1 #define PETSCMAT_DLL 2 3 /* 4 Basic functions for basic parallel dense matrices. 5 */ 6 7 8 #include "src/mat/impls/dense/mpi/mpidense.h" /*I "petscmat.h" I*/ 9 #if defined(PETSC_HAVE_PLAPACK) 10 static PetscMPIInt Plapack_nprows,Plapack_npcols,Plapack_ierror,Plapack_nb_alg; 11 static MPI_Comm Plapack_comm_2d; 12 #endif 13 14 #undef __FUNCT__ 15 #define __FUNCT__ "MatDenseGetLocalMatrix" 16 /*@ 17 18 MatDenseGetLocalMatrix - For a MATMPIDENSE or MATSEQDENSE matrix returns the sequential 19 matrix that represents the operator. For sequential matrices it returns itself. 20 21 Input Parameter: 22 . A - the Seq or MPI dense matrix 23 24 Output Parameter: 25 . B - the inner matrix 26 27 Level: intermediate 28 29 @*/ 30 PetscErrorCode MatDenseGetLocalMatrix(Mat A,Mat *B) 31 { 32 Mat_MPIDense *mat = (Mat_MPIDense*)A->data; 33 PetscErrorCode ierr; 34 PetscTruth flg; 35 36 PetscFunctionBegin; 37 ierr = PetscTypeCompare((PetscObject)A,MATMPIDENSE,&flg);CHKERRQ(ierr); 38 if (flg) { 39 *B = mat->A; 40 } else { 41 *B = A; 42 } 43 PetscFunctionReturn(0); 44 } 45 46 #undef __FUNCT__ 47 #define __FUNCT__ "MatGetRow_MPIDense" 48 PetscErrorCode MatGetRow_MPIDense(Mat A,PetscInt row,PetscInt *nz,PetscInt **idx,PetscScalar **v) 49 { 50 Mat_MPIDense *mat = (Mat_MPIDense*)A->data; 51 PetscErrorCode ierr; 52 PetscInt lrow,rstart = A->rmap->rstart,rend = A->rmap->rend; 53 54 PetscFunctionBegin; 55 if (row < rstart || row >= rend) SETERRQ(PETSC_ERR_SUP,"only local rows") 56 lrow = row - rstart; 57 ierr = MatGetRow(mat->A,lrow,nz,(const PetscInt **)idx,(const PetscScalar **)v);CHKERRQ(ierr); 58 PetscFunctionReturn(0); 59 } 60 61 #undef __FUNCT__ 62 #define __FUNCT__ "MatRestoreRow_MPIDense" 63 PetscErrorCode MatRestoreRow_MPIDense(Mat mat,PetscInt row,PetscInt *nz,PetscInt **idx,PetscScalar **v) 64 { 65 PetscErrorCode ierr; 66 67 PetscFunctionBegin; 68 if (idx) {ierr = PetscFree(*idx);CHKERRQ(ierr);} 69 if (v) {ierr = PetscFree(*v);CHKERRQ(ierr);} 70 PetscFunctionReturn(0); 71 } 72 73 EXTERN_C_BEGIN 74 #undef __FUNCT__ 75 #define __FUNCT__ "MatGetDiagonalBlock_MPIDense" 76 PetscErrorCode PETSCMAT_DLLEXPORT MatGetDiagonalBlock_MPIDense(Mat A,PetscTruth *iscopy,MatReuse reuse,Mat *B) 77 { 78 Mat_MPIDense *mdn = (Mat_MPIDense*)A->data; 79 PetscErrorCode ierr; 80 PetscInt m = A->rmap->n,rstart = A->rmap->rstart; 81 PetscScalar *array; 82 MPI_Comm comm; 83 84 PetscFunctionBegin; 85 if (A->rmap->N != A->cmap->N) SETERRQ(PETSC_ERR_SUP,"Only square matrices supported."); 86 87 /* The reuse aspect is not implemented efficiently */ 88 if (reuse) { ierr = MatDestroy(*B);CHKERRQ(ierr);} 89 90 ierr = PetscObjectGetComm((PetscObject)(mdn->A),&comm);CHKERRQ(ierr); 91 ierr = MatGetArray(mdn->A,&array);CHKERRQ(ierr); 92 ierr = MatCreate(comm,B);CHKERRQ(ierr); 93 ierr = MatSetSizes(*B,m,m,m,m);CHKERRQ(ierr); 94 ierr = MatSetType(*B,((PetscObject)mdn->A)->type_name);CHKERRQ(ierr); 95 ierr = MatSeqDenseSetPreallocation(*B,array+m*rstart);CHKERRQ(ierr); 96 ierr = MatRestoreArray(mdn->A,&array);CHKERRQ(ierr); 97 ierr = MatAssemblyBegin(*B,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 98 ierr = MatAssemblyEnd(*B,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 99 100 *iscopy = PETSC_TRUE; 101 PetscFunctionReturn(0); 102 } 103 EXTERN_C_END 104 105 #undef __FUNCT__ 106 #define __FUNCT__ "MatSetValues_MPIDense" 107 PetscErrorCode MatSetValues_MPIDense(Mat mat,PetscInt m,const PetscInt idxm[],PetscInt n,const PetscInt idxn[],const PetscScalar v[],InsertMode addv) 108 { 109 Mat_MPIDense *A = (Mat_MPIDense*)mat->data; 110 PetscErrorCode ierr; 111 PetscInt i,j,rstart = mat->rmap->rstart,rend = mat->rmap->rend,row; 112 PetscTruth roworiented = A->roworiented; 113 114 PetscFunctionBegin; 115 for (i=0; i<m; i++) { 116 if (idxm[i] < 0) continue; 117 if (idxm[i] >= mat->rmap->N) SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,"Row too large"); 118 if (idxm[i] >= rstart && idxm[i] < rend) { 119 row = idxm[i] - rstart; 120 if (roworiented) { 121 ierr = MatSetValues(A->A,1,&row,n,idxn,v+i*n,addv);CHKERRQ(ierr); 122 } else { 123 for (j=0; j<n; j++) { 124 if (idxn[j] < 0) continue; 125 if (idxn[j] >= mat->cmap->N) SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,"Column too large"); 126 ierr = MatSetValues(A->A,1,&row,1,&idxn[j],v+i+j*m,addv);CHKERRQ(ierr); 127 } 128 } 129 } else { 130 if (!A->donotstash) { 131 if (roworiented) { 132 ierr = MatStashValuesRow_Private(&mat->stash,idxm[i],n,idxn,v+i*n);CHKERRQ(ierr); 133 } else { 134 ierr = MatStashValuesCol_Private(&mat->stash,idxm[i],n,idxn,v+i,m);CHKERRQ(ierr); 135 } 136 } 137 } 138 } 139 PetscFunctionReturn(0); 140 } 141 142 #undef __FUNCT__ 143 #define __FUNCT__ "MatGetValues_MPIDense" 144 PetscErrorCode MatGetValues_MPIDense(Mat mat,PetscInt m,const PetscInt idxm[],PetscInt n,const PetscInt idxn[],PetscScalar v[]) 145 { 146 Mat_MPIDense *mdn = (Mat_MPIDense*)mat->data; 147 PetscErrorCode ierr; 148 PetscInt i,j,rstart = mat->rmap->rstart,rend = mat->rmap->rend,row; 149 150 PetscFunctionBegin; 151 for (i=0; i<m; i++) { 152 if (idxm[i] < 0) continue; /* SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,"Negative row"); */ 153 if (idxm[i] >= mat->rmap->N) SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,"Row too large"); 154 if (idxm[i] >= rstart && idxm[i] < rend) { 155 row = idxm[i] - rstart; 156 for (j=0; j<n; j++) { 157 if (idxn[j] < 0) continue; /* SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,"Negative column"); */ 158 if (idxn[j] >= mat->cmap->N) { 159 SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,"Column too large"); 160 } 161 ierr = MatGetValues(mdn->A,1,&row,1,&idxn[j],v+i*n+j);CHKERRQ(ierr); 162 } 163 } else { 164 SETERRQ(PETSC_ERR_SUP,"Only local values currently supported"); 165 } 166 } 167 PetscFunctionReturn(0); 168 } 169 170 #undef __FUNCT__ 171 #define __FUNCT__ "MatGetArray_MPIDense" 172 PetscErrorCode MatGetArray_MPIDense(Mat A,PetscScalar *array[]) 173 { 174 Mat_MPIDense *a = (Mat_MPIDense*)A->data; 175 PetscErrorCode ierr; 176 177 PetscFunctionBegin; 178 ierr = MatGetArray(a->A,array);CHKERRQ(ierr); 179 PetscFunctionReturn(0); 180 } 181 182 #undef __FUNCT__ 183 #define __FUNCT__ "MatGetSubMatrix_MPIDense" 184 static PetscErrorCode MatGetSubMatrix_MPIDense(Mat A,IS isrow,IS iscol,PetscInt cs,MatReuse scall,Mat *B) 185 { 186 Mat_MPIDense *mat = (Mat_MPIDense*)A->data,*newmatd; 187 Mat_SeqDense *lmat = (Mat_SeqDense*)mat->A->data; 188 PetscErrorCode ierr; 189 PetscInt i,j,*irow,*icol,rstart,rend,nrows,ncols,nlrows,nlcols; 190 PetscScalar *av,*bv,*v = lmat->v; 191 Mat newmat; 192 193 PetscFunctionBegin; 194 ierr = ISGetIndices(isrow,&irow);CHKERRQ(ierr); 195 ierr = ISGetIndices(iscol,&icol);CHKERRQ(ierr); 196 ierr = ISGetLocalSize(isrow,&nrows);CHKERRQ(ierr); 197 ierr = ISGetLocalSize(iscol,&ncols);CHKERRQ(ierr); 198 199 /* No parallel redistribution currently supported! Should really check each index set 200 to comfirm that it is OK. ... Currently supports only submatrix same partitioning as 201 original matrix! */ 202 203 ierr = MatGetLocalSize(A,&nlrows,&nlcols);CHKERRQ(ierr); 204 ierr = MatGetOwnershipRange(A,&rstart,&rend);CHKERRQ(ierr); 205 206 /* Check submatrix call */ 207 if (scall == MAT_REUSE_MATRIX) { 208 /* SETERRQ(PETSC_ERR_ARG_SIZ,"Reused submatrix wrong size"); */ 209 /* Really need to test rows and column sizes! */ 210 newmat = *B; 211 } else { 212 /* Create and fill new matrix */ 213 ierr = MatCreate(((PetscObject)A)->comm,&newmat);CHKERRQ(ierr); 214 ierr = MatSetSizes(newmat,nrows,cs,PETSC_DECIDE,ncols);CHKERRQ(ierr); 215 ierr = MatSetType(newmat,((PetscObject)A)->type_name);CHKERRQ(ierr); 216 ierr = MatMPIDenseSetPreallocation(newmat,PETSC_NULL);CHKERRQ(ierr); 217 } 218 219 /* Now extract the data pointers and do the copy, column at a time */ 220 newmatd = (Mat_MPIDense*)newmat->data; 221 bv = ((Mat_SeqDense *)newmatd->A->data)->v; 222 223 for (i=0; i<ncols; i++) { 224 av = v + nlrows*icol[i]; 225 for (j=0; j<nrows; j++) { 226 *bv++ = av[irow[j] - rstart]; 227 } 228 } 229 230 /* Assemble the matrices so that the correct flags are set */ 231 ierr = MatAssemblyBegin(newmat,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 232 ierr = MatAssemblyEnd(newmat,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 233 234 /* Free work space */ 235 ierr = ISRestoreIndices(isrow,&irow);CHKERRQ(ierr); 236 ierr = ISRestoreIndices(iscol,&icol);CHKERRQ(ierr); 237 *B = newmat; 238 PetscFunctionReturn(0); 239 } 240 241 #undef __FUNCT__ 242 #define __FUNCT__ "MatRestoreArray_MPIDense" 243 PetscErrorCode MatRestoreArray_MPIDense(Mat A,PetscScalar *array[]) 244 { 245 PetscFunctionBegin; 246 PetscFunctionReturn(0); 247 } 248 249 #undef __FUNCT__ 250 #define __FUNCT__ "MatAssemblyBegin_MPIDense" 251 PetscErrorCode MatAssemblyBegin_MPIDense(Mat mat,MatAssemblyType mode) 252 { 253 Mat_MPIDense *mdn = (Mat_MPIDense*)mat->data; 254 MPI_Comm comm = ((PetscObject)mat)->comm; 255 PetscErrorCode ierr; 256 PetscInt nstash,reallocs; 257 InsertMode addv; 258 259 PetscFunctionBegin; 260 /* make sure all processors are either in INSERTMODE or ADDMODE */ 261 ierr = MPI_Allreduce(&mat->insertmode,&addv,1,MPI_INT,MPI_BOR,comm);CHKERRQ(ierr); 262 if (addv == (ADD_VALUES|INSERT_VALUES)) { 263 SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Cannot mix adds/inserts on different procs"); 264 } 265 mat->insertmode = addv; /* in case this processor had no cache */ 266 267 ierr = MatStashScatterBegin_Private(mat,&mat->stash,mat->rmap->range);CHKERRQ(ierr); 268 ierr = MatStashGetInfo_Private(&mat->stash,&nstash,&reallocs);CHKERRQ(ierr); 269 ierr = PetscInfo2(mdn->A,"Stash has %D entries, uses %D mallocs.\n",nstash,reallocs);CHKERRQ(ierr); 270 PetscFunctionReturn(0); 271 } 272 273 #undef __FUNCT__ 274 #define __FUNCT__ "MatAssemblyEnd_MPIDense" 275 PetscErrorCode MatAssemblyEnd_MPIDense(Mat mat,MatAssemblyType mode) 276 { 277 Mat_MPIDense *mdn=(Mat_MPIDense*)mat->data; 278 PetscErrorCode ierr; 279 PetscInt i,*row,*col,flg,j,rstart,ncols; 280 PetscMPIInt n; 281 PetscScalar *val; 282 InsertMode addv=mat->insertmode; 283 284 PetscFunctionBegin; 285 /* wait on receives */ 286 while (1) { 287 ierr = MatStashScatterGetMesg_Private(&mat->stash,&n,&row,&col,&val,&flg);CHKERRQ(ierr); 288 if (!flg) break; 289 290 for (i=0; i<n;) { 291 /* Now identify the consecutive vals belonging to the same row */ 292 for (j=i,rstart=row[j]; j<n; j++) { if (row[j] != rstart) break; } 293 if (j < n) ncols = j-i; 294 else ncols = n-i; 295 /* Now assemble all these values with a single function call */ 296 ierr = MatSetValues_MPIDense(mat,1,row+i,ncols,col+i,val+i,addv);CHKERRQ(ierr); 297 i = j; 298 } 299 } 300 ierr = MatStashScatterEnd_Private(&mat->stash);CHKERRQ(ierr); 301 302 ierr = MatAssemblyBegin(mdn->A,mode);CHKERRQ(ierr); 303 ierr = MatAssemblyEnd(mdn->A,mode);CHKERRQ(ierr); 304 305 if (!mat->was_assembled && mode == MAT_FINAL_ASSEMBLY) { 306 ierr = MatSetUpMultiply_MPIDense(mat);CHKERRQ(ierr); 307 } 308 PetscFunctionReturn(0); 309 } 310 311 #undef __FUNCT__ 312 #define __FUNCT__ "MatZeroEntries_MPIDense" 313 PetscErrorCode MatZeroEntries_MPIDense(Mat A) 314 { 315 PetscErrorCode ierr; 316 Mat_MPIDense *l = (Mat_MPIDense*)A->data; 317 318 PetscFunctionBegin; 319 ierr = MatZeroEntries(l->A);CHKERRQ(ierr); 320 PetscFunctionReturn(0); 321 } 322 323 /* the code does not do the diagonal entries correctly unless the 324 matrix is square and the column and row owerships are identical. 325 This is a BUG. The only way to fix it seems to be to access 326 mdn->A and mdn->B directly and not through the MatZeroRows() 327 routine. 328 */ 329 #undef __FUNCT__ 330 #define __FUNCT__ "MatZeroRows_MPIDense" 331 PetscErrorCode MatZeroRows_MPIDense(Mat A,PetscInt N,const PetscInt rows[],PetscScalar diag) 332 { 333 Mat_MPIDense *l = (Mat_MPIDense*)A->data; 334 PetscErrorCode ierr; 335 PetscInt i,*owners = A->rmap->range; 336 PetscInt *nprocs,j,idx,nsends; 337 PetscInt nmax,*svalues,*starts,*owner,nrecvs; 338 PetscInt *rvalues,tag = ((PetscObject)A)->tag,count,base,slen,*source; 339 PetscInt *lens,*lrows,*values; 340 PetscMPIInt n,imdex,rank = l->rank,size = l->size; 341 MPI_Comm comm = ((PetscObject)A)->comm; 342 MPI_Request *send_waits,*recv_waits; 343 MPI_Status recv_status,*send_status; 344 PetscTruth found; 345 346 PetscFunctionBegin; 347 /* first count number of contributors to each processor */ 348 ierr = PetscMalloc(2*size*sizeof(PetscInt),&nprocs);CHKERRQ(ierr); 349 ierr = PetscMemzero(nprocs,2*size*sizeof(PetscInt));CHKERRQ(ierr); 350 ierr = PetscMalloc((N+1)*sizeof(PetscInt),&owner);CHKERRQ(ierr); /* see note*/ 351 for (i=0; i<N; i++) { 352 idx = rows[i]; 353 found = PETSC_FALSE; 354 for (j=0; j<size; j++) { 355 if (idx >= owners[j] && idx < owners[j+1]) { 356 nprocs[2*j]++; nprocs[2*j+1] = 1; owner[i] = j; found = PETSC_TRUE; break; 357 } 358 } 359 if (!found) SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,"Index out of range"); 360 } 361 nsends = 0; for (i=0; i<size; i++) { nsends += nprocs[2*i+1];} 362 363 /* inform other processors of number of messages and max length*/ 364 ierr = PetscMaxSum(comm,nprocs,&nmax,&nrecvs);CHKERRQ(ierr); 365 366 /* post receives: */ 367 ierr = PetscMalloc((nrecvs+1)*(nmax+1)*sizeof(PetscInt),&rvalues);CHKERRQ(ierr); 368 ierr = PetscMalloc((nrecvs+1)*sizeof(MPI_Request),&recv_waits);CHKERRQ(ierr); 369 for (i=0; i<nrecvs; i++) { 370 ierr = MPI_Irecv(rvalues+nmax*i,nmax,MPIU_INT,MPI_ANY_SOURCE,tag,comm,recv_waits+i);CHKERRQ(ierr); 371 } 372 373 /* do sends: 374 1) starts[i] gives the starting index in svalues for stuff going to 375 the ith processor 376 */ 377 ierr = PetscMalloc((N+1)*sizeof(PetscInt),&svalues);CHKERRQ(ierr); 378 ierr = PetscMalloc((nsends+1)*sizeof(MPI_Request),&send_waits);CHKERRQ(ierr); 379 ierr = PetscMalloc((size+1)*sizeof(PetscInt),&starts);CHKERRQ(ierr); 380 starts[0] = 0; 381 for (i=1; i<size; i++) { starts[i] = starts[i-1] + nprocs[2*i-2];} 382 for (i=0; i<N; i++) { 383 svalues[starts[owner[i]]++] = rows[i]; 384 } 385 386 starts[0] = 0; 387 for (i=1; i<size+1; i++) { starts[i] = starts[i-1] + nprocs[2*i-2];} 388 count = 0; 389 for (i=0; i<size; i++) { 390 if (nprocs[2*i+1]) { 391 ierr = MPI_Isend(svalues+starts[i],nprocs[2*i],MPIU_INT,i,tag,comm,send_waits+count++);CHKERRQ(ierr); 392 } 393 } 394 ierr = PetscFree(starts);CHKERRQ(ierr); 395 396 base = owners[rank]; 397 398 /* wait on receives */ 399 ierr = PetscMalloc(2*(nrecvs+1)*sizeof(PetscInt),&lens);CHKERRQ(ierr); 400 source = lens + nrecvs; 401 count = nrecvs; slen = 0; 402 while (count) { 403 ierr = MPI_Waitany(nrecvs,recv_waits,&imdex,&recv_status);CHKERRQ(ierr); 404 /* unpack receives into our local space */ 405 ierr = MPI_Get_count(&recv_status,MPIU_INT,&n);CHKERRQ(ierr); 406 source[imdex] = recv_status.MPI_SOURCE; 407 lens[imdex] = n; 408 slen += n; 409 count--; 410 } 411 ierr = PetscFree(recv_waits);CHKERRQ(ierr); 412 413 /* move the data into the send scatter */ 414 ierr = PetscMalloc((slen+1)*sizeof(PetscInt),&lrows);CHKERRQ(ierr); 415 count = 0; 416 for (i=0; i<nrecvs; i++) { 417 values = rvalues + i*nmax; 418 for (j=0; j<lens[i]; j++) { 419 lrows[count++] = values[j] - base; 420 } 421 } 422 ierr = PetscFree(rvalues);CHKERRQ(ierr); 423 ierr = PetscFree(lens);CHKERRQ(ierr); 424 ierr = PetscFree(owner);CHKERRQ(ierr); 425 ierr = PetscFree(nprocs);CHKERRQ(ierr); 426 427 /* actually zap the local rows */ 428 ierr = MatZeroRows(l->A,slen,lrows,diag);CHKERRQ(ierr); 429 ierr = PetscFree(lrows);CHKERRQ(ierr); 430 431 /* wait on sends */ 432 if (nsends) { 433 ierr = PetscMalloc(nsends*sizeof(MPI_Status),&send_status);CHKERRQ(ierr); 434 ierr = MPI_Waitall(nsends,send_waits,send_status);CHKERRQ(ierr); 435 ierr = PetscFree(send_status);CHKERRQ(ierr); 436 } 437 ierr = PetscFree(send_waits);CHKERRQ(ierr); 438 ierr = PetscFree(svalues);CHKERRQ(ierr); 439 440 PetscFunctionReturn(0); 441 } 442 443 #undef __FUNCT__ 444 #define __FUNCT__ "MatMult_MPIDense" 445 PetscErrorCode MatMult_MPIDense(Mat mat,Vec xx,Vec yy) 446 { 447 Mat_MPIDense *mdn = (Mat_MPIDense*)mat->data; 448 PetscErrorCode ierr; 449 450 PetscFunctionBegin; 451 ierr = VecScatterBegin(mdn->Mvctx,xx,mdn->lvec,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 452 ierr = VecScatterEnd(mdn->Mvctx,xx,mdn->lvec,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 453 ierr = MatMult_SeqDense(mdn->A,mdn->lvec,yy);CHKERRQ(ierr); 454 PetscFunctionReturn(0); 455 } 456 457 #undef __FUNCT__ 458 #define __FUNCT__ "MatMultAdd_MPIDense" 459 PetscErrorCode MatMultAdd_MPIDense(Mat mat,Vec xx,Vec yy,Vec zz) 460 { 461 Mat_MPIDense *mdn = (Mat_MPIDense*)mat->data; 462 PetscErrorCode ierr; 463 464 PetscFunctionBegin; 465 ierr = VecScatterBegin(mdn->Mvctx,xx,mdn->lvec,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 466 ierr = VecScatterEnd(mdn->Mvctx,xx,mdn->lvec,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 467 ierr = MatMultAdd_SeqDense(mdn->A,mdn->lvec,yy,zz);CHKERRQ(ierr); 468 PetscFunctionReturn(0); 469 } 470 471 #undef __FUNCT__ 472 #define __FUNCT__ "MatMultTranspose_MPIDense" 473 PetscErrorCode MatMultTranspose_MPIDense(Mat A,Vec xx,Vec yy) 474 { 475 Mat_MPIDense *a = (Mat_MPIDense*)A->data; 476 PetscErrorCode ierr; 477 PetscScalar zero = 0.0; 478 479 PetscFunctionBegin; 480 ierr = VecSet(yy,zero);CHKERRQ(ierr); 481 ierr = MatMultTranspose_SeqDense(a->A,xx,a->lvec);CHKERRQ(ierr); 482 ierr = VecScatterBegin(a->Mvctx,a->lvec,yy,ADD_VALUES,SCATTER_REVERSE);CHKERRQ(ierr); 483 ierr = VecScatterEnd(a->Mvctx,a->lvec,yy,ADD_VALUES,SCATTER_REVERSE);CHKERRQ(ierr); 484 PetscFunctionReturn(0); 485 } 486 487 #undef __FUNCT__ 488 #define __FUNCT__ "MatMultTransposeAdd_MPIDense" 489 PetscErrorCode MatMultTransposeAdd_MPIDense(Mat A,Vec xx,Vec yy,Vec zz) 490 { 491 Mat_MPIDense *a = (Mat_MPIDense*)A->data; 492 PetscErrorCode ierr; 493 494 PetscFunctionBegin; 495 ierr = VecCopy(yy,zz);CHKERRQ(ierr); 496 ierr = MatMultTranspose_SeqDense(a->A,xx,a->lvec);CHKERRQ(ierr); 497 ierr = VecScatterBegin(a->Mvctx,a->lvec,zz,ADD_VALUES,SCATTER_REVERSE);CHKERRQ(ierr); 498 ierr = VecScatterEnd(a->Mvctx,a->lvec,zz,ADD_VALUES,SCATTER_REVERSE);CHKERRQ(ierr); 499 PetscFunctionReturn(0); 500 } 501 502 #undef __FUNCT__ 503 #define __FUNCT__ "MatGetDiagonal_MPIDense" 504 PetscErrorCode MatGetDiagonal_MPIDense(Mat A,Vec v) 505 { 506 Mat_MPIDense *a = (Mat_MPIDense*)A->data; 507 Mat_SeqDense *aloc = (Mat_SeqDense*)a->A->data; 508 PetscErrorCode ierr; 509 PetscInt len,i,n,m = A->rmap->n,radd; 510 PetscScalar *x,zero = 0.0; 511 512 PetscFunctionBegin; 513 ierr = VecSet(v,zero);CHKERRQ(ierr); 514 ierr = VecGetArray(v,&x);CHKERRQ(ierr); 515 ierr = VecGetSize(v,&n);CHKERRQ(ierr); 516 if (n != A->rmap->N) SETERRQ(PETSC_ERR_ARG_SIZ,"Nonconforming mat and vec"); 517 len = PetscMin(a->A->rmap->n,a->A->cmap->n); 518 radd = A->rmap->rstart*m; 519 for (i=0; i<len; i++) { 520 x[i] = aloc->v[radd + i*m + i]; 521 } 522 ierr = VecRestoreArray(v,&x);CHKERRQ(ierr); 523 PetscFunctionReturn(0); 524 } 525 526 #undef __FUNCT__ 527 #define __FUNCT__ "MatDestroy_MPIDense" 528 PetscErrorCode MatDestroy_MPIDense(Mat mat) 529 { 530 Mat_MPIDense *mdn = (Mat_MPIDense*)mat->data; 531 PetscErrorCode ierr; 532 #if defined(PETSC_HAVE_PLAPACK) 533 Mat_Plapack *lu=(Mat_Plapack*)(mat->spptr); 534 #endif 535 536 PetscFunctionBegin; 537 538 #if defined(PETSC_USE_LOG) 539 PetscLogObjectState((PetscObject)mat,"Rows=%D, Cols=%D",mat->rmap->N,mat->cmap->N); 540 #endif 541 ierr = MatStashDestroy_Private(&mat->stash);CHKERRQ(ierr); 542 ierr = MatDestroy(mdn->A);CHKERRQ(ierr); 543 if (mdn->lvec) {ierr = VecDestroy(mdn->lvec);CHKERRQ(ierr);} 544 if (mdn->Mvctx) {ierr = VecScatterDestroy(mdn->Mvctx);CHKERRQ(ierr);} 545 #if defined(PETSC_HAVE_PLAPACK) 546 if (lu) { 547 ierr = PLA_Obj_free(&lu->A);CHKERRQ(ierr); 548 ierr = PLA_Obj_free (&lu->pivots);CHKERRQ(ierr); 549 ierr = PLA_Temp_free(&lu->templ);CHKERRQ(ierr); 550 551 if (lu->is_pla) { 552 ierr = ISDestroy(lu->is_pla);CHKERRQ(ierr); 553 ierr = ISDestroy(lu->is_petsc);CHKERRQ(ierr); 554 ierr = VecScatterDestroy(lu->ctx);CHKERRQ(ierr); 555 } 556 } 557 #endif 558 559 ierr = PetscFree(mdn);CHKERRQ(ierr); 560 ierr = PetscObjectChangeTypeName((PetscObject)mat,0);CHKERRQ(ierr); 561 ierr = PetscObjectComposeFunctionDynamic((PetscObject)mat,"MatGetDiagonalBlock_C","",PETSC_NULL);CHKERRQ(ierr); 562 ierr = PetscObjectComposeFunctionDynamic((PetscObject)mat,"MatMPIDenseSetPreallocation_C","",PETSC_NULL);CHKERRQ(ierr); 563 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatMatMult_mpiaij_mpidense_C","",PETSC_NULL);CHKERRQ(ierr); 564 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatMatMultSymbolic_mpiaij_mpidense_C","",PETSC_NULL);CHKERRQ(ierr); 565 ierr = PetscObjectComposeFunction((PetscObject)mat,"MatMatMultNumeric_mpiaij_mpidense_C","",PETSC_NULL);CHKERRQ(ierr); 566 PetscFunctionReturn(0); 567 } 568 569 #undef __FUNCT__ 570 #define __FUNCT__ "MatView_MPIDense_Binary" 571 static PetscErrorCode MatView_MPIDense_Binary(Mat mat,PetscViewer viewer) 572 { 573 Mat_MPIDense *mdn = (Mat_MPIDense*)mat->data; 574 PetscErrorCode ierr; 575 PetscViewerFormat format; 576 int fd; 577 PetscInt header[4],mmax,N = mat->cmap->N,i,j,m,k; 578 PetscMPIInt rank,tag = ((PetscObject)viewer)->tag,size; 579 PetscScalar *work,*v,*vv; 580 Mat_SeqDense *a = (Mat_SeqDense*)mdn->A->data; 581 MPI_Status status; 582 583 PetscFunctionBegin; 584 if (mdn->size == 1) { 585 ierr = MatView(mdn->A,viewer);CHKERRQ(ierr); 586 } else { 587 ierr = PetscViewerBinaryGetDescriptor(viewer,&fd);CHKERRQ(ierr); 588 ierr = MPI_Comm_rank(((PetscObject)mat)->comm,&rank);CHKERRQ(ierr); 589 ierr = MPI_Comm_size(((PetscObject)mat)->comm,&size);CHKERRQ(ierr); 590 591 ierr = PetscViewerGetFormat(viewer,&format);CHKERRQ(ierr); 592 if (format == PETSC_VIEWER_BINARY_NATIVE) { 593 594 if (!rank) { 595 /* store the matrix as a dense matrix */ 596 header[0] = MAT_FILE_COOKIE; 597 header[1] = mat->rmap->N; 598 header[2] = N; 599 header[3] = MATRIX_BINARY_FORMAT_DENSE; 600 ierr = PetscBinaryWrite(fd,header,4,PETSC_INT,PETSC_TRUE);CHKERRQ(ierr); 601 602 /* get largest work array needed for transposing array */ 603 mmax = mat->rmap->n; 604 for (i=1; i<size; i++) { 605 mmax = PetscMax(mmax,mat->rmap->range[i+1] - mat->rmap->range[i]); 606 } 607 ierr = PetscMalloc(mmax*N*sizeof(PetscScalar),&work);CHKERRQ(ierr); 608 609 /* write out local array, by rows */ 610 m = mat->rmap->n; 611 v = a->v; 612 for (j=0; j<N; j++) { 613 for (i=0; i<m; i++) { 614 work[j + i*N] = *v++; 615 } 616 } 617 ierr = PetscBinaryWrite(fd,work,m*N,PETSC_SCALAR,PETSC_FALSE);CHKERRQ(ierr); 618 /* get largest work array to receive messages from other processes, excludes process zero */ 619 mmax = 0; 620 for (i=1; i<size; i++) { 621 mmax = PetscMax(mmax,mat->rmap->range[i+1] - mat->rmap->range[i]); 622 } 623 ierr = PetscMalloc(mmax*N*sizeof(PetscScalar),&vv);CHKERRQ(ierr); 624 for(k = 1; k < size; k++) { 625 v = vv; 626 m = mat->rmap->range[k+1] - mat->rmap->range[k]; 627 ierr = MPI_Recv(v,m*N,MPIU_SCALAR,k,tag,((PetscObject)mat)->comm,&status);CHKERRQ(ierr); 628 629 for(j = 0; j < N; j++) { 630 for(i = 0; i < m; i++) { 631 work[j + i*N] = *v++; 632 } 633 } 634 ierr = PetscBinaryWrite(fd,work,m*N,PETSC_SCALAR,PETSC_FALSE);CHKERRQ(ierr); 635 } 636 ierr = PetscFree(work);CHKERRQ(ierr); 637 ierr = PetscFree(vv);CHKERRQ(ierr); 638 } else { 639 ierr = MPI_Send(a->v,mat->rmap->n*mat->cmap->N,MPIU_SCALAR,0,tag,((PetscObject)mat)->comm);CHKERRQ(ierr); 640 } 641 } else { 642 SETERRQ(PETSC_ERR_SUP,"To store a parallel dense matrix you must first call PetscViewerSetFormat(viewer,PETSC_VIEWER_BINARY_NATIVE"); 643 } 644 } 645 PetscFunctionReturn(0); 646 } 647 648 #undef __FUNCT__ 649 #define __FUNCT__ "MatView_MPIDense_ASCIIorDraworSocket" 650 static PetscErrorCode MatView_MPIDense_ASCIIorDraworSocket(Mat mat,PetscViewer viewer) 651 { 652 Mat_MPIDense *mdn = (Mat_MPIDense*)mat->data; 653 PetscErrorCode ierr; 654 PetscMPIInt size = mdn->size,rank = mdn->rank; 655 const PetscViewerType vtype; 656 PetscTruth iascii,isdraw; 657 PetscViewer sviewer; 658 PetscViewerFormat format; 659 #if defined(PETSC_HAVE_PLAPACK) 660 Mat_Plapack *lu=(Mat_Plapack*)(mat->spptr); 661 #endif 662 663 PetscFunctionBegin; 664 ierr = PetscTypeCompare((PetscObject)viewer,PETSC_VIEWER_ASCII,&iascii);CHKERRQ(ierr); 665 ierr = PetscTypeCompare((PetscObject)viewer,PETSC_VIEWER_DRAW,&isdraw);CHKERRQ(ierr); 666 if (iascii) { 667 ierr = PetscViewerGetType(viewer,&vtype);CHKERRQ(ierr); 668 ierr = PetscViewerGetFormat(viewer,&format);CHKERRQ(ierr); 669 if (format == PETSC_VIEWER_ASCII_INFO_DETAIL) { 670 MatInfo info; 671 ierr = MatGetInfo(mat,MAT_LOCAL,&info);CHKERRQ(ierr); 672 ierr = PetscViewerASCIISynchronizedPrintf(viewer," [%d] local rows %D nz %D nz alloced %D mem %D \n",rank,mat->rmap->n, 673 (PetscInt)info.nz_used,(PetscInt)info.nz_allocated,(PetscInt)info.memory);CHKERRQ(ierr); 674 ierr = PetscViewerFlush(viewer);CHKERRQ(ierr); 675 ierr = VecScatterView(mdn->Mvctx,viewer);CHKERRQ(ierr); 676 #if defined(PETSC_HAVE_PLAPACK) 677 ierr = PetscViewerASCIIPrintf(viewer,"PLAPACK run parameters:\n");CHKERRQ(ierr); 678 ierr = PetscViewerASCIIPrintf(viewer," Processor mesh: nprows %d, npcols %d\n",lu->Plapack_nprows, lu->Plapack_npcols);CHKERRQ(ierr); 679 ierr = PetscViewerASCIIPrintf(viewer," Distr. block size nb: %d \n",lu->nb);CHKERRQ(ierr); 680 ierr = PetscViewerASCIIPrintf(viewer," Error checking: %d\n",lu->Plapack_ierror);CHKERRQ(ierr); 681 ierr = PetscViewerASCIIPrintf(viewer," Algorithmic block size: %d\n",lu->Plapack_nb_alg);CHKERRQ(ierr); 682 #endif 683 PetscFunctionReturn(0); 684 } else if (format == PETSC_VIEWER_ASCII_INFO) { 685 PetscFunctionReturn(0); 686 } 687 } else if (isdraw) { 688 PetscDraw draw; 689 PetscTruth isnull; 690 691 ierr = PetscViewerDrawGetDraw(viewer,0,&draw);CHKERRQ(ierr); 692 ierr = PetscDrawIsNull(draw,&isnull);CHKERRQ(ierr); 693 if (isnull) PetscFunctionReturn(0); 694 } 695 696 if (size == 1) { 697 ierr = MatView(mdn->A,viewer);CHKERRQ(ierr); 698 } else { 699 /* assemble the entire matrix onto first processor. */ 700 Mat A; 701 PetscInt M = mat->rmap->N,N = mat->cmap->N,m,row,i,nz; 702 PetscInt *cols; 703 PetscScalar *vals; 704 705 ierr = MatCreate(((PetscObject)mat)->comm,&A);CHKERRQ(ierr); 706 if (!rank) { 707 ierr = MatSetSizes(A,M,N,M,N);CHKERRQ(ierr); 708 } else { 709 ierr = MatSetSizes(A,0,0,M,N);CHKERRQ(ierr); 710 } 711 /* Since this is a temporary matrix, MATMPIDENSE instead of ((PetscObject)A)->type_name here is probably acceptable. */ 712 ierr = MatSetType(A,MATMPIDENSE);CHKERRQ(ierr); 713 ierr = MatMPIDenseSetPreallocation(A,PETSC_NULL); 714 ierr = PetscLogObjectParent(mat,A);CHKERRQ(ierr); 715 716 /* Copy the matrix ... This isn't the most efficient means, 717 but it's quick for now */ 718 A->insertmode = INSERT_VALUES; 719 row = mat->rmap->rstart; m = mdn->A->rmap->n; 720 for (i=0; i<m; i++) { 721 ierr = MatGetRow_MPIDense(mat,row,&nz,&cols,&vals);CHKERRQ(ierr); 722 ierr = MatSetValues_MPIDense(A,1,&row,nz,cols,vals,INSERT_VALUES);CHKERRQ(ierr); 723 ierr = MatRestoreRow_MPIDense(mat,row,&nz,&cols,&vals);CHKERRQ(ierr); 724 row++; 725 } 726 727 ierr = MatAssemblyBegin(A,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 728 ierr = MatAssemblyEnd(A,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 729 ierr = PetscViewerGetSingleton(viewer,&sviewer);CHKERRQ(ierr); 730 if (!rank) { 731 ierr = MatView(((Mat_MPIDense*)(A->data))->A,sviewer);CHKERRQ(ierr); 732 } 733 ierr = PetscViewerRestoreSingleton(viewer,&sviewer);CHKERRQ(ierr); 734 ierr = PetscViewerFlush(viewer);CHKERRQ(ierr); 735 ierr = MatDestroy(A);CHKERRQ(ierr); 736 } 737 PetscFunctionReturn(0); 738 } 739 740 #undef __FUNCT__ 741 #define __FUNCT__ "MatView_MPIDense" 742 PetscErrorCode MatView_MPIDense(Mat mat,PetscViewer viewer) 743 { 744 PetscErrorCode ierr; 745 PetscTruth iascii,isbinary,isdraw,issocket; 746 747 PetscFunctionBegin; 748 749 ierr = PetscTypeCompare((PetscObject)viewer,PETSC_VIEWER_ASCII,&iascii);CHKERRQ(ierr); 750 ierr = PetscTypeCompare((PetscObject)viewer,PETSC_VIEWER_BINARY,&isbinary);CHKERRQ(ierr); 751 ierr = PetscTypeCompare((PetscObject)viewer,PETSC_VIEWER_SOCKET,&issocket);CHKERRQ(ierr); 752 ierr = PetscTypeCompare((PetscObject)viewer,PETSC_VIEWER_DRAW,&isdraw);CHKERRQ(ierr); 753 754 if (iascii || issocket || isdraw) { 755 ierr = MatView_MPIDense_ASCIIorDraworSocket(mat,viewer);CHKERRQ(ierr); 756 } else if (isbinary) { 757 ierr = MatView_MPIDense_Binary(mat,viewer);CHKERRQ(ierr); 758 } else { 759 SETERRQ1(PETSC_ERR_SUP,"Viewer type %s not supported by MPI dense matrix",((PetscObject)viewer)->type_name); 760 } 761 PetscFunctionReturn(0); 762 } 763 764 #undef __FUNCT__ 765 #define __FUNCT__ "MatGetInfo_MPIDense" 766 PetscErrorCode MatGetInfo_MPIDense(Mat A,MatInfoType flag,MatInfo *info) 767 { 768 Mat_MPIDense *mat = (Mat_MPIDense*)A->data; 769 Mat mdn = mat->A; 770 PetscErrorCode ierr; 771 PetscReal isend[5],irecv[5]; 772 773 PetscFunctionBegin; 774 info->rows_global = (double)A->rmap->N; 775 info->columns_global = (double)A->cmap->N; 776 info->rows_local = (double)A->rmap->n; 777 info->columns_local = (double)A->cmap->N; 778 info->block_size = 1.0; 779 ierr = MatGetInfo(mdn,MAT_LOCAL,info);CHKERRQ(ierr); 780 isend[0] = info->nz_used; isend[1] = info->nz_allocated; isend[2] = info->nz_unneeded; 781 isend[3] = info->memory; isend[4] = info->mallocs; 782 if (flag == MAT_LOCAL) { 783 info->nz_used = isend[0]; 784 info->nz_allocated = isend[1]; 785 info->nz_unneeded = isend[2]; 786 info->memory = isend[3]; 787 info->mallocs = isend[4]; 788 } else if (flag == MAT_GLOBAL_MAX) { 789 ierr = MPI_Allreduce(isend,irecv,5,MPIU_REAL,MPI_MAX,((PetscObject)A)->comm);CHKERRQ(ierr); 790 info->nz_used = irecv[0]; 791 info->nz_allocated = irecv[1]; 792 info->nz_unneeded = irecv[2]; 793 info->memory = irecv[3]; 794 info->mallocs = irecv[4]; 795 } else if (flag == MAT_GLOBAL_SUM) { 796 ierr = MPI_Allreduce(isend,irecv,5,MPIU_REAL,MPI_SUM,((PetscObject)A)->comm);CHKERRQ(ierr); 797 info->nz_used = irecv[0]; 798 info->nz_allocated = irecv[1]; 799 info->nz_unneeded = irecv[2]; 800 info->memory = irecv[3]; 801 info->mallocs = irecv[4]; 802 } 803 info->fill_ratio_given = 0; /* no parallel LU/ILU/Cholesky */ 804 info->fill_ratio_needed = 0; 805 info->factor_mallocs = 0; 806 PetscFunctionReturn(0); 807 } 808 809 #undef __FUNCT__ 810 #define __FUNCT__ "MatSetOption_MPIDense" 811 PetscErrorCode MatSetOption_MPIDense(Mat A,MatOption op,PetscTruth flg) 812 { 813 Mat_MPIDense *a = (Mat_MPIDense*)A->data; 814 PetscErrorCode ierr; 815 816 PetscFunctionBegin; 817 switch (op) { 818 case MAT_NEW_NONZERO_LOCATIONS: 819 case MAT_NEW_NONZERO_LOCATION_ERR: 820 case MAT_NEW_NONZERO_ALLOCATION_ERR: 821 ierr = MatSetOption(a->A,op,flg);CHKERRQ(ierr); 822 break; 823 case MAT_ROW_ORIENTED: 824 a->roworiented = flg; 825 ierr = MatSetOption(a->A,op,flg);CHKERRQ(ierr); 826 break; 827 case MAT_NEW_DIAGONALS: 828 case MAT_USE_HASH_TABLE: 829 ierr = PetscInfo1(A,"Option %s ignored\n",MatOptions[op]);CHKERRQ(ierr); 830 break; 831 case MAT_IGNORE_OFF_PROC_ENTRIES: 832 a->donotstash = flg; 833 break; 834 case MAT_SYMMETRIC: 835 case MAT_STRUCTURALLY_SYMMETRIC: 836 case MAT_HERMITIAN: 837 case MAT_SYMMETRY_ETERNAL: 838 case MAT_IGNORE_LOWER_TRIANGULAR: 839 ierr = PetscInfo1(A,"Option %s ignored\n",MatOptions[op]);CHKERRQ(ierr); 840 break; 841 default: 842 SETERRQ1(PETSC_ERR_SUP,"unknown option %s",MatOptions[op]); 843 } 844 PetscFunctionReturn(0); 845 } 846 847 848 #undef __FUNCT__ 849 #define __FUNCT__ "MatDiagonalScale_MPIDense" 850 PetscErrorCode MatDiagonalScale_MPIDense(Mat A,Vec ll,Vec rr) 851 { 852 Mat_MPIDense *mdn = (Mat_MPIDense*)A->data; 853 Mat_SeqDense *mat = (Mat_SeqDense*)mdn->A->data; 854 PetscScalar *l,*r,x,*v; 855 PetscErrorCode ierr; 856 PetscInt i,j,s2a,s3a,s2,s3,m=mdn->A->rmap->n,n=mdn->A->cmap->n; 857 858 PetscFunctionBegin; 859 ierr = MatGetLocalSize(A,&s2,&s3);CHKERRQ(ierr); 860 if (ll) { 861 ierr = VecGetLocalSize(ll,&s2a);CHKERRQ(ierr); 862 if (s2a != s2) SETERRQ2(PETSC_ERR_ARG_SIZ,"Left scaling vector non-conforming local size, %d != %d.", s2a, s2); 863 ierr = VecGetArray(ll,&l);CHKERRQ(ierr); 864 for (i=0; i<m; i++) { 865 x = l[i]; 866 v = mat->v + i; 867 for (j=0; j<n; j++) { (*v) *= x; v+= m;} 868 } 869 ierr = VecRestoreArray(ll,&l);CHKERRQ(ierr); 870 ierr = PetscLogFlops(n*m);CHKERRQ(ierr); 871 } 872 if (rr) { 873 ierr = VecGetLocalSize(rr,&s3a);CHKERRQ(ierr); 874 if (s3a != s3) SETERRQ2(PETSC_ERR_ARG_SIZ,"Right scaling vec non-conforming local size, %d != %d.", s3a, s3); 875 ierr = VecScatterBegin(mdn->Mvctx,rr,mdn->lvec,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 876 ierr = VecScatterEnd(mdn->Mvctx,rr,mdn->lvec,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 877 ierr = VecGetArray(mdn->lvec,&r);CHKERRQ(ierr); 878 for (i=0; i<n; i++) { 879 x = r[i]; 880 v = mat->v + i*m; 881 for (j=0; j<m; j++) { (*v++) *= x;} 882 } 883 ierr = VecRestoreArray(mdn->lvec,&r);CHKERRQ(ierr); 884 ierr = PetscLogFlops(n*m);CHKERRQ(ierr); 885 } 886 PetscFunctionReturn(0); 887 } 888 889 #undef __FUNCT__ 890 #define __FUNCT__ "MatNorm_MPIDense" 891 PetscErrorCode MatNorm_MPIDense(Mat A,NormType type,PetscReal *nrm) 892 { 893 Mat_MPIDense *mdn = (Mat_MPIDense*)A->data; 894 Mat_SeqDense *mat = (Mat_SeqDense*)mdn->A->data; 895 PetscErrorCode ierr; 896 PetscInt i,j; 897 PetscReal sum = 0.0; 898 PetscScalar *v = mat->v; 899 900 PetscFunctionBegin; 901 if (mdn->size == 1) { 902 ierr = MatNorm(mdn->A,type,nrm);CHKERRQ(ierr); 903 } else { 904 if (type == NORM_FROBENIUS) { 905 for (i=0; i<mdn->A->cmap->n*mdn->A->rmap->n; i++) { 906 #if defined(PETSC_USE_COMPLEX) 907 sum += PetscRealPart(PetscConj(*v)*(*v)); v++; 908 #else 909 sum += (*v)*(*v); v++; 910 #endif 911 } 912 ierr = MPI_Allreduce(&sum,nrm,1,MPIU_REAL,MPI_SUM,((PetscObject)A)->comm);CHKERRQ(ierr); 913 *nrm = sqrt(*nrm); 914 ierr = PetscLogFlops(2*mdn->A->cmap->n*mdn->A->rmap->n);CHKERRQ(ierr); 915 } else if (type == NORM_1) { 916 PetscReal *tmp,*tmp2; 917 ierr = PetscMalloc(2*A->cmap->N*sizeof(PetscReal),&tmp);CHKERRQ(ierr); 918 tmp2 = tmp + A->cmap->N; 919 ierr = PetscMemzero(tmp,2*A->cmap->N*sizeof(PetscReal));CHKERRQ(ierr); 920 *nrm = 0.0; 921 v = mat->v; 922 for (j=0; j<mdn->A->cmap->n; j++) { 923 for (i=0; i<mdn->A->rmap->n; i++) { 924 tmp[j] += PetscAbsScalar(*v); v++; 925 } 926 } 927 ierr = MPI_Allreduce(tmp,tmp2,A->cmap->N,MPIU_REAL,MPI_SUM,((PetscObject)A)->comm);CHKERRQ(ierr); 928 for (j=0; j<A->cmap->N; j++) { 929 if (tmp2[j] > *nrm) *nrm = tmp2[j]; 930 } 931 ierr = PetscFree(tmp);CHKERRQ(ierr); 932 ierr = PetscLogFlops(A->cmap->n*A->rmap->n);CHKERRQ(ierr); 933 } else if (type == NORM_INFINITY) { /* max row norm */ 934 PetscReal ntemp; 935 ierr = MatNorm(mdn->A,type,&ntemp);CHKERRQ(ierr); 936 ierr = MPI_Allreduce(&ntemp,nrm,1,MPIU_REAL,MPI_MAX,((PetscObject)A)->comm);CHKERRQ(ierr); 937 } else { 938 SETERRQ(PETSC_ERR_SUP,"No support for two norm"); 939 } 940 } 941 PetscFunctionReturn(0); 942 } 943 944 #undef __FUNCT__ 945 #define __FUNCT__ "MatTranspose_MPIDense" 946 PetscErrorCode MatTranspose_MPIDense(Mat A,MatReuse reuse,Mat *matout) 947 { 948 Mat_MPIDense *a = (Mat_MPIDense*)A->data; 949 Mat_SeqDense *Aloc = (Mat_SeqDense*)a->A->data; 950 Mat B; 951 PetscInt M = A->rmap->N,N = A->cmap->N,m,n,*rwork,rstart = A->rmap->rstart; 952 PetscErrorCode ierr; 953 PetscInt j,i; 954 PetscScalar *v; 955 956 PetscFunctionBegin; 957 if (reuse == MAT_REUSE_MATRIX && A == *matout && M != N) SETERRQ(PETSC_ERR_SUP,"Supports square matrix only in-place"); 958 if (reuse == MAT_INITIAL_MATRIX || A == *matout) { 959 ierr = MatCreate(((PetscObject)A)->comm,&B);CHKERRQ(ierr); 960 ierr = MatSetSizes(B,A->cmap->n,A->rmap->n,N,M);CHKERRQ(ierr); 961 ierr = MatSetType(B,((PetscObject)A)->type_name);CHKERRQ(ierr); 962 ierr = MatMPIDenseSetPreallocation(B,PETSC_NULL);CHKERRQ(ierr); 963 } else { 964 B = *matout; 965 } 966 967 m = a->A->rmap->n; n = a->A->cmap->n; v = Aloc->v; 968 ierr = PetscMalloc(m*sizeof(PetscInt),&rwork);CHKERRQ(ierr); 969 for (i=0; i<m; i++) rwork[i] = rstart + i; 970 for (j=0; j<n; j++) { 971 ierr = MatSetValues(B,1,&j,m,rwork,v,INSERT_VALUES);CHKERRQ(ierr); 972 v += m; 973 } 974 ierr = PetscFree(rwork);CHKERRQ(ierr); 975 ierr = MatAssemblyBegin(B,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 976 ierr = MatAssemblyEnd(B,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 977 if (reuse == MAT_INITIAL_MATRIX || *matout != A) { 978 *matout = B; 979 } else { 980 ierr = MatHeaderCopy(A,B);CHKERRQ(ierr); 981 } 982 PetscFunctionReturn(0); 983 } 984 985 #include "petscblaslapack.h" 986 #undef __FUNCT__ 987 #define __FUNCT__ "MatScale_MPIDense" 988 PetscErrorCode MatScale_MPIDense(Mat inA,PetscScalar alpha) 989 { 990 Mat_MPIDense *A = (Mat_MPIDense*)inA->data; 991 Mat_SeqDense *a = (Mat_SeqDense*)A->A->data; 992 PetscScalar oalpha = alpha; 993 PetscErrorCode ierr; 994 PetscBLASInt one = 1,nz = PetscBLASIntCast(inA->rmap->n*inA->cmap->N); 995 996 PetscFunctionBegin; 997 BLASscal_(&nz,&oalpha,a->v,&one); 998 ierr = PetscLogFlops(nz);CHKERRQ(ierr); 999 PetscFunctionReturn(0); 1000 } 1001 1002 static PetscErrorCode MatDuplicate_MPIDense(Mat,MatDuplicateOption,Mat *); 1003 1004 #undef __FUNCT__ 1005 #define __FUNCT__ "MatSetUpPreallocation_MPIDense" 1006 PetscErrorCode MatSetUpPreallocation_MPIDense(Mat A) 1007 { 1008 PetscErrorCode ierr; 1009 1010 PetscFunctionBegin; 1011 ierr = MatMPIDenseSetPreallocation(A,0);CHKERRQ(ierr); 1012 PetscFunctionReturn(0); 1013 } 1014 1015 #if defined(PETSC_HAVE_PLAPACK) 1016 1017 #undef __FUNCT__ 1018 #define __FUNCT__ "MatMPIDenseCopyToPlapack" 1019 PetscErrorCode MatMPIDenseCopyToPlapack(Mat A,Mat F) 1020 { 1021 Mat_Plapack *lu = (Mat_Plapack*)(F)->spptr; 1022 PetscErrorCode ierr; 1023 PetscInt M=A->cmap->N,m=A->rmap->n,rstart; 1024 PetscScalar *array; 1025 PetscReal one = 1.0; 1026 1027 PetscFunctionBegin; 1028 /* Copy A into F->lu->A */ 1029 ierr = PLA_Obj_set_to_zero(lu->A);CHKERRQ(ierr); 1030 ierr = PLA_API_begin();CHKERRQ(ierr); 1031 ierr = PLA_Obj_API_open(lu->A);CHKERRQ(ierr); 1032 ierr = MatGetOwnershipRange(A,&rstart,PETSC_NULL);CHKERRQ(ierr); 1033 ierr = MatGetArray(A,&array);CHKERRQ(ierr); 1034 ierr = PLA_API_axpy_matrix_to_global(m,M, &one,(void *)array,m,lu->A,rstart,0);CHKERRQ(ierr); 1035 ierr = MatRestoreArray(A,&array);CHKERRQ(ierr); 1036 ierr = PLA_Obj_API_close(lu->A);CHKERRQ(ierr); 1037 ierr = PLA_API_end();CHKERRQ(ierr); 1038 lu->rstart = rstart; 1039 PetscFunctionReturn(0); 1040 } 1041 1042 #undef __FUNCT__ 1043 #define __FUNCT__ "MatMPIDenseCopyFromPlapack" 1044 PetscErrorCode MatMPIDenseCopyFromPlapack(Mat F,Mat A) 1045 { 1046 Mat_Plapack *lu = (Mat_Plapack*)(F)->spptr; 1047 PetscErrorCode ierr; 1048 PetscInt M=A->cmap->N,m=A->rmap->n,rstart; 1049 PetscScalar *array; 1050 PetscReal one = 1.0; 1051 1052 PetscFunctionBegin; 1053 /* Copy F into A->lu->A */ 1054 ierr = MatZeroEntries(A);CHKERRQ(ierr); 1055 ierr = PLA_API_begin();CHKERRQ(ierr); 1056 ierr = PLA_Obj_API_open(lu->A);CHKERRQ(ierr); 1057 ierr = MatGetOwnershipRange(A,&rstart,PETSC_NULL);CHKERRQ(ierr); 1058 ierr = MatGetArray(A,&array);CHKERRQ(ierr); 1059 ierr = PLA_API_axpy_global_to_matrix(m,M, &one,lu->A,rstart,0,(void *)array,m);CHKERRQ(ierr); 1060 ierr = MatRestoreArray(A,&array);CHKERRQ(ierr); 1061 ierr = PLA_Obj_API_close(lu->A);CHKERRQ(ierr); 1062 ierr = PLA_API_end();CHKERRQ(ierr); 1063 lu->rstart = rstart; 1064 PetscFunctionReturn(0); 1065 } 1066 1067 #undef __FUNCT__ 1068 #define __FUNCT__ "MatMatMultNumeric_MPIDense_MPIDense" 1069 PetscErrorCode MatMatMultNumeric_MPIDense_MPIDense(Mat A,Mat B,Mat C) 1070 { 1071 PetscErrorCode ierr; 1072 Mat_Plapack *luA = (Mat_Plapack*)A->spptr; 1073 Mat_Plapack *luB = (Mat_Plapack*)B->spptr; 1074 Mat_Plapack *luC = (Mat_Plapack*)C->spptr; 1075 PLA_Obj alpha = NULL,beta = NULL; 1076 1077 PetscFunctionBegin; 1078 ierr = MatMPIDenseCopyToPlapack(A,A);CHKERRQ(ierr); 1079 ierr = MatMPIDenseCopyToPlapack(B,B);CHKERRQ(ierr); 1080 1081 /* 1082 ierr = PLA_Global_show("A = ",luA->A,"%g ","");CHKERRQ(ierr); 1083 ierr = PLA_Global_show("B = ",luB->A,"%g ","");CHKERRQ(ierr); 1084 */ 1085 1086 /* do the multiply in PLA */ 1087 ierr = PLA_Create_constants_conf_to(luA->A,NULL,NULL,&alpha);CHKERRQ(ierr); 1088 ierr = PLA_Create_constants_conf_to(luC->A,NULL,&beta,NULL);CHKERRQ(ierr); 1089 CHKMEMQ; 1090 1091 ierr = PLA_Gemm(PLA_NO_TRANSPOSE,PLA_NO_TRANSPOSE,alpha,luA->A,luB->A,beta,luC->A); /* CHKERRQ(ierr); */ 1092 CHKMEMQ; 1093 ierr = PLA_Obj_free(&alpha);CHKERRQ(ierr); 1094 ierr = PLA_Obj_free(&beta);CHKERRQ(ierr); 1095 1096 /* 1097 ierr = PLA_Global_show("C = ",luC->A,"%g ","");CHKERRQ(ierr); 1098 */ 1099 ierr = MatMPIDenseCopyFromPlapack(C,C);CHKERRQ(ierr); 1100 PetscFunctionReturn(0); 1101 } 1102 1103 #undef __FUNCT__ 1104 #define __FUNCT__ "MatMatMultSymbolic_MPIDense_MPIDense" 1105 PetscErrorCode MatMatMultSymbolic_MPIDense_MPIDense(Mat A,Mat B,PetscReal fill,Mat *C) 1106 { 1107 PetscErrorCode ierr; 1108 PetscInt m=A->rmap->n,n=B->cmap->n; 1109 Mat Cmat; 1110 1111 PetscFunctionBegin; 1112 if (A->cmap->n != B->rmap->n) SETERRQ2(PETSC_ERR_ARG_SIZ,"A->cmap->n %d != B->rmap->n %d\n",A->cmap->n,B->rmap->n); 1113 SETERRQ(PETSC_ERR_LIB,"Due to aparent bugs in PLAPACK,this is not currently supported"); 1114 ierr = MatCreate(((PetscObject)B)->comm,&Cmat);CHKERRQ(ierr); 1115 ierr = MatSetSizes(Cmat,m,n,A->rmap->N,B->cmap->N);CHKERRQ(ierr); 1116 ierr = MatSetType(Cmat,MATMPIDENSE);CHKERRQ(ierr); 1117 ierr = MatAssemblyBegin(Cmat,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1118 ierr = MatAssemblyEnd(Cmat,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1119 //ierr = MatMPIDenseCreatePlapack(A);CHKERRQ(ierr); 1120 //ierr = MatMPIDenseCreatePlapack(B);CHKERRQ(ierr); 1121 //ierr = MatMPIDenseCreatePlapack(Cmat);CHKERRQ(ierr); 1122 *C = Cmat; 1123 PetscFunctionReturn(0); 1124 } 1125 1126 #undef __FUNCT__ 1127 #define __FUNCT__ "MatMatMult_MPIDense_MPIDense" 1128 PetscErrorCode MatMatMult_MPIDense_MPIDense(Mat A,Mat B,MatReuse scall,PetscReal fill,Mat *C) 1129 { 1130 PetscErrorCode ierr; 1131 1132 PetscFunctionBegin; 1133 if (scall == MAT_INITIAL_MATRIX){ 1134 ierr = MatMatMultSymbolic_MPIDense_MPIDense(A,B,fill,C);CHKERRQ(ierr); 1135 } 1136 ierr = MatMatMultNumeric_MPIDense_MPIDense(A,B,*C);CHKERRQ(ierr); 1137 PetscFunctionReturn(0); 1138 } 1139 1140 #undef __FUNCT__ 1141 #define __FUNCT__ "MatSolve_MPIDense" 1142 PetscErrorCode MatSolve_MPIDense(Mat A,Vec b,Vec x) 1143 { 1144 MPI_Comm comm = ((PetscObject)A)->comm; 1145 Mat_Plapack *lu = (Mat_Plapack*)A->spptr; 1146 PetscErrorCode ierr; 1147 PetscInt M=A->rmap->N,m=A->rmap->n,rstart,i,j,*idx_pla,*idx_petsc,loc_m,loc_stride; 1148 PetscScalar *array; 1149 PetscReal one = 1.0; 1150 PetscMPIInt size,rank,r_rank,r_nproc,c_rank,c_nproc;; 1151 PLA_Obj v_pla = NULL; 1152 PetscScalar *loc_buf; 1153 Vec loc_x; 1154 1155 PetscFunctionBegin; 1156 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr); 1157 ierr = MPI_Comm_rank(comm,&rank);CHKERRQ(ierr); 1158 1159 /* Create PLAPACK vector objects, then copy b into PLAPACK b */ 1160 PLA_Mvector_create(lu->datatype,M,1,lu->templ,PLA_ALIGN_FIRST,&v_pla); 1161 PLA_Obj_set_to_zero(v_pla); 1162 1163 /* Copy b into rhs_pla */ 1164 PLA_API_begin(); 1165 PLA_Obj_API_open(v_pla); 1166 ierr = VecGetArray(b,&array);CHKERRQ(ierr); 1167 PLA_API_axpy_vector_to_global(m,&one,(void *)array,1,v_pla,lu->rstart); 1168 ierr = VecRestoreArray(b,&array);CHKERRQ(ierr); 1169 PLA_Obj_API_close(v_pla); 1170 PLA_API_end(); 1171 1172 if (A->factor == MAT_FACTOR_LU){ 1173 /* Apply the permutations to the right hand sides */ 1174 PLA_Apply_pivots_to_rows (v_pla,lu->pivots); 1175 1176 /* Solve L y = b, overwriting b with y */ 1177 PLA_Trsv( PLA_LOWER_TRIANGULAR,PLA_NO_TRANSPOSE,PLA_UNIT_DIAG,lu->A,v_pla ); 1178 1179 /* Solve U x = y (=b), overwriting b with x */ 1180 PLA_Trsv( PLA_UPPER_TRIANGULAR,PLA_NO_TRANSPOSE,PLA_NONUNIT_DIAG,lu->A,v_pla ); 1181 } else { /*MAT_FACTOR_CHOLESKY */ 1182 PLA_Trsv( PLA_LOWER_TRIANGULAR,PLA_NO_TRANSPOSE,PLA_NONUNIT_DIAG,lu->A,v_pla); 1183 PLA_Trsv( PLA_LOWER_TRIANGULAR,(lu->datatype == MPI_DOUBLE ? PLA_TRANSPOSE : PLA_CONJUGATE_TRANSPOSE), 1184 PLA_NONUNIT_DIAG,lu->A,v_pla); 1185 } 1186 1187 /* Copy PLAPACK x into Petsc vector x */ 1188 PLA_Obj_local_length(v_pla, &loc_m); 1189 PLA_Obj_local_buffer(v_pla, (void**)&loc_buf); 1190 PLA_Obj_local_stride(v_pla, &loc_stride); 1191 /* 1192 PetscPrintf(PETSC_COMM_SELF," [%d] b - local_m %d local_stride %d, loc_buf: %g %g, nb: %d\n",rank,loc_m,loc_stride,loc_buf[0],loc_buf[(loc_m-1)*loc_stride],lu->nb); 1193 */ 1194 ierr = VecCreateSeqWithArray(PETSC_COMM_SELF,loc_m*loc_stride,loc_buf,&loc_x);CHKERRQ(ierr); 1195 if (!lu->pla_solved){ 1196 1197 PLA_Temp_comm_row_info(lu->templ,&lu->comm_2d,&r_rank,&r_nproc); 1198 PLA_Temp_comm_col_info(lu->templ,&lu->comm_2d,&c_rank,&c_nproc); 1199 /* printf(" [%d] rank: %d %d, nproc: %d %d\n",rank,r_rank,c_rank,r_nproc,c_nproc); */ 1200 1201 /* Create IS and cts for VecScatterring */ 1202 PLA_Obj_local_length(v_pla, &loc_m); 1203 PLA_Obj_local_stride(v_pla, &loc_stride); 1204 ierr = PetscMalloc((2*loc_m+1)*sizeof(PetscInt),&idx_pla);CHKERRQ(ierr); 1205 idx_petsc = idx_pla + loc_m; 1206 1207 rstart = (r_rank*c_nproc+c_rank)*lu->nb; 1208 for (i=0; i<loc_m; i+=lu->nb){ 1209 j = 0; 1210 while (j < lu->nb && i+j < loc_m){ 1211 idx_petsc[i+j] = rstart + j; j++; 1212 } 1213 rstart += size*lu->nb; 1214 } 1215 1216 for (i=0; i<loc_m; i++) idx_pla[i] = i*loc_stride; 1217 1218 ierr = ISCreateGeneral(PETSC_COMM_SELF,loc_m,idx_pla,&lu->is_pla);CHKERRQ(ierr); 1219 ierr = ISCreateGeneral(PETSC_COMM_SELF,loc_m,idx_petsc,&lu->is_petsc);CHKERRQ(ierr); 1220 ierr = PetscFree(idx_pla);CHKERRQ(ierr); 1221 ierr = VecScatterCreate(loc_x,lu->is_pla,x,lu->is_petsc,&lu->ctx);CHKERRQ(ierr); 1222 } 1223 ierr = VecScatterBegin(lu->ctx,loc_x,x,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 1224 ierr = VecScatterEnd(lu->ctx,loc_x,x,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr); 1225 1226 /* Free data */ 1227 ierr = VecDestroy(loc_x);CHKERRQ(ierr); 1228 PLA_Obj_free(&v_pla); 1229 1230 lu->pla_solved = PETSC_TRUE; 1231 PetscFunctionReturn(0); 1232 } 1233 1234 #undef __FUNCT__ 1235 #define __FUNCT__ "MatLUFactorNumeric_MPIDense" 1236 PetscErrorCode MatLUFactorNumeric_MPIDense(Mat F,Mat A,MatFactorInfo *info) 1237 { 1238 Mat_Plapack *lu = (Mat_Plapack*)(F)->spptr; 1239 PetscErrorCode ierr; 1240 PetscInt M=A->rmap->N,m=A->rmap->n,rstart,rend; 1241 PetscInt info_pla=0; 1242 PetscScalar *array,one = 1.0; 1243 1244 PetscFunctionBegin; 1245 if (lu->mstruct == SAME_NONZERO_PATTERN){ 1246 PLA_Obj_free(&lu->A); 1247 PLA_Obj_free (&lu->pivots); 1248 } 1249 /* Create PLAPACK matrix object */ 1250 lu->A = NULL; lu->pivots = NULL; 1251 PLA_Matrix_create(lu->datatype,M,M,lu->templ,PLA_ALIGN_FIRST,PLA_ALIGN_FIRST,&lu->A); 1252 PLA_Obj_set_to_zero(lu->A); 1253 PLA_Mvector_create(MPI_INT,M,1,lu->templ,PLA_ALIGN_FIRST,&lu->pivots); 1254 1255 /* Copy A into lu->A */ 1256 PLA_API_begin(); 1257 PLA_Obj_API_open(lu->A); 1258 ierr = MatGetOwnershipRange(A,&rstart,&rend);CHKERRQ(ierr); 1259 ierr = MatGetArray(A,&array);CHKERRQ(ierr); 1260 PLA_API_axpy_matrix_to_global(m,M, &one,(void *)array,m,lu->A,rstart,0); 1261 ierr = MatRestoreArray(A,&array);CHKERRQ(ierr); 1262 PLA_Obj_API_close(lu->A); 1263 PLA_API_end(); 1264 1265 /* Factor P A -> L U overwriting lower triangular portion of A with L, upper, U */ 1266 info_pla = PLA_LU(lu->A,lu->pivots); 1267 if (info_pla != 0) 1268 SETERRQ1(PETSC_ERR_MAT_LU_ZRPVT,"Zero pivot encountered at row %d from PLA_LU()",info_pla); 1269 1270 lu->CleanUpPlapack = PETSC_TRUE; 1271 lu->rstart = rstart; 1272 lu->mstruct = SAME_NONZERO_PATTERN; 1273 F->ops->solve = MatSolve_MPIDense; 1274 F->assembled = PETSC_TRUE; /* required by -ksp_view */ 1275 PetscFunctionReturn(0); 1276 } 1277 1278 #undef __FUNCT__ 1279 #define __FUNCT__ "MatCholeskyFactorNumeric_MPIDense" 1280 PetscErrorCode MatCholeskyFactorNumeric_MPIDense(Mat F,Mat A,MatFactorInfo *info) 1281 { 1282 Mat_Plapack *lu = (Mat_Plapack*)F->spptr; 1283 PetscErrorCode ierr; 1284 PetscInt M=A->rmap->N,m=A->rmap->n,rstart,rend; 1285 PetscInt info_pla=0; 1286 PetscScalar *array,one = 1.0; 1287 1288 PetscFunctionBegin; 1289 if (lu->mstruct == SAME_NONZERO_PATTERN){ 1290 PLA_Obj_free(&lu->A); 1291 } 1292 /* Create PLAPACK matrix object */ 1293 lu->A = NULL; 1294 lu->pivots = NULL; 1295 PLA_Matrix_create(lu->datatype,M,M,lu->templ,PLA_ALIGN_FIRST,PLA_ALIGN_FIRST,&lu->A); 1296 1297 /* Copy A into lu->A */ 1298 PLA_API_begin(); 1299 PLA_Obj_API_open(lu->A); 1300 ierr = MatGetOwnershipRange(A,&rstart,&rend);CHKERRQ(ierr); 1301 ierr = MatGetArray(A,&array);CHKERRQ(ierr); 1302 PLA_API_axpy_matrix_to_global(m,M, &one,(void *)array,m,lu->A,rstart,0); 1303 ierr = MatRestoreArray(A,&array);CHKERRQ(ierr); 1304 PLA_Obj_API_close(lu->A); 1305 PLA_API_end(); 1306 1307 /* Factor P A -> Chol */ 1308 info_pla = PLA_Chol(PLA_LOWER_TRIANGULAR,lu->A); 1309 if (info_pla != 0) 1310 SETERRQ1( PETSC_ERR_MAT_CH_ZRPVT,"Nonpositive definite matrix detected at row %d from PLA_Chol()",info_pla); 1311 1312 lu->CleanUpPlapack = PETSC_TRUE; 1313 lu->rstart = rstart; 1314 lu->mstruct = SAME_NONZERO_PATTERN; 1315 F->ops->solve = MatSolve_MPIDense; 1316 F->assembled = PETSC_TRUE; /* required by -ksp_view */ 1317 PetscFunctionReturn(0); 1318 } 1319 1320 #undef __FUNCT__ 1321 #define __FUNCT__ "MatFactorSymbolic_Plapack_Private" 1322 PetscErrorCode MatFactorSymbolic_Plapack_Private(Mat F,Mat A,MatFactorInfo *info) 1323 { 1324 Mat B = F; 1325 Mat_Plapack *lu; 1326 PetscErrorCode ierr; 1327 PetscInt M=A->rmap->N; 1328 MPI_Comm comm=((PetscObject)A)->comm,comm_2d; 1329 PetscMPIInt size; 1330 PetscInt ierror; 1331 1332 PetscFunctionBegin; 1333 lu = (Mat_Plapack*)(B->spptr); 1334 1335 /* Set default Plapack parameters */ 1336 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr); 1337 lu->nprows = 1; lu->npcols = size; 1338 ierror = 0; 1339 lu->nb = M/size; 1340 if (M - lu->nb*size) lu->nb++; /* without cyclic distribution */ 1341 1342 /* Set runtime options */ 1343 ierr = PetscOptionsBegin(((PetscObject)A)->comm,((PetscObject)A)->prefix,"PLAPACK Options","Mat");CHKERRQ(ierr); 1344 ierr = PetscOptionsInt("-mat_plapack_nprows","row dimension of 2D processor mesh","None",lu->nprows,&lu->nprows,PETSC_NULL);CHKERRQ(ierr); 1345 ierr = PetscOptionsInt("-mat_plapack_npcols","column dimension of 2D processor mesh","None",lu->npcols,&lu->npcols,PETSC_NULL);CHKERRQ(ierr); 1346 1347 ierr = PetscOptionsInt("-mat_plapack_nb","block size of template vector","None",lu->nb,&lu->nb,PETSC_NULL);CHKERRQ(ierr); 1348 ierr = PetscOptionsInt("-mat_plapack_ckerror","error checking flag","None",ierror,&ierror,PETSC_NULL);CHKERRQ(ierr); 1349 if (ierror){ 1350 PLA_Set_error_checking(ierror,PETSC_TRUE,PETSC_TRUE,PETSC_FALSE ); 1351 } else { 1352 PLA_Set_error_checking(ierror,PETSC_FALSE,PETSC_FALSE,PETSC_FALSE ); 1353 } 1354 lu->ierror = ierror; 1355 1356 lu->nb_alg = 0; 1357 ierr = PetscOptionsInt("-mat_plapack_nb_alg","algorithmic block size","None",lu->nb_alg,&lu->nb_alg,PETSC_NULL);CHKERRQ(ierr); 1358 if (lu->nb_alg){ 1359 pla_Environ_set_nb_alg (PLA_OP_ALL_ALG,lu->nb_alg); 1360 } 1361 PetscOptionsEnd(); 1362 1363 1364 /* Create a 2D communicator */ 1365 PLA_Comm_1D_to_2D(comm,lu->nprows,lu->npcols,&comm_2d); 1366 lu->comm_2d = comm_2d; 1367 1368 /* Initialize PLAPACK */ 1369 PLA_Init(comm_2d); 1370 1371 /* Create object distribution template */ 1372 lu->templ = NULL; 1373 PLA_Temp_create(lu->nb, 0, &lu->templ); 1374 1375 /* Use suggested nb_alg if it is not provided by user */ 1376 if (lu->nb_alg == 0){ 1377 PLA_Environ_nb_alg(PLA_OP_PAN_PAN,lu->templ,&lu->nb_alg); 1378 pla_Environ_set_nb_alg(PLA_OP_ALL_ALG,lu->nb_alg); 1379 } 1380 1381 /* Set the datatype */ 1382 #if defined(PETSC_USE_COMPLEX) 1383 lu->datatype = MPI_DOUBLE_COMPLEX; 1384 #else 1385 lu->datatype = MPI_DOUBLE; 1386 #endif 1387 1388 lu->pla_solved = PETSC_FALSE; /* MatSolve_Plapack() is called yet */ 1389 lu->mstruct = DIFFERENT_NONZERO_PATTERN; 1390 lu->CleanUpPlapack = PETSC_TRUE; 1391 PetscFunctionReturn(0); 1392 } 1393 1394 /* Note the Petsc perm permutation is ignored */ 1395 #undef __FUNCT__ 1396 #define __FUNCT__ "MatCholeskyFactorSymbolic_MPIDense" 1397 PetscErrorCode MatCholeskyFactorSymbolic_MPIDense(Mat F,Mat A,IS perm,MatFactorInfo *info) 1398 { 1399 PetscErrorCode ierr; 1400 PetscTruth issymmetric,set; 1401 1402 PetscFunctionBegin; 1403 ierr = MatIsSymmetricKnown(A,&set,&issymmetric); CHKERRQ(ierr); 1404 if (!set || !issymmetric) SETERRQ(PETSC_ERR_USER,"Matrix must be set as MAT_SYMMETRIC for CholeskyFactor()"); 1405 ierr = MatFactorSymbolic_Plapack_Private(F,A,info);CHKERRQ(ierr); 1406 F->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_MPIDense; 1407 PetscFunctionReturn(0); 1408 } 1409 1410 /* Note the Petsc r and c permutations are ignored */ 1411 #undef __FUNCT__ 1412 #define __FUNCT__ "MatLUFactorSymbolic_MPIDense" 1413 PetscErrorCode MatLUFactorSymbolic_MPIDense(Mat F,Mat A,IS r,IS c,MatFactorInfo *info) 1414 { 1415 PetscErrorCode ierr; 1416 PetscInt M = A->rmap->N; 1417 Mat_Plapack *lu; 1418 1419 PetscFunctionBegin; 1420 ierr = MatFactorSymbolic_Plapack_Private(F,A,info);CHKERRQ(ierr); 1421 lu = (Mat_Plapack*)F->spptr; 1422 ierr = PLA_Mvector_create(MPI_INT,M,1,lu->templ,PLA_ALIGN_FIRST,&lu->pivots);CHKERRQ(ierr); 1423 F->ops->lufactornumeric = MatLUFactorNumeric_MPIDense; 1424 PetscFunctionReturn(0); 1425 } 1426 1427 #undef __FUNCT__ 1428 #define __FUNCT__ "MatGetFactor_mpidense_plapack" 1429 PetscErrorCode MatGetFactor_mpidense_plapack(Mat A,MatFactorType ftype,Mat *F) 1430 { 1431 PetscErrorCode ierr; 1432 Mat_Plapack *lu; 1433 PetscMPIInt size; 1434 PetscInt M; 1435 1436 PetscFunctionBegin; 1437 /* Create the factorization matrix */ 1438 ierr = MatCreate(((PetscObject)A)->comm,F);CHKERRQ(ierr); 1439 ierr = MatSetSizes(*F,A->rmap->n,A->cmap->n,A->rmap->N,A->cmap->N);CHKERRQ(ierr); 1440 ierr = MatSetType(*F,((PetscObject)A)->type_name);CHKERRQ(ierr); 1441 ierr = PetscNewLog(A,Mat_Plapack,&lu);CHKERRQ(ierr); 1442 A->spptr = (void*)lu; 1443 1444 lu = (Mat_Plapack*)(A->spptr); 1445 1446 /* Set default Plapack parameters */ 1447 ierr = MPI_Comm_size(((PetscObject)A)->comm,&size);CHKERRQ(ierr); 1448 lu->nb = M/size; 1449 if (M - lu->nb*size) lu->nb++; /* without cyclic distribution */ 1450 1451 /* Set runtime options */ 1452 ierr = PetscOptionsBegin(((PetscObject)A)->comm,((PetscObject)A)->prefix,"PLAPACK Options","Mat");CHKERRQ(ierr); 1453 ierr = PetscOptionsInt("-mat_plapack_nb","block size of template vector","None",lu->nb,&lu->nb,PETSC_NULL);CHKERRQ(ierr); 1454 PetscOptionsEnd(); 1455 1456 /* Create object distribution template */ 1457 lu->templ = NULL; 1458 ierr = PLA_Temp_create(lu->nb, 0, &lu->templ);CHKERRQ(ierr); 1459 1460 /* Set the datatype */ 1461 #if defined(PETSC_USE_COMPLEX) 1462 lu->datatype = MPI_DOUBLE_COMPLEX; 1463 #else 1464 lu->datatype = MPI_DOUBLE; 1465 #endif 1466 1467 ierr = PLA_Matrix_create(lu->datatype,M,A->cmap->N,lu->templ,PLA_ALIGN_FIRST,PLA_ALIGN_FIRST,&lu->A);CHKERRQ(ierr); 1468 1469 1470 lu->pla_solved = PETSC_FALSE; /* MatSolve_Plapack() is called yet */ 1471 1472 if (ftype == MAT_FACTOR_LU) { 1473 (*F)->ops->lufactorsymbolic = MatLUFactorSymbolic_MPIDense; 1474 } else if (ftype == MAT_FACTOR_CHOLESKY) { 1475 (*F)->ops->choleskyfactorsymbolic = MatCholeskyFactorSymbolic_MPIDense; 1476 } else SETERRQ(PETSC_ERR_SUP,"No incomplete factorizations for dense matrices"); 1477 (*F)->factor = ftype; 1478 PetscFunctionReturn(0); 1479 } 1480 #endif 1481 1482 /* -------------------------------------------------------------------*/ 1483 static struct _MatOps MatOps_Values = {MatSetValues_MPIDense, 1484 MatGetRow_MPIDense, 1485 MatRestoreRow_MPIDense, 1486 MatMult_MPIDense, 1487 /* 4*/ MatMultAdd_MPIDense, 1488 MatMultTranspose_MPIDense, 1489 MatMultTransposeAdd_MPIDense, 1490 0, 1491 0, 1492 0, 1493 /*10*/ 0, 1494 0, 1495 0, 1496 0, 1497 MatTranspose_MPIDense, 1498 /*15*/ MatGetInfo_MPIDense, 1499 MatEqual_MPIDense, 1500 MatGetDiagonal_MPIDense, 1501 MatDiagonalScale_MPIDense, 1502 MatNorm_MPIDense, 1503 /*20*/ MatAssemblyBegin_MPIDense, 1504 MatAssemblyEnd_MPIDense, 1505 0, 1506 MatSetOption_MPIDense, 1507 MatZeroEntries_MPIDense, 1508 /*25*/ MatZeroRows_MPIDense, 1509 0, 1510 0, 1511 0, 1512 0, 1513 /*30*/ MatSetUpPreallocation_MPIDense, 1514 0, 1515 0, 1516 MatGetArray_MPIDense, 1517 MatRestoreArray_MPIDense, 1518 /*35*/ MatDuplicate_MPIDense, 1519 0, 1520 0, 1521 0, 1522 0, 1523 /*40*/ 0, 1524 MatGetSubMatrices_MPIDense, 1525 0, 1526 MatGetValues_MPIDense, 1527 0, 1528 /*45*/ 0, 1529 MatScale_MPIDense, 1530 0, 1531 0, 1532 0, 1533 /*50*/ 0, 1534 0, 1535 0, 1536 0, 1537 0, 1538 /*55*/ 0, 1539 0, 1540 0, 1541 0, 1542 0, 1543 /*60*/ MatGetSubMatrix_MPIDense, 1544 MatDestroy_MPIDense, 1545 MatView_MPIDense, 1546 0, 1547 0, 1548 /*65*/ 0, 1549 0, 1550 0, 1551 0, 1552 0, 1553 /*70*/ 0, 1554 0, 1555 0, 1556 0, 1557 0, 1558 /*75*/ 0, 1559 0, 1560 0, 1561 0, 1562 0, 1563 /*80*/ 0, 1564 0, 1565 0, 1566 0, 1567 /*84*/ MatLoad_MPIDense, 1568 0, 1569 0, 1570 0, 1571 0, 1572 0, 1573 /*90*/ 1574 #if defined(PETSC_HAVE_PLAPACK) 1575 MatMatMult_MPIDense_MPIDense, 1576 MatMatMultSymbolic_MPIDense_MPIDense, 1577 MatMatMultNumeric_MPIDense_MPIDense, 1578 #else 1579 0, 1580 0, 1581 0, 1582 #endif 1583 0, 1584 /*95*/ 0, 1585 0, 1586 0, 1587 0}; 1588 1589 EXTERN_C_BEGIN 1590 #undef __FUNCT__ 1591 #define __FUNCT__ "MatMPIDenseSetPreallocation_MPIDense" 1592 PetscErrorCode PETSCMAT_DLLEXPORT MatMPIDenseSetPreallocation_MPIDense(Mat mat,PetscScalar *data) 1593 { 1594 Mat_MPIDense *a; 1595 PetscErrorCode ierr; 1596 1597 PetscFunctionBegin; 1598 mat->preallocated = PETSC_TRUE; 1599 /* Note: For now, when data is specified above, this assumes the user correctly 1600 allocates the local dense storage space. We should add error checking. */ 1601 1602 a = (Mat_MPIDense*)mat->data; 1603 ierr = MatCreate(PETSC_COMM_SELF,&a->A);CHKERRQ(ierr); 1604 ierr = MatSetSizes(a->A,mat->rmap->n,mat->cmap->N,mat->rmap->n,mat->cmap->N);CHKERRQ(ierr); 1605 ierr = MatSetType(a->A,MATSEQDENSE);CHKERRQ(ierr); 1606 ierr = MatSeqDenseSetPreallocation(a->A,data);CHKERRQ(ierr); 1607 ierr = PetscLogObjectParent(mat,a->A);CHKERRQ(ierr); 1608 PetscFunctionReturn(0); 1609 } 1610 EXTERN_C_END 1611 1612 /*MC 1613 MATMPIDENSE - MATMPIDENSE = "mpidense" - A matrix type to be used for distributed dense matrices. 1614 1615 Options Database Keys: 1616 . -mat_type mpidense - sets the matrix type to "mpidense" during a call to MatSetFromOptions() 1617 1618 Level: beginner 1619 1620 MATMPIDENSE matrices may use direct solvers (LU, Cholesky, and QR) 1621 for parallel dense matrices via the external package PLAPACK, if PLAPACK is installed 1622 (run config/configure.py with the option --download-plapack) 1623 1624 1625 Options Database Keys: 1626 . -mat_plapack_nprows <n> - number of rows in processor partition 1627 . -mat_plapack_npcols <n> - number of columns in processor partition 1628 . -mat_plapack_nb <n> - block size of template vector 1629 . -mat_plapack_nb_alg <n> - algorithmic block size 1630 - -mat_plapack_ckerror <n> - error checking flag 1631 1632 .seealso: MatCreateMPIDense(), MATDENSE, MATSEQDENSE 1633 M*/ 1634 1635 EXTERN_C_BEGIN 1636 #undef __FUNCT__ 1637 #define __FUNCT__ "MatCreate_MPIDense" 1638 PetscErrorCode PETSCMAT_DLLEXPORT MatCreate_MPIDense(Mat mat) 1639 { 1640 Mat_MPIDense *a; 1641 PetscErrorCode ierr; 1642 1643 PetscFunctionBegin; 1644 ierr = PetscNewLog(mat,Mat_MPIDense,&a);CHKERRQ(ierr); 1645 mat->data = (void*)a; 1646 ierr = PetscMemcpy(mat->ops,&MatOps_Values,sizeof(struct _MatOps));CHKERRQ(ierr); 1647 mat->mapping = 0; 1648 1649 mat->insertmode = NOT_SET_VALUES; 1650 ierr = MPI_Comm_rank(((PetscObject)mat)->comm,&a->rank);CHKERRQ(ierr); 1651 ierr = MPI_Comm_size(((PetscObject)mat)->comm,&a->size);CHKERRQ(ierr); 1652 1653 mat->rmap->bs = mat->cmap->bs = 1; 1654 ierr = PetscMapSetUp(mat->rmap);CHKERRQ(ierr); 1655 ierr = PetscMapSetUp(mat->cmap);CHKERRQ(ierr); 1656 a->nvec = mat->cmap->n; 1657 1658 /* build cache for off array entries formed */ 1659 a->donotstash = PETSC_FALSE; 1660 ierr = MatStashCreate_Private(((PetscObject)mat)->comm,1,&mat->stash);CHKERRQ(ierr); 1661 1662 /* stuff used for matrix vector multiply */ 1663 a->lvec = 0; 1664 a->Mvctx = 0; 1665 a->roworiented = PETSC_TRUE; 1666 1667 ierr = PetscObjectComposeFunctionDynamic((PetscObject)mat,"MatGetDiagonalBlock_C", 1668 "MatGetDiagonalBlock_MPIDense", 1669 MatGetDiagonalBlock_MPIDense);CHKERRQ(ierr); 1670 ierr = PetscObjectComposeFunctionDynamic((PetscObject)mat,"MatMPIDenseSetPreallocation_C", 1671 "MatMPIDenseSetPreallocation_MPIDense", 1672 MatMPIDenseSetPreallocation_MPIDense);CHKERRQ(ierr); 1673 ierr = PetscObjectComposeFunctionDynamic((PetscObject)mat,"MatMatMult_mpiaij_mpidense_C", 1674 "MatMatMult_MPIAIJ_MPIDense", 1675 MatMatMult_MPIAIJ_MPIDense);CHKERRQ(ierr); 1676 ierr = PetscObjectComposeFunctionDynamic((PetscObject)mat,"MatMatMultSymbolic_mpiaij_mpidense_C", 1677 "MatMatMultSymbolic_MPIAIJ_MPIDense", 1678 MatMatMultSymbolic_MPIAIJ_MPIDense);CHKERRQ(ierr); 1679 ierr = PetscObjectComposeFunctionDynamic((PetscObject)mat,"MatMatMultNumeric_mpiaij_mpidense_C", 1680 "MatMatMultNumeric_MPIAIJ_MPIDense", 1681 MatMatMultNumeric_MPIAIJ_MPIDense);CHKERRQ(ierr); 1682 #if defined(PETSC_HAVE_PLAPACK) 1683 ierr = PetscObjectComposeFunctionDynamic((PetscObject)mat,"MatGetFactor_mpidense_plapack_C", 1684 "MatGetFactor_mpidense_plapack", 1685 MatGetFactor_mpidense_plapack);CHKERRQ(ierr); 1686 #if defined(PETSC_HAVE_PLAPACK) 1687 ierr = PetscPLAPACKInitializePackage(((PetscObject)mat)->comm);CHKERRQ(ierr); 1688 #endif 1689 1690 #endif 1691 ierr = PetscObjectChangeTypeName((PetscObject)mat,MATMPIDENSE);CHKERRQ(ierr); 1692 1693 PetscFunctionReturn(0); 1694 } 1695 EXTERN_C_END 1696 1697 /*MC 1698 MATDENSE - MATDENSE = "dense" - A matrix type to be used for dense matrices. 1699 1700 This matrix type is identical to MATSEQDENSE when constructed with a single process communicator, 1701 and MATMPIDENSE otherwise. 1702 1703 Options Database Keys: 1704 . -mat_type dense - sets the matrix type to "dense" during a call to MatSetFromOptions() 1705 1706 Level: beginner 1707 1708 1709 .seealso: MatCreateMPIDense,MATSEQDENSE,MATMPIDENSE 1710 M*/ 1711 1712 EXTERN_C_BEGIN 1713 #undef __FUNCT__ 1714 #define __FUNCT__ "MatCreate_Dense" 1715 PetscErrorCode PETSCMAT_DLLEXPORT MatCreate_Dense(Mat A) 1716 { 1717 PetscErrorCode ierr; 1718 PetscMPIInt size; 1719 1720 PetscFunctionBegin; 1721 ierr = MPI_Comm_size(((PetscObject)A)->comm,&size);CHKERRQ(ierr); 1722 if (size == 1) { 1723 ierr = MatSetType(A,MATSEQDENSE);CHKERRQ(ierr); 1724 } else { 1725 ierr = MatSetType(A,MATMPIDENSE);CHKERRQ(ierr); 1726 } 1727 PetscFunctionReturn(0); 1728 } 1729 EXTERN_C_END 1730 1731 #undef __FUNCT__ 1732 #define __FUNCT__ "MatMPIDenseSetPreallocation" 1733 /*@C 1734 MatMPIDenseSetPreallocation - Sets the array used to store the matrix entries 1735 1736 Not collective 1737 1738 Input Parameters: 1739 . A - the matrix 1740 - data - optional location of matrix data. Set data=PETSC_NULL for PETSc 1741 to control all matrix memory allocation. 1742 1743 Notes: 1744 The dense format is fully compatible with standard Fortran 77 1745 storage by columns. 1746 1747 The data input variable is intended primarily for Fortran programmers 1748 who wish to allocate their own matrix memory space. Most users should 1749 set data=PETSC_NULL. 1750 1751 Level: intermediate 1752 1753 .keywords: matrix,dense, parallel 1754 1755 .seealso: MatCreate(), MatCreateSeqDense(), MatSetValues() 1756 @*/ 1757 PetscErrorCode PETSCMAT_DLLEXPORT MatMPIDenseSetPreallocation(Mat mat,PetscScalar *data) 1758 { 1759 PetscErrorCode ierr,(*f)(Mat,PetscScalar *); 1760 1761 PetscFunctionBegin; 1762 ierr = PetscObjectQueryFunction((PetscObject)mat,"MatMPIDenseSetPreallocation_C",(void (**)(void))&f);CHKERRQ(ierr); 1763 if (f) { 1764 ierr = (*f)(mat,data);CHKERRQ(ierr); 1765 } 1766 PetscFunctionReturn(0); 1767 } 1768 1769 #undef __FUNCT__ 1770 #define __FUNCT__ "MatCreateMPIDense" 1771 /*@C 1772 MatCreateMPIDense - Creates a sparse parallel matrix in dense format. 1773 1774 Collective on MPI_Comm 1775 1776 Input Parameters: 1777 + comm - MPI communicator 1778 . m - number of local rows (or PETSC_DECIDE to have calculated if M is given) 1779 . n - number of local columns (or PETSC_DECIDE to have calculated if N is given) 1780 . M - number of global rows (or PETSC_DECIDE to have calculated if m is given) 1781 . N - number of global columns (or PETSC_DECIDE to have calculated if n is given) 1782 - data - optional location of matrix data. Set data=PETSC_NULL (PETSC_NULL_SCALAR for Fortran users) for PETSc 1783 to control all matrix memory allocation. 1784 1785 Output Parameter: 1786 . A - the matrix 1787 1788 Notes: 1789 The dense format is fully compatible with standard Fortran 77 1790 storage by columns. 1791 1792 The data input variable is intended primarily for Fortran programmers 1793 who wish to allocate their own matrix memory space. Most users should 1794 set data=PETSC_NULL (PETSC_NULL_SCALAR for Fortran users). 1795 1796 The user MUST specify either the local or global matrix dimensions 1797 (possibly both). 1798 1799 Level: intermediate 1800 1801 .keywords: matrix,dense, parallel 1802 1803 .seealso: MatCreate(), MatCreateSeqDense(), MatSetValues() 1804 @*/ 1805 PetscErrorCode PETSCMAT_DLLEXPORT MatCreateMPIDense(MPI_Comm comm,PetscInt m,PetscInt n,PetscInt M,PetscInt N,PetscScalar *data,Mat *A) 1806 { 1807 PetscErrorCode ierr; 1808 PetscMPIInt size; 1809 1810 PetscFunctionBegin; 1811 ierr = MatCreate(comm,A);CHKERRQ(ierr); 1812 ierr = MatSetSizes(*A,m,n,M,N);CHKERRQ(ierr); 1813 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr); 1814 if (size > 1) { 1815 ierr = MatSetType(*A,MATMPIDENSE);CHKERRQ(ierr); 1816 ierr = MatMPIDenseSetPreallocation(*A,data);CHKERRQ(ierr); 1817 } else { 1818 ierr = MatSetType(*A,MATSEQDENSE);CHKERRQ(ierr); 1819 ierr = MatSeqDenseSetPreallocation(*A,data);CHKERRQ(ierr); 1820 } 1821 PetscFunctionReturn(0); 1822 } 1823 1824 #undef __FUNCT__ 1825 #define __FUNCT__ "MatDuplicate_MPIDense" 1826 static PetscErrorCode MatDuplicate_MPIDense(Mat A,MatDuplicateOption cpvalues,Mat *newmat) 1827 { 1828 Mat mat; 1829 Mat_MPIDense *a,*oldmat = (Mat_MPIDense*)A->data; 1830 PetscErrorCode ierr; 1831 1832 PetscFunctionBegin; 1833 *newmat = 0; 1834 ierr = MatCreate(((PetscObject)A)->comm,&mat);CHKERRQ(ierr); 1835 ierr = MatSetSizes(mat,A->rmap->n,A->cmap->n,A->rmap->N,A->cmap->N);CHKERRQ(ierr); 1836 ierr = MatSetType(mat,((PetscObject)A)->type_name);CHKERRQ(ierr); 1837 a = (Mat_MPIDense*)mat->data; 1838 ierr = PetscMemcpy(mat->ops,A->ops,sizeof(struct _MatOps));CHKERRQ(ierr); 1839 mat->factor = A->factor; 1840 mat->assembled = PETSC_TRUE; 1841 mat->preallocated = PETSC_TRUE; 1842 1843 mat->rmap->rstart = A->rmap->rstart; 1844 mat->rmap->rend = A->rmap->rend; 1845 a->size = oldmat->size; 1846 a->rank = oldmat->rank; 1847 mat->insertmode = NOT_SET_VALUES; 1848 a->nvec = oldmat->nvec; 1849 a->donotstash = oldmat->donotstash; 1850 1851 ierr = PetscMemcpy(mat->rmap->range,A->rmap->range,(a->size+1)*sizeof(PetscInt));CHKERRQ(ierr); 1852 ierr = PetscMemcpy(mat->cmap->range,A->cmap->range,(a->size+1)*sizeof(PetscInt));CHKERRQ(ierr); 1853 ierr = MatStashCreate_Private(((PetscObject)A)->comm,1,&mat->stash);CHKERRQ(ierr); 1854 1855 ierr = MatSetUpMultiply_MPIDense(mat);CHKERRQ(ierr); 1856 ierr = MatDuplicate(oldmat->A,cpvalues,&a->A);CHKERRQ(ierr); 1857 ierr = PetscLogObjectParent(mat,a->A);CHKERRQ(ierr); 1858 1859 #if defined(PETSC_HAVE_PLAPACK) 1860 ierr = PetscMemcpy(mat->spptr,A->spptr,sizeof(Mat_Plapack));CHKERRQ(ierr); 1861 #endif 1862 *newmat = mat; 1863 PetscFunctionReturn(0); 1864 } 1865 1866 #include "petscsys.h" 1867 1868 #undef __FUNCT__ 1869 #define __FUNCT__ "MatLoad_MPIDense_DenseInFile" 1870 PetscErrorCode MatLoad_MPIDense_DenseInFile(MPI_Comm comm,PetscInt fd,PetscInt M,PetscInt N, const MatType type,Mat *newmat) 1871 { 1872 PetscErrorCode ierr; 1873 PetscMPIInt rank,size; 1874 PetscInt *rowners,i,m,nz,j; 1875 PetscScalar *array,*vals,*vals_ptr; 1876 MPI_Status status; 1877 1878 PetscFunctionBegin; 1879 ierr = MPI_Comm_rank(comm,&rank);CHKERRQ(ierr); 1880 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr); 1881 1882 /* determine ownership of all rows */ 1883 m = M/size + ((M % size) > rank); 1884 ierr = PetscMalloc((size+2)*sizeof(PetscInt),&rowners);CHKERRQ(ierr); 1885 ierr = MPI_Allgather(&m,1,MPIU_INT,rowners+1,1,MPIU_INT,comm);CHKERRQ(ierr); 1886 rowners[0] = 0; 1887 for (i=2; i<=size; i++) { 1888 rowners[i] += rowners[i-1]; 1889 } 1890 1891 ierr = MatCreate(comm,newmat);CHKERRQ(ierr); 1892 ierr = MatSetSizes(*newmat,m,PETSC_DECIDE,M,N);CHKERRQ(ierr); 1893 ierr = MatSetType(*newmat,type);CHKERRQ(ierr); 1894 ierr = MatMPIDenseSetPreallocation(*newmat,PETSC_NULL);CHKERRQ(ierr); 1895 ierr = MatGetArray(*newmat,&array);CHKERRQ(ierr); 1896 1897 if (!rank) { 1898 ierr = PetscMalloc(m*N*sizeof(PetscScalar),&vals);CHKERRQ(ierr); 1899 1900 /* read in my part of the matrix numerical values */ 1901 ierr = PetscBinaryRead(fd,vals,m*N,PETSC_SCALAR);CHKERRQ(ierr); 1902 1903 /* insert into matrix-by row (this is why cannot directly read into array */ 1904 vals_ptr = vals; 1905 for (i=0; i<m; i++) { 1906 for (j=0; j<N; j++) { 1907 array[i + j*m] = *vals_ptr++; 1908 } 1909 } 1910 1911 /* read in other processors and ship out */ 1912 for (i=1; i<size; i++) { 1913 nz = (rowners[i+1] - rowners[i])*N; 1914 ierr = PetscBinaryRead(fd,vals,nz,PETSC_SCALAR);CHKERRQ(ierr); 1915 ierr = MPI_Send(vals,nz,MPIU_SCALAR,i,((PetscObject)(*newmat))->tag,comm);CHKERRQ(ierr); 1916 } 1917 } else { 1918 /* receive numeric values */ 1919 ierr = PetscMalloc(m*N*sizeof(PetscScalar),&vals);CHKERRQ(ierr); 1920 1921 /* receive message of values*/ 1922 ierr = MPI_Recv(vals,m*N,MPIU_SCALAR,0,((PetscObject)(*newmat))->tag,comm,&status);CHKERRQ(ierr); 1923 1924 /* insert into matrix-by row (this is why cannot directly read into array */ 1925 vals_ptr = vals; 1926 for (i=0; i<m; i++) { 1927 for (j=0; j<N; j++) { 1928 array[i + j*m] = *vals_ptr++; 1929 } 1930 } 1931 } 1932 ierr = PetscFree(rowners);CHKERRQ(ierr); 1933 ierr = PetscFree(vals);CHKERRQ(ierr); 1934 ierr = MatAssemblyBegin(*newmat,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1935 ierr = MatAssemblyEnd(*newmat,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1936 PetscFunctionReturn(0); 1937 } 1938 1939 #undef __FUNCT__ 1940 #define __FUNCT__ "MatLoad_MPIDense" 1941 PetscErrorCode MatLoad_MPIDense(PetscViewer viewer,const MatType type,Mat *newmat) 1942 { 1943 Mat A; 1944 PetscScalar *vals,*svals; 1945 MPI_Comm comm = ((PetscObject)viewer)->comm; 1946 MPI_Status status; 1947 PetscMPIInt rank,size,tag = ((PetscObject)viewer)->tag,*rowners,*sndcounts,m,maxnz; 1948 PetscInt header[4],*rowlengths = 0,M,N,*cols; 1949 PetscInt *ourlens,*procsnz = 0,*offlens,jj,*mycols,*smycols; 1950 PetscInt i,nz,j,rstart,rend; 1951 int fd; 1952 PetscErrorCode ierr; 1953 1954 PetscFunctionBegin; 1955 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr); 1956 ierr = MPI_Comm_rank(comm,&rank);CHKERRQ(ierr); 1957 if (!rank) { 1958 ierr = PetscViewerBinaryGetDescriptor(viewer,&fd);CHKERRQ(ierr); 1959 ierr = PetscBinaryRead(fd,(char *)header,4,PETSC_INT);CHKERRQ(ierr); 1960 if (header[0] != MAT_FILE_COOKIE) SETERRQ(PETSC_ERR_FILE_UNEXPECTED,"not matrix object"); 1961 } 1962 1963 ierr = MPI_Bcast(header+1,3,MPIU_INT,0,comm);CHKERRQ(ierr); 1964 M = header[1]; N = header[2]; nz = header[3]; 1965 1966 /* 1967 Handle case where matrix is stored on disk as a dense matrix 1968 */ 1969 if (nz == MATRIX_BINARY_FORMAT_DENSE) { 1970 ierr = MatLoad_MPIDense_DenseInFile(comm,fd,M,N,type,newmat);CHKERRQ(ierr); 1971 PetscFunctionReturn(0); 1972 } 1973 1974 /* determine ownership of all rows */ 1975 m = PetscMPIIntCast(M/size + ((M % size) > rank)); 1976 ierr = PetscMalloc((size+2)*sizeof(PetscMPIInt),&rowners);CHKERRQ(ierr); 1977 ierr = MPI_Allgather(&m,1,MPI_INT,rowners+1,1,MPI_INT,comm);CHKERRQ(ierr); 1978 rowners[0] = 0; 1979 for (i=2; i<=size; i++) { 1980 rowners[i] += rowners[i-1]; 1981 } 1982 rstart = rowners[rank]; 1983 rend = rowners[rank+1]; 1984 1985 /* distribute row lengths to all processors */ 1986 ierr = PetscMalloc(2*(rend-rstart+1)*sizeof(PetscInt),&ourlens);CHKERRQ(ierr); 1987 offlens = ourlens + (rend-rstart); 1988 if (!rank) { 1989 ierr = PetscMalloc(M*sizeof(PetscInt),&rowlengths);CHKERRQ(ierr); 1990 ierr = PetscBinaryRead(fd,rowlengths,M,PETSC_INT);CHKERRQ(ierr); 1991 ierr = PetscMalloc(size*sizeof(PetscMPIInt),&sndcounts);CHKERRQ(ierr); 1992 for (i=0; i<size; i++) sndcounts[i] = rowners[i+1] - rowners[i]; 1993 ierr = MPI_Scatterv(rowlengths,sndcounts,rowners,MPIU_INT,ourlens,rend-rstart,MPIU_INT,0,comm);CHKERRQ(ierr); 1994 ierr = PetscFree(sndcounts);CHKERRQ(ierr); 1995 } else { 1996 ierr = MPI_Scatterv(0,0,0,MPIU_INT,ourlens,rend-rstart,MPIU_INT,0,comm);CHKERRQ(ierr); 1997 } 1998 1999 if (!rank) { 2000 /* calculate the number of nonzeros on each processor */ 2001 ierr = PetscMalloc(size*sizeof(PetscInt),&procsnz);CHKERRQ(ierr); 2002 ierr = PetscMemzero(procsnz,size*sizeof(PetscInt));CHKERRQ(ierr); 2003 for (i=0; i<size; i++) { 2004 for (j=rowners[i]; j< rowners[i+1]; j++) { 2005 procsnz[i] += rowlengths[j]; 2006 } 2007 } 2008 ierr = PetscFree(rowlengths);CHKERRQ(ierr); 2009 2010 /* determine max buffer needed and allocate it */ 2011 maxnz = 0; 2012 for (i=0; i<size; i++) { 2013 maxnz = PetscMax(maxnz,procsnz[i]); 2014 } 2015 ierr = PetscMalloc(maxnz*sizeof(PetscInt),&cols);CHKERRQ(ierr); 2016 2017 /* read in my part of the matrix column indices */ 2018 nz = procsnz[0]; 2019 ierr = PetscMalloc(nz*sizeof(PetscInt),&mycols);CHKERRQ(ierr); 2020 ierr = PetscBinaryRead(fd,mycols,nz,PETSC_INT);CHKERRQ(ierr); 2021 2022 /* read in every one elses and ship off */ 2023 for (i=1; i<size; i++) { 2024 nz = procsnz[i]; 2025 ierr = PetscBinaryRead(fd,cols,nz,PETSC_INT);CHKERRQ(ierr); 2026 ierr = MPI_Send(cols,nz,MPIU_INT,i,tag,comm);CHKERRQ(ierr); 2027 } 2028 ierr = PetscFree(cols);CHKERRQ(ierr); 2029 } else { 2030 /* determine buffer space needed for message */ 2031 nz = 0; 2032 for (i=0; i<m; i++) { 2033 nz += ourlens[i]; 2034 } 2035 ierr = PetscMalloc((nz+1)*sizeof(PetscInt),&mycols);CHKERRQ(ierr); 2036 2037 /* receive message of column indices*/ 2038 ierr = MPI_Recv(mycols,nz,MPIU_INT,0,tag,comm,&status);CHKERRQ(ierr); 2039 ierr = MPI_Get_count(&status,MPIU_INT,&maxnz);CHKERRQ(ierr); 2040 if (maxnz != nz) SETERRQ(PETSC_ERR_FILE_UNEXPECTED,"something is wrong with file"); 2041 } 2042 2043 /* loop over local rows, determining number of off diagonal entries */ 2044 ierr = PetscMemzero(offlens,m*sizeof(PetscInt));CHKERRQ(ierr); 2045 jj = 0; 2046 for (i=0; i<m; i++) { 2047 for (j=0; j<ourlens[i]; j++) { 2048 if (mycols[jj] < rstart || mycols[jj] >= rend) offlens[i]++; 2049 jj++; 2050 } 2051 } 2052 2053 /* create our matrix */ 2054 for (i=0; i<m; i++) { 2055 ourlens[i] -= offlens[i]; 2056 } 2057 ierr = MatCreate(comm,newmat);CHKERRQ(ierr); 2058 ierr = MatSetSizes(*newmat,m,PETSC_DECIDE,M,N);CHKERRQ(ierr); 2059 ierr = MatSetType(*newmat,type);CHKERRQ(ierr); 2060 ierr = MatMPIDenseSetPreallocation(*newmat,PETSC_NULL);CHKERRQ(ierr); 2061 A = *newmat; 2062 for (i=0; i<m; i++) { 2063 ourlens[i] += offlens[i]; 2064 } 2065 2066 if (!rank) { 2067 ierr = PetscMalloc(maxnz*sizeof(PetscScalar),&vals);CHKERRQ(ierr); 2068 2069 /* read in my part of the matrix numerical values */ 2070 nz = procsnz[0]; 2071 ierr = PetscBinaryRead(fd,vals,nz,PETSC_SCALAR);CHKERRQ(ierr); 2072 2073 /* insert into matrix */ 2074 jj = rstart; 2075 smycols = mycols; 2076 svals = vals; 2077 for (i=0; i<m; i++) { 2078 ierr = MatSetValues(A,1,&jj,ourlens[i],smycols,svals,INSERT_VALUES);CHKERRQ(ierr); 2079 smycols += ourlens[i]; 2080 svals += ourlens[i]; 2081 jj++; 2082 } 2083 2084 /* read in other processors and ship out */ 2085 for (i=1; i<size; i++) { 2086 nz = procsnz[i]; 2087 ierr = PetscBinaryRead(fd,vals,nz,PETSC_SCALAR);CHKERRQ(ierr); 2088 ierr = MPI_Send(vals,nz,MPIU_SCALAR,i,((PetscObject)A)->tag,comm);CHKERRQ(ierr); 2089 } 2090 ierr = PetscFree(procsnz);CHKERRQ(ierr); 2091 } else { 2092 /* receive numeric values */ 2093 ierr = PetscMalloc((nz+1)*sizeof(PetscScalar),&vals);CHKERRQ(ierr); 2094 2095 /* receive message of values*/ 2096 ierr = MPI_Recv(vals,nz,MPIU_SCALAR,0,((PetscObject)A)->tag,comm,&status);CHKERRQ(ierr); 2097 ierr = MPI_Get_count(&status,MPIU_SCALAR,&maxnz);CHKERRQ(ierr); 2098 if (maxnz != nz) SETERRQ(PETSC_ERR_FILE_UNEXPECTED,"something is wrong with file"); 2099 2100 /* insert into matrix */ 2101 jj = rstart; 2102 smycols = mycols; 2103 svals = vals; 2104 for (i=0; i<m; i++) { 2105 ierr = MatSetValues(A,1,&jj,ourlens[i],smycols,svals,INSERT_VALUES);CHKERRQ(ierr); 2106 smycols += ourlens[i]; 2107 svals += ourlens[i]; 2108 jj++; 2109 } 2110 } 2111 ierr = PetscFree(ourlens);CHKERRQ(ierr); 2112 ierr = PetscFree(vals);CHKERRQ(ierr); 2113 ierr = PetscFree(mycols);CHKERRQ(ierr); 2114 ierr = PetscFree(rowners);CHKERRQ(ierr); 2115 2116 ierr = MatAssemblyBegin(A,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 2117 ierr = MatAssemblyEnd(A,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 2118 PetscFunctionReturn(0); 2119 } 2120 2121 #undef __FUNCT__ 2122 #define __FUNCT__ "MatEqual_MPIDense" 2123 PetscErrorCode MatEqual_MPIDense(Mat A,Mat B,PetscTruth *flag) 2124 { 2125 Mat_MPIDense *matB = (Mat_MPIDense*)B->data,*matA = (Mat_MPIDense*)A->data; 2126 Mat a,b; 2127 PetscTruth flg; 2128 PetscErrorCode ierr; 2129 2130 PetscFunctionBegin; 2131 a = matA->A; 2132 b = matB->A; 2133 ierr = MatEqual(a,b,&flg);CHKERRQ(ierr); 2134 ierr = MPI_Allreduce(&flg,flag,1,MPI_INT,MPI_LAND,((PetscObject)A)->comm);CHKERRQ(ierr); 2135 PetscFunctionReturn(0); 2136 } 2137 2138 #if defined(PETSC_HAVE_PLAPACK) 2139 2140 #undef __FUNCT__ 2141 #define __FUNCT__ "PetscPLAPACKFinalizePackage" 2142 /*@C 2143 PetscPLAPACKFinalizePackage - This function destroys everything in the Petsc interface to PLAPACK. 2144 Level: developer 2145 2146 .keywords: Petsc, destroy, package, PLAPACK 2147 .seealso: PetscFinalize() 2148 @*/ 2149 PetscErrorCode PETSC_DLLEXPORT PetscPLAPACKFinalizePackage(void) 2150 { 2151 PetscErrorCode ierr; 2152 2153 PetscFunctionBegin; 2154 ierr = PLA_Finalize();CHKERRQ(ierr); 2155 PetscFunctionReturn(0); 2156 } 2157 2158 #undef __FUNCT__ 2159 #define __FUNCT__ "PetscPLAPACKInitializePackage" 2160 /*@C 2161 PetscPLAPACKInitializePackage - This function initializes everything in the Petsc interface to PLAPACK. It is 2162 called from MatCreate_MPIDense() the first time an MPI dense matrix is called. 2163 2164 Input Parameter: 2165 . comm - the communicator the matrix lives on 2166 2167 Level: developer 2168 2169 Notes: PLAPACK does not have a good fit with MPI communicators; all (parallel) PLAPACK objects have to live in the 2170 same communicator (because there is some global state that is initialized and used for all matrices). In addition if 2171 PLAPACK is initialized (that is the initial matrices created) are on subcommunicators of MPI_COMM_WORLD, these subcommunicators 2172 cannot overlap. 2173 2174 .keywords: Petsc, initialize, package, PLAPACK 2175 .seealso: PetscInitializePackage(), PetscInitialize() 2176 @*/ 2177 PetscErrorCode PETSC_DLLEXPORT PetscPLAPACKInitializePackage(MPI_Comm comm) 2178 { 2179 PetscMPIInt size; 2180 PetscErrorCode ierr; 2181 2182 PetscFunctionBegin; 2183 if (!PLA_Initialized(PETSC_NULL)) { 2184 2185 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr); 2186 Plapack_nprows = 1; 2187 Plapack_npcols = size; 2188 2189 ierr = PetscOptionsBegin(comm,PETSC_NULL,"PLAPACK Options","Mat");CHKERRQ(ierr); 2190 ierr = PetscOptionsInt("-plapack_nprows","row dimension of 2D processor mesh","None",Plapack_nprows,&Plapack_nprows,PETSC_NULL);CHKERRQ(ierr); 2191 ierr = PetscOptionsInt("-plapack_npcols","column dimension of 2D processor mesh","None",Plapack_npcols,&Plapack_npcols,PETSC_NULL);CHKERRQ(ierr); 2192 #if defined(PETSC_USE_DEBUG) 2193 Plapack_ierror = 3; 2194 #else 2195 Plapack_ierror = 0; 2196 #endif 2197 ierr = PetscOptionsInt("-plapack_ckerror","error checking flag","None",Plapack_ierror,&Plapack_ierror,PETSC_NULL);CHKERRQ(ierr); 2198 if (Plapack_ierror){ 2199 ierr = PLA_Set_error_checking(Plapack_ierror,PETSC_TRUE,PETSC_TRUE,PETSC_FALSE );CHKERRQ(ierr); 2200 } else { 2201 ierr = PLA_Set_error_checking(Plapack_ierror,PETSC_FALSE,PETSC_FALSE,PETSC_FALSE );CHKERRQ(ierr); 2202 } 2203 2204 Plapack_nb_alg = 0; 2205 ierr = PetscOptionsInt("-plapack_nb_alg","algorithmic block size","None",Plapack_nb_alg,&Plapack_nb_alg,PETSC_NULL);CHKERRQ(ierr); 2206 if (Plapack_nb_alg) { 2207 ierr = pla_Environ_set_nb_alg (PLA_OP_ALL_ALG,Plapack_nb_alg);CHKERRQ(ierr); 2208 } 2209 PetscOptionsEnd(); 2210 2211 ierr = PLA_Comm_1D_to_2D(comm,Plapack_nprows,Plapack_npcols,&Plapack_comm_2d);CHKERRQ(ierr); 2212 ierr = PLA_Init(Plapack_comm_2d);CHKERRQ(ierr); 2213 ierr = PetscRegisterFinalize(PetscPLAPACKFinalizePackage);CHKERRQ(ierr); 2214 } 2215 PetscFunctionReturn(0); 2216 } 2217 2218 2219 #endif 2220