1 2 #include "src/mat/impls/baij/mpi/mpibaij.h" /*I "petscmat.h" I*/ 3 #include "mpisbaij.h" 4 #include "src/mat/impls/sbaij/seq/sbaij.h" 5 6 EXTERN PetscErrorCode MatSetUpMultiply_MPISBAIJ(Mat); 7 EXTERN PetscErrorCode MatSetUpMultiply_MPISBAIJ_2comm(Mat); 8 EXTERN PetscErrorCode DisAssemble_MPISBAIJ(Mat); 9 EXTERN PetscErrorCode MatIncreaseOverlap_MPISBAIJ(Mat,PetscInt,IS[],PetscInt); 10 EXTERN PetscErrorCode MatGetValues_SeqSBAIJ(Mat,PetscInt,const PetscInt[],PetscInt,const PetscInt[],PetscScalar []); 11 EXTERN PetscErrorCode MatGetValues_SeqBAIJ(Mat,PetscInt,const PetscInt[],PetscInt,const PetscInt[],PetscScalar []); 12 EXTERN PetscErrorCode MatSetValues_SeqSBAIJ(Mat,PetscInt,const PetscInt [],PetscInt,const PetscInt [],const PetscScalar [],InsertMode); 13 EXTERN PetscErrorCode MatSetValuesBlocked_SeqSBAIJ(Mat,PetscInt,const PetscInt[],PetscInt,const PetscInt[],const PetscScalar[],InsertMode); 14 EXTERN PetscErrorCode MatSetValuesBlocked_SeqBAIJ(Mat,PetscInt,const PetscInt[],PetscInt,const PetscInt[],const PetscScalar[],InsertMode); 15 EXTERN PetscErrorCode MatGetRow_SeqSBAIJ(Mat,PetscInt,PetscInt*,PetscInt**,PetscScalar**); 16 EXTERN PetscErrorCode MatRestoreRow_SeqSBAIJ(Mat,PetscInt,PetscInt*,PetscInt**,PetscScalar**); 17 EXTERN PetscErrorCode MatPrintHelp_SeqSBAIJ(Mat); 18 EXTERN PetscErrorCode MatZeroRows_SeqSBAIJ(Mat,IS,PetscScalar*); 19 EXTERN PetscErrorCode MatZeroRows_SeqBAIJ(Mat,IS,PetscScalar *); 20 EXTERN PetscErrorCode MatGetRowMax_MPISBAIJ(Mat,Vec); 21 EXTERN PetscErrorCode MatRelax_MPISBAIJ(Mat,Vec,PetscReal,MatSORType,PetscReal,PetscInt,PetscInt,Vec); 22 23 /* UGLY, ugly, ugly 24 When MatScalar == PetscScalar the function MatSetValuesBlocked_MPIBAIJ_MatScalar() does 25 not exist. Otherwise ..._MatScalar() takes matrix elements in single precision and 26 inserts them into the single precision data structure. The function MatSetValuesBlocked_MPIBAIJ() 27 converts the entries into single precision and then calls ..._MatScalar() to put them 28 into the single precision data structures. 29 */ 30 #if defined(PETSC_USE_MAT_SINGLE) 31 EXTERN PetscErrorCode MatSetValuesBlocked_SeqSBAIJ_MatScalar(Mat,PetscInt,const PetscInt[],PetscInt,const PetscInt[],const MatScalar[],InsertMode); 32 EXTERN PetscErrorCode MatSetValues_MPISBAIJ_MatScalar(Mat,PetscInt,const PetscInt[],PetscInt,const PetscInt[],const MatScalar[],InsertMode); 33 EXTERN PetscErrorCode MatSetValuesBlocked_MPISBAIJ_MatScalar(Mat,PetscInt,const PetscInt[],PetscInt,const PetscInt[],const MatScalar[],InsertMode); 34 EXTERN PetscErrorCode MatSetValues_MPISBAIJ_HT_MatScalar(Mat,PetscInt,const PetscInt[],PetscInt,const PetscInt[],const MatScalar[],InsertMode); 35 EXTERN PetscErrorCode MatSetValuesBlocked_MPISBAIJ_HT_MatScalar(Mat,PetscInt,const PetscInt[],PetscInt,const PetscInt[],const MatScalar[],InsertMode); 36 #else 37 #define MatSetValuesBlocked_SeqSBAIJ_MatScalar MatSetValuesBlocked_SeqSBAIJ 38 #define MatSetValues_MPISBAIJ_MatScalar MatSetValues_MPISBAIJ 39 #define MatSetValuesBlocked_MPISBAIJ_MatScalar MatSetValuesBlocked_MPISBAIJ 40 #define MatSetValues_MPISBAIJ_HT_MatScalar MatSetValues_MPISBAIJ_HT 41 #define MatSetValuesBlocked_MPISBAIJ_HT_MatScalar MatSetValuesBlocked_MPISBAIJ_HT 42 #endif 43 44 EXTERN_C_BEGIN 45 #undef __FUNCT__ 46 #define __FUNCT__ "MatStoreValues_MPISBAIJ" 47 PetscErrorCode MatStoreValues_MPISBAIJ(Mat mat) 48 { 49 Mat_MPISBAIJ *aij = (Mat_MPISBAIJ *)mat->data; 50 PetscErrorCode ierr; 51 52 PetscFunctionBegin; 53 ierr = MatStoreValues(aij->A);CHKERRQ(ierr); 54 ierr = MatStoreValues(aij->B);CHKERRQ(ierr); 55 PetscFunctionReturn(0); 56 } 57 EXTERN_C_END 58 59 EXTERN_C_BEGIN 60 #undef __FUNCT__ 61 #define __FUNCT__ "MatRetrieveValues_MPISBAIJ" 62 PetscErrorCode MatRetrieveValues_MPISBAIJ(Mat mat) 63 { 64 Mat_MPISBAIJ *aij = (Mat_MPISBAIJ *)mat->data; 65 PetscErrorCode ierr; 66 67 PetscFunctionBegin; 68 ierr = MatRetrieveValues(aij->A);CHKERRQ(ierr); 69 ierr = MatRetrieveValues(aij->B);CHKERRQ(ierr); 70 PetscFunctionReturn(0); 71 } 72 EXTERN_C_END 73 74 75 #define CHUNKSIZE 10 76 77 #define MatSetValues_SeqSBAIJ_A_Private(row,col,value,addv) \ 78 { \ 79 \ 80 brow = row/bs; \ 81 rp = aj + ai[brow]; ap = aa + bs2*ai[brow]; \ 82 rmax = aimax[brow]; nrow = ailen[brow]; \ 83 bcol = col/bs; \ 84 ridx = row % bs; cidx = col % bs; \ 85 low = 0; high = nrow; \ 86 while (high-low > 3) { \ 87 t = (low+high)/2; \ 88 if (rp[t] > bcol) high = t; \ 89 else low = t; \ 90 } \ 91 for (_i=low; _i<high; _i++) { \ 92 if (rp[_i] > bcol) break; \ 93 if (rp[_i] == bcol) { \ 94 bap = ap + bs2*_i + bs*cidx + ridx; \ 95 if (addv == ADD_VALUES) *bap += value; \ 96 else *bap = value; \ 97 goto a_noinsert; \ 98 } \ 99 } \ 100 if (a->nonew == 1) goto a_noinsert; \ 101 else if (a->nonew == -1) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"Inserting a new nonzero (%D, %D) into matrix", row, col); \ 102 if (nrow >= rmax) { \ 103 /* there is no extra room in row, therefore enlarge */ \ 104 PetscInt new_nz = ai[a->mbs] + CHUNKSIZE,len,*new_i,*new_j; \ 105 MatScalar *new_a; \ 106 \ 107 if (a->nonew == -2) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"Inserting a new nonzero (%D, %D) in the matrix", row, col); \ 108 \ 109 /* malloc new storage space */ \ 110 len = new_nz*(sizeof(PetscInt)+bs2*sizeof(MatScalar))+(a->mbs+1)*sizeof(PetscInt); \ 111 ierr = PetscMalloc(len,&new_a);CHKERRQ(ierr); \ 112 new_j = (PetscInt*)(new_a + bs2*new_nz); \ 113 new_i = new_j + new_nz; \ 114 \ 115 /* copy over old data into new slots */ \ 116 for (ii=0; ii<brow+1; ii++) {new_i[ii] = ai[ii];} \ 117 for (ii=brow+1; ii<a->mbs+1; ii++) {new_i[ii] = ai[ii]+CHUNKSIZE;} \ 118 ierr = PetscMemcpy(new_j,aj,(ai[brow]+nrow)*sizeof(PetscInt));CHKERRQ(ierr); \ 119 len = (new_nz - CHUNKSIZE - ai[brow] - nrow); \ 120 ierr = PetscMemcpy(new_j+ai[brow]+nrow+CHUNKSIZE,aj+ai[brow]+nrow,len*sizeof(PetscInt));CHKERRQ(ierr); \ 121 ierr = PetscMemcpy(new_a,aa,(ai[brow]+nrow)*bs2*sizeof(MatScalar));CHKERRQ(ierr); \ 122 ierr = PetscMemzero(new_a+bs2*(ai[brow]+nrow),bs2*CHUNKSIZE*sizeof(PetscScalar));CHKERRQ(ierr); \ 123 ierr = PetscMemcpy(new_a+bs2*(ai[brow]+nrow+CHUNKSIZE), \ 124 aa+bs2*(ai[brow]+nrow),bs2*len*sizeof(MatScalar));CHKERRQ(ierr); \ 125 /* free up old matrix storage */ \ 126 ierr = PetscFree(a->a);CHKERRQ(ierr); \ 127 if (!a->singlemalloc) { \ 128 ierr = PetscFree(a->i);CHKERRQ(ierr); \ 129 ierr = PetscFree(a->j);CHKERRQ(ierr);\ 130 } \ 131 aa = a->a = new_a; ai = a->i = new_i; aj = a->j = new_j; \ 132 a->singlemalloc = PETSC_TRUE; \ 133 \ 134 rp = aj + ai[brow]; ap = aa + bs2*ai[brow]; \ 135 rmax = aimax[brow] = aimax[brow] + CHUNKSIZE; \ 136 ierr = PetscLogObjectMemory(A,CHUNKSIZE*(sizeof(PetscInt) + bs2*sizeof(MatScalar)));CHKERRQ(ierr); \ 137 a->maxnz += bs2*CHUNKSIZE; \ 138 a->reallocs++; \ 139 a->nz++; \ 140 } \ 141 N = nrow++ - 1; \ 142 /* shift up all the later entries in this row */ \ 143 for (ii=N; ii>=_i; ii--) { \ 144 rp[ii+1] = rp[ii]; \ 145 ierr = PetscMemcpy(ap+bs2*(ii+1),ap+bs2*(ii),bs2*sizeof(MatScalar));CHKERRQ(ierr); \ 146 } \ 147 if (N>=_i) { ierr = PetscMemzero(ap+bs2*_i,bs2*sizeof(MatScalar));CHKERRQ(ierr); } \ 148 rp[_i] = bcol; \ 149 ap[bs2*_i + bs*cidx + ridx] = value; \ 150 a_noinsert:; \ 151 ailen[brow] = nrow; \ 152 } 153 #ifndef MatSetValues_SeqBAIJ_B_Private 154 #define MatSetValues_SeqSBAIJ_B_Private(row,col,value,addv) \ 155 { \ 156 brow = row/bs; \ 157 rp = bj + bi[brow]; ap = ba + bs2*bi[brow]; \ 158 rmax = bimax[brow]; nrow = bilen[brow]; \ 159 bcol = col/bs; \ 160 ridx = row % bs; cidx = col % bs; \ 161 low = 0; high = nrow; \ 162 while (high-low > 3) { \ 163 t = (low+high)/2; \ 164 if (rp[t] > bcol) high = t; \ 165 else low = t; \ 166 } \ 167 for (_i=low; _i<high; _i++) { \ 168 if (rp[_i] > bcol) break; \ 169 if (rp[_i] == bcol) { \ 170 bap = ap + bs2*_i + bs*cidx + ridx; \ 171 if (addv == ADD_VALUES) *bap += value; \ 172 else *bap = value; \ 173 goto b_noinsert; \ 174 } \ 175 } \ 176 if (b->nonew == 1) goto b_noinsert; \ 177 else if (b->nonew == -1) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"Inserting a new nonzero (%D, %D) into matrix", row, col); \ 178 if (nrow >= rmax) { \ 179 /* there is no extra room in row, therefore enlarge */ \ 180 PetscInt new_nz = bi[b->mbs] + CHUNKSIZE,len,*new_i,*new_j; \ 181 MatScalar *new_a; \ 182 \ 183 if (b->nonew == -2) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"Inserting a new nonzero (%D, %D) in the matrix", row, col); \ 184 \ 185 /* malloc new storage space */ \ 186 len = new_nz*(sizeof(PetscInt)+bs2*sizeof(MatScalar))+(b->mbs+1)*sizeof(PetscInt); \ 187 ierr = PetscMalloc(len,&new_a);CHKERRQ(ierr); \ 188 new_j = (PetscInt*)(new_a + bs2*new_nz); \ 189 new_i = new_j + new_nz; \ 190 \ 191 /* copy over old data into new slots */ \ 192 for (ii=0; ii<brow+1; ii++) {new_i[ii] = bi[ii];} \ 193 for (ii=brow+1; ii<b->mbs+1; ii++) {new_i[ii] = bi[ii]+CHUNKSIZE;} \ 194 ierr = PetscMemcpy(new_j,bj,(bi[brow]+nrow)*sizeof(PetscInt));CHKERRQ(ierr); \ 195 len = (new_nz - CHUNKSIZE - bi[brow] - nrow); \ 196 ierr = PetscMemcpy(new_j+bi[brow]+nrow+CHUNKSIZE,bj+bi[brow]+nrow,len*sizeof(PetscInt));CHKERRQ(ierr); \ 197 ierr = PetscMemcpy(new_a,ba,(bi[brow]+nrow)*bs2*sizeof(MatScalar));CHKERRQ(ierr); \ 198 ierr = PetscMemzero(new_a+bs2*(bi[brow]+nrow),bs2*CHUNKSIZE*sizeof(MatScalar));CHKERRQ(ierr); \ 199 ierr = PetscMemcpy(new_a+bs2*(bi[brow]+nrow+CHUNKSIZE), \ 200 ba+bs2*(bi[brow]+nrow),bs2*len*sizeof(MatScalar));CHKERRQ(ierr); \ 201 /* free up old matrix storage */ \ 202 ierr = PetscFree(b->a);CHKERRQ(ierr); \ 203 if (!b->singlemalloc) { \ 204 ierr = PetscFree(b->i);CHKERRQ(ierr); \ 205 ierr = PetscFree(b->j);CHKERRQ(ierr); \ 206 } \ 207 ba = b->a = new_a; bi = b->i = new_i; bj = b->j = new_j; \ 208 b->singlemalloc = PETSC_TRUE; \ 209 \ 210 rp = bj + bi[brow]; ap = ba + bs2*bi[brow]; \ 211 rmax = bimax[brow] = bimax[brow] + CHUNKSIZE; \ 212 ierr = PetscLogObjectMemory(B,CHUNKSIZE*(sizeof(PetscInt) + bs2*sizeof(MatScalar)));CHKERRQ(ierr); \ 213 b->maxnz += bs2*CHUNKSIZE; \ 214 b->reallocs++; \ 215 b->nz++; \ 216 } \ 217 N = nrow++ - 1; \ 218 /* shift up all the later entries in this row */ \ 219 for (ii=N; ii>=_i; ii--) { \ 220 rp[ii+1] = rp[ii]; \ 221 ierr = PetscMemcpy(ap+bs2*(ii+1),ap+bs2*(ii),bs2*sizeof(MatScalar));CHKERRQ(ierr); \ 222 } \ 223 if (N>=_i) { ierr = PetscMemzero(ap+bs2*_i,bs2*sizeof(MatScalar));CHKERRQ(ierr);} \ 224 rp[_i] = bcol; \ 225 ap[bs2*_i + bs*cidx + ridx] = value; \ 226 b_noinsert:; \ 227 bilen[brow] = nrow; \ 228 } 229 #endif 230 231 #if defined(PETSC_USE_MAT_SINGLE) 232 #undef __FUNCT__ 233 #define __FUNCT__ "MatSetValues_MPISBAIJ" 234 PetscErrorCode MatSetValues_MPISBAIJ(Mat mat,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],const PetscScalar v[],InsertMode addv) 235 { 236 Mat_MPISBAIJ *b = (Mat_MPISBAIJ*)mat->data; 237 PetscErrorCode ierr; 238 PetscInt i,N = m*n; 239 MatScalar *vsingle; 240 241 PetscFunctionBegin; 242 if (N > b->setvalueslen) { 243 if (b->setvaluescopy) {ierr = PetscFree(b->setvaluescopy);CHKERRQ(ierr);} 244 ierr = PetscMalloc(N*sizeof(MatScalar),&b->setvaluescopy);CHKERRQ(ierr); 245 b->setvalueslen = N; 246 } 247 vsingle = b->setvaluescopy; 248 249 for (i=0; i<N; i++) { 250 vsingle[i] = v[i]; 251 } 252 ierr = MatSetValues_MPISBAIJ_MatScalar(mat,m,im,n,in,vsingle,addv);CHKERRQ(ierr); 253 PetscFunctionReturn(0); 254 } 255 256 #undef __FUNCT__ 257 #define __FUNCT__ "MatSetValuesBlocked_MPISBAIJ" 258 PetscErrorCode MatSetValuesBlocked_MPISBAIJ(Mat mat,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],const PetscScalar v[],InsertMode addv) 259 { 260 Mat_MPIBAIJ *b = (Mat_MPIBAIJ*)mat->data; 261 PetscErrorCode ierr; 262 PetscInt i,N = m*n*b->bs2; 263 MatScalar *vsingle; 264 265 PetscFunctionBegin; 266 if (N > b->setvalueslen) { 267 if (b->setvaluescopy) {ierr = PetscFree(b->setvaluescopy);CHKERRQ(ierr);} 268 ierr = PetscMalloc(N*sizeof(MatScalar),&b->setvaluescopy);CHKERRQ(ierr); 269 b->setvalueslen = N; 270 } 271 vsingle = b->setvaluescopy; 272 for (i=0; i<N; i++) { 273 vsingle[i] = v[i]; 274 } 275 ierr = MatSetValuesBlocked_MPISBAIJ_MatScalar(mat,m,im,n,in,vsingle,addv);CHKERRQ(ierr); 276 PetscFunctionReturn(0); 277 } 278 #endif 279 280 /* Only add/insert a(i,j) with i<=j (blocks). 281 Any a(i,j) with i>j input by user is ingored. 282 */ 283 #undef __FUNCT__ 284 #define __FUNCT__ "MatSetValues_MPIBAIJ_MatScalar" 285 PetscErrorCode MatSetValues_MPISBAIJ_MatScalar(Mat mat,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],const MatScalar v[],InsertMode addv) 286 { 287 Mat_MPISBAIJ *baij = (Mat_MPISBAIJ*)mat->data; 288 MatScalar value; 289 PetscTruth roworiented = baij->roworiented; 290 PetscErrorCode ierr; 291 PetscInt i,j,row,col; 292 PetscInt rstart_orig=baij->rstart_bs; 293 PetscInt rend_orig=baij->rend_bs,cstart_orig=baij->cstart_bs; 294 PetscInt cend_orig=baij->cend_bs,bs=mat->bs; 295 296 /* Some Variables required in the macro */ 297 Mat A = baij->A; 298 Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)(A)->data; 299 PetscInt *aimax=a->imax,*ai=a->i,*ailen=a->ilen,*aj=a->j; 300 MatScalar *aa=a->a; 301 302 Mat B = baij->B; 303 Mat_SeqBAIJ *b = (Mat_SeqBAIJ*)(B)->data; 304 PetscInt *bimax=b->imax,*bi=b->i,*bilen=b->ilen,*bj=b->j; 305 MatScalar *ba=b->a; 306 307 PetscInt *rp,ii,nrow,_i,rmax,N,brow,bcol; 308 PetscInt low,high,t,ridx,cidx,bs2=a->bs2; 309 MatScalar *ap,*bap; 310 311 /* for stash */ 312 PetscInt n_loc, *in_loc=0; 313 MatScalar *v_loc=0; 314 315 PetscFunctionBegin; 316 317 if(!baij->donotstash){ 318 ierr = PetscMalloc(n*sizeof(PetscInt),&in_loc);CHKERRQ(ierr); 319 ierr = PetscMalloc(n*sizeof(MatScalar),&v_loc);CHKERRQ(ierr); 320 } 321 322 for (i=0; i<m; i++) { 323 if (im[i] < 0) continue; 324 #if defined(PETSC_USE_DEBUG) 325 if (im[i] >= mat->M) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"Row too large: row %D max %D",im[i],mat->M-1); 326 #endif 327 if (im[i] >= rstart_orig && im[i] < rend_orig) { /* this processor entry */ 328 row = im[i] - rstart_orig; /* local row index */ 329 for (j=0; j<n; j++) { 330 if (im[i]/bs > in[j]/bs) continue; /* ignore lower triangular blocks */ 331 if (in[j] >= cstart_orig && in[j] < cend_orig){ /* diag entry (A) */ 332 col = in[j] - cstart_orig; /* local col index */ 333 brow = row/bs; bcol = col/bs; 334 if (brow > bcol) continue; /* ignore lower triangular blocks of A */ 335 if (roworiented) value = v[i*n+j]; else value = v[i+j*m]; 336 MatSetValues_SeqSBAIJ_A_Private(row,col,value,addv); 337 /* ierr = MatSetValues_SeqBAIJ(baij->A,1,&row,1,&col,&value,addv);CHKERRQ(ierr); */ 338 } else if (in[j] < 0) continue; 339 #if defined(PETSC_USE_DEBUG) 340 else if (in[j] >= mat->N) {SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"Column too large: col %D max %D",in[j],mat->N-1);} 341 #endif 342 else { /* off-diag entry (B) */ 343 if (mat->was_assembled) { 344 if (!baij->colmap) { 345 ierr = CreateColmap_MPIBAIJ_Private(mat);CHKERRQ(ierr); 346 } 347 #if defined (PETSC_USE_CTABLE) 348 ierr = PetscTableFind(baij->colmap,in[j]/bs + 1,&col);CHKERRQ(ierr); 349 col = col - 1; 350 #else 351 col = baij->colmap[in[j]/bs] - 1; 352 #endif 353 if (col < 0 && !((Mat_SeqSBAIJ*)(baij->A->data))->nonew) { 354 ierr = DisAssemble_MPISBAIJ(mat);CHKERRQ(ierr); 355 col = in[j]; 356 /* Reinitialize the variables required by MatSetValues_SeqBAIJ_B_Private() */ 357 B = baij->B; 358 b = (Mat_SeqBAIJ*)(B)->data; 359 bimax=b->imax;bi=b->i;bilen=b->ilen;bj=b->j; 360 ba=b->a; 361 } else col += in[j]%bs; 362 } else col = in[j]; 363 if (roworiented) value = v[i*n+j]; else value = v[i+j*m]; 364 MatSetValues_SeqSBAIJ_B_Private(row,col,value,addv); 365 /* ierr = MatSetValues_SeqBAIJ(baij->B,1,&row,1,&col,&value,addv);CHKERRQ(ierr); */ 366 } 367 } 368 } else { /* off processor entry */ 369 if (!baij->donotstash) { 370 n_loc = 0; 371 for (j=0; j<n; j++){ 372 if (im[i]/bs > in[j]/bs) continue; /* ignore lower triangular blocks */ 373 in_loc[n_loc] = in[j]; 374 if (roworiented) { 375 v_loc[n_loc] = v[i*n+j]; 376 } else { 377 v_loc[n_loc] = v[j*m+i]; 378 } 379 n_loc++; 380 } 381 ierr = MatStashValuesRow_Private(&mat->stash,im[i],n_loc,in_loc,v_loc);CHKERRQ(ierr); 382 } 383 } 384 } 385 386 if(!baij->donotstash){ 387 ierr = PetscFree(in_loc);CHKERRQ(ierr); 388 ierr = PetscFree(v_loc);CHKERRQ(ierr); 389 } 390 PetscFunctionReturn(0); 391 } 392 393 #undef __FUNCT__ 394 #define __FUNCT__ "MatSetValuesBlocked_MPISBAIJ_MatScalar" 395 PetscErrorCode MatSetValuesBlocked_MPISBAIJ_MatScalar(Mat mat,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],const MatScalar v[],InsertMode addv) 396 { 397 Mat_MPISBAIJ *baij = (Mat_MPISBAIJ*)mat->data; 398 const MatScalar *value; 399 MatScalar *barray=baij->barray; 400 PetscTruth roworiented = baij->roworiented; 401 PetscErrorCode ierr; 402 PetscInt i,j,ii,jj,row,col,rstart=baij->rstart; 403 PetscInt rend=baij->rend,cstart=baij->cstart,stepval; 404 PetscInt cend=baij->cend,bs=mat->bs,bs2=baij->bs2; 405 406 PetscFunctionBegin; 407 if(!barray) { 408 ierr = PetscMalloc(bs2*sizeof(MatScalar),&barray);CHKERRQ(ierr); 409 baij->barray = barray; 410 } 411 412 if (roworiented) { 413 stepval = (n-1)*bs; 414 } else { 415 stepval = (m-1)*bs; 416 } 417 for (i=0; i<m; i++) { 418 if (im[i] < 0) continue; 419 #if defined(PETSC_USE_DEBUG) 420 if (im[i] >= baij->Mbs) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"Row too large, row %D max %D",im[i],baij->Mbs-1); 421 #endif 422 if (im[i] >= rstart && im[i] < rend) { 423 row = im[i] - rstart; 424 for (j=0; j<n; j++) { 425 /* If NumCol = 1 then a copy is not required */ 426 if ((roworiented) && (n == 1)) { 427 barray = (MatScalar*) v + i*bs2; 428 } else if((!roworiented) && (m == 1)) { 429 barray = (MatScalar*) v + j*bs2; 430 } else { /* Here a copy is required */ 431 if (roworiented) { 432 value = v + i*(stepval+bs)*bs + j*bs; 433 } else { 434 value = v + j*(stepval+bs)*bs + i*bs; 435 } 436 for (ii=0; ii<bs; ii++,value+=stepval) { 437 for (jj=0; jj<bs; jj++) { 438 *barray++ = *value++; 439 } 440 } 441 barray -=bs2; 442 } 443 444 if (in[j] >= cstart && in[j] < cend){ 445 col = in[j] - cstart; 446 ierr = MatSetValuesBlocked_SeqSBAIJ(baij->A,1,&row,1,&col,barray,addv);CHKERRQ(ierr); 447 } 448 else if (in[j] < 0) continue; 449 #if defined(PETSC_USE_DEBUG) 450 else if (in[j] >= baij->Nbs) {SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"Column too large, col %D max %D",in[j],baij->Nbs-1);} 451 #endif 452 else { 453 if (mat->was_assembled) { 454 if (!baij->colmap) { 455 ierr = CreateColmap_MPIBAIJ_Private(mat);CHKERRQ(ierr); 456 } 457 458 #if defined(PETSC_USE_DEBUG) 459 #if defined (PETSC_USE_CTABLE) 460 { PetscInt data; 461 ierr = PetscTableFind(baij->colmap,in[j]+1,&data);CHKERRQ(ierr); 462 if ((data - 1) % bs) SETERRQ(PETSC_ERR_PLIB,"Incorrect colmap"); 463 } 464 #else 465 if ((baij->colmap[in[j]] - 1) % bs) SETERRQ(PETSC_ERR_PLIB,"Incorrect colmap"); 466 #endif 467 #endif 468 #if defined (PETSC_USE_CTABLE) 469 ierr = PetscTableFind(baij->colmap,in[j]+1,&col);CHKERRQ(ierr); 470 col = (col - 1)/bs; 471 #else 472 col = (baij->colmap[in[j]] - 1)/bs; 473 #endif 474 if (col < 0 && !((Mat_SeqBAIJ*)(baij->A->data))->nonew) { 475 ierr = DisAssemble_MPISBAIJ(mat);CHKERRQ(ierr); 476 col = in[j]; 477 } 478 } 479 else col = in[j]; 480 ierr = MatSetValuesBlocked_SeqBAIJ(baij->B,1,&row,1,&col,barray,addv);CHKERRQ(ierr); 481 } 482 } 483 } else { 484 if (!baij->donotstash) { 485 if (roworiented) { 486 ierr = MatStashValuesRowBlocked_Private(&mat->bstash,im[i],n,in,v,m,n,i);CHKERRQ(ierr); 487 } else { 488 ierr = MatStashValuesColBlocked_Private(&mat->bstash,im[i],n,in,v,m,n,i);CHKERRQ(ierr); 489 } 490 } 491 } 492 } 493 PetscFunctionReturn(0); 494 } 495 496 #undef __FUNCT__ 497 #define __FUNCT__ "MatGetValues_MPISBAIJ" 498 PetscErrorCode MatGetValues_MPISBAIJ(Mat mat,PetscInt m,const PetscInt idxm[],PetscInt n,const PetscInt idxn[],PetscScalar v[]) 499 { 500 Mat_MPISBAIJ *baij = (Mat_MPISBAIJ*)mat->data; 501 PetscErrorCode ierr; 502 PetscInt bs=mat->bs,i,j,bsrstart = baij->rstart*bs,bsrend = baij->rend*bs; 503 PetscInt bscstart = baij->cstart*bs,bscend = baij->cend*bs,row,col,data; 504 505 PetscFunctionBegin; 506 for (i=0; i<m; i++) { 507 if (idxm[i] < 0) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"Negative row: %D",idxm[i]); 508 if (idxm[i] >= mat->M) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"Row too large: row %D max %D",idxm[i],mat->M-1); 509 if (idxm[i] >= bsrstart && idxm[i] < bsrend) { 510 row = idxm[i] - bsrstart; 511 for (j=0; j<n; j++) { 512 if (idxn[j] < 0) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"Negative column %D",idxn[j]); 513 if (idxn[j] >= mat->N) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"Column too large: col %D max %D",idxn[j],mat->N-1); 514 if (idxn[j] >= bscstart && idxn[j] < bscend){ 515 col = idxn[j] - bscstart; 516 ierr = MatGetValues_SeqSBAIJ(baij->A,1,&row,1,&col,v+i*n+j);CHKERRQ(ierr); 517 } else { 518 if (!baij->colmap) { 519 ierr = CreateColmap_MPIBAIJ_Private(mat);CHKERRQ(ierr); 520 } 521 #if defined (PETSC_USE_CTABLE) 522 ierr = PetscTableFind(baij->colmap,idxn[j]/bs+1,&data);CHKERRQ(ierr); 523 data --; 524 #else 525 data = baij->colmap[idxn[j]/bs]-1; 526 #endif 527 if((data < 0) || (baij->garray[data/bs] != idxn[j]/bs)) *(v+i*n+j) = 0.0; 528 else { 529 col = data + idxn[j]%bs; 530 ierr = MatGetValues_SeqBAIJ(baij->B,1,&row,1,&col,v+i*n+j);CHKERRQ(ierr); 531 } 532 } 533 } 534 } else { 535 SETERRQ(PETSC_ERR_SUP,"Only local values currently supported"); 536 } 537 } 538 PetscFunctionReturn(0); 539 } 540 541 #undef __FUNCT__ 542 #define __FUNCT__ "MatNorm_MPISBAIJ" 543 PetscErrorCode MatNorm_MPISBAIJ(Mat mat,NormType type,PetscReal *norm) 544 { 545 Mat_MPISBAIJ *baij = (Mat_MPISBAIJ*)mat->data; 546 PetscErrorCode ierr; 547 PetscReal sum[2],*lnorm2; 548 549 PetscFunctionBegin; 550 if (baij->size == 1) { 551 ierr = MatNorm(baij->A,type,norm);CHKERRQ(ierr); 552 } else { 553 if (type == NORM_FROBENIUS) { 554 ierr = PetscMalloc(2*sizeof(PetscReal),&lnorm2);CHKERRQ(ierr); 555 ierr = MatNorm(baij->A,type,lnorm2);CHKERRQ(ierr); 556 *lnorm2 = (*lnorm2)*(*lnorm2); lnorm2++; /* squar power of norm(A) */ 557 ierr = MatNorm(baij->B,type,lnorm2);CHKERRQ(ierr); 558 *lnorm2 = (*lnorm2)*(*lnorm2); lnorm2--; /* squar power of norm(B) */ 559 ierr = MPI_Allreduce(lnorm2,&sum,2,MPIU_REAL,MPI_SUM,mat->comm);CHKERRQ(ierr); 560 *norm = sqrt(sum[0] + 2*sum[1]); 561 ierr = PetscFree(lnorm2);CHKERRQ(ierr); 562 } else { 563 SETERRQ(PETSC_ERR_SUP,"No support for this norm yet"); 564 } 565 } 566 PetscFunctionReturn(0); 567 } 568 569 #undef __FUNCT__ 570 #define __FUNCT__ "MatAssemblyBegin_MPISBAIJ" 571 PetscErrorCode MatAssemblyBegin_MPISBAIJ(Mat mat,MatAssemblyType mode) 572 { 573 Mat_MPISBAIJ *baij = (Mat_MPISBAIJ*)mat->data; 574 PetscErrorCode ierr; 575 PetscInt nstash,reallocs; 576 InsertMode addv; 577 578 PetscFunctionBegin; 579 if (baij->donotstash) { 580 PetscFunctionReturn(0); 581 } 582 583 /* make sure all processors are either in INSERTMODE or ADDMODE */ 584 ierr = MPI_Allreduce(&mat->insertmode,&addv,1,MPI_INT,MPI_BOR,mat->comm);CHKERRQ(ierr); 585 if (addv == (ADD_VALUES|INSERT_VALUES)) { 586 SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Some processors inserted others added"); 587 } 588 mat->insertmode = addv; /* in case this processor had no cache */ 589 590 ierr = MatStashScatterBegin_Private(&mat->stash,baij->rowners_bs);CHKERRQ(ierr); 591 ierr = MatStashScatterBegin_Private(&mat->bstash,baij->rowners);CHKERRQ(ierr); 592 ierr = MatStashGetInfo_Private(&mat->stash,&nstash,&reallocs);CHKERRQ(ierr); 593 PetscLogInfo(0,"MatAssemblyBegin_MPISBAIJ:Stash has %D entries,uses %D mallocs.\n",nstash,reallocs); 594 ierr = MatStashGetInfo_Private(&mat->stash,&nstash,&reallocs);CHKERRQ(ierr); 595 PetscLogInfo(0,"MatAssemblyBegin_MPISBAIJ:Block-Stash has %D entries, uses %D mallocs.\n",nstash,reallocs); 596 PetscFunctionReturn(0); 597 } 598 599 #undef __FUNCT__ 600 #define __FUNCT__ "MatAssemblyEnd_MPISBAIJ" 601 PetscErrorCode MatAssemblyEnd_MPISBAIJ(Mat mat,MatAssemblyType mode) 602 { 603 Mat_MPISBAIJ *baij=(Mat_MPISBAIJ*)mat->data; 604 Mat_SeqSBAIJ *a=(Mat_SeqSBAIJ*)baij->A->data; 605 Mat_SeqBAIJ *b=(Mat_SeqBAIJ*)baij->B->data; 606 PetscErrorCode ierr; 607 PetscInt i,j,rstart,ncols,flg,bs2=baij->bs2; 608 PetscInt *row,*col,other_disassembled; 609 PetscMPIInt n; 610 PetscTruth r1,r2,r3; 611 MatScalar *val; 612 InsertMode addv = mat->insertmode; 613 614 PetscFunctionBegin; 615 616 if (!baij->donotstash) { 617 while (1) { 618 ierr = MatStashScatterGetMesg_Private(&mat->stash,&n,&row,&col,&val,&flg);CHKERRQ(ierr); 619 if (!flg) break; 620 621 for (i=0; i<n;) { 622 /* Now identify the consecutive vals belonging to the same row */ 623 for (j=i,rstart=row[j]; j<n; j++) { if (row[j] != rstart) break; } 624 if (j < n) ncols = j-i; 625 else ncols = n-i; 626 /* Now assemble all these values with a single function call */ 627 ierr = MatSetValues_MPISBAIJ_MatScalar(mat,1,row+i,ncols,col+i,val+i,addv);CHKERRQ(ierr); 628 i = j; 629 } 630 } 631 ierr = MatStashScatterEnd_Private(&mat->stash);CHKERRQ(ierr); 632 /* Now process the block-stash. Since the values are stashed column-oriented, 633 set the roworiented flag to column oriented, and after MatSetValues() 634 restore the original flags */ 635 r1 = baij->roworiented; 636 r2 = a->roworiented; 637 r3 = b->roworiented; 638 baij->roworiented = PETSC_FALSE; 639 a->roworiented = PETSC_FALSE; 640 b->roworiented = PETSC_FALSE; 641 while (1) { 642 ierr = MatStashScatterGetMesg_Private(&mat->bstash,&n,&row,&col,&val,&flg);CHKERRQ(ierr); 643 if (!flg) break; 644 645 for (i=0; i<n;) { 646 /* Now identify the consecutive vals belonging to the same row */ 647 for (j=i,rstart=row[j]; j<n; j++) { if (row[j] != rstart) break; } 648 if (j < n) ncols = j-i; 649 else ncols = n-i; 650 ierr = MatSetValuesBlocked_MPISBAIJ_MatScalar(mat,1,row+i,ncols,col+i,val+i*bs2,addv);CHKERRQ(ierr); 651 i = j; 652 } 653 } 654 ierr = MatStashScatterEnd_Private(&mat->bstash);CHKERRQ(ierr); 655 baij->roworiented = r1; 656 a->roworiented = r2; 657 b->roworiented = r3; 658 } 659 660 ierr = MatAssemblyBegin(baij->A,mode);CHKERRQ(ierr); 661 ierr = MatAssemblyEnd(baij->A,mode);CHKERRQ(ierr); 662 663 /* determine if any processor has disassembled, if so we must 664 also disassemble ourselfs, in order that we may reassemble. */ 665 /* 666 if nonzero structure of submatrix B cannot change then we know that 667 no processor disassembled thus we can skip this stuff 668 */ 669 if (!((Mat_SeqBAIJ*)baij->B->data)->nonew) { 670 ierr = MPI_Allreduce(&mat->was_assembled,&other_disassembled,1,MPI_INT,MPI_PROD,mat->comm);CHKERRQ(ierr); 671 if (mat->was_assembled && !other_disassembled) { 672 ierr = DisAssemble_MPISBAIJ(mat);CHKERRQ(ierr); 673 } 674 } 675 676 if (!mat->was_assembled && mode == MAT_FINAL_ASSEMBLY) { 677 ierr = MatSetUpMultiply_MPISBAIJ(mat);CHKERRQ(ierr); /* setup Mvctx and sMvctx */ 678 } 679 b->compressedrow.use = PETSC_TRUE; 680 ierr = MatAssemblyBegin(baij->B,mode);CHKERRQ(ierr); 681 ierr = MatAssemblyEnd(baij->B,mode);CHKERRQ(ierr); 682 683 if (baij->rowvalues) { 684 ierr = PetscFree(baij->rowvalues);CHKERRQ(ierr); 685 baij->rowvalues = 0; 686 } 687 688 PetscFunctionReturn(0); 689 } 690 691 #undef __FUNCT__ 692 #define __FUNCT__ "MatView_MPISBAIJ_ASCIIorDraworSocket" 693 static PetscErrorCode MatView_MPISBAIJ_ASCIIorDraworSocket(Mat mat,PetscViewer viewer) 694 { 695 Mat_MPISBAIJ *baij = (Mat_MPISBAIJ*)mat->data; 696 PetscErrorCode ierr; 697 PetscInt bs = mat->bs; 698 PetscMPIInt size = baij->size,rank = baij->rank; 699 PetscTruth iascii,isdraw; 700 PetscViewer sviewer; 701 PetscViewerFormat format; 702 703 PetscFunctionBegin; 704 ierr = PetscTypeCompare((PetscObject)viewer,PETSC_VIEWER_ASCII,&iascii);CHKERRQ(ierr); 705 ierr = PetscTypeCompare((PetscObject)viewer,PETSC_VIEWER_DRAW,&isdraw);CHKERRQ(ierr); 706 if (iascii) { 707 ierr = PetscViewerGetFormat(viewer,&format);CHKERRQ(ierr); 708 if (format == PETSC_VIEWER_ASCII_INFO_DETAIL) { 709 MatInfo info; 710 ierr = MPI_Comm_rank(mat->comm,&rank);CHKERRQ(ierr); 711 ierr = MatGetInfo(mat,MAT_LOCAL,&info);CHKERRQ(ierr); 712 ierr = PetscViewerASCIISynchronizedPrintf(viewer,"[%d] Local rows %D nz %D nz alloced %D bs %D mem %D\n", 713 rank,mat->m,(PetscInt)info.nz_used*bs,(PetscInt)info.nz_allocated*bs, 714 mat->bs,(PetscInt)info.memory);CHKERRQ(ierr); 715 ierr = MatGetInfo(baij->A,MAT_LOCAL,&info);CHKERRQ(ierr); 716 ierr = PetscViewerASCIISynchronizedPrintf(viewer,"[%d] on-diagonal part: nz %D \n",rank,(PetscInt)info.nz_used*bs);CHKERRQ(ierr); 717 ierr = MatGetInfo(baij->B,MAT_LOCAL,&info);CHKERRQ(ierr); 718 ierr = PetscViewerASCIISynchronizedPrintf(viewer,"[%d] off-diagonal part: nz %D \n",rank,(PetscInt)info.nz_used*bs);CHKERRQ(ierr); 719 ierr = PetscViewerFlush(viewer);CHKERRQ(ierr); 720 ierr = VecScatterView(baij->Mvctx,viewer);CHKERRQ(ierr); 721 PetscFunctionReturn(0); 722 } else if (format == PETSC_VIEWER_ASCII_INFO) { 723 ierr = PetscViewerASCIIPrintf(viewer," block size is %D\n",bs);CHKERRQ(ierr); 724 PetscFunctionReturn(0); 725 } 726 } 727 728 if (isdraw) { 729 PetscDraw draw; 730 PetscTruth isnull; 731 ierr = PetscViewerDrawGetDraw(viewer,0,&draw);CHKERRQ(ierr); 732 ierr = PetscDrawIsNull(draw,&isnull);CHKERRQ(ierr); if (isnull) PetscFunctionReturn(0); 733 } 734 735 if (size == 1) { 736 ierr = PetscObjectSetName((PetscObject)baij->A,mat->name);CHKERRQ(ierr); 737 ierr = MatView(baij->A,viewer);CHKERRQ(ierr); 738 } else { 739 /* assemble the entire matrix onto first processor. */ 740 Mat A; 741 Mat_SeqSBAIJ *Aloc; 742 Mat_SeqBAIJ *Bloc; 743 PetscInt M = mat->M,N = mat->N,*ai,*aj,col,i,j,k,*rvals,mbs = baij->mbs; 744 MatScalar *a; 745 746 /* Should this be the same type as mat? */ 747 if (!rank) { 748 ierr = MatCreate(mat->comm,M,N,M,N,&A);CHKERRQ(ierr); 749 } else { 750 ierr = MatCreate(mat->comm,0,0,M,N,&A);CHKERRQ(ierr); 751 } 752 ierr = MatSetType(A,MATMPISBAIJ);CHKERRQ(ierr); 753 ierr = MatMPISBAIJSetPreallocation(A,mat->bs,0,PETSC_NULL,0,PETSC_NULL);CHKERRQ(ierr); 754 ierr = PetscLogObjectParent(mat,A);CHKERRQ(ierr); 755 756 /* copy over the A part */ 757 Aloc = (Mat_SeqSBAIJ*)baij->A->data; 758 ai = Aloc->i; aj = Aloc->j; a = Aloc->a; 759 ierr = PetscMalloc(bs*sizeof(PetscInt),&rvals);CHKERRQ(ierr); 760 761 for (i=0; i<mbs; i++) { 762 rvals[0] = bs*(baij->rstart + i); 763 for (j=1; j<bs; j++) { rvals[j] = rvals[j-1] + 1; } 764 for (j=ai[i]; j<ai[i+1]; j++) { 765 col = (baij->cstart+aj[j])*bs; 766 for (k=0; k<bs; k++) { 767 ierr = MatSetValues_MPISBAIJ_MatScalar(A,bs,rvals,1,&col,a,INSERT_VALUES);CHKERRQ(ierr); 768 col++; a += bs; 769 } 770 } 771 } 772 /* copy over the B part */ 773 Bloc = (Mat_SeqBAIJ*)baij->B->data; 774 ai = Bloc->i; aj = Bloc->j; a = Bloc->a; 775 for (i=0; i<mbs; i++) { 776 rvals[0] = bs*(baij->rstart + i); 777 for (j=1; j<bs; j++) { rvals[j] = rvals[j-1] + 1; } 778 for (j=ai[i]; j<ai[i+1]; j++) { 779 col = baij->garray[aj[j]]*bs; 780 for (k=0; k<bs; k++) { 781 ierr = MatSetValues_MPISBAIJ_MatScalar(A,bs,rvals,1,&col,a,INSERT_VALUES);CHKERRQ(ierr); 782 col++; a += bs; 783 } 784 } 785 } 786 ierr = PetscFree(rvals);CHKERRQ(ierr); 787 ierr = MatAssemblyBegin(A,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 788 ierr = MatAssemblyEnd(A,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 789 /* 790 Everyone has to call to draw the matrix since the graphics waits are 791 synchronized across all processors that share the PetscDraw object 792 */ 793 ierr = PetscViewerGetSingleton(viewer,&sviewer);CHKERRQ(ierr); 794 if (!rank) { 795 ierr = PetscObjectSetName((PetscObject)((Mat_MPISBAIJ*)(A->data))->A,mat->name);CHKERRQ(ierr); 796 ierr = MatView(((Mat_MPISBAIJ*)(A->data))->A,sviewer);CHKERRQ(ierr); 797 } 798 ierr = PetscViewerRestoreSingleton(viewer,&sviewer);CHKERRQ(ierr); 799 ierr = MatDestroy(A);CHKERRQ(ierr); 800 } 801 PetscFunctionReturn(0); 802 } 803 804 #undef __FUNCT__ 805 #define __FUNCT__ "MatView_MPISBAIJ" 806 PetscErrorCode MatView_MPISBAIJ(Mat mat,PetscViewer viewer) 807 { 808 PetscErrorCode ierr; 809 PetscTruth iascii,isdraw,issocket,isbinary; 810 811 PetscFunctionBegin; 812 ierr = PetscTypeCompare((PetscObject)viewer,PETSC_VIEWER_ASCII,&iascii);CHKERRQ(ierr); 813 ierr = PetscTypeCompare((PetscObject)viewer,PETSC_VIEWER_DRAW,&isdraw);CHKERRQ(ierr); 814 ierr = PetscTypeCompare((PetscObject)viewer,PETSC_VIEWER_SOCKET,&issocket);CHKERRQ(ierr); 815 ierr = PetscTypeCompare((PetscObject)viewer,PETSC_VIEWER_BINARY,&isbinary);CHKERRQ(ierr); 816 if (iascii || isdraw || issocket || isbinary) { 817 ierr = MatView_MPISBAIJ_ASCIIorDraworSocket(mat,viewer);CHKERRQ(ierr); 818 } else { 819 SETERRQ1(PETSC_ERR_SUP,"Viewer type %s not supported by MPISBAIJ matrices",((PetscObject)viewer)->type_name); 820 } 821 PetscFunctionReturn(0); 822 } 823 824 #undef __FUNCT__ 825 #define __FUNCT__ "MatDestroy_MPISBAIJ" 826 PetscErrorCode MatDestroy_MPISBAIJ(Mat mat) 827 { 828 Mat_MPISBAIJ *baij = (Mat_MPISBAIJ*)mat->data; 829 PetscErrorCode ierr; 830 831 PetscFunctionBegin; 832 #if defined(PETSC_USE_LOG) 833 PetscLogObjectState((PetscObject)mat,"Rows=%D,Cols=%D",mat->M,mat->N); 834 #endif 835 ierr = MatStashDestroy_Private(&mat->stash);CHKERRQ(ierr); 836 ierr = MatStashDestroy_Private(&mat->bstash);CHKERRQ(ierr); 837 ierr = PetscFree(baij->rowners);CHKERRQ(ierr); 838 ierr = MatDestroy(baij->A);CHKERRQ(ierr); 839 ierr = MatDestroy(baij->B);CHKERRQ(ierr); 840 #if defined (PETSC_USE_CTABLE) 841 if (baij->colmap) {ierr = PetscTableDelete(baij->colmap);CHKERRQ(ierr);} 842 #else 843 if (baij->colmap) {ierr = PetscFree(baij->colmap);CHKERRQ(ierr);} 844 #endif 845 if (baij->garray) {ierr = PetscFree(baij->garray);CHKERRQ(ierr);} 846 if (baij->lvec) {ierr = VecDestroy(baij->lvec);CHKERRQ(ierr);} 847 if (baij->Mvctx) {ierr = VecScatterDestroy(baij->Mvctx);CHKERRQ(ierr);} 848 if (baij->slvec0) { 849 ierr = VecDestroy(baij->slvec0);CHKERRQ(ierr); 850 ierr = VecDestroy(baij->slvec0b);CHKERRQ(ierr); 851 } 852 if (baij->slvec1) { 853 ierr = VecDestroy(baij->slvec1);CHKERRQ(ierr); 854 ierr = VecDestroy(baij->slvec1a);CHKERRQ(ierr); 855 ierr = VecDestroy(baij->slvec1b);CHKERRQ(ierr); 856 } 857 if (baij->sMvctx) {ierr = VecScatterDestroy(baij->sMvctx);CHKERRQ(ierr);} 858 if (baij->rowvalues) {ierr = PetscFree(baij->rowvalues);CHKERRQ(ierr);} 859 if (baij->barray) {ierr = PetscFree(baij->barray);CHKERRQ(ierr);} 860 if (baij->hd) {ierr = PetscFree(baij->hd);CHKERRQ(ierr);} 861 #if defined(PETSC_USE_MAT_SINGLE) 862 if (baij->setvaluescopy) {ierr = PetscFree(baij->setvaluescopy);CHKERRQ(ierr);} 863 #endif 864 ierr = PetscFree(baij);CHKERRQ(ierr); 865 866 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatStoreValues_C","",PETSC_NULL);CHKERRQ(ierr); 867 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatRetrieveValues_C","",PETSC_NULL);CHKERRQ(ierr); 868 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatGetDiagonalBlock_C","",PETSC_NULL);CHKERRQ(ierr); 869 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatMPISBAIJSetPreallocation_C","",PETSC_NULL);CHKERRQ(ierr); 870 PetscFunctionReturn(0); 871 } 872 873 #undef __FUNCT__ 874 #define __FUNCT__ "MatMult_MPISBAIJ" 875 PetscErrorCode MatMult_MPISBAIJ(Mat A,Vec xx,Vec yy) 876 { 877 Mat_MPISBAIJ *a = (Mat_MPISBAIJ*)A->data; 878 PetscErrorCode ierr; 879 PetscInt nt,mbs=a->mbs,bs=A->bs; 880 PetscScalar *x,*from,zero=0.0; 881 882 PetscFunctionBegin; 883 ierr = VecGetLocalSize(xx,&nt);CHKERRQ(ierr); 884 if (nt != A->n) { 885 SETERRQ(PETSC_ERR_ARG_SIZ,"Incompatible partition of A and xx"); 886 } 887 ierr = VecGetLocalSize(yy,&nt);CHKERRQ(ierr); 888 if (nt != A->m) { 889 SETERRQ(PETSC_ERR_ARG_SIZ,"Incompatible parition of A and yy"); 890 } 891 892 /* diagonal part */ 893 ierr = (*a->A->ops->mult)(a->A,xx,a->slvec1a);CHKERRQ(ierr); 894 ierr = VecSet(&zero,a->slvec1b);CHKERRQ(ierr); 895 896 /* subdiagonal part */ 897 ierr = (*a->B->ops->multtranspose)(a->B,xx,a->slvec0b);CHKERRQ(ierr); 898 899 /* copy x into the vec slvec0 */ 900 ierr = VecGetArray(a->slvec0,&from);CHKERRQ(ierr); 901 ierr = VecGetArray(xx,&x);CHKERRQ(ierr); 902 ierr = PetscMemcpy(from,x,bs*mbs*sizeof(MatScalar));CHKERRQ(ierr); 903 ierr = VecRestoreArray(a->slvec0,&from);CHKERRQ(ierr); 904 905 ierr = VecScatterBegin(a->slvec0,a->slvec1,ADD_VALUES,SCATTER_FORWARD,a->sMvctx);CHKERRQ(ierr); 906 ierr = VecRestoreArray(xx,&x);CHKERRQ(ierr); 907 ierr = VecScatterEnd(a->slvec0,a->slvec1,ADD_VALUES,SCATTER_FORWARD,a->sMvctx);CHKERRQ(ierr); 908 909 /* supperdiagonal part */ 910 ierr = (*a->B->ops->multadd)(a->B,a->slvec1b,a->slvec1a,yy);CHKERRQ(ierr); 911 912 PetscFunctionReturn(0); 913 } 914 915 #undef __FUNCT__ 916 #define __FUNCT__ "MatMult_MPISBAIJ_2comm" 917 PetscErrorCode MatMult_MPISBAIJ_2comm(Mat A,Vec xx,Vec yy) 918 { 919 Mat_MPISBAIJ *a = (Mat_MPISBAIJ*)A->data; 920 PetscErrorCode ierr; 921 PetscInt nt; 922 923 PetscFunctionBegin; 924 ierr = VecGetLocalSize(xx,&nt);CHKERRQ(ierr); 925 if (nt != A->n) { 926 SETERRQ(PETSC_ERR_ARG_SIZ,"Incompatible partition of A and xx"); 927 } 928 ierr = VecGetLocalSize(yy,&nt);CHKERRQ(ierr); 929 if (nt != A->m) { 930 SETERRQ(PETSC_ERR_ARG_SIZ,"Incompatible parition of A and yy"); 931 } 932 933 ierr = VecScatterBegin(xx,a->lvec,INSERT_VALUES,SCATTER_FORWARD,a->Mvctx);CHKERRQ(ierr); 934 /* do diagonal part */ 935 ierr = (*a->A->ops->mult)(a->A,xx,yy);CHKERRQ(ierr); 936 /* do supperdiagonal part */ 937 ierr = VecScatterEnd(xx,a->lvec,INSERT_VALUES,SCATTER_FORWARD,a->Mvctx);CHKERRQ(ierr); 938 ierr = (*a->B->ops->multadd)(a->B,a->lvec,yy,yy);CHKERRQ(ierr); 939 /* do subdiagonal part */ 940 ierr = (*a->B->ops->multtranspose)(a->B,xx,a->lvec);CHKERRQ(ierr); 941 ierr = VecScatterBegin(a->lvec,yy,ADD_VALUES,SCATTER_REVERSE,a->Mvctx);CHKERRQ(ierr); 942 ierr = VecScatterEnd(a->lvec,yy,ADD_VALUES,SCATTER_REVERSE,a->Mvctx);CHKERRQ(ierr); 943 944 PetscFunctionReturn(0); 945 } 946 947 #undef __FUNCT__ 948 #define __FUNCT__ "MatMultAdd_MPISBAIJ" 949 PetscErrorCode MatMultAdd_MPISBAIJ(Mat A,Vec xx,Vec yy,Vec zz) 950 { 951 Mat_MPISBAIJ *a = (Mat_MPISBAIJ*)A->data; 952 PetscErrorCode ierr; 953 PetscInt mbs=a->mbs,bs=A->bs; 954 PetscScalar *x,*from,zero=0.0; 955 956 PetscFunctionBegin; 957 /* 958 PetscSynchronizedPrintf(A->comm," MatMultAdd is called ...\n"); 959 PetscSynchronizedFlush(A->comm); 960 */ 961 /* diagonal part */ 962 ierr = (*a->A->ops->multadd)(a->A,xx,yy,a->slvec1a);CHKERRQ(ierr); 963 ierr = VecSet(&zero,a->slvec1b);CHKERRQ(ierr); 964 965 /* subdiagonal part */ 966 ierr = (*a->B->ops->multtranspose)(a->B,xx,a->slvec0b);CHKERRQ(ierr); 967 968 /* copy x into the vec slvec0 */ 969 ierr = VecGetArray(a->slvec0,&from);CHKERRQ(ierr); 970 ierr = VecGetArray(xx,&x);CHKERRQ(ierr); 971 ierr = PetscMemcpy(from,x,bs*mbs*sizeof(MatScalar));CHKERRQ(ierr); 972 ierr = VecRestoreArray(a->slvec0,&from);CHKERRQ(ierr); 973 974 ierr = VecScatterBegin(a->slvec0,a->slvec1,ADD_VALUES,SCATTER_FORWARD,a->sMvctx);CHKERRQ(ierr); 975 ierr = VecRestoreArray(xx,&x);CHKERRQ(ierr); 976 ierr = VecScatterEnd(a->slvec0,a->slvec1,ADD_VALUES,SCATTER_FORWARD,a->sMvctx);CHKERRQ(ierr); 977 978 /* supperdiagonal part */ 979 ierr = (*a->B->ops->multadd)(a->B,a->slvec1b,a->slvec1a,zz);CHKERRQ(ierr); 980 981 PetscFunctionReturn(0); 982 } 983 984 #undef __FUNCT__ 985 #define __FUNCT__ "MatMultAdd_MPISBAIJ_2comm" 986 PetscErrorCode MatMultAdd_MPISBAIJ_2comm(Mat A,Vec xx,Vec yy,Vec zz) 987 { 988 Mat_MPISBAIJ *a = (Mat_MPISBAIJ*)A->data; 989 PetscErrorCode ierr; 990 991 PetscFunctionBegin; 992 ierr = VecScatterBegin(xx,a->lvec,INSERT_VALUES,SCATTER_FORWARD,a->Mvctx);CHKERRQ(ierr); 993 /* do diagonal part */ 994 ierr = (*a->A->ops->multadd)(a->A,xx,yy,zz);CHKERRQ(ierr); 995 /* do supperdiagonal part */ 996 ierr = VecScatterEnd(xx,a->lvec,INSERT_VALUES,SCATTER_FORWARD,a->Mvctx);CHKERRQ(ierr); 997 ierr = (*a->B->ops->multadd)(a->B,a->lvec,zz,zz);CHKERRQ(ierr); 998 999 /* do subdiagonal part */ 1000 ierr = (*a->B->ops->multtranspose)(a->B,xx,a->lvec);CHKERRQ(ierr); 1001 ierr = VecScatterBegin(a->lvec,zz,ADD_VALUES,SCATTER_REVERSE,a->Mvctx);CHKERRQ(ierr); 1002 ierr = VecScatterEnd(a->lvec,zz,ADD_VALUES,SCATTER_REVERSE,a->Mvctx);CHKERRQ(ierr); 1003 1004 PetscFunctionReturn(0); 1005 } 1006 1007 #undef __FUNCT__ 1008 #define __FUNCT__ "MatMultTranspose_MPISBAIJ" 1009 PetscErrorCode MatMultTranspose_MPISBAIJ(Mat A,Vec xx,Vec yy) 1010 { 1011 PetscErrorCode ierr; 1012 1013 PetscFunctionBegin; 1014 ierr = MatMult(A,xx,yy);CHKERRQ(ierr); 1015 PetscFunctionReturn(0); 1016 } 1017 1018 #undef __FUNCT__ 1019 #define __FUNCT__ "MatMultTransposeAdd_MPISBAIJ" 1020 PetscErrorCode MatMultTransposeAdd_MPISBAIJ(Mat A,Vec xx,Vec yy,Vec zz) 1021 { 1022 PetscErrorCode ierr; 1023 1024 PetscFunctionBegin; 1025 ierr = MatMultAdd(A,xx,yy,zz);CHKERRQ(ierr); 1026 PetscFunctionReturn(0); 1027 } 1028 1029 /* 1030 This only works correctly for square matrices where the subblock A->A is the 1031 diagonal block 1032 */ 1033 #undef __FUNCT__ 1034 #define __FUNCT__ "MatGetDiagonal_MPISBAIJ" 1035 PetscErrorCode MatGetDiagonal_MPISBAIJ(Mat A,Vec v) 1036 { 1037 Mat_MPISBAIJ *a = (Mat_MPISBAIJ*)A->data; 1038 PetscErrorCode ierr; 1039 1040 PetscFunctionBegin; 1041 /* if (a->M != a->N) SETERRQ(PETSC_ERR_SUP,"Supports only square matrix where A->A is diag block"); */ 1042 ierr = MatGetDiagonal(a->A,v);CHKERRQ(ierr); 1043 PetscFunctionReturn(0); 1044 } 1045 1046 #undef __FUNCT__ 1047 #define __FUNCT__ "MatScale_MPISBAIJ" 1048 PetscErrorCode MatScale_MPISBAIJ(const PetscScalar *aa,Mat A) 1049 { 1050 Mat_MPISBAIJ *a = (Mat_MPISBAIJ*)A->data; 1051 PetscErrorCode ierr; 1052 1053 PetscFunctionBegin; 1054 ierr = MatScale(aa,a->A);CHKERRQ(ierr); 1055 ierr = MatScale(aa,a->B);CHKERRQ(ierr); 1056 PetscFunctionReturn(0); 1057 } 1058 1059 #undef __FUNCT__ 1060 #define __FUNCT__ "MatGetRow_MPISBAIJ" 1061 PetscErrorCode MatGetRow_MPISBAIJ(Mat matin,PetscInt row,PetscInt *nz,PetscInt **idx,PetscScalar **v) 1062 { 1063 PetscFunctionBegin; 1064 SETERRQ(PETSC_ERR_SUP,"MatGetRow is not supported for SBAIJ matrix format"); 1065 PetscFunctionReturn(0); 1066 } 1067 1068 #undef __FUNCT__ 1069 #define __FUNCT__ "MatRestoreRow_MPISBAIJ" 1070 PetscErrorCode MatRestoreRow_MPISBAIJ(Mat mat,PetscInt row,PetscInt *nz,PetscInt **idx,PetscScalar **v) 1071 { 1072 Mat_MPISBAIJ *baij = (Mat_MPISBAIJ*)mat->data; 1073 1074 PetscFunctionBegin; 1075 if (!baij->getrowactive) { 1076 SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"MatGetRow() must be called first"); 1077 } 1078 baij->getrowactive = PETSC_FALSE; 1079 PetscFunctionReturn(0); 1080 } 1081 1082 #undef __FUNCT__ 1083 #define __FUNCT__ "MatZeroEntries_MPISBAIJ" 1084 PetscErrorCode MatZeroEntries_MPISBAIJ(Mat A) 1085 { 1086 Mat_MPISBAIJ *l = (Mat_MPISBAIJ*)A->data; 1087 PetscErrorCode ierr; 1088 1089 PetscFunctionBegin; 1090 ierr = MatZeroEntries(l->A);CHKERRQ(ierr); 1091 ierr = MatZeroEntries(l->B);CHKERRQ(ierr); 1092 PetscFunctionReturn(0); 1093 } 1094 1095 #undef __FUNCT__ 1096 #define __FUNCT__ "MatGetInfo_MPISBAIJ" 1097 PetscErrorCode MatGetInfo_MPISBAIJ(Mat matin,MatInfoType flag,MatInfo *info) 1098 { 1099 Mat_MPISBAIJ *a = (Mat_MPISBAIJ*)matin->data; 1100 Mat A = a->A,B = a->B; 1101 PetscErrorCode ierr; 1102 PetscReal isend[5],irecv[5]; 1103 1104 PetscFunctionBegin; 1105 info->block_size = (PetscReal)matin->bs; 1106 ierr = MatGetInfo(A,MAT_LOCAL,info);CHKERRQ(ierr); 1107 isend[0] = info->nz_used; isend[1] = info->nz_allocated; isend[2] = info->nz_unneeded; 1108 isend[3] = info->memory; isend[4] = info->mallocs; 1109 ierr = MatGetInfo(B,MAT_LOCAL,info);CHKERRQ(ierr); 1110 isend[0] += info->nz_used; isend[1] += info->nz_allocated; isend[2] += info->nz_unneeded; 1111 isend[3] += info->memory; isend[4] += info->mallocs; 1112 if (flag == MAT_LOCAL) { 1113 info->nz_used = isend[0]; 1114 info->nz_allocated = isend[1]; 1115 info->nz_unneeded = isend[2]; 1116 info->memory = isend[3]; 1117 info->mallocs = isend[4]; 1118 } else if (flag == MAT_GLOBAL_MAX) { 1119 ierr = MPI_Allreduce(isend,irecv,5,MPIU_REAL,MPI_MAX,matin->comm);CHKERRQ(ierr); 1120 info->nz_used = irecv[0]; 1121 info->nz_allocated = irecv[1]; 1122 info->nz_unneeded = irecv[2]; 1123 info->memory = irecv[3]; 1124 info->mallocs = irecv[4]; 1125 } else if (flag == MAT_GLOBAL_SUM) { 1126 ierr = MPI_Allreduce(isend,irecv,5,MPIU_REAL,MPI_SUM,matin->comm);CHKERRQ(ierr); 1127 info->nz_used = irecv[0]; 1128 info->nz_allocated = irecv[1]; 1129 info->nz_unneeded = irecv[2]; 1130 info->memory = irecv[3]; 1131 info->mallocs = irecv[4]; 1132 } else { 1133 SETERRQ1(PETSC_ERR_ARG_WRONG,"Unknown MatInfoType argument %d",(int)flag); 1134 } 1135 info->rows_global = (PetscReal)A->M; 1136 info->columns_global = (PetscReal)A->N; 1137 info->rows_local = (PetscReal)A->m; 1138 info->columns_local = (PetscReal)A->N; 1139 info->fill_ratio_given = 0; /* no parallel LU/ILU/Cholesky */ 1140 info->fill_ratio_needed = 0; 1141 info->factor_mallocs = 0; 1142 PetscFunctionReturn(0); 1143 } 1144 1145 #undef __FUNCT__ 1146 #define __FUNCT__ "MatSetOption_MPISBAIJ" 1147 PetscErrorCode MatSetOption_MPISBAIJ(Mat A,MatOption op) 1148 { 1149 Mat_MPISBAIJ *a = (Mat_MPISBAIJ*)A->data; 1150 PetscErrorCode ierr; 1151 1152 PetscFunctionBegin; 1153 switch (op) { 1154 case MAT_NO_NEW_NONZERO_LOCATIONS: 1155 case MAT_YES_NEW_NONZERO_LOCATIONS: 1156 case MAT_COLUMNS_UNSORTED: 1157 case MAT_COLUMNS_SORTED: 1158 case MAT_NEW_NONZERO_ALLOCATION_ERR: 1159 case MAT_KEEP_ZEROED_ROWS: 1160 case MAT_NEW_NONZERO_LOCATION_ERR: 1161 ierr = MatSetOption(a->A,op);CHKERRQ(ierr); 1162 ierr = MatSetOption(a->B,op);CHKERRQ(ierr); 1163 break; 1164 case MAT_ROW_ORIENTED: 1165 a->roworiented = PETSC_TRUE; 1166 ierr = MatSetOption(a->A,op);CHKERRQ(ierr); 1167 ierr = MatSetOption(a->B,op);CHKERRQ(ierr); 1168 break; 1169 case MAT_ROWS_SORTED: 1170 case MAT_ROWS_UNSORTED: 1171 case MAT_YES_NEW_DIAGONALS: 1172 PetscLogInfo(A,"Info:MatSetOption_MPIBAIJ:Option ignored\n"); 1173 break; 1174 case MAT_COLUMN_ORIENTED: 1175 a->roworiented = PETSC_FALSE; 1176 ierr = MatSetOption(a->A,op);CHKERRQ(ierr); 1177 ierr = MatSetOption(a->B,op);CHKERRQ(ierr); 1178 break; 1179 case MAT_IGNORE_OFF_PROC_ENTRIES: 1180 a->donotstash = PETSC_TRUE; 1181 break; 1182 case MAT_NO_NEW_DIAGONALS: 1183 SETERRQ(PETSC_ERR_SUP,"MAT_NO_NEW_DIAGONALS"); 1184 case MAT_USE_HASH_TABLE: 1185 a->ht_flag = PETSC_TRUE; 1186 break; 1187 case MAT_NOT_SYMMETRIC: 1188 case MAT_NOT_STRUCTURALLY_SYMMETRIC: 1189 case MAT_HERMITIAN: 1190 SETERRQ(PETSC_ERR_SUP,"Matrix must be symmetric"); 1191 case MAT_SYMMETRIC: 1192 case MAT_STRUCTURALLY_SYMMETRIC: 1193 case MAT_NOT_HERMITIAN: 1194 case MAT_SYMMETRY_ETERNAL: 1195 case MAT_NOT_SYMMETRY_ETERNAL: 1196 break; 1197 default: 1198 SETERRQ(PETSC_ERR_SUP,"unknown option"); 1199 } 1200 PetscFunctionReturn(0); 1201 } 1202 1203 #undef __FUNCT__ 1204 #define __FUNCT__ "MatTranspose_MPISBAIJ" 1205 PetscErrorCode MatTranspose_MPISBAIJ(Mat A,Mat *B) 1206 { 1207 PetscErrorCode ierr; 1208 PetscFunctionBegin; 1209 ierr = MatDuplicate(A,MAT_COPY_VALUES,B);CHKERRQ(ierr); 1210 PetscFunctionReturn(0); 1211 } 1212 1213 #undef __FUNCT__ 1214 #define __FUNCT__ "MatDiagonalScale_MPISBAIJ" 1215 PetscErrorCode MatDiagonalScale_MPISBAIJ(Mat mat,Vec ll,Vec rr) 1216 { 1217 Mat_MPISBAIJ *baij = (Mat_MPISBAIJ*)mat->data; 1218 Mat a=baij->A, b=baij->B; 1219 PetscErrorCode ierr; 1220 PetscInt nv,m,n; 1221 PetscTruth flg; 1222 1223 PetscFunctionBegin; 1224 if (ll != rr){ 1225 ierr = VecEqual(ll,rr,&flg);CHKERRQ(ierr); 1226 if (!flg) 1227 SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,"For symmetric format, left and right scaling vectors must be same\n"); 1228 } 1229 if (!ll) PetscFunctionReturn(0); 1230 1231 ierr = MatGetLocalSize(mat,&m,&n);CHKERRQ(ierr); 1232 if (m != n) SETERRQ2(PETSC_ERR_ARG_SIZ,"For symmetric format, local size %d %d must be same",m,n); 1233 1234 ierr = VecGetLocalSize(rr,&nv);CHKERRQ(ierr); 1235 if (nv!=n) SETERRQ(PETSC_ERR_ARG_SIZ,"Left and right vector non-conforming local size"); 1236 1237 ierr = VecScatterBegin(rr,baij->lvec,INSERT_VALUES,SCATTER_FORWARD,baij->Mvctx);CHKERRQ(ierr); 1238 1239 /* left diagonalscale the off-diagonal part */ 1240 ierr = (*b->ops->diagonalscale)(b,ll,PETSC_NULL);CHKERRQ(ierr); 1241 1242 /* scale the diagonal part */ 1243 ierr = (*a->ops->diagonalscale)(a,ll,rr);CHKERRQ(ierr); 1244 1245 /* right diagonalscale the off-diagonal part */ 1246 ierr = VecScatterEnd(rr,baij->lvec,INSERT_VALUES,SCATTER_FORWARD,baij->Mvctx);CHKERRQ(ierr); 1247 ierr = (*b->ops->diagonalscale)(b,PETSC_NULL,baij->lvec);CHKERRQ(ierr); 1248 PetscFunctionReturn(0); 1249 } 1250 1251 #undef __FUNCT__ 1252 #define __FUNCT__ "MatPrintHelp_MPISBAIJ" 1253 PetscErrorCode MatPrintHelp_MPISBAIJ(Mat A) 1254 { 1255 Mat_MPISBAIJ *a = (Mat_MPISBAIJ*)A->data; 1256 MPI_Comm comm = A->comm; 1257 static PetscTruth called = PETSC_FALSE; 1258 PetscErrorCode ierr; 1259 1260 PetscFunctionBegin; 1261 if (!a->rank) { 1262 ierr = MatPrintHelp_SeqSBAIJ(a->A);CHKERRQ(ierr); 1263 } 1264 if (called) {PetscFunctionReturn(0);} else called = PETSC_TRUE; 1265 ierr = (*PetscHelpPrintf)(comm," Options for MATMPISBAIJ matrix format (the defaults):\n");CHKERRQ(ierr); 1266 ierr = (*PetscHelpPrintf)(comm," -mat_use_hash_table <factor>: Use hashtable for efficient matrix assembly\n");CHKERRQ(ierr); 1267 PetscFunctionReturn(0); 1268 } 1269 1270 #undef __FUNCT__ 1271 #define __FUNCT__ "MatSetUnfactored_MPISBAIJ" 1272 PetscErrorCode MatSetUnfactored_MPISBAIJ(Mat A) 1273 { 1274 Mat_MPISBAIJ *a = (Mat_MPISBAIJ*)A->data; 1275 PetscErrorCode ierr; 1276 1277 PetscFunctionBegin; 1278 ierr = MatSetUnfactored(a->A);CHKERRQ(ierr); 1279 PetscFunctionReturn(0); 1280 } 1281 1282 static PetscErrorCode MatDuplicate_MPISBAIJ(Mat,MatDuplicateOption,Mat *); 1283 1284 #undef __FUNCT__ 1285 #define __FUNCT__ "MatEqual_MPISBAIJ" 1286 PetscErrorCode MatEqual_MPISBAIJ(Mat A,Mat B,PetscTruth *flag) 1287 { 1288 Mat_MPISBAIJ *matB = (Mat_MPISBAIJ*)B->data,*matA = (Mat_MPISBAIJ*)A->data; 1289 Mat a,b,c,d; 1290 PetscTruth flg; 1291 PetscErrorCode ierr; 1292 1293 PetscFunctionBegin; 1294 a = matA->A; b = matA->B; 1295 c = matB->A; d = matB->B; 1296 1297 ierr = MatEqual(a,c,&flg);CHKERRQ(ierr); 1298 if (flg) { 1299 ierr = MatEqual(b,d,&flg);CHKERRQ(ierr); 1300 } 1301 ierr = MPI_Allreduce(&flg,flag,1,MPI_INT,MPI_LAND,A->comm);CHKERRQ(ierr); 1302 PetscFunctionReturn(0); 1303 } 1304 1305 #undef __FUNCT__ 1306 #define __FUNCT__ "MatSetUpPreallocation_MPISBAIJ" 1307 PetscErrorCode MatSetUpPreallocation_MPISBAIJ(Mat A) 1308 { 1309 PetscErrorCode ierr; 1310 1311 PetscFunctionBegin; 1312 ierr = MatMPISBAIJSetPreallocation(A,1,PETSC_DEFAULT,0,PETSC_DEFAULT,0);CHKERRQ(ierr); 1313 PetscFunctionReturn(0); 1314 } 1315 1316 #undef __FUNCT__ 1317 #define __FUNCT__ "MatGetSubMatrices_MPISBAIJ" 1318 PetscErrorCode MatGetSubMatrices_MPISBAIJ(Mat A,PetscInt n,const IS irow[],const IS icol[],MatReuse scall,Mat *B[]) 1319 { 1320 PetscErrorCode ierr; 1321 PetscInt i; 1322 PetscTruth flg; 1323 1324 PetscFunctionBegin; 1325 for (i=0; i<n; i++) { 1326 ierr = ISEqual(irow[i],icol[i],&flg);CHKERRQ(ierr); 1327 if (!flg) { 1328 SETERRQ(PETSC_ERR_SUP,"Can only get symmetric submatrix for MPISBAIJ matrices"); 1329 } 1330 } 1331 ierr = MatGetSubMatrices_MPIBAIJ(A,n,irow,icol,scall,B);CHKERRQ(ierr); 1332 PetscFunctionReturn(0); 1333 } 1334 1335 1336 /* -------------------------------------------------------------------*/ 1337 static struct _MatOps MatOps_Values = { 1338 MatSetValues_MPISBAIJ, 1339 MatGetRow_MPISBAIJ, 1340 MatRestoreRow_MPISBAIJ, 1341 MatMult_MPISBAIJ, 1342 /* 4*/ MatMultAdd_MPISBAIJ, 1343 MatMultTranspose_MPISBAIJ, 1344 MatMultTransposeAdd_MPISBAIJ, 1345 0, 1346 0, 1347 0, 1348 /*10*/ 0, 1349 0, 1350 0, 1351 MatRelax_MPISBAIJ, 1352 MatTranspose_MPISBAIJ, 1353 /*15*/ MatGetInfo_MPISBAIJ, 1354 MatEqual_MPISBAIJ, 1355 MatGetDiagonal_MPISBAIJ, 1356 MatDiagonalScale_MPISBAIJ, 1357 MatNorm_MPISBAIJ, 1358 /*20*/ MatAssemblyBegin_MPISBAIJ, 1359 MatAssemblyEnd_MPISBAIJ, 1360 0, 1361 MatSetOption_MPISBAIJ, 1362 MatZeroEntries_MPISBAIJ, 1363 /*25*/ 0, 1364 0, 1365 0, 1366 0, 1367 0, 1368 /*30*/ MatSetUpPreallocation_MPISBAIJ, 1369 0, 1370 0, 1371 0, 1372 0, 1373 /*35*/ MatDuplicate_MPISBAIJ, 1374 0, 1375 0, 1376 0, 1377 0, 1378 /*40*/ 0, 1379 MatGetSubMatrices_MPISBAIJ, 1380 MatIncreaseOverlap_MPISBAIJ, 1381 MatGetValues_MPISBAIJ, 1382 0, 1383 /*45*/ MatPrintHelp_MPISBAIJ, 1384 MatScale_MPISBAIJ, 1385 0, 1386 0, 1387 0, 1388 /*50*/ 0, 1389 0, 1390 0, 1391 0, 1392 0, 1393 /*55*/ 0, 1394 0, 1395 MatSetUnfactored_MPISBAIJ, 1396 0, 1397 MatSetValuesBlocked_MPISBAIJ, 1398 /*60*/ 0, 1399 0, 1400 0, 1401 MatGetPetscMaps_Petsc, 1402 0, 1403 /*65*/ 0, 1404 0, 1405 0, 1406 0, 1407 0, 1408 /*70*/ MatGetRowMax_MPISBAIJ, 1409 0, 1410 0, 1411 0, 1412 0, 1413 /*75*/ 0, 1414 0, 1415 0, 1416 0, 1417 0, 1418 /*80*/ 0, 1419 0, 1420 0, 1421 0, 1422 MatLoad_MPISBAIJ, 1423 /*85*/ 0, 1424 0, 1425 0, 1426 0, 1427 0, 1428 /*90*/ 0, 1429 0, 1430 0, 1431 0, 1432 0, 1433 /*95*/ 0, 1434 0, 1435 0, 1436 0}; 1437 1438 1439 EXTERN_C_BEGIN 1440 #undef __FUNCT__ 1441 #define __FUNCT__ "MatGetDiagonalBlock_MPISBAIJ" 1442 PetscErrorCode MatGetDiagonalBlock_MPISBAIJ(Mat A,PetscTruth *iscopy,MatReuse reuse,Mat *a) 1443 { 1444 PetscFunctionBegin; 1445 *a = ((Mat_MPISBAIJ *)A->data)->A; 1446 *iscopy = PETSC_FALSE; 1447 PetscFunctionReturn(0); 1448 } 1449 EXTERN_C_END 1450 1451 EXTERN_C_BEGIN 1452 #undef __FUNCT__ 1453 #define __FUNCT__ "MatMPISBAIJSetPreallocation_MPISBAIJ" 1454 PetscErrorCode MatMPISBAIJSetPreallocation_MPISBAIJ(Mat B,PetscInt bs,PetscInt d_nz,PetscInt *d_nnz,PetscInt o_nz,PetscInt *o_nnz) 1455 { 1456 Mat_MPISBAIJ *b; 1457 PetscErrorCode ierr; 1458 PetscInt i,mbs,Mbs; 1459 1460 PetscFunctionBegin; 1461 ierr = PetscOptionsGetInt(B->prefix,"-mat_block_size",&bs,PETSC_NULL);CHKERRQ(ierr); 1462 1463 if (bs < 1) SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,"Invalid block size specified, must be positive"); 1464 if (d_nz == PETSC_DECIDE || d_nz == PETSC_DEFAULT) d_nz = 3; 1465 if (o_nz == PETSC_DECIDE || o_nz == PETSC_DEFAULT) o_nz = 1; 1466 if (d_nz < 0) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"d_nz cannot be less than 0: value %D",d_nz); 1467 if (o_nz < 0) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"o_nz cannot be less than 0: value %D",o_nz); 1468 if (d_nnz) { 1469 for (i=0; i<B->m/bs; i++) { 1470 if (d_nnz[i] < 0) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"d_nnz cannot be less than -1: local row %D value %D",i,d_nnz[i]); 1471 } 1472 } 1473 if (o_nnz) { 1474 for (i=0; i<B->m/bs; i++) { 1475 if (o_nnz[i] < 0) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"o_nnz cannot be less than -1: local row %D value %D",i,o_nnz[i]); 1476 } 1477 } 1478 B->preallocated = PETSC_TRUE; 1479 ierr = PetscSplitOwnershipBlock(B->comm,bs,&B->m,&B->M);CHKERRQ(ierr); 1480 ierr = PetscSplitOwnershipBlock(B->comm,bs,&B->n,&B->N);CHKERRQ(ierr); 1481 ierr = PetscMapCreateMPI(B->comm,B->m,B->M,&B->rmap);CHKERRQ(ierr); 1482 ierr = PetscMapCreateMPI(B->comm,B->m,B->M,&B->cmap);CHKERRQ(ierr); 1483 1484 b = (Mat_MPISBAIJ*)B->data; 1485 mbs = B->m/bs; 1486 Mbs = B->M/bs; 1487 if (mbs*bs != B->m) { 1488 SETERRQ2(PETSC_ERR_ARG_SIZ,"No of local rows %D must be divisible by blocksize %D",B->m,bs); 1489 } 1490 1491 B->bs = bs; 1492 b->bs2 = bs*bs; 1493 b->mbs = mbs; 1494 b->nbs = mbs; 1495 b->Mbs = Mbs; 1496 b->Nbs = Mbs; 1497 1498 ierr = MPI_Allgather(&b->mbs,1,MPIU_INT,b->rowners+1,1,MPIU_INT,B->comm);CHKERRQ(ierr); 1499 b->rowners[0] = 0; 1500 for (i=2; i<=b->size; i++) { 1501 b->rowners[i] += b->rowners[i-1]; 1502 } 1503 b->rstart = b->rowners[b->rank]; 1504 b->rend = b->rowners[b->rank+1]; 1505 b->cstart = b->rstart; 1506 b->cend = b->rend; 1507 for (i=0; i<=b->size; i++) { 1508 b->rowners_bs[i] = b->rowners[i]*bs; 1509 } 1510 b->rstart_bs = b-> rstart*bs; 1511 b->rend_bs = b->rend*bs; 1512 1513 b->cstart_bs = b->cstart*bs; 1514 b->cend_bs = b->cend*bs; 1515 1516 ierr = MatCreate(PETSC_COMM_SELF,B->m,B->m,B->m,B->m,&b->A);CHKERRQ(ierr); 1517 ierr = MatSetType(b->A,MATSEQSBAIJ);CHKERRQ(ierr); 1518 ierr = MatSeqSBAIJSetPreallocation(b->A,bs,d_nz,d_nnz);CHKERRQ(ierr); 1519 ierr = PetscLogObjectParent(B,b->A);CHKERRQ(ierr); 1520 1521 ierr = MatCreate(PETSC_COMM_SELF,B->m,B->M,B->m,B->M,&b->B);CHKERRQ(ierr); 1522 ierr = MatSetType(b->B,MATSEQBAIJ);CHKERRQ(ierr); 1523 ierr = MatSeqBAIJSetPreallocation(b->B,bs,o_nz,o_nnz);CHKERRQ(ierr); 1524 ierr = PetscLogObjectParent(B,b->B);CHKERRQ(ierr); 1525 1526 /* build cache for off array entries formed */ 1527 ierr = MatStashCreate_Private(B->comm,bs,&B->bstash);CHKERRQ(ierr); 1528 1529 PetscFunctionReturn(0); 1530 } 1531 EXTERN_C_END 1532 1533 /*MC 1534 MATMPISBAIJ - MATMPISBAIJ = "mpisbaij" - A matrix type to be used for distributed symmetric sparse block matrices, 1535 based on block compressed sparse row format. Only the upper triangular portion of the matrix is stored. 1536 1537 Options Database Keys: 1538 . -mat_type mpisbaij - sets the matrix type to "mpisbaij" during a call to MatSetFromOptions() 1539 1540 Level: beginner 1541 1542 .seealso: MatCreateMPISBAIJ 1543 M*/ 1544 1545 EXTERN_C_BEGIN 1546 #undef __FUNCT__ 1547 #define __FUNCT__ "MatCreate_MPISBAIJ" 1548 PetscErrorCode MatCreate_MPISBAIJ(Mat B) 1549 { 1550 Mat_MPISBAIJ *b; 1551 PetscErrorCode ierr; 1552 PetscTruth flg; 1553 1554 PetscFunctionBegin; 1555 1556 ierr = PetscNew(Mat_MPISBAIJ,&b);CHKERRQ(ierr); 1557 B->data = (void*)b; 1558 ierr = PetscMemcpy(B->ops,&MatOps_Values,sizeof(struct _MatOps));CHKERRQ(ierr); 1559 1560 B->ops->destroy = MatDestroy_MPISBAIJ; 1561 B->ops->view = MatView_MPISBAIJ; 1562 B->mapping = 0; 1563 B->factor = 0; 1564 B->assembled = PETSC_FALSE; 1565 1566 B->insertmode = NOT_SET_VALUES; 1567 ierr = MPI_Comm_rank(B->comm,&b->rank);CHKERRQ(ierr); 1568 ierr = MPI_Comm_size(B->comm,&b->size);CHKERRQ(ierr); 1569 1570 /* build local table of row and column ownerships */ 1571 ierr = PetscMalloc(3*(b->size+2)*sizeof(PetscInt),&b->rowners);CHKERRQ(ierr); 1572 b->cowners = b->rowners + b->size + 2; 1573 b->rowners_bs = b->cowners + b->size + 2; 1574 ierr = PetscLogObjectMemory(B,3*(b->size+2)*sizeof(PetscInt)+sizeof(struct _p_Mat)+sizeof(Mat_MPISBAIJ));CHKERRQ(ierr); 1575 1576 /* build cache for off array entries formed */ 1577 ierr = MatStashCreate_Private(B->comm,1,&B->stash);CHKERRQ(ierr); 1578 b->donotstash = PETSC_FALSE; 1579 b->colmap = PETSC_NULL; 1580 b->garray = PETSC_NULL; 1581 b->roworiented = PETSC_TRUE; 1582 1583 #if defined(PETSC_USE_MAT_SINGLE) 1584 /* stuff for MatSetValues_XXX in single precision */ 1585 b->setvalueslen = 0; 1586 b->setvaluescopy = PETSC_NULL; 1587 #endif 1588 1589 /* stuff used in block assembly */ 1590 b->barray = 0; 1591 1592 /* stuff used for matrix vector multiply */ 1593 b->lvec = 0; 1594 b->Mvctx = 0; 1595 b->slvec0 = 0; 1596 b->slvec0b = 0; 1597 b->slvec1 = 0; 1598 b->slvec1a = 0; 1599 b->slvec1b = 0; 1600 b->sMvctx = 0; 1601 1602 /* stuff for MatGetRow() */ 1603 b->rowindices = 0; 1604 b->rowvalues = 0; 1605 b->getrowactive = PETSC_FALSE; 1606 1607 /* hash table stuff */ 1608 b->ht = 0; 1609 b->hd = 0; 1610 b->ht_size = 0; 1611 b->ht_flag = PETSC_FALSE; 1612 b->ht_fact = 0; 1613 b->ht_total_ct = 0; 1614 b->ht_insert_ct = 0; 1615 1616 ierr = PetscOptionsHasName(B->prefix,"-mat_use_hash_table",&flg);CHKERRQ(ierr); 1617 if (flg) { 1618 PetscReal fact = 1.39; 1619 ierr = MatSetOption(B,MAT_USE_HASH_TABLE);CHKERRQ(ierr); 1620 ierr = PetscOptionsGetReal(B->prefix,"-mat_use_hash_table",&fact,PETSC_NULL);CHKERRQ(ierr); 1621 if (fact <= 1.0) fact = 1.39; 1622 ierr = MatMPIBAIJSetHashTableFactor(B,fact);CHKERRQ(ierr); 1623 PetscLogInfo(0,"MatCreateMPISBAIJ:Hash table Factor used %5.2f\n",fact); 1624 } 1625 ierr = PetscObjectComposeFunctionDynamic((PetscObject)B,"MatStoreValues_C", 1626 "MatStoreValues_MPISBAIJ", 1627 MatStoreValues_MPISBAIJ);CHKERRQ(ierr); 1628 ierr = PetscObjectComposeFunctionDynamic((PetscObject)B,"MatRetrieveValues_C", 1629 "MatRetrieveValues_MPISBAIJ", 1630 MatRetrieveValues_MPISBAIJ);CHKERRQ(ierr); 1631 ierr = PetscObjectComposeFunctionDynamic((PetscObject)B,"MatGetDiagonalBlock_C", 1632 "MatGetDiagonalBlock_MPISBAIJ", 1633 MatGetDiagonalBlock_MPISBAIJ);CHKERRQ(ierr); 1634 ierr = PetscObjectComposeFunctionDynamic((PetscObject)B,"MatMPISBAIJSetPreallocation_C", 1635 "MatMPISBAIJSetPreallocation_MPISBAIJ", 1636 MatMPISBAIJSetPreallocation_MPISBAIJ);CHKERRQ(ierr); 1637 B->symmetric = PETSC_TRUE; 1638 B->structurally_symmetric = PETSC_TRUE; 1639 B->symmetric_set = PETSC_TRUE; 1640 B->structurally_symmetric_set = PETSC_TRUE; 1641 PetscFunctionReturn(0); 1642 } 1643 EXTERN_C_END 1644 1645 /*MC 1646 MATSBAIJ - MATSBAIJ = "sbaij" - A matrix type to be used for symmetric block sparse matrices. 1647 1648 This matrix type is identical to MATSEQSBAIJ when constructed with a single process communicator, 1649 and MATMPISBAIJ otherwise. 1650 1651 Options Database Keys: 1652 . -mat_type sbaij - sets the matrix type to "sbaij" during a call to MatSetFromOptions() 1653 1654 Level: beginner 1655 1656 .seealso: MatCreateMPISBAIJ,MATSEQSBAIJ,MATMPISBAIJ 1657 M*/ 1658 1659 EXTERN_C_BEGIN 1660 #undef __FUNCT__ 1661 #define __FUNCT__ "MatCreate_SBAIJ" 1662 PetscErrorCode MatCreate_SBAIJ(Mat A) 1663 { 1664 PetscErrorCode ierr; 1665 PetscMPIInt size; 1666 1667 PetscFunctionBegin; 1668 ierr = PetscObjectChangeTypeName((PetscObject)A,MATSBAIJ);CHKERRQ(ierr); 1669 ierr = MPI_Comm_size(A->comm,&size);CHKERRQ(ierr); 1670 if (size == 1) { 1671 ierr = MatSetType(A,MATSEQSBAIJ);CHKERRQ(ierr); 1672 } else { 1673 ierr = MatSetType(A,MATMPISBAIJ);CHKERRQ(ierr); 1674 } 1675 PetscFunctionReturn(0); 1676 } 1677 EXTERN_C_END 1678 1679 #undef __FUNCT__ 1680 #define __FUNCT__ "MatMPISBAIJSetPreallocation" 1681 /*@C 1682 MatMPISBAIJSetPreallocation - For good matrix assembly performance 1683 the user should preallocate the matrix storage by setting the parameters 1684 d_nz (or d_nnz) and o_nz (or o_nnz). By setting these parameters accurately, 1685 performance can be increased by more than a factor of 50. 1686 1687 Collective on Mat 1688 1689 Input Parameters: 1690 + A - the matrix 1691 . bs - size of blockk 1692 . d_nz - number of block nonzeros per block row in diagonal portion of local 1693 submatrix (same for all local rows) 1694 . d_nnz - array containing the number of block nonzeros in the various block rows 1695 in the upper triangular and diagonal part of the in diagonal portion of the local 1696 (possibly different for each block row) or PETSC_NULL. You must leave room 1697 for the diagonal entry even if it is zero. 1698 . o_nz - number of block nonzeros per block row in the off-diagonal portion of local 1699 submatrix (same for all local rows). 1700 - o_nnz - array containing the number of nonzeros in the various block rows of the 1701 off-diagonal portion of the local submatrix (possibly different for 1702 each block row) or PETSC_NULL. 1703 1704 1705 Options Database Keys: 1706 . -mat_no_unroll - uses code that does not unroll the loops in the 1707 block calculations (much slower) 1708 . -mat_block_size - size of the blocks to use 1709 1710 Notes: 1711 1712 If PETSC_DECIDE or PETSC_DETERMINE is used for a particular argument on one processor 1713 than it must be used on all processors that share the object for that argument. 1714 1715 If the *_nnz parameter is given then the *_nz parameter is ignored 1716 1717 Storage Information: 1718 For a square global matrix we define each processor's diagonal portion 1719 to be its local rows and the corresponding columns (a square submatrix); 1720 each processor's off-diagonal portion encompasses the remainder of the 1721 local matrix (a rectangular submatrix). 1722 1723 The user can specify preallocated storage for the diagonal part of 1724 the local submatrix with either d_nz or d_nnz (not both). Set 1725 d_nz=PETSC_DEFAULT and d_nnz=PETSC_NULL for PETSc to control dynamic 1726 memory allocation. Likewise, specify preallocated storage for the 1727 off-diagonal part of the local submatrix with o_nz or o_nnz (not both). 1728 1729 Consider a processor that owns rows 3, 4 and 5 of a parallel matrix. In 1730 the figure below we depict these three local rows and all columns (0-11). 1731 1732 .vb 1733 0 1 2 3 4 5 6 7 8 9 10 11 1734 ------------------- 1735 row 3 | o o o d d d o o o o o o 1736 row 4 | o o o d d d o o o o o o 1737 row 5 | o o o d d d o o o o o o 1738 ------------------- 1739 .ve 1740 1741 Thus, any entries in the d locations are stored in the d (diagonal) 1742 submatrix, and any entries in the o locations are stored in the 1743 o (off-diagonal) submatrix. Note that the d matrix is stored in 1744 MatSeqSBAIJ format and the o submatrix in MATSEQBAIJ format. 1745 1746 Now d_nz should indicate the number of block nonzeros per row in the upper triangular 1747 plus the diagonal part of the d matrix, 1748 and o_nz should indicate the number of block nonzeros per row in the o matrix. 1749 In general, for PDE problems in which most nonzeros are near the diagonal, 1750 one expects d_nz >> o_nz. For large problems you MUST preallocate memory 1751 or you will get TERRIBLE performance; see the users' manual chapter on 1752 matrices. 1753 1754 Level: intermediate 1755 1756 .keywords: matrix, block, aij, compressed row, sparse, parallel 1757 1758 .seealso: MatCreate(), MatCreateSeqSBAIJ(), MatSetValues(), MatCreateMPIBAIJ() 1759 @*/ 1760 PetscErrorCode MatMPISBAIJSetPreallocation(Mat B,PetscInt bs,PetscInt d_nz,const PetscInt d_nnz[],PetscInt o_nz,const PetscInt o_nnz[]) 1761 { 1762 PetscErrorCode ierr,(*f)(Mat,PetscInt,PetscInt,const PetscInt[],PetscInt,const PetscInt[]); 1763 1764 PetscFunctionBegin; 1765 ierr = PetscObjectQueryFunction((PetscObject)B,"MatMPISBAIJSetPreallocation_C",(void (**)(void))&f);CHKERRQ(ierr); 1766 if (f) { 1767 ierr = (*f)(B,bs,d_nz,d_nnz,o_nz,o_nnz);CHKERRQ(ierr); 1768 } 1769 PetscFunctionReturn(0); 1770 } 1771 1772 #undef __FUNCT__ 1773 #define __FUNCT__ "MatCreateMPISBAIJ" 1774 /*@C 1775 MatCreateMPISBAIJ - Creates a sparse parallel matrix in symmetric block AIJ format 1776 (block compressed row). For good matrix assembly performance 1777 the user should preallocate the matrix storage by setting the parameters 1778 d_nz (or d_nnz) and o_nz (or o_nnz). By setting these parameters accurately, 1779 performance can be increased by more than a factor of 50. 1780 1781 Collective on MPI_Comm 1782 1783 Input Parameters: 1784 + comm - MPI communicator 1785 . bs - size of blockk 1786 . m - number of local rows (or PETSC_DECIDE to have calculated if M is given) 1787 This value should be the same as the local size used in creating the 1788 y vector for the matrix-vector product y = Ax. 1789 . n - number of local columns (or PETSC_DECIDE to have calculated if N is given) 1790 This value should be the same as the local size used in creating the 1791 x vector for the matrix-vector product y = Ax. 1792 . M - number of global rows (or PETSC_DETERMINE to have calculated if m is given) 1793 . N - number of global columns (or PETSC_DETERMINE to have calculated if n is given) 1794 . d_nz - number of block nonzeros per block row in diagonal portion of local 1795 submatrix (same for all local rows) 1796 . d_nnz - array containing the number of block nonzeros in the various block rows 1797 in the upper triangular portion of the in diagonal portion of the local 1798 (possibly different for each block block row) or PETSC_NULL. 1799 You must leave room for the diagonal entry even if it is zero. 1800 . o_nz - number of block nonzeros per block row in the off-diagonal portion of local 1801 submatrix (same for all local rows). 1802 - o_nnz - array containing the number of nonzeros in the various block rows of the 1803 off-diagonal portion of the local submatrix (possibly different for 1804 each block row) or PETSC_NULL. 1805 1806 Output Parameter: 1807 . A - the matrix 1808 1809 Options Database Keys: 1810 . -mat_no_unroll - uses code that does not unroll the loops in the 1811 block calculations (much slower) 1812 . -mat_block_size - size of the blocks to use 1813 . -mat_mpi - use the parallel matrix data structures even on one processor 1814 (defaults to using SeqBAIJ format on one processor) 1815 1816 Notes: 1817 The number of rows and columns must be divisible by blocksize. 1818 1819 The user MUST specify either the local or global matrix dimensions 1820 (possibly both). 1821 1822 If PETSC_DECIDE or PETSC_DETERMINE is used for a particular argument on one processor 1823 than it must be used on all processors that share the object for that argument. 1824 1825 If the *_nnz parameter is given then the *_nz parameter is ignored 1826 1827 Storage Information: 1828 For a square global matrix we define each processor's diagonal portion 1829 to be its local rows and the corresponding columns (a square submatrix); 1830 each processor's off-diagonal portion encompasses the remainder of the 1831 local matrix (a rectangular submatrix). 1832 1833 The user can specify preallocated storage for the diagonal part of 1834 the local submatrix with either d_nz or d_nnz (not both). Set 1835 d_nz=PETSC_DEFAULT and d_nnz=PETSC_NULL for PETSc to control dynamic 1836 memory allocation. Likewise, specify preallocated storage for the 1837 off-diagonal part of the local submatrix with o_nz or o_nnz (not both). 1838 1839 Consider a processor that owns rows 3, 4 and 5 of a parallel matrix. In 1840 the figure below we depict these three local rows and all columns (0-11). 1841 1842 .vb 1843 0 1 2 3 4 5 6 7 8 9 10 11 1844 ------------------- 1845 row 3 | o o o d d d o o o o o o 1846 row 4 | o o o d d d o o o o o o 1847 row 5 | o o o d d d o o o o o o 1848 ------------------- 1849 .ve 1850 1851 Thus, any entries in the d locations are stored in the d (diagonal) 1852 submatrix, and any entries in the o locations are stored in the 1853 o (off-diagonal) submatrix. Note that the d matrix is stored in 1854 MatSeqSBAIJ format and the o submatrix in MATSEQBAIJ format. 1855 1856 Now d_nz should indicate the number of block nonzeros per row in the upper triangular 1857 plus the diagonal part of the d matrix, 1858 and o_nz should indicate the number of block nonzeros per row in the o matrix. 1859 In general, for PDE problems in which most nonzeros are near the diagonal, 1860 one expects d_nz >> o_nz. For large problems you MUST preallocate memory 1861 or you will get TERRIBLE performance; see the users' manual chapter on 1862 matrices. 1863 1864 Level: intermediate 1865 1866 .keywords: matrix, block, aij, compressed row, sparse, parallel 1867 1868 .seealso: MatCreate(), MatCreateSeqSBAIJ(), MatSetValues(), MatCreateMPIBAIJ() 1869 @*/ 1870 1871 PetscErrorCode MatCreateMPISBAIJ(MPI_Comm comm,PetscInt bs,PetscInt m,PetscInt n,PetscInt M,PetscInt N,PetscInt d_nz,const PetscInt d_nnz[],PetscInt o_nz,const PetscInt o_nnz[],Mat *A) 1872 { 1873 PetscErrorCode ierr; 1874 PetscMPIInt size; 1875 1876 PetscFunctionBegin; 1877 ierr = MatCreate(comm,m,n,M,N,A);CHKERRQ(ierr); 1878 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr); 1879 if (size > 1) { 1880 ierr = MatSetType(*A,MATMPISBAIJ);CHKERRQ(ierr); 1881 ierr = MatMPISBAIJSetPreallocation(*A,bs,d_nz,d_nnz,o_nz,o_nnz);CHKERRQ(ierr); 1882 } else { 1883 ierr = MatSetType(*A,MATSEQSBAIJ);CHKERRQ(ierr); 1884 ierr = MatSeqSBAIJSetPreallocation(*A,bs,d_nz,d_nnz);CHKERRQ(ierr); 1885 } 1886 PetscFunctionReturn(0); 1887 } 1888 1889 1890 #undef __FUNCT__ 1891 #define __FUNCT__ "MatDuplicate_MPISBAIJ" 1892 static PetscErrorCode MatDuplicate_MPISBAIJ(Mat matin,MatDuplicateOption cpvalues,Mat *newmat) 1893 { 1894 Mat mat; 1895 Mat_MPISBAIJ *a,*oldmat = (Mat_MPISBAIJ*)matin->data; 1896 PetscErrorCode ierr; 1897 PetscInt len=0,nt,bs=matin->bs,mbs=oldmat->mbs; 1898 PetscScalar *array; 1899 1900 PetscFunctionBegin; 1901 *newmat = 0; 1902 ierr = MatCreate(matin->comm,matin->m,matin->n,matin->M,matin->N,&mat);CHKERRQ(ierr); 1903 ierr = MatSetType(mat,matin->type_name);CHKERRQ(ierr); 1904 ierr = PetscMemcpy(mat->ops,matin->ops,sizeof(struct _MatOps));CHKERRQ(ierr); 1905 1906 mat->factor = matin->factor; 1907 mat->preallocated = PETSC_TRUE; 1908 mat->assembled = PETSC_TRUE; 1909 mat->insertmode = NOT_SET_VALUES; 1910 1911 a = (Mat_MPISBAIJ*)mat->data; 1912 mat->bs = matin->bs; 1913 a->bs2 = oldmat->bs2; 1914 a->mbs = oldmat->mbs; 1915 a->nbs = oldmat->nbs; 1916 a->Mbs = oldmat->Mbs; 1917 a->Nbs = oldmat->Nbs; 1918 1919 a->rstart = oldmat->rstart; 1920 a->rend = oldmat->rend; 1921 a->cstart = oldmat->cstart; 1922 a->cend = oldmat->cend; 1923 a->size = oldmat->size; 1924 a->rank = oldmat->rank; 1925 a->donotstash = oldmat->donotstash; 1926 a->roworiented = oldmat->roworiented; 1927 a->rowindices = 0; 1928 a->rowvalues = 0; 1929 a->getrowactive = PETSC_FALSE; 1930 a->barray = 0; 1931 a->rstart_bs = oldmat->rstart_bs; 1932 a->rend_bs = oldmat->rend_bs; 1933 a->cstart_bs = oldmat->cstart_bs; 1934 a->cend_bs = oldmat->cend_bs; 1935 1936 /* hash table stuff */ 1937 a->ht = 0; 1938 a->hd = 0; 1939 a->ht_size = 0; 1940 a->ht_flag = oldmat->ht_flag; 1941 a->ht_fact = oldmat->ht_fact; 1942 a->ht_total_ct = 0; 1943 a->ht_insert_ct = 0; 1944 1945 ierr = PetscMemcpy(a->rowners,oldmat->rowners,3*(a->size+2)*sizeof(PetscInt));CHKERRQ(ierr); 1946 ierr = MatStashCreate_Private(matin->comm,1,&mat->stash);CHKERRQ(ierr); 1947 ierr = MatStashCreate_Private(matin->comm,matin->bs,&mat->bstash);CHKERRQ(ierr); 1948 if (oldmat->colmap) { 1949 #if defined (PETSC_USE_CTABLE) 1950 ierr = PetscTableCreateCopy(oldmat->colmap,&a->colmap);CHKERRQ(ierr); 1951 #else 1952 ierr = PetscMalloc((a->Nbs)*sizeof(PetscInt),&a->colmap);CHKERRQ(ierr); 1953 ierr = PetscLogObjectMemory(mat,(a->Nbs)*sizeof(PetscInt));CHKERRQ(ierr); 1954 ierr = PetscMemcpy(a->colmap,oldmat->colmap,(a->Nbs)*sizeof(PetscInt));CHKERRQ(ierr); 1955 #endif 1956 } else a->colmap = 0; 1957 1958 if (oldmat->garray && (len = ((Mat_SeqBAIJ*)(oldmat->B->data))->nbs)) { 1959 ierr = PetscMalloc(len*sizeof(PetscInt),&a->garray);CHKERRQ(ierr); 1960 ierr = PetscLogObjectMemory(mat,len*sizeof(PetscInt));CHKERRQ(ierr); 1961 ierr = PetscMemcpy(a->garray,oldmat->garray,len*sizeof(PetscInt));CHKERRQ(ierr); 1962 } else a->garray = 0; 1963 1964 ierr = VecDuplicate(oldmat->lvec,&a->lvec);CHKERRQ(ierr); 1965 ierr = PetscLogObjectParent(mat,a->lvec);CHKERRQ(ierr); 1966 ierr = VecScatterCopy(oldmat->Mvctx,&a->Mvctx);CHKERRQ(ierr); 1967 ierr = PetscLogObjectParent(mat,a->Mvctx);CHKERRQ(ierr); 1968 1969 ierr = VecDuplicate(oldmat->slvec0,&a->slvec0);CHKERRQ(ierr); 1970 ierr = PetscLogObjectParent(mat,a->slvec0);CHKERRQ(ierr); 1971 ierr = VecDuplicate(oldmat->slvec1,&a->slvec1);CHKERRQ(ierr); 1972 ierr = PetscLogObjectParent(mat,a->slvec1);CHKERRQ(ierr); 1973 1974 ierr = VecGetLocalSize(a->slvec1,&nt);CHKERRQ(ierr); 1975 ierr = VecGetArray(a->slvec1,&array);CHKERRQ(ierr); 1976 ierr = VecCreateSeqWithArray(PETSC_COMM_SELF,bs*mbs,array,&a->slvec1a);CHKERRQ(ierr); 1977 ierr = VecCreateSeqWithArray(PETSC_COMM_SELF,nt-bs*mbs,array+bs*mbs,&a->slvec1b);CHKERRQ(ierr); 1978 ierr = VecRestoreArray(a->slvec1,&array);CHKERRQ(ierr); 1979 ierr = VecGetArray(a->slvec0,&array);CHKERRQ(ierr); 1980 ierr = VecCreateSeqWithArray(PETSC_COMM_SELF,nt-bs*mbs,array+bs*mbs,&a->slvec0b);CHKERRQ(ierr); 1981 ierr = VecRestoreArray(a->slvec0,&array);CHKERRQ(ierr); 1982 ierr = PetscLogObjectParent(mat,a->slvec0);CHKERRQ(ierr); 1983 ierr = PetscLogObjectParent(mat,a->slvec1);CHKERRQ(ierr); 1984 ierr = PetscLogObjectParent(mat,a->slvec0b);CHKERRQ(ierr); 1985 ierr = PetscLogObjectParent(mat,a->slvec1a);CHKERRQ(ierr); 1986 ierr = PetscLogObjectParent(mat,a->slvec1b);CHKERRQ(ierr); 1987 1988 /* ierr = VecScatterCopy(oldmat->sMvctx,&a->sMvctx); - not written yet, replaced by the lazy trick: */ 1989 ierr = PetscObjectReference((PetscObject)oldmat->sMvctx);CHKERRQ(ierr); 1990 a->sMvctx = oldmat->sMvctx; 1991 ierr = PetscLogObjectParent(mat,a->sMvctx);CHKERRQ(ierr); 1992 1993 ierr = MatDuplicate(oldmat->A,cpvalues,&a->A);CHKERRQ(ierr); 1994 ierr = PetscLogObjectParent(mat,a->A);CHKERRQ(ierr); 1995 ierr = MatDuplicate(oldmat->B,cpvalues,&a->B);CHKERRQ(ierr); 1996 ierr = PetscLogObjectParent(mat,a->B);CHKERRQ(ierr); 1997 ierr = PetscFListDuplicate(mat->qlist,&matin->qlist);CHKERRQ(ierr); 1998 *newmat = mat; 1999 PetscFunctionReturn(0); 2000 } 2001 2002 #include "petscsys.h" 2003 2004 #undef __FUNCT__ 2005 #define __FUNCT__ "MatLoad_MPISBAIJ" 2006 PetscErrorCode MatLoad_MPISBAIJ(PetscViewer viewer,const MatType type,Mat *newmat) 2007 { 2008 Mat A; 2009 PetscErrorCode ierr; 2010 PetscInt i,nz,j,rstart,rend; 2011 PetscScalar *vals,*buf; 2012 MPI_Comm comm = ((PetscObject)viewer)->comm; 2013 MPI_Status status; 2014 PetscMPIInt rank,size,tag = ((PetscObject)viewer)->tag,*sndcounts = 0,*browners,maxnz,*rowners; 2015 PetscInt header[4],*rowlengths = 0,M,N,m,*cols; 2016 PetscInt *locrowlens,*procsnz = 0,jj,*mycols,*ibuf; 2017 PetscInt bs=1,Mbs,mbs,extra_rows; 2018 PetscInt *dlens,*odlens,*mask,*masked1,*masked2,rowcount,odcount; 2019 PetscInt dcount,kmax,k,nzcount,tmp; 2020 int fd; 2021 2022 PetscFunctionBegin; 2023 ierr = PetscOptionsGetInt(PETSC_NULL,"-matload_block_size",&bs,PETSC_NULL);CHKERRQ(ierr); 2024 2025 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr); 2026 ierr = MPI_Comm_rank(comm,&rank);CHKERRQ(ierr); 2027 if (!rank) { 2028 ierr = PetscViewerBinaryGetDescriptor(viewer,&fd);CHKERRQ(ierr); 2029 ierr = PetscBinaryRead(fd,(char *)header,4,PETSC_INT);CHKERRQ(ierr); 2030 if (header[0] != MAT_FILE_COOKIE) SETERRQ(PETSC_ERR_FILE_UNEXPECTED,"not matrix object"); 2031 if (header[3] < 0) { 2032 SETERRQ(PETSC_ERR_FILE_UNEXPECTED,"Matrix stored in special format, cannot load as MPISBAIJ"); 2033 } 2034 } 2035 2036 ierr = MPI_Bcast(header+1,3,MPIU_INT,0,comm);CHKERRQ(ierr); 2037 M = header[1]; N = header[2]; 2038 2039 if (M != N) SETERRQ(PETSC_ERR_SUP,"Can only do square matrices"); 2040 2041 /* 2042 This code adds extra rows to make sure the number of rows is 2043 divisible by the blocksize 2044 */ 2045 Mbs = M/bs; 2046 extra_rows = bs - M + bs*(Mbs); 2047 if (extra_rows == bs) extra_rows = 0; 2048 else Mbs++; 2049 if (extra_rows &&!rank) { 2050 PetscLogInfo(0,"MatLoad_MPISBAIJ:Padding loaded matrix to match blocksize\n"); 2051 } 2052 2053 /* determine ownership of all rows */ 2054 mbs = Mbs/size + ((Mbs % size) > rank); 2055 m = mbs*bs; 2056 ierr = PetscMalloc(2*(size+2)*sizeof(PetscMPIInt),&rowners);CHKERRQ(ierr); 2057 browners = rowners + size + 1; 2058 ierr = MPI_Allgather(&mbs,1,MPI_INT,rowners+1,1,MPI_INT,comm);CHKERRQ(ierr); 2059 rowners[0] = 0; 2060 for (i=2; i<=size; i++) rowners[i] += rowners[i-1]; 2061 for (i=0; i<=size; i++) browners[i] = rowners[i]*bs; 2062 rstart = rowners[rank]; 2063 rend = rowners[rank+1]; 2064 2065 /* distribute row lengths to all processors */ 2066 ierr = PetscMalloc((rend-rstart)*bs*sizeof(PetscInt),&locrowlens);CHKERRQ(ierr); 2067 if (!rank) { 2068 ierr = PetscMalloc((M+extra_rows)*sizeof(PetscInt),&rowlengths);CHKERRQ(ierr); 2069 ierr = PetscBinaryRead(fd,rowlengths,M,PETSC_INT);CHKERRQ(ierr); 2070 for (i=0; i<extra_rows; i++) rowlengths[M+i] = 1; 2071 ierr = PetscMalloc(size*sizeof(PetscMPIInt),&sndcounts);CHKERRQ(ierr); 2072 for (i=0; i<size; i++) sndcounts[i] = browners[i+1] - browners[i]; 2073 ierr = MPI_Scatterv(rowlengths,sndcounts,browners,MPIU_INT,locrowlens,(rend-rstart)*bs,MPIU_INT,0,comm);CHKERRQ(ierr); 2074 ierr = PetscFree(sndcounts);CHKERRQ(ierr); 2075 } else { 2076 ierr = MPI_Scatterv(0,0,0,MPIU_INT,locrowlens,(rend-rstart)*bs,MPIU_INT,0,comm);CHKERRQ(ierr); 2077 } 2078 2079 if (!rank) { /* procs[0] */ 2080 /* calculate the number of nonzeros on each processor */ 2081 ierr = PetscMalloc(size*sizeof(PetscInt),&procsnz);CHKERRQ(ierr); 2082 ierr = PetscMemzero(procsnz,size*sizeof(PetscInt));CHKERRQ(ierr); 2083 for (i=0; i<size; i++) { 2084 for (j=rowners[i]*bs; j< rowners[i+1]*bs; j++) { 2085 procsnz[i] += rowlengths[j]; 2086 } 2087 } 2088 ierr = PetscFree(rowlengths);CHKERRQ(ierr); 2089 2090 /* determine max buffer needed and allocate it */ 2091 maxnz = 0; 2092 for (i=0; i<size; i++) { 2093 maxnz = PetscMax(maxnz,procsnz[i]); 2094 } 2095 ierr = PetscMalloc(maxnz*sizeof(PetscInt),&cols);CHKERRQ(ierr); 2096 2097 /* read in my part of the matrix column indices */ 2098 nz = procsnz[0]; 2099 ierr = PetscMalloc(nz*sizeof(PetscInt),&ibuf);CHKERRQ(ierr); 2100 mycols = ibuf; 2101 if (size == 1) nz -= extra_rows; 2102 ierr = PetscBinaryRead(fd,mycols,nz,PETSC_INT);CHKERRQ(ierr); 2103 if (size == 1) for (i=0; i< extra_rows; i++) { mycols[nz+i] = M+i; } 2104 2105 /* read in every ones (except the last) and ship off */ 2106 for (i=1; i<size-1; i++) { 2107 nz = procsnz[i]; 2108 ierr = PetscBinaryRead(fd,cols,nz,PETSC_INT);CHKERRQ(ierr); 2109 ierr = MPI_Send(cols,nz,MPIU_INT,i,tag,comm);CHKERRQ(ierr); 2110 } 2111 /* read in the stuff for the last proc */ 2112 if (size != 1) { 2113 nz = procsnz[size-1] - extra_rows; /* the extra rows are not on the disk */ 2114 ierr = PetscBinaryRead(fd,cols,nz,PETSC_INT);CHKERRQ(ierr); 2115 for (i=0; i<extra_rows; i++) cols[nz+i] = M+i; 2116 ierr = MPI_Send(cols,nz+extra_rows,MPIU_INT,size-1,tag,comm);CHKERRQ(ierr); 2117 } 2118 ierr = PetscFree(cols);CHKERRQ(ierr); 2119 } else { /* procs[i], i>0 */ 2120 /* determine buffer space needed for message */ 2121 nz = 0; 2122 for (i=0; i<m; i++) { 2123 nz += locrowlens[i]; 2124 } 2125 ierr = PetscMalloc(nz*sizeof(PetscInt),&ibuf);CHKERRQ(ierr); 2126 mycols = ibuf; 2127 /* receive message of column indices*/ 2128 ierr = MPI_Recv(mycols,nz,MPIU_INT,0,tag,comm,&status);CHKERRQ(ierr); 2129 ierr = MPI_Get_count(&status,MPIU_INT,&maxnz);CHKERRQ(ierr); 2130 if (maxnz != nz) SETERRQ(PETSC_ERR_FILE_UNEXPECTED,"something is wrong with file"); 2131 } 2132 2133 /* loop over local rows, determining number of off diagonal entries */ 2134 ierr = PetscMalloc(2*(rend-rstart+1)*sizeof(PetscInt),&dlens);CHKERRQ(ierr); 2135 odlens = dlens + (rend-rstart); 2136 ierr = PetscMalloc(3*Mbs*sizeof(PetscInt),&mask);CHKERRQ(ierr); 2137 ierr = PetscMemzero(mask,3*Mbs*sizeof(PetscInt));CHKERRQ(ierr); 2138 masked1 = mask + Mbs; 2139 masked2 = masked1 + Mbs; 2140 rowcount = 0; nzcount = 0; 2141 for (i=0; i<mbs; i++) { 2142 dcount = 0; 2143 odcount = 0; 2144 for (j=0; j<bs; j++) { 2145 kmax = locrowlens[rowcount]; 2146 for (k=0; k<kmax; k++) { 2147 tmp = mycols[nzcount++]/bs; /* block col. index */ 2148 if (!mask[tmp]) { 2149 mask[tmp] = 1; 2150 if (tmp < rstart || tmp >= rend) masked2[odcount++] = tmp; /* entry in off-diag portion */ 2151 else masked1[dcount++] = tmp; /* entry in diag portion */ 2152 } 2153 } 2154 rowcount++; 2155 } 2156 2157 dlens[i] = dcount; /* d_nzz[i] */ 2158 odlens[i] = odcount; /* o_nzz[i] */ 2159 2160 /* zero out the mask elements we set */ 2161 for (j=0; j<dcount; j++) mask[masked1[j]] = 0; 2162 for (j=0; j<odcount; j++) mask[masked2[j]] = 0; 2163 } 2164 2165 /* create our matrix */ 2166 ierr = MatCreate(comm,m,m,PETSC_DETERMINE,PETSC_DETERMINE,&A);CHKERRQ(ierr); 2167 ierr = MatSetType(A,type);CHKERRQ(ierr); 2168 ierr = MatMPISBAIJSetPreallocation(A,bs,0,dlens,0,odlens);CHKERRQ(ierr); 2169 ierr = MatSetOption(A,MAT_COLUMNS_SORTED);CHKERRQ(ierr); 2170 2171 if (!rank) { 2172 ierr = PetscMalloc(maxnz*sizeof(PetscScalar),&buf);CHKERRQ(ierr); 2173 /* read in my part of the matrix numerical values */ 2174 nz = procsnz[0]; 2175 vals = buf; 2176 mycols = ibuf; 2177 if (size == 1) nz -= extra_rows; 2178 ierr = PetscBinaryRead(fd,vals,nz,PETSC_SCALAR);CHKERRQ(ierr); 2179 if (size == 1) for (i=0; i< extra_rows; i++) { vals[nz+i] = 1.0; } 2180 2181 /* insert into matrix */ 2182 jj = rstart*bs; 2183 for (i=0; i<m; i++) { 2184 ierr = MatSetValues(A,1,&jj,locrowlens[i],mycols,vals,INSERT_VALUES);CHKERRQ(ierr); 2185 mycols += locrowlens[i]; 2186 vals += locrowlens[i]; 2187 jj++; 2188 } 2189 2190 /* read in other processors (except the last one) and ship out */ 2191 for (i=1; i<size-1; i++) { 2192 nz = procsnz[i]; 2193 vals = buf; 2194 ierr = PetscBinaryRead(fd,vals,nz,PETSC_SCALAR);CHKERRQ(ierr); 2195 ierr = MPI_Send(vals,nz,MPIU_SCALAR,i,A->tag,comm);CHKERRQ(ierr); 2196 } 2197 /* the last proc */ 2198 if (size != 1){ 2199 nz = procsnz[i] - extra_rows; 2200 vals = buf; 2201 ierr = PetscBinaryRead(fd,vals,nz,PETSC_SCALAR);CHKERRQ(ierr); 2202 for (i=0; i<extra_rows; i++) vals[nz+i] = 1.0; 2203 ierr = MPI_Send(vals,nz+extra_rows,MPIU_SCALAR,size-1,A->tag,comm);CHKERRQ(ierr); 2204 } 2205 ierr = PetscFree(procsnz);CHKERRQ(ierr); 2206 2207 } else { 2208 /* receive numeric values */ 2209 ierr = PetscMalloc(nz*sizeof(PetscScalar),&buf);CHKERRQ(ierr); 2210 2211 /* receive message of values*/ 2212 vals = buf; 2213 mycols = ibuf; 2214 ierr = MPI_Recv(vals,nz,MPIU_SCALAR,0,A->tag,comm,&status);CHKERRQ(ierr); 2215 ierr = MPI_Get_count(&status,MPIU_SCALAR,&maxnz);CHKERRQ(ierr); 2216 if (maxnz != nz) SETERRQ(PETSC_ERR_FILE_UNEXPECTED,"something is wrong with file"); 2217 2218 /* insert into matrix */ 2219 jj = rstart*bs; 2220 for (i=0; i<m; i++) { 2221 ierr = MatSetValues_MPISBAIJ(A,1,&jj,locrowlens[i],mycols,vals,INSERT_VALUES);CHKERRQ(ierr); 2222 mycols += locrowlens[i]; 2223 vals += locrowlens[i]; 2224 jj++; 2225 } 2226 } 2227 2228 ierr = PetscFree(locrowlens);CHKERRQ(ierr); 2229 ierr = PetscFree(buf);CHKERRQ(ierr); 2230 ierr = PetscFree(ibuf);CHKERRQ(ierr); 2231 ierr = PetscFree(rowners);CHKERRQ(ierr); 2232 ierr = PetscFree(dlens);CHKERRQ(ierr); 2233 ierr = PetscFree(mask);CHKERRQ(ierr); 2234 ierr = MatAssemblyBegin(A,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 2235 ierr = MatAssemblyEnd(A,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 2236 *newmat = A; 2237 PetscFunctionReturn(0); 2238 } 2239 2240 #undef __FUNCT__ 2241 #define __FUNCT__ "MatMPISBAIJSetHashTableFactor" 2242 /*XXXXX@ 2243 MatMPISBAIJSetHashTableFactor - Sets the factor required to compute the size of the HashTable. 2244 2245 Input Parameters: 2246 . mat - the matrix 2247 . fact - factor 2248 2249 Collective on Mat 2250 2251 Level: advanced 2252 2253 Notes: 2254 This can also be set by the command line option: -mat_use_hash_table fact 2255 2256 .keywords: matrix, hashtable, factor, HT 2257 2258 .seealso: MatSetOption() 2259 @XXXXX*/ 2260 2261 2262 #undef __FUNCT__ 2263 #define __FUNCT__ "MatGetRowMax_MPISBAIJ" 2264 PetscErrorCode MatGetRowMax_MPISBAIJ(Mat A,Vec v) 2265 { 2266 Mat_MPISBAIJ *a = (Mat_MPISBAIJ*)A->data; 2267 Mat_SeqBAIJ *b = (Mat_SeqBAIJ*)(a->B)->data; 2268 PetscReal atmp; 2269 PetscReal *work,*svalues,*rvalues; 2270 PetscErrorCode ierr; 2271 PetscInt i,bs,mbs,*bi,*bj,brow,j,ncols,krow,kcol,col,row,Mbs,bcol; 2272 PetscMPIInt rank,size; 2273 PetscInt *rowners_bs,dest,count,source; 2274 PetscScalar *va; 2275 MatScalar *ba; 2276 MPI_Status stat; 2277 2278 PetscFunctionBegin; 2279 ierr = MatGetRowMax(a->A,v);CHKERRQ(ierr); 2280 ierr = VecGetArray(v,&va);CHKERRQ(ierr); 2281 2282 ierr = MPI_Comm_size(A->comm,&size);CHKERRQ(ierr); 2283 ierr = MPI_Comm_rank(A->comm,&rank);CHKERRQ(ierr); 2284 2285 bs = A->bs; 2286 mbs = a->mbs; 2287 Mbs = a->Mbs; 2288 ba = b->a; 2289 bi = b->i; 2290 bj = b->j; 2291 2292 /* find ownerships */ 2293 rowners_bs = a->rowners_bs; 2294 2295 /* each proc creates an array to be distributed */ 2296 ierr = PetscMalloc(bs*Mbs*sizeof(PetscReal),&work);CHKERRQ(ierr); 2297 ierr = PetscMemzero(work,bs*Mbs*sizeof(PetscReal));CHKERRQ(ierr); 2298 2299 /* row_max for B */ 2300 if (rank != size-1){ 2301 for (i=0; i<mbs; i++) { 2302 ncols = bi[1] - bi[0]; bi++; 2303 brow = bs*i; 2304 for (j=0; j<ncols; j++){ 2305 bcol = bs*(*bj); 2306 for (kcol=0; kcol<bs; kcol++){ 2307 col = bcol + kcol; /* local col index */ 2308 col += rowners_bs[rank+1]; /* global col index */ 2309 for (krow=0; krow<bs; krow++){ 2310 atmp = PetscAbsScalar(*ba); ba++; 2311 row = brow + krow; /* local row index */ 2312 if (PetscRealPart(va[row]) < atmp) va[row] = atmp; 2313 if (work[col] < atmp) work[col] = atmp; 2314 } 2315 } 2316 bj++; 2317 } 2318 } 2319 2320 /* send values to its owners */ 2321 for (dest=rank+1; dest<size; dest++){ 2322 svalues = work + rowners_bs[dest]; 2323 count = rowners_bs[dest+1]-rowners_bs[dest]; 2324 ierr = MPI_Send(svalues,count,MPIU_REAL,dest,rank,A->comm);CHKERRQ(ierr); 2325 } 2326 } 2327 2328 /* receive values */ 2329 if (rank){ 2330 rvalues = work; 2331 count = rowners_bs[rank+1]-rowners_bs[rank]; 2332 for (source=0; source<rank; source++){ 2333 ierr = MPI_Recv(rvalues,count,MPIU_REAL,MPI_ANY_SOURCE,MPI_ANY_TAG,A->comm,&stat);CHKERRQ(ierr); 2334 /* process values */ 2335 for (i=0; i<count; i++){ 2336 if (PetscRealPart(va[i]) < rvalues[i]) va[i] = rvalues[i]; 2337 } 2338 } 2339 } 2340 2341 ierr = VecRestoreArray(v,&va);CHKERRQ(ierr); 2342 ierr = PetscFree(work);CHKERRQ(ierr); 2343 PetscFunctionReturn(0); 2344 } 2345 2346 #undef __FUNCT__ 2347 #define __FUNCT__ "MatRelax_MPISBAIJ" 2348 PetscErrorCode MatRelax_MPISBAIJ(Mat matin,Vec bb,PetscReal omega,MatSORType flag,PetscReal fshift,PetscInt its,PetscInt lits,Vec xx) 2349 { 2350 Mat_MPISBAIJ *mat = (Mat_MPISBAIJ*)matin->data; 2351 PetscErrorCode ierr; 2352 PetscInt mbs=mat->mbs,bs=matin->bs; 2353 PetscScalar mone=-1.0,*x,*b,*ptr,zero=0.0; 2354 Vec bb1; 2355 2356 PetscFunctionBegin; 2357 if (its <= 0 || lits <= 0) SETERRQ2(PETSC_ERR_ARG_WRONG,"Relaxation requires global its %D and local its %D both positive",its,lits); 2358 if (bs > 1) 2359 SETERRQ(PETSC_ERR_SUP,"SSOR for block size > 1 is not yet implemented"); 2360 2361 if ((flag & SOR_LOCAL_SYMMETRIC_SWEEP) == SOR_LOCAL_SYMMETRIC_SWEEP){ 2362 if ( flag & SOR_ZERO_INITIAL_GUESS ) { 2363 ierr = (*mat->A->ops->relax)(mat->A,bb,omega,flag,fshift,lits,lits,xx);CHKERRQ(ierr); 2364 its--; 2365 } 2366 2367 ierr = VecDuplicate(bb,&bb1);CHKERRQ(ierr); 2368 while (its--){ 2369 2370 /* lower triangular part: slvec0b = - B^T*xx */ 2371 ierr = (*mat->B->ops->multtranspose)(mat->B,xx,mat->slvec0b);CHKERRQ(ierr); 2372 2373 /* copy xx into slvec0a */ 2374 ierr = VecGetArray(mat->slvec0,&ptr);CHKERRQ(ierr); 2375 ierr = VecGetArray(xx,&x);CHKERRQ(ierr); 2376 ierr = PetscMemcpy(ptr,x,bs*mbs*sizeof(MatScalar));CHKERRQ(ierr); 2377 ierr = VecRestoreArray(mat->slvec0,&ptr);CHKERRQ(ierr); 2378 2379 ierr = VecScale(&mone,mat->slvec0);CHKERRQ(ierr); 2380 2381 /* copy bb into slvec1a */ 2382 ierr = VecGetArray(mat->slvec1,&ptr);CHKERRQ(ierr); 2383 ierr = VecGetArray(bb,&b);CHKERRQ(ierr); 2384 ierr = PetscMemcpy(ptr,b,bs*mbs*sizeof(MatScalar));CHKERRQ(ierr); 2385 ierr = VecRestoreArray(mat->slvec1,&ptr);CHKERRQ(ierr); 2386 2387 /* set slvec1b = 0 */ 2388 ierr = VecSet(&zero,mat->slvec1b);CHKERRQ(ierr); 2389 2390 ierr = VecScatterBegin(mat->slvec0,mat->slvec1,ADD_VALUES,SCATTER_FORWARD,mat->sMvctx);CHKERRQ(ierr); 2391 ierr = VecRestoreArray(xx,&x);CHKERRQ(ierr); 2392 ierr = VecRestoreArray(bb,&b);CHKERRQ(ierr); 2393 ierr = VecScatterEnd(mat->slvec0,mat->slvec1,ADD_VALUES,SCATTER_FORWARD,mat->sMvctx);CHKERRQ(ierr); 2394 2395 /* upper triangular part: bb1 = bb1 - B*x */ 2396 ierr = (*mat->B->ops->multadd)(mat->B,mat->slvec1b,mat->slvec1a,bb1);CHKERRQ(ierr); 2397 2398 /* local diagonal sweep */ 2399 ierr = (*mat->A->ops->relax)(mat->A,bb1,omega,SOR_SYMMETRIC_SWEEP,fshift,lits,lits,xx);CHKERRQ(ierr); 2400 } 2401 ierr = VecDestroy(bb1);CHKERRQ(ierr); 2402 } else { 2403 SETERRQ(PETSC_ERR_SUP,"MatSORType is not supported for SBAIJ matrix format"); 2404 } 2405 PetscFunctionReturn(0); 2406 } 2407 2408 #undef __FUNCT__ 2409 #define __FUNCT__ "MatRelax_MPISBAIJ_2comm" 2410 PetscErrorCode MatRelax_MPISBAIJ_2comm(Mat matin,Vec bb,PetscReal omega,MatSORType flag,PetscReal fshift,PetscInt its,PetscInt lits,Vec xx) 2411 { 2412 Mat_MPISBAIJ *mat = (Mat_MPISBAIJ*)matin->data; 2413 PetscErrorCode ierr; 2414 PetscScalar mone=-1.0; 2415 Vec lvec1,bb1; 2416 2417 PetscFunctionBegin; 2418 if (its <= 0 || lits <= 0) SETERRQ2(PETSC_ERR_ARG_WRONG,"Relaxation requires global its %D and local its %D both positive",its,lits); 2419 if (matin->bs > 1) 2420 SETERRQ(PETSC_ERR_SUP,"SSOR for block size > 1 is not yet implemented"); 2421 2422 if ((flag & SOR_LOCAL_SYMMETRIC_SWEEP) == SOR_LOCAL_SYMMETRIC_SWEEP){ 2423 if ( flag & SOR_ZERO_INITIAL_GUESS ) { 2424 ierr = (*mat->A->ops->relax)(mat->A,bb,omega,flag,fshift,lits,lits,xx);CHKERRQ(ierr); 2425 its--; 2426 } 2427 2428 ierr = VecDuplicate(mat->lvec,&lvec1);CHKERRQ(ierr); 2429 ierr = VecDuplicate(bb,&bb1);CHKERRQ(ierr); 2430 while (its--){ 2431 ierr = VecScatterBegin(xx,mat->lvec,INSERT_VALUES,SCATTER_FORWARD,mat->Mvctx);CHKERRQ(ierr); 2432 2433 /* lower diagonal part: bb1 = bb - B^T*xx */ 2434 ierr = (*mat->B->ops->multtranspose)(mat->B,xx,lvec1);CHKERRQ(ierr); 2435 ierr = VecScale(&mone,lvec1);CHKERRQ(ierr); 2436 2437 ierr = VecScatterEnd(xx,mat->lvec,INSERT_VALUES,SCATTER_FORWARD,mat->Mvctx);CHKERRQ(ierr); 2438 ierr = VecCopy(bb,bb1);CHKERRQ(ierr); 2439 ierr = VecScatterBegin(lvec1,bb1,ADD_VALUES,SCATTER_REVERSE,mat->Mvctx);CHKERRQ(ierr); 2440 2441 /* upper diagonal part: bb1 = bb1 - B*x */ 2442 ierr = VecScale(&mone,mat->lvec);CHKERRQ(ierr); 2443 ierr = (*mat->B->ops->multadd)(mat->B,mat->lvec,bb1,bb1);CHKERRQ(ierr); 2444 2445 ierr = VecScatterEnd(lvec1,bb1,ADD_VALUES,SCATTER_REVERSE,mat->Mvctx);CHKERRQ(ierr); 2446 2447 /* diagonal sweep */ 2448 ierr = (*mat->A->ops->relax)(mat->A,bb1,omega,SOR_SYMMETRIC_SWEEP,fshift,lits,lits,xx);CHKERRQ(ierr); 2449 } 2450 ierr = VecDestroy(lvec1);CHKERRQ(ierr); 2451 ierr = VecDestroy(bb1);CHKERRQ(ierr); 2452 } else { 2453 SETERRQ(PETSC_ERR_SUP,"MatSORType is not supported for SBAIJ matrix format"); 2454 } 2455 PetscFunctionReturn(0); 2456 } 2457 2458