1 #ifdef PETSC_RCS_HEADER 2 static char vcid[] = "$Id: mpibaij.c,v 1.104 1998/02/18 20:37:17 balay Exp balay $"; 3 #endif 4 5 #include "pinclude/pviewer.h" 6 #include "src/mat/impls/baij/mpi/mpibaij.h" 7 #include "src/vec/vecimpl.h" 8 9 10 extern int MatSetUpMultiply_MPIBAIJ(Mat); 11 extern int DisAssemble_MPIBAIJ(Mat); 12 extern int MatIncreaseOverlap_MPIBAIJ(Mat,int,IS *,int); 13 extern int MatGetSubMatrices_MPIBAIJ(Mat,int,IS *,IS *,MatGetSubMatrixCall,Mat **); 14 15 /* 16 Local utility routine that creates a mapping from the global column 17 number to the local number in the off-diagonal part of the local 18 storage of the matrix. This is done in a non scable way since the 19 length of colmap equals the global matrix length. 20 */ 21 #undef __FUNC__ 22 #define __FUNC__ "CreateColmap_MPIBAIJ_Private" 23 static int CreateColmap_MPIBAIJ_Private(Mat mat) 24 { 25 Mat_MPIBAIJ *baij = (Mat_MPIBAIJ *) mat->data; 26 Mat_SeqBAIJ *B = (Mat_SeqBAIJ*) baij->B->data; 27 int nbs = B->nbs,i,bs=B->bs;; 28 29 PetscFunctionBegin; 30 baij->colmap = (int *) PetscMalloc((baij->Nbs+1)*sizeof(int));CHKPTRQ(baij->colmap); 31 PLogObjectMemory(mat,baij->Nbs*sizeof(int)); 32 PetscMemzero(baij->colmap,baij->Nbs*sizeof(int)); 33 for ( i=0; i<nbs; i++ ) baij->colmap[baij->garray[i]] = i*bs+1; 34 PetscFunctionReturn(0); 35 } 36 37 #define CHUNKSIZE 10 38 39 #define MatSetValues_SeqBAIJ_A_Private(row,col,value,addv) \ 40 { \ 41 \ 42 brow = row/bs; \ 43 rp = aj + ai[brow]; ap = aa + bs2*ai[brow]; \ 44 rmax = aimax[brow]; nrow = ailen[brow]; \ 45 bcol = col/bs; \ 46 ridx = row % bs; cidx = col % bs; \ 47 low = 0; high = nrow; \ 48 while (high-low > 3) { \ 49 t = (low+high)/2; \ 50 if (rp[t] > bcol) high = t; \ 51 else low = t; \ 52 } \ 53 for ( _i=low; _i<high; _i++ ) { \ 54 if (rp[_i] > bcol) break; \ 55 if (rp[_i] == bcol) { \ 56 bap = ap + bs2*_i + bs*cidx + ridx; \ 57 if (addv == ADD_VALUES) *bap += value; \ 58 else *bap = value; \ 59 goto a_noinsert; \ 60 } \ 61 } \ 62 if (a->nonew == 1) goto a_noinsert; \ 63 else if (a->nonew == -1) SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,0,"Inserting a new nonzero into matrix"); \ 64 if (nrow >= rmax) { \ 65 /* there is no extra room in row, therefore enlarge */ \ 66 int new_nz = ai[a->mbs] + CHUNKSIZE,len,*new_i,*new_j; \ 67 Scalar *new_a; \ 68 \ 69 if (a->nonew == -2) SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,0,"Inserting a new nonzero in the matrix"); \ 70 \ 71 /* malloc new storage space */ \ 72 len = new_nz*(sizeof(int)+bs2*sizeof(Scalar))+(a->mbs+1)*sizeof(int); \ 73 new_a = (Scalar *) PetscMalloc( len ); CHKPTRQ(new_a); \ 74 new_j = (int *) (new_a + bs2*new_nz); \ 75 new_i = new_j + new_nz; \ 76 \ 77 /* copy over old data into new slots */ \ 78 for ( ii=0; ii<brow+1; ii++ ) {new_i[ii] = ai[ii];} \ 79 for ( ii=brow+1; ii<a->mbs+1; ii++ ) {new_i[ii] = ai[ii]+CHUNKSIZE;} \ 80 PetscMemcpy(new_j,aj,(ai[brow]+nrow)*sizeof(int)); \ 81 len = (new_nz - CHUNKSIZE - ai[brow] - nrow); \ 82 PetscMemcpy(new_j+ai[brow]+nrow+CHUNKSIZE,aj+ai[brow]+nrow, \ 83 len*sizeof(int)); \ 84 PetscMemcpy(new_a,aa,(ai[brow]+nrow)*bs2*sizeof(Scalar)); \ 85 PetscMemzero(new_a+bs2*(ai[brow]+nrow),bs2*CHUNKSIZE*sizeof(Scalar)); \ 86 PetscMemcpy(new_a+bs2*(ai[brow]+nrow+CHUNKSIZE), \ 87 aa+bs2*(ai[brow]+nrow),bs2*len*sizeof(Scalar)); \ 88 /* free up old matrix storage */ \ 89 PetscFree(a->a); \ 90 if (!a->singlemalloc) {PetscFree(a->i);PetscFree(a->j);} \ 91 aa = a->a = new_a; ai = a->i = new_i; aj = a->j = new_j; \ 92 a->singlemalloc = 1; \ 93 \ 94 rp = aj + ai[brow]; ap = aa + bs2*ai[brow]; \ 95 rmax = aimax[brow] = aimax[brow] + CHUNKSIZE; \ 96 PLogObjectMemory(A,CHUNKSIZE*(sizeof(int) + bs2*sizeof(Scalar))); \ 97 a->maxnz += bs2*CHUNKSIZE; \ 98 a->reallocs++; \ 99 a->nz++; \ 100 } \ 101 N = nrow++ - 1; \ 102 /* shift up all the later entries in this row */ \ 103 for ( ii=N; ii>=_i; ii-- ) { \ 104 rp[ii+1] = rp[ii]; \ 105 PetscMemcpy(ap+bs2*(ii+1),ap+bs2*(ii),bs2*sizeof(Scalar)); \ 106 } \ 107 if (N>=_i) PetscMemzero(ap+bs2*_i,bs2*sizeof(Scalar)); \ 108 rp[_i] = bcol; \ 109 ap[bs2*_i + bs*cidx + ridx] = value; \ 110 a_noinsert:; \ 111 ailen[brow] = nrow; \ 112 } 113 114 #define MatSetValues_SeqBAIJ_B_Private(row,col,value,addv) \ 115 { \ 116 \ 117 brow = row/bs; \ 118 rp = bj + bi[brow]; ap = ba + bs2*bi[brow]; \ 119 rmax = bimax[brow]; nrow = bilen[brow]; \ 120 bcol = col/bs; \ 121 ridx = row % bs; cidx = col % bs; \ 122 low = 0; high = nrow; \ 123 while (high-low > 3) { \ 124 t = (low+high)/2; \ 125 if (rp[t] > bcol) high = t; \ 126 else low = t; \ 127 } \ 128 for ( _i=low; _i<high; _i++ ) { \ 129 if (rp[_i] > bcol) break; \ 130 if (rp[_i] == bcol) { \ 131 bap = ap + bs2*_i + bs*cidx + ridx; \ 132 if (addv == ADD_VALUES) *bap += value; \ 133 else *bap = value; \ 134 goto b_noinsert; \ 135 } \ 136 } \ 137 if (b->nonew == 1) goto b_noinsert; \ 138 else if (b->nonew == -1) SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,0,"Inserting a new nonzero into matrix"); \ 139 if (nrow >= rmax) { \ 140 /* there is no extra room in row, therefore enlarge */ \ 141 int new_nz = bi[b->mbs] + CHUNKSIZE,len,*new_i,*new_j; \ 142 Scalar *new_a; \ 143 \ 144 if (b->nonew == -2) SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,0,"Inserting a new nonzero in the matrix"); \ 145 \ 146 /* malloc new storage space */ \ 147 len = new_nz*(sizeof(int)+bs2*sizeof(Scalar))+(b->mbs+1)*sizeof(int); \ 148 new_a = (Scalar *) PetscMalloc( len ); CHKPTRQ(new_a); \ 149 new_j = (int *) (new_a + bs2*new_nz); \ 150 new_i = new_j + new_nz; \ 151 \ 152 /* copy over old data into new slots */ \ 153 for ( ii=0; ii<brow+1; ii++ ) {new_i[ii] = bi[ii];} \ 154 for ( ii=brow+1; ii<b->mbs+1; ii++ ) {new_i[ii] = bi[ii]+CHUNKSIZE;} \ 155 PetscMemcpy(new_j,bj,(bi[brow]+nrow)*sizeof(int)); \ 156 len = (new_nz - CHUNKSIZE - bi[brow] - nrow); \ 157 PetscMemcpy(new_j+bi[brow]+nrow+CHUNKSIZE,bj+bi[brow]+nrow, \ 158 len*sizeof(int)); \ 159 PetscMemcpy(new_a,ba,(bi[brow]+nrow)*bs2*sizeof(Scalar)); \ 160 PetscMemzero(new_a+bs2*(bi[brow]+nrow),bs2*CHUNKSIZE*sizeof(Scalar)); \ 161 PetscMemcpy(new_a+bs2*(bi[brow]+nrow+CHUNKSIZE), \ 162 ba+bs2*(bi[brow]+nrow),bs2*len*sizeof(Scalar)); \ 163 /* free up old matrix storage */ \ 164 PetscFree(b->a); \ 165 if (!b->singlemalloc) {PetscFree(b->i);PetscFree(b->j);} \ 166 ba = b->a = new_a; bi = b->i = new_i; bj = b->j = new_j; \ 167 b->singlemalloc = 1; \ 168 \ 169 rp = bj + bi[brow]; ap = ba + bs2*bi[brow]; \ 170 rmax = bimax[brow] = bimax[brow] + CHUNKSIZE; \ 171 PLogObjectMemory(B,CHUNKSIZE*(sizeof(int) + bs2*sizeof(Scalar))); \ 172 b->maxnz += bs2*CHUNKSIZE; \ 173 b->reallocs++; \ 174 b->nz++; \ 175 } \ 176 N = nrow++ - 1; \ 177 /* shift up all the later entries in this row */ \ 178 for ( ii=N; ii>=_i; ii-- ) { \ 179 rp[ii+1] = rp[ii]; \ 180 PetscMemcpy(ap+bs2*(ii+1),ap+bs2*(ii),bs2*sizeof(Scalar)); \ 181 } \ 182 if (N>=_i) PetscMemzero(ap+bs2*_i,bs2*sizeof(Scalar)); \ 183 rp[_i] = bcol; \ 184 ap[bs2*_i + bs*cidx + ridx] = value; \ 185 b_noinsert:; \ 186 bilen[brow] = nrow; \ 187 } 188 189 #undef __FUNC__ 190 #define __FUNC__ "MatSetValues_MPIBAIJ" 191 int MatSetValues_MPIBAIJ(Mat mat,int m,int *im,int n,int *in,Scalar *v,InsertMode addv) 192 { 193 Mat_MPIBAIJ *baij = (Mat_MPIBAIJ *) mat->data; 194 Scalar value; 195 int ierr,i,j,row,col; 196 int roworiented = baij->roworiented,rstart_orig=baij->rstart_bs ; 197 int rend_orig=baij->rend_bs,cstart_orig=baij->cstart_bs; 198 int cend_orig=baij->cend_bs,bs=baij->bs; 199 200 /* Some Variables required in the macro */ 201 Mat A = baij->A; 202 Mat_SeqBAIJ *a = (Mat_SeqBAIJ *) (A)->data; 203 int *aimax=a->imax,*ai=a->i,*ailen=a->ilen,*aj=a->j; 204 Scalar *aa=a->a; 205 206 Mat B = baij->B; 207 Mat_SeqBAIJ *b = (Mat_SeqBAIJ *) (B)->data; 208 int *bimax=b->imax,*bi=b->i,*bilen=b->ilen,*bj=b->j; 209 Scalar *ba=b->a; 210 211 int *rp,ii,nrow,_i,rmax,N,brow,bcol; 212 int low,high,t,ridx,cidx,bs2=a->bs2; 213 Scalar *ap,*bap; 214 215 PetscFunctionBegin; 216 for ( i=0; i<m; i++ ) { 217 #if defined(USE_PETSC_BOPT_g) 218 if (im[i] < 0) SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,0,"Negative row"); 219 if (im[i] >= baij->M) SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,0,"Row too large"); 220 #endif 221 if (im[i] >= rstart_orig && im[i] < rend_orig) { 222 row = im[i] - rstart_orig; 223 for ( j=0; j<n; j++ ) { 224 if (in[j] >= cstart_orig && in[j] < cend_orig){ 225 col = in[j] - cstart_orig; 226 if (roworiented) value = v[i*n+j]; else value = v[i+j*m]; 227 MatSetValues_SeqBAIJ_A_Private(row,col,value,addv); 228 /* ierr = MatSetValues_SeqBAIJ(baij->A,1,&row,1,&col,&value,addv);CHKERRQ(ierr); */ 229 } 230 #if defined(USE_PETSC_BOPT_g) 231 else if (in[j] < 0) {SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,0,"Negative column");} 232 else if (in[j] >= baij->N) {SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,0,"Col too large");} 233 #endif 234 else { 235 if (mat->was_assembled) { 236 if (!baij->colmap) { 237 ierr = CreateColmap_MPIBAIJ_Private(mat);CHKERRQ(ierr); 238 } 239 col = baij->colmap[in[j]/bs] - 1 + in[j]%bs; 240 if (col < 0 && !((Mat_SeqBAIJ*)(baij->A->data))->nonew) { 241 ierr = DisAssemble_MPIBAIJ(mat); CHKERRQ(ierr); 242 col = in[j]; 243 /* Reinitialize the variables required by MatSetValues_SeqBAIJ_B_Private() */ 244 B = baij->B; 245 b = (Mat_SeqBAIJ *) (B)->data; 246 bimax=b->imax;bi=b->i;bilen=b->ilen;bj=b->j; 247 ba=b->a; 248 } 249 } else col = in[j]; 250 if (roworiented) value = v[i*n+j]; else value = v[i+j*m]; 251 MatSetValues_SeqBAIJ_B_Private(row,col,value,addv); 252 /* ierr = MatSetValues_SeqBAIJ(baij->B,1,&row,1,&col,&value,addv);CHKERRQ(ierr); */ 253 } 254 } 255 } else { 256 if (roworiented && !baij->donotstash) { 257 ierr = StashValues_Private(&baij->stash,im[i],n,in,v+i*n,addv);CHKERRQ(ierr); 258 } else { 259 if (!baij->donotstash) { 260 row = im[i]; 261 for ( j=0; j<n; j++ ) { 262 ierr = StashValues_Private(&baij->stash,row,1,in+j,v+i+j*m,addv);CHKERRQ(ierr); 263 } 264 } 265 } 266 } 267 } 268 PetscFunctionReturn(0); 269 } 270 271 extern int MatSetValuesBlocked_SeqBAIJ(Mat,int,int*,int,int*,Scalar*,InsertMode); 272 #undef __FUNC__ 273 #define __FUNC__ "MatSetValuesBlocked_MPIBAIJ" 274 int MatSetValuesBlocked_MPIBAIJ(Mat mat,int m,int *im,int n,int *in,Scalar *v,InsertMode addv) 275 { 276 Mat_MPIBAIJ *baij = (Mat_MPIBAIJ *) mat->data; 277 Scalar *value,*barray=baij->barray; 278 int ierr,i,j,ii,jj,row,col,k,l; 279 int roworiented = baij->roworiented,rstart=baij->rstart ; 280 int rend=baij->rend,cstart=baij->cstart,stepval; 281 int cend=baij->cend,bs=baij->bs,bs2=baij->bs2; 282 283 if(!barray) { 284 baij->barray = barray = (Scalar*) PetscMalloc(bs2*sizeof(Scalar)); CHKPTRQ(barray); 285 } 286 287 if (roworiented) { 288 stepval = (n-1)*bs; 289 } else { 290 stepval = (m-1)*bs; 291 } 292 for ( i=0; i<m; i++ ) { 293 #if defined(USE_PETSC_BOPT_g) 294 if (im[i] < 0) SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,0,"Negative row"); 295 if (im[i] >= baij->Mbs) SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,0,"Row too large"); 296 #endif 297 if (im[i] >= rstart && im[i] < rend) { 298 row = im[i] - rstart; 299 for ( j=0; j<n; j++ ) { 300 /* If NumCol = 1 then a copy is not required */ 301 if ((roworiented) && (n == 1)) { 302 barray = v + i*bs2; 303 } else if((!roworiented) && (m == 1)) { 304 barray = v + j*bs2; 305 } else { /* Here a copy is required */ 306 if (roworiented) { 307 value = v + i*(stepval+bs)*bs + j*bs; 308 } else { 309 value = v + j*(stepval+bs)*bs + i*bs; 310 } 311 for ( ii=0; ii<bs; ii++,value+=stepval ) { 312 for (jj=0; jj<bs; jj++ ) { 313 *barray++ = *value++; 314 } 315 } 316 barray -=bs2; 317 } 318 319 if (in[j] >= cstart && in[j] < cend){ 320 col = in[j] - cstart; 321 ierr = MatSetValuesBlocked_SeqBAIJ(baij->A,1,&row,1,&col,barray,addv);CHKERRQ(ierr); 322 } 323 #if defined(USE_PETSC_BOPT_g) 324 else if (in[j] < 0) {SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,0,"Negative column");} 325 else if (in[j] >= baij->Nbs) {SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,0,"Column too large");} 326 #endif 327 else { 328 if (mat->was_assembled) { 329 if (!baij->colmap) { 330 ierr = CreateColmap_MPIBAIJ_Private(mat);CHKERRQ(ierr); 331 } 332 333 #if defined(USE_PETSC_BOPT_g) 334 if ((baij->colmap[in[j]] - 1) % bs) {SETERRQ(PETSC_ERR_PLIB,0,"Incorrect colmap");} 335 #endif 336 col = (baij->colmap[in[j]] - 1)/bs; 337 if (col < 0 && !((Mat_SeqBAIJ*)(baij->A->data))->nonew) { 338 ierr = DisAssemble_MPIBAIJ(mat); CHKERRQ(ierr); 339 col = in[j]; 340 } 341 } 342 else col = in[j]; 343 ierr = MatSetValuesBlocked_SeqBAIJ(baij->B,1,&row,1,&col,barray,addv);CHKERRQ(ierr); 344 } 345 } 346 } else { 347 if (!baij->donotstash) { 348 if (roworiented ) { 349 row = im[i]*bs; 350 value = v + i*(stepval+bs)*bs; 351 for ( j=0; j<bs; j++,row++ ) { 352 for ( k=0; k<n; k++ ) { 353 for ( col=in[k]*bs,l=0; l<bs; l++,col++) { 354 ierr = StashValues_Private(&baij->stash,row,1,&col,value++,addv);CHKERRQ(ierr); 355 } 356 } 357 } 358 } else { 359 for ( j=0; j<n; j++ ) { 360 value = v + j*(stepval+bs)*bs + i*bs; 361 col = in[j]*bs; 362 for ( k=0; k<bs; k++,col++,value+=stepval) { 363 for ( row = im[i]*bs,l=0; l<bs; l++,row++) { 364 ierr = StashValues_Private(&baij->stash,row,1,&col,value++,addv);CHKERRQ(ierr); 365 } 366 } 367 } 368 } 369 } 370 } 371 } 372 PetscFunctionReturn(0); 373 } 374 #include <math.h> 375 #define HASH_KEY 0.6180339887 376 /* #define HASH1(size,key) ((int)((size)*fmod(((key)*HASH_KEY),1))) */ 377 #define HASH(size,key,tmp) (tmp = (key)*HASH_KEY,(int)((size)*(tmp-(int)tmp))) 378 /* #define HASH(size,key,tmp) ((int)((size)*fmod(((key)*HASH_KEY),1))) */ 379 #undef __FUNC__ 380 #define __FUNC__ "MatSetValues_MPIBAIJ_HT" 381 int MatSetValues_MPIBAIJ_HT(Mat mat,int m,int *im,int n,int *in,Scalar *v,InsertMode addv) 382 { 383 Mat_MPIBAIJ *baij = (Mat_MPIBAIJ *) mat->data; 384 int ierr,i,j,row,col; 385 int roworiented = baij->roworiented,rstart_orig=baij->rstart_bs ; 386 int rend_orig=baij->rend_bs,Nbs=baij->Nbs; 387 int h1,key,size=baij->ht_size,bs=baij->bs,*HT=baij->ht,idx; 388 double tmp; 389 Scalar ** HD = baij->hd,value; 390 #if defined(USE_PETSC_BOPT_g) 391 int total_ct=baij->ht_total_ct,insert_ct=baij->ht_insert_ct; 392 #endif 393 394 PetscFunctionBegin; 395 396 for ( i=0; i<m; i++ ) { 397 #if defined(USE_PETSC_BOPT_g) 398 if (im[i] < 0) SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,0,"Negative row"); 399 if (im[i] >= baij->M) SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,0,"Row too large"); 400 #endif 401 row = im[i]; 402 if (row >= rstart_orig && row < rend_orig) { 403 for ( j=0; j<n; j++ ) { 404 col = in[j]; 405 if (roworiented) value = v[i*n+j]; else value = v[i+j*m]; 406 /* Look up into the Hash Table */ 407 key = (row/bs)*Nbs+(col/bs)+1; 408 h1 = HASH(size,key,tmp); 409 410 411 idx = h1; 412 #if defined(USE_PETSC_BOPT_g) 413 insert_ct++; 414 total_ct++; 415 if (HT[idx] != key) { 416 for ( idx=h1; (idx<size) && (HT[idx]!=key); idx++,total_ct++); 417 if (idx == size) { 418 for ( idx=0; (idx<h1) && (HT[idx]!=key); idx++,total_ct++); 419 if (idx == h1) { 420 SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,0,"(row,col) has no entry in the hash table"); 421 } 422 } 423 } 424 #else 425 if (HT[idx] != key) { 426 for ( idx=h1; (idx<size) && (HT[idx]!=key); idx++); 427 if (idx == size) { 428 for ( idx=0; (idx<h1) && (HT[idx]!=key); idx++); 429 if (idx == h1) { 430 SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,0,"(row,col) has no entry in the hash table"); 431 } 432 } 433 } 434 #endif 435 /* A HASH table entry is found, so insert the values at the correct address */ 436 if (addv == ADD_VALUES) *(HD[idx]+ (col % bs)*bs + (row % bs)) += value; 437 else *(HD[idx]+ (col % bs)*bs + (row % bs)) = value; 438 } 439 } else { 440 if (roworiented && !baij->donotstash) { 441 ierr = StashValues_Private(&baij->stash,im[i],n,in,v+i*n,addv);CHKERRQ(ierr); 442 } else { 443 if (!baij->donotstash) { 444 row = im[i]; 445 for ( j=0; j<n; j++ ) { 446 ierr = StashValues_Private(&baij->stash,row,1,in+j,v+i+j*m,addv);CHKERRQ(ierr); 447 } 448 } 449 } 450 } 451 } 452 #if defined(USE_PETSC_BOPT_g) 453 baij->ht_total_ct = total_ct; 454 baij->ht_insert_ct = insert_ct; 455 #endif 456 PetscFunctionReturn(0); 457 } 458 459 #undef __FUNC__ 460 #define __FUNC__ "MatSetValuesBlocked_MPIBAIJ_HT" 461 int MatSetValuesBlocked_MPIBAIJ_HT(Mat mat,int m,int *im,int n,int *in,Scalar *v,InsertMode addv) 462 { 463 Mat_MPIBAIJ *baij = (Mat_MPIBAIJ *) mat->data; 464 int ierr,i,j,ii,jj,row,col,k,l; 465 int roworiented = baij->roworiented,rstart=baij->rstart ; 466 int rend=baij->rend,stepval,bs=baij->bs,bs2=baij->bs2; 467 int h1,key,size=baij->ht_size,idx,*HT=baij->ht,Nbs=baij->Nbs; 468 double tmp; 469 Scalar ** HD = baij->hd,*value,*v_t,*baij_a; 470 #if defined(USE_PETSC_BOPT_g) 471 int total_ct=baij->ht_total_ct,insert_ct=baij->ht_insert_ct; 472 #endif 473 474 PetscFunctionBegin; 475 476 if (roworiented) { 477 stepval = (n-1)*bs; 478 } else { 479 stepval = (m-1)*bs; 480 } 481 for ( i=0; i<m; i++ ) { 482 #if defined(USE_PETSC_BOPT_g) 483 if (im[i] < 0) SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,0,"Negative row"); 484 if (im[i] >= baij->Mbs) SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,0,"Row too large"); 485 #endif 486 row = im[i]; 487 v_t = v + i*bs2; 488 if (row >= rstart && row < rend) { 489 for ( j=0; j<n; j++ ) { 490 col = in[j]; 491 492 /* Look up into the Hash Table */ 493 key = row*Nbs+col+1; 494 h1 = HASH(size,key,tmp); 495 496 idx = h1; 497 #if defined(USE_PETSC_BOPT_g) 498 total_ct++; 499 insert_ct++; 500 if (HT[idx] != key) { 501 for ( idx=h1; (idx<size) && (HT[idx]!=key); idx++,total_ct++); 502 if (idx == size) { 503 for ( idx=0; (idx<h1) && (HT[idx]!=key); idx++,total_ct++); 504 if (idx == h1) { 505 SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,0,"(row,col) has no entry in the hash table"); 506 } 507 } 508 } 509 #else 510 if (HT[idx] != key) { 511 for ( idx=h1; (idx<size) && (HT[idx]!=key); idx++); 512 if (idx == size) { 513 for ( idx=0; (idx<h1) && (HT[idx]!=key); idx++); 514 if (idx == h1) { 515 SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,0,"(row,col) has no entry in the hash table"); 516 } 517 } 518 } 519 #endif 520 baij_a = HD[idx]; 521 if (roworiented) { 522 /*value = v + i*(stepval+bs)*bs + j*bs;*/ 523 /* value = v + (i*(stepval+bs)+j)*bs; */ 524 value = v_t; 525 v_t += bs; 526 if (addv == ADD_VALUES) { 527 for ( ii=0; ii<bs; ii++,value+=stepval) { 528 for ( jj=ii; jj<bs2; jj+=bs ) { 529 baij_a[jj] += *value++; 530 } 531 } 532 } else { 533 for ( ii=0; ii<bs; ii++,value+=stepval) { 534 for ( jj=ii; jj<bs2; jj+=bs ) { 535 baij_a[jj] = *value++; 536 } 537 } 538 } 539 } else { 540 value = v + j*(stepval+bs)*bs + i*bs; 541 if (addv == ADD_VALUES) { 542 for ( ii=0; ii<bs; ii++,value+=stepval,baij_a+=bs ) { 543 for ( jj=0; jj<bs; jj++ ) { 544 baij_a[jj] += *value++; 545 } 546 } 547 } else { 548 for ( ii=0; ii<bs; ii++,value+=stepval,baij_a+=bs ) { 549 for ( jj=0; jj<bs; jj++ ) { 550 baij_a[jj] = *value++; 551 } 552 } 553 } 554 } 555 } 556 } else { 557 if (!baij->donotstash) { 558 if (roworiented ) { 559 row = im[i]*bs; 560 value = v + i*(stepval+bs)*bs; 561 for ( j=0; j<bs; j++,row++ ) { 562 for ( k=0; k<n; k++ ) { 563 for ( col=in[k]*bs,l=0; l<bs; l++,col++) { 564 ierr = StashValues_Private(&baij->stash,row,1,&col,value++,addv);CHKERRQ(ierr); 565 } 566 } 567 } 568 } else { 569 for ( j=0; j<n; j++ ) { 570 value = v + j*(stepval+bs)*bs + i*bs; 571 col = in[j]*bs; 572 for ( k=0; k<bs; k++,col++,value+=stepval) { 573 for ( row = im[i]*bs,l=0; l<bs; l++,row++) { 574 ierr = StashValues_Private(&baij->stash,row,1,&col,value++,addv);CHKERRQ(ierr); 575 } 576 } 577 } 578 } 579 } 580 } 581 } 582 #if defined(USE_PETSC_BOPT_g) 583 baij->ht_total_ct = total_ct; 584 baij->ht_insert_ct = insert_ct; 585 #endif 586 PetscFunctionReturn(0); 587 } 588 589 #undef __FUNC__ 590 #define __FUNC__ "MatGetValues_MPIBAIJ" 591 int MatGetValues_MPIBAIJ(Mat mat,int m,int *idxm,int n,int *idxn,Scalar *v) 592 { 593 Mat_MPIBAIJ *baij = (Mat_MPIBAIJ *) mat->data; 594 int bs=baij->bs,ierr,i,j, bsrstart = baij->rstart*bs, bsrend = baij->rend*bs; 595 int bscstart = baij->cstart*bs, bscend = baij->cend*bs,row,col; 596 597 PetscFunctionBegin; 598 for ( i=0; i<m; i++ ) { 599 if (idxm[i] < 0) SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,0,"Negative row"); 600 if (idxm[i] >= baij->M) SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,0,"Row too large"); 601 if (idxm[i] >= bsrstart && idxm[i] < bsrend) { 602 row = idxm[i] - bsrstart; 603 for ( j=0; j<n; j++ ) { 604 if (idxn[j] < 0) SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,0,"Negative column"); 605 if (idxn[j] >= baij->N) SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,0,"Column too large"); 606 if (idxn[j] >= bscstart && idxn[j] < bscend){ 607 col = idxn[j] - bscstart; 608 ierr = MatGetValues(baij->A,1,&row,1,&col,v+i*n+j); CHKERRQ(ierr); 609 } else { 610 if (!baij->colmap) { 611 ierr = CreateColmap_MPIBAIJ_Private(mat);CHKERRQ(ierr); 612 } 613 if((baij->colmap[idxn[j]/bs]-1 < 0) || 614 (baij->garray[(baij->colmap[idxn[j]/bs]-1)/bs] != idxn[j]/bs)) *(v+i*n+j) = 0.0; 615 else { 616 col = (baij->colmap[idxn[j]/bs]-1) + idxn[j]%bs; 617 ierr = MatGetValues(baij->B,1,&row,1,&col,v+i*n+j); CHKERRQ(ierr); 618 } 619 } 620 } 621 } else { 622 SETERRQ(PETSC_ERR_SUP,0,"Only local values currently supported"); 623 } 624 } 625 PetscFunctionReturn(0); 626 } 627 628 #undef __FUNC__ 629 #define __FUNC__ "MatNorm_MPIBAIJ" 630 int MatNorm_MPIBAIJ(Mat mat,NormType type,double *norm) 631 { 632 Mat_MPIBAIJ *baij = (Mat_MPIBAIJ *) mat->data; 633 Mat_SeqBAIJ *amat = (Mat_SeqBAIJ*) baij->A->data, *bmat = (Mat_SeqBAIJ*) baij->B->data; 634 int ierr, i,bs2=baij->bs2; 635 double sum = 0.0; 636 Scalar *v; 637 638 PetscFunctionBegin; 639 if (baij->size == 1) { 640 ierr = MatNorm(baij->A,type,norm); CHKERRQ(ierr); 641 } else { 642 if (type == NORM_FROBENIUS) { 643 v = amat->a; 644 for (i=0; i<amat->nz*bs2; i++ ) { 645 #if defined(USE_PETSC_COMPLEX) 646 sum += real(conj(*v)*(*v)); v++; 647 #else 648 sum += (*v)*(*v); v++; 649 #endif 650 } 651 v = bmat->a; 652 for (i=0; i<bmat->nz*bs2; i++ ) { 653 #if defined(USE_PETSC_COMPLEX) 654 sum += real(conj(*v)*(*v)); v++; 655 #else 656 sum += (*v)*(*v); v++; 657 #endif 658 } 659 ierr = MPI_Allreduce(&sum,norm,1,MPI_DOUBLE,MPI_SUM,mat->comm);CHKERRQ(ierr); 660 *norm = sqrt(*norm); 661 } else { 662 SETERRQ(PETSC_ERR_SUP,0,"No support for this norm yet"); 663 } 664 } 665 PetscFunctionReturn(0); 666 } 667 668 #undef __FUNC__ 669 #define __FUNC__ "MatAssemblyBegin_MPIBAIJ" 670 int MatAssemblyBegin_MPIBAIJ(Mat mat,MatAssemblyType mode) 671 { 672 Mat_MPIBAIJ *baij = (Mat_MPIBAIJ *) mat->data; 673 MPI_Comm comm = mat->comm; 674 int size = baij->size, *owners = baij->rowners,bs=baij->bs; 675 int rank = baij->rank,tag = mat->tag, *owner,*starts,count,ierr; 676 MPI_Request *send_waits,*recv_waits; 677 int *nprocs,i,j,idx,*procs,nsends,nreceives,nmax,*work; 678 InsertMode addv; 679 Scalar *rvalues,*svalues; 680 681 PetscFunctionBegin; 682 /* make sure all processors are either in INSERTMODE or ADDMODE */ 683 ierr = MPI_Allreduce(&mat->insertmode,&addv,1,MPI_INT,MPI_BOR,comm);CHKERRQ(ierr); 684 if (addv == (ADD_VALUES|INSERT_VALUES)) { 685 SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Some processors inserted others added"); 686 } 687 mat->insertmode = addv; /* in case this processor had no cache */ 688 689 /* first count number of contributors to each processor */ 690 nprocs = (int *) PetscMalloc( 2*size*sizeof(int) ); CHKPTRQ(nprocs); 691 PetscMemzero(nprocs,2*size*sizeof(int)); procs = nprocs + size; 692 owner = (int *) PetscMalloc( (baij->stash.n+1)*sizeof(int) ); CHKPTRQ(owner); 693 for ( i=0; i<baij->stash.n; i++ ) { 694 idx = baij->stash.idx[i]; 695 for ( j=0; j<size; j++ ) { 696 if (idx >= owners[j]*bs && idx < owners[j+1]*bs) { 697 nprocs[j]++; procs[j] = 1; owner[i] = j; break; 698 } 699 } 700 } 701 nsends = 0; for ( i=0; i<size; i++ ) { nsends += procs[i];} 702 703 /* inform other processors of number of messages and max length*/ 704 work = (int *) PetscMalloc( size*sizeof(int) ); CHKPTRQ(work); 705 ierr = MPI_Allreduce(procs, work,size,MPI_INT,MPI_SUM,comm);CHKERRQ(ierr); 706 nreceives = work[rank]; 707 ierr = MPI_Allreduce( nprocs, work,size,MPI_INT,MPI_MAX,comm);CHKERRQ(ierr); 708 nmax = work[rank]; 709 PetscFree(work); 710 711 /* post receives: 712 1) each message will consist of ordered pairs 713 (global index,value) we store the global index as a double 714 to simplify the message passing. 715 2) since we don't know how long each individual message is we 716 allocate the largest needed buffer for each receive. Potentially 717 this is a lot of wasted space. 718 719 720 This could be done better. 721 */ 722 rvalues = (Scalar *) PetscMalloc(3*(nreceives+1)*(nmax+1)*sizeof(Scalar));CHKPTRQ(rvalues); 723 recv_waits = (MPI_Request *) PetscMalloc((nreceives+1)*sizeof(MPI_Request));CHKPTRQ(recv_waits); 724 for ( i=0; i<nreceives; i++ ) { 725 ierr = MPI_Irecv(rvalues+3*nmax*i,3*nmax,MPIU_SCALAR,MPI_ANY_SOURCE,tag,comm,recv_waits+i);CHKERRQ(ierr); 726 } 727 728 /* do sends: 729 1) starts[i] gives the starting index in svalues for stuff going to 730 the ith processor 731 */ 732 svalues = (Scalar *) PetscMalloc(3*(baij->stash.n+1)*sizeof(Scalar));CHKPTRQ(svalues); 733 send_waits = (MPI_Request *) PetscMalloc((nsends+1)*sizeof(MPI_Request));CHKPTRQ(send_waits); 734 starts = (int *) PetscMalloc( size*sizeof(int) ); CHKPTRQ(starts); 735 starts[0] = 0; 736 for ( i=1; i<size; i++ ) { starts[i] = starts[i-1] + nprocs[i-1];} 737 for ( i=0; i<baij->stash.n; i++ ) { 738 svalues[3*starts[owner[i]]] = (Scalar) baij->stash.idx[i]; 739 svalues[3*starts[owner[i]]+1] = (Scalar) baij->stash.idy[i]; 740 svalues[3*(starts[owner[i]]++)+2] = baij->stash.array[i]; 741 } 742 PetscFree(owner); 743 starts[0] = 0; 744 for ( i=1; i<size; i++ ) { starts[i] = starts[i-1] + nprocs[i-1];} 745 count = 0; 746 for ( i=0; i<size; i++ ) { 747 if (procs[i]) { 748 ierr = MPI_Isend(svalues+3*starts[i],3*nprocs[i],MPIU_SCALAR,i,tag,comm,send_waits+count++);CHKERRQ(ierr); 749 } 750 } 751 PetscFree(starts); PetscFree(nprocs); 752 753 /* Free cache space */ 754 PLogInfo(0,"MatAssemblyBegin_MPIBAIJ:Number of off-processor values %d\n",baij->stash.n); 755 ierr = StashDestroy_Private(&baij->stash); CHKERRQ(ierr); 756 757 baij->svalues = svalues; baij->rvalues = rvalues; 758 baij->nsends = nsends; baij->nrecvs = nreceives; 759 baij->send_waits = send_waits; baij->recv_waits = recv_waits; 760 baij->rmax = nmax; 761 762 PetscFunctionReturn(0); 763 } 764 765 /* 766 Creates the hash table, and sets the table 767 This table is created only once. 768 If new entried need to be added to the matrix 769 then the hash table has to be destroyed and 770 recreated. 771 */ 772 #undef __FUNC__ 773 #define __FUNC__ "MatCreateHashTable_MPIBAIJ_Private" 774 int MatCreateHashTable_MPIBAIJ_Private(Mat mat,double factor) 775 { 776 Mat_MPIBAIJ *baij = (Mat_MPIBAIJ *) mat->data; 777 Mat A = baij->A, B=baij->B; 778 Mat_SeqBAIJ *a=(Mat_SeqBAIJ *)A->data, *b=(Mat_SeqBAIJ *)B->data; 779 int i,j,k,nz=a->nz+b->nz,h1,*ai=a->i,*aj=a->j,*bi=b->i,*bj=b->j; 780 int size,bs2=baij->bs2,rstart=baij->rstart; 781 int cstart=baij->cstart,*garray=baij->garray,row,col,Nbs=baij->Nbs; 782 int *HT,key; 783 Scalar **HD; 784 double tmp; 785 #if defined(USE_PETSC_BOPT_g) 786 int ct=0,max=0; 787 #endif 788 789 PetscFunctionBegin; 790 baij->ht_size=(int)(factor*nz); 791 size = baij->ht_size; 792 793 if (baij->ht) { 794 PetscFunctionReturn(0); 795 } 796 797 /* Allocate Memory for Hash Table */ 798 baij->hd = (Scalar**)PetscMalloc((size)*(sizeof(int)+sizeof(Scalar*))+1); CHKPTRQ(baij->hd); 799 baij->ht = (int*)(baij->hd + size); 800 HD = baij->hd; 801 HT = baij->ht; 802 803 804 PetscMemzero(HD,size*(sizeof(int)+sizeof(Scalar*))); 805 806 807 /* Loop Over A */ 808 for ( i=0; i<a->mbs; i++ ) { 809 for ( j=ai[i]; j<ai[i+1]; j++ ) { 810 row = i+rstart; 811 col = aj[j]+cstart; 812 813 key = row*Nbs + col + 1; 814 h1 = HASH(size,key,tmp); 815 for ( k=0; k<size; k++ ){ 816 if (HT[(h1+k)%size] == 0.0) { 817 HT[(h1+k)%size] = key; 818 HD[(h1+k)%size] = a->a + j*bs2; 819 break; 820 #if defined(USE_PETSC_BOPT_g) 821 } else { 822 ct++; 823 #endif 824 } 825 } 826 #if defined(USE_PETSC_BOPT_g) 827 if (k> max) max = k; 828 #endif 829 } 830 } 831 /* Loop Over B */ 832 for ( i=0; i<b->mbs; i++ ) { 833 for ( j=bi[i]; j<bi[i+1]; j++ ) { 834 row = i+rstart; 835 col = garray[bj[j]]; 836 key = row*Nbs + col + 1; 837 h1 = HASH(size,key,tmp); 838 for ( k=0; k<size; k++ ){ 839 if (HT[(h1+k)%size] == 0.0) { 840 HT[(h1+k)%size] = key; 841 HD[(h1+k)%size] = b->a + j*bs2; 842 break; 843 #if defined(USE_PETSC_BOPT_g) 844 } else { 845 ct++; 846 #endif 847 } 848 } 849 #if defined(USE_PETSC_BOPT_g) 850 if (k> max) max = k; 851 #endif 852 } 853 } 854 855 /* Print Summary */ 856 #if defined(USE_PETSC_BOPT_g) 857 for ( i=0,j=0; i<size; i++) 858 if (HT[i]) {j++;} 859 PLogInfo(0,"MatCreateHashTable_MPIBAIJ_Private: Average Search = %5.2f,max search = %d\n", 860 (j== 0)? 0.0:((double)(ct+j))/j,max); 861 #endif 862 PetscFunctionReturn(0); 863 } 864 865 #undef __FUNC__ 866 #define __FUNC__ "MatAssemblyEnd_MPIBAIJ" 867 int MatAssemblyEnd_MPIBAIJ(Mat mat,MatAssemblyType mode) 868 { 869 Mat_MPIBAIJ *baij = (Mat_MPIBAIJ *) mat->data; 870 MPI_Status *send_status,recv_status; 871 int imdex,nrecvs = baij->nrecvs, count = nrecvs, i, n, ierr; 872 int bs=baij->bs,row,col,other_disassembled,flg; 873 Scalar *values,val; 874 InsertMode addv = mat->insertmode; 875 876 PetscFunctionBegin; 877 /* wait on receives */ 878 while (count) { 879 ierr = MPI_Waitany(nrecvs,baij->recv_waits,&imdex,&recv_status);CHKERRQ(ierr); 880 /* unpack receives into our local space */ 881 values = baij->rvalues + 3*imdex*baij->rmax; 882 ierr = MPI_Get_count(&recv_status,MPIU_SCALAR,&n);CHKERRQ(ierr); 883 n = n/3; 884 for ( i=0; i<n; i++ ) { 885 row = (int) PetscReal(values[3*i]) - baij->rstart*bs; 886 col = (int) PetscReal(values[3*i+1]); 887 val = values[3*i+2]; 888 if (col >= baij->cstart*bs && col < baij->cend*bs) { 889 col -= baij->cstart*bs; 890 ierr = MatSetValues(baij->A,1,&row,1,&col,&val,addv); CHKERRQ(ierr) 891 } else { 892 if (mat->was_assembled) { 893 if (!baij->colmap) { 894 ierr = CreateColmap_MPIBAIJ_Private(mat); CHKERRQ(ierr); 895 } 896 col = (baij->colmap[col/bs]) - 1 + col%bs; 897 if (col < 0 && !((Mat_SeqBAIJ*)(baij->A->data))->nonew) { 898 ierr = DisAssemble_MPIBAIJ(mat); CHKERRQ(ierr); 899 col = (int) PetscReal(values[3*i+1]); 900 } 901 } 902 ierr = MatSetValues(baij->B,1,&row,1,&col,&val,addv); CHKERRQ(ierr) 903 } 904 } 905 count--; 906 } 907 PetscFree(baij->recv_waits); PetscFree(baij->rvalues); 908 909 /* wait on sends */ 910 if (baij->nsends) { 911 send_status = (MPI_Status *) PetscMalloc(baij->nsends*sizeof(MPI_Status));CHKPTRQ(send_status); 912 ierr = MPI_Waitall(baij->nsends,baij->send_waits,send_status);CHKERRQ(ierr); 913 PetscFree(send_status); 914 } 915 PetscFree(baij->send_waits); PetscFree(baij->svalues); 916 917 ierr = MatAssemblyBegin(baij->A,mode); CHKERRQ(ierr); 918 ierr = MatAssemblyEnd(baij->A,mode); CHKERRQ(ierr); 919 920 /* determine if any processor has disassembled, if so we must 921 also disassemble ourselfs, in order that we may reassemble. */ 922 ierr = MPI_Allreduce(&mat->was_assembled,&other_disassembled,1,MPI_INT,MPI_PROD,mat->comm);CHKERRQ(ierr); 923 if (mat->was_assembled && !other_disassembled) { 924 ierr = DisAssemble_MPIBAIJ(mat); CHKERRQ(ierr); 925 } 926 927 if (!mat->was_assembled && mode == MAT_FINAL_ASSEMBLY) { 928 ierr = MatSetUpMultiply_MPIBAIJ(mat); CHKERRQ(ierr); 929 } 930 ierr = MatAssemblyBegin(baij->B,mode); CHKERRQ(ierr); 931 ierr = MatAssemblyEnd(baij->B,mode); CHKERRQ(ierr); 932 933 #if defined(USE_PETSC_BOPT_g) 934 if (baij->ht && mode== MAT_FINAL_ASSEMBLY) { 935 PLogInfo(0,"MatAssemblyEnd_MPIBAIJ:Average Hash Table Search in MatSetValues = %5.2f\n", 936 ((double)baij->ht_total_ct)/baij->ht_insert_ct); 937 baij->ht_total_ct = 0; 938 baij->ht_insert_ct = 0; 939 } 940 #endif 941 if (baij->ht_flag && !baij->ht && mode == MAT_FINAL_ASSEMBLY) { 942 ierr = MatCreateHashTable_MPIBAIJ_Private(mat,baij->ht_fact); CHKERRQ(ierr); 943 mat->ops.setvalues = MatSetValues_MPIBAIJ_HT; 944 mat->ops.setvaluesblocked = MatSetValuesBlocked_MPIBAIJ_HT; 945 } 946 947 if (baij->rowvalues) {PetscFree(baij->rowvalues); baij->rowvalues = 0;} 948 PetscFunctionReturn(0); 949 } 950 951 #undef __FUNC__ 952 #define __FUNC__ "MatView_MPIBAIJ_Binary" 953 static int MatView_MPIBAIJ_Binary(Mat mat,Viewer viewer) 954 { 955 Mat_MPIBAIJ *baij = (Mat_MPIBAIJ *) mat->data; 956 int ierr; 957 958 PetscFunctionBegin; 959 if (baij->size == 1) { 960 ierr = MatView(baij->A,viewer); CHKERRQ(ierr); 961 } else SETERRQ(PETSC_ERR_SUP,0,"Only uniprocessor output supported"); 962 PetscFunctionReturn(0); 963 } 964 965 #undef __FUNC__ 966 #define __FUNC__ "MatView_MPIBAIJ_ASCIIorDraworMatlab" 967 static int MatView_MPIBAIJ_ASCIIorDraworMatlab(Mat mat,Viewer viewer) 968 { 969 Mat_MPIBAIJ *baij = (Mat_MPIBAIJ *) mat->data; 970 int ierr, format,rank,bs = baij->bs; 971 FILE *fd; 972 ViewerType vtype; 973 974 PetscFunctionBegin; 975 ierr = ViewerGetType(viewer,&vtype); CHKERRQ(ierr); 976 if (vtype == ASCII_FILES_VIEWER || vtype == ASCII_FILE_VIEWER) { 977 ierr = ViewerGetFormat(viewer,&format); 978 if (format == VIEWER_FORMAT_ASCII_INFO_LONG) { 979 MatInfo info; 980 MPI_Comm_rank(mat->comm,&rank); 981 ierr = ViewerASCIIGetPointer(viewer,&fd); CHKERRQ(ierr); 982 ierr = MatGetInfo(mat,MAT_LOCAL,&info); 983 PetscSequentialPhaseBegin(mat->comm,1); 984 fprintf(fd,"[%d] Local rows %d nz %d nz alloced %d bs %d mem %d\n", 985 rank,baij->m,(int)info.nz_used*bs,(int)info.nz_allocated*bs, 986 baij->bs,(int)info.memory); 987 ierr = MatGetInfo(baij->A,MAT_LOCAL,&info); 988 fprintf(fd,"[%d] on-diagonal part: nz %d \n",rank,(int)info.nz_used*bs); 989 ierr = MatGetInfo(baij->B,MAT_LOCAL,&info); 990 fprintf(fd,"[%d] off-diagonal part: nz %d \n",rank,(int)info.nz_used*bs); 991 fflush(fd); 992 PetscSequentialPhaseEnd(mat->comm,1); 993 ierr = VecScatterView(baij->Mvctx,viewer); CHKERRQ(ierr); 994 PetscFunctionReturn(0); 995 } else if (format == VIEWER_FORMAT_ASCII_INFO) { 996 PetscPrintf(mat->comm," block size is %d\n",bs); 997 PetscFunctionReturn(0); 998 } 999 } 1000 1001 if (vtype == DRAW_VIEWER) { 1002 Draw draw; 1003 PetscTruth isnull; 1004 ierr = ViewerDrawGetDraw(viewer,&draw); CHKERRQ(ierr); 1005 ierr = DrawIsNull(draw,&isnull); CHKERRQ(ierr); if (isnull) PetscFunctionReturn(0); 1006 } 1007 1008 if (vtype == ASCII_FILE_VIEWER) { 1009 ierr = ViewerASCIIGetPointer(viewer,&fd); CHKERRQ(ierr); 1010 PetscSequentialPhaseBegin(mat->comm,1); 1011 fprintf(fd,"[%d] rows %d starts %d ends %d cols %d starts %d ends %d\n", 1012 baij->rank,baij->m,baij->rstart*bs,baij->rend*bs,baij->n, 1013 baij->cstart*bs,baij->cend*bs); 1014 ierr = MatView(baij->A,viewer); CHKERRQ(ierr); 1015 ierr = MatView(baij->B,viewer); CHKERRQ(ierr); 1016 fflush(fd); 1017 PetscSequentialPhaseEnd(mat->comm,1); 1018 } else { 1019 int size = baij->size; 1020 rank = baij->rank; 1021 if (size == 1) { 1022 ierr = MatView(baij->A,viewer); CHKERRQ(ierr); 1023 } else { 1024 /* assemble the entire matrix onto first processor. */ 1025 Mat A; 1026 Mat_SeqBAIJ *Aloc; 1027 int M = baij->M, N = baij->N,*ai,*aj,row,col,i,j,k,*rvals; 1028 int mbs=baij->mbs; 1029 Scalar *a; 1030 1031 if (!rank) { 1032 ierr = MatCreateMPIBAIJ(mat->comm,baij->bs,M,N,M,N,0,PETSC_NULL,0,PETSC_NULL,&A);CHKERRQ(ierr); 1033 } else { 1034 ierr = MatCreateMPIBAIJ(mat->comm,baij->bs,0,0,M,N,0,PETSC_NULL,0,PETSC_NULL,&A);CHKERRQ(ierr); 1035 } 1036 PLogObjectParent(mat,A); 1037 1038 /* copy over the A part */ 1039 Aloc = (Mat_SeqBAIJ*) baij->A->data; 1040 ai = Aloc->i; aj = Aloc->j; a = Aloc->a; 1041 row = baij->rstart; 1042 rvals = (int *) PetscMalloc(bs*sizeof(int)); CHKPTRQ(rvals); 1043 1044 for ( i=0; i<mbs; i++ ) { 1045 rvals[0] = bs*(baij->rstart + i); 1046 for ( j=1; j<bs; j++ ) { rvals[j] = rvals[j-1] + 1; } 1047 for ( j=ai[i]; j<ai[i+1]; j++ ) { 1048 col = (baij->cstart+aj[j])*bs; 1049 for (k=0; k<bs; k++ ) { 1050 ierr = MatSetValues(A,bs,rvals,1,&col,a,INSERT_VALUES);CHKERRQ(ierr); 1051 col++; a += bs; 1052 } 1053 } 1054 } 1055 /* copy over the B part */ 1056 Aloc = (Mat_SeqBAIJ*) baij->B->data; 1057 ai = Aloc->i; aj = Aloc->j; a = Aloc->a; 1058 row = baij->rstart*bs; 1059 for ( i=0; i<mbs; i++ ) { 1060 rvals[0] = bs*(baij->rstart + i); 1061 for ( j=1; j<bs; j++ ) { rvals[j] = rvals[j-1] + 1; } 1062 for ( j=ai[i]; j<ai[i+1]; j++ ) { 1063 col = baij->garray[aj[j]]*bs; 1064 for (k=0; k<bs; k++ ) { 1065 ierr = MatSetValues(A,bs,rvals,1,&col,a,INSERT_VALUES);CHKERRQ(ierr); 1066 col++; a += bs; 1067 } 1068 } 1069 } 1070 PetscFree(rvals); 1071 ierr = MatAssemblyBegin(A,MAT_FINAL_ASSEMBLY); CHKERRQ(ierr); 1072 ierr = MatAssemblyEnd(A,MAT_FINAL_ASSEMBLY); CHKERRQ(ierr); 1073 /* 1074 Everyone has to call to draw the matrix since the graphics waits are 1075 synchronized across all processors that share the Draw object 1076 */ 1077 if (!rank || vtype == DRAW_VIEWER) { 1078 ierr = MatView(((Mat_MPIBAIJ*)(A->data))->A,viewer); CHKERRQ(ierr); 1079 } 1080 ierr = MatDestroy(A); CHKERRQ(ierr); 1081 } 1082 } 1083 PetscFunctionReturn(0); 1084 } 1085 1086 1087 1088 #undef __FUNC__ 1089 #define __FUNC__ "MatView_MPIBAIJ" 1090 int MatView_MPIBAIJ(PetscObject obj,Viewer viewer) 1091 { 1092 Mat mat = (Mat) obj; 1093 int ierr; 1094 ViewerType vtype; 1095 1096 PetscFunctionBegin; 1097 ierr = ViewerGetType(viewer,&vtype); CHKERRQ(ierr); 1098 if (vtype == ASCII_FILE_VIEWER || vtype == ASCII_FILES_VIEWER || 1099 vtype == DRAW_VIEWER || vtype == MATLAB_VIEWER) { 1100 ierr = MatView_MPIBAIJ_ASCIIorDraworMatlab(mat,viewer); CHKERRQ(ierr); 1101 } else if (vtype == BINARY_FILE_VIEWER) { 1102 ierr = MatView_MPIBAIJ_Binary(mat,viewer);CHKERRQ(ierr); 1103 } 1104 PetscFunctionReturn(0); 1105 } 1106 1107 #undef __FUNC__ 1108 #define __FUNC__ "MatDestroy_MPIBAIJ" 1109 int MatDestroy_MPIBAIJ(PetscObject obj) 1110 { 1111 Mat mat = (Mat) obj; 1112 Mat_MPIBAIJ *baij = (Mat_MPIBAIJ *) mat->data; 1113 int ierr; 1114 1115 PetscFunctionBegin; 1116 #if defined(USE_PETSC_LOG) 1117 PLogObjectState(obj,"Rows=%d, Cols=%d",baij->M,baij->N); 1118 #endif 1119 1120 ierr = StashDestroy_Private(&baij->stash); CHKERRQ(ierr); 1121 PetscFree(baij->rowners); 1122 ierr = MatDestroy(baij->A); CHKERRQ(ierr); 1123 ierr = MatDestroy(baij->B); CHKERRQ(ierr); 1124 if (baij->colmap) PetscFree(baij->colmap); 1125 if (baij->garray) PetscFree(baij->garray); 1126 if (baij->lvec) VecDestroy(baij->lvec); 1127 if (baij->Mvctx) VecScatterDestroy(baij->Mvctx); 1128 if (baij->rowvalues) PetscFree(baij->rowvalues); 1129 if (baij->barray) PetscFree(baij->barray); 1130 if (baij->hd) PetscFree(baij->hd); 1131 PetscFree(baij); 1132 PLogObjectDestroy(mat); 1133 PetscHeaderDestroy(mat); 1134 PetscFunctionReturn(0); 1135 } 1136 1137 #undef __FUNC__ 1138 #define __FUNC__ "MatMult_MPIBAIJ" 1139 int MatMult_MPIBAIJ(Mat A,Vec xx,Vec yy) 1140 { 1141 Mat_MPIBAIJ *a = (Mat_MPIBAIJ *) A->data; 1142 int ierr, nt; 1143 1144 PetscFunctionBegin; 1145 VecGetLocalSize_Fast(xx,nt); 1146 if (nt != a->n) { 1147 SETERRQ(PETSC_ERR_ARG_SIZ,0,"Incompatible partition of A and xx"); 1148 } 1149 VecGetLocalSize_Fast(yy,nt); 1150 if (nt != a->m) { 1151 SETERRQ(PETSC_ERR_ARG_SIZ,0,"Incompatible parition of A and yy"); 1152 } 1153 ierr = VecScatterBegin(xx,a->lvec,INSERT_VALUES,SCATTER_FORWARD,a->Mvctx);CHKERRQ(ierr); 1154 ierr = (*a->A->ops.mult)(a->A,xx,yy); CHKERRQ(ierr); 1155 ierr = VecScatterEnd(xx,a->lvec,INSERT_VALUES,SCATTER_FORWARD,a->Mvctx);CHKERRQ(ierr); 1156 ierr = (*a->B->ops.multadd)(a->B,a->lvec,yy,yy); CHKERRQ(ierr); 1157 ierr = VecScatterPostRecvs(xx,a->lvec,INSERT_VALUES,SCATTER_FORWARD,a->Mvctx);CHKERRQ(ierr); 1158 PetscFunctionReturn(0); 1159 } 1160 1161 #undef __FUNC__ 1162 #define __FUNC__ "MatMultAdd_MPIBAIJ" 1163 int MatMultAdd_MPIBAIJ(Mat A,Vec xx,Vec yy,Vec zz) 1164 { 1165 Mat_MPIBAIJ *a = (Mat_MPIBAIJ *) A->data; 1166 int ierr; 1167 1168 PetscFunctionBegin; 1169 ierr = VecScatterBegin(xx,a->lvec,INSERT_VALUES,SCATTER_FORWARD,a->Mvctx);CHKERRQ(ierr); 1170 ierr = (*a->A->ops.multadd)(a->A,xx,yy,zz); CHKERRQ(ierr); 1171 ierr = VecScatterEnd(xx,a->lvec,INSERT_VALUES,SCATTER_FORWARD,a->Mvctx);CHKERRQ(ierr); 1172 ierr = (*a->B->ops.multadd)(a->B,a->lvec,zz,zz); CHKERRQ(ierr); 1173 PetscFunctionReturn(0); 1174 } 1175 1176 #undef __FUNC__ 1177 #define __FUNC__ "MatMultTrans_MPIBAIJ" 1178 int MatMultTrans_MPIBAIJ(Mat A,Vec xx,Vec yy) 1179 { 1180 Mat_MPIBAIJ *a = (Mat_MPIBAIJ *) A->data; 1181 int ierr; 1182 1183 PetscFunctionBegin; 1184 /* do nondiagonal part */ 1185 ierr = (*a->B->ops.multtrans)(a->B,xx,a->lvec); CHKERRQ(ierr); 1186 /* send it on its way */ 1187 ierr = VecScatterBegin(a->lvec,yy,ADD_VALUES,SCATTER_REVERSE,a->Mvctx);CHKERRQ(ierr); 1188 /* do local part */ 1189 ierr = (*a->A->ops.multtrans)(a->A,xx,yy); CHKERRQ(ierr); 1190 /* receive remote parts: note this assumes the values are not actually */ 1191 /* inserted in yy until the next line, which is true for my implementation*/ 1192 /* but is not perhaps always true. */ 1193 ierr = VecScatterEnd(a->lvec,yy,ADD_VALUES,SCATTER_REVERSE,a->Mvctx);CHKERRQ(ierr); 1194 PetscFunctionReturn(0); 1195 } 1196 1197 #undef __FUNC__ 1198 #define __FUNC__ "MatMultTransAdd_MPIBAIJ" 1199 int MatMultTransAdd_MPIBAIJ(Mat A,Vec xx,Vec yy,Vec zz) 1200 { 1201 Mat_MPIBAIJ *a = (Mat_MPIBAIJ *) A->data; 1202 int ierr; 1203 1204 PetscFunctionBegin; 1205 /* do nondiagonal part */ 1206 ierr = (*a->B->ops.multtrans)(a->B,xx,a->lvec); CHKERRQ(ierr); 1207 /* send it on its way */ 1208 ierr = VecScatterBegin(a->lvec,zz,ADD_VALUES,SCATTER_REVERSE,a->Mvctx); CHKERRQ(ierr); 1209 /* do local part */ 1210 ierr = (*a->A->ops.multtransadd)(a->A,xx,yy,zz); CHKERRQ(ierr); 1211 /* receive remote parts: note this assumes the values are not actually */ 1212 /* inserted in yy until the next line, which is true for my implementation*/ 1213 /* but is not perhaps always true. */ 1214 ierr = VecScatterEnd(a->lvec,zz,ADD_VALUES,SCATTER_REVERSE,a->Mvctx); CHKERRQ(ierr); 1215 PetscFunctionReturn(0); 1216 } 1217 1218 /* 1219 This only works correctly for square matrices where the subblock A->A is the 1220 diagonal block 1221 */ 1222 #undef __FUNC__ 1223 #define __FUNC__ "MatGetDiagonal_MPIBAIJ" 1224 int MatGetDiagonal_MPIBAIJ(Mat A,Vec v) 1225 { 1226 Mat_MPIBAIJ *a = (Mat_MPIBAIJ *) A->data; 1227 int ierr; 1228 1229 PetscFunctionBegin; 1230 if (a->M != a->N) SETERRQ(PETSC_ERR_SUP,0,"Supports only square matrix where A->A is diag block"); 1231 ierr = MatGetDiagonal(a->A,v);CHKERRQ(ierr); 1232 PetscFunctionReturn(0); 1233 } 1234 1235 #undef __FUNC__ 1236 #define __FUNC__ "MatScale_MPIBAIJ" 1237 int MatScale_MPIBAIJ(Scalar *aa,Mat A) 1238 { 1239 Mat_MPIBAIJ *a = (Mat_MPIBAIJ *) A->data; 1240 int ierr; 1241 1242 PetscFunctionBegin; 1243 ierr = MatScale(aa,a->A); CHKERRQ(ierr); 1244 ierr = MatScale(aa,a->B); CHKERRQ(ierr); 1245 PetscFunctionReturn(0); 1246 } 1247 1248 #undef __FUNC__ 1249 #define __FUNC__ "MatGetSize_MPIBAIJ" 1250 int MatGetSize_MPIBAIJ(Mat matin,int *m,int *n) 1251 { 1252 Mat_MPIBAIJ *mat = (Mat_MPIBAIJ *) matin->data; 1253 1254 PetscFunctionBegin; 1255 if (m) *m = mat->M; 1256 if (n) *n = mat->N; 1257 PetscFunctionReturn(0); 1258 } 1259 1260 #undef __FUNC__ 1261 #define __FUNC__ "MatGetLocalSize_MPIBAIJ" 1262 int MatGetLocalSize_MPIBAIJ(Mat matin,int *m,int *n) 1263 { 1264 Mat_MPIBAIJ *mat = (Mat_MPIBAIJ *) matin->data; 1265 1266 PetscFunctionBegin; 1267 *m = mat->m; *n = mat->N; 1268 PetscFunctionReturn(0); 1269 } 1270 1271 #undef __FUNC__ 1272 #define __FUNC__ "MatGetOwnershipRange_MPIBAIJ" 1273 int MatGetOwnershipRange_MPIBAIJ(Mat matin,int *m,int *n) 1274 { 1275 Mat_MPIBAIJ *mat = (Mat_MPIBAIJ *) matin->data; 1276 1277 PetscFunctionBegin; 1278 *m = mat->rstart*mat->bs; *n = mat->rend*mat->bs; 1279 PetscFunctionReturn(0); 1280 } 1281 1282 extern int MatGetRow_SeqBAIJ(Mat,int,int*,int**,Scalar**); 1283 extern int MatRestoreRow_SeqBAIJ(Mat,int,int*,int**,Scalar**); 1284 1285 #undef __FUNC__ 1286 #define __FUNC__ "MatGetRow_MPIBAIJ" 1287 int MatGetRow_MPIBAIJ(Mat matin,int row,int *nz,int **idx,Scalar **v) 1288 { 1289 Mat_MPIBAIJ *mat = (Mat_MPIBAIJ *) matin->data; 1290 Scalar *vworkA, *vworkB, **pvA, **pvB,*v_p; 1291 int bs = mat->bs, bs2 = mat->bs2, i, ierr, *cworkA, *cworkB, **pcA, **pcB; 1292 int nztot, nzA, nzB, lrow, brstart = mat->rstart*bs, brend = mat->rend*bs; 1293 int *cmap, *idx_p,cstart = mat->cstart; 1294 1295 PetscFunctionBegin; 1296 if (mat->getrowactive == PETSC_TRUE) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"Already active"); 1297 mat->getrowactive = PETSC_TRUE; 1298 1299 if (!mat->rowvalues && (idx || v)) { 1300 /* 1301 allocate enough space to hold information from the longest row. 1302 */ 1303 Mat_SeqBAIJ *Aa = (Mat_SeqBAIJ *) mat->A->data,*Ba = (Mat_SeqBAIJ *) mat->B->data; 1304 int max = 1,mbs = mat->mbs,tmp; 1305 for ( i=0; i<mbs; i++ ) { 1306 tmp = Aa->i[i+1] - Aa->i[i] + Ba->i[i+1] - Ba->i[i]; 1307 if (max < tmp) { max = tmp; } 1308 } 1309 mat->rowvalues = (Scalar *) PetscMalloc( max*bs2*(sizeof(int)+sizeof(Scalar))); 1310 CHKPTRQ(mat->rowvalues); 1311 mat->rowindices = (int *) (mat->rowvalues + max*bs2); 1312 } 1313 1314 if (row < brstart || row >= brend) SETERRQ(PETSC_ERR_SUP,0,"Only local rows") 1315 lrow = row - brstart; 1316 1317 pvA = &vworkA; pcA = &cworkA; pvB = &vworkB; pcB = &cworkB; 1318 if (!v) {pvA = 0; pvB = 0;} 1319 if (!idx) {pcA = 0; if (!v) pcB = 0;} 1320 ierr = (*mat->A->ops.getrow)(mat->A,lrow,&nzA,pcA,pvA); CHKERRQ(ierr); 1321 ierr = (*mat->B->ops.getrow)(mat->B,lrow,&nzB,pcB,pvB); CHKERRQ(ierr); 1322 nztot = nzA + nzB; 1323 1324 cmap = mat->garray; 1325 if (v || idx) { 1326 if (nztot) { 1327 /* Sort by increasing column numbers, assuming A and B already sorted */ 1328 int imark = -1; 1329 if (v) { 1330 *v = v_p = mat->rowvalues; 1331 for ( i=0; i<nzB; i++ ) { 1332 if (cmap[cworkB[i]/bs] < cstart) v_p[i] = vworkB[i]; 1333 else break; 1334 } 1335 imark = i; 1336 for ( i=0; i<nzA; i++ ) v_p[imark+i] = vworkA[i]; 1337 for ( i=imark; i<nzB; i++ ) v_p[nzA+i] = vworkB[i]; 1338 } 1339 if (idx) { 1340 *idx = idx_p = mat->rowindices; 1341 if (imark > -1) { 1342 for ( i=0; i<imark; i++ ) { 1343 idx_p[i] = cmap[cworkB[i]/bs]*bs + cworkB[i]%bs; 1344 } 1345 } else { 1346 for ( i=0; i<nzB; i++ ) { 1347 if (cmap[cworkB[i]/bs] < cstart) 1348 idx_p[i] = cmap[cworkB[i]/bs]*bs + cworkB[i]%bs ; 1349 else break; 1350 } 1351 imark = i; 1352 } 1353 for ( i=0; i<nzA; i++ ) idx_p[imark+i] = cstart*bs + cworkA[i]; 1354 for ( i=imark; i<nzB; i++ ) idx_p[nzA+i] = cmap[cworkB[i]/bs]*bs + cworkB[i]%bs ; 1355 } 1356 } else { 1357 if (idx) *idx = 0; 1358 if (v) *v = 0; 1359 } 1360 } 1361 *nz = nztot; 1362 ierr = (*mat->A->ops.restorerow)(mat->A,lrow,&nzA,pcA,pvA); CHKERRQ(ierr); 1363 ierr = (*mat->B->ops.restorerow)(mat->B,lrow,&nzB,pcB,pvB); CHKERRQ(ierr); 1364 PetscFunctionReturn(0); 1365 } 1366 1367 #undef __FUNC__ 1368 #define __FUNC__ "MatRestoreRow_MPIBAIJ" 1369 int MatRestoreRow_MPIBAIJ(Mat mat,int row,int *nz,int **idx,Scalar **v) 1370 { 1371 Mat_MPIBAIJ *baij = (Mat_MPIBAIJ *) mat->data; 1372 1373 PetscFunctionBegin; 1374 if (baij->getrowactive == PETSC_FALSE) { 1375 SETERRQ(PETSC_ERR_ARG_WRONGSTATE,0,"MatGetRow not called"); 1376 } 1377 baij->getrowactive = PETSC_FALSE; 1378 PetscFunctionReturn(0); 1379 } 1380 1381 #undef __FUNC__ 1382 #define __FUNC__ "MatGetBlockSize_MPIBAIJ" 1383 int MatGetBlockSize_MPIBAIJ(Mat mat,int *bs) 1384 { 1385 Mat_MPIBAIJ *baij = (Mat_MPIBAIJ *) mat->data; 1386 1387 PetscFunctionBegin; 1388 *bs = baij->bs; 1389 PetscFunctionReturn(0); 1390 } 1391 1392 #undef __FUNC__ 1393 #define __FUNC__ "MatZeroEntries_MPIBAIJ" 1394 int MatZeroEntries_MPIBAIJ(Mat A) 1395 { 1396 Mat_MPIBAIJ *l = (Mat_MPIBAIJ *) A->data; 1397 int ierr; 1398 1399 PetscFunctionBegin; 1400 ierr = MatZeroEntries(l->A); CHKERRQ(ierr); 1401 ierr = MatZeroEntries(l->B); CHKERRQ(ierr); 1402 PetscFunctionReturn(0); 1403 } 1404 1405 #undef __FUNC__ 1406 #define __FUNC__ "MatGetInfo_MPIBAIJ" 1407 int MatGetInfo_MPIBAIJ(Mat matin,MatInfoType flag,MatInfo *info) 1408 { 1409 Mat_MPIBAIJ *a = (Mat_MPIBAIJ *) matin->data; 1410 Mat A = a->A, B = a->B; 1411 int ierr; 1412 double isend[5], irecv[5]; 1413 1414 PetscFunctionBegin; 1415 info->block_size = (double)a->bs; 1416 ierr = MatGetInfo(A,MAT_LOCAL,info); CHKERRQ(ierr); 1417 isend[0] = info->nz_used; isend[1] = info->nz_allocated; isend[2] = info->memory; 1418 ierr = MatGetInfo(B,MAT_LOCAL,info); CHKERRQ(ierr); 1419 isend[0] += info->nz_used; isend[1] += info->nz_allocated; isend[2] += info->memory; 1420 if (flag == MAT_LOCAL) { 1421 info->nz_used = isend[0]; 1422 info->nz_allocated = isend[1]; 1423 info->nz_unneeded = isend[2]; 1424 info->memory = isend[3]; 1425 info->mallocs = isend[4]; 1426 } else if (flag == MAT_GLOBAL_MAX) { 1427 ierr = MPI_Allreduce(isend,irecv,5,MPI_INT,MPI_MAX,matin->comm);CHKERRQ(ierr); 1428 info->nz_used = irecv[0]; 1429 info->nz_allocated = irecv[1]; 1430 info->nz_unneeded = irecv[2]; 1431 info->memory = irecv[3]; 1432 info->mallocs = irecv[4]; 1433 } else if (flag == MAT_GLOBAL_SUM) { 1434 ierr = MPI_Allreduce(isend,irecv,5,MPI_INT,MPI_SUM,matin->comm);CHKERRQ(ierr); 1435 info->nz_used = irecv[0]; 1436 info->nz_allocated = irecv[1]; 1437 info->nz_unneeded = irecv[2]; 1438 info->memory = irecv[3]; 1439 info->mallocs = irecv[4]; 1440 } 1441 info->rows_global = (double)a->M; 1442 info->columns_global = (double)a->N; 1443 info->rows_local = (double)a->m; 1444 info->columns_local = (double)a->N; 1445 info->fill_ratio_given = 0; /* no parallel LU/ILU/Cholesky */ 1446 info->fill_ratio_needed = 0; 1447 info->factor_mallocs = 0; 1448 PetscFunctionReturn(0); 1449 } 1450 1451 #undef __FUNC__ 1452 #define __FUNC__ "MatSetOption_MPIBAIJ" 1453 int MatSetOption_MPIBAIJ(Mat A,MatOption op) 1454 { 1455 Mat_MPIBAIJ *a = (Mat_MPIBAIJ *) A->data; 1456 1457 PetscFunctionBegin; 1458 if (op == MAT_NO_NEW_NONZERO_LOCATIONS || 1459 op == MAT_YES_NEW_NONZERO_LOCATIONS || 1460 op == MAT_COLUMNS_UNSORTED || 1461 op == MAT_COLUMNS_SORTED || 1462 op == MAT_NEW_NONZERO_ALLOCATION_ERROR || 1463 op == MAT_NEW_NONZERO_LOCATION_ERROR) { 1464 MatSetOption(a->A,op); 1465 MatSetOption(a->B,op); 1466 } else if (op == MAT_ROW_ORIENTED) { 1467 a->roworiented = 1; 1468 MatSetOption(a->A,op); 1469 MatSetOption(a->B,op); 1470 } else if (op == MAT_ROWS_SORTED || 1471 op == MAT_ROWS_UNSORTED || 1472 op == MAT_SYMMETRIC || 1473 op == MAT_STRUCTURALLY_SYMMETRIC || 1474 op == MAT_YES_NEW_DIAGONALS) 1475 PLogInfo(A,"Info:MatSetOption_MPIBAIJ:Option ignored\n"); 1476 else if (op == MAT_COLUMN_ORIENTED) { 1477 a->roworiented = 0; 1478 MatSetOption(a->A,op); 1479 MatSetOption(a->B,op); 1480 } else if (op == MAT_IGNORE_OFF_PROC_ENTRIES) { 1481 a->donotstash = 1; 1482 } else if (op == MAT_NO_NEW_DIAGONALS) { 1483 SETERRQ(PETSC_ERR_SUP,0,"MAT_NO_NEW_DIAGONALS"); 1484 } else if (op == MAT_USE_HASH_TABLE) { 1485 a->ht_flag = 1; 1486 } else { 1487 SETERRQ(PETSC_ERR_SUP,0,"unknown option"); 1488 } 1489 PetscFunctionReturn(0); 1490 } 1491 1492 #undef __FUNC__ 1493 #define __FUNC__ "MatTranspose_MPIBAIJ(" 1494 int MatTranspose_MPIBAIJ(Mat A,Mat *matout) 1495 { 1496 Mat_MPIBAIJ *baij = (Mat_MPIBAIJ *) A->data; 1497 Mat_SeqBAIJ *Aloc; 1498 Mat B; 1499 int ierr,M=baij->M,N=baij->N,*ai,*aj,row,i,*rvals,j,k,col; 1500 int bs=baij->bs,mbs=baij->mbs; 1501 Scalar *a; 1502 1503 PetscFunctionBegin; 1504 if (matout == PETSC_NULL && M != N) SETERRQ(PETSC_ERR_ARG_SIZ,0,"Square matrix only for in-place"); 1505 ierr = MatCreateMPIBAIJ(A->comm,baij->bs,PETSC_DECIDE,PETSC_DECIDE,N,M,0,PETSC_NULL,0,PETSC_NULL,&B); 1506 CHKERRQ(ierr); 1507 1508 /* copy over the A part */ 1509 Aloc = (Mat_SeqBAIJ*) baij->A->data; 1510 ai = Aloc->i; aj = Aloc->j; a = Aloc->a; 1511 row = baij->rstart; 1512 rvals = (int *) PetscMalloc(bs*sizeof(int)); CHKPTRQ(rvals); 1513 1514 for ( i=0; i<mbs; i++ ) { 1515 rvals[0] = bs*(baij->rstart + i); 1516 for ( j=1; j<bs; j++ ) { rvals[j] = rvals[j-1] + 1; } 1517 for ( j=ai[i]; j<ai[i+1]; j++ ) { 1518 col = (baij->cstart+aj[j])*bs; 1519 for (k=0; k<bs; k++ ) { 1520 ierr = MatSetValues(B,1,&col,bs,rvals,a,INSERT_VALUES);CHKERRQ(ierr); 1521 col++; a += bs; 1522 } 1523 } 1524 } 1525 /* copy over the B part */ 1526 Aloc = (Mat_SeqBAIJ*) baij->B->data; 1527 ai = Aloc->i; aj = Aloc->j; a = Aloc->a; 1528 row = baij->rstart*bs; 1529 for ( i=0; i<mbs; i++ ) { 1530 rvals[0] = bs*(baij->rstart + i); 1531 for ( j=1; j<bs; j++ ) { rvals[j] = rvals[j-1] + 1; } 1532 for ( j=ai[i]; j<ai[i+1]; j++ ) { 1533 col = baij->garray[aj[j]]*bs; 1534 for (k=0; k<bs; k++ ) { 1535 ierr = MatSetValues(B,1,&col,bs,rvals,a,INSERT_VALUES);CHKERRQ(ierr); 1536 col++; a += bs; 1537 } 1538 } 1539 } 1540 PetscFree(rvals); 1541 ierr = MatAssemblyBegin(B,MAT_FINAL_ASSEMBLY); CHKERRQ(ierr); 1542 ierr = MatAssemblyEnd(B,MAT_FINAL_ASSEMBLY); CHKERRQ(ierr); 1543 1544 if (matout != PETSC_NULL) { 1545 *matout = B; 1546 } else { 1547 /* This isn't really an in-place transpose .... but free data structures from baij */ 1548 PetscFree(baij->rowners); 1549 ierr = MatDestroy(baij->A); CHKERRQ(ierr); 1550 ierr = MatDestroy(baij->B); CHKERRQ(ierr); 1551 if (baij->colmap) PetscFree(baij->colmap); 1552 if (baij->garray) PetscFree(baij->garray); 1553 if (baij->lvec) VecDestroy(baij->lvec); 1554 if (baij->Mvctx) VecScatterDestroy(baij->Mvctx); 1555 PetscFree(baij); 1556 PetscMemcpy(A,B,sizeof(struct _p_Mat)); 1557 PetscHeaderDestroy(B); 1558 } 1559 PetscFunctionReturn(0); 1560 } 1561 1562 #undef __FUNC__ 1563 #define __FUNC__ "MatDiagonalScale_MPIBAIJ" 1564 int MatDiagonalScale_MPIBAIJ(Mat A,Vec ll,Vec rr) 1565 { 1566 Mat a = ((Mat_MPIBAIJ *) A->data)->A; 1567 Mat b = ((Mat_MPIBAIJ *) A->data)->B; 1568 int ierr,s1,s2,s3; 1569 1570 PetscFunctionBegin; 1571 if (ll) { 1572 ierr = VecGetLocalSize(ll,&s1); CHKERRQ(ierr); 1573 ierr = MatGetLocalSize(A,&s2,&s3); CHKERRQ(ierr); 1574 if (s1!=s2) SETERRQ(PETSC_ERR_ARG_SIZ,0,"non-conforming local sizes"); 1575 ierr = MatDiagonalScale(a,ll,0); CHKERRQ(ierr); 1576 ierr = MatDiagonalScale(b,ll,0); CHKERRQ(ierr); 1577 } 1578 if (rr) SETERRQ(PETSC_ERR_SUP,0,"not supported for right vector"); 1579 PetscFunctionReturn(0); 1580 } 1581 1582 #undef __FUNC__ 1583 #define __FUNC__ "MatZeroRows_MPIBAIJ" 1584 int MatZeroRows_MPIBAIJ(Mat A,IS is,Scalar *diag) 1585 { 1586 Mat_MPIBAIJ *l = (Mat_MPIBAIJ *) A->data; 1587 int i,ierr,N, *rows,*owners = l->rowners,size = l->size; 1588 int *procs,*nprocs,j,found,idx,nsends,*work,row; 1589 int nmax,*svalues,*starts,*owner,nrecvs,rank = l->rank; 1590 int *rvalues,tag = A->tag,count,base,slen,n,*source; 1591 int *lens,imdex,*lrows,*values,bs=l->bs,rstart_bs=l->rstart_bs; 1592 MPI_Comm comm = A->comm; 1593 MPI_Request *send_waits,*recv_waits; 1594 MPI_Status recv_status,*send_status; 1595 IS istmp; 1596 1597 PetscFunctionBegin; 1598 ierr = ISGetSize(is,&N); CHKERRQ(ierr); 1599 ierr = ISGetIndices(is,&rows); CHKERRQ(ierr); 1600 1601 /* first count number of contributors to each processor */ 1602 nprocs = (int *) PetscMalloc( 2*size*sizeof(int) ); CHKPTRQ(nprocs); 1603 PetscMemzero(nprocs,2*size*sizeof(int)); procs = nprocs + size; 1604 owner = (int *) PetscMalloc((N+1)*sizeof(int)); CHKPTRQ(owner); /* see note*/ 1605 for ( i=0; i<N; i++ ) { 1606 idx = rows[i]; 1607 found = 0; 1608 for ( j=0; j<size; j++ ) { 1609 if (idx >= owners[j]*bs && idx < owners[j+1]*bs) { 1610 nprocs[j]++; procs[j] = 1; owner[i] = j; found = 1; break; 1611 } 1612 } 1613 if (!found) SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,0,"Index out of range"); 1614 } 1615 nsends = 0; for ( i=0; i<size; i++ ) { nsends += procs[i];} 1616 1617 /* inform other processors of number of messages and max length*/ 1618 work = (int *) PetscMalloc( size*sizeof(int) ); CHKPTRQ(work); 1619 ierr = MPI_Allreduce( procs, work,size,MPI_INT,MPI_SUM,comm);CHKERRQ(ierr); 1620 nrecvs = work[rank]; 1621 ierr = MPI_Allreduce( nprocs, work,size,MPI_INT,MPI_MAX,comm);CHKERRQ(ierr); 1622 nmax = work[rank]; 1623 PetscFree(work); 1624 1625 /* post receives: */ 1626 rvalues = (int *) PetscMalloc((nrecvs+1)*(nmax+1)*sizeof(int)); CHKPTRQ(rvalues); 1627 recv_waits = (MPI_Request *) PetscMalloc((nrecvs+1)*sizeof(MPI_Request));CHKPTRQ(recv_waits); 1628 for ( i=0; i<nrecvs; i++ ) { 1629 ierr = MPI_Irecv(rvalues+nmax*i,nmax,MPI_INT,MPI_ANY_SOURCE,tag,comm,recv_waits+i);CHKERRQ(ierr); 1630 } 1631 1632 /* do sends: 1633 1) starts[i] gives the starting index in svalues for stuff going to 1634 the ith processor 1635 */ 1636 svalues = (int *) PetscMalloc( (N+1)*sizeof(int) ); CHKPTRQ(svalues); 1637 send_waits = (MPI_Request *) PetscMalloc( (nsends+1)*sizeof(MPI_Request));CHKPTRQ(send_waits); 1638 starts = (int *) PetscMalloc( (size+1)*sizeof(int) ); CHKPTRQ(starts); 1639 starts[0] = 0; 1640 for ( i=1; i<size; i++ ) { starts[i] = starts[i-1] + nprocs[i-1];} 1641 for ( i=0; i<N; i++ ) { 1642 svalues[starts[owner[i]]++] = rows[i]; 1643 } 1644 ISRestoreIndices(is,&rows); 1645 1646 starts[0] = 0; 1647 for ( i=1; i<size+1; i++ ) { starts[i] = starts[i-1] + nprocs[i-1];} 1648 count = 0; 1649 for ( i=0; i<size; i++ ) { 1650 if (procs[i]) { 1651 ierr = MPI_Isend(svalues+starts[i],nprocs[i],MPI_INT,i,tag,comm,send_waits+count++);CHKERRQ(ierr); 1652 } 1653 } 1654 PetscFree(starts); 1655 1656 base = owners[rank]*bs; 1657 1658 /* wait on receives */ 1659 lens = (int *) PetscMalloc( 2*(nrecvs+1)*sizeof(int) ); CHKPTRQ(lens); 1660 source = lens + nrecvs; 1661 count = nrecvs; slen = 0; 1662 while (count) { 1663 ierr = MPI_Waitany(nrecvs,recv_waits,&imdex,&recv_status);CHKERRQ(ierr); 1664 /* unpack receives into our local space */ 1665 ierr = MPI_Get_count(&recv_status,MPI_INT,&n);CHKERRQ(ierr); 1666 source[imdex] = recv_status.MPI_SOURCE; 1667 lens[imdex] = n; 1668 slen += n; 1669 count--; 1670 } 1671 PetscFree(recv_waits); 1672 1673 /* move the data into the send scatter */ 1674 lrows = (int *) PetscMalloc( (slen+1)*sizeof(int) ); CHKPTRQ(lrows); 1675 count = 0; 1676 for ( i=0; i<nrecvs; i++ ) { 1677 values = rvalues + i*nmax; 1678 for ( j=0; j<lens[i]; j++ ) { 1679 lrows[count++] = values[j] - base; 1680 } 1681 } 1682 PetscFree(rvalues); PetscFree(lens); 1683 PetscFree(owner); PetscFree(nprocs); 1684 1685 /* actually zap the local rows */ 1686 ierr = ISCreateGeneral(PETSC_COMM_SELF,slen,lrows,&istmp);CHKERRQ(ierr); 1687 PLogObjectParent(A,istmp); 1688 1689 ierr = MatZeroRows(l->A,istmp,0); CHKERRQ(ierr); 1690 ierr = MatZeroRows(l->B,istmp,0); CHKERRQ(ierr); 1691 ierr = ISDestroy(istmp); CHKERRQ(ierr); 1692 1693 if (diag) { 1694 for ( i = 0; i < slen; i++ ) { 1695 row = lrows[i] + rstart_bs; 1696 ierr = MatSetValues(A,1,&row,1,&row,diag,INSERT_VALUES); CHKERRQ(ierr); 1697 } 1698 ierr = MatAssemblyBegin(A,MAT_FINAL_ASSEMBLY); CHKERRQ(ierr); 1699 ierr = MatAssemblyEnd(A,MAT_FINAL_ASSEMBLY); CHKERRQ(ierr); 1700 } 1701 PetscFree(lrows); 1702 1703 /* wait on sends */ 1704 if (nsends) { 1705 send_status = (MPI_Status *) PetscMalloc(nsends*sizeof(MPI_Status));CHKPTRQ(send_status); 1706 ierr = MPI_Waitall(nsends,send_waits,send_status);CHKERRQ(ierr); 1707 PetscFree(send_status); 1708 } 1709 PetscFree(send_waits); PetscFree(svalues); 1710 1711 PetscFunctionReturn(0); 1712 } 1713 extern int MatPrintHelp_SeqBAIJ(Mat); 1714 #undef __FUNC__ 1715 #define __FUNC__ "MatPrintHelp_MPIBAIJ" 1716 int MatPrintHelp_MPIBAIJ(Mat A) 1717 { 1718 Mat_MPIBAIJ *a = (Mat_MPIBAIJ*) A->data; 1719 MPI_Comm comm = A->comm; 1720 static int called = 0; 1721 int ierr; 1722 1723 PetscFunctionBegin; 1724 if (!a->rank) { 1725 ierr = MatPrintHelp_SeqBAIJ(a->A);CHKERRQ(ierr); 1726 } 1727 if (called) {PetscFunctionReturn(0);} else called = 1; 1728 (*PetscHelpPrintf)(comm," Options for MATMPIBAIJ matrix format (the defaults):\n"); 1729 (*PetscHelpPrintf)(comm," -mat_use_hash_table <factor>: Use hashtable for efficient matrix assembly\n"); 1730 PetscFunctionReturn(0); 1731 } 1732 1733 #undef __FUNC__ 1734 #define __FUNC__ "MatSetUnfactored_MPIBAIJ" 1735 int MatSetUnfactored_MPIBAIJ(Mat A) 1736 { 1737 Mat_MPIBAIJ *a = (Mat_MPIBAIJ*) A->data; 1738 int ierr; 1739 1740 PetscFunctionBegin; 1741 ierr = MatSetUnfactored(a->A); CHKERRQ(ierr); 1742 PetscFunctionReturn(0); 1743 } 1744 1745 static int MatConvertSameType_MPIBAIJ(Mat,Mat *,int); 1746 1747 /* -------------------------------------------------------------------*/ 1748 static struct _MatOps MatOps = { 1749 MatSetValues_MPIBAIJ,MatGetRow_MPIBAIJ,MatRestoreRow_MPIBAIJ,MatMult_MPIBAIJ, 1750 MatMultAdd_MPIBAIJ,MatMultTrans_MPIBAIJ,MatMultTransAdd_MPIBAIJ,0, 1751 0,0,0,0, 1752 0,0,MatTranspose_MPIBAIJ,MatGetInfo_MPIBAIJ, 1753 0,MatGetDiagonal_MPIBAIJ,MatDiagonalScale_MPIBAIJ,MatNorm_MPIBAIJ, 1754 MatAssemblyBegin_MPIBAIJ,MatAssemblyEnd_MPIBAIJ,0,MatSetOption_MPIBAIJ, 1755 MatZeroEntries_MPIBAIJ,MatZeroRows_MPIBAIJ,0, 1756 0,0,0,MatGetSize_MPIBAIJ, 1757 MatGetLocalSize_MPIBAIJ,MatGetOwnershipRange_MPIBAIJ,0,0, 1758 0,0,MatConvertSameType_MPIBAIJ,0,0, 1759 0,0,0,MatGetSubMatrices_MPIBAIJ, 1760 MatIncreaseOverlap_MPIBAIJ,MatGetValues_MPIBAIJ,0,MatPrintHelp_MPIBAIJ, 1761 MatScale_MPIBAIJ,0,0,0,MatGetBlockSize_MPIBAIJ, 1762 0,0,0,0,0,0,MatSetUnfactored_MPIBAIJ,0,MatSetValuesBlocked_MPIBAIJ}; 1763 1764 1765 #undef __FUNC__ 1766 #define __FUNC__ "MatCreateMPIBAIJ" 1767 /*@C 1768 MatCreateMPIBAIJ - Creates a sparse parallel matrix in block AIJ format 1769 (block compressed row). For good matrix assembly performance 1770 the user should preallocate the matrix storage by setting the parameters 1771 d_nz (or d_nnz) and o_nz (or o_nnz). By setting these parameters accurately, 1772 performance can be increased by more than a factor of 50. 1773 1774 Input Parameters: 1775 . comm - MPI communicator 1776 . bs - size of blockk 1777 . m - number of local rows (or PETSC_DECIDE to have calculated if M is given) 1778 This value should be the same as the local size used in creating the 1779 y vector for the matrix-vector product y = Ax. 1780 . n - number of local columns (or PETSC_DECIDE to have calculated if N is given) 1781 This value should be the same as the local size used in creating the 1782 x vector for the matrix-vector product y = Ax. 1783 . M - number of global rows (or PETSC_DECIDE to have calculated if m is given) 1784 . N - number of global columns (or PETSC_DECIDE to have calculated if n is given) 1785 . d_nz - number of block nonzeros per block row in diagonal portion of local 1786 submatrix (same for all local rows) 1787 . d_nzz - array containing the number of block nonzeros in the various block rows 1788 of the in diagonal portion of the local (possibly different for each block 1789 row) or PETSC_NULL. You must leave room for the diagonal entry even if 1790 it is zero. 1791 . o_nz - number of block nonzeros per block row in the off-diagonal portion of local 1792 submatrix (same for all local rows). 1793 . o_nzz - array containing the number of nonzeros in the various block rows of the 1794 off-diagonal portion of the local submatrix (possibly different for 1795 each block row) or PETSC_NULL. 1796 1797 Output Parameter: 1798 . A - the matrix 1799 1800 Notes: 1801 The user MUST specify either the local or global matrix dimensions 1802 (possibly both). 1803 1804 Storage Information: 1805 For a square global matrix we define each processor's diagonal portion 1806 to be its local rows and the corresponding columns (a square submatrix); 1807 each processor's off-diagonal portion encompasses the remainder of the 1808 local matrix (a rectangular submatrix). 1809 1810 The user can specify preallocated storage for the diagonal part of 1811 the local submatrix with either d_nz or d_nnz (not both). Set 1812 d_nz=PETSC_DEFAULT and d_nnz=PETSC_NULL for PETSc to control dynamic 1813 memory allocation. Likewise, specify preallocated storage for the 1814 off-diagonal part of the local submatrix with o_nz or o_nnz (not both). 1815 1816 Consider a processor that owns rows 3, 4 and 5 of a parallel matrix. In 1817 the figure below we depict these three local rows and all columns (0-11). 1818 1819 $ 0 1 2 3 4 5 6 7 8 9 10 11 1820 $ ------------------- 1821 $ row 3 | o o o d d d o o o o o o 1822 $ row 4 | o o o d d d o o o o o o 1823 $ row 5 | o o o d d d o o o o o o 1824 $ ------------------- 1825 $ 1826 1827 Thus, any entries in the d locations are stored in the d (diagonal) 1828 submatrix, and any entries in the o locations are stored in the 1829 o (off-diagonal) submatrix. Note that the d and the o submatrices are 1830 stored simply in the MATSEQBAIJ format for compressed row storage. 1831 1832 Now d_nz should indicate the number of block nonzeros per row in the d matrix, 1833 and o_nz should indicate the number of block nonzeros per row in the o matrix. 1834 In general, for PDE problems in which most nonzeros are near the diagonal, 1835 one expects d_nz >> o_nz. For large problems you MUST preallocate memory 1836 or you will get TERRIBLE performance; see the users' manual chapter on 1837 matrices. 1838 1839 .keywords: matrix, block, aij, compressed row, sparse, parallel 1840 1841 .seealso: MatCreate(), MatCreateSeqBAIJ(), MatSetValues() 1842 @*/ 1843 int MatCreateMPIBAIJ(MPI_Comm comm,int bs,int m,int n,int M,int N, 1844 int d_nz,int *d_nnz,int o_nz,int *o_nnz,Mat *A) 1845 { 1846 Mat B; 1847 Mat_MPIBAIJ *b; 1848 int ierr, i,sum[2],work[2],mbs,nbs,Mbs=PETSC_DECIDE,Nbs=PETSC_DECIDE,size,flg; 1849 1850 PetscFunctionBegin; 1851 if (bs < 1) SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,0,"Invalid block size specified, must be positive"); 1852 1853 MPI_Comm_size(comm,&size); 1854 if (size == 1) { 1855 if (M == PETSC_DECIDE) M = m; 1856 if (N == PETSC_DECIDE) N = n; 1857 ierr = MatCreateSeqBAIJ(comm,bs,M,N,d_nz,d_nnz,A); CHKERRQ(ierr); 1858 PetscFunctionReturn(0); 1859 } 1860 1861 *A = 0; 1862 PetscHeaderCreate(B,_p_Mat,MAT_COOKIE,MATMPIBAIJ,comm,MatDestroy,MatView); 1863 PLogObjectCreate(B); 1864 B->data = (void *) (b = PetscNew(Mat_MPIBAIJ)); CHKPTRQ(b); 1865 PetscMemzero(b,sizeof(Mat_MPIBAIJ)); 1866 PetscMemcpy(&B->ops,&MatOps,sizeof(struct _MatOps)); 1867 1868 B->destroy = MatDestroy_MPIBAIJ; 1869 B->view = MatView_MPIBAIJ; 1870 B->mapping = 0; 1871 B->factor = 0; 1872 B->assembled = PETSC_FALSE; 1873 1874 B->insertmode = NOT_SET_VALUES; 1875 MPI_Comm_rank(comm,&b->rank); 1876 MPI_Comm_size(comm,&b->size); 1877 1878 if ( m == PETSC_DECIDE && (d_nnz != PETSC_NULL || o_nnz != PETSC_NULL)) { 1879 SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,0,"Cannot have PETSC_DECIDE rows but set d_nnz or o_nnz"); 1880 } 1881 if ( M == PETSC_DECIDE && m == PETSC_DECIDE) { 1882 SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,0,"either M or m should be specified"); 1883 } 1884 if ( N == PETSC_DECIDE && n == PETSC_DECIDE) { 1885 SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,0,"either N or n should be specified"); 1886 } 1887 if ( M != PETSC_DECIDE && m != PETSC_DECIDE) M = PETSC_DECIDE; 1888 if ( N != PETSC_DECIDE && n != PETSC_DECIDE) N = PETSC_DECIDE; 1889 1890 if (M == PETSC_DECIDE || N == PETSC_DECIDE) { 1891 work[0] = m; work[1] = n; 1892 mbs = m/bs; nbs = n/bs; 1893 ierr = MPI_Allreduce( work, sum,2,MPI_INT,MPI_SUM,comm );CHKERRQ(ierr); 1894 if (M == PETSC_DECIDE) {M = sum[0]; Mbs = M/bs;} 1895 if (N == PETSC_DECIDE) {N = sum[1]; Nbs = N/bs;} 1896 } 1897 if (m == PETSC_DECIDE) { 1898 Mbs = M/bs; 1899 if (Mbs*bs != M) SETERRQ(PETSC_ERR_ARG_SIZ,0,"No of global rows must be divisible by blocksize"); 1900 mbs = Mbs/b->size + ((Mbs % b->size) > b->rank); 1901 m = mbs*bs; 1902 } 1903 if (n == PETSC_DECIDE) { 1904 Nbs = N/bs; 1905 if (Nbs*bs != N) SETERRQ(PETSC_ERR_ARG_SIZ,0,"No of global cols must be divisible by blocksize"); 1906 nbs = Nbs/b->size + ((Nbs % b->size) > b->rank); 1907 n = nbs*bs; 1908 } 1909 if (mbs*bs != m || nbs*bs != n) { 1910 SETERRQ(PETSC_ERR_ARG_SIZ,0,"No of local rows, cols must be divisible by blocksize"); 1911 } 1912 1913 b->m = m; B->m = m; 1914 b->n = n; B->n = n; 1915 b->N = N; B->N = N; 1916 b->M = M; B->M = M; 1917 b->bs = bs; 1918 b->bs2 = bs*bs; 1919 b->mbs = mbs; 1920 b->nbs = nbs; 1921 b->Mbs = Mbs; 1922 b->Nbs = Nbs; 1923 1924 /* build local table of row and column ownerships */ 1925 b->rowners = (int *) PetscMalloc(2*(b->size+2)*sizeof(int)); CHKPTRQ(b->rowners); 1926 PLogObjectMemory(B,2*(b->size+2)*sizeof(int)+sizeof(struct _p_Mat)+sizeof(Mat_MPIBAIJ)); 1927 b->cowners = b->rowners + b->size + 2; 1928 ierr = MPI_Allgather(&mbs,1,MPI_INT,b->rowners+1,1,MPI_INT,comm);CHKERRQ(ierr); 1929 b->rowners[0] = 0; 1930 for ( i=2; i<=b->size; i++ ) { 1931 b->rowners[i] += b->rowners[i-1]; 1932 } 1933 b->rstart = b->rowners[b->rank]; 1934 b->rend = b->rowners[b->rank+1]; 1935 b->rstart_bs = b->rstart * bs; 1936 b->rend_bs = b->rend * bs; 1937 1938 ierr = MPI_Allgather(&nbs,1,MPI_INT,b->cowners+1,1,MPI_INT,comm);CHKERRQ(ierr); 1939 b->cowners[0] = 0; 1940 for ( i=2; i<=b->size; i++ ) { 1941 b->cowners[i] += b->cowners[i-1]; 1942 } 1943 b->cstart = b->cowners[b->rank]; 1944 b->cend = b->cowners[b->rank+1]; 1945 b->cstart_bs = b->cstart * bs; 1946 b->cend_bs = b->cend * bs; 1947 1948 1949 if (d_nz == PETSC_DEFAULT) d_nz = 5; 1950 ierr = MatCreateSeqBAIJ(PETSC_COMM_SELF,bs,m,n,d_nz,d_nnz,&b->A); CHKERRQ(ierr); 1951 PLogObjectParent(B,b->A); 1952 if (o_nz == PETSC_DEFAULT) o_nz = 0; 1953 ierr = MatCreateSeqBAIJ(PETSC_COMM_SELF,bs,m,N,o_nz,o_nnz,&b->B); CHKERRQ(ierr); 1954 PLogObjectParent(B,b->B); 1955 1956 /* build cache for off array entries formed */ 1957 ierr = StashBuild_Private(&b->stash); CHKERRQ(ierr); 1958 b->donotstash = 0; 1959 b->colmap = 0; 1960 b->garray = 0; 1961 b->roworiented = 1; 1962 1963 /* stuff used in block assembly */ 1964 b->barray = 0; 1965 1966 /* stuff used for matrix vector multiply */ 1967 b->lvec = 0; 1968 b->Mvctx = 0; 1969 1970 /* stuff for MatGetRow() */ 1971 b->rowindices = 0; 1972 b->rowvalues = 0; 1973 b->getrowactive = PETSC_FALSE; 1974 1975 /* hash table stuff */ 1976 b->ht = 0; 1977 b->hd = 0; 1978 b->ht_size = 0; 1979 b->ht_flag = 0; 1980 b->ht_fact = 0; 1981 b->ht_total_ct = 0; 1982 b->ht_insert_ct = 0; 1983 1984 *A = B; 1985 ierr = OptionsHasName(PETSC_NULL,"-mat_use_hash_table",&flg); CHKERRQ(ierr); 1986 if (flg) { 1987 double fact = 1.39; 1988 ierr = MatSetOption(B,MAT_USE_HASH_TABLE); CHKERRQ(ierr); 1989 ierr = OptionsGetDouble(PETSC_NULL,"-mat_use_hash_table",&fact,&flg); CHKERRQ(ierr); 1990 if (fact <= 1.0) fact = 1.39; 1991 ierr = MatMPIBAIJSetHashTableFactor(B,fact); CHKERRQ(ierr); 1992 PLogInfo(0,"MatCreateMPIBAIJ:Hash table Factor used %5.2f\n",fact); 1993 } 1994 PetscFunctionReturn(0); 1995 } 1996 1997 #undef __FUNC__ 1998 #define __FUNC__ "MatConvertSameType_MPIBAIJ" 1999 static int MatConvertSameType_MPIBAIJ(Mat matin,Mat *newmat,int cpvalues) 2000 { 2001 Mat mat; 2002 Mat_MPIBAIJ *a,*oldmat = (Mat_MPIBAIJ *) matin->data; 2003 int ierr, len=0, flg; 2004 2005 PetscFunctionBegin; 2006 *newmat = 0; 2007 PetscHeaderCreate(mat,_p_Mat,MAT_COOKIE,MATMPIBAIJ,matin->comm,MatDestroy,MatView); 2008 PLogObjectCreate(mat); 2009 mat->data = (void *) (a = PetscNew(Mat_MPIBAIJ)); CHKPTRQ(a); 2010 PetscMemcpy(&mat->ops,&MatOps,sizeof(struct _MatOps)); 2011 mat->destroy = MatDestroy_MPIBAIJ; 2012 mat->view = MatView_MPIBAIJ; 2013 mat->factor = matin->factor; 2014 mat->assembled = PETSC_TRUE; 2015 2016 a->m = mat->m = oldmat->m; 2017 a->n = mat->n = oldmat->n; 2018 a->M = mat->M = oldmat->M; 2019 a->N = mat->N = oldmat->N; 2020 2021 a->bs = oldmat->bs; 2022 a->bs2 = oldmat->bs2; 2023 a->mbs = oldmat->mbs; 2024 a->nbs = oldmat->nbs; 2025 a->Mbs = oldmat->Mbs; 2026 a->Nbs = oldmat->Nbs; 2027 2028 a->rstart = oldmat->rstart; 2029 a->rend = oldmat->rend; 2030 a->cstart = oldmat->cstart; 2031 a->cend = oldmat->cend; 2032 a->size = oldmat->size; 2033 a->rank = oldmat->rank; 2034 mat->insertmode = NOT_SET_VALUES; 2035 a->rowvalues = 0; 2036 a->getrowactive = PETSC_FALSE; 2037 a->barray = 0; 2038 2039 /* hash table stuff */ 2040 a->ht = 0; 2041 a->hd = 0; 2042 a->ht_size = 0; 2043 a->ht_flag = oldmat->ht_flag; 2044 a->ht_fact = oldmat->ht_fact; 2045 a->ht_total_ct = 0; 2046 a->ht_insert_ct = 0; 2047 2048 2049 a->rowners = (int *) PetscMalloc(2*(a->size+2)*sizeof(int)); CHKPTRQ(a->rowners); 2050 PLogObjectMemory(mat,2*(a->size+2)*sizeof(int)+sizeof(struct _p_Mat)+sizeof(Mat_MPIBAIJ)); 2051 a->cowners = a->rowners + a->size + 2; 2052 PetscMemcpy(a->rowners,oldmat->rowners,2*(a->size+2)*sizeof(int)); 2053 ierr = StashInitialize_Private(&a->stash); CHKERRQ(ierr); 2054 if (oldmat->colmap) { 2055 a->colmap = (int *) PetscMalloc((a->Nbs)*sizeof(int));CHKPTRQ(a->colmap); 2056 PLogObjectMemory(mat,(a->Nbs)*sizeof(int)); 2057 PetscMemcpy(a->colmap,oldmat->colmap,(a->Nbs)*sizeof(int)); 2058 } else a->colmap = 0; 2059 if (oldmat->garray && (len = ((Mat_SeqBAIJ *) (oldmat->B->data))->nbs)) { 2060 a->garray = (int *) PetscMalloc(len*sizeof(int)); CHKPTRQ(a->garray); 2061 PLogObjectMemory(mat,len*sizeof(int)); 2062 PetscMemcpy(a->garray,oldmat->garray,len*sizeof(int)); 2063 } else a->garray = 0; 2064 2065 ierr = VecDuplicate(oldmat->lvec,&a->lvec); CHKERRQ(ierr); 2066 PLogObjectParent(mat,a->lvec); 2067 ierr = VecScatterCopy(oldmat->Mvctx,&a->Mvctx); CHKERRQ(ierr); 2068 PLogObjectParent(mat,a->Mvctx); 2069 ierr = MatConvert(oldmat->A,MATSAME,&a->A); CHKERRQ(ierr); 2070 PLogObjectParent(mat,a->A); 2071 ierr = MatConvert(oldmat->B,MATSAME,&a->B); CHKERRQ(ierr); 2072 PLogObjectParent(mat,a->B); 2073 ierr = OptionsHasName(PETSC_NULL,"-help",&flg); CHKERRQ(ierr); 2074 if (flg) { 2075 ierr = MatPrintHelp(mat); CHKERRQ(ierr); 2076 } 2077 *newmat = mat; 2078 PetscFunctionReturn(0); 2079 } 2080 2081 #include "sys.h" 2082 2083 #undef __FUNC__ 2084 #define __FUNC__ "MatLoad_MPIBAIJ" 2085 int MatLoad_MPIBAIJ(Viewer viewer,MatType type,Mat *newmat) 2086 { 2087 Mat A; 2088 int i, nz, ierr, j,rstart, rend, fd; 2089 Scalar *vals,*buf; 2090 MPI_Comm comm = ((PetscObject)viewer)->comm; 2091 MPI_Status status; 2092 int header[4],rank,size,*rowlengths = 0,M,N,m,*rowners,*browners,maxnz,*cols; 2093 int *locrowlens,*sndcounts = 0,*procsnz = 0, jj,*mycols,*ibuf; 2094 int flg,tag = ((PetscObject)viewer)->tag,bs=1,bs2,Mbs,mbs,extra_rows; 2095 int *dlens,*odlens,*mask,*masked1,*masked2,rowcount,odcount; 2096 int dcount,kmax,k,nzcount,tmp; 2097 2098 PetscFunctionBegin; 2099 ierr = OptionsGetInt(PETSC_NULL,"-matload_block_size",&bs,&flg);CHKERRQ(ierr); 2100 bs2 = bs*bs; 2101 2102 MPI_Comm_size(comm,&size); MPI_Comm_rank(comm,&rank); 2103 if (!rank) { 2104 ierr = ViewerBinaryGetDescriptor(viewer,&fd); CHKERRQ(ierr); 2105 ierr = PetscBinaryRead(fd,(char *)header,4,PETSC_INT); CHKERRQ(ierr); 2106 if (header[0] != MAT_COOKIE) SETERRQ(PETSC_ERR_FILE_UNEXPECTED,0,"not matrix object"); 2107 if (header[3] < 0) { 2108 SETERRQ(PETSC_ERR_FILE_UNEXPECTED,1,"Matrix stored in special format, cannot load as MPIBAIJ"); 2109 } 2110 } 2111 2112 ierr = MPI_Bcast(header+1,3,MPI_INT,0,comm);CHKERRQ(ierr); 2113 M = header[1]; N = header[2]; 2114 2115 if (M != N) SETERRQ(PETSC_ERR_SUP,0,"Can only do square matrices"); 2116 2117 /* 2118 This code adds extra rows to make sure the number of rows is 2119 divisible by the blocksize 2120 */ 2121 Mbs = M/bs; 2122 extra_rows = bs - M + bs*(Mbs); 2123 if (extra_rows == bs) extra_rows = 0; 2124 else Mbs++; 2125 if (extra_rows &&!rank) { 2126 PLogInfo(0,"MatLoad_MPIBAIJ:Padding loaded matrix to match blocksize\n"); 2127 } 2128 2129 /* determine ownership of all rows */ 2130 mbs = Mbs/size + ((Mbs % size) > rank); 2131 m = mbs * bs; 2132 rowners = (int *) PetscMalloc(2*(size+2)*sizeof(int)); CHKPTRQ(rowners); 2133 browners = rowners + size + 1; 2134 ierr = MPI_Allgather(&mbs,1,MPI_INT,rowners+1,1,MPI_INT,comm);CHKERRQ(ierr); 2135 rowners[0] = 0; 2136 for ( i=2; i<=size; i++ ) rowners[i] += rowners[i-1]; 2137 for ( i=0; i<=size; i++ ) browners[i] = rowners[i]*bs; 2138 rstart = rowners[rank]; 2139 rend = rowners[rank+1]; 2140 2141 /* distribute row lengths to all processors */ 2142 locrowlens = (int*) PetscMalloc( (rend-rstart)*bs*sizeof(int) ); CHKPTRQ(locrowlens); 2143 if (!rank) { 2144 rowlengths = (int*) PetscMalloc( (M+extra_rows)*sizeof(int) ); CHKPTRQ(rowlengths); 2145 ierr = PetscBinaryRead(fd,rowlengths,M,PETSC_INT); CHKERRQ(ierr); 2146 for ( i=0; i<extra_rows; i++ ) rowlengths[M+i] = 1; 2147 sndcounts = (int*) PetscMalloc( size*sizeof(int) ); CHKPTRQ(sndcounts); 2148 for ( i=0; i<size; i++ ) sndcounts[i] = browners[i+1] - browners[i]; 2149 ierr = MPI_Scatterv(rowlengths,sndcounts,browners,MPI_INT,locrowlens,(rend-rstart)*bs,MPI_INT,0,comm);CHKERRQ(ierr); 2150 PetscFree(sndcounts); 2151 } else { 2152 ierr = MPI_Scatterv(0,0,0,MPI_INT,locrowlens,(rend-rstart)*bs,MPI_INT, 0,comm);CHKERRQ(ierr); 2153 } 2154 2155 if (!rank) { 2156 /* calculate the number of nonzeros on each processor */ 2157 procsnz = (int*) PetscMalloc( size*sizeof(int) ); CHKPTRQ(procsnz); 2158 PetscMemzero(procsnz,size*sizeof(int)); 2159 for ( i=0; i<size; i++ ) { 2160 for ( j=rowners[i]*bs; j< rowners[i+1]*bs; j++ ) { 2161 procsnz[i] += rowlengths[j]; 2162 } 2163 } 2164 PetscFree(rowlengths); 2165 2166 /* determine max buffer needed and allocate it */ 2167 maxnz = 0; 2168 for ( i=0; i<size; i++ ) { 2169 maxnz = PetscMax(maxnz,procsnz[i]); 2170 } 2171 cols = (int *) PetscMalloc( maxnz*sizeof(int) ); CHKPTRQ(cols); 2172 2173 /* read in my part of the matrix column indices */ 2174 nz = procsnz[0]; 2175 ibuf = (int *) PetscMalloc( nz*sizeof(int) ); CHKPTRQ(ibuf); 2176 mycols = ibuf; 2177 if (size == 1) nz -= extra_rows; 2178 ierr = PetscBinaryRead(fd,mycols,nz,PETSC_INT); CHKERRQ(ierr); 2179 if (size == 1) for (i=0; i< extra_rows; i++) { mycols[nz+i] = M+i; } 2180 2181 /* read in every ones (except the last) and ship off */ 2182 for ( i=1; i<size-1; i++ ) { 2183 nz = procsnz[i]; 2184 ierr = PetscBinaryRead(fd,cols,nz,PETSC_INT); CHKERRQ(ierr); 2185 ierr = MPI_Send(cols,nz,MPI_INT,i,tag,comm);CHKERRQ(ierr); 2186 } 2187 /* read in the stuff for the last proc */ 2188 if ( size != 1 ) { 2189 nz = procsnz[size-1] - extra_rows; /* the extra rows are not on the disk */ 2190 ierr = PetscBinaryRead(fd,cols,nz,PETSC_INT); CHKERRQ(ierr); 2191 for ( i=0; i<extra_rows; i++ ) cols[nz+i] = M+i; 2192 ierr = MPI_Send(cols,nz+extra_rows,MPI_INT,size-1,tag,comm);CHKERRQ(ierr); 2193 } 2194 PetscFree(cols); 2195 } else { 2196 /* determine buffer space needed for message */ 2197 nz = 0; 2198 for ( i=0; i<m; i++ ) { 2199 nz += locrowlens[i]; 2200 } 2201 ibuf = (int*) PetscMalloc( nz*sizeof(int) ); CHKPTRQ(ibuf); 2202 mycols = ibuf; 2203 /* receive message of column indices*/ 2204 ierr = MPI_Recv(mycols,nz,MPI_INT,0,tag,comm,&status);CHKERRQ(ierr); 2205 ierr = MPI_Get_count(&status,MPI_INT,&maxnz);CHKERRQ(ierr); 2206 if (maxnz != nz) SETERRQ(PETSC_ERR_FILE_UNEXPECTED,0,"something is wrong with file"); 2207 } 2208 2209 /* loop over local rows, determining number of off diagonal entries */ 2210 dlens = (int *) PetscMalloc( 2*(rend-rstart+1)*sizeof(int) ); CHKPTRQ(dlens); 2211 odlens = dlens + (rend-rstart); 2212 mask = (int *) PetscMalloc( 3*Mbs*sizeof(int) ); CHKPTRQ(mask); 2213 PetscMemzero(mask,3*Mbs*sizeof(int)); 2214 masked1 = mask + Mbs; 2215 masked2 = masked1 + Mbs; 2216 rowcount = 0; nzcount = 0; 2217 for ( i=0; i<mbs; i++ ) { 2218 dcount = 0; 2219 odcount = 0; 2220 for ( j=0; j<bs; j++ ) { 2221 kmax = locrowlens[rowcount]; 2222 for ( k=0; k<kmax; k++ ) { 2223 tmp = mycols[nzcount++]/bs; 2224 if (!mask[tmp]) { 2225 mask[tmp] = 1; 2226 if (tmp < rstart || tmp >= rend ) masked2[odcount++] = tmp; 2227 else masked1[dcount++] = tmp; 2228 } 2229 } 2230 rowcount++; 2231 } 2232 2233 dlens[i] = dcount; 2234 odlens[i] = odcount; 2235 2236 /* zero out the mask elements we set */ 2237 for ( j=0; j<dcount; j++ ) mask[masked1[j]] = 0; 2238 for ( j=0; j<odcount; j++ ) mask[masked2[j]] = 0; 2239 } 2240 2241 /* create our matrix */ 2242 ierr = MatCreateMPIBAIJ(comm,bs,m,PETSC_DECIDE,M+extra_rows,N+extra_rows,0,dlens,0,odlens,newmat); 2243 CHKERRQ(ierr); 2244 A = *newmat; 2245 MatSetOption(A,MAT_COLUMNS_SORTED); 2246 2247 if (!rank) { 2248 buf = (Scalar *) PetscMalloc( maxnz*sizeof(Scalar) ); CHKPTRQ(buf); 2249 /* read in my part of the matrix numerical values */ 2250 nz = procsnz[0]; 2251 vals = buf; 2252 mycols = ibuf; 2253 if (size == 1) nz -= extra_rows; 2254 ierr = PetscBinaryRead(fd,vals,nz,PETSC_SCALAR); CHKERRQ(ierr); 2255 if (size == 1) for (i=0; i< extra_rows; i++) { vals[nz+i] = 1.0; } 2256 2257 /* insert into matrix */ 2258 jj = rstart*bs; 2259 for ( i=0; i<m; i++ ) { 2260 ierr = MatSetValues(A,1,&jj,locrowlens[i],mycols,vals,INSERT_VALUES);CHKERRQ(ierr); 2261 mycols += locrowlens[i]; 2262 vals += locrowlens[i]; 2263 jj++; 2264 } 2265 /* read in other processors (except the last one) and ship out */ 2266 for ( i=1; i<size-1; i++ ) { 2267 nz = procsnz[i]; 2268 vals = buf; 2269 ierr = PetscBinaryRead(fd,vals,nz,PETSC_SCALAR); CHKERRQ(ierr); 2270 ierr = MPI_Send(vals,nz,MPIU_SCALAR,i,A->tag,comm);CHKERRQ(ierr); 2271 } 2272 /* the last proc */ 2273 if ( size != 1 ){ 2274 nz = procsnz[i] - extra_rows; 2275 vals = buf; 2276 ierr = PetscBinaryRead(fd,vals,nz,PETSC_SCALAR); CHKERRQ(ierr); 2277 for ( i=0; i<extra_rows; i++ ) vals[nz+i] = 1.0; 2278 ierr = MPI_Send(vals,nz+extra_rows,MPIU_SCALAR,size-1,A->tag,comm);CHKERRQ(ierr); 2279 } 2280 PetscFree(procsnz); 2281 } else { 2282 /* receive numeric values */ 2283 buf = (Scalar*) PetscMalloc( nz*sizeof(Scalar) ); CHKPTRQ(buf); 2284 2285 /* receive message of values*/ 2286 vals = buf; 2287 mycols = ibuf; 2288 ierr = MPI_Recv(vals,nz,MPIU_SCALAR,0,A->tag,comm,&status);CHKERRQ(ierr); 2289 ierr = MPI_Get_count(&status,MPIU_SCALAR,&maxnz);CHKERRQ(ierr); 2290 if (maxnz != nz) SETERRQ(PETSC_ERR_FILE_UNEXPECTED,0,"something is wrong with file"); 2291 2292 /* insert into matrix */ 2293 jj = rstart*bs; 2294 for ( i=0; i<m; i++ ) { 2295 ierr = MatSetValues(A,1,&jj,locrowlens[i],mycols,vals,INSERT_VALUES);CHKERRQ(ierr); 2296 mycols += locrowlens[i]; 2297 vals += locrowlens[i]; 2298 jj++; 2299 } 2300 } 2301 PetscFree(locrowlens); 2302 PetscFree(buf); 2303 PetscFree(ibuf); 2304 PetscFree(rowners); 2305 PetscFree(dlens); 2306 PetscFree(mask); 2307 ierr = MatAssemblyBegin(A,MAT_FINAL_ASSEMBLY); CHKERRQ(ierr); 2308 ierr = MatAssemblyEnd(A,MAT_FINAL_ASSEMBLY); CHKERRQ(ierr); 2309 PetscFunctionReturn(0); 2310 } 2311 2312 2313 2314 #undef __FUNC__ 2315 #define __FUNC__ "MatMPIBAIJSetHashTableFactor" 2316 /*@ 2317 MatMPIBAIJSetHashTableFactor - Sets the factor required to compute the size of the HashTable. 2318 2319 Input Parameters: 2320 . mat - the matrix 2321 . fact - factor 2322 2323 Notes: 2324 This can also be set by the command line option: -mat_use_hash_table fact 2325 2326 .keywords: matrix, hashtable, factor, HT 2327 2328 .seealso: MatSetOption() 2329 @*/ 2330 int MatMPIBAIJSetHashTableFactor(Mat mat,double fact) 2331 { 2332 int ierr; 2333 Mat_MPIBAIJ *baij; 2334 2335 PetscFunctionBegin; 2336 PetscValidHeaderSpecific(mat,MAT_COOKIE); 2337 if (mat->type != MATMPIBAIJ) { 2338 SETERRQ(PETSC_ERR_ARG_WRONG,1,"Incorrect matrix type. Use MPIBAIJ only."); 2339 } 2340 baij = (Mat_MPIBAIJ*) mat->data; 2341 baij->ht_fact = fact; 2342 PetscFunctionReturn(0); 2343 } 2344