1 2 #include <../src/mat/impls/baij/mpi/mpibaij.h> /*I "petscmat.h" I*/ 3 #include <../src/mat/impls/sbaij/mpi/mpisbaij.h> 4 #include <../src/mat/impls/sbaij/seq/sbaij.h> 5 #include <petscblaslapack.h> 6 7 extern PetscErrorCode MatSetUpMultiply_MPISBAIJ(Mat); 8 extern PetscErrorCode MatSetUpMultiply_MPISBAIJ_2comm(Mat); 9 extern PetscErrorCode MatDisAssemble_MPISBAIJ(Mat); 10 extern PetscErrorCode MatIncreaseOverlap_MPISBAIJ(Mat,PetscInt,IS[],PetscInt); 11 extern PetscErrorCode MatGetValues_SeqSBAIJ(Mat,PetscInt,const PetscInt[],PetscInt,const PetscInt[],PetscScalar []); 12 extern PetscErrorCode MatGetValues_SeqBAIJ(Mat,PetscInt,const PetscInt[],PetscInt,const PetscInt[],PetscScalar []); 13 extern PetscErrorCode MatSetValues_SeqSBAIJ(Mat,PetscInt,const PetscInt [],PetscInt,const PetscInt [],const PetscScalar [],InsertMode); 14 extern PetscErrorCode MatSetValuesBlocked_SeqSBAIJ(Mat,PetscInt,const PetscInt[],PetscInt,const PetscInt[],const PetscScalar[],InsertMode); 15 extern PetscErrorCode MatSetValuesBlocked_SeqBAIJ(Mat,PetscInt,const PetscInt[],PetscInt,const PetscInt[],const PetscScalar[],InsertMode); 16 extern PetscErrorCode MatGetRow_SeqSBAIJ(Mat,PetscInt,PetscInt*,PetscInt**,PetscScalar**); 17 extern PetscErrorCode MatRestoreRow_SeqSBAIJ(Mat,PetscInt,PetscInt*,PetscInt**,PetscScalar**); 18 extern PetscErrorCode MatZeroRows_SeqSBAIJ(Mat,IS,PetscScalar*,Vec,Vec); 19 extern PetscErrorCode MatZeroRows_SeqBAIJ(Mat,IS,PetscScalar*,Vec,Vec); 20 extern PetscErrorCode MatGetRowMaxAbs_MPISBAIJ(Mat,Vec,PetscInt[]); 21 extern PetscErrorCode MatSOR_MPISBAIJ(Mat,Vec,PetscReal,MatSORType,PetscReal,PetscInt,PetscInt,Vec); 22 23 #undef __FUNCT__ 24 #define __FUNCT__ "MatStoreValues_MPISBAIJ" 25 PetscErrorCode MatStoreValues_MPISBAIJ(Mat mat) 26 { 27 Mat_MPISBAIJ *aij = (Mat_MPISBAIJ*)mat->data; 28 PetscErrorCode ierr; 29 30 PetscFunctionBegin; 31 ierr = MatStoreValues(aij->A);CHKERRQ(ierr); 32 ierr = MatStoreValues(aij->B);CHKERRQ(ierr); 33 PetscFunctionReturn(0); 34 } 35 36 #undef __FUNCT__ 37 #define __FUNCT__ "MatRetrieveValues_MPISBAIJ" 38 PetscErrorCode MatRetrieveValues_MPISBAIJ(Mat mat) 39 { 40 Mat_MPISBAIJ *aij = (Mat_MPISBAIJ*)mat->data; 41 PetscErrorCode ierr; 42 43 PetscFunctionBegin; 44 ierr = MatRetrieveValues(aij->A);CHKERRQ(ierr); 45 ierr = MatRetrieveValues(aij->B);CHKERRQ(ierr); 46 PetscFunctionReturn(0); 47 } 48 49 #define MatSetValues_SeqSBAIJ_A_Private(row,col,value,addv) \ 50 { \ 51 \ 52 brow = row/bs; \ 53 rp = aj + ai[brow]; ap = aa + bs2*ai[brow]; \ 54 rmax = aimax[brow]; nrow = ailen[brow]; \ 55 bcol = col/bs; \ 56 ridx = row % bs; cidx = col % bs; \ 57 low = 0; high = nrow; \ 58 while (high-low > 3) { \ 59 t = (low+high)/2; \ 60 if (rp[t] > bcol) high = t; \ 61 else low = t; \ 62 } \ 63 for (_i=low; _i<high; _i++) { \ 64 if (rp[_i] > bcol) break; \ 65 if (rp[_i] == bcol) { \ 66 bap = ap + bs2*_i + bs*cidx + ridx; \ 67 if (addv == ADD_VALUES) *bap += value; \ 68 else *bap = value; \ 69 goto a_noinsert; \ 70 } \ 71 } \ 72 if (a->nonew == 1) goto a_noinsert; \ 73 if (a->nonew == -1) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Inserting a new nonzero (%D, %D) into matrix", row, col); \ 74 MatSeqXAIJReallocateAIJ(A,a->mbs,bs2,nrow,brow,bcol,rmax,aa,ai,aj,rp,ap,aimax,a->nonew,MatScalar); \ 75 N = nrow++ - 1; \ 76 /* shift up all the later entries in this row */ \ 77 for (ii=N; ii>=_i; ii--) { \ 78 rp[ii+1] = rp[ii]; \ 79 ierr = PetscMemcpy(ap+bs2*(ii+1),ap+bs2*(ii),bs2*sizeof(MatScalar));CHKERRQ(ierr); \ 80 } \ 81 if (N>=_i) { ierr = PetscMemzero(ap+bs2*_i,bs2*sizeof(MatScalar));CHKERRQ(ierr); } \ 82 rp[_i] = bcol; \ 83 ap[bs2*_i + bs*cidx + ridx] = value; \ 84 a_noinsert:; \ 85 ailen[brow] = nrow; \ 86 } 87 88 #define MatSetValues_SeqSBAIJ_B_Private(row,col,value,addv) \ 89 { \ 90 brow = row/bs; \ 91 rp = bj + bi[brow]; ap = ba + bs2*bi[brow]; \ 92 rmax = bimax[brow]; nrow = bilen[brow]; \ 93 bcol = col/bs; \ 94 ridx = row % bs; cidx = col % bs; \ 95 low = 0; high = nrow; \ 96 while (high-low > 3) { \ 97 t = (low+high)/2; \ 98 if (rp[t] > bcol) high = t; \ 99 else low = t; \ 100 } \ 101 for (_i=low; _i<high; _i++) { \ 102 if (rp[_i] > bcol) break; \ 103 if (rp[_i] == bcol) { \ 104 bap = ap + bs2*_i + bs*cidx + ridx; \ 105 if (addv == ADD_VALUES) *bap += value; \ 106 else *bap = value; \ 107 goto b_noinsert; \ 108 } \ 109 } \ 110 if (b->nonew == 1) goto b_noinsert; \ 111 if (b->nonew == -1) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Inserting a new nonzero (%D, %D) into matrix", row, col); \ 112 MatSeqXAIJReallocateAIJ(B,b->mbs,bs2,nrow,brow,bcol,rmax,ba,bi,bj,rp,ap,bimax,b->nonew,MatScalar); \ 113 N = nrow++ - 1; \ 114 /* shift up all the later entries in this row */ \ 115 for (ii=N; ii>=_i; ii--) { \ 116 rp[ii+1] = rp[ii]; \ 117 ierr = PetscMemcpy(ap+bs2*(ii+1),ap+bs2*(ii),bs2*sizeof(MatScalar));CHKERRQ(ierr); \ 118 } \ 119 if (N>=_i) { ierr = PetscMemzero(ap+bs2*_i,bs2*sizeof(MatScalar));CHKERRQ(ierr);} \ 120 rp[_i] = bcol; \ 121 ap[bs2*_i + bs*cidx + ridx] = value; \ 122 b_noinsert:; \ 123 bilen[brow] = nrow; \ 124 } 125 126 /* Only add/insert a(i,j) with i<=j (blocks). 127 Any a(i,j) with i>j input by user is ingored. 128 */ 129 #undef __FUNCT__ 130 #define __FUNCT__ "MatSetValues_MPISBAIJ" 131 PetscErrorCode MatSetValues_MPISBAIJ(Mat mat,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],const PetscScalar v[],InsertMode addv) 132 { 133 Mat_MPISBAIJ *baij = (Mat_MPISBAIJ*)mat->data; 134 MatScalar value; 135 PetscBool roworiented = baij->roworiented; 136 PetscErrorCode ierr; 137 PetscInt i,j,row,col; 138 PetscInt rstart_orig=mat->rmap->rstart; 139 PetscInt rend_orig =mat->rmap->rend,cstart_orig=mat->cmap->rstart; 140 PetscInt cend_orig =mat->cmap->rend,bs=mat->rmap->bs; 141 142 /* Some Variables required in the macro */ 143 Mat A = baij->A; 144 Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)(A)->data; 145 PetscInt *aimax=a->imax,*ai=a->i,*ailen=a->ilen,*aj=a->j; 146 MatScalar *aa =a->a; 147 148 Mat B = baij->B; 149 Mat_SeqBAIJ *b = (Mat_SeqBAIJ*)(B)->data; 150 PetscInt *bimax=b->imax,*bi=b->i,*bilen=b->ilen,*bj=b->j; 151 MatScalar *ba =b->a; 152 153 PetscInt *rp,ii,nrow,_i,rmax,N,brow,bcol; 154 PetscInt low,high,t,ridx,cidx,bs2=a->bs2; 155 MatScalar *ap,*bap; 156 157 /* for stash */ 158 PetscInt n_loc, *in_loc = NULL; 159 MatScalar *v_loc = NULL; 160 161 PetscFunctionBegin; 162 if (v) PetscValidScalarPointer(v,6); 163 if (!baij->donotstash) { 164 if (n > baij->n_loc) { 165 ierr = PetscFree(baij->in_loc);CHKERRQ(ierr); 166 ierr = PetscFree(baij->v_loc);CHKERRQ(ierr); 167 ierr = PetscMalloc1(n,&baij->in_loc);CHKERRQ(ierr); 168 ierr = PetscMalloc1(n,&baij->v_loc);CHKERRQ(ierr); 169 170 baij->n_loc = n; 171 } 172 in_loc = baij->in_loc; 173 v_loc = baij->v_loc; 174 } 175 176 for (i=0; i<m; i++) { 177 if (im[i] < 0) continue; 178 #if defined(PETSC_USE_DEBUG) 179 if (im[i] >= mat->rmap->N) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row too large: row %D max %D",im[i],mat->rmap->N-1); 180 #endif 181 if (im[i] >= rstart_orig && im[i] < rend_orig) { /* this processor entry */ 182 row = im[i] - rstart_orig; /* local row index */ 183 for (j=0; j<n; j++) { 184 if (im[i]/bs > in[j]/bs) { 185 if (a->ignore_ltriangular) { 186 continue; /* ignore lower triangular blocks */ 187 } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_USER,"Lower triangular value cannot be set for sbaij format. Ignoring these values, run with -mat_ignore_lower_triangular or call MatSetOption(mat,MAT_IGNORE_LOWER_TRIANGULAR,PETSC_TRUE)"); 188 } 189 if (in[j] >= cstart_orig && in[j] < cend_orig) { /* diag entry (A) */ 190 col = in[j] - cstart_orig; /* local col index */ 191 brow = row/bs; bcol = col/bs; 192 if (brow > bcol) continue; /* ignore lower triangular blocks of A */ 193 if (roworiented) value = v[i*n+j]; 194 else value = v[i+j*m]; 195 MatSetValues_SeqSBAIJ_A_Private(row,col,value,addv); 196 /* ierr = MatSetValues_SeqBAIJ(baij->A,1,&row,1,&col,&value,addv);CHKERRQ(ierr); */ 197 } else if (in[j] < 0) continue; 198 #if defined(PETSC_USE_DEBUG) 199 else if (in[j] >= mat->cmap->N) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Column too large: col %D max %D",in[j],mat->cmap->N-1); 200 #endif 201 else { /* off-diag entry (B) */ 202 if (mat->was_assembled) { 203 if (!baij->colmap) { 204 ierr = MatCreateColmap_MPIBAIJ_Private(mat);CHKERRQ(ierr); 205 } 206 #if defined(PETSC_USE_CTABLE) 207 ierr = PetscTableFind(baij->colmap,in[j]/bs + 1,&col);CHKERRQ(ierr); 208 col = col - 1; 209 #else 210 col = baij->colmap[in[j]/bs] - 1; 211 #endif 212 if (col < 0 && !((Mat_SeqSBAIJ*)(baij->A->data))->nonew) { 213 ierr = MatDisAssemble_MPISBAIJ(mat);CHKERRQ(ierr); 214 col = in[j]; 215 /* Reinitialize the variables required by MatSetValues_SeqBAIJ_B_Private() */ 216 B = baij->B; 217 b = (Mat_SeqBAIJ*)(B)->data; 218 bimax= b->imax;bi=b->i;bilen=b->ilen;bj=b->j; 219 ba = b->a; 220 } else col += in[j]%bs; 221 } else col = in[j]; 222 if (roworiented) value = v[i*n+j]; 223 else value = v[i+j*m]; 224 MatSetValues_SeqSBAIJ_B_Private(row,col,value,addv); 225 /* ierr = MatSetValues_SeqBAIJ(baij->B,1,&row,1,&col,&value,addv);CHKERRQ(ierr); */ 226 } 227 } 228 } else { /* off processor entry */ 229 if (mat->nooffprocentries) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Setting off process row %D even though MatSetOption(,MAT_NO_OFF_PROC_ENTRIES,PETSC_TRUE) was set",im[i]); 230 if (!baij->donotstash) { 231 mat->assembled = PETSC_FALSE; 232 n_loc = 0; 233 for (j=0; j<n; j++) { 234 if (im[i]/bs > in[j]/bs) continue; /* ignore lower triangular blocks */ 235 in_loc[n_loc] = in[j]; 236 if (roworiented) { 237 v_loc[n_loc] = v[i*n+j]; 238 } else { 239 v_loc[n_loc] = v[j*m+i]; 240 } 241 n_loc++; 242 } 243 ierr = MatStashValuesRow_Private(&mat->stash,im[i],n_loc,in_loc,v_loc,PETSC_FALSE);CHKERRQ(ierr); 244 } 245 } 246 } 247 PetscFunctionReturn(0); 248 } 249 250 #undef __FUNCT__ 251 #define __FUNCT__ "MatSetValuesBlocked_MPISBAIJ" 252 PetscErrorCode MatSetValuesBlocked_MPISBAIJ(Mat mat,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],const MatScalar v[],InsertMode addv) 253 { 254 Mat_MPISBAIJ *baij = (Mat_MPISBAIJ*)mat->data; 255 const MatScalar *value; 256 MatScalar *barray =baij->barray; 257 PetscBool roworiented = baij->roworiented,ignore_ltriangular = ((Mat_SeqSBAIJ*)baij->A->data)->ignore_ltriangular; 258 PetscErrorCode ierr; 259 PetscInt i,j,ii,jj,row,col,rstart=baij->rstartbs; 260 PetscInt rend=baij->rendbs,cstart=baij->rstartbs,stepval; 261 PetscInt cend=baij->rendbs,bs=mat->rmap->bs,bs2=baij->bs2; 262 263 PetscFunctionBegin; 264 if (!barray) { 265 ierr = PetscMalloc1(bs2,&barray);CHKERRQ(ierr); 266 baij->barray = barray; 267 } 268 269 if (roworiented) { 270 stepval = (n-1)*bs; 271 } else { 272 stepval = (m-1)*bs; 273 } 274 for (i=0; i<m; i++) { 275 if (im[i] < 0) continue; 276 #if defined(PETSC_USE_DEBUG) 277 if (im[i] >= baij->Mbs) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row too large, row %D max %D",im[i],baij->Mbs-1); 278 #endif 279 if (im[i] >= rstart && im[i] < rend) { 280 row = im[i] - rstart; 281 for (j=0; j<n; j++) { 282 if (im[i] > in[j]) { 283 if (ignore_ltriangular) continue; /* ignore lower triangular blocks */ 284 else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_USER,"Lower triangular value cannot be set for sbaij format. Ignoring these values, run with -mat_ignore_lower_triangular or call MatSetOption(mat,MAT_IGNORE_LOWER_TRIANGULAR,PETSC_TRUE)"); 285 } 286 /* If NumCol = 1 then a copy is not required */ 287 if ((roworiented) && (n == 1)) { 288 barray = (MatScalar*) v + i*bs2; 289 } else if ((!roworiented) && (m == 1)) { 290 barray = (MatScalar*) v + j*bs2; 291 } else { /* Here a copy is required */ 292 if (roworiented) { 293 value = v + i*(stepval+bs)*bs + j*bs; 294 } else { 295 value = v + j*(stepval+bs)*bs + i*bs; 296 } 297 for (ii=0; ii<bs; ii++,value+=stepval) { 298 for (jj=0; jj<bs; jj++) { 299 *barray++ = *value++; 300 } 301 } 302 barray -=bs2; 303 } 304 305 if (in[j] >= cstart && in[j] < cend) { 306 col = in[j] - cstart; 307 ierr = MatSetValuesBlocked_SeqSBAIJ(baij->A,1,&row,1,&col,barray,addv);CHKERRQ(ierr); 308 } else if (in[j] < 0) continue; 309 #if defined(PETSC_USE_DEBUG) 310 else if (in[j] >= baij->Nbs) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Column too large, col %D max %D",in[j],baij->Nbs-1); 311 #endif 312 else { 313 if (mat->was_assembled) { 314 if (!baij->colmap) { 315 ierr = MatCreateColmap_MPIBAIJ_Private(mat);CHKERRQ(ierr); 316 } 317 318 #if defined(PETSC_USE_DEBUG) 319 #if defined(PETSC_USE_CTABLE) 320 { PetscInt data; 321 ierr = PetscTableFind(baij->colmap,in[j]+1,&data);CHKERRQ(ierr); 322 if ((data - 1) % bs) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Incorrect colmap"); 323 } 324 #else 325 if ((baij->colmap[in[j]] - 1) % bs) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Incorrect colmap"); 326 #endif 327 #endif 328 #if defined(PETSC_USE_CTABLE) 329 ierr = PetscTableFind(baij->colmap,in[j]+1,&col);CHKERRQ(ierr); 330 col = (col - 1)/bs; 331 #else 332 col = (baij->colmap[in[j]] - 1)/bs; 333 #endif 334 if (col < 0 && !((Mat_SeqBAIJ*)(baij->A->data))->nonew) { 335 ierr = MatDisAssemble_MPISBAIJ(mat);CHKERRQ(ierr); 336 col = in[j]; 337 } 338 } else col = in[j]; 339 ierr = MatSetValuesBlocked_SeqBAIJ(baij->B,1,&row,1,&col,barray,addv);CHKERRQ(ierr); 340 } 341 } 342 } else { 343 if (mat->nooffprocentries) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Setting off process row %D even though MatSetOption(,MAT_NO_OFF_PROC_ENTRIES,PETSC_TRUE) was set",im[i]); 344 if (!baij->donotstash) { 345 if (roworiented) { 346 ierr = MatStashValuesRowBlocked_Private(&mat->bstash,im[i],n,in,v,m,n,i);CHKERRQ(ierr); 347 } else { 348 ierr = MatStashValuesColBlocked_Private(&mat->bstash,im[i],n,in,v,m,n,i);CHKERRQ(ierr); 349 } 350 } 351 } 352 } 353 PetscFunctionReturn(0); 354 } 355 356 #undef __FUNCT__ 357 #define __FUNCT__ "MatGetValues_MPISBAIJ" 358 PetscErrorCode MatGetValues_MPISBAIJ(Mat mat,PetscInt m,const PetscInt idxm[],PetscInt n,const PetscInt idxn[],PetscScalar v[]) 359 { 360 Mat_MPISBAIJ *baij = (Mat_MPISBAIJ*)mat->data; 361 PetscErrorCode ierr; 362 PetscInt bs = mat->rmap->bs,i,j,bsrstart = mat->rmap->rstart,bsrend = mat->rmap->rend; 363 PetscInt bscstart = mat->cmap->rstart,bscend = mat->cmap->rend,row,col,data; 364 365 PetscFunctionBegin; 366 for (i=0; i<m; i++) { 367 if (idxm[i] < 0) continue; /* SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative row: %D",idxm[i]); */ 368 if (idxm[i] >= mat->rmap->N) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row too large: row %D max %D",idxm[i],mat->rmap->N-1); 369 if (idxm[i] >= bsrstart && idxm[i] < bsrend) { 370 row = idxm[i] - bsrstart; 371 for (j=0; j<n; j++) { 372 if (idxn[j] < 0) continue; /* SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative column %D",idxn[j]); */ 373 if (idxn[j] >= mat->cmap->N) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Column too large: col %D max %D",idxn[j],mat->cmap->N-1); 374 if (idxn[j] >= bscstart && idxn[j] < bscend) { 375 col = idxn[j] - bscstart; 376 ierr = MatGetValues_SeqSBAIJ(baij->A,1,&row,1,&col,v+i*n+j);CHKERRQ(ierr); 377 } else { 378 if (!baij->colmap) { 379 ierr = MatCreateColmap_MPIBAIJ_Private(mat);CHKERRQ(ierr); 380 } 381 #if defined(PETSC_USE_CTABLE) 382 ierr = PetscTableFind(baij->colmap,idxn[j]/bs+1,&data);CHKERRQ(ierr); 383 data--; 384 #else 385 data = baij->colmap[idxn[j]/bs]-1; 386 #endif 387 if ((data < 0) || (baij->garray[data/bs] != idxn[j]/bs)) *(v+i*n+j) = 0.0; 388 else { 389 col = data + idxn[j]%bs; 390 ierr = MatGetValues_SeqBAIJ(baij->B,1,&row,1,&col,v+i*n+j);CHKERRQ(ierr); 391 } 392 } 393 } 394 } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Only local values currently supported"); 395 } 396 PetscFunctionReturn(0); 397 } 398 399 #undef __FUNCT__ 400 #define __FUNCT__ "MatNorm_MPISBAIJ" 401 PetscErrorCode MatNorm_MPISBAIJ(Mat mat,NormType type,PetscReal *norm) 402 { 403 Mat_MPISBAIJ *baij = (Mat_MPISBAIJ*)mat->data; 404 PetscErrorCode ierr; 405 PetscReal sum[2],*lnorm2; 406 407 PetscFunctionBegin; 408 if (baij->size == 1) { 409 ierr = MatNorm(baij->A,type,norm);CHKERRQ(ierr); 410 } else { 411 if (type == NORM_FROBENIUS) { 412 ierr = PetscMalloc1(2,&lnorm2);CHKERRQ(ierr); 413 ierr = MatNorm(baij->A,type,lnorm2);CHKERRQ(ierr); 414 *lnorm2 = (*lnorm2)*(*lnorm2); lnorm2++; /* squar power of norm(A) */ 415 ierr = MatNorm(baij->B,type,lnorm2);CHKERRQ(ierr); 416 *lnorm2 = (*lnorm2)*(*lnorm2); lnorm2--; /* squar power of norm(B) */ 417 ierr = MPI_Allreduce(lnorm2,sum,2,MPIU_REAL,MPIU_SUM,PetscObjectComm((PetscObject)mat));CHKERRQ(ierr); 418 *norm = PetscSqrtReal(sum[0] + 2*sum[1]); 419 ierr = PetscFree(lnorm2);CHKERRQ(ierr); 420 } else if (type == NORM_INFINITY || type == NORM_1) { /* max row/column sum */ 421 Mat_SeqSBAIJ *amat=(Mat_SeqSBAIJ*)baij->A->data; 422 Mat_SeqBAIJ *bmat=(Mat_SeqBAIJ*)baij->B->data; 423 PetscReal *rsum,*rsum2,vabs; 424 PetscInt *jj,*garray=baij->garray,rstart=baij->rstartbs,nz; 425 PetscInt brow,bcol,col,bs=baij->A->rmap->bs,row,grow,gcol,mbs=amat->mbs; 426 MatScalar *v; 427 428 ierr = PetscMalloc2(mat->cmap->N,&rsum,mat->cmap->N,&rsum2);CHKERRQ(ierr); 429 ierr = PetscMemzero(rsum,mat->cmap->N*sizeof(PetscReal));CHKERRQ(ierr); 430 /* Amat */ 431 v = amat->a; jj = amat->j; 432 for (brow=0; brow<mbs; brow++) { 433 grow = bs*(rstart + brow); 434 nz = amat->i[brow+1] - amat->i[brow]; 435 for (bcol=0; bcol<nz; bcol++) { 436 gcol = bs*(rstart + *jj); jj++; 437 for (col=0; col<bs; col++) { 438 for (row=0; row<bs; row++) { 439 vabs = PetscAbsScalar(*v); v++; 440 rsum[gcol+col] += vabs; 441 /* non-diagonal block */ 442 if (bcol > 0 && vabs > 0.0) rsum[grow+row] += vabs; 443 } 444 } 445 } 446 } 447 /* Bmat */ 448 v = bmat->a; jj = bmat->j; 449 for (brow=0; brow<mbs; brow++) { 450 grow = bs*(rstart + brow); 451 nz = bmat->i[brow+1] - bmat->i[brow]; 452 for (bcol=0; bcol<nz; bcol++) { 453 gcol = bs*garray[*jj]; jj++; 454 for (col=0; col<bs; col++) { 455 for (row=0; row<bs; row++) { 456 vabs = PetscAbsScalar(*v); v++; 457 rsum[gcol+col] += vabs; 458 rsum[grow+row] += vabs; 459 } 460 } 461 } 462 } 463 ierr = MPI_Allreduce(rsum,rsum2,mat->cmap->N,MPIU_REAL,MPIU_SUM,PetscObjectComm((PetscObject)mat));CHKERRQ(ierr); 464 *norm = 0.0; 465 for (col=0; col<mat->cmap->N; col++) { 466 if (rsum2[col] > *norm) *norm = rsum2[col]; 467 } 468 ierr = PetscFree2(rsum,rsum2);CHKERRQ(ierr); 469 } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"No support for this norm yet"); 470 } 471 PetscFunctionReturn(0); 472 } 473 474 #undef __FUNCT__ 475 #define __FUNCT__ "MatAssemblyBegin_MPISBAIJ" 476 PetscErrorCode MatAssemblyBegin_MPISBAIJ(Mat mat,MatAssemblyType mode) 477 { 478 Mat_MPISBAIJ *baij = (Mat_MPISBAIJ*)mat->data; 479 PetscErrorCode ierr; 480 PetscInt nstash,reallocs; 481 InsertMode addv; 482 483 PetscFunctionBegin; 484 if (baij->donotstash || mat->nooffprocentries) PetscFunctionReturn(0); 485 486 /* make sure all processors are either in INSERTMODE or ADDMODE */ 487 ierr = MPI_Allreduce((PetscEnum*)&mat->insertmode,(PetscEnum*)&addv,1,MPIU_ENUM,MPI_BOR,PetscObjectComm((PetscObject)mat));CHKERRQ(ierr); 488 if (addv == (ADD_VALUES|INSERT_VALUES)) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Some processors inserted others added"); 489 mat->insertmode = addv; /* in case this processor had no cache */ 490 491 ierr = MatStashScatterBegin_Private(mat,&mat->stash,mat->rmap->range);CHKERRQ(ierr); 492 ierr = MatStashScatterBegin_Private(mat,&mat->bstash,baij->rangebs);CHKERRQ(ierr); 493 ierr = MatStashGetInfo_Private(&mat->stash,&nstash,&reallocs);CHKERRQ(ierr); 494 ierr = PetscInfo2(mat,"Stash has %D entries,uses %D mallocs.\n",nstash,reallocs);CHKERRQ(ierr); 495 ierr = MatStashGetInfo_Private(&mat->stash,&nstash,&reallocs);CHKERRQ(ierr); 496 ierr = PetscInfo2(mat,"Block-Stash has %D entries, uses %D mallocs.\n",nstash,reallocs);CHKERRQ(ierr); 497 PetscFunctionReturn(0); 498 } 499 500 #undef __FUNCT__ 501 #define __FUNCT__ "MatAssemblyEnd_MPISBAIJ" 502 PetscErrorCode MatAssemblyEnd_MPISBAIJ(Mat mat,MatAssemblyType mode) 503 { 504 Mat_MPISBAIJ *baij=(Mat_MPISBAIJ*)mat->data; 505 Mat_SeqSBAIJ *a =(Mat_SeqSBAIJ*)baij->A->data; 506 PetscErrorCode ierr; 507 PetscInt i,j,rstart,ncols,flg,bs2=baij->bs2; 508 PetscInt *row,*col; 509 PetscBool other_disassembled; 510 PetscMPIInt n; 511 PetscBool r1,r2,r3; 512 MatScalar *val; 513 InsertMode addv = mat->insertmode; 514 515 /* do not use 'b=(Mat_SeqBAIJ*)baij->B->data' as B can be reset in disassembly */ 516 PetscFunctionBegin; 517 if (!baij->donotstash && !mat->nooffprocentries) { 518 while (1) { 519 ierr = MatStashScatterGetMesg_Private(&mat->stash,&n,&row,&col,&val,&flg);CHKERRQ(ierr); 520 if (!flg) break; 521 522 for (i=0; i<n;) { 523 /* Now identify the consecutive vals belonging to the same row */ 524 for (j=i,rstart=row[j]; j<n; j++) { 525 if (row[j] != rstart) break; 526 } 527 if (j < n) ncols = j-i; 528 else ncols = n-i; 529 /* Now assemble all these values with a single function call */ 530 ierr = MatSetValues_MPISBAIJ(mat,1,row+i,ncols,col+i,val+i,addv);CHKERRQ(ierr); 531 i = j; 532 } 533 } 534 ierr = MatStashScatterEnd_Private(&mat->stash);CHKERRQ(ierr); 535 /* Now process the block-stash. Since the values are stashed column-oriented, 536 set the roworiented flag to column oriented, and after MatSetValues() 537 restore the original flags */ 538 r1 = baij->roworiented; 539 r2 = a->roworiented; 540 r3 = ((Mat_SeqBAIJ*)baij->B->data)->roworiented; 541 542 baij->roworiented = PETSC_FALSE; 543 a->roworiented = PETSC_FALSE; 544 545 ((Mat_SeqBAIJ*)baij->B->data)->roworiented = PETSC_FALSE; /* b->roworinted */ 546 while (1) { 547 ierr = MatStashScatterGetMesg_Private(&mat->bstash,&n,&row,&col,&val,&flg);CHKERRQ(ierr); 548 if (!flg) break; 549 550 for (i=0; i<n;) { 551 /* Now identify the consecutive vals belonging to the same row */ 552 for (j=i,rstart=row[j]; j<n; j++) { 553 if (row[j] != rstart) break; 554 } 555 if (j < n) ncols = j-i; 556 else ncols = n-i; 557 ierr = MatSetValuesBlocked_MPISBAIJ(mat,1,row+i,ncols,col+i,val+i*bs2,addv);CHKERRQ(ierr); 558 i = j; 559 } 560 } 561 ierr = MatStashScatterEnd_Private(&mat->bstash);CHKERRQ(ierr); 562 563 baij->roworiented = r1; 564 a->roworiented = r2; 565 566 ((Mat_SeqBAIJ*)baij->B->data)->roworiented = r3; /* b->roworinted */ 567 } 568 569 ierr = MatAssemblyBegin(baij->A,mode);CHKERRQ(ierr); 570 ierr = MatAssemblyEnd(baij->A,mode);CHKERRQ(ierr); 571 572 /* determine if any processor has disassembled, if so we must 573 also disassemble ourselfs, in order that we may reassemble. */ 574 /* 575 if nonzero structure of submatrix B cannot change then we know that 576 no processor disassembled thus we can skip this stuff 577 */ 578 if (!((Mat_SeqBAIJ*)baij->B->data)->nonew) { 579 ierr = MPI_Allreduce(&mat->was_assembled,&other_disassembled,1,MPIU_BOOL,MPI_PROD,PetscObjectComm((PetscObject)mat));CHKERRQ(ierr); 580 if (mat->was_assembled && !other_disassembled) { 581 ierr = MatDisAssemble_MPISBAIJ(mat);CHKERRQ(ierr); 582 } 583 } 584 585 if (!mat->was_assembled && mode == MAT_FINAL_ASSEMBLY) { 586 ierr = MatSetUpMultiply_MPISBAIJ(mat);CHKERRQ(ierr); /* setup Mvctx and sMvctx */ 587 } 588 ierr = MatAssemblyBegin(baij->B,mode);CHKERRQ(ierr); 589 ierr = MatAssemblyEnd(baij->B,mode);CHKERRQ(ierr); 590 591 ierr = PetscFree2(baij->rowvalues,baij->rowindices);CHKERRQ(ierr); 592 593 baij->rowvalues = 0; 594 PetscFunctionReturn(0); 595 } 596 597 extern PetscErrorCode MatSetValues_MPIBAIJ(Mat,PetscInt,const PetscInt[],PetscInt,const PetscInt[],const PetscScalar[],InsertMode); 598 #include <petscdraw.h> 599 #undef __FUNCT__ 600 #define __FUNCT__ "MatView_MPISBAIJ_ASCIIorDraworSocket" 601 static PetscErrorCode MatView_MPISBAIJ_ASCIIorDraworSocket(Mat mat,PetscViewer viewer) 602 { 603 Mat_MPISBAIJ *baij = (Mat_MPISBAIJ*)mat->data; 604 PetscErrorCode ierr; 605 PetscInt bs = mat->rmap->bs; 606 PetscMPIInt size = baij->size,rank = baij->rank; 607 PetscBool iascii,isdraw; 608 PetscViewer sviewer; 609 PetscViewerFormat format; 610 611 PetscFunctionBegin; 612 ierr = PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&iascii);CHKERRQ(ierr); 613 ierr = PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERDRAW,&isdraw);CHKERRQ(ierr); 614 if (iascii) { 615 ierr = PetscViewerGetFormat(viewer,&format);CHKERRQ(ierr); 616 if (format == PETSC_VIEWER_ASCII_INFO_DETAIL) { 617 MatInfo info; 618 ierr = MPI_Comm_rank(PetscObjectComm((PetscObject)mat),&rank);CHKERRQ(ierr); 619 ierr = MatGetInfo(mat,MAT_LOCAL,&info);CHKERRQ(ierr); 620 ierr = PetscViewerASCIISynchronizedAllow(viewer,PETSC_TRUE);CHKERRQ(ierr); 621 ierr = PetscViewerASCIISynchronizedPrintf(viewer,"[%d] Local rows %D nz %D nz alloced %D bs %D mem %D\n",rank,mat->rmap->n,(PetscInt)info.nz_used,(PetscInt)info.nz_allocated,mat->rmap->bs,(PetscInt)info.memory);CHKERRQ(ierr); 622 ierr = MatGetInfo(baij->A,MAT_LOCAL,&info);CHKERRQ(ierr); 623 ierr = PetscViewerASCIISynchronizedPrintf(viewer,"[%d] on-diagonal part: nz %D \n",rank,(PetscInt)info.nz_used);CHKERRQ(ierr); 624 ierr = MatGetInfo(baij->B,MAT_LOCAL,&info);CHKERRQ(ierr); 625 ierr = PetscViewerASCIISynchronizedPrintf(viewer,"[%d] off-diagonal part: nz %D \n",rank,(PetscInt)info.nz_used);CHKERRQ(ierr); 626 ierr = PetscViewerFlush(viewer);CHKERRQ(ierr); 627 ierr = PetscViewerASCIISynchronizedAllow(viewer,PETSC_FALSE);CHKERRQ(ierr); 628 ierr = PetscViewerASCIIPrintf(viewer,"Information on VecScatter used in matrix-vector product: \n");CHKERRQ(ierr); 629 ierr = VecScatterView(baij->Mvctx,viewer);CHKERRQ(ierr); 630 PetscFunctionReturn(0); 631 } else if (format == PETSC_VIEWER_ASCII_INFO) { 632 ierr = PetscViewerASCIIPrintf(viewer," block size is %D\n",bs);CHKERRQ(ierr); 633 PetscFunctionReturn(0); 634 } else if (format == PETSC_VIEWER_ASCII_FACTOR_INFO) { 635 PetscFunctionReturn(0); 636 } 637 } 638 639 if (isdraw) { 640 PetscDraw draw; 641 PetscBool isnull; 642 ierr = PetscViewerDrawGetDraw(viewer,0,&draw);CHKERRQ(ierr); 643 ierr = PetscDrawIsNull(draw,&isnull);CHKERRQ(ierr); if (isnull) PetscFunctionReturn(0); 644 } 645 646 if (size == 1) { 647 ierr = PetscObjectSetName((PetscObject)baij->A,((PetscObject)mat)->name);CHKERRQ(ierr); 648 ierr = MatView(baij->A,viewer);CHKERRQ(ierr); 649 } else { 650 /* assemble the entire matrix onto first processor. */ 651 Mat A; 652 Mat_SeqSBAIJ *Aloc; 653 Mat_SeqBAIJ *Bloc; 654 PetscInt M = mat->rmap->N,N = mat->cmap->N,*ai,*aj,col,i,j,k,*rvals,mbs = baij->mbs; 655 MatScalar *a; 656 657 /* Should this be the same type as mat? */ 658 ierr = MatCreate(PetscObjectComm((PetscObject)mat),&A);CHKERRQ(ierr); 659 if (!rank) { 660 ierr = MatSetSizes(A,M,N,M,N);CHKERRQ(ierr); 661 } else { 662 ierr = MatSetSizes(A,0,0,M,N);CHKERRQ(ierr); 663 } 664 ierr = MatSetType(A,MATMPISBAIJ);CHKERRQ(ierr); 665 ierr = MatMPISBAIJSetPreallocation(A,mat->rmap->bs,0,NULL,0,NULL);CHKERRQ(ierr); 666 ierr = MatSetOption(A,MAT_NEW_NONZERO_LOCATION_ERR,PETSC_FALSE);CHKERRQ(ierr); 667 ierr = PetscLogObjectParent((PetscObject)mat,(PetscObject)A);CHKERRQ(ierr); 668 669 /* copy over the A part */ 670 Aloc = (Mat_SeqSBAIJ*)baij->A->data; 671 ai = Aloc->i; aj = Aloc->j; a = Aloc->a; 672 ierr = PetscMalloc1(bs,&rvals);CHKERRQ(ierr); 673 674 for (i=0; i<mbs; i++) { 675 rvals[0] = bs*(baij->rstartbs + i); 676 for (j=1; j<bs; j++) rvals[j] = rvals[j-1] + 1; 677 for (j=ai[i]; j<ai[i+1]; j++) { 678 col = (baij->cstartbs+aj[j])*bs; 679 for (k=0; k<bs; k++) { 680 ierr = MatSetValues_MPISBAIJ(A,bs,rvals,1,&col,a,INSERT_VALUES);CHKERRQ(ierr); 681 col++; 682 a += bs; 683 } 684 } 685 } 686 /* copy over the B part */ 687 Bloc = (Mat_SeqBAIJ*)baij->B->data; 688 ai = Bloc->i; aj = Bloc->j; a = Bloc->a; 689 for (i=0; i<mbs; i++) { 690 691 rvals[0] = bs*(baij->rstartbs + i); 692 for (j=1; j<bs; j++) rvals[j] = rvals[j-1] + 1; 693 for (j=ai[i]; j<ai[i+1]; j++) { 694 col = baij->garray[aj[j]]*bs; 695 for (k=0; k<bs; k++) { 696 ierr = MatSetValues_MPIBAIJ(A,bs,rvals,1,&col,a,INSERT_VALUES);CHKERRQ(ierr); 697 col++; 698 a += bs; 699 } 700 } 701 } 702 ierr = PetscFree(rvals);CHKERRQ(ierr); 703 ierr = MatAssemblyBegin(A,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 704 ierr = MatAssemblyEnd(A,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 705 /* 706 Everyone has to call to draw the matrix since the graphics waits are 707 synchronized across all processors that share the PetscDraw object 708 */ 709 ierr = PetscViewerGetSingleton(viewer,&sviewer);CHKERRQ(ierr); 710 if (!rank) { 711 ierr = PetscObjectSetName((PetscObject)((Mat_MPISBAIJ*)(A->data))->A,((PetscObject)mat)->name);CHKERRQ(ierr); 712 /* Set the type name to MATMPISBAIJ so that the correct type can be printed out by PetscObjectPrintClassNamePrefixType() in MatView_SeqSBAIJ_ASCII()*/ 713 ierr = PetscStrcpy(((PetscObject)((Mat_MPISBAIJ*)(A->data))->A)->type_name,MATMPISBAIJ);CHKERRQ(ierr); 714 ierr = MatView(((Mat_MPISBAIJ*)(A->data))->A,sviewer);CHKERRQ(ierr); 715 } 716 ierr = PetscViewerRestoreSingleton(viewer,&sviewer);CHKERRQ(ierr); 717 ierr = MatDestroy(&A);CHKERRQ(ierr); 718 } 719 PetscFunctionReturn(0); 720 } 721 722 #undef __FUNCT__ 723 #define __FUNCT__ "MatView_MPISBAIJ" 724 PetscErrorCode MatView_MPISBAIJ(Mat mat,PetscViewer viewer) 725 { 726 PetscErrorCode ierr; 727 PetscBool iascii,isdraw,issocket,isbinary; 728 729 PetscFunctionBegin; 730 ierr = PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&iascii);CHKERRQ(ierr); 731 ierr = PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERDRAW,&isdraw);CHKERRQ(ierr); 732 ierr = PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERSOCKET,&issocket);CHKERRQ(ierr); 733 ierr = PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERBINARY,&isbinary);CHKERRQ(ierr); 734 if (iascii || isdraw || issocket || isbinary) { 735 ierr = MatView_MPISBAIJ_ASCIIorDraworSocket(mat,viewer);CHKERRQ(ierr); 736 } 737 PetscFunctionReturn(0); 738 } 739 740 #undef __FUNCT__ 741 #define __FUNCT__ "MatDestroy_MPISBAIJ" 742 PetscErrorCode MatDestroy_MPISBAIJ(Mat mat) 743 { 744 Mat_MPISBAIJ *baij = (Mat_MPISBAIJ*)mat->data; 745 PetscErrorCode ierr; 746 747 PetscFunctionBegin; 748 #if defined(PETSC_USE_LOG) 749 PetscLogObjectState((PetscObject)mat,"Rows=%D,Cols=%D",mat->rmap->N,mat->cmap->N); 750 #endif 751 ierr = MatStashDestroy_Private(&mat->stash);CHKERRQ(ierr); 752 ierr = MatStashDestroy_Private(&mat->bstash);CHKERRQ(ierr); 753 ierr = MatDestroy(&baij->A);CHKERRQ(ierr); 754 ierr = MatDestroy(&baij->B);CHKERRQ(ierr); 755 #if defined(PETSC_USE_CTABLE) 756 ierr = PetscTableDestroy(&baij->colmap);CHKERRQ(ierr); 757 #else 758 ierr = PetscFree(baij->colmap);CHKERRQ(ierr); 759 #endif 760 ierr = PetscFree(baij->garray);CHKERRQ(ierr); 761 ierr = VecDestroy(&baij->lvec);CHKERRQ(ierr); 762 ierr = VecScatterDestroy(&baij->Mvctx);CHKERRQ(ierr); 763 ierr = VecDestroy(&baij->slvec0);CHKERRQ(ierr); 764 ierr = VecDestroy(&baij->slvec0b);CHKERRQ(ierr); 765 ierr = VecDestroy(&baij->slvec1);CHKERRQ(ierr); 766 ierr = VecDestroy(&baij->slvec1a);CHKERRQ(ierr); 767 ierr = VecDestroy(&baij->slvec1b);CHKERRQ(ierr); 768 ierr = VecScatterDestroy(&baij->sMvctx);CHKERRQ(ierr); 769 ierr = PetscFree2(baij->rowvalues,baij->rowindices);CHKERRQ(ierr); 770 ierr = PetscFree(baij->barray);CHKERRQ(ierr); 771 ierr = PetscFree(baij->hd);CHKERRQ(ierr); 772 ierr = VecDestroy(&baij->diag);CHKERRQ(ierr); 773 ierr = VecDestroy(&baij->bb1);CHKERRQ(ierr); 774 ierr = VecDestroy(&baij->xx1);CHKERRQ(ierr); 775 #if defined(PETSC_USE_REAL_MAT_SINGLE) 776 ierr = PetscFree(baij->setvaluescopy);CHKERRQ(ierr); 777 #endif 778 ierr = PetscFree(baij->in_loc);CHKERRQ(ierr); 779 ierr = PetscFree(baij->v_loc);CHKERRQ(ierr); 780 ierr = PetscFree(baij->rangebs);CHKERRQ(ierr); 781 ierr = PetscFree(mat->data);CHKERRQ(ierr); 782 783 ierr = PetscObjectChangeTypeName((PetscObject)mat,0);CHKERRQ(ierr); 784 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatStoreValues_C",NULL);CHKERRQ(ierr); 785 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatRetrieveValues_C",NULL);CHKERRQ(ierr); 786 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatGetDiagonalBlock_C",NULL);CHKERRQ(ierr); 787 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatMPISBAIJSetPreallocation_C",NULL);CHKERRQ(ierr); 788 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatConvert_mpisbaij_mpisbstrm_C",NULL);CHKERRQ(ierr); 789 PetscFunctionReturn(0); 790 } 791 792 #undef __FUNCT__ 793 #define __FUNCT__ "MatMult_MPISBAIJ_Hermitian" 794 PetscErrorCode MatMult_MPISBAIJ_Hermitian(Mat A,Vec xx,Vec yy) 795 { 796 Mat_MPISBAIJ *a = (Mat_MPISBAIJ*)A->data; 797 PetscErrorCode ierr; 798 PetscInt nt,mbs=a->mbs,bs=A->rmap->bs; 799 PetscScalar *x,*from; 800 801 PetscFunctionBegin; 802 ierr = VecGetLocalSize(xx,&nt);CHKERRQ(ierr); 803 if (nt != A->cmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Incompatible partition of A and xx"); 804 805 /* diagonal part */ 806 ierr = (*a->A->ops->mult)(a->A,xx,a->slvec1a);CHKERRQ(ierr); 807 ierr = VecSet(a->slvec1b,0.0);CHKERRQ(ierr); 808 809 /* subdiagonal part */ 810 ierr = (*a->B->ops->multhermitiantranspose)(a->B,xx,a->slvec0b);CHKERRQ(ierr); 811 812 /* copy x into the vec slvec0 */ 813 ierr = VecGetArray(a->slvec0,&from);CHKERRQ(ierr); 814 ierr = VecGetArray(xx,&x);CHKERRQ(ierr); 815 816 ierr = PetscMemcpy(from,x,bs*mbs*sizeof(MatScalar));CHKERRQ(ierr); 817 ierr = VecRestoreArray(a->slvec0,&from);CHKERRQ(ierr); 818 ierr = VecRestoreArray(xx,&x);CHKERRQ(ierr); 819 820 ierr = VecScatterBegin(a->sMvctx,a->slvec0,a->slvec1,ADD_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 821 ierr = VecScatterEnd(a->sMvctx,a->slvec0,a->slvec1,ADD_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 822 /* supperdiagonal part */ 823 ierr = (*a->B->ops->multadd)(a->B,a->slvec1b,a->slvec1a,yy);CHKERRQ(ierr); 824 PetscFunctionReturn(0); 825 } 826 827 #undef __FUNCT__ 828 #define __FUNCT__ "MatMult_MPISBAIJ" 829 PetscErrorCode MatMult_MPISBAIJ(Mat A,Vec xx,Vec yy) 830 { 831 Mat_MPISBAIJ *a = (Mat_MPISBAIJ*)A->data; 832 PetscErrorCode ierr; 833 PetscInt nt,mbs=a->mbs,bs=A->rmap->bs; 834 PetscScalar *x,*from; 835 836 PetscFunctionBegin; 837 ierr = VecGetLocalSize(xx,&nt);CHKERRQ(ierr); 838 if (nt != A->cmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Incompatible partition of A and xx"); 839 840 /* diagonal part */ 841 ierr = (*a->A->ops->mult)(a->A,xx,a->slvec1a);CHKERRQ(ierr); 842 ierr = VecSet(a->slvec1b,0.0);CHKERRQ(ierr); 843 844 /* subdiagonal part */ 845 ierr = (*a->B->ops->multtranspose)(a->B,xx,a->slvec0b);CHKERRQ(ierr); 846 847 /* copy x into the vec slvec0 */ 848 ierr = VecGetArray(a->slvec0,&from);CHKERRQ(ierr); 849 ierr = VecGetArray(xx,&x);CHKERRQ(ierr); 850 851 ierr = PetscMemcpy(from,x,bs*mbs*sizeof(MatScalar));CHKERRQ(ierr); 852 ierr = VecRestoreArray(a->slvec0,&from);CHKERRQ(ierr); 853 ierr = VecRestoreArray(xx,&x);CHKERRQ(ierr); 854 855 ierr = VecScatterBegin(a->sMvctx,a->slvec0,a->slvec1,ADD_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 856 ierr = VecScatterEnd(a->sMvctx,a->slvec0,a->slvec1,ADD_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 857 /* supperdiagonal part */ 858 ierr = (*a->B->ops->multadd)(a->B,a->slvec1b,a->slvec1a,yy);CHKERRQ(ierr); 859 PetscFunctionReturn(0); 860 } 861 862 #undef __FUNCT__ 863 #define __FUNCT__ "MatMult_MPISBAIJ_2comm" 864 PetscErrorCode MatMult_MPISBAIJ_2comm(Mat A,Vec xx,Vec yy) 865 { 866 Mat_MPISBAIJ *a = (Mat_MPISBAIJ*)A->data; 867 PetscErrorCode ierr; 868 PetscInt nt; 869 870 PetscFunctionBegin; 871 ierr = VecGetLocalSize(xx,&nt);CHKERRQ(ierr); 872 if (nt != A->cmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Incompatible partition of A and xx"); 873 874 ierr = VecGetLocalSize(yy,&nt);CHKERRQ(ierr); 875 if (nt != A->rmap->N) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Incompatible parition of A and yy"); 876 877 ierr = VecScatterBegin(a->Mvctx,xx,a->lvec,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 878 /* do diagonal part */ 879 ierr = (*a->A->ops->mult)(a->A,xx,yy);CHKERRQ(ierr); 880 /* do supperdiagonal part */ 881 ierr = VecScatterEnd(a->Mvctx,xx,a->lvec,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 882 ierr = (*a->B->ops->multadd)(a->B,a->lvec,yy,yy);CHKERRQ(ierr); 883 /* do subdiagonal part */ 884 ierr = (*a->B->ops->multtranspose)(a->B,xx,a->lvec);CHKERRQ(ierr); 885 ierr = VecScatterBegin(a->Mvctx,a->lvec,yy,ADD_VALUES,SCATTER_REVERSE);CHKERRQ(ierr); 886 ierr = VecScatterEnd(a->Mvctx,a->lvec,yy,ADD_VALUES,SCATTER_REVERSE);CHKERRQ(ierr); 887 PetscFunctionReturn(0); 888 } 889 890 #undef __FUNCT__ 891 #define __FUNCT__ "MatMultAdd_MPISBAIJ" 892 PetscErrorCode MatMultAdd_MPISBAIJ(Mat A,Vec xx,Vec yy,Vec zz) 893 { 894 Mat_MPISBAIJ *a = (Mat_MPISBAIJ*)A->data; 895 PetscErrorCode ierr; 896 PetscInt mbs=a->mbs,bs=A->rmap->bs; 897 PetscScalar *x,*from,zero=0.0; 898 899 PetscFunctionBegin; 900 /* 901 PetscSynchronizedPrintf(PetscObjectComm((PetscObject)A)," MatMultAdd is called ...\n"); 902 PetscSynchronizedFlush(PetscObjectComm((PetscObject)A),PETSC_STDOUT); 903 */ 904 /* diagonal part */ 905 ierr = (*a->A->ops->multadd)(a->A,xx,yy,a->slvec1a);CHKERRQ(ierr); 906 ierr = VecSet(a->slvec1b,zero);CHKERRQ(ierr); 907 908 /* subdiagonal part */ 909 ierr = (*a->B->ops->multtranspose)(a->B,xx,a->slvec0b);CHKERRQ(ierr); 910 911 /* copy x into the vec slvec0 */ 912 ierr = VecGetArray(a->slvec0,&from);CHKERRQ(ierr); 913 ierr = VecGetArray(xx,&x);CHKERRQ(ierr); 914 ierr = PetscMemcpy(from,x,bs*mbs*sizeof(MatScalar));CHKERRQ(ierr); 915 ierr = VecRestoreArray(a->slvec0,&from);CHKERRQ(ierr); 916 917 ierr = VecScatterBegin(a->sMvctx,a->slvec0,a->slvec1,ADD_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 918 ierr = VecRestoreArray(xx,&x);CHKERRQ(ierr); 919 ierr = VecScatterEnd(a->sMvctx,a->slvec0,a->slvec1,ADD_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 920 921 /* supperdiagonal part */ 922 ierr = (*a->B->ops->multadd)(a->B,a->slvec1b,a->slvec1a,zz);CHKERRQ(ierr); 923 PetscFunctionReturn(0); 924 } 925 926 #undef __FUNCT__ 927 #define __FUNCT__ "MatMultAdd_MPISBAIJ_2comm" 928 PetscErrorCode MatMultAdd_MPISBAIJ_2comm(Mat A,Vec xx,Vec yy,Vec zz) 929 { 930 Mat_MPISBAIJ *a = (Mat_MPISBAIJ*)A->data; 931 PetscErrorCode ierr; 932 933 PetscFunctionBegin; 934 ierr = VecScatterBegin(a->Mvctx,xx,a->lvec,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 935 /* do diagonal part */ 936 ierr = (*a->A->ops->multadd)(a->A,xx,yy,zz);CHKERRQ(ierr); 937 /* do supperdiagonal part */ 938 ierr = VecScatterEnd(a->Mvctx,xx,a->lvec,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 939 ierr = (*a->B->ops->multadd)(a->B,a->lvec,zz,zz);CHKERRQ(ierr); 940 941 /* do subdiagonal part */ 942 ierr = (*a->B->ops->multtranspose)(a->B,xx,a->lvec);CHKERRQ(ierr); 943 ierr = VecScatterBegin(a->Mvctx,a->lvec,zz,ADD_VALUES,SCATTER_REVERSE);CHKERRQ(ierr); 944 ierr = VecScatterEnd(a->Mvctx,a->lvec,zz,ADD_VALUES,SCATTER_REVERSE);CHKERRQ(ierr); 945 PetscFunctionReturn(0); 946 } 947 948 /* 949 This only works correctly for square matrices where the subblock A->A is the 950 diagonal block 951 */ 952 #undef __FUNCT__ 953 #define __FUNCT__ "MatGetDiagonal_MPISBAIJ" 954 PetscErrorCode MatGetDiagonal_MPISBAIJ(Mat A,Vec v) 955 { 956 Mat_MPISBAIJ *a = (Mat_MPISBAIJ*)A->data; 957 PetscErrorCode ierr; 958 959 PetscFunctionBegin; 960 /* if (a->rmap->N != a->cmap->N) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Supports only square matrix where A->A is diag block"); */ 961 ierr = MatGetDiagonal(a->A,v);CHKERRQ(ierr); 962 PetscFunctionReturn(0); 963 } 964 965 #undef __FUNCT__ 966 #define __FUNCT__ "MatScale_MPISBAIJ" 967 PetscErrorCode MatScale_MPISBAIJ(Mat A,PetscScalar aa) 968 { 969 Mat_MPISBAIJ *a = (Mat_MPISBAIJ*)A->data; 970 PetscErrorCode ierr; 971 972 PetscFunctionBegin; 973 ierr = MatScale(a->A,aa);CHKERRQ(ierr); 974 ierr = MatScale(a->B,aa);CHKERRQ(ierr); 975 PetscFunctionReturn(0); 976 } 977 978 #undef __FUNCT__ 979 #define __FUNCT__ "MatGetRow_MPISBAIJ" 980 PetscErrorCode MatGetRow_MPISBAIJ(Mat matin,PetscInt row,PetscInt *nz,PetscInt **idx,PetscScalar **v) 981 { 982 Mat_MPISBAIJ *mat = (Mat_MPISBAIJ*)matin->data; 983 PetscScalar *vworkA,*vworkB,**pvA,**pvB,*v_p; 984 PetscErrorCode ierr; 985 PetscInt bs = matin->rmap->bs,bs2 = mat->bs2,i,*cworkA,*cworkB,**pcA,**pcB; 986 PetscInt nztot,nzA,nzB,lrow,brstart = matin->rmap->rstart,brend = matin->rmap->rend; 987 PetscInt *cmap,*idx_p,cstart = mat->rstartbs; 988 989 PetscFunctionBegin; 990 if (mat->getrowactive) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Already active"); 991 mat->getrowactive = PETSC_TRUE; 992 993 if (!mat->rowvalues && (idx || v)) { 994 /* 995 allocate enough space to hold information from the longest row. 996 */ 997 Mat_SeqSBAIJ *Aa = (Mat_SeqSBAIJ*)mat->A->data; 998 Mat_SeqBAIJ *Ba = (Mat_SeqBAIJ*)mat->B->data; 999 PetscInt max = 1,mbs = mat->mbs,tmp; 1000 for (i=0; i<mbs; i++) { 1001 tmp = Aa->i[i+1] - Aa->i[i] + Ba->i[i+1] - Ba->i[i]; /* row length */ 1002 if (max < tmp) max = tmp; 1003 } 1004 ierr = PetscMalloc2(max*bs2,&mat->rowvalues,max*bs2,&mat->rowindices);CHKERRQ(ierr); 1005 } 1006 1007 if (row < brstart || row >= brend) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Only local rows"); 1008 lrow = row - brstart; /* local row index */ 1009 1010 pvA = &vworkA; pcA = &cworkA; pvB = &vworkB; pcB = &cworkB; 1011 if (!v) {pvA = 0; pvB = 0;} 1012 if (!idx) {pcA = 0; if (!v) pcB = 0;} 1013 ierr = (*mat->A->ops->getrow)(mat->A,lrow,&nzA,pcA,pvA);CHKERRQ(ierr); 1014 ierr = (*mat->B->ops->getrow)(mat->B,lrow,&nzB,pcB,pvB);CHKERRQ(ierr); 1015 nztot = nzA + nzB; 1016 1017 cmap = mat->garray; 1018 if (v || idx) { 1019 if (nztot) { 1020 /* Sort by increasing column numbers, assuming A and B already sorted */ 1021 PetscInt imark = -1; 1022 if (v) { 1023 *v = v_p = mat->rowvalues; 1024 for (i=0; i<nzB; i++) { 1025 if (cmap[cworkB[i]/bs] < cstart) v_p[i] = vworkB[i]; 1026 else break; 1027 } 1028 imark = i; 1029 for (i=0; i<nzA; i++) v_p[imark+i] = vworkA[i]; 1030 for (i=imark; i<nzB; i++) v_p[nzA+i] = vworkB[i]; 1031 } 1032 if (idx) { 1033 *idx = idx_p = mat->rowindices; 1034 if (imark > -1) { 1035 for (i=0; i<imark; i++) { 1036 idx_p[i] = cmap[cworkB[i]/bs]*bs + cworkB[i]%bs; 1037 } 1038 } else { 1039 for (i=0; i<nzB; i++) { 1040 if (cmap[cworkB[i]/bs] < cstart) idx_p[i] = cmap[cworkB[i]/bs]*bs + cworkB[i]%bs; 1041 else break; 1042 } 1043 imark = i; 1044 } 1045 for (i=0; i<nzA; i++) idx_p[imark+i] = cstart*bs + cworkA[i]; 1046 for (i=imark; i<nzB; i++) idx_p[nzA+i] = cmap[cworkB[i]/bs]*bs + cworkB[i]%bs ; 1047 } 1048 } else { 1049 if (idx) *idx = 0; 1050 if (v) *v = 0; 1051 } 1052 } 1053 *nz = nztot; 1054 ierr = (*mat->A->ops->restorerow)(mat->A,lrow,&nzA,pcA,pvA);CHKERRQ(ierr); 1055 ierr = (*mat->B->ops->restorerow)(mat->B,lrow,&nzB,pcB,pvB);CHKERRQ(ierr); 1056 PetscFunctionReturn(0); 1057 } 1058 1059 #undef __FUNCT__ 1060 #define __FUNCT__ "MatRestoreRow_MPISBAIJ" 1061 PetscErrorCode MatRestoreRow_MPISBAIJ(Mat mat,PetscInt row,PetscInt *nz,PetscInt **idx,PetscScalar **v) 1062 { 1063 Mat_MPISBAIJ *baij = (Mat_MPISBAIJ*)mat->data; 1064 1065 PetscFunctionBegin; 1066 if (!baij->getrowactive) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"MatGetRow() must be called first"); 1067 baij->getrowactive = PETSC_FALSE; 1068 PetscFunctionReturn(0); 1069 } 1070 1071 #undef __FUNCT__ 1072 #define __FUNCT__ "MatGetRowUpperTriangular_MPISBAIJ" 1073 PetscErrorCode MatGetRowUpperTriangular_MPISBAIJ(Mat A) 1074 { 1075 Mat_MPISBAIJ *a = (Mat_MPISBAIJ*)A->data; 1076 Mat_SeqSBAIJ *aA = (Mat_SeqSBAIJ*)a->A->data; 1077 1078 PetscFunctionBegin; 1079 aA->getrow_utriangular = PETSC_TRUE; 1080 PetscFunctionReturn(0); 1081 } 1082 #undef __FUNCT__ 1083 #define __FUNCT__ "MatRestoreRowUpperTriangular_MPISBAIJ" 1084 PetscErrorCode MatRestoreRowUpperTriangular_MPISBAIJ(Mat A) 1085 { 1086 Mat_MPISBAIJ *a = (Mat_MPISBAIJ*)A->data; 1087 Mat_SeqSBAIJ *aA = (Mat_SeqSBAIJ*)a->A->data; 1088 1089 PetscFunctionBegin; 1090 aA->getrow_utriangular = PETSC_FALSE; 1091 PetscFunctionReturn(0); 1092 } 1093 1094 #undef __FUNCT__ 1095 #define __FUNCT__ "MatRealPart_MPISBAIJ" 1096 PetscErrorCode MatRealPart_MPISBAIJ(Mat A) 1097 { 1098 Mat_MPISBAIJ *a = (Mat_MPISBAIJ*)A->data; 1099 PetscErrorCode ierr; 1100 1101 PetscFunctionBegin; 1102 ierr = MatRealPart(a->A);CHKERRQ(ierr); 1103 ierr = MatRealPart(a->B);CHKERRQ(ierr); 1104 PetscFunctionReturn(0); 1105 } 1106 1107 #undef __FUNCT__ 1108 #define __FUNCT__ "MatImaginaryPart_MPISBAIJ" 1109 PetscErrorCode MatImaginaryPart_MPISBAIJ(Mat A) 1110 { 1111 Mat_MPISBAIJ *a = (Mat_MPISBAIJ*)A->data; 1112 PetscErrorCode ierr; 1113 1114 PetscFunctionBegin; 1115 ierr = MatImaginaryPart(a->A);CHKERRQ(ierr); 1116 ierr = MatImaginaryPart(a->B);CHKERRQ(ierr); 1117 PetscFunctionReturn(0); 1118 } 1119 1120 #undef __FUNCT__ 1121 #define __FUNCT__ "MatZeroEntries_MPISBAIJ" 1122 PetscErrorCode MatZeroEntries_MPISBAIJ(Mat A) 1123 { 1124 Mat_MPISBAIJ *l = (Mat_MPISBAIJ*)A->data; 1125 PetscErrorCode ierr; 1126 1127 PetscFunctionBegin; 1128 ierr = MatZeroEntries(l->A);CHKERRQ(ierr); 1129 ierr = MatZeroEntries(l->B);CHKERRQ(ierr); 1130 PetscFunctionReturn(0); 1131 } 1132 1133 #undef __FUNCT__ 1134 #define __FUNCT__ "MatGetInfo_MPISBAIJ" 1135 PetscErrorCode MatGetInfo_MPISBAIJ(Mat matin,MatInfoType flag,MatInfo *info) 1136 { 1137 Mat_MPISBAIJ *a = (Mat_MPISBAIJ*)matin->data; 1138 Mat A = a->A,B = a->B; 1139 PetscErrorCode ierr; 1140 PetscReal isend[5],irecv[5]; 1141 1142 PetscFunctionBegin; 1143 info->block_size = (PetscReal)matin->rmap->bs; 1144 1145 ierr = MatGetInfo(A,MAT_LOCAL,info);CHKERRQ(ierr); 1146 1147 isend[0] = info->nz_used; isend[1] = info->nz_allocated; isend[2] = info->nz_unneeded; 1148 isend[3] = info->memory; isend[4] = info->mallocs; 1149 1150 ierr = MatGetInfo(B,MAT_LOCAL,info);CHKERRQ(ierr); 1151 1152 isend[0] += info->nz_used; isend[1] += info->nz_allocated; isend[2] += info->nz_unneeded; 1153 isend[3] += info->memory; isend[4] += info->mallocs; 1154 if (flag == MAT_LOCAL) { 1155 info->nz_used = isend[0]; 1156 info->nz_allocated = isend[1]; 1157 info->nz_unneeded = isend[2]; 1158 info->memory = isend[3]; 1159 info->mallocs = isend[4]; 1160 } else if (flag == MAT_GLOBAL_MAX) { 1161 ierr = MPI_Allreduce(isend,irecv,5,MPIU_REAL,MPIU_MAX,PetscObjectComm((PetscObject)matin));CHKERRQ(ierr); 1162 1163 info->nz_used = irecv[0]; 1164 info->nz_allocated = irecv[1]; 1165 info->nz_unneeded = irecv[2]; 1166 info->memory = irecv[3]; 1167 info->mallocs = irecv[4]; 1168 } else if (flag == MAT_GLOBAL_SUM) { 1169 ierr = MPI_Allreduce(isend,irecv,5,MPIU_REAL,MPIU_SUM,PetscObjectComm((PetscObject)matin));CHKERRQ(ierr); 1170 1171 info->nz_used = irecv[0]; 1172 info->nz_allocated = irecv[1]; 1173 info->nz_unneeded = irecv[2]; 1174 info->memory = irecv[3]; 1175 info->mallocs = irecv[4]; 1176 } else SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Unknown MatInfoType argument %d",(int)flag); 1177 info->fill_ratio_given = 0; /* no parallel LU/ILU/Cholesky */ 1178 info->fill_ratio_needed = 0; 1179 info->factor_mallocs = 0; 1180 PetscFunctionReturn(0); 1181 } 1182 1183 #undef __FUNCT__ 1184 #define __FUNCT__ "MatSetOption_MPISBAIJ" 1185 PetscErrorCode MatSetOption_MPISBAIJ(Mat A,MatOption op,PetscBool flg) 1186 { 1187 Mat_MPISBAIJ *a = (Mat_MPISBAIJ*)A->data; 1188 Mat_SeqSBAIJ *aA = (Mat_SeqSBAIJ*)a->A->data; 1189 PetscErrorCode ierr; 1190 1191 PetscFunctionBegin; 1192 switch (op) { 1193 case MAT_NEW_NONZERO_LOCATIONS: 1194 case MAT_NEW_NONZERO_ALLOCATION_ERR: 1195 case MAT_UNUSED_NONZERO_LOCATION_ERR: 1196 case MAT_KEEP_NONZERO_PATTERN: 1197 case MAT_NEW_NONZERO_LOCATION_ERR: 1198 ierr = MatSetOption(a->A,op,flg);CHKERRQ(ierr); 1199 ierr = MatSetOption(a->B,op,flg);CHKERRQ(ierr); 1200 break; 1201 case MAT_ROW_ORIENTED: 1202 a->roworiented = flg; 1203 1204 ierr = MatSetOption(a->A,op,flg);CHKERRQ(ierr); 1205 ierr = MatSetOption(a->B,op,flg);CHKERRQ(ierr); 1206 break; 1207 case MAT_NEW_DIAGONALS: 1208 ierr = PetscInfo1(A,"Option %s ignored\n",MatOptions[op]);CHKERRQ(ierr); 1209 break; 1210 case MAT_IGNORE_OFF_PROC_ENTRIES: 1211 a->donotstash = flg; 1212 break; 1213 case MAT_USE_HASH_TABLE: 1214 a->ht_flag = flg; 1215 break; 1216 case MAT_HERMITIAN: 1217 if (!A->assembled) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Must call MatAssemblyEnd() first"); 1218 ierr = MatSetOption(a->A,op,flg);CHKERRQ(ierr); 1219 1220 A->ops->mult = MatMult_MPISBAIJ_Hermitian; 1221 break; 1222 case MAT_SPD: 1223 A->spd_set = PETSC_TRUE; 1224 A->spd = flg; 1225 if (flg) { 1226 A->symmetric = PETSC_TRUE; 1227 A->structurally_symmetric = PETSC_TRUE; 1228 A->symmetric_set = PETSC_TRUE; 1229 A->structurally_symmetric_set = PETSC_TRUE; 1230 } 1231 break; 1232 case MAT_SYMMETRIC: 1233 ierr = MatSetOption(a->A,op,flg);CHKERRQ(ierr); 1234 break; 1235 case MAT_STRUCTURALLY_SYMMETRIC: 1236 ierr = MatSetOption(a->A,op,flg);CHKERRQ(ierr); 1237 break; 1238 case MAT_SYMMETRY_ETERNAL: 1239 if (!flg) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Matrix must be symmetric"); 1240 ierr = PetscInfo1(A,"Option %s ignored\n",MatOptions[op]);CHKERRQ(ierr); 1241 break; 1242 case MAT_IGNORE_LOWER_TRIANGULAR: 1243 aA->ignore_ltriangular = flg; 1244 break; 1245 case MAT_ERROR_LOWER_TRIANGULAR: 1246 aA->ignore_ltriangular = flg; 1247 break; 1248 case MAT_GETROW_UPPERTRIANGULAR: 1249 aA->getrow_utriangular = flg; 1250 break; 1251 default: 1252 SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"unknown option %d",op); 1253 } 1254 PetscFunctionReturn(0); 1255 } 1256 1257 #undef __FUNCT__ 1258 #define __FUNCT__ "MatTranspose_MPISBAIJ" 1259 PetscErrorCode MatTranspose_MPISBAIJ(Mat A,MatReuse reuse,Mat *B) 1260 { 1261 PetscErrorCode ierr; 1262 1263 PetscFunctionBegin; 1264 if (MAT_INITIAL_MATRIX || *B != A) { 1265 ierr = MatDuplicate(A,MAT_COPY_VALUES,B);CHKERRQ(ierr); 1266 } 1267 PetscFunctionReturn(0); 1268 } 1269 1270 #undef __FUNCT__ 1271 #define __FUNCT__ "MatDiagonalScale_MPISBAIJ" 1272 PetscErrorCode MatDiagonalScale_MPISBAIJ(Mat mat,Vec ll,Vec rr) 1273 { 1274 Mat_MPISBAIJ *baij = (Mat_MPISBAIJ*)mat->data; 1275 Mat a = baij->A, b=baij->B; 1276 PetscErrorCode ierr; 1277 PetscInt nv,m,n; 1278 PetscBool flg; 1279 1280 PetscFunctionBegin; 1281 if (ll != rr) { 1282 ierr = VecEqual(ll,rr,&flg);CHKERRQ(ierr); 1283 if (!flg) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"For symmetric format, left and right scaling vectors must be same\n"); 1284 } 1285 if (!ll) PetscFunctionReturn(0); 1286 1287 ierr = MatGetLocalSize(mat,&m,&n);CHKERRQ(ierr); 1288 if (m != n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"For symmetric format, local size %d %d must be same",m,n); 1289 1290 ierr = VecGetLocalSize(rr,&nv);CHKERRQ(ierr); 1291 if (nv!=n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Left and right vector non-conforming local size"); 1292 1293 ierr = VecScatterBegin(baij->Mvctx,rr,baij->lvec,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 1294 1295 /* left diagonalscale the off-diagonal part */ 1296 ierr = (*b->ops->diagonalscale)(b,ll,NULL);CHKERRQ(ierr); 1297 1298 /* scale the diagonal part */ 1299 ierr = (*a->ops->diagonalscale)(a,ll,rr);CHKERRQ(ierr); 1300 1301 /* right diagonalscale the off-diagonal part */ 1302 ierr = VecScatterEnd(baij->Mvctx,rr,baij->lvec,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 1303 ierr = (*b->ops->diagonalscale)(b,NULL,baij->lvec);CHKERRQ(ierr); 1304 PetscFunctionReturn(0); 1305 } 1306 1307 #undef __FUNCT__ 1308 #define __FUNCT__ "MatSetUnfactored_MPISBAIJ" 1309 PetscErrorCode MatSetUnfactored_MPISBAIJ(Mat A) 1310 { 1311 Mat_MPISBAIJ *a = (Mat_MPISBAIJ*)A->data; 1312 PetscErrorCode ierr; 1313 1314 PetscFunctionBegin; 1315 ierr = MatSetUnfactored(a->A);CHKERRQ(ierr); 1316 PetscFunctionReturn(0); 1317 } 1318 1319 static PetscErrorCode MatDuplicate_MPISBAIJ(Mat,MatDuplicateOption,Mat*); 1320 1321 #undef __FUNCT__ 1322 #define __FUNCT__ "MatEqual_MPISBAIJ" 1323 PetscErrorCode MatEqual_MPISBAIJ(Mat A,Mat B,PetscBool *flag) 1324 { 1325 Mat_MPISBAIJ *matB = (Mat_MPISBAIJ*)B->data,*matA = (Mat_MPISBAIJ*)A->data; 1326 Mat a,b,c,d; 1327 PetscBool flg; 1328 PetscErrorCode ierr; 1329 1330 PetscFunctionBegin; 1331 a = matA->A; b = matA->B; 1332 c = matB->A; d = matB->B; 1333 1334 ierr = MatEqual(a,c,&flg);CHKERRQ(ierr); 1335 if (flg) { 1336 ierr = MatEqual(b,d,&flg);CHKERRQ(ierr); 1337 } 1338 ierr = MPI_Allreduce(&flg,flag,1,MPIU_BOOL,MPI_LAND,PetscObjectComm((PetscObject)A));CHKERRQ(ierr); 1339 PetscFunctionReturn(0); 1340 } 1341 1342 #undef __FUNCT__ 1343 #define __FUNCT__ "MatCopy_MPISBAIJ" 1344 PetscErrorCode MatCopy_MPISBAIJ(Mat A,Mat B,MatStructure str) 1345 { 1346 PetscErrorCode ierr; 1347 Mat_MPISBAIJ *a = (Mat_MPISBAIJ*)A->data; 1348 Mat_MPISBAIJ *b = (Mat_MPISBAIJ*)B->data; 1349 1350 PetscFunctionBegin; 1351 /* If the two matrices don't have the same copy implementation, they aren't compatible for fast copy. */ 1352 if ((str != SAME_NONZERO_PATTERN) || (A->ops->copy != B->ops->copy)) { 1353 ierr = MatGetRowUpperTriangular(A);CHKERRQ(ierr); 1354 ierr = MatCopy_Basic(A,B,str);CHKERRQ(ierr); 1355 ierr = MatRestoreRowUpperTriangular(A);CHKERRQ(ierr); 1356 } else { 1357 ierr = MatCopy(a->A,b->A,str);CHKERRQ(ierr); 1358 ierr = MatCopy(a->B,b->B,str);CHKERRQ(ierr); 1359 } 1360 PetscFunctionReturn(0); 1361 } 1362 1363 #undef __FUNCT__ 1364 #define __FUNCT__ "MatSetUp_MPISBAIJ" 1365 PetscErrorCode MatSetUp_MPISBAIJ(Mat A) 1366 { 1367 PetscErrorCode ierr; 1368 1369 PetscFunctionBegin; 1370 ierr = MatMPISBAIJSetPreallocation(A,A->rmap->bs,PETSC_DEFAULT,0,PETSC_DEFAULT,0);CHKERRQ(ierr); 1371 PetscFunctionReturn(0); 1372 } 1373 1374 #undef __FUNCT__ 1375 #define __FUNCT__ "MatAXPY_MPISBAIJ" 1376 PetscErrorCode MatAXPY_MPISBAIJ(Mat Y,PetscScalar a,Mat X,MatStructure str) 1377 { 1378 PetscErrorCode ierr; 1379 Mat_MPISBAIJ *xx=(Mat_MPISBAIJ*)X->data,*yy=(Mat_MPISBAIJ*)Y->data; 1380 PetscBLASInt bnz,one=1; 1381 Mat_SeqSBAIJ *xa,*ya; 1382 Mat_SeqBAIJ *xb,*yb; 1383 1384 PetscFunctionBegin; 1385 if (str == SAME_NONZERO_PATTERN) { 1386 PetscScalar alpha = a; 1387 xa = (Mat_SeqSBAIJ*)xx->A->data; 1388 ya = (Mat_SeqSBAIJ*)yy->A->data; 1389 ierr = PetscBLASIntCast(xa->nz,&bnz);CHKERRQ(ierr); 1390 PetscStackCallBLAS("BLASaxpy",BLASaxpy_(&bnz,&alpha,xa->a,&one,ya->a,&one)); 1391 xb = (Mat_SeqBAIJ*)xx->B->data; 1392 yb = (Mat_SeqBAIJ*)yy->B->data; 1393 ierr = PetscBLASIntCast(xb->nz,&bnz);CHKERRQ(ierr); 1394 PetscStackCallBLAS("BLASaxpy",BLASaxpy_(&bnz,&alpha,xb->a,&one,yb->a,&one)); 1395 } else { 1396 ierr = MatGetRowUpperTriangular(X);CHKERRQ(ierr); 1397 ierr = MatAXPY_Basic(Y,a,X,str);CHKERRQ(ierr); 1398 ierr = MatRestoreRowUpperTriangular(X);CHKERRQ(ierr); 1399 } 1400 PetscFunctionReturn(0); 1401 } 1402 1403 #undef __FUNCT__ 1404 #define __FUNCT__ "MatGetSubMatrices_MPISBAIJ" 1405 PetscErrorCode MatGetSubMatrices_MPISBAIJ(Mat A,PetscInt n,const IS irow[],const IS icol[],MatReuse scall,Mat *B[]) 1406 { 1407 PetscErrorCode ierr; 1408 PetscInt i; 1409 PetscBool flg; 1410 1411 PetscFunctionBegin; 1412 for (i=0; i<n; i++) { 1413 ierr = ISEqual(irow[i],icol[i],&flg);CHKERRQ(ierr); 1414 if (!flg) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Can only get symmetric submatrix for MPISBAIJ matrices"); 1415 } 1416 ierr = MatGetSubMatrices_MPIBAIJ(A,n,irow,icol,scall,B);CHKERRQ(ierr); 1417 PetscFunctionReturn(0); 1418 } 1419 1420 1421 /* -------------------------------------------------------------------*/ 1422 static struct _MatOps MatOps_Values = {MatSetValues_MPISBAIJ, 1423 MatGetRow_MPISBAIJ, 1424 MatRestoreRow_MPISBAIJ, 1425 MatMult_MPISBAIJ, 1426 /* 4*/ MatMultAdd_MPISBAIJ, 1427 MatMult_MPISBAIJ, /* transpose versions are same as non-transpose */ 1428 MatMultAdd_MPISBAIJ, 1429 0, 1430 0, 1431 0, 1432 /* 10*/ 0, 1433 0, 1434 0, 1435 MatSOR_MPISBAIJ, 1436 MatTranspose_MPISBAIJ, 1437 /* 15*/ MatGetInfo_MPISBAIJ, 1438 MatEqual_MPISBAIJ, 1439 MatGetDiagonal_MPISBAIJ, 1440 MatDiagonalScale_MPISBAIJ, 1441 MatNorm_MPISBAIJ, 1442 /* 20*/ MatAssemblyBegin_MPISBAIJ, 1443 MatAssemblyEnd_MPISBAIJ, 1444 MatSetOption_MPISBAIJ, 1445 MatZeroEntries_MPISBAIJ, 1446 /* 24*/ 0, 1447 0, 1448 0, 1449 0, 1450 0, 1451 /* 29*/ MatSetUp_MPISBAIJ, 1452 0, 1453 0, 1454 0, 1455 0, 1456 /* 34*/ MatDuplicate_MPISBAIJ, 1457 0, 1458 0, 1459 0, 1460 0, 1461 /* 39*/ MatAXPY_MPISBAIJ, 1462 MatGetSubMatrices_MPISBAIJ, 1463 MatIncreaseOverlap_MPISBAIJ, 1464 MatGetValues_MPISBAIJ, 1465 MatCopy_MPISBAIJ, 1466 /* 44*/ 0, 1467 MatScale_MPISBAIJ, 1468 0, 1469 0, 1470 0, 1471 /* 49*/ 0, 1472 0, 1473 0, 1474 0, 1475 0, 1476 /* 54*/ 0, 1477 0, 1478 MatSetUnfactored_MPISBAIJ, 1479 0, 1480 MatSetValuesBlocked_MPISBAIJ, 1481 /* 59*/ 0, 1482 0, 1483 0, 1484 0, 1485 0, 1486 /* 64*/ 0, 1487 0, 1488 0, 1489 0, 1490 0, 1491 /* 69*/ MatGetRowMaxAbs_MPISBAIJ, 1492 0, 1493 0, 1494 0, 1495 0, 1496 /* 74*/ 0, 1497 0, 1498 0, 1499 0, 1500 0, 1501 /* 79*/ 0, 1502 0, 1503 0, 1504 0, 1505 MatLoad_MPISBAIJ, 1506 /* 84*/ 0, 1507 0, 1508 0, 1509 0, 1510 0, 1511 /* 89*/ 0, 1512 0, 1513 0, 1514 0, 1515 0, 1516 /* 94*/ 0, 1517 0, 1518 0, 1519 0, 1520 0, 1521 /* 99*/ 0, 1522 0, 1523 0, 1524 0, 1525 0, 1526 /*104*/ 0, 1527 MatRealPart_MPISBAIJ, 1528 MatImaginaryPart_MPISBAIJ, 1529 MatGetRowUpperTriangular_MPISBAIJ, 1530 MatRestoreRowUpperTriangular_MPISBAIJ, 1531 /*109*/ 0, 1532 0, 1533 0, 1534 0, 1535 0, 1536 /*114*/ 0, 1537 0, 1538 0, 1539 0, 1540 0, 1541 /*119*/ 0, 1542 0, 1543 0, 1544 0, 1545 0, 1546 /*124*/ 0, 1547 0, 1548 0, 1549 0, 1550 0, 1551 /*129*/ 0, 1552 0, 1553 0, 1554 0, 1555 0, 1556 /*134*/ 0, 1557 0, 1558 0, 1559 0, 1560 0, 1561 /*139*/ 0, 1562 0, 1563 0 1564 }; 1565 1566 1567 #undef __FUNCT__ 1568 #define __FUNCT__ "MatGetDiagonalBlock_MPISBAIJ" 1569 PetscErrorCode MatGetDiagonalBlock_MPISBAIJ(Mat A,Mat *a) 1570 { 1571 PetscFunctionBegin; 1572 *a = ((Mat_MPISBAIJ*)A->data)->A; 1573 PetscFunctionReturn(0); 1574 } 1575 1576 #undef __FUNCT__ 1577 #define __FUNCT__ "MatMPISBAIJSetPreallocation_MPISBAIJ" 1578 PetscErrorCode MatMPISBAIJSetPreallocation_MPISBAIJ(Mat B,PetscInt bs,PetscInt d_nz,const PetscInt *d_nnz,PetscInt o_nz,const PetscInt *o_nnz) 1579 { 1580 Mat_MPISBAIJ *b; 1581 PetscErrorCode ierr; 1582 PetscInt i,mbs,Mbs; 1583 1584 PetscFunctionBegin; 1585 ierr = PetscLayoutSetBlockSize(B->rmap,bs);CHKERRQ(ierr); 1586 ierr = PetscLayoutSetBlockSize(B->cmap,bs);CHKERRQ(ierr); 1587 ierr = PetscLayoutSetUp(B->rmap);CHKERRQ(ierr); 1588 ierr = PetscLayoutSetUp(B->cmap);CHKERRQ(ierr); 1589 ierr = PetscLayoutGetBlockSize(B->rmap,&bs);CHKERRQ(ierr); 1590 1591 b = (Mat_MPISBAIJ*)B->data; 1592 mbs = B->rmap->n/bs; 1593 Mbs = B->rmap->N/bs; 1594 if (mbs*bs != B->rmap->n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"No of local rows %D must be divisible by blocksize %D",B->rmap->N,bs); 1595 1596 B->rmap->bs = bs; 1597 b->bs2 = bs*bs; 1598 b->mbs = mbs; 1599 b->nbs = mbs; 1600 b->Mbs = Mbs; 1601 b->Nbs = Mbs; 1602 1603 for (i=0; i<=b->size; i++) { 1604 b->rangebs[i] = B->rmap->range[i]/bs; 1605 } 1606 b->rstartbs = B->rmap->rstart/bs; 1607 b->rendbs = B->rmap->rend/bs; 1608 1609 b->cstartbs = B->cmap->rstart/bs; 1610 b->cendbs = B->cmap->rend/bs; 1611 1612 if (!B->preallocated) { 1613 ierr = MatCreate(PETSC_COMM_SELF,&b->A);CHKERRQ(ierr); 1614 ierr = MatSetSizes(b->A,B->rmap->n,B->cmap->n,B->rmap->n,B->cmap->n);CHKERRQ(ierr); 1615 ierr = MatSetType(b->A,MATSEQSBAIJ);CHKERRQ(ierr); 1616 ierr = PetscLogObjectParent((PetscObject)B,(PetscObject)b->A);CHKERRQ(ierr); 1617 ierr = MatCreate(PETSC_COMM_SELF,&b->B);CHKERRQ(ierr); 1618 ierr = MatSetSizes(b->B,B->rmap->n,B->cmap->N,B->rmap->n,B->cmap->N);CHKERRQ(ierr); 1619 ierr = MatSetType(b->B,MATSEQBAIJ);CHKERRQ(ierr); 1620 ierr = PetscLogObjectParent((PetscObject)B,(PetscObject)b->B);CHKERRQ(ierr); 1621 ierr = MatStashCreate_Private(PetscObjectComm((PetscObject)B),bs,&B->bstash);CHKERRQ(ierr); 1622 } 1623 1624 ierr = MatSeqSBAIJSetPreallocation(b->A,bs,d_nz,d_nnz);CHKERRQ(ierr); 1625 ierr = MatSeqBAIJSetPreallocation(b->B,bs,o_nz,o_nnz);CHKERRQ(ierr); 1626 1627 B->preallocated = PETSC_TRUE; 1628 PetscFunctionReturn(0); 1629 } 1630 1631 #undef __FUNCT__ 1632 #define __FUNCT__ "MatMPISBAIJSetPreallocationCSR_MPISBAIJ" 1633 PetscErrorCode MatMPISBAIJSetPreallocationCSR_MPISBAIJ(Mat B,PetscInt bs,const PetscInt ii[],const PetscInt jj[],const PetscScalar V[]) 1634 { 1635 PetscInt m,rstart,cstart,cend; 1636 PetscInt i,j,d,nz,nz_max=0,*d_nnz=0,*o_nnz=0; 1637 const PetscInt *JJ =0; 1638 PetscScalar *values=0; 1639 PetscErrorCode ierr; 1640 1641 PetscFunctionBegin; 1642 if (bs < 1) SETERRQ1(PetscObjectComm((PetscObject)B),PETSC_ERR_ARG_OUTOFRANGE,"Invalid block size specified, must be positive but it is %D",bs); 1643 ierr = PetscLayoutSetBlockSize(B->rmap,bs);CHKERRQ(ierr); 1644 ierr = PetscLayoutSetBlockSize(B->cmap,bs);CHKERRQ(ierr); 1645 ierr = PetscLayoutSetUp(B->rmap);CHKERRQ(ierr); 1646 ierr = PetscLayoutSetUp(B->cmap);CHKERRQ(ierr); 1647 ierr = PetscLayoutGetBlockSize(B->rmap,&bs);CHKERRQ(ierr); 1648 m = B->rmap->n/bs; 1649 rstart = B->rmap->rstart/bs; 1650 cstart = B->cmap->rstart/bs; 1651 cend = B->cmap->rend/bs; 1652 1653 if (ii[0]) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"ii[0] must be 0 but it is %D",ii[0]); 1654 ierr = PetscMalloc2(m,&d_nnz,m,&o_nnz);CHKERRQ(ierr); 1655 for (i=0; i<m; i++) { 1656 nz = ii[i+1] - ii[i]; 1657 if (nz < 0) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Local row %D has a negative number of columns %D",i,nz); 1658 nz_max = PetscMax(nz_max,nz); 1659 JJ = jj + ii[i]; 1660 for (j=0; j<nz; j++) { 1661 if (*JJ >= cstart) break; 1662 JJ++; 1663 } 1664 d = 0; 1665 for (; j<nz; j++) { 1666 if (*JJ++ >= cend) break; 1667 d++; 1668 } 1669 d_nnz[i] = d; 1670 o_nnz[i] = nz - d; 1671 } 1672 ierr = MatMPISBAIJSetPreallocation(B,bs,0,d_nnz,0,o_nnz);CHKERRQ(ierr); 1673 ierr = PetscFree2(d_nnz,o_nnz);CHKERRQ(ierr); 1674 1675 values = (PetscScalar*)V; 1676 if (!values) { 1677 ierr = PetscMalloc1(bs*bs*nz_max,&values);CHKERRQ(ierr); 1678 ierr = PetscMemzero(values,bs*bs*nz_max*sizeof(PetscScalar));CHKERRQ(ierr); 1679 } 1680 for (i=0; i<m; i++) { 1681 PetscInt row = i + rstart; 1682 PetscInt ncols = ii[i+1] - ii[i]; 1683 const PetscInt *icols = jj + ii[i]; 1684 const PetscScalar *svals = values + (V ? (bs*bs*ii[i]) : 0); 1685 ierr = MatSetValuesBlocked_MPISBAIJ(B,1,&row,ncols,icols,svals,INSERT_VALUES);CHKERRQ(ierr); 1686 } 1687 1688 if (!V) { ierr = PetscFree(values);CHKERRQ(ierr); } 1689 ierr = MatAssemblyBegin(B,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1690 ierr = MatAssemblyEnd(B,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1691 ierr = MatSetOption(B,MAT_NEW_NONZERO_LOCATION_ERR,PETSC_TRUE);CHKERRQ(ierr); 1692 PetscFunctionReturn(0); 1693 } 1694 1695 #if defined(PETSC_HAVE_MUMPS) 1696 PETSC_EXTERN PetscErrorCode MatGetFactor_sbaij_mumps(Mat,MatFactorType,Mat*); 1697 #endif 1698 #if defined(PETSC_HAVE_PASTIX) 1699 PETSC_EXTERN PetscErrorCode MatGetFactor_mpisbaij_pastix(Mat,MatFactorType,Mat*); 1700 #endif 1701 1702 /*MC 1703 MATMPISBAIJ - MATMPISBAIJ = "mpisbaij" - A matrix type to be used for distributed symmetric sparse block matrices, 1704 based on block compressed sparse row format. Only the upper triangular portion of the "diagonal" portion of 1705 the matrix is stored. 1706 1707 For complex numbers by default this matrix is symmetric, NOT Hermitian symmetric. To make it Hermitian symmetric you 1708 can call MatSetOption(Mat, MAT_HERMITIAN); 1709 1710 Options Database Keys: 1711 . -mat_type mpisbaij - sets the matrix type to "mpisbaij" during a call to MatSetFromOptions() 1712 1713 Level: beginner 1714 1715 .seealso: MatCreateMPISBAIJ 1716 M*/ 1717 1718 PETSC_EXTERN PetscErrorCode MatConvert_MPISBAIJ_MPISBSTRM(Mat,MatType,MatReuse,Mat*); 1719 1720 #undef __FUNCT__ 1721 #define __FUNCT__ "MatCreate_MPISBAIJ" 1722 PETSC_EXTERN PetscErrorCode MatCreate_MPISBAIJ(Mat B) 1723 { 1724 Mat_MPISBAIJ *b; 1725 PetscErrorCode ierr; 1726 PetscBool flg; 1727 1728 PetscFunctionBegin; 1729 ierr = PetscNewLog(B,&b);CHKERRQ(ierr); 1730 B->data = (void*)b; 1731 ierr = PetscMemcpy(B->ops,&MatOps_Values,sizeof(struct _MatOps));CHKERRQ(ierr); 1732 1733 B->ops->destroy = MatDestroy_MPISBAIJ; 1734 B->ops->view = MatView_MPISBAIJ; 1735 B->assembled = PETSC_FALSE; 1736 B->insertmode = NOT_SET_VALUES; 1737 1738 ierr = MPI_Comm_rank(PetscObjectComm((PetscObject)B),&b->rank);CHKERRQ(ierr); 1739 ierr = MPI_Comm_size(PetscObjectComm((PetscObject)B),&b->size);CHKERRQ(ierr); 1740 1741 /* build local table of row and column ownerships */ 1742 ierr = PetscMalloc1((b->size+2),&b->rangebs);CHKERRQ(ierr); 1743 1744 /* build cache for off array entries formed */ 1745 ierr = MatStashCreate_Private(PetscObjectComm((PetscObject)B),1,&B->stash);CHKERRQ(ierr); 1746 1747 b->donotstash = PETSC_FALSE; 1748 b->colmap = NULL; 1749 b->garray = NULL; 1750 b->roworiented = PETSC_TRUE; 1751 1752 /* stuff used in block assembly */ 1753 b->barray = 0; 1754 1755 /* stuff used for matrix vector multiply */ 1756 b->lvec = 0; 1757 b->Mvctx = 0; 1758 b->slvec0 = 0; 1759 b->slvec0b = 0; 1760 b->slvec1 = 0; 1761 b->slvec1a = 0; 1762 b->slvec1b = 0; 1763 b->sMvctx = 0; 1764 1765 /* stuff for MatGetRow() */ 1766 b->rowindices = 0; 1767 b->rowvalues = 0; 1768 b->getrowactive = PETSC_FALSE; 1769 1770 /* hash table stuff */ 1771 b->ht = 0; 1772 b->hd = 0; 1773 b->ht_size = 0; 1774 b->ht_flag = PETSC_FALSE; 1775 b->ht_fact = 0; 1776 b->ht_total_ct = 0; 1777 b->ht_insert_ct = 0; 1778 1779 /* stuff for MatGetSubMatrices_MPIBAIJ_local() */ 1780 b->ijonly = PETSC_FALSE; 1781 1782 b->in_loc = 0; 1783 b->v_loc = 0; 1784 b->n_loc = 0; 1785 ierr = PetscOptionsBegin(PetscObjectComm((PetscObject)B),NULL,"Options for loading MPISBAIJ matrix 1","Mat");CHKERRQ(ierr); 1786 ierr = PetscOptionsBool("-mat_use_hash_table","Use hash table to save memory in constructing matrix","MatSetOption",PETSC_FALSE,&flg,NULL);CHKERRQ(ierr); 1787 if (flg) { 1788 PetscReal fact = 1.39; 1789 ierr = MatSetOption(B,MAT_USE_HASH_TABLE,PETSC_TRUE);CHKERRQ(ierr); 1790 ierr = PetscOptionsReal("-mat_use_hash_table","Use hash table factor","MatMPIBAIJSetHashTableFactor",fact,&fact,NULL);CHKERRQ(ierr); 1791 if (fact <= 1.0) fact = 1.39; 1792 ierr = MatMPIBAIJSetHashTableFactor(B,fact);CHKERRQ(ierr); 1793 ierr = PetscInfo1(B,"Hash table Factor used %5.2f\n",fact);CHKERRQ(ierr); 1794 } 1795 ierr = PetscOptionsEnd();CHKERRQ(ierr); 1796 1797 #if defined(PETSC_HAVE_PASTIX) 1798 ierr = PetscObjectComposeFunction((PetscObject)B,"MatGetFactor_pastix_C",MatGetFactor_mpisbaij_pastix);CHKERRQ(ierr); 1799 #endif 1800 #if defined(PETSC_HAVE_MUMPS) 1801 ierr = PetscObjectComposeFunction((PetscObject)B,"MatGetFactor_mumps_C",MatGetFactor_sbaij_mumps);CHKERRQ(ierr); 1802 #endif 1803 ierr = PetscObjectComposeFunction((PetscObject)B,"MatStoreValues_C",MatStoreValues_MPISBAIJ);CHKERRQ(ierr); 1804 ierr = PetscObjectComposeFunction((PetscObject)B,"MatRetrieveValues_C",MatRetrieveValues_MPISBAIJ);CHKERRQ(ierr); 1805 ierr = PetscObjectComposeFunction((PetscObject)B,"MatGetDiagonalBlock_C",MatGetDiagonalBlock_MPISBAIJ);CHKERRQ(ierr); 1806 ierr = PetscObjectComposeFunction((PetscObject)B,"MatMPISBAIJSetPreallocation_C",MatMPISBAIJSetPreallocation_MPISBAIJ);CHKERRQ(ierr); 1807 ierr = PetscObjectComposeFunction((PetscObject)B,"MatMPISBAIJSetPreallocationCSR_C",MatMPISBAIJSetPreallocationCSR_MPISBAIJ);CHKERRQ(ierr); 1808 ierr = PetscObjectComposeFunction((PetscObject)B,"MatConvert_mpisbaij_mpisbstrm_C",MatConvert_MPISBAIJ_MPISBSTRM);CHKERRQ(ierr); 1809 1810 B->symmetric = PETSC_TRUE; 1811 B->structurally_symmetric = PETSC_TRUE; 1812 B->symmetric_set = PETSC_TRUE; 1813 B->structurally_symmetric_set = PETSC_TRUE; 1814 1815 ierr = PetscObjectChangeTypeName((PetscObject)B,MATMPISBAIJ);CHKERRQ(ierr); 1816 PetscFunctionReturn(0); 1817 } 1818 1819 /*MC 1820 MATSBAIJ - MATSBAIJ = "sbaij" - A matrix type to be used for symmetric block sparse matrices. 1821 1822 This matrix type is identical to MATSEQSBAIJ when constructed with a single process communicator, 1823 and MATMPISBAIJ otherwise. 1824 1825 Options Database Keys: 1826 . -mat_type sbaij - sets the matrix type to "sbaij" during a call to MatSetFromOptions() 1827 1828 Level: beginner 1829 1830 .seealso: MatCreateMPISBAIJ,MATSEQSBAIJ,MATMPISBAIJ 1831 M*/ 1832 1833 #undef __FUNCT__ 1834 #define __FUNCT__ "MatMPISBAIJSetPreallocation" 1835 /*@C 1836 MatMPISBAIJSetPreallocation - For good matrix assembly performance 1837 the user should preallocate the matrix storage by setting the parameters 1838 d_nz (or d_nnz) and o_nz (or o_nnz). By setting these parameters accurately, 1839 performance can be increased by more than a factor of 50. 1840 1841 Collective on Mat 1842 1843 Input Parameters: 1844 + A - the matrix 1845 . bs - size of blockk 1846 . d_nz - number of block nonzeros per block row in diagonal portion of local 1847 submatrix (same for all local rows) 1848 . d_nnz - array containing the number of block nonzeros in the various block rows 1849 in the upper triangular and diagonal part of the in diagonal portion of the local 1850 (possibly different for each block row) or NULL. If you plan to factor the matrix you must leave room 1851 for the diagonal entry and set a value even if it is zero. 1852 . o_nz - number of block nonzeros per block row in the off-diagonal portion of local 1853 submatrix (same for all local rows). 1854 - o_nnz - array containing the number of nonzeros in the various block rows of the 1855 off-diagonal portion of the local submatrix that is right of the diagonal 1856 (possibly different for each block row) or NULL. 1857 1858 1859 Options Database Keys: 1860 . -mat_no_unroll - uses code that does not unroll the loops in the 1861 block calculations (much slower) 1862 . -mat_block_size - size of the blocks to use 1863 1864 Notes: 1865 1866 If PETSC_DECIDE or PETSC_DETERMINE is used for a particular argument on one processor 1867 than it must be used on all processors that share the object for that argument. 1868 1869 If the *_nnz parameter is given then the *_nz parameter is ignored 1870 1871 Storage Information: 1872 For a square global matrix we define each processor's diagonal portion 1873 to be its local rows and the corresponding columns (a square submatrix); 1874 each processor's off-diagonal portion encompasses the remainder of the 1875 local matrix (a rectangular submatrix). 1876 1877 The user can specify preallocated storage for the diagonal part of 1878 the local submatrix with either d_nz or d_nnz (not both). Set 1879 d_nz=PETSC_DEFAULT and d_nnz=NULL for PETSc to control dynamic 1880 memory allocation. Likewise, specify preallocated storage for the 1881 off-diagonal part of the local submatrix with o_nz or o_nnz (not both). 1882 1883 You can call MatGetInfo() to get information on how effective the preallocation was; 1884 for example the fields mallocs,nz_allocated,nz_used,nz_unneeded; 1885 You can also run with the option -info and look for messages with the string 1886 malloc in them to see if additional memory allocation was needed. 1887 1888 Consider a processor that owns rows 3, 4 and 5 of a parallel matrix. In 1889 the figure below we depict these three local rows and all columns (0-11). 1890 1891 .vb 1892 0 1 2 3 4 5 6 7 8 9 10 11 1893 -------------------------- 1894 row 3 |. . . d d d o o o o o o 1895 row 4 |. . . d d d o o o o o o 1896 row 5 |. . . d d d o o o o o o 1897 -------------------------- 1898 .ve 1899 1900 Thus, any entries in the d locations are stored in the d (diagonal) 1901 submatrix, and any entries in the o locations are stored in the 1902 o (off-diagonal) submatrix. Note that the d matrix is stored in 1903 MatSeqSBAIJ format and the o submatrix in MATSEQBAIJ format. 1904 1905 Now d_nz should indicate the number of block nonzeros per row in the upper triangular 1906 plus the diagonal part of the d matrix, 1907 and o_nz should indicate the number of block nonzeros per row in the o matrix 1908 1909 In general, for PDE problems in which most nonzeros are near the diagonal, 1910 one expects d_nz >> o_nz. For large problems you MUST preallocate memory 1911 or you will get TERRIBLE performance; see the users' manual chapter on 1912 matrices. 1913 1914 Level: intermediate 1915 1916 .keywords: matrix, block, aij, compressed row, sparse, parallel 1917 1918 .seealso: MatCreate(), MatCreateSeqSBAIJ(), MatSetValues(), MatCreateBAIJ(), PetscSplitOwnership() 1919 @*/ 1920 PetscErrorCode MatMPISBAIJSetPreallocation(Mat B,PetscInt bs,PetscInt d_nz,const PetscInt d_nnz[],PetscInt o_nz,const PetscInt o_nnz[]) 1921 { 1922 PetscErrorCode ierr; 1923 1924 PetscFunctionBegin; 1925 PetscValidHeaderSpecific(B,MAT_CLASSID,1); 1926 PetscValidType(B,1); 1927 PetscValidLogicalCollectiveInt(B,bs,2); 1928 ierr = PetscTryMethod(B,"MatMPISBAIJSetPreallocation_C",(Mat,PetscInt,PetscInt,const PetscInt[],PetscInt,const PetscInt[]),(B,bs,d_nz,d_nnz,o_nz,o_nnz));CHKERRQ(ierr); 1929 PetscFunctionReturn(0); 1930 } 1931 1932 #undef __FUNCT__ 1933 #define __FUNCT__ "MatCreateSBAIJ" 1934 /*@C 1935 MatCreateSBAIJ - Creates a sparse parallel matrix in symmetric block AIJ format 1936 (block compressed row). For good matrix assembly performance 1937 the user should preallocate the matrix storage by setting the parameters 1938 d_nz (or d_nnz) and o_nz (or o_nnz). By setting these parameters accurately, 1939 performance can be increased by more than a factor of 50. 1940 1941 Collective on MPI_Comm 1942 1943 Input Parameters: 1944 + comm - MPI communicator 1945 . bs - size of blockk 1946 . m - number of local rows (or PETSC_DECIDE to have calculated if M is given) 1947 This value should be the same as the local size used in creating the 1948 y vector for the matrix-vector product y = Ax. 1949 . n - number of local columns (or PETSC_DECIDE to have calculated if N is given) 1950 This value should be the same as the local size used in creating the 1951 x vector for the matrix-vector product y = Ax. 1952 . M - number of global rows (or PETSC_DETERMINE to have calculated if m is given) 1953 . N - number of global columns (or PETSC_DETERMINE to have calculated if n is given) 1954 . d_nz - number of block nonzeros per block row in diagonal portion of local 1955 submatrix (same for all local rows) 1956 . d_nnz - array containing the number of block nonzeros in the various block rows 1957 in the upper triangular portion of the in diagonal portion of the local 1958 (possibly different for each block block row) or NULL. 1959 If you plan to factor the matrix you must leave room for the diagonal entry and 1960 set its value even if it is zero. 1961 . o_nz - number of block nonzeros per block row in the off-diagonal portion of local 1962 submatrix (same for all local rows). 1963 - o_nnz - array containing the number of nonzeros in the various block rows of the 1964 off-diagonal portion of the local submatrix (possibly different for 1965 each block row) or NULL. 1966 1967 Output Parameter: 1968 . A - the matrix 1969 1970 Options Database Keys: 1971 . -mat_no_unroll - uses code that does not unroll the loops in the 1972 block calculations (much slower) 1973 . -mat_block_size - size of the blocks to use 1974 . -mat_mpi - use the parallel matrix data structures even on one processor 1975 (defaults to using SeqBAIJ format on one processor) 1976 1977 It is recommended that one use the MatCreate(), MatSetType() and/or MatSetFromOptions(), 1978 MatXXXXSetPreallocation() paradgm instead of this routine directly. 1979 [MatXXXXSetPreallocation() is, for example, MatSeqAIJSetPreallocation] 1980 1981 Notes: 1982 The number of rows and columns must be divisible by blocksize. 1983 This matrix type does not support complex Hermitian operation. 1984 1985 The user MUST specify either the local or global matrix dimensions 1986 (possibly both). 1987 1988 If PETSC_DECIDE or PETSC_DETERMINE is used for a particular argument on one processor 1989 than it must be used on all processors that share the object for that argument. 1990 1991 If the *_nnz parameter is given then the *_nz parameter is ignored 1992 1993 Storage Information: 1994 For a square global matrix we define each processor's diagonal portion 1995 to be its local rows and the corresponding columns (a square submatrix); 1996 each processor's off-diagonal portion encompasses the remainder of the 1997 local matrix (a rectangular submatrix). 1998 1999 The user can specify preallocated storage for the diagonal part of 2000 the local submatrix with either d_nz or d_nnz (not both). Set 2001 d_nz=PETSC_DEFAULT and d_nnz=NULL for PETSc to control dynamic 2002 memory allocation. Likewise, specify preallocated storage for the 2003 off-diagonal part of the local submatrix with o_nz or o_nnz (not both). 2004 2005 Consider a processor that owns rows 3, 4 and 5 of a parallel matrix. In 2006 the figure below we depict these three local rows and all columns (0-11). 2007 2008 .vb 2009 0 1 2 3 4 5 6 7 8 9 10 11 2010 -------------------------- 2011 row 3 |. . . d d d o o o o o o 2012 row 4 |. . . d d d o o o o o o 2013 row 5 |. . . d d d o o o o o o 2014 -------------------------- 2015 .ve 2016 2017 Thus, any entries in the d locations are stored in the d (diagonal) 2018 submatrix, and any entries in the o locations are stored in the 2019 o (off-diagonal) submatrix. Note that the d matrix is stored in 2020 MatSeqSBAIJ format and the o submatrix in MATSEQBAIJ format. 2021 2022 Now d_nz should indicate the number of block nonzeros per row in the upper triangular 2023 plus the diagonal part of the d matrix, 2024 and o_nz should indicate the number of block nonzeros per row in the o matrix. 2025 In general, for PDE problems in which most nonzeros are near the diagonal, 2026 one expects d_nz >> o_nz. For large problems you MUST preallocate memory 2027 or you will get TERRIBLE performance; see the users' manual chapter on 2028 matrices. 2029 2030 Level: intermediate 2031 2032 .keywords: matrix, block, aij, compressed row, sparse, parallel 2033 2034 .seealso: MatCreate(), MatCreateSeqSBAIJ(), MatSetValues(), MatCreateBAIJ() 2035 @*/ 2036 2037 PetscErrorCode MatCreateSBAIJ(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) 2038 { 2039 PetscErrorCode ierr; 2040 PetscMPIInt size; 2041 2042 PetscFunctionBegin; 2043 ierr = MatCreate(comm,A);CHKERRQ(ierr); 2044 ierr = MatSetSizes(*A,m,n,M,N);CHKERRQ(ierr); 2045 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr); 2046 if (size > 1) { 2047 ierr = MatSetType(*A,MATMPISBAIJ);CHKERRQ(ierr); 2048 ierr = MatMPISBAIJSetPreallocation(*A,bs,d_nz,d_nnz,o_nz,o_nnz);CHKERRQ(ierr); 2049 } else { 2050 ierr = MatSetType(*A,MATSEQSBAIJ);CHKERRQ(ierr); 2051 ierr = MatSeqSBAIJSetPreallocation(*A,bs,d_nz,d_nnz);CHKERRQ(ierr); 2052 } 2053 PetscFunctionReturn(0); 2054 } 2055 2056 2057 #undef __FUNCT__ 2058 #define __FUNCT__ "MatDuplicate_MPISBAIJ" 2059 static PetscErrorCode MatDuplicate_MPISBAIJ(Mat matin,MatDuplicateOption cpvalues,Mat *newmat) 2060 { 2061 Mat mat; 2062 Mat_MPISBAIJ *a,*oldmat = (Mat_MPISBAIJ*)matin->data; 2063 PetscErrorCode ierr; 2064 PetscInt len=0,nt,bs=matin->rmap->bs,mbs=oldmat->mbs; 2065 PetscScalar *array; 2066 2067 PetscFunctionBegin; 2068 *newmat = 0; 2069 2070 ierr = MatCreate(PetscObjectComm((PetscObject)matin),&mat);CHKERRQ(ierr); 2071 ierr = MatSetSizes(mat,matin->rmap->n,matin->cmap->n,matin->rmap->N,matin->cmap->N);CHKERRQ(ierr); 2072 ierr = MatSetType(mat,((PetscObject)matin)->type_name);CHKERRQ(ierr); 2073 ierr = PetscMemcpy(mat->ops,matin->ops,sizeof(struct _MatOps));CHKERRQ(ierr); 2074 ierr = PetscLayoutReference(matin->rmap,&mat->rmap);CHKERRQ(ierr); 2075 ierr = PetscLayoutReference(matin->cmap,&mat->cmap);CHKERRQ(ierr); 2076 2077 mat->factortype = matin->factortype; 2078 mat->preallocated = PETSC_TRUE; 2079 mat->assembled = PETSC_TRUE; 2080 mat->insertmode = NOT_SET_VALUES; 2081 2082 a = (Mat_MPISBAIJ*)mat->data; 2083 a->bs2 = oldmat->bs2; 2084 a->mbs = oldmat->mbs; 2085 a->nbs = oldmat->nbs; 2086 a->Mbs = oldmat->Mbs; 2087 a->Nbs = oldmat->Nbs; 2088 2089 2090 a->size = oldmat->size; 2091 a->rank = oldmat->rank; 2092 a->donotstash = oldmat->donotstash; 2093 a->roworiented = oldmat->roworiented; 2094 a->rowindices = 0; 2095 a->rowvalues = 0; 2096 a->getrowactive = PETSC_FALSE; 2097 a->barray = 0; 2098 a->rstartbs = oldmat->rstartbs; 2099 a->rendbs = oldmat->rendbs; 2100 a->cstartbs = oldmat->cstartbs; 2101 a->cendbs = oldmat->cendbs; 2102 2103 /* hash table stuff */ 2104 a->ht = 0; 2105 a->hd = 0; 2106 a->ht_size = 0; 2107 a->ht_flag = oldmat->ht_flag; 2108 a->ht_fact = oldmat->ht_fact; 2109 a->ht_total_ct = 0; 2110 a->ht_insert_ct = 0; 2111 2112 ierr = PetscMemcpy(a->rangebs,oldmat->rangebs,(a->size+2)*sizeof(PetscInt));CHKERRQ(ierr); 2113 if (oldmat->colmap) { 2114 #if defined(PETSC_USE_CTABLE) 2115 ierr = PetscTableCreateCopy(oldmat->colmap,&a->colmap);CHKERRQ(ierr); 2116 #else 2117 ierr = PetscMalloc1((a->Nbs),&a->colmap);CHKERRQ(ierr); 2118 ierr = PetscLogObjectMemory((PetscObject)mat,(a->Nbs)*sizeof(PetscInt));CHKERRQ(ierr); 2119 ierr = PetscMemcpy(a->colmap,oldmat->colmap,(a->Nbs)*sizeof(PetscInt));CHKERRQ(ierr); 2120 #endif 2121 } else a->colmap = 0; 2122 2123 if (oldmat->garray && (len = ((Mat_SeqBAIJ*)(oldmat->B->data))->nbs)) { 2124 ierr = PetscMalloc1(len,&a->garray);CHKERRQ(ierr); 2125 ierr = PetscLogObjectMemory((PetscObject)mat,len*sizeof(PetscInt));CHKERRQ(ierr); 2126 ierr = PetscMemcpy(a->garray,oldmat->garray,len*sizeof(PetscInt));CHKERRQ(ierr); 2127 } else a->garray = 0; 2128 2129 ierr = MatStashCreate_Private(PetscObjectComm((PetscObject)matin),matin->rmap->bs,&mat->bstash);CHKERRQ(ierr); 2130 ierr = VecDuplicate(oldmat->lvec,&a->lvec);CHKERRQ(ierr); 2131 ierr = PetscLogObjectParent((PetscObject)mat,(PetscObject)a->lvec);CHKERRQ(ierr); 2132 ierr = VecScatterCopy(oldmat->Mvctx,&a->Mvctx);CHKERRQ(ierr); 2133 ierr = PetscLogObjectParent((PetscObject)mat,(PetscObject)a->Mvctx);CHKERRQ(ierr); 2134 2135 ierr = VecDuplicate(oldmat->slvec0,&a->slvec0);CHKERRQ(ierr); 2136 ierr = PetscLogObjectParent((PetscObject)mat,(PetscObject)a->slvec0);CHKERRQ(ierr); 2137 ierr = VecDuplicate(oldmat->slvec1,&a->slvec1);CHKERRQ(ierr); 2138 ierr = PetscLogObjectParent((PetscObject)mat,(PetscObject)a->slvec1);CHKERRQ(ierr); 2139 2140 ierr = VecGetLocalSize(a->slvec1,&nt);CHKERRQ(ierr); 2141 ierr = VecGetArray(a->slvec1,&array);CHKERRQ(ierr); 2142 ierr = VecCreateSeqWithArray(PETSC_COMM_SELF,1,bs*mbs,array,&a->slvec1a);CHKERRQ(ierr); 2143 ierr = VecCreateSeqWithArray(PETSC_COMM_SELF,1,nt-bs*mbs,array+bs*mbs,&a->slvec1b);CHKERRQ(ierr); 2144 ierr = VecRestoreArray(a->slvec1,&array);CHKERRQ(ierr); 2145 ierr = VecGetArray(a->slvec0,&array);CHKERRQ(ierr); 2146 ierr = VecCreateSeqWithArray(PETSC_COMM_SELF,1,nt-bs*mbs,array+bs*mbs,&a->slvec0b);CHKERRQ(ierr); 2147 ierr = VecRestoreArray(a->slvec0,&array);CHKERRQ(ierr); 2148 ierr = PetscLogObjectParent((PetscObject)mat,(PetscObject)a->slvec0);CHKERRQ(ierr); 2149 ierr = PetscLogObjectParent((PetscObject)mat,(PetscObject)a->slvec1);CHKERRQ(ierr); 2150 ierr = PetscLogObjectParent((PetscObject)mat,(PetscObject)a->slvec0b);CHKERRQ(ierr); 2151 ierr = PetscLogObjectParent((PetscObject)mat,(PetscObject)a->slvec1a);CHKERRQ(ierr); 2152 ierr = PetscLogObjectParent((PetscObject)mat,(PetscObject)a->slvec1b);CHKERRQ(ierr); 2153 2154 /* ierr = VecScatterCopy(oldmat->sMvctx,&a->sMvctx); - not written yet, replaced by the lazy trick: */ 2155 ierr = PetscObjectReference((PetscObject)oldmat->sMvctx);CHKERRQ(ierr); 2156 a->sMvctx = oldmat->sMvctx; 2157 ierr = PetscLogObjectParent((PetscObject)mat,(PetscObject)a->sMvctx);CHKERRQ(ierr); 2158 2159 ierr = MatDuplicate(oldmat->A,cpvalues,&a->A);CHKERRQ(ierr); 2160 ierr = PetscLogObjectParent((PetscObject)mat,(PetscObject)a->A);CHKERRQ(ierr); 2161 ierr = MatDuplicate(oldmat->B,cpvalues,&a->B);CHKERRQ(ierr); 2162 ierr = PetscLogObjectParent((PetscObject)mat,(PetscObject)a->B);CHKERRQ(ierr); 2163 ierr = PetscFunctionListDuplicate(((PetscObject)matin)->qlist,&((PetscObject)mat)->qlist);CHKERRQ(ierr); 2164 *newmat = mat; 2165 PetscFunctionReturn(0); 2166 } 2167 2168 #undef __FUNCT__ 2169 #define __FUNCT__ "MatLoad_MPISBAIJ" 2170 PetscErrorCode MatLoad_MPISBAIJ(Mat newmat,PetscViewer viewer) 2171 { 2172 PetscErrorCode ierr; 2173 PetscInt i,nz,j,rstart,rend; 2174 PetscScalar *vals,*buf; 2175 MPI_Comm comm; 2176 MPI_Status status; 2177 PetscMPIInt rank,size,tag = ((PetscObject)viewer)->tag,*sndcounts = 0,*browners,maxnz,*rowners,mmbs; 2178 PetscInt header[4],*rowlengths = 0,M,N,m,*cols,*locrowlens; 2179 PetscInt *procsnz = 0,jj,*mycols,*ibuf; 2180 PetscInt bs =1,Mbs,mbs,extra_rows; 2181 PetscInt *dlens,*odlens,*mask,*masked1,*masked2,rowcount,odcount; 2182 PetscInt dcount,kmax,k,nzcount,tmp,sizesset=1,grows,gcols; 2183 int fd; 2184 2185 PetscFunctionBegin; 2186 ierr = PetscObjectGetComm((PetscObject)viewer,&comm);CHKERRQ(ierr); 2187 ierr = PetscOptionsBegin(comm,NULL,"Options for loading MPISBAIJ matrix 2","Mat");CHKERRQ(ierr); 2188 ierr = PetscOptionsInt("-matload_block_size","Set the blocksize used to store the matrix","MatLoad",bs,&bs,NULL);CHKERRQ(ierr); 2189 ierr = PetscOptionsEnd();CHKERRQ(ierr); 2190 2191 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr); 2192 ierr = MPI_Comm_rank(comm,&rank);CHKERRQ(ierr); 2193 if (!rank) { 2194 ierr = PetscViewerBinaryGetDescriptor(viewer,&fd);CHKERRQ(ierr); 2195 ierr = PetscBinaryRead(fd,(char*)header,4,PETSC_INT);CHKERRQ(ierr); 2196 if (header[0] != MAT_FILE_CLASSID) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"not matrix object"); 2197 if (header[3] < 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"Matrix stored in special format, cannot load as MPISBAIJ"); 2198 } 2199 2200 if (newmat->rmap->n < 0 && newmat->rmap->N < 0 && newmat->cmap->n < 0 && newmat->cmap->N < 0) sizesset = 0; 2201 2202 ierr = MPI_Bcast(header+1,3,MPIU_INT,0,comm);CHKERRQ(ierr); 2203 M = header[1]; 2204 N = header[2]; 2205 2206 /* If global rows/cols are set to PETSC_DECIDE, set it to the sizes given in the file */ 2207 if (sizesset && newmat->rmap->N < 0) newmat->rmap->N = M; 2208 if (sizesset && newmat->cmap->N < 0) newmat->cmap->N = N; 2209 2210 /* If global sizes are set, check if they are consistent with that given in the file */ 2211 if (sizesset) { 2212 ierr = MatGetSize(newmat,&grows,&gcols);CHKERRQ(ierr); 2213 } 2214 if (sizesset && newmat->rmap->N != grows) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED, "Inconsistent # of rows:Matrix in file has (%d) and input matrix has (%d)",M,grows); 2215 if (sizesset && newmat->cmap->N != gcols) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED, "Inconsistent # of cols:Matrix in file has (%d) and input matrix has (%d)",N,gcols); 2216 2217 if (M != N) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Can only do square matrices"); 2218 2219 /* 2220 This code adds extra rows to make sure the number of rows is 2221 divisible by the blocksize 2222 */ 2223 Mbs = M/bs; 2224 extra_rows = bs - M + bs*(Mbs); 2225 if (extra_rows == bs) extra_rows = 0; 2226 else Mbs++; 2227 if (extra_rows &&!rank) { 2228 ierr = PetscInfo(viewer,"Padding loaded matrix to match blocksize\n");CHKERRQ(ierr); 2229 } 2230 2231 /* determine ownership of all rows */ 2232 if (newmat->rmap->n < 0) { /* PETSC_DECIDE */ 2233 mbs = Mbs/size + ((Mbs % size) > rank); 2234 m = mbs*bs; 2235 } else { /* User Set */ 2236 m = newmat->rmap->n; 2237 mbs = m/bs; 2238 } 2239 ierr = PetscMalloc2(size+1,&rowners,size+1,&browners);CHKERRQ(ierr); 2240 ierr = PetscMPIIntCast(mbs,&mmbs);CHKERRQ(ierr); 2241 ierr = MPI_Allgather(&mmbs,1,MPI_INT,rowners+1,1,MPI_INT,comm);CHKERRQ(ierr); 2242 rowners[0] = 0; 2243 for (i=2; i<=size; i++) rowners[i] += rowners[i-1]; 2244 for (i=0; i<=size; i++) browners[i] = rowners[i]*bs; 2245 rstart = rowners[rank]; 2246 rend = rowners[rank+1]; 2247 2248 /* distribute row lengths to all processors */ 2249 ierr = PetscMalloc1((rend-rstart)*bs,&locrowlens);CHKERRQ(ierr); 2250 if (!rank) { 2251 ierr = PetscMalloc1((M+extra_rows),&rowlengths);CHKERRQ(ierr); 2252 ierr = PetscBinaryRead(fd,rowlengths,M,PETSC_INT);CHKERRQ(ierr); 2253 for (i=0; i<extra_rows; i++) rowlengths[M+i] = 1; 2254 ierr = PetscMalloc1(size,&sndcounts);CHKERRQ(ierr); 2255 for (i=0; i<size; i++) sndcounts[i] = browners[i+1] - browners[i]; 2256 ierr = MPI_Scatterv(rowlengths,sndcounts,browners,MPIU_INT,locrowlens,(rend-rstart)*bs,MPIU_INT,0,comm);CHKERRQ(ierr); 2257 ierr = PetscFree(sndcounts);CHKERRQ(ierr); 2258 } else { 2259 ierr = MPI_Scatterv(0,0,0,MPIU_INT,locrowlens,(rend-rstart)*bs,MPIU_INT,0,comm);CHKERRQ(ierr); 2260 } 2261 2262 if (!rank) { /* procs[0] */ 2263 /* calculate the number of nonzeros on each processor */ 2264 ierr = PetscMalloc1(size,&procsnz);CHKERRQ(ierr); 2265 ierr = PetscMemzero(procsnz,size*sizeof(PetscInt));CHKERRQ(ierr); 2266 for (i=0; i<size; i++) { 2267 for (j=rowners[i]*bs; j< rowners[i+1]*bs; j++) { 2268 procsnz[i] += rowlengths[j]; 2269 } 2270 } 2271 ierr = PetscFree(rowlengths);CHKERRQ(ierr); 2272 2273 /* determine max buffer needed and allocate it */ 2274 maxnz = 0; 2275 for (i=0; i<size; i++) { 2276 maxnz = PetscMax(maxnz,procsnz[i]); 2277 } 2278 ierr = PetscMalloc1(maxnz,&cols);CHKERRQ(ierr); 2279 2280 /* read in my part of the matrix column indices */ 2281 nz = procsnz[0]; 2282 ierr = PetscMalloc1(nz,&ibuf);CHKERRQ(ierr); 2283 mycols = ibuf; 2284 if (size == 1) nz -= extra_rows; 2285 ierr = PetscBinaryRead(fd,mycols,nz,PETSC_INT);CHKERRQ(ierr); 2286 if (size == 1) { 2287 for (i=0; i< extra_rows; i++) mycols[nz+i] = M+i; 2288 } 2289 2290 /* read in every ones (except the last) and ship off */ 2291 for (i=1; i<size-1; i++) { 2292 nz = procsnz[i]; 2293 ierr = PetscBinaryRead(fd,cols,nz,PETSC_INT);CHKERRQ(ierr); 2294 ierr = MPI_Send(cols,nz,MPIU_INT,i,tag,comm);CHKERRQ(ierr); 2295 } 2296 /* read in the stuff for the last proc */ 2297 if (size != 1) { 2298 nz = procsnz[size-1] - extra_rows; /* the extra rows are not on the disk */ 2299 ierr = PetscBinaryRead(fd,cols,nz,PETSC_INT);CHKERRQ(ierr); 2300 for (i=0; i<extra_rows; i++) cols[nz+i] = M+i; 2301 ierr = MPI_Send(cols,nz+extra_rows,MPIU_INT,size-1,tag,comm);CHKERRQ(ierr); 2302 } 2303 ierr = PetscFree(cols);CHKERRQ(ierr); 2304 } else { /* procs[i], i>0 */ 2305 /* determine buffer space needed for message */ 2306 nz = 0; 2307 for (i=0; i<m; i++) nz += locrowlens[i]; 2308 ierr = PetscMalloc1(nz,&ibuf);CHKERRQ(ierr); 2309 mycols = ibuf; 2310 /* receive message of column indices*/ 2311 ierr = MPI_Recv(mycols,nz,MPIU_INT,0,tag,comm,&status);CHKERRQ(ierr); 2312 ierr = MPI_Get_count(&status,MPIU_INT,&maxnz);CHKERRQ(ierr); 2313 if (maxnz != nz) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"something is wrong with file"); 2314 } 2315 2316 /* loop over local rows, determining number of off diagonal entries */ 2317 ierr = PetscMalloc2(rend-rstart,&dlens,rend-rstart,&odlens);CHKERRQ(ierr); 2318 ierr = PetscMalloc3(Mbs,&mask,Mbs,&masked1,Mbs,&masked2);CHKERRQ(ierr); 2319 ierr = PetscMemzero(mask,Mbs*sizeof(PetscInt));CHKERRQ(ierr); 2320 ierr = PetscMemzero(masked1,Mbs*sizeof(PetscInt));CHKERRQ(ierr); 2321 ierr = PetscMemzero(masked2,Mbs*sizeof(PetscInt));CHKERRQ(ierr); 2322 rowcount = 0; 2323 nzcount = 0; 2324 for (i=0; i<mbs; i++) { 2325 dcount = 0; 2326 odcount = 0; 2327 for (j=0; j<bs; j++) { 2328 kmax = locrowlens[rowcount]; 2329 for (k=0; k<kmax; k++) { 2330 tmp = mycols[nzcount++]/bs; /* block col. index */ 2331 if (!mask[tmp]) { 2332 mask[tmp] = 1; 2333 if (tmp < rstart || tmp >= rend) masked2[odcount++] = tmp; /* entry in off-diag portion */ 2334 else masked1[dcount++] = tmp; /* entry in diag portion */ 2335 } 2336 } 2337 rowcount++; 2338 } 2339 2340 dlens[i] = dcount; /* d_nzz[i] */ 2341 odlens[i] = odcount; /* o_nzz[i] */ 2342 2343 /* zero out the mask elements we set */ 2344 for (j=0; j<dcount; j++) mask[masked1[j]] = 0; 2345 for (j=0; j<odcount; j++) mask[masked2[j]] = 0; 2346 } 2347 if (!sizesset) { 2348 ierr = MatSetSizes(newmat,m,m,M+extra_rows,N+extra_rows);CHKERRQ(ierr); 2349 } 2350 ierr = MatMPISBAIJSetPreallocation(newmat,bs,0,dlens,0,odlens);CHKERRQ(ierr); 2351 ierr = MatSetOption(newmat,MAT_IGNORE_LOWER_TRIANGULAR,PETSC_TRUE);CHKERRQ(ierr); 2352 2353 if (!rank) { 2354 ierr = PetscMalloc1(maxnz,&buf);CHKERRQ(ierr); 2355 /* read in my part of the matrix numerical values */ 2356 nz = procsnz[0]; 2357 vals = buf; 2358 mycols = ibuf; 2359 if (size == 1) nz -= extra_rows; 2360 ierr = PetscBinaryRead(fd,vals,nz,PETSC_SCALAR);CHKERRQ(ierr); 2361 if (size == 1) { 2362 for (i=0; i< extra_rows; i++) vals[nz+i] = 1.0; 2363 } 2364 2365 /* insert into matrix */ 2366 jj = rstart*bs; 2367 for (i=0; i<m; i++) { 2368 ierr = MatSetValues(newmat,1,&jj,locrowlens[i],mycols,vals,INSERT_VALUES);CHKERRQ(ierr); 2369 mycols += locrowlens[i]; 2370 vals += locrowlens[i]; 2371 jj++; 2372 } 2373 2374 /* read in other processors (except the last one) and ship out */ 2375 for (i=1; i<size-1; i++) { 2376 nz = procsnz[i]; 2377 vals = buf; 2378 ierr = PetscBinaryRead(fd,vals,nz,PETSC_SCALAR);CHKERRQ(ierr); 2379 ierr = MPI_Send(vals,nz,MPIU_SCALAR,i,((PetscObject)newmat)->tag,comm);CHKERRQ(ierr); 2380 } 2381 /* the last proc */ 2382 if (size != 1) { 2383 nz = procsnz[i] - extra_rows; 2384 vals = buf; 2385 ierr = PetscBinaryRead(fd,vals,nz,PETSC_SCALAR);CHKERRQ(ierr); 2386 for (i=0; i<extra_rows; i++) vals[nz+i] = 1.0; 2387 ierr = MPI_Send(vals,nz+extra_rows,MPIU_SCALAR,size-1,((PetscObject)newmat)->tag,comm);CHKERRQ(ierr); 2388 } 2389 ierr = PetscFree(procsnz);CHKERRQ(ierr); 2390 2391 } else { 2392 /* receive numeric values */ 2393 ierr = PetscMalloc1(nz,&buf);CHKERRQ(ierr); 2394 2395 /* receive message of values*/ 2396 vals = buf; 2397 mycols = ibuf; 2398 ierr = MPI_Recv(vals,nz,MPIU_SCALAR,0,((PetscObject)newmat)->tag,comm,&status);CHKERRQ(ierr); 2399 ierr = MPI_Get_count(&status,MPIU_SCALAR,&maxnz);CHKERRQ(ierr); 2400 if (maxnz != nz) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"something is wrong with file"); 2401 2402 /* insert into matrix */ 2403 jj = rstart*bs; 2404 for (i=0; i<m; i++) { 2405 ierr = MatSetValues_MPISBAIJ(newmat,1,&jj,locrowlens[i],mycols,vals,INSERT_VALUES);CHKERRQ(ierr); 2406 mycols += locrowlens[i]; 2407 vals += locrowlens[i]; 2408 jj++; 2409 } 2410 } 2411 2412 ierr = PetscFree(locrowlens);CHKERRQ(ierr); 2413 ierr = PetscFree(buf);CHKERRQ(ierr); 2414 ierr = PetscFree(ibuf);CHKERRQ(ierr); 2415 ierr = PetscFree2(rowners,browners);CHKERRQ(ierr); 2416 ierr = PetscFree2(dlens,odlens);CHKERRQ(ierr); 2417 ierr = PetscFree3(mask,masked1,masked2);CHKERRQ(ierr); 2418 ierr = MatAssemblyBegin(newmat,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 2419 ierr = MatAssemblyEnd(newmat,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 2420 PetscFunctionReturn(0); 2421 } 2422 2423 #undef __FUNCT__ 2424 #define __FUNCT__ "MatMPISBAIJSetHashTableFactor" 2425 /*XXXXX@ 2426 MatMPISBAIJSetHashTableFactor - Sets the factor required to compute the size of the HashTable. 2427 2428 Input Parameters: 2429 . mat - the matrix 2430 . fact - factor 2431 2432 Not Collective on Mat, each process can have a different hash factor 2433 2434 Level: advanced 2435 2436 Notes: 2437 This can also be set by the command line option: -mat_use_hash_table fact 2438 2439 .keywords: matrix, hashtable, factor, HT 2440 2441 .seealso: MatSetOption() 2442 @XXXXX*/ 2443 2444 2445 #undef __FUNCT__ 2446 #define __FUNCT__ "MatGetRowMaxAbs_MPISBAIJ" 2447 PetscErrorCode MatGetRowMaxAbs_MPISBAIJ(Mat A,Vec v,PetscInt idx[]) 2448 { 2449 Mat_MPISBAIJ *a = (Mat_MPISBAIJ*)A->data; 2450 Mat_SeqBAIJ *b = (Mat_SeqBAIJ*)(a->B)->data; 2451 PetscReal atmp; 2452 PetscReal *work,*svalues,*rvalues; 2453 PetscErrorCode ierr; 2454 PetscInt i,bs,mbs,*bi,*bj,brow,j,ncols,krow,kcol,col,row,Mbs,bcol; 2455 PetscMPIInt rank,size; 2456 PetscInt *rowners_bs,dest,count,source; 2457 PetscScalar *va; 2458 MatScalar *ba; 2459 MPI_Status stat; 2460 2461 PetscFunctionBegin; 2462 if (idx) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Send email to petsc-maint@mcs.anl.gov"); 2463 ierr = MatGetRowMaxAbs(a->A,v,NULL);CHKERRQ(ierr); 2464 ierr = VecGetArray(v,&va);CHKERRQ(ierr); 2465 2466 ierr = MPI_Comm_size(PetscObjectComm((PetscObject)A),&size);CHKERRQ(ierr); 2467 ierr = MPI_Comm_rank(PetscObjectComm((PetscObject)A),&rank);CHKERRQ(ierr); 2468 2469 bs = A->rmap->bs; 2470 mbs = a->mbs; 2471 Mbs = a->Mbs; 2472 ba = b->a; 2473 bi = b->i; 2474 bj = b->j; 2475 2476 /* find ownerships */ 2477 rowners_bs = A->rmap->range; 2478 2479 /* each proc creates an array to be distributed */ 2480 ierr = PetscMalloc1(bs*Mbs,&work);CHKERRQ(ierr); 2481 ierr = PetscMemzero(work,bs*Mbs*sizeof(PetscReal));CHKERRQ(ierr); 2482 2483 /* row_max for B */ 2484 if (rank != size-1) { 2485 for (i=0; i<mbs; i++) { 2486 ncols = bi[1] - bi[0]; bi++; 2487 brow = bs*i; 2488 for (j=0; j<ncols; j++) { 2489 bcol = bs*(*bj); 2490 for (kcol=0; kcol<bs; kcol++) { 2491 col = bcol + kcol; /* local col index */ 2492 col += rowners_bs[rank+1]; /* global col index */ 2493 for (krow=0; krow<bs; krow++) { 2494 atmp = PetscAbsScalar(*ba); ba++; 2495 row = brow + krow; /* local row index */ 2496 if (PetscRealPart(va[row]) < atmp) va[row] = atmp; 2497 if (work[col] < atmp) work[col] = atmp; 2498 } 2499 } 2500 bj++; 2501 } 2502 } 2503 2504 /* send values to its owners */ 2505 for (dest=rank+1; dest<size; dest++) { 2506 svalues = work + rowners_bs[dest]; 2507 count = rowners_bs[dest+1]-rowners_bs[dest]; 2508 ierr = MPI_Send(svalues,count,MPIU_REAL,dest,rank,PetscObjectComm((PetscObject)A));CHKERRQ(ierr); 2509 } 2510 } 2511 2512 /* receive values */ 2513 if (rank) { 2514 rvalues = work; 2515 count = rowners_bs[rank+1]-rowners_bs[rank]; 2516 for (source=0; source<rank; source++) { 2517 ierr = MPI_Recv(rvalues,count,MPIU_REAL,MPI_ANY_SOURCE,MPI_ANY_TAG,PetscObjectComm((PetscObject)A),&stat);CHKERRQ(ierr); 2518 /* process values */ 2519 for (i=0; i<count; i++) { 2520 if (PetscRealPart(va[i]) < rvalues[i]) va[i] = rvalues[i]; 2521 } 2522 } 2523 } 2524 2525 ierr = VecRestoreArray(v,&va);CHKERRQ(ierr); 2526 ierr = PetscFree(work);CHKERRQ(ierr); 2527 PetscFunctionReturn(0); 2528 } 2529 2530 #undef __FUNCT__ 2531 #define __FUNCT__ "MatSOR_MPISBAIJ" 2532 PetscErrorCode MatSOR_MPISBAIJ(Mat matin,Vec bb,PetscReal omega,MatSORType flag,PetscReal fshift,PetscInt its,PetscInt lits,Vec xx) 2533 { 2534 Mat_MPISBAIJ *mat = (Mat_MPISBAIJ*)matin->data; 2535 PetscErrorCode ierr; 2536 PetscInt mbs=mat->mbs,bs=matin->rmap->bs; 2537 PetscScalar *x,*ptr,*from; 2538 Vec bb1; 2539 const PetscScalar *b; 2540 2541 PetscFunctionBegin; 2542 if (its <= 0 || lits <= 0) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Relaxation requires global its %D and local its %D both positive",its,lits); 2543 if (bs > 1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"SSOR for block size > 1 is not yet implemented"); 2544 2545 if (flag == SOR_APPLY_UPPER) { 2546 ierr = (*mat->A->ops->sor)(mat->A,bb,omega,flag,fshift,lits,1,xx);CHKERRQ(ierr); 2547 PetscFunctionReturn(0); 2548 } 2549 2550 if ((flag & SOR_LOCAL_SYMMETRIC_SWEEP) == SOR_LOCAL_SYMMETRIC_SWEEP) { 2551 if (flag & SOR_ZERO_INITIAL_GUESS) { 2552 ierr = (*mat->A->ops->sor)(mat->A,bb,omega,flag,fshift,lits,lits,xx);CHKERRQ(ierr); 2553 its--; 2554 } 2555 2556 ierr = VecDuplicate(bb,&bb1);CHKERRQ(ierr); 2557 while (its--) { 2558 2559 /* lower triangular part: slvec0b = - B^T*xx */ 2560 ierr = (*mat->B->ops->multtranspose)(mat->B,xx,mat->slvec0b);CHKERRQ(ierr); 2561 2562 /* copy xx into slvec0a */ 2563 ierr = VecGetArray(mat->slvec0,&ptr);CHKERRQ(ierr); 2564 ierr = VecGetArray(xx,&x);CHKERRQ(ierr); 2565 ierr = PetscMemcpy(ptr,x,bs*mbs*sizeof(MatScalar));CHKERRQ(ierr); 2566 ierr = VecRestoreArray(mat->slvec0,&ptr);CHKERRQ(ierr); 2567 2568 ierr = VecScale(mat->slvec0,-1.0);CHKERRQ(ierr); 2569 2570 /* copy bb into slvec1a */ 2571 ierr = VecGetArray(mat->slvec1,&ptr);CHKERRQ(ierr); 2572 ierr = VecGetArrayRead(bb,&b);CHKERRQ(ierr); 2573 ierr = PetscMemcpy(ptr,b,bs*mbs*sizeof(MatScalar));CHKERRQ(ierr); 2574 ierr = VecRestoreArray(mat->slvec1,&ptr);CHKERRQ(ierr); 2575 2576 /* set slvec1b = 0 */ 2577 ierr = VecSet(mat->slvec1b,0.0);CHKERRQ(ierr); 2578 2579 ierr = VecScatterBegin(mat->sMvctx,mat->slvec0,mat->slvec1,ADD_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 2580 ierr = VecRestoreArray(xx,&x);CHKERRQ(ierr); 2581 ierr = VecRestoreArrayRead(bb,&b);CHKERRQ(ierr); 2582 ierr = VecScatterEnd(mat->sMvctx,mat->slvec0,mat->slvec1,ADD_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 2583 2584 /* upper triangular part: bb1 = bb1 - B*x */ 2585 ierr = (*mat->B->ops->multadd)(mat->B,mat->slvec1b,mat->slvec1a,bb1);CHKERRQ(ierr); 2586 2587 /* local diagonal sweep */ 2588 ierr = (*mat->A->ops->sor)(mat->A,bb1,omega,SOR_SYMMETRIC_SWEEP,fshift,lits,lits,xx);CHKERRQ(ierr); 2589 } 2590 ierr = VecDestroy(&bb1);CHKERRQ(ierr); 2591 } else if ((flag & SOR_LOCAL_FORWARD_SWEEP) && (its == 1) && (flag & SOR_ZERO_INITIAL_GUESS)) { 2592 ierr = (*mat->A->ops->sor)(mat->A,bb,omega,flag,fshift,lits,1,xx);CHKERRQ(ierr); 2593 } else if ((flag & SOR_LOCAL_BACKWARD_SWEEP) && (its == 1) && (flag & SOR_ZERO_INITIAL_GUESS)) { 2594 ierr = (*mat->A->ops->sor)(mat->A,bb,omega,flag,fshift,lits,1,xx);CHKERRQ(ierr); 2595 } else if (flag & SOR_EISENSTAT) { 2596 Vec xx1; 2597 PetscBool hasop; 2598 const PetscScalar *diag; 2599 PetscScalar *sl,scale = (omega - 2.0)/omega; 2600 PetscInt i,n; 2601 2602 if (!mat->xx1) { 2603 ierr = VecDuplicate(bb,&mat->xx1);CHKERRQ(ierr); 2604 ierr = VecDuplicate(bb,&mat->bb1);CHKERRQ(ierr); 2605 } 2606 xx1 = mat->xx1; 2607 bb1 = mat->bb1; 2608 2609 ierr = (*mat->A->ops->sor)(mat->A,bb,omega,(MatSORType)(SOR_ZERO_INITIAL_GUESS | SOR_LOCAL_BACKWARD_SWEEP),fshift,lits,1,xx);CHKERRQ(ierr); 2610 2611 if (!mat->diag) { 2612 /* this is wrong for same matrix with new nonzero values */ 2613 ierr = MatGetVecs(matin,&mat->diag,NULL);CHKERRQ(ierr); 2614 ierr = MatGetDiagonal(matin,mat->diag);CHKERRQ(ierr); 2615 } 2616 ierr = MatHasOperation(matin,MATOP_MULT_DIAGONAL_BLOCK,&hasop);CHKERRQ(ierr); 2617 2618 if (hasop) { 2619 ierr = MatMultDiagonalBlock(matin,xx,bb1);CHKERRQ(ierr); 2620 ierr = VecAYPX(mat->slvec1a,scale,bb);CHKERRQ(ierr); 2621 } else { 2622 /* 2623 These two lines are replaced by code that may be a bit faster for a good compiler 2624 ierr = VecPointwiseMult(mat->slvec1a,mat->diag,xx);CHKERRQ(ierr); 2625 ierr = VecAYPX(mat->slvec1a,scale,bb);CHKERRQ(ierr); 2626 */ 2627 ierr = VecGetArray(mat->slvec1a,&sl);CHKERRQ(ierr); 2628 ierr = VecGetArrayRead(mat->diag,&diag);CHKERRQ(ierr); 2629 ierr = VecGetArrayRead(bb,&b);CHKERRQ(ierr); 2630 ierr = VecGetArray(xx,&x);CHKERRQ(ierr); 2631 ierr = VecGetLocalSize(xx,&n);CHKERRQ(ierr); 2632 if (omega == 1.0) { 2633 for (i=0; i<n; i++) sl[i] = b[i] - diag[i]*x[i]; 2634 ierr = PetscLogFlops(2.0*n);CHKERRQ(ierr); 2635 } else { 2636 for (i=0; i<n; i++) sl[i] = b[i] + scale*diag[i]*x[i]; 2637 ierr = PetscLogFlops(3.0*n);CHKERRQ(ierr); 2638 } 2639 ierr = VecRestoreArray(mat->slvec1a,&sl);CHKERRQ(ierr); 2640 ierr = VecRestoreArrayRead(mat->diag,&diag);CHKERRQ(ierr); 2641 ierr = VecRestoreArrayRead(bb,&b);CHKERRQ(ierr); 2642 ierr = VecRestoreArray(xx,&x);CHKERRQ(ierr); 2643 } 2644 2645 /* multiply off-diagonal portion of matrix */ 2646 ierr = VecSet(mat->slvec1b,0.0);CHKERRQ(ierr); 2647 ierr = (*mat->B->ops->multtranspose)(mat->B,xx,mat->slvec0b);CHKERRQ(ierr); 2648 ierr = VecGetArray(mat->slvec0,&from);CHKERRQ(ierr); 2649 ierr = VecGetArray(xx,&x);CHKERRQ(ierr); 2650 ierr = PetscMemcpy(from,x,bs*mbs*sizeof(MatScalar));CHKERRQ(ierr); 2651 ierr = VecRestoreArray(mat->slvec0,&from);CHKERRQ(ierr); 2652 ierr = VecRestoreArray(xx,&x);CHKERRQ(ierr); 2653 ierr = VecScatterBegin(mat->sMvctx,mat->slvec0,mat->slvec1,ADD_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 2654 ierr = VecScatterEnd(mat->sMvctx,mat->slvec0,mat->slvec1,ADD_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 2655 ierr = (*mat->B->ops->multadd)(mat->B,mat->slvec1b,mat->slvec1a,mat->slvec1a);CHKERRQ(ierr); 2656 2657 /* local sweep */ 2658 ierr = (*mat->A->ops->sor)(mat->A,mat->slvec1a,omega,(MatSORType)(SOR_ZERO_INITIAL_GUESS | SOR_LOCAL_FORWARD_SWEEP),fshift,lits,1,xx1);CHKERRQ(ierr); 2659 ierr = VecAXPY(xx,1.0,xx1);CHKERRQ(ierr); 2660 } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"MatSORType is not supported for SBAIJ matrix format"); 2661 PetscFunctionReturn(0); 2662 } 2663 2664 #undef __FUNCT__ 2665 #define __FUNCT__ "MatSOR_MPISBAIJ_2comm" 2666 PetscErrorCode MatSOR_MPISBAIJ_2comm(Mat matin,Vec bb,PetscReal omega,MatSORType flag,PetscReal fshift,PetscInt its,PetscInt lits,Vec xx) 2667 { 2668 Mat_MPISBAIJ *mat = (Mat_MPISBAIJ*)matin->data; 2669 PetscErrorCode ierr; 2670 Vec lvec1,bb1; 2671 2672 PetscFunctionBegin; 2673 if (its <= 0 || lits <= 0) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Relaxation requires global its %D and local its %D both positive",its,lits); 2674 if (matin->rmap->bs > 1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"SSOR for block size > 1 is not yet implemented"); 2675 2676 if ((flag & SOR_LOCAL_SYMMETRIC_SWEEP) == SOR_LOCAL_SYMMETRIC_SWEEP) { 2677 if (flag & SOR_ZERO_INITIAL_GUESS) { 2678 ierr = (*mat->A->ops->sor)(mat->A,bb,omega,flag,fshift,lits,lits,xx);CHKERRQ(ierr); 2679 its--; 2680 } 2681 2682 ierr = VecDuplicate(mat->lvec,&lvec1);CHKERRQ(ierr); 2683 ierr = VecDuplicate(bb,&bb1);CHKERRQ(ierr); 2684 while (its--) { 2685 ierr = VecScatterBegin(mat->Mvctx,xx,mat->lvec,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 2686 2687 /* lower diagonal part: bb1 = bb - B^T*xx */ 2688 ierr = (*mat->B->ops->multtranspose)(mat->B,xx,lvec1);CHKERRQ(ierr); 2689 ierr = VecScale(lvec1,-1.0);CHKERRQ(ierr); 2690 2691 ierr = VecScatterEnd(mat->Mvctx,xx,mat->lvec,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 2692 ierr = VecCopy(bb,bb1);CHKERRQ(ierr); 2693 ierr = VecScatterBegin(mat->Mvctx,lvec1,bb1,ADD_VALUES,SCATTER_REVERSE);CHKERRQ(ierr); 2694 2695 /* upper diagonal part: bb1 = bb1 - B*x */ 2696 ierr = VecScale(mat->lvec,-1.0);CHKERRQ(ierr); 2697 ierr = (*mat->B->ops->multadd)(mat->B,mat->lvec,bb1,bb1);CHKERRQ(ierr); 2698 2699 ierr = VecScatterEnd(mat->Mvctx,lvec1,bb1,ADD_VALUES,SCATTER_REVERSE);CHKERRQ(ierr); 2700 2701 /* diagonal sweep */ 2702 ierr = (*mat->A->ops->sor)(mat->A,bb1,omega,SOR_SYMMETRIC_SWEEP,fshift,lits,lits,xx);CHKERRQ(ierr); 2703 } 2704 ierr = VecDestroy(&lvec1);CHKERRQ(ierr); 2705 ierr = VecDestroy(&bb1);CHKERRQ(ierr); 2706 } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"MatSORType is not supported for SBAIJ matrix format"); 2707 PetscFunctionReturn(0); 2708 } 2709 2710 #undef __FUNCT__ 2711 #define __FUNCT__ "MatCreateMPISBAIJWithArrays" 2712 /*@ 2713 MatCreateMPISBAIJWithArrays - creates a MPI SBAIJ matrix using arrays that contain in standard 2714 CSR format the local rows. 2715 2716 Collective on MPI_Comm 2717 2718 Input Parameters: 2719 + comm - MPI communicator 2720 . bs - the block size, only a block size of 1 is supported 2721 . m - number of local rows (Cannot be PETSC_DECIDE) 2722 . n - This value should be the same as the local size used in creating the 2723 x vector for the matrix-vector product y = Ax. (or PETSC_DECIDE to have 2724 calculated if N is given) For square matrices n is almost always m. 2725 . M - number of global rows (or PETSC_DETERMINE to have calculated if m is given) 2726 . N - number of global columns (or PETSC_DETERMINE to have calculated if n is given) 2727 . i - row indices 2728 . j - column indices 2729 - a - matrix values 2730 2731 Output Parameter: 2732 . mat - the matrix 2733 2734 Level: intermediate 2735 2736 Notes: 2737 The i, j, and a arrays ARE copied by this routine into the internal format used by PETSc; 2738 thus you CANNOT change the matrix entries by changing the values of a[] after you have 2739 called this routine. Use MatCreateMPIAIJWithSplitArrays() to avoid needing to copy the arrays. 2740 2741 The i and j indices are 0 based, and i indices are indices corresponding to the local j array. 2742 2743 .keywords: matrix, aij, compressed row, sparse, parallel 2744 2745 .seealso: MatCreate(), MatCreateSeqAIJ(), MatSetValues(), MatMPIAIJSetPreallocation(), MatMPIAIJSetPreallocationCSR(), 2746 MPIAIJ, MatCreateAIJ(), MatCreateMPIAIJWithSplitArrays() 2747 @*/ 2748 PetscErrorCode MatCreateMPISBAIJWithArrays(MPI_Comm comm,PetscInt bs,PetscInt m,PetscInt n,PetscInt M,PetscInt N,const PetscInt i[],const PetscInt j[],const PetscScalar a[],Mat *mat) 2749 { 2750 PetscErrorCode ierr; 2751 2752 2753 PetscFunctionBegin; 2754 if (i[0]) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"i (row indices) must start with 0"); 2755 if (m < 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"local number of rows (m) cannot be PETSC_DECIDE, or negative"); 2756 ierr = MatCreate(comm,mat);CHKERRQ(ierr); 2757 ierr = MatSetSizes(*mat,m,n,M,N);CHKERRQ(ierr); 2758 ierr = MatSetType(*mat,MATMPISBAIJ);CHKERRQ(ierr); 2759 ierr = MatMPISBAIJSetPreallocationCSR(*mat,bs,i,j,a);CHKERRQ(ierr); 2760 PetscFunctionReturn(0); 2761 } 2762 2763 2764 #undef __FUNCT__ 2765 #define __FUNCT__ "MatMPISBAIJSetPreallocationCSR" 2766 /*@C 2767 MatMPISBAIJSetPreallocationCSR - Allocates memory for a sparse parallel matrix in BAIJ format 2768 (the default parallel PETSc format). 2769 2770 Collective on MPI_Comm 2771 2772 Input Parameters: 2773 + A - the matrix 2774 . bs - the block size 2775 . i - the indices into j for the start of each local row (starts with zero) 2776 . j - the column indices for each local row (starts with zero) these must be sorted for each row 2777 - v - optional values in the matrix 2778 2779 Level: developer 2780 2781 .keywords: matrix, aij, compressed row, sparse, parallel 2782 2783 .seealso: MatCreate(), MatCreateSeqAIJ(), MatSetValues(), MatMPIBAIJSetPreallocation(), MatCreateAIJ(), MPIAIJ 2784 @*/ 2785 PetscErrorCode MatMPISBAIJSetPreallocationCSR(Mat B,PetscInt bs,const PetscInt i[],const PetscInt j[], const PetscScalar v[]) 2786 { 2787 PetscErrorCode ierr; 2788 2789 PetscFunctionBegin; 2790 ierr = PetscTryMethod(B,"MatMPISBAIJSetPreallocationCSR_C",(Mat,PetscInt,const PetscInt[],const PetscInt[],const PetscScalar[]),(B,bs,i,j,v));CHKERRQ(ierr); 2791 PetscFunctionReturn(0); 2792 } 2793 2794 2795