g,ND d) { - int k,m; + int l,i,k,m; - for ( k = 0, m = 1; l > m; k++, m <<= 2 ); - /* 4^(k-1) < l <= 4^k */ + if ( !d ) + return; + l = LEN(d); + for ( k = 0, m = 1; l > m; k++, m <<= 1 ); + /* 2^(k-1) < l <= 2^k (=m) */ + d = nd_merge(g->body[k],d); + for ( ; d && LEN(d) > m; k++, m <<= 1 ) { + g->body[k] = 0; + d = nd_merge(g->body[k+1],d); + } + g->body[k] = d; + g->m = MAX(g->m,k); +} + +void add_pbucket(int mod,PGeoBucket g,ND d) +{ + int l,i,k,m; + + if ( !d ) + return; + l = LEN(d); + for ( k = 0, m = 1; l > m; k++, m <<= 1 ); + /* 2^(k-1) < l <= 2^k (=m) */ d = nd_add(mod,g->body[k],d); - for ( ; d && nd_length(d) > 1<<(2*k); k++ ) { + for ( ; d && LEN(d) > m; k++, m <<= 1 ) { g->body[k] = 0; d = nd_add(mod,g->body[k+1],d); } @@ -1205,11 +1679,50 @@ void add_pbucket(int mod,PGeoBucket g,ND d,int l) g->m = MAX(g->m,k); } -/* XXX not completed */ +void mulq_pbucket(PGeoBucket g,Q c) +{ + int k; + + for ( k = 0; k <= g->m; k++ ) + nd_mul_c_q(g->body[k],c); +} + +NM remove_head_pbucket_symbolic(PGeoBucket g) +{ + int j,i,k,c; + NM head; + + k = g->m; + j = -1; + for ( i = 0; i <= k; i++ ) { + if ( !g->body[i] ) continue; + if ( j < 0 ) j = i; + else { + c = DL_COMPARE(HDL(g->body[i]),HDL(g->body[j])); + if ( c > 0 ) + j = i; + else if ( c == 0 ) + g->body[i] = nd_remove_head(g->body[i]); + } + } + if ( j < 0 ) return 0; + else { + head = BDY(g->body[j]); + if ( !NEXT(head) ) { + FREEND(g->body[j]); + g->body[j] = 0; + } else { + BDY(g->body[j]) = NEXT(head); + LEN(g->body[j])--; + } + return head; + } +} + int head_pbucket(int mod,PGeoBucket g) { int j,i,c,k,nv,sum; - unsigned int *di,*dj; + UINT *di,*dj; ND gi,gj; k = g->m; @@ -1224,33 +1737,22 @@ int head_pbucket(int mod,PGeoBucket g) dj = HDL(gj); sum = HCM(gj); } else { - di = HDL(gi); - nv = NV(gi); - if ( HTD(gi) > HTD(gj) ) - c = 1; - else if ( HTD(gi) < HTD(gj) ) - c = -1; - else - c = ndl_compare(di,dj); + c = DL_COMPARE(HDL(gi),dj); if ( c > 0 ) { - if ( sum ) - HCM(gj) = sum; - else - g->body[j] = nd_remove_head(gj); + if ( sum ) HCM(gj) = sum; + else g->body[j] = nd_remove_head(gj); j = i; gj = g->body[j]; dj = HDL(gj); sum = HCM(gj); } else if ( c == 0 ) { sum = sum+HCM(gi)-mod; - if ( sum < 0 ) - sum += mod; + if ( sum < 0 ) sum += mod; g->body[i] = nd_remove_head(gi); } } } - if ( j < 0 ) - return -1; + if ( j < 0 ) return -1; else if ( sum ) { HCM(gj) = sum; return j; @@ -1259,24 +1761,70 @@ int head_pbucket(int mod,PGeoBucket g) } } +int head_pbucket_q(PGeoBucket g) +{ + int j,i,c,k,nv; + Q sum,t; + ND gi,gj; + + k = g->m; + while ( 1 ) { + j = -1; + for ( i = 0; i <= k; i++ ) { + if ( !(gi = g->body[i]) ) continue; + if ( j < 0 ) { + j = i; + gj = g->body[j]; + sum = HCQ(gj); + } else { + nv = NV(gi); + c = DL_COMPARE(HDL(gi),HDL(gj)); + if ( c > 0 ) { + if ( sum ) HCQ(gj) = sum; + else g->body[j] = nd_remove_head(gj); + j = i; + gj = g->body[j]; + sum = HCQ(gj); + } else if ( c == 0 ) { + addq(sum,HCQ(gi),&t); + sum = t; + g->body[i] = nd_remove_head(gi); + } + } + } + if ( j < 0 ) return -1; + else if ( sum ) { + HCQ(gj) = sum; + return j; + } else + g->body[j] = nd_remove_head(gj); + } +} + ND normalize_pbucket(int mod,PGeoBucket g) { int i; ND r,t; r = 0; - for ( i = 0; i <= g->m; i++ ) + for ( i = 0; i <= g->m; i++ ) { r = nd_add(mod,r,g->body[i]); + g->body[i] = 0; + } + g->m = -1; return r; } -NODE nd_gb(int m) +/* return value = 0 => input is not a GB */ + +NODE nd_gb(int m,int checkonly) { int i,nh,sugar,stat; NODE r,g,t; ND_pairs d; ND_pairs l; ND h,nf; + NDV nfv; g = 0; d = 0; for ( i = 0; i < nd_psn; i++ ) { @@ -1291,20 +1839,27 @@ again: sugar = SG(l); fprintf(asir_out,"%d",sugar); } - stat = nd_sp(m,l,&h); + stat = nd_sp(m,0,l,&h); if ( !stat ) { NEXT(l) = d; d = l; d = nd_reconstruct(m,0,d); goto again; } - stat = nd_nf(m,h,!Top,&nf); +#if USE_GEOBUCKET + stat = m?nd_nf_pbucket(m,h,nd_ps,!Top,&nf):nd_nf(m,h,nd_ps,!Top,&nf); +#else + stat = nd_nf(m,h,nd_ps,!Top,&nf); +#endif if ( !stat ) { NEXT(l) = d; d = l; d = nd_reconstruct(m,0,d); goto again; } else if ( nf ) { + if ( checkonly ) return 0; printf("+"); fflush(stdout); - nh = nd_newps(m,nf); + nd_removecont(m,nf); + nfv = ndtondv(m,nf); nd_free(nf); + nh = ndv_newps(nfv,0); d = update_pairs(d,g,nh); g = update_base(g,nh); FREENDP(l); @@ -1313,12 +1868,7 @@ again: FREENDP(l); } } - if ( m ) - for ( t = g; t; t = NEXT(t) ) - BDY(t) = (pointer)nd_ps[(int)BDY(t)]; - else - for ( t = g; t; t = NEXT(t) ) - BDY(t) = (pointer)nd_psq[(int)BDY(t)]; + for ( t = g; t; t = NEXT(t) ) BDY(t) = (pointer)nd_ps[(int)BDY(t)]; return g; } @@ -1329,6 +1879,7 @@ NODE nd_gb_trace(int m) ND_pairs d; ND_pairs l; ND h,nf,nfq; + NDV nfv,nfqv; g = 0; d = 0; for ( i = 0; i < nd_psn; i++ ) { @@ -1343,24 +1894,33 @@ again: sugar = SG(l); fprintf(asir_out,"%d",sugar); } - stat = nd_sp(m,l,&h); + stat = nd_sp(m,0,l,&h); if ( !stat ) { NEXT(l) = d; d = l; d = nd_reconstruct(m,1,d); goto again; } - stat = nd_nf(m,h,!Top,&nf); +#if USE_GEOBUCKET + stat = nd_nf_pbucket(m,h,nd_ps,!Top,&nf); +#else + stat = nd_nf(m,h,nd_ps,!Top,&nf); +#endif if ( !stat ) { NEXT(l) = d; d = l; d = nd_reconstruct(m,1,d); goto again; } else if ( nf ) { /* overflow does not occur */ - nd_sp(0,l,&h); - nd_nf(0,h,!Top,&nfq); + nd_sp(0,1,l,&h); + nd_nf(0,h,nd_ps_trace,!Top,&nfq); if ( nfq ) { + /* failure; m|HC(nfq) */ + if ( !rem(NM(HCQ(nfq)),m) ) return 0; + printf("+"); fflush(stdout); - nh = nd_newps_trace(m,nf,nfq); + nd_removecont(m,nf); nfv = ndtondv(m,nf); nd_free(nf); + nd_removecont(0,nfq); nfqv = ndtondv(0,nfq); nd_free(nfq); + nh = ndv_newps(nfv,nfqv); d = update_pairs(d,g,nh); g = update_base(g,nh); } else { @@ -1372,61 +1932,69 @@ again: FREENDP(l); } for ( t = g; t; t = NEXT(t) ) - BDY(t) = (pointer)nd_psq[(int)BDY(t)]; + BDY(t) = (pointer)nd_ps_trace[(int)BDY(t)]; return g; } int ndv_compare(NDV *p1,NDV *p2) { - int td1,td2; - - td1 = HTD(*p1); td2 = HTD(*p2); - if ( td1 > td2 ) return 1; - else if ( td1 < td2 ) return -1; - else return ndl_compare(HDL(*p1),HDL(*p2)); + return DL_COMPARE(HDL(*p1),HDL(*p2)); } int ndv_compare_rev(NDV *p1,NDV *p2) { - return -ndv_compare(p1,p2); + return -DL_COMPARE(HDL(*p1),HDL(*p2)); } -NODE nd_reduceall(int m,NODE f) +NODE ndv_reduceall(int m,NODE f) { int i,j,n,stat; NDV *w,*ps; ND nf,g; NODE t,a0,a; + struct oBaseSet base; + UINT **bound; for ( n = 0, t = f; t; t = NEXT(t), n++ ); ps = (NDV *)ALLOCA(n*sizeof(NDV)); - for ( i = 0, t = f; i < n; i++, t = NEXT(t) ) - ps[i] = (NDV)BDY(t); + bound = (UINT **)ALLOCA(n*sizeof(UINT *)); + for ( i = 0, t = f; i < n; i++, t = NEXT(t) ) ps[i] = (NDV)BDY(t); qsort(ps,n,sizeof(NDV),(int (*)(const void *,const void *))ndv_compare); - w = (NDV *)ALLOCA((n-1)*sizeof(NDV)); - for ( i = 0; i < n; i++ ) { - for ( j = 0; j < i; j++ ) - w[j] = (NDV)ps[j]; - for ( j = i+1; j < n; j++ ) - w[j-1] = ps[j]; + for ( i = 0; i < n; i++ ) bound[i] = ndv_compute_bound(ps[i]); + base.ps = (NDV *)ALLOCA((n-1)*sizeof(NDV)); + base.bound = (UINT **)ALLOCA((n-1)*sizeof(UINT *)); + base.len = n-1; + i = 0; + while ( i < n ) { + for ( j = 0; j < i; j++ ) { + base.ps[j] = ps[j]; base.bound[j] = bound[j]; + } + for ( j = i+1; j < n; j++ ) { + base.ps[j-1] = ps[j]; base.bound[j-1] = bound[j]; + } g = ndvtond(m,ps[i]); - stat = nd_nf_direct(m,g,w,n-1,1,&nf); + stat = nd_nf_direct(m,g,&base,1,&nf); if ( !stat ) nd_reconstruct_direct(m,ps,n); else if ( !nf ) { printf("."); fflush(stdout); ndv_free(ps[i]); - for ( j = i+1; j < n; j++ ) - ps[j-1] = ps[j]; + for ( j = i+1; j < n; j++ ) { + ps[j-1] = ps[j]; bound[j-1] = bound[j]; + } n--; + base.len = n-1; } else { printf("."); fflush(stdout); ndv_free(ps[i]); nd_removecont(m,nf); ps[i] = ndtondv(m,nf); + bound[i] = ndv_compute_bound(ps[i]); nd_free(nf); + i++; } } + printf("\n"); for ( a0 = 0, i = 0; i < n; i++ ) { NEXTNODE(a0,a); BDY(a) = (pointer)ps[i]; @@ -1444,27 +2012,26 @@ ND_pairs update_pairs( ND_pairs d, NODE /* of index */ d1 = nd_newpairs(g,t); d1 = crit_M(d1); d1 = crit_F(d1); - prev = 0; cur = head = d1; - while ( cur ) { - if ( crit_2( cur->i1,cur->i2 ) ) { - remove = cur; - if ( !prev ) { - head = cur = NEXT(cur); + if ( do_weyl ) + head = d1; + else { + prev = 0; cur = head = d1; + while ( cur ) { + if ( crit_2( cur->i1,cur->i2 ) ) { + remove = cur; + if ( !prev ) head = cur = NEXT(cur); + else cur = NEXT(prev) = NEXT(cur); + FREENDP(remove); } else { - cur = NEXT(prev) = NEXT(cur); + prev = cur; cur = NEXT(cur); } - FREENDP(remove); - } else { - prev = cur; - cur = NEXT(cur); } } if ( !d ) return head; else { nd = d; - while ( NEXT(nd) ) - nd = NEXT(nd); + while ( NEXT(nd) ) nd = NEXT(nd); NEXT(nd) = head; return d; } @@ -1473,21 +2040,19 @@ ND_pairs update_pairs( ND_pairs d, NODE /* of index */ ND_pairs nd_newpairs( NODE g, int t ) { NODE h; - unsigned int *dl; - int td,ts,s; + UINT *dl; + int ts,s; ND_pairs r,r0; dl = DL(nd_psh[t]); - td = TD(nd_psh[t]); - ts = SG(nd_psh[t]) - td; + ts = SG(nd_psh[t]) - TD(dl); for ( r0 = 0, h = g; h; h = NEXT(h) ) { NEXTND_pairs(r0,r); r->i1 = (int)BDY(h); r->i2 = t; ndl_lcm(DL(nd_psh[r->i1]),dl,r->lcm); - TD(r) = ndl_td(r->lcm); - s = SG(nd_psh[r->i1])-TD(nd_psh[r->i1]); - SG(r) = MAX(s,ts) + TD(r); + s = SG(nd_psh[r->i1])-TD(DL(nd_psh[r->i1])); + SG(r) = MAX(s,ts) + TD(LCM(r)); } NEXT(r) = 0; return r0; @@ -1496,14 +2061,14 @@ ND_pairs nd_newpairs( NODE g, int t ) ND_pairs crit_B( ND_pairs d, int s ) { ND_pairs cur,head,prev,remove; - unsigned int *t,*tl,*lcm; + UINT *t,*tl,*lcm; int td,tdl; if ( !d ) return 0; t = DL(nd_psh[s]); prev = 0; head = cur = d; - lcm = (unsigned int *)ALLOCA(nd_wpd*sizeof(unsigned int)); + lcm = (UINT *)ALLOCA(nd_wpd*sizeof(UINT)); while ( cur ) { tl = cur->lcm; if ( ndl_reducible(tl,t) @@ -1517,8 +2082,7 @@ ND_pairs crit_B( ND_pairs d, int s ) } FREENDP(remove); } else { - prev = cur; - cur = NEXT(cur); + prev = cur; cur = NEXT(cur); } } return head; @@ -1527,29 +2091,23 @@ ND_pairs crit_B( ND_pairs d, int s ) ND_pairs crit_M( ND_pairs d1 ) { ND_pairs e,d2,d3,dd,p; - unsigned int *id,*jd; - int itd,jtd; + UINT *id,*jd; for ( dd = 0, e = d1; e; e = d3 ) { if ( !(d2 = NEXT(e)) ) { NEXT(e) = dd; return e; } - id = e->lcm; - itd = TD(e); + id = LCM(e); for ( d3 = 0; d2; d2 = p ) { - p = NEXT(d2), - jd = d2->lcm; - jtd = TD(d2); - if ( jtd == itd ) - if ( id == jd ); - else if ( ndl_reducible(jd,id) ) continue; - else if ( ndl_reducible(id,jd) ) goto delit; - else ; - else if ( jtd > itd ) + p = NEXT(d2); + jd = LCM(d2); + if ( ndl_equal(jd,id) ) + ; + else if ( TD(jd) > TD(id) ) if ( ndl_reducible(jd,id) ) continue; else ; - else if ( ndl_reducible(id,jd ) ) goto delit; + else if ( ndl_reducible(id,jd) ) goto delit; NEXT(d2) = d3; d3 = d2; } @@ -1616,25 +2174,21 @@ int crit_2( int dp1, int dp2 ) return ndl_disjoint(DL(nd_psh[dp1]),DL(nd_psh[dp2])); } -static ND_pairs equivalent_pairs( ND_pairs d1, ND_pairs *prest ) +ND_pairs equivalent_pairs( ND_pairs d1, ND_pairs *prest ) { ND_pairs w,p,r,s; - unsigned int *d; - int td; + UINT *d; w = d1; - d = w->lcm; - td = TD(w); + d = LCM(w); s = NEXT(w); NEXT(w) = 0; for ( r = 0; s; s = p ) { p = NEXT(s); - if ( td == TD(s) && ndl_equal(d,s->lcm) ) { - NEXT(s) = w; - w = s; + if ( ndl_equal(d,LCM(s)) ) { + NEXT(s) = w; w = s; } else { - NEXT(s) = r; - r = s; + NEXT(s) = r; r = s; } } *prest = r; @@ -1643,16 +2197,13 @@ static ND_pairs equivalent_pairs( ND_pairs d1, ND_pair NODE update_base(NODE nd,int ndp) { - unsigned int *dl, *dln; + UINT *dl, *dln; NODE last, p, head; - int td,tdn; dl = DL(nd_psh[ndp]); - td = TD(nd_psh[ndp]); for ( head = last = 0, p = nd; p; ) { dln = DL(nd_psh[(int)BDY(p)]); - tdn = TD(nd_psh[(int)BDY(p)]); - if ( tdn >= td && ndl_reducible( dln, dl ) ) { + if ( ndl_reducible( dln, dl ) ) { p = NEXT(p); if ( last ) NEXT(last) = p; } else { @@ -1667,46 +2218,20 @@ NODE update_base(NODE nd,int ndp) ND_pairs nd_minp( ND_pairs d, ND_pairs *prest ) { ND_pairs m,ml,p,l; - unsigned int *lcm; - int s,td,len,tlen,c; + UINT *lcm; + int s,td,len,tlen,c,c1; if ( !(p = NEXT(m = d)) ) { *prest = p; NEXT(m) = 0; return m; } - lcm = m->lcm; s = SG(m); - td = TD(m); - len = nd_psl[m->i1]+nd_psl[m->i2]; - for ( ml = 0, l = m; p; p = NEXT(l = p) ) { - if (SG(p) < s) - goto find; - else if ( SG(p) == s ) { - if ( TD(p) < td ) - goto find; - else if ( TD(p) == td ) { - c = ndl_compare(p->lcm,lcm); - if ( c < 0 ) - goto find; -#if 0 - else if ( c == 0 ) { - tlen = nd_psl[p->i1]+nd_psl[p->i2]; - if ( tlen < len ) - goto find; - } -#endif - } + for ( ml = 0, l = m; p; p = NEXT(l = p) ) + if ( (SG(p) < s) + || ((SG(p) == s) && (DL_COMPARE(LCM(p),LCM(m)) < 0)) ) { + ml = l; m = p; s = SG(m); } - continue; -find: - ml = l; - m = p; - lcm = m->lcm; - s = SG(m); - td = TD(m); - len = tlen; - } if ( !ml ) *prest = NEXT(m); else { NEXT(ml) = NEXT(m); @@ -1716,37 +2241,31 @@ find: return m; } -int nd_newps(int mod,ND a) +ND_pairs nd_minsugarp( ND_pairs d, ND_pairs *prest ) { - int len; - RHist r; - NDV b; + int msugar; + ND_pairs t,dm0,dm,dr0,dr; - if ( nd_psn == nd_pslen ) { - nd_pslen *= 2; - nd_psl = (int *)REALLOC((char *)nd_psl,nd_pslen*sizeof(int)); - nd_ps = (NDV *)REALLOC((char *)nd_ps,nd_pslen*sizeof(NDV)); - nd_psq = (NDV *)REALLOC((char *)nd_psq,nd_pslen*sizeof(NDV)); - nd_psh = (RHist *)REALLOC((char *)nd_psh,nd_pslen*sizeof(RHist)); - nd_bound = (unsigned int **) - REALLOC((char *)nd_bound,nd_pslen*sizeof(unsigned int *)); - } - nd_removecont(mod,a); - nd_bound[nd_psn] = nd_compute_bound(a); - NEWRHist(r); SG(r) = SG(a); TD(r) = HTD(a); ndl_copy(HDL(a),DL(r)); - nd_psh[nd_psn] = r; - b = ndtondv(mod,a); - len = LEN(b); - if ( mod ) - nd_ps[nd_psn] = b; - else - nd_psq[nd_psn] = b; - nd_psl[nd_psn] = len; - nd_free(a); - return nd_psn++; + for ( msugar = SG(d), t = NEXT(d); t; t = NEXT(t) ) + if ( SG(t) < msugar ) msugar = SG(t); + dm0 = 0; dr0 = 0; + for ( t = d; t; t = NEXT(t) ) + if ( SG(t) == msugar ) { + if ( dm0 ) NEXT(dm) = t; + else dm0 = t; + dm = t; + } else { + if ( dr0 ) NEXT(dr) = t; + else dr0 = t; + dr = t; + } + NEXT(dm) = 0; + if ( dr0 ) NEXT(dr) = 0; + *prest = dr0; + return dm0; } -int nd_newps_trace(int mod,ND nf,ND nfq) +int ndv_newps(NDV a,NDV aq) { int len; RHist r; @@ -1754,368 +2273,259 @@ int nd_newps_trace(int mod,ND nf,ND nfq) if ( nd_psn == nd_pslen ) { nd_pslen *= 2; - nd_psl = (int *)REALLOC((char *)nd_psl,nd_pslen*sizeof(int)); nd_ps = (NDV *)REALLOC((char *)nd_ps,nd_pslen*sizeof(NDV)); - nd_psq = (NDV *)REALLOC((char *)nd_psq,nd_pslen*sizeof(NDV)); + nd_ps_trace = (NDV *)REALLOC((char *)nd_ps_trace,nd_pslen*sizeof(NDV)); nd_psh = (RHist *)REALLOC((char *)nd_psh,nd_pslen*sizeof(RHist)); - nd_bound = (unsigned int **) - REALLOC((char *)nd_bound,nd_pslen*sizeof(unsigned int *)); + nd_bound = (UINT **) + REALLOC((char *)nd_bound,nd_pslen*sizeof(UINT *)); } - nd_removecont(mod,nf); - nd_ps[nd_psn] = ndtondv(mod,nf); - - nd_removecont(0,nfq); - nd_psq[nd_psn] = ndtondv(0,nfq); - - nd_bound[nd_psn] = nd_compute_bound(nfq); - NEWRHist(r); SG(r) = SG(nf); TD(r) = HTD(nf); ndl_copy(HDL(nf),DL(r)); - nd_psh[nd_psn] = r; - - len = LEN(nd_psq[nd_psn]); - nd_psl[nd_psn] = len; - - nd_free(nf); nd_free(nfq); + NEWRHist(r); nd_psh[nd_psn] = r; + nd_ps[nd_psn] = a; + if ( aq ) { + nd_ps_trace[nd_psn] = aq; + nd_bound[nd_psn] = ndv_compute_bound(aq); + SG(r) = SG(aq); ndl_copy(HDL(aq),DL(r)); + } else { + nd_bound[nd_psn] = ndv_compute_bound(a); + SG(r) = SG(a); ndl_copy(HDL(a),DL(r)); + } return nd_psn++; } -void nd_setup(int mod,NODE f) +void ndv_setup(int mod,int trace,NODE f) { int i,j,td,len,max; NODE s,s0,f0; - unsigned int *d; + UINT *d; RHist r; - NDV a; + NDV a,am; nd_found = 0; nd_notfirst = 0; nd_create = 0; nd_psn = length(f); nd_pslen = 2*nd_psn; - nd_psl = (int *)MALLOC(nd_pslen*sizeof(int)); nd_ps = (NDV *)MALLOC(nd_pslen*sizeof(NDV)); - nd_psq = (NDV *)MALLOC(nd_pslen*sizeof(NDV)); + nd_ps_trace = (NDV *)MALLOC(nd_pslen*sizeof(NDV)); nd_psh = (RHist *)MALLOC(nd_pslen*sizeof(RHist)); - nd_bound = (unsigned int **)MALLOC(nd_pslen*sizeof(unsigned int *)); - for ( max = 0, i = 0, s = f; i < nd_psn; i++, s = NEXT(s) ) { - nd_bound[i] = d = dp_compute_bound((DP)BDY(s)); - for ( j = 0; j < nd_nvar; j++ ) - max = MAX(d[j],max); - } - if ( !nd_red ) - nd_red = (RHist *)MALLOC(REDTAB_LEN*sizeof(RHist)); - bzero(nd_red,REDTAB_LEN*sizeof(RHist)); + nd_bound = (UINT **)MALLOC(nd_pslen*sizeof(UINT *)); - if ( max < 2 ) - nd_bpe = 2; - else if ( max < 4 ) - nd_bpe = 4; - else if ( max < 64 ) - nd_bpe = 6; - else if ( max < 256 ) - nd_bpe = 8; - else if ( max < 65536 ) - nd_bpe = 16; - else - nd_bpe = 32; - - nd_setup_parameters(); - nd_free_private_storage(); - len = 0; - for ( i = 0; i < nd_psn; i++, f = NEXT(f) ) { - NEWRHist(r); - a = dptondv(mod,(DP)BDY(f)); - ndv_removecont(mod,a); - len = MAX(len,LEN(a)); - SG(r) = HTD(a); TD(r) = HTD(a); ndl_copy(HDL(a),DL(r)); - if ( mod ) - nd_ps[i] = a; - else - nd_psq[i] = a; - nd_psh[i] = r; - } -} - -void nd_setup_trace(int mod,NODE f) -{ - int i,j,td,len,max; - NODE s,s0,f0; - unsigned int *d; - RHist r; - NDV a; - - nd_found = 0; nd_notfirst = 0; nd_create = 0; - - nd_psn = length(f); nd_pslen = 2*nd_psn; - nd_psl = (int *)MALLOC(nd_pslen*sizeof(int)); - nd_ps = (NDV *)MALLOC(nd_pslen*sizeof(NDV)); - nd_psq = (NDV *)MALLOC(nd_pslen*sizeof(NDV)); - nd_psh = (RHist *)MALLOC(nd_pslen*sizeof(RHist)); - nd_bound = (unsigned int **)MALLOC(nd_pslen*sizeof(unsigned int *)); - for ( max = 0, i = 0, s = f; i < nd_psn; i++, s = NEXT(s) ) { - nd_bound[i] = d = dp_compute_bound((DP)BDY(s)); - for ( j = 0; j < nd_nvar; j++ ) - max = MAX(d[j],max); - } if ( !nd_red ) nd_red = (RHist *)MALLOC(REDTAB_LEN*sizeof(RHist)); bzero(nd_red,REDTAB_LEN*sizeof(RHist)); - - if ( max < 2 ) - nd_bpe = 2; - else if ( max < 4 ) - nd_bpe = 4; - else if ( max < 64 ) - nd_bpe = 6; - else if ( max < 256 ) - nd_bpe = 8; - else if ( max < 65536 ) - nd_bpe = 16; - else - nd_bpe = 32; - - nd_setup_parameters(); nd_free_private_storage(); - len = 0; for ( i = 0; i < nd_psn; i++, f = NEXT(f) ) { - a = dptondv(mod,(DP)BDY(f)); ndv_removecont(mod,a); nd_ps[i] = a; - a = dptondv(0,(DP)BDY(f)); ndv_removecont(0,a); nd_psq[i] = a; - NEWRHist(r); - len = MAX(len,LEN(a)); - SG(r) = HTD(a); TD(r) = HTD(a); ndl_copy(HDL(a),DL(r)); + if ( trace ) { + a = nd_ps_trace[i] = ndv_dup(0,(NDV)BDY(f)); + ndv_removecont(0,a); + am = nd_ps[i] = ndv_dup(mod,a); + ndv_mod(mod,am); + ndv_removecont(mod,am); + } else { + a = nd_ps[i] = ndv_dup(mod,(NDV)BDY(f)); + if ( mod ) ndv_mod(mod,a); + ndv_removecont(mod,a); + } + NEWRHist(r); SG(r) = HTD(a); ndl_copy(HDL(a),DL(r)); + nd_bound[i] = ndv_compute_bound(a); nd_psh[i] = r; } } -void nd_gr(LIST f,LIST v,int m,struct order_spec *ord,LIST *rp) +void nd_gr(LIST f,LIST v,int m,int f4,struct order_spec *ord,LIST *rp) { - struct order_spec ord1; - VL fv,vv,vc; + VL tv,fv,vv,vc; NODE fd,fd0,r,r0,t,x,s,xx; - DP a,b,c; + int e,max,nvar; + ND b; get_vars((Obj)f,&fv); pltovl(v,&vv); - nd_nvar = length(vv); - if ( ord->id ) - error("nd_gr : unsupported order"); - switch ( ord->ord.simple ) { - case 0: - is_rlex = 1; - break; - case 1: - is_rlex = 0; - break; - default: - error("nd_gr : unsupported order"); - } - initd(ord); + nvar = length(vv); + nd_init_ord(ord); + for ( t = BDY(f), max = 0; t; t = NEXT(t) ) + for ( tv = vv; tv; tv = NEXT(tv) ) { + e = getdeg(tv->v,(P)BDY(t)); + max = MAX(e,max); + } + nd_setup_parameters(nvar,max); for ( fd0 = 0, t = BDY(f); t; t = NEXT(t) ) { - ptod(CO,vv,(P)BDY(t),&b); - NEXTNODE(fd0,fd); BDY(fd) = (pointer)b; + b = (pointer)ptondv(CO,vv,(P)BDY(t)); + if ( b ) { NEXTNODE(fd0,fd); BDY(fd) = (pointer)b; } } if ( fd0 ) NEXT(fd) = 0; - nd_setup(m,fd0); - x = nd_gb(m); - fprintf(asir_out,"found=%d,notfirst=%d,create=%d\n", - nd_found,nd_notfirst,nd_create); - x = nd_reduceall(m,x); + ndv_setup(m,0,fd0); + x = f4?nd_f4(m):nd_gb(m,0); + x = ndv_reducebase(x); + x = ndv_reduceall(m,x); for ( r0 = 0, t = x; t; t = NEXT(t) ) { NEXTNODE(r0,r); - if ( m ) { - a = ndvtodp(m,BDY(t)); - _dtop_mod(CO,vv,a,(P *)&BDY(r)); - } else { - a = ndvtodp(m,BDY(t)); - dtop(CO,vv,a,(P *)&BDY(r)); - } + BDY(r) = ndvtop(m,CO,vv,BDY(t)); } if ( r0 ) NEXT(r) = 0; MKLIST(*rp,r0); } -void nd_gr_trace(LIST f,LIST v,int m,int homo,struct order_spec *ord,LIST *rp) +void nd_gr_trace(LIST f,LIST v,int trace,int homo,struct order_spec *ord,LIST *rp) { struct order_spec ord1; - VL fv,vv,vc; - NODE fd,fd0,r,r0,t,x,s,xx,t0; - DP a,b,c,h; - NDV *w; - int len,i,j; + VL tv,fv,vv,vc; + NODE fd,fd0,in0,in,r,r0,t,s,cand; + int m,nocheck,nvar,mindex,e,max; + NDV c; + NMV a; + P p; + EPOS oepos; + int obpe,oadv,wmax,i,len,cbpe; get_vars((Obj)f,&fv); pltovl(v,&vv); - nd_nvar = length(vv); - if ( ord->id ) - error("nd_gr : unsupported order"); - initd(ord); - if ( homo ) { - homogenize_order(ord,nd_nvar,&ord1); - switch ( ord1.ord.simple ) { - case 0: is_rlex = 1; break; - case 1: is_rlex = 0; break; - default: error("nd_gr : unsupported order"); + nvar = length(vv); + nocheck = 0; + mindex = 0; + + /* setup modulus */ + if ( trace < 0 ) { + trace = -trace; + nocheck = 1; + } + m = trace > 1 ? trace : get_lprime(mindex); + for ( t = BDY(f), max = 0; t; t = NEXT(t) ) + for ( tv = vv; tv; tv = NEXT(tv) ) { + e = getdeg(tv->v,(P)BDY(t)); + max = MAX(e,max); } - for ( fd0 = 0, t = BDY(f); t; t = NEXT(t) ) { - ptod(CO,vv,(P)BDY(t),&c); - if ( c ) { - dp_homo(c,&h); NEXTNODE(fd0,fd); BDY(fd) = (pointer)h; - } + nd_init_ord(ord); + nd_setup_parameters(nvar,max); + obpe = nd_bpe; oadv = nmv_adv; oepos = nd_epos; + for ( in0 = 0, fd0 = 0, t = BDY(f); t; t = NEXT(t) ) { + c = ptondv(CO,vv,(P)BDY(t)); + if ( c ) { + NEXTNODE(in0,in); BDY(in) = (pointer)c; + NEXTNODE(fd0,fd); BDY(fd) = (pointer)ndv_dup(0,c); } - if ( fd0 ) NEXT(fd) = 0; - initd(&ord1); - nd_nvar++; - nd_setup_trace(m,fd0); - x = nd_gb_trace(m); - - /* dehomogenization */ - for ( t = x; t; t = NEXT(t) ) - ndv_dehomogenize((NDV)BDY(t)); - nd_nvar--; - nd_setup_parameters(); - initd(ord); - len = length(x); - w = (NDV *)ALLOCA(len*sizeof(NDV)); - for ( i = 0, t = x; i < len; i++, t = NEXT(t) ) w[i] = BDY(t); - for ( i = 0; i < len; i++ ) { - for ( j = 0; j < i; j++ ) { - if ( w[i] && w[j] ) - if ( ndl_reducible(HDL(w[i]),HDL(w[j])) ) w[i] = 0; - else if ( ndl_reducible(HDL(w[j]),HDL(w[i])) ) w[j] = 0; - } + } + if ( in0 ) NEXT(in) = 0; + if ( fd0 ) NEXT(fd) = 0; + if ( homo ) { + for ( t = in0, wmax = 0; t; t = NEXT(t) ) { + c = (NDV)BDY(t); len = LEN(c); + for ( a = BDY(c), i = 0; i < len; i++, NMV_ADV(a) ) + wmax = MAX(TD(DL(a)),wmax); } - for ( i = len-1, t0 = 0; i >= 0; i-- ) { - if ( w[i] ) { NEXTNODE(t0,t); BDY(t) = (pointer)w[i]; } + homogenize_order(ord,nvar,&ord1); + nd_init_ord(&ord1); + nd_setup_parameters(nvar+1,wmax); + for ( t = fd0; t; t = NEXT(t) ) + ndv_homogenize((NDV)BDY(t),obpe,oadv,oepos); + } + while ( 1 ) { + ndv_setup(m,1,fd0); + cand = nd_gb_trace(m); + if ( !cand ) { + /* failure */ + if ( trace > 1 ) { *rp = 0; return; } + else m = get_lprime(++mindex); + continue; } - NEXT(t) = 0; x = t0; - } else { - switch ( ord->ord.simple ) { - case 0: is_rlex = 1; break; - case 1: is_rlex = 0; break; - default: error("nd_gr : unsupported order"); + if ( homo ) { + /* dehomogenization */ + for ( t = cand; t; t = NEXT(t) ) ndv_dehomogenize((NDV)BDY(t),ord); + nd_init_ord(ord); + nd_setup_parameters(nvar,0); } - for ( fd0 = 0, t = BDY(f); t; t = NEXT(t) ) { - ptod(CO,vv,(P)BDY(t),&c); - if ( c ) { - NEXTNODE(fd0,fd); BDY(fd) = (pointer)c; + cand = ndv_reducebase(cand); + cand = ndv_reduceall(0,cand); + if ( nocheck ) + break; + cbpe = nd_bpe; + if ( ndv_check_candidate(in0,obpe,oadv,oepos,cand) ) + /* success */ + break; + else if ( trace > 1 ) { + /* failure */ + *rp = 0; return; + } else { + /* try the next modulus */ + m = get_lprime(++mindex); + /* reset the parameters */ + if ( homo ) { + nd_init_ord(&ord1); + nd_setup_parameters(nvar+1,wmax); + } else { + nd_init_ord(ord); + nd_setup_parameters(nvar,max); } } - if ( fd0 ) NEXT(fd) = 0; - /* setup over GF(m) */ - nd_setup_trace(m,fd0); - x = nd_gb_trace(m); } - fprintf(asir_out,"found=%d,notfirst=%d,create=%d\n", - nd_found,nd_notfirst,nd_create); - x = nd_reduceall(0,x); - for ( r0 = 0; x; x = NEXT(x) ) { - NEXTNODE(r0,r); - a = ndvtodp(0,(NDV)BDY(x)); - dtop(CO,vv,a,(P *)&BDY(r)); - } - if ( r0 ) NEXT(r) = 0; - MKLIST(*rp,r0); + /* dp->p */ + nd_bpe = cbpe; + nd_setup_parameters(0,0); + for ( r = cand; r; r = NEXT(r) ) BDY(r) = (pointer)ndvtop(0,CO,vv,BDY(r)); + MKLIST(*rp,cand); } -void dltondl(int n,DL dl,unsigned int *r) +void dltondl(int n,DL dl,UINT *r) { - unsigned int *d; - int i; + UINT *d; + int i,j,l,s,ord_l; + struct order_pair *op; d = dl->d; - bzero(r,nd_wpd*sizeof(unsigned int)); - if ( is_rlex ) - for ( i = 0; i < n; i++ ) - r[(n-1-i)/nd_epw] |= (d[i]<<((nd_epw-((n-1-i)%nd_epw)-1)*nd_bpe)); - else - for ( i = 0; i < n; i++ ) - r[i/nd_epw] |= d[i]<<((nd_epw-(i%nd_epw)-1)*nd_bpe); + for ( i = 0; i < nd_wpd; i++ ) r[i] = 0; + if ( nd_blockmask ) { + l = nd_blockmask->n; + op = nd_blockmask->order_pair; + for ( j = 0, s = 0; j < l; j++ ) { + ord_l = op[j].length; + for ( i = 0; i < ord_l; i++, s++ ) PUT_EXP(r,s,d[s]); + } + TD(r) = ndl_weight(r); + ndl_weight_mask(r); + } else { + for ( i = 0; i < n; i++ ) PUT_EXP(r,i,d[i]); + TD(r) = ndl_weight(r); + } } -DL ndltodl(int n,int td,unsigned int *ndl) +DL ndltodl(int n,UINT *ndl) { DL dl; int *d; - int i; + int i,j,l,s,ord_l; + struct order_pair *op; NEWDL(dl,n); - TD(dl) = td; + dl->td = TD(ndl); d = dl->d; - if ( is_rlex ) - for ( i = 0; i < n; i++ ) - d[i] = (ndl[(n-1-i)/nd_epw]>>((nd_epw-((n-1-i)%nd_epw)-1)*nd_bpe)) - &((1<>((nd_epw-(i%nd_epw)-1)*nd_bpe)) - &((1<n; + op = nd_blockmask->order_pair; + for ( j = 0, s = 0; j < l; j++ ) { + ord_l = op[j].length; + for ( i = 0; i < ord_l; i++, s++ ) d[s] = GET_EXP(ndl,s); + } + } else { + for ( i = 0; i < n; i++ ) d[i] = GET_EXP(ndl,i); + } return dl; } -ND dptond(int mod,DP p) +void ndl_print(UINT *dl) { - ND d; - NM m0,m; - MP t; int n; + int i,j,l,ord_l,s,s0; + struct order_pair *op; - if ( !p ) - return 0; - n = NV(p); - m0 = 0; - for ( t = BDY(p); t; t = NEXT(t) ) { - NEXTNM(m0,m); - if ( mod ) - CM(m) = ITOS(C(t)); - else - CQ(m) = (Q)C(t); - TD(m) = TD(DL(t)); - dltondl(n,DL(t),DL(m)); - } - NEXT(m) = 0; - MKND(n,m0,d); - NV(d) = n; - SG(d) = SG(p); - return d; -} - -DP ndtodp(int mod,ND p) -{ - DP d; - MP m0,m; - NM t; - int n; - - if ( !p ) - return 0; - n = NV(p); - m0 = 0; - for ( t = BDY(p); t; t = NEXT(t) ) { - NEXTMP(m0,m); - if ( mod ) - C(m) = STOI(CM(t)); - else - C(m) = (P)CQ(t); - DL(m) = ndltodl(n,TD(t),DL(t)); - } - NEXT(m) = 0; - MKDP(n,m0,d); - SG(d) = SG(p); - return d; -} - -void ndl_print(unsigned int *dl) -{ - int n; - int i; - n = nd_nvar; printf("<<"); - if ( is_rlex ) - for ( i = 0; i < n; i++ ) - printf(i==n-1?"%d":"%d,", - (dl[(n-1-i)/nd_epw]>>((nd_epw-((n-1-i)%nd_epw)-1)*nd_bpe)) - &((1<>((nd_epw-(i%nd_epw)-1)*nd_bpe)) - &((1<n; + op = nd_blockmask->order_pair; + for ( j = 0, s = s0 = 0; j < l; j++ ) { + ord_l = op[j].length; + for ( i = 0; i < ord_l; i++, s++ ) + printf(s==n-1?"%d":"%d,",GET_EXP(dl,s)); + } + } else { + for ( i = 0; i < n; i++ ) printf(i==n-1?"%d":"%d,",GET_EXP(dl,i)); + } printf(">>"); } @@ -2155,9 +2565,7 @@ void ndp_print(ND_pairs d) { ND_pairs t; - for ( t = d; t; t = NEXT(t) ) { - printf("%d,%d ",t->i1,t->i2); - } + for ( t = d; t; t = NEXT(t) ) printf("%d,%d ",t->i1,t->i2); printf("\n"); } @@ -2170,15 +2578,13 @@ void nd_removecont(int mod,ND p) struct oVECT v; N q,r; - if ( mod ) - nd_mul_c(mod,p,invm(HCM(p),mod)); + if ( mod ) nd_mul_c(mod,p,invm(HCM(p),mod)); else { for ( m = BDY(p), n = 0; m; m = NEXT(m), n++ ); w = (Q *)ALLOCA(n*sizeof(Q)); v.len = n; v.body = (pointer *)w; - for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) - w[i] = CQ(m); + for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) w[i] = CQ(m); removecont_array(w,n); for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) CQ(m) = w[i]; } @@ -2204,10 +2610,8 @@ void nd_removecont2(ND p1,ND p2) w = (Q *)ALLOCA(n*sizeof(Q)); v.len = n; v.body = (pointer *)w; - for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) - w[i] = CQ(m); - for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) - w[i] = CQ(m); + for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) w[i] = CQ(m); + for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) w[i] = CQ(m); removecont_array(w,n); for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) CQ(m) = w[i]; for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) CQ(m) = w[i]; @@ -2225,8 +2629,7 @@ void ndv_removecont(int mod,NDV p) else { len = p->len; w = (Q *)ALLOCA(len*sizeof(Q)); - for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) - w[i] = CQ(m); + for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) w[i] = CQ(m); sortbynm(w,len); qltozl(w,len,&dvr); for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) { @@ -2235,25 +2638,47 @@ void ndv_removecont(int mod,NDV p) } } -void ndv_dehomogenize(NDV p) +void ndv_homogenize(NDV p,int obpe,int oadv,EPOS oepos) { - int i,len,newnvar,newwpd,newadv; + int len,i,max; + NMV m,mr0,mr,t; + + len = p->len; + for ( m = BDY(p), i = 0, max = 0; i < len; NMV_OADV(m), i++ ) + max = MAX(max,TD(DL(m))); + mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p); + m = (NMV)((char *)mr0+(len-1)*oadv); + mr = (NMV)((char *)mr0+(len-1)*nmv_adv); + t = (NMV)ALLOCA(nmv_adv); + for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) { + ndl_homogenize(DL(m),DL(t),obpe,oepos,max); + CQ(mr) = CQ(m); + ndl_copy(DL(t),DL(mr)); + } + NV(p)++; + BDY(p) = mr0; +} + +void ndv_dehomogenize(NDV p,struct order_spec *ord) +{ + int i,j,adj,len,newnvar,newwpd,newadv,newexporigin; Q *w; Q dvr,t; NMV m,r; - unsigned int *d; -#define NEWADV(m) (m = (NMV)(((char *)m)+newadv)) len = p->len; newnvar = nd_nvar-1; - newwpd = newnvar/nd_epw+(newnvar%nd_epw?1:0); + newexporigin = nd_get_exporigin(ord); + newwpd = newnvar/nd_epw+(newnvar%nd_epw?1:0)+newexporigin; for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) - TD(m) -= ndl_dehomogenize(DL(m)); + ndl_dehomogenize(DL(m)); if ( newwpd != nd_wpd ) { - d = (unsigned int *)ALLOCA(newwpd*sizeof(unsigned int)); - newadv = sizeof(struct oNMV)+(newwpd-1)*sizeof(unsigned int); - for ( m = r = BDY(p), i = 0; i < len; NMV_ADV(m), NEWADV(r), i++ ) { - CQ(r) = CQ(m); TD(r) = TD(m); ndl_copy(DL(m),d); ndl_copy(d,DL(r)); + newadv = sizeof(struct oNMV)+(newwpd-1)*sizeof(UINT); + for ( m = r = BDY(p), i = 0; i < len; NMV_ADV(m), NDV_NADV(r), i++ ) { + CQ(r) = CQ(m); + for ( j = 0; j < newexporigin; j++ ) DL(r)[j] = DL(m)[j]; + adj = nd_exporigin-newexporigin; + for ( ; j < newwpd; j++ ) DL(r)[j] = DL(m)[j+adj]; } } NV(p)--; @@ -2276,9 +2701,7 @@ void removecont_array(Q *c,int n) NTOQ(qn,SGN(c[i])*SGN(d0),q[i]); NTOQ(rn,SGN(c[i]),r[i]); } - for ( i = 0; i < n; i++ ) - if ( r[i] ) - break; + for ( i = 0; i < n; i++ ) if ( r[i] ) break; if ( i < n ) { v.id = O_VECT; v.len = n; v.body = (pointer *)r; igcdv(&v,&d1); @@ -2293,8 +2716,7 @@ void removecont_array(Q *c,int n) q[i] = u; } } - for ( i = 0; i < n; i++ ) - c[i] = q[i]; + for ( i = 0; i < n; i++ ) c[i] = q[i]; } void nd_mul_c(int mod,ND p,int mul) @@ -2302,8 +2724,7 @@ void nd_mul_c(int mod,ND p,int mul) NM m; int c,c1; - if ( !p ) - return; + if ( !p ) return; for ( m = BDY(p); m; m = NEXT(m) ) { c1 = CM(m); DMAR(c1,mul,0,mod,c); @@ -2316,19 +2737,28 @@ void nd_mul_c_q(ND p,Q mul) NM m; Q c; - if ( !p ) - return; + if ( !p ) return; for ( m = BDY(p); m; m = NEXT(m) ) { mulq(CQ(m),mul,&c); CQ(m) = c; } } +void nd_mul_c_p(VL vl,ND p,P mul) +{ + NM m; + P c; + + if ( !p ) return; + for ( m = BDY(p); m; m = NEXT(m) ) { + mulp(vl,CP(m),mul,&c); CP(m) = c; + } +} + void nd_free(ND p) { NM t,s; - if ( !p ) - return; + if ( !p ) return; t = BDY(p); while ( t ) { s = NEXT(t); @@ -2343,73 +2773,88 @@ void ndv_free(NDV p) GC_free(BDY(p)); } -void nd_append_red(unsigned int *d,int td,int i) +void nd_append_red(UINT *d,int i) { RHist m,m0; int h; NEWRHist(m); - h = ndl_hash_value(td,d); + h = ndl_hash_value(d); m->index = i; - TD(m) = td; ndl_copy(d,DL(m)); NEXT(m) = nd_red[h]; nd_red[h] = m; } -unsigned int *dp_compute_bound(DP p) +UINT *ndv_compute_bound(NDV p) { - unsigned int *d,*d1,*d2,*t; - MP m; - int i,l; + UINT *d1,*d2,*t; + UINT u; + int i,j,k,l,len,ind; + NMV m; if ( !p ) return 0; - d1 = (unsigned int *)ALLOCA(nd_nvar*sizeof(unsigned int)); - d2 = (unsigned int *)ALLOCA(nd_nvar*sizeof(unsigned int)); - m = BDY(p); - bcopy(DL(m)->d,d1,nd_nvar*sizeof(unsigned int)); - for ( m = NEXT(BDY(p)); m; m = NEXT(m) ) { - d = DL(m)->d; - for ( i = 0; i < nd_nvar; i++ ) - d2[i] = d[i] > d1[i] ? d[i] : d1[i]; + d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT)); + d2 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT)); + len = LEN(p); + m = BDY(p); ndl_copy(DL(m),d1); NMV_ADV(m); + for ( i = 1; i < len; i++, NMV_ADV(m) ) { + ndl_lcm(DL(m),d1,d2); t = d1; d1 = d2; d2 = t; } - l = (nd_nvar+31); - t = (unsigned int *)MALLOC_ATOMIC(l*sizeof(unsigned int)); - bzero(t,l*sizeof(unsigned int)); - bcopy(d1,t,nd_nvar*sizeof(unsigned int)); + l = nd_nvar+31; + t = (UINT *)MALLOC_ATOMIC(l*sizeof(UINT)); + for ( i = nd_exporigin, ind = 0; i < nd_wpd; i++ ) { + u = d1[i]; + k = (nd_epw-1)*nd_bpe; + for ( j = 0; j < nd_epw; j++, k -= nd_bpe, ind++ ) + t[ind] = (u>>k)&nd_mask0; + } + for ( ; ind < l; ind++ ) t[ind] = 0; return t; } -unsigned int *nd_compute_bound(ND p) +int nd_get_exporigin(struct order_spec *ord) { - unsigned int *d1,*d2,*t; - int i,l; - NM m; - - if ( !p ) - return 0; - d1 = (unsigned int *)ALLOCA(nd_wpd*sizeof(unsigned int)); - d2 = (unsigned int *)ALLOCA(nd_wpd*sizeof(unsigned int)); - bcopy(HDL(p),d1,nd_wpd*sizeof(unsigned int)); - for ( m = NEXT(BDY(p)); m; m = NEXT(m) ) { - ndl_lcm(DL(m),d1,d2); - t = d1; d1 = d2; d2 = t; + switch ( ord->id ) { + case 0: + return 1; + case 1: + /* block order */ + /* d[0]:weight d[1]:w0,...,d[nd_exporigin-1]:w(n-1) */ + return ord->ord.block.length+1; + case 2: + error("nd_get_exporigin : matrix order is not supported yet."); } - l = nd_nvar+31; - t = (unsigned int *)MALLOC_ATOMIC(l*sizeof(unsigned int)); - bzero(t,l*sizeof(unsigned int)); - for ( i = 0; i < nd_nvar; i++ ) - t[i] = (d1[i/nd_epw]>>((nd_epw-(i%nd_epw)-1)*nd_bpe))&nd_mask0; - return t; } -void nd_setup_parameters() { - int i; +void nd_setup_parameters(int nvar,int max) { + int i,j,n,elen,ord_o,ord_l,l,s; + struct order_pair *op; - nd_epw = (sizeof(unsigned int)*8)/nd_bpe; - nd_wpd = nd_nvar/nd_epw+(nd_nvar%nd_epw?1:0); + /* if max == 0, don't touch nd_bpe */ + if ( max > 0 ) { + if ( max < 2 ) nd_bpe = 1; + if ( max < 4 ) nd_bpe = 2; + else if ( max < 8 ) nd_bpe = 3; + else if ( max < 16 ) nd_bpe = 4; + else if ( max < 32 ) nd_bpe = 5; + else if ( max < 64 ) nd_bpe = 6; + else if ( max < 256 ) nd_bpe = 8; + else if ( max < 1024 ) nd_bpe = 10; + else if ( max < 65536 ) nd_bpe = 16; + else nd_bpe = 32; + } + /* nvar == 0, don't touch nd_nvar */ + if ( nvar > 0 ) nd_nvar = nvar; + + nd_epw = (sizeof(UINT)*8)/nd_bpe; + elen = nd_nvar/nd_epw+(nd_nvar%nd_epw?1:0); + + nd_exporigin = nd_get_exporigin(nd_ord); + nd_wpd = nd_exporigin+elen; + if ( nd_bpe < 32 ) { nd_mask0 = (1< realloc nd_ps and pd_psq */ - ND_pairs nd_reconstruct(int mod,int trace,ND_pairs d) { - int i,obpe,oadv; + int i,obpe,oadv,h; NM prev_nm_free_list; RHist mr0,mr; RHist r; + RHist *old_red; ND_pairs s0,s,t,prev_ndp_free_list; + EPOS oepos; obpe = nd_bpe; oadv = nmv_adv; - if ( obpe < 4 ) - nd_bpe = 4; - else if ( obpe < 6 ) - nd_bpe = 6; - else if ( obpe < 8 ) - nd_bpe = 8; - else if ( obpe < 16 ) - nd_bpe = 16; - else if ( obpe < 32 ) - nd_bpe = 32; - else - error("nd_reconstruct : exponent too large"); + oepos = nd_epos; + if ( obpe < 2 ) nd_bpe = 2; + else if ( obpe < 3 ) nd_bpe = 3; + else if ( obpe < 4 ) nd_bpe = 4; + else if ( obpe < 5 ) nd_bpe = 5; + else if ( obpe < 6 ) nd_bpe = 6; + else if ( obpe < 8 ) nd_bpe = 8; + else if ( obpe < 10 ) nd_bpe = 10; + else if ( obpe < 16 ) nd_bpe = 16; + else if ( obpe < 32 ) nd_bpe = 32; + else error("nd_reconstruct : exponent too large"); - nd_setup_parameters(); + nd_setup_parameters(0,0); prev_nm_free_list = _nm_free_list; prev_ndp_free_list = _ndp_free_list; _nm_free_list = 0; _ndp_free_list = 0; - if ( mod != 0 ) + for ( i = nd_psn-1; i >= 0; i-- ) ndv_realloc(nd_ps[i],obpe,oadv,oepos); + if ( trace ) for ( i = nd_psn-1; i >= 0; i-- ) - ndv_realloc(nd_ps[i],obpe,oadv); - if ( !mod || trace ) - for ( i = nd_psn-1; i >= 0; i-- ) - ndv_realloc(nd_psq[i],obpe,oadv); + ndv_realloc(nd_ps_trace[i],obpe,oadv,oepos); s0 = 0; for ( t = d; t; t = NEXT(t) ) { NEXTND_pairs(s0,s); s->i1 = t->i1; s->i2 = t->i2; - TD(s) = TD(t); SG(s) = SG(t); - ndl_dup(obpe,t->lcm,s->lcm); + ndl_reconstruct(LCM(t),LCM(s),obpe,oepos); } + + old_red = (RHist *)ALLOCA(REDTAB_LEN*sizeof(RHist)); for ( i = 0; i < REDTAB_LEN; i++ ) { - for ( mr0 = 0, r = nd_red[i]; r; r = NEXT(r) ) { - NEXTRHist(mr0,mr); + old_red[i] = nd_red[i]; + nd_red[i] = 0; + } + for ( i = 0; i < REDTAB_LEN; i++ ) + for ( r = old_red[i]; r; r = NEXT(r) ) { + NEWRHist(mr); mr->index = r->index; SG(mr) = SG(r); - TD(mr) = TD(r); - ndl_dup(obpe,DL(r),DL(mr)); + ndl_reconstruct(DL(r),DL(mr),obpe,oepos); + h = ndl_hash_value(DL(mr)); + NEXT(mr) = nd_red[h]; + nd_red[h] = mr; } - if ( mr0 ) NEXT(mr) = 0; - nd_red[i] = mr0; - } + for ( i = 0; i < REDTAB_LEN; i++ ) old_red[i] = 0; + old_red = 0; for ( i = 0; i < nd_psn; i++ ) { NEWRHist(r); SG(r) = SG(nd_psh[i]); - TD(r) = TD(nd_psh[i]); ndl_dup(obpe,DL(nd_psh[i]),DL(r)); + ndl_reconstruct(DL(nd_psh[i]),DL(r),obpe,oepos); nd_psh[i] = r; } if ( s0 ) NEXT(s) = 0; @@ -2495,135 +2945,118 @@ ND_pairs nd_reconstruct(int mod,int trace,ND_pairs d) void nd_reconstruct_direct(int mod,NDV *ps,int len) { - int i,obpe,oadv; + int i,obpe,oadv,h; + UINT **bound; NM prev_nm_free_list; RHist mr0,mr; RHist r; + RHist *old_red; ND_pairs s0,s,t,prev_ndp_free_list; + EPOS oepos; obpe = nd_bpe; oadv = nmv_adv; - if ( obpe < 4 ) - nd_bpe = 4; - else if ( obpe < 6 ) - nd_bpe = 6; - else if ( obpe < 8 ) - nd_bpe = 8; - else if ( obpe < 16 ) - nd_bpe = 16; - else if ( obpe < 32 ) - nd_bpe = 32; - else - error("nd_reconstruct_direct : exponent too large"); + oepos = nd_epos; + if ( obpe < 2 ) nd_bpe = 2; + else if ( obpe < 3 ) nd_bpe = 3; + else if ( obpe < 4 ) nd_bpe = 4; + else if ( obpe < 5 ) nd_bpe = 5; + else if ( obpe < 6 ) nd_bpe = 6; + else if ( obpe < 8 ) nd_bpe = 8; + else if ( obpe < 10 ) nd_bpe = 10; + else if ( obpe < 16 ) nd_bpe = 16; + else if ( obpe < 32 ) nd_bpe = 32; + else error("nd_reconstruct_direct : exponent too large"); - nd_setup_parameters(); + nd_setup_parameters(0,0); prev_nm_free_list = _nm_free_list; prev_ndp_free_list = _ndp_free_list; - _nm_free_list = 0; - _ndp_free_list = 0; - if ( mod != 0 ) - for ( i = len-1; i >= 0; i-- ) - ndv_realloc(ps[i],obpe,oadv); + _nm_free_list = 0; _ndp_free_list = 0; + for ( i = len-1; i >= 0; i-- ) ndv_realloc(ps[i],obpe,oadv,oepos); prev_nm_free_list = 0; prev_ndp_free_list = 0; GC_gcollect(); } -void ndl_dup(int obpe,unsigned int *d,unsigned int *r) +void ndl_reconstruct(UINT *d,UINT *r,int obpe,EPOS oepos) { - int n,i,ei,oepw,cepw,cbpe; + int n,i,ei,oepw,omask0,j,s,ord_l,l; + struct order_pair *op; n = nd_nvar; - oepw = (sizeof(unsigned int)*8)/obpe; - cepw = nd_epw; - cbpe = nd_bpe; - for ( i = 0; i < nd_wpd; i++ ) - r[i] = 0; - if ( is_rlex ) - for ( i = 0; i < n; i++ ) { - ei = (d[(n-1-i)/oepw]>>((oepw-((n-1-i)%oepw)-1)*obpe)) - &((1<n; + op = nd_blockmask->order_pair; + for ( i = 1; i < nd_exporigin; i++ ) + r[i] = d[i]; + for ( j = 0, s = 0; j < l; j++ ) { + ord_l = op[j].length; + for ( i = 0; i < ord_l; i++, s++ ) { + ei = GET_EXP_OLD(d,s); + PUT_EXP(r,s,ei); + } } - else + } else { for ( i = 0; i < n; i++ ) { - ei = (d[i/oepw]>>((oepw-(i%oepw)-1)*obpe)) - &((1<i1]; p2 = nd_ps[p->i2]; + if ( trace ) { + p1 = nd_ps_trace[p->i1]; p2 = nd_ps_trace[p->i2]; } else { - p1 = nd_psq[p->i1]; p2 = nd_psq[p->i2]; + p1 = nd_ps[p->i1]; p2 = nd_ps[p->i2]; } - lcm = p->lcm; - td = TD(p); + lcm = LCM(p); NEWNM(m); CQ(m) = HCQ(p2); - TD(m) = td-HTD(p1); ndl_sub(lcm,HDL(p1),DL(m)); + ndl_sub(lcm,HDL(p1),DL(m)); if ( ndl_check_bound2(p->i1,DL(m)) ) return 0; - t1 = ndv_mul_nm(mod,p1,m); - if ( mod ) - CM(m) = mod-HCM(p1); - else - chsgnq(HCQ(p1),&CQ(m)); - TD(m) = td-HTD(p2); ndl_sub(lcm,HDL(p2),DL(m)); + t1 = ndv_mul_nm(mod,m,p1); + if ( mod ) CM(m) = mod-HCM(p1); + else chsgnq(HCQ(p1),&CQ(m)); + ndl_sub(lcm,HDL(p2),DL(m)); if ( ndl_check_bound2(p->i2,DL(m)) ) { nd_free(t1); return 0; } - t2 = ndv_mul_nm(mod,p2,m); + t2 = ndv_mul_nm(mod,m,p2); *rp = nd_add(mod,t1,t2); FREENM(m); return 1; @@ -2634,13 +3067,10 @@ void ndv_mul_c(int mod,NDV p,int mul) NMV m; int c,c1,len,i; - if ( !p ) - return; + if ( !p ) return; len = LEN(p); for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) { - c1 = CM(m); - DMAR(c1,mul,0,mod,c); - CM(m) = c; + c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c; } } @@ -2650,30 +3080,194 @@ void ndv_mul_c_q(NDV p,Q mul) Q c; int len,i; - if ( !p ) - return; + if ( !p ) return; len = LEN(p); for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) { mulq(CQ(m),mul,&c); CQ(m) = c; } } -ND ndv_mul_nm(int mod,NDV p,NM m0) +ND weyl_ndv_mul_nm(int mod,NM m0,NDV p) { + int n2,i,j,l,n,tlen; + UINT *d0; + NM *tab,*psum; + ND s,r; + NM t; + NMV m1; + + if ( !p ) return 0; + n = NV(p); n2 = n>>1; + d0 = DL(m0); + l = LEN(p); + for ( i = 0, tlen = 1; i < n2; i++ ) tlen *= (GET_EXP(d0,n2+i)+1); + tab = (NM *)ALLOCA(tlen*sizeof(NM)); + psum = (NM *)ALLOCA(tlen*sizeof(NM)); + for ( i = 0; i < tlen; i++ ) psum[i] = 0; + m1 = (NMV)(((char *)BDY(p))+nmv_adv*(l-1)); + for ( i = l-1; i >= 0; i--, NMV_PREV(m1) ) { + /* m0(NM) * m1(NMV) => tab(NM) */ + weyl_mul_nm_nmv(n,mod,m0,m1,tab,tlen); + for ( j = 0; j < tlen; j++ ) { + if ( tab[j] ) { + NEXT(tab[j]) = psum[j]; psum[j] = tab[j]; + } + } + } + for ( i = tlen-1, r = 0; i >= 0; i-- ) + if ( psum[i] ) { + for ( j = 0, t = psum[i]; t; t = NEXT(t), j++ ); + MKND(n,psum[i],j,s); + r = nd_add(mod,r,s); + } + if ( r ) SG(r) = SG(p)+TD(d0); + return r; +} + +/* product of monomials */ +/* XXX block order is not handled correctly */ + +void weyl_mul_nm_nmv(int n,int mod,NM m0,NMV m1,NM *tab,int tlen) { + int i,n2,j,s,curlen,homo,h,a,b,k,l,u,min; + UINT *d0,*d1,*d,*dt,*ctab; + Q *ctab_q; + Q q,q1; + UINT c0,c1,c; + NM *p; + NM m,t; + + for ( i = 0; i < tlen; i++ ) tab[i] = 0; + if ( !m0 || !m1 ) return; + d0 = DL(m0); d1 = DL(m1); n2 = n>>1; + NEWNM(m); d = DL(m); + if ( mod ) { + c0 = CM(m0); c1 = CM(m1); DMAR(c0,c1,0,mod,c); CM(m) = c; + } else + mulq(CQ(m0),CQ(m1),&CQ(m)); + for ( i = 0; i < nd_wpd; i++ ) d[i] = 0; + homo = n&1 ? 1 : 0; + if ( homo ) { + /* offset of h-degree */ + h = GET_EXP(d0,n-1)+GET_EXP(d1,n-1); + PUT_EXP(DL(m),n-1,h); + TD(DL(m)) = h; + if ( nd_blockmask ) ndl_weight_mask(DL(m)); + } + tab[0] = m; + NEWNM(m); d = DL(m); + for ( i = 0, curlen = 1; i < n2; i++ ) { + a = GET_EXP(d0,i); b = GET_EXP(d1,n2+i); + k = GET_EXP(d0,n2+i); l = GET_EXP(d1,i); + /* 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; j < curlen; j++ ) + if ( t = tab[j] ) { + dt = DL(t); + PUT_EXP(dt,i,a); PUT_EXP(dt,n2+i,b); TD(dt) += s; + if ( nd_blockmask ) ndl_weight_mask(dt); + } + curlen *= k+1; + continue; + } + min = MIN(k,l); + if ( mod ) { + ctab = (UINT *)ALLOCA((min+1)*sizeof(UINT)); + mkwcm(k,l,mod,ctab); + } else { + ctab_q = (Q *)ALLOCA((min+1)*sizeof(Q)); + mkwc(k,l,ctab_q); + } + for ( j = min; j >= 0; j-- ) { + for ( u = 0; u < nd_wpd; u++ ) d[u] = 0; + PUT_EXP(d,i,a-j); PUT_EXP(d,n2+i,b-j); + h = MUL_WEIGHT(a-j,i)+MUL_WEIGHT(b-j,n2+i); + if ( homo ) { + TD(d) = s; + PUT_EXP(d,n-1,s-h); + } else TD(d) = h; + if ( nd_blockmask ) ndl_weight_mask(d); + if ( mod ) c = ctab[j]; + else q = ctab_q[j]; + p = tab+curlen*j; + if ( j == 0 ) { + for ( u = 0; u < curlen; u++, p++ ) { + if ( tab[u] ) { + ndl_addto(DL(tab[u]),d); + if ( mod ) { + c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(tab[u]) = c1; + } else { + mulq(CQ(tab[u]),q,&q1); CQ(tab[u]) = q1; + } + } + } + } else { + for ( u = 0; u < curlen; u++, p++ ) { + if ( tab[u] ) { + NEWNM(t); + ndl_add(DL(tab[u]),d,DL(t)); + if ( mod ) { + c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(t) = c1; + } else + mulq(CQ(tab[u]),q,&CQ(t)); + *p = t; + } + } + } + } + curlen *= k+1; + } + FREENM(m); +} + +ND ndv_mul_nm_symbolic(NM m0,NDV p) +{ NM mr,mr0; NMV m; - unsigned int *d,*dt,*dm; + UINT *d,*dt,*dm; int c,n,td,i,c1,c2,len; Q q; ND r; - if ( !p ) - return 0; + if ( !p ) return 0; else { n = NV(p); m = BDY(p); - d = DL(m0); td = TD(m0); + d = DL(m0); len = LEN(p); mr0 = 0; + td = TD(d); + c = CM(m0); + for ( i = 0; i < len; i++, NMV_ADV(m) ) { + NEXTNM(mr0,mr); + CM(mr) = 1; + ndl_add(DL(m),d,DL(mr)); + } + NEXT(mr) = 0; + MKND(NV(p),mr0,len,r); + SG(r) = SG(p) + TD(d); + return r; + } +} + +ND ndv_mul_nm(int mod,NM m0,NDV p) +{ + NM mr,mr0; + NMV m; + UINT *d,*dt,*dm; + int c,n,td,i,c1,c2,len; + Q q; + ND r; + + if ( !p ) return 0; + else if ( do_weyl ) + return weyl_ndv_mul_nm(mod,m0,p); + else { + n = NV(p); m = BDY(p); + d = DL(m0); + len = LEN(p); + mr0 = 0; + td = TD(d); if ( mod ) { c = CM(m0); for ( i = 0; i < len; i++, NMV_ADV(m) ) { @@ -2681,7 +3275,6 @@ ND ndv_mul_nm(int mod,NDV p,NM m0) c1 = CM(m); DMAR(c1,c,0,mod,c2); CM(mr) = c2; - TD(mr) = TD(m)+td; ndl_add(DL(m),d,DL(mr)); } } else { @@ -2689,63 +3282,68 @@ ND ndv_mul_nm(int mod,NDV p,NM m0) for ( i = 0; i < len; i++, NMV_ADV(m) ) { NEXTNM(mr0,mr); mulq(CQ(m),q,&CQ(mr)); - TD(mr) = TD(m)+td; ndl_add(DL(m),d,DL(mr)); } } NEXT(mr) = 0; - MKND(NV(p),mr0,r); - SG(r) = SG(p) + td; + MKND(NV(p),mr0,len,r); + SG(r) = SG(p) + TD(d); return r; } } -void ndv_realloc(NDV p,int obpe,int oadv) +void ndv_realloc(NDV p,int obpe,int oadv,EPOS oepos) { NMV m,mr,mr0,t; int len,i,k; -#define NMV_OPREV(m) (m = (NMV)(((char *)m)-oadv)) -#define NMV_PREV(m) (m = (NMV)(((char *)m)-nmv_adv)) + if ( !p ) return; + m = BDY(p); len = LEN(p); + mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p); + m = (NMV)((char *)mr0+(len-1)*oadv); + mr = (NMV)((char *)mr0+(len-1)*nmv_adv); + t = (NMV)ALLOCA(nmv_adv); + for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) { + CQ(t) = CQ(m); + for ( k = 0; k < nd_wpd; k++ ) DL(t)[k] = 0; + ndl_reconstruct(DL(m),DL(t),obpe,oepos); + CQ(mr) = CQ(t); + ndl_copy(DL(t),DL(mr)); + } + BDY(p) = mr0; +} - if ( p ) { - m = BDY(p); len = LEN(p); - if ( nmv_adv > oadv ) - mr0 = (NMV)REALLOC(BDY(p),len*nmv_adv); - else - mr0 = BDY(p); - m = (NMV)((char *)mr0+(len-1)*oadv); - mr = (NMV)((char *)mr0+(len-1)*nmv_adv); - t = (NMV)ALLOCA(nmv_adv); - for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) { - CQ(t) = CQ(m); - TD(t) = TD(m); - for ( k = 0; k < nd_wpd; k++ ) DL(t)[k] = 0; - ndl_dup(obpe,DL(m),DL(t)); - CQ(mr) = CQ(t); - TD(mr) = TD(t); - ndl_copy(DL(t),DL(mr)); - } - BDY(p) = mr0; +NDV ndv_dup_realloc(NDV p,int obpe,int oadv,EPOS oepos) +{ + NMV m,mr,mr0; + int len,i; + NDV r; + + if ( !p ) return 0; + m = BDY(p); len = LEN(p); + mr0 = mr = (NMV)MALLOC(len*nmv_adv); + for ( i = 0; i < len; i++, NMV_OADV(m), NMV_ADV(mr) ) { + ndl_zero(DL(mr)); + ndl_reconstruct(DL(m),DL(mr),obpe,oepos); + CQ(mr) = CQ(m); } + MKNDV(NV(p),mr0,len,r); + SG(r) = SG(p); + return r; } -NDV ndtondv(int mod,ND p) +/* duplicate p */ + +NDV ndv_dup(int mod,NDV p) { NDV d; - NMV m,m0; - NM t; + NMV t,m,m0; int i,len; - if ( !p ) - return 0; - len = nd_length(p); - if ( mod ) - m0 = m = (NMV)MALLOC_ATOMIC(len*nmv_adv); - else - m0 = m = (NMV)MALLOC(len*nmv_adv); - for ( t = BDY(p), i = 0; t; t = NEXT(t), i++, NMV_ADV(m) ) { - TD(m) = TD(t); + if ( !p ) return 0; + len = LEN(p); + m0 = m = (NMV)(mod?MALLOC_ATOMIC(len*nmv_adv):MALLOC(len*nmv_adv)); + for ( t = BDY(p), i = 0; i < len; i++, NMV_ADV(t), NMV_ADV(m) ) { ndl_copy(DL(t),DL(m)); CQ(m) = CQ(t); } @@ -2754,81 +3352,177 @@ NDV ndtondv(int mod,ND p) return d; } -ND ndvtond(int mod,NDV p) +ND nd_dup(ND p) { ND d; - NM m,m0; - NMV t; - int i,len; + NM t,m,m0; - if ( !p ) - return 0; - m0 = 0; - len = p->len; - for ( t = BDY(p), i = 0; i < len; NMV_ADV(t), i++ ) { + if ( !p ) return 0; + for ( m0 = 0, t = BDY(p); t; t = NEXT(t) ) { NEXTNM(m0,m); - TD(m) = TD(t); ndl_copy(DL(t),DL(m)); CQ(m) = CQ(t); } - NEXT(m) = 0; - MKND(NV(p),m0,d); + if ( m0 ) NEXT(m) = 0; + MKND(NV(p),m0,LEN(p),d); SG(d) = SG(p); return d; } -NDV dptondv(int mod,DP p) +/* XXX if p->len == 0 then it represents 0 */ + +void ndv_mod(int mod,NDV p) { - NDV d; - NMV m0,m; - MP t; - DP q; - int l,i,n; + NMV t,d; + int r; + int i,len,dlen; + if ( !p ) return; + len = LEN(p); + dlen = 0; + for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) { + r = rem(NM(CQ(t)),mod); + if ( r ) { + if ( SGN(CQ(t)) < 0 ) + r = mod-r; + CM(d) = r; + ndl_copy(DL(t),DL(d)); + NMV_ADV(d); + dlen++; + } + } + LEN(p) = dlen; +} + +NDV ptondv(VL vl,VL dvl,P p) +{ + ND nd; + + nd = ptond(vl,dvl,p); + return ndtondv(0,nd); +} + +ND ptond(VL vl,VL dvl,P p) +{ + int n,i,j,k,e; + VL tvl; + V v; + DCP dc; + DCP *w; + ND r,s,t,u; + P x; + int c; + UINT *d; + NM m,m0; + if ( !p ) return 0; - for ( t = BDY(p), l = 0; t; t = NEXT(t), l++ ); - if ( mod ) { - _dp_mod(p,mod,0,&q); p = q; - m0 = m = (NMV)MALLOC_ATOMIC(l*nmv_adv); - } else - m0 = m = (NMV)MALLOC(l*nmv_adv); - n = NV(p); - for ( t = BDY(p), i = 0; i < l; i++, t = NEXT(t), NMV_ADV(m) ) { - if ( mod ) - CM(m) = ITOS(C(t)); - else - CQ(m) = (Q)C(t); - TD(m) = TD(DL(t)); - dltondl(n,DL(t),DL(m)); + else if ( NUM(p) ) { + NEWNM(m); + ndl_zero(DL(m)); + CQ(m) = (Q)p; + NEXT(m) = 0; + MKND(nd_nvar,m,1,r); + SG(r) = 0; + return r; + } else { + for ( dc = DC(p), k = 0; dc; dc = NEXT(dc), k++ ); + w = (DCP *)ALLOCA(k*sizeof(DCP)); + for ( dc = DC(p), j = 0; j < k; dc = NEXT(dc), j++ ) w[j] = dc; + for ( i = 0, tvl = dvl, v = VR(p); + vl && tvl->v != v; tvl = NEXT(tvl), i++ ); + if ( !tvl ) { + for ( j = k-1, s = 0, MKV(v,x); j >= 0; j-- ) { + t = ptond(vl,dvl,COEF(w[j])); + pwrp(vl,x,DEG(w[j]),&p); + nd_mul_c_p(CO,t,p); s = nd_add(0,s,t); + } + return s; + } else { + NEWNM(m0); d = DL(m0); + for ( j = k-1, s = 0; j >= 0; j-- ) { + ndl_zero(d); e = QTOS(DEG(w[j])); PUT_EXP(d,i,e); + TD(d) = MUL_WEIGHT(e,i); + if ( nd_blockmask) ndl_weight_mask(d); + t = ptond(vl,dvl,COEF(w[j])); + for ( m = BDY(t); m; m = NEXT(m) ) + ndl_addto(DL(m),d); + SG(t) += TD(d); + s = nd_add(0,s,t); + } + FREENM(m0); + return s; + } } - MKNDV(n,m0,l,d); +} + +P ndvtop(int mod,VL vl,VL dvl,NDV p) +{ + VL tvl; + int len,n,j,i,e; + NMV m; + Q q; + P c; + UINT *d; + P s,r,u,t,w; + + if ( !p ) return 0; + else { + len = LEN(p); + n = NV(p); + m = (NMV)(((char *)BDY(p))+nmv_adv*(len-1)); + for ( j = len-1, s = 0; j >= 0; j--, NMV_PREV(m) ) { + if ( mod ) { + STOQ(CM(m),q); c = (P)q; + } else + c = CP(m); + d = DL(m); + for ( i = 0, t = c, tvl = dvl; i < n; tvl = NEXT(tvl), i++ ) { + MKV(tvl->v,r); e = GET_EXP(d,i); STOQ(e,q); + pwrp(vl,r,q,&u); mulp(vl,t,u,&w); t = w; + } + addp(vl,s,t,&u); s = u; + } + return s; + } +} + +NDV ndtondv(int mod,ND p) +{ + NDV d; + NMV m,m0; + NM t; + int i,len; + + if ( !p ) return 0; + len = LEN(p); + m0 = m = (NMV)(mod?MALLOC_ATOMIC(len*nmv_adv):MALLOC(len*nmv_adv)); + for ( t = BDY(p), i = 0; t; t = NEXT(t), i++, NMV_ADV(m) ) { + ndl_copy(DL(t),DL(m)); + CQ(m) = CQ(t); + } + MKNDV(NV(p),m0,len,d); SG(d) = SG(p); return d; } -DP ndvtodp(int mod,NDV p) +ND ndvtond(int mod,NDV p) { - DP d; - MP m0,m; + ND d; + NM m,m0; NMV t; - int len,i,n; + int i,len; - if ( !p ) - return 0; + if ( !p ) return 0; m0 = 0; - len = LEN(p); - n = NV(p); - for ( t = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) { - NEXTMP(m0,m); - if ( mod ) - C(m) = STOI(CM(t)); - else - C(m) = (P)CQ(t); - DL(m) = ndltodl(n,TD(t),DL(t)); + len = p->len; + for ( t = BDY(p), i = 0; i < len; NMV_ADV(t), i++ ) { + NEXTNM(m0,m); + ndl_copy(DL(t),DL(m)); + CQ(m) = CQ(t); } NEXT(m) = 0; - MKDP(NV(p),m0,d); + MKND(NV(p),m0,len,d); SG(d) = SG(p); return d; } @@ -2838,8 +3532,7 @@ void ndv_print(NDV p) NMV m; int i,len; - if ( !p ) - printf("0\n"); + if ( !p ) printf("0\n"); else { len = LEN(p); for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) { @@ -2855,8 +3548,7 @@ void ndv_print_q(NDV p) NMV m; int i,len; - if ( !p ) - printf("0\n"); + if ( !p ) printf("0\n"); else { len = LEN(p); for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) { @@ -2869,14 +3561,452 @@ void ndv_print_q(NDV p) } } -int nd_equal(ND a,ND b) +NODE ndv_reducebase(NODE x) { - NM ma,mb; + int len,i,j; + NDV *w; + NODE t,t0; - if ( (!a && b)||(a&& !b)||(NV(a) != NV(b))||(SG(a) != SG(b)) ) return 0; - for ( ma = BDY(a), mb = BDY(b); ma && mb; ma = NEXT(ma), mb = NEXT(mb) ) { - if ( (TD(ma) != TD(mb)) || !ndl_equal(DL(ma),DL(mb)) ) return 0; + len = length(x); + w = (NDV *)ALLOCA(len*sizeof(NDV)); + for ( i = 0, t = x; i < len; i++, t = NEXT(t) ) w[i] = BDY(t); + for ( i = 0; i < len; i++ ) { + for ( j = 0; j < i; j++ ) { + if ( w[i] && w[j] ) + if ( ndl_reducible(HDL(w[i]),HDL(w[j])) ) w[i] = 0; + else if ( ndl_reducible(HDL(w[j]),HDL(w[i])) ) w[j] = 0; + } } - if ( !ma && !mb ) return 1; - else return 0; + for ( i = len-1, t0 = 0; i >= 0; i-- ) { + if ( w[i] ) { NEXTNODE(t0,t); BDY(t) = (pointer)w[i]; } + } + NEXT(t) = 0; x = t0; + return x; +} + +/* XXX incomplete */ + +void nd_init_ord(struct order_spec *ord) +{ + switch ( ord->id ) { + case 0: + switch ( ord->ord.simple ) { + case 0: + nd_dcomp = 1; + nd_isrlex = 1; + break; + case 1: + nd_dcomp = 1; + nd_isrlex = 0; + break; + case 2: + nd_dcomp = 0; + nd_isrlex = 0; + ndl_compare_function = ndl_lex_compare; + break; + case 11: + /* XXX */ + nd_dcomp = 0; + nd_isrlex = 1; + ndl_compare_function = ndl_ww_lex_compare; + break; + default: + error("nd_gr : unsupported order"); + } + break; + case 1: + /* XXX */ + nd_dcomp = -1; + nd_isrlex = 0; + ndl_compare_function = ndl_block_compare; + break; + case 2: + error("nd_init_ord : matrix order is not supported yet."); + break; + } + nd_ord = ord; +} + +BlockMask nd_create_blockmask(struct order_spec *ord) +{ + int n,i,j,s,l; + UINT *t; + BlockMask bm; + + if ( !ord->id ) + return 0; + n = ord->ord.block.length; + bm = (BlockMask)MALLOC(sizeof(struct oBlockMask)); + bm->n = n; + bm->order_pair = ord->ord.block.order_pair; + bm->mask = (UINT **)MALLOC(n*sizeof(UINT *)); + for ( i = 0, s = 0; i < n; i++ ) { + bm->mask[i] = t = (UINT *)MALLOC_ATOMIC(nd_wpd*sizeof(UINT)); + for ( j = 0; j < nd_wpd; j++ ) t[j] = 0; + l = bm->order_pair[i].length; + for ( j = 0; j < l; j++, s++ ) PUT_EXP(t,s,nd_mask0); + } + return bm; +} + +EPOS nd_create_epos(struct order_spec *ord) +{ + int i,j,l,s,ord_l,ord_o; + EPOS epos; + struct order_pair *op; + + epos = (EPOS)MALLOC_ATOMIC(nd_nvar*sizeof(struct oEPOS)); + switch ( ord->id ) { + case 0: + if ( nd_isrlex ) { + for ( i = 0; i < nd_nvar; i++ ) { + epos[i].i = nd_exporigin + (nd_nvar-1-i)/nd_epw; + epos[i].s = (nd_epw-((nd_nvar-1-i)%nd_epw)-1)*nd_bpe; + } + } else { + for ( i = 0; i < nd_nvar; i++ ) { + epos[i].i = nd_exporigin + i/nd_epw; + epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe; + } + } + break; + case 1: + /* block order */ + l = ord->ord.block.length; + op = ord->ord.block.order_pair; + for ( j = 0, s = 0; j < l; j++ ) { + ord_o = op[j].order; + ord_l = op[j].length; + if ( !ord_o ) + for ( i = 0; i < ord_l; i++ ) { + epos[s+i].i = nd_exporigin + (s+ord_l-i-1)/nd_epw; + epos[s+i].s = (nd_epw-((s+ord_l-i-1)%nd_epw)-1)*nd_bpe; + } + else + for ( i = 0; i < ord_l; i++ ) { + epos[s+i].i = nd_exporigin + (s+i)/nd_epw; + epos[s+i].s = (nd_epw-((s+i)%nd_epw)-1)*nd_bpe; + } + s += ord_l; + } + break; + case 2: + error("nd_create_epos : matrix order is not supported yet."); + } + return epos; +} + +/* external interface */ + +void nd_nf_p(P f,LIST g,LIST v,int m,struct order_spec *ord,P *rp) +{ + NODE t,in0,in; + ND nd,nf; + NDV ndv; + VL vv,tv; + int stat,nvar,max,e; + + pltovl(v,&vv); + nvar = length(vv); + + /* get the degree bound */ + for ( t = BDY(g), max = 0; t; t = NEXT(t) ) + for ( tv = vv; tv; tv = NEXT(tv) ) { + e = getdeg(tv->v,(P)BDY(t)); + max = MAX(e,max); + } + for ( tv = vv; tv; tv = NEXT(tv) ) { + e = getdeg(tv->v,f); + max = MAX(e,max); + } + + nd_init_ord(ord); + nd_setup_parameters(nvar,max); + + /* conversion to ndv */ + for ( in0 = 0, t = BDY(g); t; t = NEXT(t) ) { + NEXTNODE(in0,in); + BDY(in) = (pointer)ptondv(CO,vv,(P)BDY(t)); + } + NEXTNODE(in0,in); + BDY(in) = (pointer)ptondv(CO,vv,f); + NEXT(in) = 0; + + ndv_setup(m,0,in0); + nd_psn--; + nd_scale=2; + while ( 1 ) { + nd = (pointer)ndvtond(m,nd_ps[nd_psn]); + stat = nd_nf(m,nd,nd_ps,1,&nf); + if ( !stat ) { + nd_psn++; + nd_reconstruct(m,0,0); + nd_psn--; + } else + break; + } + *rp = ndvtop(m,CO,vv,ndtondv(m,nf)); +} + +int nd_to_vect(int mod,UINT *s0,int n,ND d,UINT *r) +{ + NM m; + UINT *t,*s; + int i; + + for ( i = 0; i < n; i++ ) r[i] = 0; + for ( i = 0, s = s0, m = BDY(d); m; m = NEXT(m) ) { + t = DL(m); + for ( ; !ndl_equal(t,s); s += nd_wpd, i++ ); + r[i] = CM(m); + } + for ( i = 0; !r[i]; i++ ); + return i; +} + +int nm_ind_pair_to_vect(int mod,UINT *s0,int n,NM_ind_pair pair,UINT *r) +{ + NM m; + NMV mr; + UINT *d,*t,*s; + NDV p; + int i,j,len; + + m = pair->mul; + d = DL(m); + p = nd_ps[pair->index]; + t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT)); + for ( i = 0; i < n; i++ ) r[i] = 0; + len = LEN(p); + for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) { + ndl_add(d,DL(mr),t); + for ( ; !ndl_equal(t,s); s += nd_wpd, i++ ); + r[i] = CM(mr); + } + for ( i = 0; !r[i]; i++ ); + return i; +} + +void nm_ind_pair_to_vect_compress(int mod,UINT *s0,int n,NM_ind_pair pair,int *ind) +{ + NM m; + NMV mr; + UINT *d,*t,*s; + NDV p; + int i,j,len; + + m = pair->mul; + d = DL(m); + p = nd_ps[pair->index]; + len = LEN(p); + t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT)); + for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) { + ndl_add(d,DL(mr),t); + for ( ; !ndl_equal(t,s); s += nd_wpd, i++ ); + ind[j] = i; + } +} + + +void ndv_reduce_vect(int m,UINT *svect,int **imat,NODE rp0) +{ + int i,j,k,len,pos; + UINT c,c1,c2,c3; + UINT *ivect; + NDV redv; + NMV mr0,mr; + NODE rp; + + for ( rp = rp0, i = 0; rp; i++, rp = NEXT(rp) ) { + ivect = imat[i]; + k = ivect[0]; + if ( c = svect[k] ) { + c = m-c; + redv = nd_ps[((NM_ind_pair)BDY(rp))->index]; + len = LEN(redv); + mr0 = BDY(redv); + for ( j = 0, mr = mr0; j < len; j++, NMV_ADV(mr) ) { + pos = ivect[j]; + c1 = CM(mr); + c2 = svect[pos]; DMAR(c1,c,c2,m,c3); + svect[pos] = c3; + } + } + } +} + +NDV vect_to_ndv(UINT *vect,int spcol,int col,int *rhead,UINT *s0vect) +{ + int j,k,len; + UINT *p; + UINT c; + NDV r; + NMV mr0,mr; + + for ( j = 0, len = 0; j < spcol; j++ ) if ( vect[j] ) len++; + if ( !len ) return 0; + else { + mr0 = (NMV)MALLOC_ATOMIC(nmv_adv*len); + mr = mr0; + p = s0vect; + for ( j = k = 0; j < col; j++, p += nd_wpd ) + if ( !rhead[j] ) { + if ( c = vect[k++] ) { + ndl_copy(p,DL(mr)); CM(mr) = c; NMV_ADV(mr); + } + } + MKNDV(nd_nvar,mr0,len,r); + return r; + } +} + +NODE nd_sp_f4(int m,ND_pairs l,PGeoBucket bucket) +{ + ND_pairs t; + NODE sp0,sp; + int stat; + ND spol; + + sp0 = 0; + for ( t = l; t; t = NEXT(t) ) { + stat = nd_sp(m,0,t,&spol); + if ( !stat ) return 0; + if ( spol ) { + NEXTNODE(sp0,sp); BDY(sp) = (pointer)nd_dup(spol); + add_pbucket_symbolic(bucket,spol); + } + } + return sp0; +} + +int nd_symbolic_preproc(PGeoBucket bucket,UINT **s0vect,NODE *r) +{ + NODE rp0,rp; + NM mul,head,s0,s; + int index,col,i; + RHist h; + UINT *s0v,*p; + NM_ind_pair pair; + ND red; + + s0 = 0; rp0 = 0; col = 0; + while ( 1 ) { + head = remove_head_pbucket_symbolic(bucket); + if ( !head ) break; + if ( !s0 ) s0 = head; + else NEXT(s) = head; + s = head; + index = ndl_find_reducer(DL(head)); + if ( index >= 0 ) { + h = nd_psh[index]; + NEWNM(mul); + ndl_sub(DL(head),DL(h),DL(mul)); + if ( ndl_check_bound2(index,DL(mul)) ) return 0; + MKNM_ind_pair(pair,mul,index); + red = ndv_mul_nm_symbolic(mul,nd_ps[index]); + add_pbucket_symbolic(bucket,nd_remove_head(red)); + NEXTNODE(rp0,rp); BDY(rp) = (pointer)pair; + } + col++; + } + NEXT(rp) = 0; NEXT(s) = 0; + s0v = (UINT *)MALLOC_ATOMIC(col*nd_wpd*sizeof(UINT)); + for ( i = 0, p = s0v, s = s0; i < col; + i++, p += nd_wpd, s = NEXT(s) ) ndl_copy(DL(s),p); + *s0vect = s0v; + *r = rp0; + return col; +} + +NODE nd_f4(int m) +{ + int i,nh,stat,index; + NODE r,g; + ND_pairs d,l,t; + ND spol,red; + NDV nf,redv; + NM s0,s; + NODE sp0,sp,rp0,rp; + int nsp,nred,col,rank,len,k,j,a; + UINT c; + UINT **spmat; + UINT *s0vect,*svect,*p,*v; + int *colstat; + int **imat; + int *rhead; + int spcol,sprow; + int sugar; + PGeoBucket bucket; + + if ( !m ) + error("nd_f4 : not implemented"); + + g = 0; d = 0; + for ( i = 0; i < nd_psn; i++ ) { + d = update_pairs(d,g,i); + g = update_base(g,i); + } + while ( d ) { + l = nd_minsugarp(d,&d); + sugar = SG(l); + bucket = create_pbucket(); + if ( !(sp0 = nd_sp_f4(m,l,bucket)) + || !(col = nd_symbolic_preproc(bucket,&s0vect,&rp0)) ) { + for ( t = l; NEXT(t); t = NEXT(t) ); + NEXT(t) = d; d = l; + d = nd_reconstruct(m,0,d); + continue; + } + + nsp = length(sp0); nred = length(rp0); spcol = col-nred; + imat = (int **)MALLOC(nred*sizeof(int *)); + rhead = (int *)MALLOC_ATOMIC(col*sizeof(int)); + for ( i = 0; i < col; i++ ) rhead[i] = 0; + + /* construction of index arrays */ + for ( rp = rp0, i = 0; rp; i++, rp = NEXT(rp) ) { + redv = nd_ps[((NM_ind_pair)BDY(rp))->index]; + imat[i] = (int *)MALLOC_ATOMIC(LEN(redv)*sizeof(int)); + nm_ind_pair_to_vect_compress(m,s0vect,col, + (NM_ind_pair)BDY(rp),imat[i]); + rhead[imat[i][0]] = 1; + } + + /* elimination (1st step) */ + spmat = (UINT **)MALLOC(nsp*sizeof(UINT *)); + svect = (UINT *)MALLOC_ATOMIC(col*sizeof(UINT)); + for ( a = sprow = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) { + nd_to_vect(m,s0vect,col,BDY(sp),svect); + ndv_reduce_vect(m,svect,imat,rp0); + for ( i = 0; i < col; i++ ) if ( svect[i] ) break; + if ( i < col ) { + spmat[sprow] = v = (UINT *)MALLOC_ATOMIC(spcol*sizeof(UINT)); + for ( j = k = 0; j < col; j++ ) + if ( !rhead[j] ) v[k++] = svect[j]; + sprow++; + } + } + /* free index arrays */ + for ( i = 0; i < nred; i++ ) GC_free(imat[i]); + + /* elimination (2nd step) */ + colstat = (int *)ALLOCA(spcol*sizeof(int)); + rank = generic_gauss_elim_mod(spmat,sprow,spcol,m,colstat); + + fprintf(asir_out,"sugar=%d,rank=%d\n",sugar,rank); fflush(asir_out); + + /* adding new bases */ + for ( i = sprow-1; i >= rank; i-- ) GC_free(spmat[i]); + for ( ; i >= 0; i-- ) { + nf = vect_to_ndv(spmat[i],spcol,col,rhead,s0vect); + SG(nf) = sugar; + ndv_removecont(m,nf); + nh = ndv_newps(nf,0); + d = update_pairs(d,g,nh); + g = update_base(g,nh); + GC_free(spmat[i]); + } + } + for ( r = g; r; r = NEXT(r) ) BDY(r) = (pointer)nd_ps[(int)BDY(r)]; + return g; }