1 2 /* 3 Defines matrix-matrix product routines for pairs of MPIAIJ matrices 4 C = A * B 5 */ 6 #include <../src/mat/impls/aij/seq/aij.h> /*I "petscmat.h" I*/ 7 #include <../src/mat/utils/freespace.h> 8 #include <../src/mat/impls/aij/mpi/mpiaij.h> 9 #include <petscbt.h> 10 #include <../src/mat/impls/dense/mpi/mpidense.h> 11 #include <petsc/private/vecimpl.h> 12 13 #if defined(PETSC_HAVE_HYPRE) 14 PETSC_INTERN PetscErrorCode MatMatMultSymbolic_AIJ_AIJ_wHYPRE(Mat,Mat,PetscReal,Mat*); 15 #endif 16 17 PETSC_INTERN PetscErrorCode MatMatMult_MPIAIJ_MPIAIJ(Mat A,Mat B,MatReuse scall,PetscReal fill, Mat *C) 18 { 19 PetscErrorCode ierr; 20 #if defined(PETSC_HAVE_HYPRE) 21 const char *algTypes[4] = {"scalable","nonscalable","seqmpi","hypre"}; 22 PetscInt nalg = 4; 23 #else 24 const char *algTypes[3] = {"scalable","nonscalable","seqmpi"}; 25 PetscInt nalg = 3; 26 #endif 27 PetscInt alg = 1; /* set nonscalable algorithm as default */ 28 MPI_Comm comm; 29 PetscBool flg; 30 31 PetscFunctionBegin; 32 if (scall == MAT_INITIAL_MATRIX) { 33 ierr = PetscObjectGetComm((PetscObject)A,&comm);CHKERRQ(ierr); 34 if (A->cmap->rstart != B->rmap->rstart || A->cmap->rend != B->rmap->rend) SETERRQ4(comm,PETSC_ERR_ARG_SIZ,"Matrix local dimensions are incompatible, (%D, %D) != (%D,%D)",A->cmap->rstart,A->cmap->rend,B->rmap->rstart,B->rmap->rend); 35 36 ierr = PetscObjectOptionsBegin((PetscObject)A);CHKERRQ(ierr); 37 PetscOptionsObject->alreadyprinted = PETSC_FALSE; /* a hack to ensure the option shows in '-help' */ 38 ierr = PetscOptionsEList("-matmatmult_via","Algorithmic approach","MatMatMult",algTypes,nalg,algTypes[1],&alg,&flg);CHKERRQ(ierr); 39 ierr = PetscOptionsEnd();CHKERRQ(ierr); 40 41 if (!flg && B->cmap->N > 100000) { /* may switch to scalable algorithm as default */ 42 MatInfo Ainfo,Binfo; 43 PetscInt nz_local; 44 PetscBool alg_scalable_loc=PETSC_FALSE,alg_scalable; 45 46 ierr = MatGetInfo(A,MAT_LOCAL,&Ainfo);CHKERRQ(ierr); 47 ierr = MatGetInfo(B,MAT_LOCAL,&Binfo);CHKERRQ(ierr); 48 nz_local = (PetscInt)(Ainfo.nz_allocated + Binfo.nz_allocated); 49 50 if (B->cmap->N > fill*nz_local) alg_scalable_loc = PETSC_TRUE; 51 ierr = MPIU_Allreduce(&alg_scalable_loc,&alg_scalable,1,MPIU_BOOL,MPI_LOR,comm);CHKERRQ(ierr); 52 53 if (alg_scalable) { 54 alg = 0; /* scalable algorithm would 50% slower than nonscalable algorithm */ 55 ierr = PetscInfo2(B,"Use scalable algorithm, BN %D, fill*nz_allocated %g\n",B->cmap->N,fill*nz_local);CHKERRQ(ierr); 56 } 57 } 58 59 ierr = PetscLogEventBegin(MAT_MatMultSymbolic,A,B,0,0);CHKERRQ(ierr); 60 switch (alg) { 61 case 1: 62 ierr = MatMatMultSymbolic_MPIAIJ_MPIAIJ_nonscalable(A,B,fill,C);CHKERRQ(ierr); 63 break; 64 case 2: 65 ierr = MatMatMultSymbolic_MPIAIJ_MPIAIJ_seqMPI(A,B,fill,C);CHKERRQ(ierr); 66 break; 67 #if defined(PETSC_HAVE_HYPRE) 68 case 3: 69 ierr = MatMatMultSymbolic_AIJ_AIJ_wHYPRE(A,B,fill,C);CHKERRQ(ierr); 70 break; 71 #endif 72 default: 73 ierr = MatMatMultSymbolic_MPIAIJ_MPIAIJ(A,B,fill,C);CHKERRQ(ierr); 74 break; 75 } 76 ierr = PetscLogEventEnd(MAT_MatMultSymbolic,A,B,0,0);CHKERRQ(ierr); 77 } 78 ierr = PetscLogEventBegin(MAT_MatMultNumeric,A,B,0,0);CHKERRQ(ierr); 79 ierr = (*(*C)->ops->matmultnumeric)(A,B,*C);CHKERRQ(ierr); 80 ierr = PetscLogEventEnd(MAT_MatMultNumeric,A,B,0,0);CHKERRQ(ierr); 81 PetscFunctionReturn(0); 82 } 83 84 PetscErrorCode MatDestroy_MPIAIJ_MatMatMult(Mat A) 85 { 86 PetscErrorCode ierr; 87 Mat_MPIAIJ *a = (Mat_MPIAIJ*)A->data; 88 Mat_PtAPMPI *ptap = a->ptap; 89 90 PetscFunctionBegin; 91 ierr = PetscFree2(ptap->startsj_s,ptap->startsj_r);CHKERRQ(ierr); 92 ierr = PetscFree(ptap->bufa);CHKERRQ(ierr); 93 ierr = MatDestroy(&ptap->P_loc);CHKERRQ(ierr); 94 ierr = MatDestroy(&ptap->P_oth);CHKERRQ(ierr); 95 ierr = MatDestroy(&ptap->Pt);CHKERRQ(ierr); 96 ierr = PetscFree(ptap->api);CHKERRQ(ierr); 97 ierr = PetscFree(ptap->apj);CHKERRQ(ierr); 98 ierr = PetscFree(ptap->apa);CHKERRQ(ierr); 99 ierr = ptap->destroy(A);CHKERRQ(ierr); 100 ierr = PetscFree(ptap);CHKERRQ(ierr); 101 PetscFunctionReturn(0); 102 } 103 104 PetscErrorCode MatDuplicate_MPIAIJ_MatMatMult(Mat A, MatDuplicateOption op, Mat *M) 105 { 106 PetscErrorCode ierr; 107 Mat_MPIAIJ *a = (Mat_MPIAIJ*)A->data; 108 Mat_PtAPMPI *ptap = a->ptap; 109 110 PetscFunctionBegin; 111 ierr = (*ptap->duplicate)(A,op,M);CHKERRQ(ierr); 112 113 (*M)->ops->destroy = ptap->destroy; /* = MatDestroy_MPIAIJ, *M doesn't duplicate A's special structure! */ 114 (*M)->ops->duplicate = ptap->duplicate; /* = MatDuplicate_MPIAIJ */ 115 PetscFunctionReturn(0); 116 } 117 118 PetscErrorCode MatMatMultNumeric_MPIAIJ_MPIAIJ_nonscalable(Mat A,Mat P,Mat C) 119 { 120 PetscErrorCode ierr; 121 Mat_MPIAIJ *a =(Mat_MPIAIJ*)A->data,*c=(Mat_MPIAIJ*)C->data; 122 Mat_SeqAIJ *ad =(Mat_SeqAIJ*)(a->A)->data,*ao=(Mat_SeqAIJ*)(a->B)->data; 123 Mat_SeqAIJ *cd =(Mat_SeqAIJ*)(c->A)->data,*co=(Mat_SeqAIJ*)(c->B)->data; 124 PetscScalar *cda=cd->a,*coa=co->a; 125 Mat_SeqAIJ *p_loc,*p_oth; 126 PetscScalar *apa,*ca; 127 PetscInt cm =C->rmap->n; 128 Mat_PtAPMPI *ptap=c->ptap; 129 PetscInt *api,*apj,*apJ,i,k; 130 PetscInt cstart=C->cmap->rstart; 131 PetscInt cdnz,conz,k0,k1; 132 MPI_Comm comm; 133 PetscMPIInt size; 134 135 PetscFunctionBegin; 136 ierr = PetscObjectGetComm((PetscObject)A,&comm);CHKERRQ(ierr); 137 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr); 138 139 /* 1) get P_oth = ptap->P_oth and P_loc = ptap->P_loc */ 140 /*-----------------------------------------------------*/ 141 /* update numerical values of P_oth and P_loc */ 142 ierr = MatGetBrowsOfAoCols_MPIAIJ(A,P,MAT_REUSE_MATRIX,&ptap->startsj_s,&ptap->startsj_r,&ptap->bufa,&ptap->P_oth);CHKERRQ(ierr); 143 ierr = MatMPIAIJGetLocalMat(P,MAT_REUSE_MATRIX,&ptap->P_loc);CHKERRQ(ierr); 144 145 /* 2) compute numeric C_loc = A_loc*P = Ad*P_loc + Ao*P_oth */ 146 /*----------------------------------------------------------*/ 147 /* get data from symbolic products */ 148 p_loc = (Mat_SeqAIJ*)(ptap->P_loc)->data; 149 p_oth = NULL; 150 if (size >1) { 151 p_oth = (Mat_SeqAIJ*)(ptap->P_oth)->data; 152 } 153 154 /* get apa for storing dense row A[i,:]*P */ 155 apa = ptap->apa; 156 157 api = ptap->api; 158 apj = ptap->apj; 159 for (i=0; i<cm; i++) { 160 /* compute apa = A[i,:]*P */ 161 AProw_nonscalable(i,ad,ao,p_loc,p_oth,apa); 162 163 /* set values in C */ 164 apJ = apj + api[i]; 165 cdnz = cd->i[i+1] - cd->i[i]; 166 conz = co->i[i+1] - co->i[i]; 167 168 /* 1st off-diagoanl part of C */ 169 ca = coa + co->i[i]; 170 k = 0; 171 for (k0=0; k0<conz; k0++) { 172 if (apJ[k] >= cstart) break; 173 ca[k0] = apa[apJ[k]]; 174 apa[apJ[k++]] = 0.0; 175 } 176 177 /* diagonal part of C */ 178 ca = cda + cd->i[i]; 179 for (k1=0; k1<cdnz; k1++) { 180 ca[k1] = apa[apJ[k]]; 181 apa[apJ[k++]] = 0.0; 182 } 183 184 /* 2nd off-diagoanl part of C */ 185 ca = coa + co->i[i]; 186 for (; k0<conz; k0++) { 187 ca[k0] = apa[apJ[k]]; 188 apa[apJ[k++]] = 0.0; 189 } 190 } 191 ierr = MatAssemblyBegin(C,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 192 ierr = MatAssemblyEnd(C,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 193 PetscFunctionReturn(0); 194 } 195 196 PetscErrorCode MatMatMultSymbolic_MPIAIJ_MPIAIJ_nonscalable(Mat A,Mat P,PetscReal fill,Mat *C) 197 { 198 PetscErrorCode ierr; 199 MPI_Comm comm; 200 PetscMPIInt size; 201 Mat Cmpi; 202 Mat_PtAPMPI *ptap; 203 PetscFreeSpaceList free_space=NULL,current_space=NULL; 204 Mat_MPIAIJ *a =(Mat_MPIAIJ*)A->data,*c; 205 Mat_SeqAIJ *ad =(Mat_SeqAIJ*)(a->A)->data,*ao=(Mat_SeqAIJ*)(a->B)->data,*p_loc,*p_oth; 206 PetscInt *pi_loc,*pj_loc,*pi_oth,*pj_oth,*dnz,*onz; 207 PetscInt *adi=ad->i,*adj=ad->j,*aoi=ao->i,*aoj=ao->j,rstart=A->rmap->rstart; 208 PetscInt *lnk,i,pnz,row,*api,*apj,*Jptr,apnz,nspacedouble=0,j,nzi; 209 PetscInt am=A->rmap->n,pN=P->cmap->N,pn=P->cmap->n,pm=P->rmap->n; 210 PetscBT lnkbt; 211 PetscScalar *apa; 212 PetscReal afill; 213 214 PetscFunctionBegin; 215 ierr = PetscObjectGetComm((PetscObject)A,&comm);CHKERRQ(ierr); 216 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr); 217 218 /* create struct Mat_PtAPMPI and attached it to C later */ 219 ierr = PetscNew(&ptap);CHKERRQ(ierr); 220 221 /* get P_oth by taking rows of P (= non-zero cols of local A) from other processors */ 222 ierr = MatGetBrowsOfAoCols_MPIAIJ(A,P,MAT_INITIAL_MATRIX,&ptap->startsj_s,&ptap->startsj_r,&ptap->bufa,&ptap->P_oth);CHKERRQ(ierr); 223 224 /* get P_loc by taking all local rows of P */ 225 ierr = MatMPIAIJGetLocalMat(P,MAT_INITIAL_MATRIX,&ptap->P_loc);CHKERRQ(ierr); 226 227 p_loc = (Mat_SeqAIJ*)(ptap->P_loc)->data; 228 pi_loc = p_loc->i; pj_loc = p_loc->j; 229 if (size > 1) { 230 p_oth = (Mat_SeqAIJ*)(ptap->P_oth)->data; 231 pi_oth = p_oth->i; pj_oth = p_oth->j; 232 } else { 233 p_oth = NULL; 234 pi_oth = NULL; pj_oth = NULL; 235 } 236 237 /* first, compute symbolic AP = A_loc*P = A_diag*P_loc + A_off*P_oth */ 238 /*-------------------------------------------------------------------*/ 239 ierr = PetscMalloc1(am+2,&api);CHKERRQ(ierr); 240 ptap->api = api; 241 api[0] = 0; 242 243 /* create and initialize a linked list */ 244 ierr = PetscLLCondensedCreate(pN,pN,&lnk,&lnkbt);CHKERRQ(ierr); 245 246 /* Initial FreeSpace size is fill*(nnz(A)+nnz(P)) */ 247 ierr = PetscFreeSpaceGet(PetscRealIntMultTruncate(fill,PetscIntSumTruncate(adi[am],PetscIntSumTruncate(aoi[am],pi_loc[pm]))),&free_space);CHKERRQ(ierr); 248 current_space = free_space; 249 250 ierr = MatPreallocateInitialize(comm,am,pn,dnz,onz);CHKERRQ(ierr); 251 for (i=0; i<am; i++) { 252 /* diagonal portion of A */ 253 nzi = adi[i+1] - adi[i]; 254 for (j=0; j<nzi; j++) { 255 row = *adj++; 256 pnz = pi_loc[row+1] - pi_loc[row]; 257 Jptr = pj_loc + pi_loc[row]; 258 /* add non-zero cols of P into the sorted linked list lnk */ 259 ierr = PetscLLCondensedAddSorted(pnz,Jptr,lnk,lnkbt);CHKERRQ(ierr); 260 } 261 /* off-diagonal portion of A */ 262 nzi = aoi[i+1] - aoi[i]; 263 for (j=0; j<nzi; j++) { 264 row = *aoj++; 265 pnz = pi_oth[row+1] - pi_oth[row]; 266 Jptr = pj_oth + pi_oth[row]; 267 ierr = PetscLLCondensedAddSorted(pnz,Jptr,lnk,lnkbt);CHKERRQ(ierr); 268 } 269 270 apnz = lnk[0]; 271 api[i+1] = api[i] + apnz; 272 273 /* if free space is not available, double the total space in the list */ 274 if (current_space->local_remaining<apnz) { 275 ierr = PetscFreeSpaceGet(PetscIntSumTruncate(apnz,current_space->total_array_size),¤t_space);CHKERRQ(ierr); 276 nspacedouble++; 277 } 278 279 /* Copy data into free space, then initialize lnk */ 280 ierr = PetscLLCondensedClean(pN,apnz,current_space->array,lnk,lnkbt);CHKERRQ(ierr); 281 ierr = MatPreallocateSet(i+rstart,apnz,current_space->array,dnz,onz);CHKERRQ(ierr); 282 283 current_space->array += apnz; 284 current_space->local_used += apnz; 285 current_space->local_remaining -= apnz; 286 } 287 288 /* Allocate space for apj, initialize apj, and */ 289 /* destroy list of free space and other temporary array(s) */ 290 ierr = PetscMalloc1(api[am]+1,&ptap->apj);CHKERRQ(ierr); 291 apj = ptap->apj; 292 ierr = PetscFreeSpaceContiguous(&free_space,ptap->apj);CHKERRQ(ierr); 293 ierr = PetscLLDestroy(lnk,lnkbt);CHKERRQ(ierr); 294 295 /* malloc apa to store dense row A[i,:]*P */ 296 ierr = PetscCalloc1(pN,&apa);CHKERRQ(ierr); 297 298 ptap->apa = apa; 299 300 /* create and assemble symbolic parallel matrix Cmpi */ 301 /*----------------------------------------------------*/ 302 ierr = MatCreate(comm,&Cmpi);CHKERRQ(ierr); 303 ierr = MatSetSizes(Cmpi,am,pn,PETSC_DETERMINE,PETSC_DETERMINE);CHKERRQ(ierr); 304 ierr = MatSetBlockSizesFromMats(Cmpi,A,P);CHKERRQ(ierr); 305 306 ierr = MatSetType(Cmpi,MATMPIAIJ);CHKERRQ(ierr); 307 ierr = MatMPIAIJSetPreallocation(Cmpi,0,dnz,0,onz);CHKERRQ(ierr); 308 309 ierr = MatSetValues_MPIAIJ_CopyFromCSRFormat_Symbolic(Cmpi, apj, api, dnz, onz);CHKERRQ(ierr); 310 ierr = MatAssemblyBegin(Cmpi,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 311 ierr = MatAssemblyEnd(Cmpi,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 312 ierr = MatPreallocateFinalize(dnz,onz);CHKERRQ(ierr); 313 314 ptap->destroy = Cmpi->ops->destroy; 315 ptap->duplicate = Cmpi->ops->duplicate; 316 Cmpi->ops->matmultnumeric = MatMatMultNumeric_MPIAIJ_MPIAIJ_nonscalable; 317 Cmpi->ops->destroy = MatDestroy_MPIAIJ_MatMatMult; 318 Cmpi->ops->duplicate = MatDuplicate_MPIAIJ_MatMatMult; 319 320 /* attach the supporting struct to Cmpi for reuse */ 321 c = (Mat_MPIAIJ*)Cmpi->data; 322 c->ptap = ptap; 323 324 *C = Cmpi; 325 326 /* set MatInfo */ 327 afill = (PetscReal)api[am]/(adi[am]+aoi[am]+pi_loc[pm]+1) + 1.e-5; 328 if (afill < 1.0) afill = 1.0; 329 Cmpi->info.mallocs = nspacedouble; 330 Cmpi->info.fill_ratio_given = fill; 331 Cmpi->info.fill_ratio_needed = afill; 332 333 #if defined(PETSC_USE_INFO) 334 if (api[am]) { 335 ierr = PetscInfo3(Cmpi,"Reallocs %D; Fill ratio: given %g needed %g.\n",nspacedouble,(double)fill,(double)afill);CHKERRQ(ierr); 336 ierr = PetscInfo1(Cmpi,"Use MatMatMult(A,B,MatReuse,%g,&C) for best performance.;\n",(double)afill);CHKERRQ(ierr); 337 } else { 338 ierr = PetscInfo(Cmpi,"Empty matrix product\n");CHKERRQ(ierr); 339 } 340 #endif 341 PetscFunctionReturn(0); 342 } 343 344 PETSC_INTERN PetscErrorCode MatMatMult_MPIAIJ_MPIDense(Mat A,Mat B,MatReuse scall,PetscReal fill,Mat *C) 345 { 346 PetscErrorCode ierr; 347 348 PetscFunctionBegin; 349 if (scall == MAT_INITIAL_MATRIX) { 350 ierr = PetscLogEventBegin(MAT_MatMultSymbolic,A,B,0,0);CHKERRQ(ierr); 351 ierr = MatMatMultSymbolic_MPIAIJ_MPIDense(A,B,fill,C);CHKERRQ(ierr); 352 ierr = PetscLogEventEnd(MAT_MatMultSymbolic,A,B,0,0);CHKERRQ(ierr); 353 } 354 ierr = PetscLogEventBegin(MAT_MatMultNumeric,A,B,0,0);CHKERRQ(ierr); 355 ierr = MatMatMultNumeric_MPIAIJ_MPIDense(A,B,*C);CHKERRQ(ierr); 356 ierr = PetscLogEventEnd(MAT_MatMultNumeric,A,B,0,0);CHKERRQ(ierr); 357 PetscFunctionReturn(0); 358 } 359 360 typedef struct { 361 Mat workB; 362 PetscScalar *rvalues,*svalues; 363 MPI_Request *rwaits,*swaits; 364 } MPIAIJ_MPIDense; 365 366 PetscErrorCode MatMPIAIJ_MPIDenseDestroy(void *ctx) 367 { 368 MPIAIJ_MPIDense *contents = (MPIAIJ_MPIDense*) ctx; 369 PetscErrorCode ierr; 370 371 PetscFunctionBegin; 372 ierr = MatDestroy(&contents->workB);CHKERRQ(ierr); 373 ierr = PetscFree4(contents->rvalues,contents->svalues,contents->rwaits,contents->swaits);CHKERRQ(ierr); 374 ierr = PetscFree(contents);CHKERRQ(ierr); 375 PetscFunctionReturn(0); 376 } 377 378 /* 379 This is a "dummy function" that handles the case where matrix C was created as a dense matrix 380 directly by the user and passed to MatMatMult() with the MAT_REUSE_MATRIX option 381 382 It is the same as MatMatMultSymbolic_MPIAIJ_MPIDense() except does not create C 383 */ 384 PetscErrorCode MatMatMultNumeric_MPIDense(Mat A,Mat B,Mat C) 385 { 386 PetscErrorCode ierr; 387 PetscBool flg; 388 Mat_MPIAIJ *aij = (Mat_MPIAIJ*) A->data; 389 PetscInt nz = aij->B->cmap->n; 390 PetscContainer container; 391 MPIAIJ_MPIDense *contents; 392 VecScatter ctx = aij->Mvctx; 393 VecScatter_MPI_General *from = (VecScatter_MPI_General*) ctx->fromdata; 394 VecScatter_MPI_General *to = (VecScatter_MPI_General*) ctx->todata; 395 396 PetscFunctionBegin; 397 ierr = PetscObjectTypeCompare((PetscObject)B,MATMPIDENSE,&flg);CHKERRQ(ierr); 398 if (!flg) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Second matrix must be mpidense"); 399 400 /* Handle case where where user provided the final C matrix rather than calling MatMatMult() with MAT_INITIAL_MATRIX*/ 401 ierr = PetscObjectTypeCompare((PetscObject)A,MATMPIAIJ,&flg);CHKERRQ(ierr); 402 if (!flg) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"First matrix must be MPIAIJ"); 403 404 C->ops->matmultnumeric = MatMatMultNumeric_MPIAIJ_MPIDense; 405 406 ierr = PetscNew(&contents);CHKERRQ(ierr); 407 /* Create work matrix used to store off processor rows of B needed for local product */ 408 ierr = MatCreateSeqDense(PETSC_COMM_SELF,nz,B->cmap->N,NULL,&contents->workB);CHKERRQ(ierr); 409 /* Create work arrays needed */ 410 ierr = PetscMalloc4(B->cmap->N*from->starts[from->n],&contents->rvalues, 411 B->cmap->N*to->starts[to->n],&contents->svalues, 412 from->n,&contents->rwaits, 413 to->n,&contents->swaits);CHKERRQ(ierr); 414 415 ierr = PetscContainerCreate(PetscObjectComm((PetscObject)A),&container);CHKERRQ(ierr); 416 ierr = PetscContainerSetPointer(container,contents);CHKERRQ(ierr); 417 ierr = PetscContainerSetUserDestroy(container,MatMPIAIJ_MPIDenseDestroy);CHKERRQ(ierr); 418 ierr = PetscObjectCompose((PetscObject)C,"workB",(PetscObject)container);CHKERRQ(ierr); 419 ierr = PetscContainerDestroy(&container);CHKERRQ(ierr); 420 421 ierr = (*C->ops->matmultnumeric)(A,B,C);CHKERRQ(ierr); 422 PetscFunctionReturn(0); 423 } 424 425 PetscErrorCode MatMatMultSymbolic_MPIAIJ_MPIDense(Mat A,Mat B,PetscReal fill,Mat *C) 426 { 427 PetscErrorCode ierr; 428 Mat_MPIAIJ *aij = (Mat_MPIAIJ*) A->data; 429 PetscInt nz = aij->B->cmap->n; 430 PetscContainer container; 431 MPIAIJ_MPIDense *contents; 432 VecScatter ctx = aij->Mvctx; 433 VecScatter_MPI_General *from = (VecScatter_MPI_General*) ctx->fromdata; 434 VecScatter_MPI_General *to = (VecScatter_MPI_General*) ctx->todata; 435 PetscInt m = A->rmap->n,n=B->cmap->n; 436 437 PetscFunctionBegin; 438 ierr = MatCreate(PetscObjectComm((PetscObject)B),C);CHKERRQ(ierr); 439 ierr = MatSetSizes(*C,m,n,A->rmap->N,B->cmap->N);CHKERRQ(ierr); 440 ierr = MatSetBlockSizesFromMats(*C,A,B);CHKERRQ(ierr); 441 ierr = MatSetType(*C,MATMPIDENSE);CHKERRQ(ierr); 442 ierr = MatMPIDenseSetPreallocation(*C,NULL);CHKERRQ(ierr); 443 ierr = MatAssemblyBegin(*C,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 444 ierr = MatAssemblyEnd(*C,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 445 446 (*C)->ops->matmultnumeric = MatMatMultNumeric_MPIAIJ_MPIDense; 447 448 ierr = PetscNew(&contents);CHKERRQ(ierr); 449 /* Create work matrix used to store off processor rows of B needed for local product */ 450 ierr = MatCreateSeqDense(PETSC_COMM_SELF,nz,B->cmap->N,NULL,&contents->workB);CHKERRQ(ierr); 451 /* Create work arrays needed */ 452 ierr = PetscMalloc4(B->cmap->N*from->starts[from->n],&contents->rvalues, 453 B->cmap->N*to->starts[to->n],&contents->svalues, 454 from->n,&contents->rwaits, 455 to->n,&contents->swaits);CHKERRQ(ierr); 456 457 ierr = PetscContainerCreate(PetscObjectComm((PetscObject)A),&container);CHKERRQ(ierr); 458 ierr = PetscContainerSetPointer(container,contents);CHKERRQ(ierr); 459 ierr = PetscContainerSetUserDestroy(container,MatMPIAIJ_MPIDenseDestroy);CHKERRQ(ierr); 460 ierr = PetscObjectCompose((PetscObject)(*C),"workB",(PetscObject)container);CHKERRQ(ierr); 461 ierr = PetscContainerDestroy(&container);CHKERRQ(ierr); 462 PetscFunctionReturn(0); 463 } 464 465 /* 466 Performs an efficient scatter on the rows of B needed by this process; this is 467 a modification of the VecScatterBegin_() routines. 468 */ 469 PetscErrorCode MatMPIDenseScatter(Mat A,Mat B,Mat C,Mat *outworkB) 470 { 471 Mat_MPIAIJ *aij = (Mat_MPIAIJ*)A->data; 472 PetscErrorCode ierr; 473 PetscScalar *b,*w,*svalues,*rvalues; 474 VecScatter ctx = aij->Mvctx; 475 VecScatter_MPI_General *from = (VecScatter_MPI_General*) ctx->fromdata; 476 VecScatter_MPI_General *to = (VecScatter_MPI_General*) ctx->todata; 477 PetscInt i,j,k; 478 PetscInt *sindices,*sstarts,*rindices,*rstarts; 479 PetscMPIInt *sprocs,*rprocs,nrecvs; 480 MPI_Request *swaits,*rwaits; 481 MPI_Comm comm; 482 PetscMPIInt tag = ((PetscObject)ctx)->tag,ncols = B->cmap->N, nrows = aij->B->cmap->n,imdex,nrowsB = B->rmap->n; 483 MPI_Status status; 484 MPIAIJ_MPIDense *contents; 485 PetscContainer container; 486 Mat workB; 487 488 PetscFunctionBegin; 489 ierr = PetscObjectGetComm((PetscObject)A,&comm);CHKERRQ(ierr); 490 ierr = PetscObjectQuery((PetscObject)C,"workB",(PetscObject*)&container);CHKERRQ(ierr); 491 if (!container) SETERRQ(comm,PETSC_ERR_PLIB,"Container does not exist"); 492 ierr = PetscContainerGetPointer(container,(void**)&contents);CHKERRQ(ierr); 493 494 workB = *outworkB = contents->workB; 495 if (nrows != workB->rmap->n) SETERRQ2(comm,PETSC_ERR_PLIB,"Number of rows of workB %D not equal to columns of aij->B %D",nrows,workB->cmap->n); 496 sindices = to->indices; 497 sstarts = to->starts; 498 sprocs = to->procs; 499 swaits = contents->swaits; 500 svalues = contents->svalues; 501 502 rindices = from->indices; 503 rstarts = from->starts; 504 rprocs = from->procs; 505 rwaits = contents->rwaits; 506 rvalues = contents->rvalues; 507 508 ierr = MatDenseGetArray(B,&b);CHKERRQ(ierr); 509 ierr = MatDenseGetArray(workB,&w);CHKERRQ(ierr); 510 511 for (i=0; i<from->n; i++) { 512 ierr = MPI_Irecv(rvalues+ncols*rstarts[i],ncols*(rstarts[i+1]-rstarts[i]),MPIU_SCALAR,rprocs[i],tag,comm,rwaits+i);CHKERRQ(ierr); 513 } 514 515 for (i=0; i<to->n; i++) { 516 /* pack a message at a time */ 517 for (j=0; j<sstarts[i+1]-sstarts[i]; j++) { 518 for (k=0; k<ncols; k++) { 519 svalues[ncols*(sstarts[i] + j) + k] = b[sindices[sstarts[i]+j] + nrowsB*k]; 520 } 521 } 522 ierr = MPI_Isend(svalues+ncols*sstarts[i],ncols*(sstarts[i+1]-sstarts[i]),MPIU_SCALAR,sprocs[i],tag,comm,swaits+i);CHKERRQ(ierr); 523 } 524 525 nrecvs = from->n; 526 while (nrecvs) { 527 ierr = MPI_Waitany(from->n,rwaits,&imdex,&status);CHKERRQ(ierr); 528 nrecvs--; 529 /* unpack a message at a time */ 530 for (j=0; j<rstarts[imdex+1]-rstarts[imdex]; j++) { 531 for (k=0; k<ncols; k++) { 532 w[rindices[rstarts[imdex]+j] + nrows*k] = rvalues[ncols*(rstarts[imdex] + j) + k]; 533 } 534 } 535 } 536 if (to->n) {ierr = MPI_Waitall(to->n,swaits,to->sstatus);CHKERRQ(ierr);} 537 538 ierr = MatDenseRestoreArray(B,&b);CHKERRQ(ierr); 539 ierr = MatDenseRestoreArray(workB,&w);CHKERRQ(ierr); 540 ierr = MatAssemblyBegin(workB,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 541 ierr = MatAssemblyEnd(workB,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 542 PetscFunctionReturn(0); 543 } 544 extern PetscErrorCode MatMatMultNumericAdd_SeqAIJ_SeqDense(Mat,Mat,Mat); 545 546 PetscErrorCode MatMatMultNumeric_MPIAIJ_MPIDense(Mat A,Mat B,Mat C) 547 { 548 PetscErrorCode ierr; 549 Mat_MPIAIJ *aij = (Mat_MPIAIJ*)A->data; 550 Mat_MPIDense *bdense = (Mat_MPIDense*)B->data; 551 Mat_MPIDense *cdense = (Mat_MPIDense*)C->data; 552 Mat workB; 553 554 PetscFunctionBegin; 555 /* diagonal block of A times all local rows of B*/ 556 ierr = MatMatMultNumeric_SeqAIJ_SeqDense(aij->A,bdense->A,cdense->A);CHKERRQ(ierr); 557 558 /* get off processor parts of B needed to complete the product */ 559 ierr = MatMPIDenseScatter(A,B,C,&workB);CHKERRQ(ierr); 560 561 /* off-diagonal block of A times nonlocal rows of B */ 562 ierr = MatMatMultNumericAdd_SeqAIJ_SeqDense(aij->B,workB,cdense->A);CHKERRQ(ierr); 563 ierr = MatAssemblyBegin(C,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 564 ierr = MatAssemblyEnd(C,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 565 PetscFunctionReturn(0); 566 } 567 568 PetscErrorCode MatMatMultNumeric_MPIAIJ_MPIAIJ(Mat A,Mat P,Mat C) 569 { 570 PetscErrorCode ierr; 571 Mat_MPIAIJ *a = (Mat_MPIAIJ*)A->data,*c=(Mat_MPIAIJ*)C->data; 572 Mat_SeqAIJ *ad = (Mat_SeqAIJ*)(a->A)->data,*ao=(Mat_SeqAIJ*)(a->B)->data; 573 Mat_SeqAIJ *cd = (Mat_SeqAIJ*)(c->A)->data,*co=(Mat_SeqAIJ*)(c->B)->data; 574 PetscInt *adi = ad->i,*adj,*aoi=ao->i,*aoj; 575 PetscScalar *ada,*aoa,*cda=cd->a,*coa=co->a; 576 Mat_SeqAIJ *p_loc,*p_oth; 577 PetscInt *pi_loc,*pj_loc,*pi_oth,*pj_oth,*pj; 578 PetscScalar *pa_loc,*pa_oth,*pa,valtmp,*ca; 579 PetscInt cm = C->rmap->n,anz,pnz; 580 Mat_PtAPMPI *ptap = c->ptap; 581 PetscScalar *apa_sparse = ptap->apa; 582 PetscInt *api,*apj,*apJ,i,j,k,row; 583 PetscInt cstart = C->cmap->rstart; 584 PetscInt cdnz,conz,k0,k1,nextp; 585 MPI_Comm comm; 586 PetscMPIInt size; 587 588 PetscFunctionBegin; 589 ierr = PetscObjectGetComm((PetscObject)A,&comm);CHKERRQ(ierr); 590 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr); 591 592 /* 1) get P_oth = ptap->P_oth and P_loc = ptap->P_loc */ 593 /*-----------------------------------------------------*/ 594 /* update numerical values of P_oth and P_loc */ 595 ierr = MatGetBrowsOfAoCols_MPIAIJ(A,P,MAT_REUSE_MATRIX,&ptap->startsj_s,&ptap->startsj_r,&ptap->bufa,&ptap->P_oth);CHKERRQ(ierr); 596 ierr = MatMPIAIJGetLocalMat(P,MAT_REUSE_MATRIX,&ptap->P_loc);CHKERRQ(ierr); 597 598 /* 2) compute numeric C_loc = A_loc*P = Ad*P_loc + Ao*P_oth */ 599 /*----------------------------------------------------------*/ 600 /* get data from symbolic products */ 601 p_loc = (Mat_SeqAIJ*)(ptap->P_loc)->data; 602 pi_loc = p_loc->i; pj_loc = p_loc->j; pa_loc = p_loc->a; 603 if (size >1) { 604 p_oth = (Mat_SeqAIJ*)(ptap->P_oth)->data; 605 pi_oth = p_oth->i; pj_oth = p_oth->j; pa_oth = p_oth->a; 606 } else { 607 p_oth = NULL; pi_oth = NULL; pj_oth = NULL; pa_oth = NULL; 608 } 609 610 api = ptap->api; 611 apj = ptap->apj; 612 for (i=0; i<cm; i++) { 613 apJ = apj + api[i]; 614 615 /* diagonal portion of A */ 616 anz = adi[i+1] - adi[i]; 617 adj = ad->j + adi[i]; 618 ada = ad->a + adi[i]; 619 for (j=0; j<anz; j++) { 620 row = adj[j]; 621 pnz = pi_loc[row+1] - pi_loc[row]; 622 pj = pj_loc + pi_loc[row]; 623 pa = pa_loc + pi_loc[row]; 624 /* perform sparse axpy */ 625 valtmp = ada[j]; 626 nextp = 0; 627 for (k=0; nextp<pnz; k++) { 628 if (apJ[k] == pj[nextp]) { /* column of AP == column of P */ 629 apa_sparse[k] += valtmp*pa[nextp++]; 630 } 631 } 632 ierr = PetscLogFlops(2.0*pnz);CHKERRQ(ierr); 633 } 634 635 /* off-diagonal portion of A */ 636 anz = aoi[i+1] - aoi[i]; 637 aoj = ao->j + aoi[i]; 638 aoa = ao->a + aoi[i]; 639 for (j=0; j<anz; j++) { 640 row = aoj[j]; 641 pnz = pi_oth[row+1] - pi_oth[row]; 642 pj = pj_oth + pi_oth[row]; 643 pa = pa_oth + pi_oth[row]; 644 /* perform sparse axpy */ 645 valtmp = aoa[j]; 646 nextp = 0; 647 for (k=0; nextp<pnz; k++) { 648 if (apJ[k] == pj[nextp]) { /* column of AP == column of P */ 649 apa_sparse[k] += valtmp*pa[nextp++]; 650 } 651 } 652 ierr = PetscLogFlops(2.0*pnz);CHKERRQ(ierr); 653 } 654 655 /* set values in C */ 656 cdnz = cd->i[i+1] - cd->i[i]; 657 conz = co->i[i+1] - co->i[i]; 658 659 /* 1st off-diagoanl part of C */ 660 ca = coa + co->i[i]; 661 k = 0; 662 for (k0=0; k0<conz; k0++) { 663 if (apJ[k] >= cstart) break; 664 ca[k0] = apa_sparse[k]; 665 apa_sparse[k] = 0.0; 666 k++; 667 } 668 669 /* diagonal part of C */ 670 ca = cda + cd->i[i]; 671 for (k1=0; k1<cdnz; k1++) { 672 ca[k1] = apa_sparse[k]; 673 apa_sparse[k] = 0.0; 674 k++; 675 } 676 677 /* 2nd off-diagoanl part of C */ 678 ca = coa + co->i[i]; 679 for (; k0<conz; k0++) { 680 ca[k0] = apa_sparse[k]; 681 apa_sparse[k] = 0.0; 682 k++; 683 } 684 } 685 ierr = MatAssemblyBegin(C,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 686 ierr = MatAssemblyEnd(C,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 687 PetscFunctionReturn(0); 688 } 689 690 /* same as MatMatMultSymbolic_MPIAIJ_MPIAIJ_nonscalable(), except using LLCondensed to avoid O(BN) memory requirement */ 691 PetscErrorCode MatMatMultSymbolic_MPIAIJ_MPIAIJ(Mat A,Mat P,PetscReal fill,Mat *C) 692 { 693 PetscErrorCode ierr; 694 MPI_Comm comm; 695 PetscMPIInt size; 696 Mat Cmpi; 697 Mat_PtAPMPI *ptap; 698 PetscFreeSpaceList free_space = NULL,current_space=NULL; 699 Mat_MPIAIJ *a = (Mat_MPIAIJ*)A->data,*c; 700 Mat_SeqAIJ *ad = (Mat_SeqAIJ*)(a->A)->data,*ao=(Mat_SeqAIJ*)(a->B)->data,*p_loc,*p_oth; 701 PetscInt *pi_loc,*pj_loc,*pi_oth,*pj_oth,*dnz,*onz; 702 PetscInt *adi=ad->i,*adj=ad->j,*aoi=ao->i,*aoj=ao->j,rstart=A->rmap->rstart; 703 PetscInt i,pnz,row,*api,*apj,*Jptr,apnz,nspacedouble=0,j,nzi,*lnk,apnz_max; 704 PetscInt am=A->rmap->n,pN=P->cmap->N,pn=P->cmap->n,pm=P->rmap->n; 705 PetscReal afill; 706 PetscScalar *apa; 707 PetscTable ta; 708 709 PetscFunctionBegin; 710 ierr = PetscObjectGetComm((PetscObject)A,&comm);CHKERRQ(ierr); 711 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr); 712 713 /* create struct Mat_PtAPMPI and attached it to C later */ 714 ierr = PetscNew(&ptap);CHKERRQ(ierr); 715 716 /* get P_oth by taking rows of P (= non-zero cols of local A) from other processors */ 717 ierr = MatGetBrowsOfAoCols_MPIAIJ(A,P,MAT_INITIAL_MATRIX,&ptap->startsj_s,&ptap->startsj_r,&ptap->bufa,&ptap->P_oth);CHKERRQ(ierr); 718 719 /* get P_loc by taking all local rows of P */ 720 ierr = MatMPIAIJGetLocalMat(P,MAT_INITIAL_MATRIX,&ptap->P_loc);CHKERRQ(ierr); 721 722 p_loc = (Mat_SeqAIJ*)(ptap->P_loc)->data; 723 pi_loc = p_loc->i; pj_loc = p_loc->j; 724 if (size > 1) { 725 p_oth = (Mat_SeqAIJ*)(ptap->P_oth)->data; 726 pi_oth = p_oth->i; pj_oth = p_oth->j; 727 } else { 728 p_oth = NULL; 729 pi_oth = NULL; pj_oth = NULL; 730 } 731 732 /* first, compute symbolic AP = A_loc*P = A_diag*P_loc + A_off*P_oth */ 733 /*-------------------------------------------------------------------*/ 734 ierr = PetscMalloc1(am+2,&api);CHKERRQ(ierr); 735 ptap->api = api; 736 api[0] = 0; 737 738 /* create and initialize a linked list */ 739 ierr = PetscTableCreate(pn,pN,&ta);CHKERRQ(ierr); 740 741 /* Calculate apnz_max */ 742 apnz_max = 0; 743 for (i=0; i<am; i++) { 744 ierr = PetscTableRemoveAll(ta);CHKERRQ(ierr); 745 /* diagonal portion of A */ 746 nzi = adi[i+1] - adi[i]; 747 Jptr = adj+adi[i]; /* cols of A_diag */ 748 MatMergeRows_SeqAIJ(p_loc,nzi,Jptr,ta); 749 ierr = PetscTableGetCount(ta,&apnz);CHKERRQ(ierr); 750 if (apnz_max < apnz) apnz_max = apnz; 751 752 /* off-diagonal portion of A */ 753 nzi = aoi[i+1] - aoi[i]; 754 Jptr = aoj+aoi[i]; /* cols of A_off */ 755 MatMergeRows_SeqAIJ(p_oth,nzi,Jptr,ta); 756 ierr = PetscTableGetCount(ta,&apnz);CHKERRQ(ierr); 757 if (apnz_max < apnz) apnz_max = apnz; 758 } 759 ierr = PetscTableDestroy(&ta);CHKERRQ(ierr); 760 ierr = PetscLLCondensedCreate_Scalable(apnz_max,&lnk);CHKERRQ(ierr); 761 762 /* Initial FreeSpace size is fill*(nnz(A)+nnz(P)) */ 763 ierr = PetscFreeSpaceGet(PetscRealIntMultTruncate(fill,PetscIntSumTruncate(adi[am],PetscIntSumTruncate(aoi[am],pi_loc[pm]))),&free_space);CHKERRQ(ierr); 764 current_space = free_space; 765 ierr = MatPreallocateInitialize(comm,am,pn,dnz,onz);CHKERRQ(ierr); 766 for (i=0; i<am; i++) { 767 /* diagonal portion of A */ 768 nzi = adi[i+1] - adi[i]; 769 for (j=0; j<nzi; j++) { 770 row = *adj++; 771 pnz = pi_loc[row+1] - pi_loc[row]; 772 Jptr = pj_loc + pi_loc[row]; 773 /* add non-zero cols of P into the sorted linked list lnk */ 774 ierr = PetscLLCondensedAddSorted_Scalable(pnz,Jptr,lnk);CHKERRQ(ierr); 775 } 776 /* off-diagonal portion of A */ 777 nzi = aoi[i+1] - aoi[i]; 778 for (j=0; j<nzi; j++) { 779 row = *aoj++; 780 pnz = pi_oth[row+1] - pi_oth[row]; 781 Jptr = pj_oth + pi_oth[row]; 782 ierr = PetscLLCondensedAddSorted_Scalable(pnz,Jptr,lnk);CHKERRQ(ierr); 783 } 784 apnz = *lnk; 785 api[i+1] = api[i] + apnz; 786 787 /* if free space is not available, double the total space in the list */ 788 if (current_space->local_remaining<apnz) { 789 ierr = PetscFreeSpaceGet(PetscIntSumTruncate(apnz,current_space->total_array_size),¤t_space);CHKERRQ(ierr); 790 nspacedouble++; 791 } 792 793 /* Copy data into free space, then initialize lnk */ 794 ierr = PetscLLCondensedClean_Scalable(apnz,current_space->array,lnk);CHKERRQ(ierr); 795 ierr = MatPreallocateSet(i+rstart,apnz,current_space->array,dnz,onz);CHKERRQ(ierr); 796 797 current_space->array += apnz; 798 current_space->local_used += apnz; 799 current_space->local_remaining -= apnz; 800 } 801 802 /* Allocate space for apj, initialize apj, and */ 803 /* destroy list of free space and other temporary array(s) */ 804 ierr = PetscMalloc1(api[am]+1,&ptap->apj);CHKERRQ(ierr); 805 apj = ptap->apj; 806 ierr = PetscFreeSpaceContiguous(&free_space,ptap->apj);CHKERRQ(ierr); 807 ierr = PetscLLCondensedDestroy_Scalable(lnk);CHKERRQ(ierr); 808 809 /* create and assemble symbolic parallel matrix Cmpi */ 810 /*----------------------------------------------------*/ 811 ierr = MatCreate(comm,&Cmpi);CHKERRQ(ierr); 812 ierr = MatSetSizes(Cmpi,am,pn,PETSC_DETERMINE,PETSC_DETERMINE);CHKERRQ(ierr); 813 ierr = MatSetBlockSizesFromMats(Cmpi,A,P);CHKERRQ(ierr); 814 ierr = MatSetType(Cmpi,MATMPIAIJ);CHKERRQ(ierr); 815 ierr = MatMPIAIJSetPreallocation(Cmpi,0,dnz,0,onz);CHKERRQ(ierr); 816 817 /* malloc apa for assembly Cmpi */ 818 ierr = PetscCalloc1(apnz_max,&apa);CHKERRQ(ierr); 819 ptap->apa = apa; 820 821 ierr = MatSetValues_MPIAIJ_CopyFromCSRFormat_Symbolic(Cmpi, apj, api, dnz, onz);CHKERRQ(ierr); 822 ierr = MatAssemblyBegin(Cmpi,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 823 ierr = MatAssemblyEnd(Cmpi,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 824 ierr = MatPreallocateFinalize(dnz,onz);CHKERRQ(ierr); 825 826 ptap->destroy = Cmpi->ops->destroy; 827 ptap->duplicate = Cmpi->ops->duplicate; 828 Cmpi->ops->matmultnumeric = MatMatMultNumeric_MPIAIJ_MPIAIJ; 829 Cmpi->ops->destroy = MatDestroy_MPIAIJ_MatMatMult; 830 Cmpi->ops->duplicate = MatDuplicate_MPIAIJ_MatMatMult; 831 832 /* attach the supporting struct to Cmpi for reuse */ 833 c = (Mat_MPIAIJ*)Cmpi->data; 834 c->ptap = ptap; 835 836 *C = Cmpi; 837 838 /* set MatInfo */ 839 afill = (PetscReal)api[am]/(adi[am]+aoi[am]+pi_loc[pm]+1) + 1.e-5; 840 if (afill < 1.0) afill = 1.0; 841 Cmpi->info.mallocs = nspacedouble; 842 Cmpi->info.fill_ratio_given = fill; 843 Cmpi->info.fill_ratio_needed = afill; 844 845 #if defined(PETSC_USE_INFO) 846 if (api[am]) { 847 ierr = PetscInfo3(Cmpi,"Reallocs %D; Fill ratio: given %g needed %g.\n",nspacedouble,(double)fill,(double)afill);CHKERRQ(ierr); 848 ierr = PetscInfo1(Cmpi,"Use MatMatMult(A,B,MatReuse,%g,&C) for best performance.;\n",(double)afill);CHKERRQ(ierr); 849 } else { 850 ierr = PetscInfo(Cmpi,"Empty matrix product\n");CHKERRQ(ierr); 851 } 852 #endif 853 PetscFunctionReturn(0); 854 } 855 856 /* This function is needed for the seqMPI matrix-matrix multiplication. */ 857 /* Three input arrays are merged to one output array. The size of the */ 858 /* output array is also output. Duplicate entries only show up once. */ 859 static void Merge3SortedArrays(PetscInt size1, PetscInt *in1, 860 PetscInt size2, PetscInt *in2, 861 PetscInt size3, PetscInt *in3, 862 PetscInt *size4, PetscInt *out) 863 { 864 int i = 0, j = 0, k = 0, l = 0; 865 866 /* Traverse all three arrays */ 867 while (i<size1 && j<size2 && k<size3) { 868 if (in1[i] < in2[j] && in1[i] < in3[k]) { 869 out[l++] = in1[i++]; 870 } 871 else if(in2[j] < in1[i] && in2[j] < in3[k]) { 872 out[l++] = in2[j++]; 873 } 874 else if(in3[k] < in1[i] && in3[k] < in2[j]) { 875 out[l++] = in3[k++]; 876 } 877 else if(in1[i] == in2[j] && in1[i] < in3[k]) { 878 out[l++] = in1[i]; 879 i++, j++; 880 } 881 else if(in1[i] == in3[k] && in1[i] < in2[j]) { 882 out[l++] = in1[i]; 883 i++, k++; 884 } 885 else if(in3[k] == in2[j] && in2[j] < in1[i]) { 886 out[l++] = in2[j]; 887 k++, j++; 888 } 889 else if(in1[i] == in2[j] && in1[i] == in3[k]) { 890 out[l++] = in1[i]; 891 i++, j++, k++; 892 } 893 } 894 895 /* Traverse two remaining arrays */ 896 while (i<size1 && j<size2) { 897 if (in1[i] < in2[j]) { 898 out[l++] = in1[i++]; 899 } 900 else if(in1[i] > in2[j]) { 901 out[l++] = in2[j++]; 902 } 903 else { 904 out[l++] = in1[i]; 905 i++, j++; 906 } 907 } 908 909 while (i<size1 && k<size3) { 910 if (in1[i] < in3[k]) { 911 out[l++] = in1[i++]; 912 } 913 else if(in1[i] > in3[k]) { 914 out[l++] = in3[k++]; 915 } 916 else { 917 out[l++] = in1[i]; 918 i++, k++; 919 } 920 } 921 922 while (k<size3 && j<size2) { 923 if (in3[k] < in2[j]) { 924 out[l++] = in3[k++]; 925 } 926 else if(in3[k] > in2[j]) { 927 out[l++] = in2[j++]; 928 } 929 else { 930 out[l++] = in3[k]; 931 k++, j++; 932 } 933 } 934 935 /* Traverse one remaining array */ 936 while (i<size1) out[l++] = in1[i++]; 937 while (j<size2) out[l++] = in2[j++]; 938 while (k<size3) out[l++] = in3[k++]; 939 940 *size4 = l; 941 } 942 943 /* This matrix-matrix multiplication algorithm divides the multiplication into three multiplications and */ 944 /* adds up the products. Two of these three multiplications are performed with existing (sequential) */ 945 /* matrix-matrix multiplications. */ 946 #undef __FUNCT__ 947 #define __FUNCT__ "MatMatMultSymbolic_MPIAIJ_MPIAIJ_seqMPI" 948 PetscErrorCode MatMatMultSymbolic_MPIAIJ_MPIAIJ_seqMPI(Mat A, Mat P, PetscReal fill, Mat *C) 949 { 950 PetscErrorCode ierr; 951 MPI_Comm comm; 952 PetscMPIInt size; 953 Mat Cmpi; 954 Mat_PtAPMPI *ptap; 955 PetscFreeSpaceList free_space_diag=NULL, current_space=NULL; 956 Mat_MPIAIJ *a =(Mat_MPIAIJ*)A->data; 957 Mat_SeqAIJ *ad =(Mat_SeqAIJ*)(a->A)->data,*ao=(Mat_SeqAIJ*)(a->B)->data,*p_loc; 958 Mat_MPIAIJ *p =(Mat_MPIAIJ*)P->data; 959 Mat_MPIAIJ *c; 960 Mat_SeqAIJ *adpd_seq, *p_off, *aopoth_seq; 961 PetscInt adponz, adpdnz; 962 PetscInt *pi_loc,*dnz,*onz; 963 PetscInt *adi=ad->i,*adj=ad->j,*aoi=ao->i,rstart=A->rmap->rstart; 964 PetscInt *lnk,i, i1=0,pnz,row,*adpoi,*adpoj, *api, *adpoJ, *aopJ, *apJ,*Jptr, aopnz, nspacedouble=0,j,nzi, 965 *apj,apnz, *adpdi, *adpdj, *adpdJ, *poff_i, *poff_j, *j_temp, *aopothi, *aopothj; 966 PetscInt am=A->rmap->n,pN=P->cmap->N,pn=P->cmap->n,pm=P->rmap->n, p_colstart, p_colend; 967 PetscBT lnkbt; 968 PetscScalar *apa; 969 PetscReal afill; 970 PetscMPIInt rank; 971 Mat adpd, aopoth; 972 973 PetscFunctionBegin; 974 ierr = PetscObjectGetComm((PetscObject)A,&comm);CHKERRQ(ierr); 975 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr); 976 ierr = MPI_Comm_rank(comm, &rank);CHKERRQ(ierr); 977 ierr = MatGetOwnershipRangeColumn(P, &p_colstart, &p_colend); CHKERRQ(ierr); 978 979 /* create struct Mat_PtAPMPI and attached it to C later */ 980 ierr = PetscNew(&ptap);CHKERRQ(ierr); 981 982 /* get P_oth by taking rows of P (= non-zero cols of local A) from other processors */ 983 ierr = MatGetBrowsOfAoCols_MPIAIJ(A,P,MAT_INITIAL_MATRIX,&ptap->startsj_s,&ptap->startsj_r,&ptap->bufa,&ptap->P_oth);CHKERRQ(ierr); 984 985 /* get P_loc by taking all local rows of P */ 986 ierr = MatMPIAIJGetLocalMat(P,MAT_INITIAL_MATRIX,&ptap->P_loc);CHKERRQ(ierr); 987 988 989 p_loc = (Mat_SeqAIJ*)(ptap->P_loc)->data; 990 pi_loc = p_loc->i; 991 992 /* Allocate memory for the i arrays of the matrices A*P, A_diag*P_off and A_offd * P */ 993 ierr = PetscMalloc1(am+2,&api);CHKERRQ(ierr); 994 ierr = PetscMalloc1(am+2,&adpoi);CHKERRQ(ierr); 995 996 adpoi[0] = 0; 997 ptap->api = api; 998 api[0] = 0; 999 1000 /* create and initialize a linked list, will be used for both A_diag * P_loc_off and A_offd * P_oth */ 1001 ierr = PetscLLCondensedCreate(pN,pN,&lnk,&lnkbt);CHKERRQ(ierr); 1002 ierr = MatPreallocateInitialize(comm,am,pn,dnz,onz);CHKERRQ(ierr); 1003 1004 /* Symbolic calc of A_loc_diag * P_loc_diag */ 1005 ierr = MatMatMultSymbolic_SeqAIJ_SeqAIJ(a->A, p->A, fill, &adpd);CHKERRQ(ierr); 1006 adpd_seq = (Mat_SeqAIJ*)((adpd)->data); 1007 adpdi = adpd_seq->i; adpdj = adpd_seq->j; 1008 p_off = (Mat_SeqAIJ*)((p->B)->data); 1009 poff_i = p_off->i; poff_j = p_off->j; 1010 1011 /* j_temp stores indices of a result row before they are added to the linked list */ 1012 ierr = PetscMalloc1(pN+2,&j_temp);CHKERRQ(ierr); 1013 1014 1015 /* Symbolic calc of the A_diag * p_loc_off */ 1016 /* Initial FreeSpace size is fill*(nnz(A)+nnz(P)) */ 1017 ierr = PetscFreeSpaceGet(PetscRealIntMultTruncate(fill,PetscIntSumTruncate(adi[am],PetscIntSumTruncate(aoi[am],pi_loc[pm]))),&free_space_diag);CHKERRQ(ierr); 1018 current_space = free_space_diag; 1019 1020 for (i=0; i<am; i++) { 1021 /* A_diag * P_loc_off */ 1022 nzi = adi[i+1] - adi[i]; 1023 for (j=0; j<nzi; j++) { 1024 row = *adj++; 1025 pnz = poff_i[row+1] - poff_i[row]; 1026 Jptr = poff_j + poff_i[row]; 1027 for(i1 = 0; i1 < pnz; i1++) { 1028 j_temp[i1] = p->garray[Jptr[i1]]; 1029 } 1030 /* add non-zero cols of P into the sorted linked list lnk */ 1031 ierr = PetscLLCondensedAddSorted(pnz,j_temp,lnk,lnkbt);CHKERRQ(ierr); 1032 } 1033 1034 adponz = lnk[0]; 1035 adpoi[i+1] = adpoi[i] + adponz; 1036 1037 /* if free space is not available, double the total space in the list */ 1038 if (current_space->local_remaining<adponz) { 1039 ierr = PetscFreeSpaceGet(PetscIntSumTruncate(adponz,current_space->total_array_size),¤t_space);CHKERRQ(ierr); 1040 nspacedouble++; 1041 } 1042 1043 /* Copy data into free space, then initialize lnk */ 1044 ierr = PetscLLCondensedClean(pN,adponz,current_space->array,lnk,lnkbt);CHKERRQ(ierr); 1045 1046 current_space->array += adponz; 1047 current_space->local_used += adponz; 1048 current_space->local_remaining -= adponz; 1049 } 1050 1051 /* Symbolic calc of A_off * P_oth */ 1052 ierr = MatMatMultSymbolic_SeqAIJ_SeqAIJ(a->B, ptap->P_oth, fill, &aopoth);CHKERRQ(ierr); 1053 aopoth_seq = (Mat_SeqAIJ*)((aopoth)->data); 1054 aopothi = aopoth_seq->i; aopothj = aopoth_seq->j; 1055 1056 /* Allocate space for apj, adpj, aopj, ... */ 1057 /* destroy lists of free space and other temporary array(s) */ 1058 1059 ierr = PetscMalloc1(aopothi[am] + adpoi[am] + adpdi[am]+2, &ptap->apj);CHKERRQ(ierr); 1060 ierr = PetscMalloc1(adpoi[am]+2, &adpoj);CHKERRQ(ierr); 1061 1062 /* Copy from linked list to j-array */ 1063 ierr = PetscFreeSpaceContiguous(&free_space_diag,adpoj);CHKERRQ(ierr); 1064 ierr = PetscLLDestroy(lnk,lnkbt);CHKERRQ(ierr); 1065 1066 adpoJ = adpoj; 1067 adpdJ = adpdj; 1068 aopJ = aopothj; 1069 apj = ptap->apj; 1070 apJ = apj; /* still empty */ 1071 1072 /* Merge j-arrays of A_off * P, A_diag * P_loc_off, and */ 1073 /* A_diag * P_loc_diag to get A*P */ 1074 for (i = 0; i < am; i++) { 1075 aopnz = aopothi[i+1] - aopothi[i]; 1076 adponz = adpoi[i+1] - adpoi[i]; 1077 adpdnz = adpdi[i+1] - adpdi[i]; 1078 1079 /* Correct indices from A_diag*P_diag */ 1080 for(i1 = 0; i1 < adpdnz; i1++) { 1081 adpdJ[i1] += p_colstart; 1082 } 1083 /* Merge j-arrays of A_diag * P_loc_off and A_diag * P_loc_diag and A_off * P_oth */ 1084 Merge3SortedArrays(adponz, adpoJ, adpdnz, adpdJ, aopnz, aopJ, &apnz, apJ); 1085 ierr = MatPreallocateSet(i+rstart, apnz, apJ, dnz, onz); CHKERRQ(ierr); 1086 1087 aopJ += aopnz; 1088 adpoJ += adponz; 1089 adpdJ += adpdnz; 1090 apJ += apnz; 1091 api[i+1] = api[i] + apnz; 1092 } 1093 1094 /* malloc apa to store dense row A[i,:]*P */ 1095 ierr = PetscCalloc1(pN+2,&apa);CHKERRQ(ierr); 1096 1097 ptap->apa = apa; 1098 /* create and assemble symbolic parallel matrix Cmpi */ 1099 ierr = MatCreate(comm,&Cmpi);CHKERRQ(ierr); 1100 ierr = MatSetSizes(Cmpi,am,pn,PETSC_DETERMINE,PETSC_DETERMINE);CHKERRQ(ierr); 1101 ierr = MatSetBlockSizesFromMats(Cmpi,A,P);CHKERRQ(ierr); 1102 1103 ierr = MatSetType(Cmpi,MATMPIAIJ);CHKERRQ(ierr); 1104 ierr = MatMPIAIJSetPreallocation(Cmpi,0,dnz,0,onz);CHKERRQ(ierr); 1105 1106 1107 ierr = MatSetValues_MPIAIJ_CopyFromCSRFormat_Symbolic(Cmpi, apj, api, dnz, onz);CHKERRQ(ierr); 1108 ierr = MatAssemblyBegin(Cmpi,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1109 ierr = MatAssemblyEnd(Cmpi,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1110 ierr = MatPreallocateFinalize(dnz,onz);CHKERRQ(ierr); 1111 1112 1113 ptap->destroy = Cmpi->ops->destroy; 1114 ptap->duplicate = Cmpi->ops->duplicate; 1115 Cmpi->ops->matmultnumeric = MatMatMultNumeric_MPIAIJ_MPIAIJ_nonscalable; 1116 Cmpi->ops->destroy = MatDestroy_MPIAIJ_MatMatMult; 1117 Cmpi->ops->duplicate = MatDuplicate_MPIAIJ_MatMatMult; 1118 1119 /* attach the supporting struct to Cmpi for reuse */ 1120 c = (Mat_MPIAIJ*)Cmpi->data; 1121 c->ptap = ptap; 1122 *C = Cmpi; 1123 1124 /* set MatInfo */ 1125 afill = (PetscReal)api[am]/(adi[am]+aoi[am]+pi_loc[pm]+1) + 1.e-5; 1126 if (afill < 1.0) afill = 1.0; 1127 Cmpi->info.mallocs = nspacedouble; 1128 Cmpi->info.fill_ratio_given = fill; 1129 Cmpi->info.fill_ratio_needed = afill; 1130 1131 #if defined(PETSC_USE_INFO) 1132 if (api[am]) { 1133 ierr = PetscInfo3(Cmpi,"Reallocs %D; Fill ratio: given %g needed %g.\n",nspacedouble,(double)fill,(double)afill);CHKERRQ(ierr); 1134 ierr = PetscInfo1(Cmpi,"Use MatMatMult(A,B,MatReuse,%g,&C) for best performance.;\n",(double)afill);CHKERRQ(ierr); 1135 } else { 1136 ierr = PetscInfo(Cmpi,"Empty matrix product\n");CHKERRQ(ierr); 1137 } 1138 #endif 1139 1140 ierr = MatDestroy(&aopoth);CHKERRQ(ierr); 1141 ierr = MatDestroy(&adpd);CHKERRQ(ierr); 1142 ierr = PetscFree(j_temp);CHKERRQ(ierr); 1143 ierr = PetscFree(adpoj);CHKERRQ(ierr); 1144 ierr = PetscFree(adpoi);CHKERRQ(ierr); 1145 PetscFunctionReturn(0); 1146 } 1147 1148 1149 /*-------------------------------------------------------------------------*/ 1150 PetscErrorCode MatTransposeMatMult_MPIAIJ_MPIAIJ(Mat P,Mat A,MatReuse scall,PetscReal fill,Mat *C) 1151 { 1152 PetscErrorCode ierr; 1153 const char *algTypes[3] = {"scalable","nonscalable","matmatmult"}; 1154 PetscInt aN=A->cmap->N,alg=1; /* set default algorithm */ 1155 PetscBool flg; 1156 1157 PetscFunctionBegin; 1158 if (scall == MAT_INITIAL_MATRIX) { 1159 ierr = PetscObjectOptionsBegin((PetscObject)A);CHKERRQ(ierr); 1160 PetscOptionsObject->alreadyprinted = PETSC_FALSE; /* a hack to ensure the option shows in '-help' */ 1161 ierr = PetscOptionsEList("-mattransposematmult_via","Algorithmic approach","MatTransposeMatMult",algTypes,3,algTypes[1],&alg,&flg);CHKERRQ(ierr); 1162 ierr = PetscOptionsEnd();CHKERRQ(ierr); 1163 1164 ierr = PetscLogEventBegin(MAT_TransposeMatMultSymbolic,P,A,0,0);CHKERRQ(ierr); 1165 switch (alg) { 1166 case 1: 1167 if (!flg && aN > 100000) { /* may switch to scalable algorithm as default */ 1168 MatInfo Ainfo,Pinfo; 1169 PetscInt nz_local; 1170 PetscBool alg_scalable_loc=PETSC_FALSE,alg_scalable; 1171 MPI_Comm comm; 1172 1173 ierr = MatGetInfo(A,MAT_LOCAL,&Ainfo);CHKERRQ(ierr); 1174 ierr = MatGetInfo(P,MAT_LOCAL,&Pinfo);CHKERRQ(ierr); 1175 nz_local = (PetscInt)(Ainfo.nz_allocated + Pinfo.nz_allocated); /* estimated local nonzero entries */ 1176 1177 if (aN > fill*nz_local) alg_scalable_loc = PETSC_TRUE; 1178 ierr = PetscObjectGetComm((PetscObject)A,&comm);CHKERRQ(ierr); 1179 ierr = MPIU_Allreduce(&alg_scalable_loc,&alg_scalable,1,MPIU_BOOL,MPI_LOR,comm);CHKERRQ(ierr); 1180 1181 if (alg_scalable) { 1182 alg = 0; /* scalable algorithm would slower than nonscalable algorithm */ 1183 ierr = MatTransposeMatMultSymbolic_MPIAIJ_MPIAIJ(P,A,fill,C);CHKERRQ(ierr); 1184 break; 1185 } 1186 } 1187 ierr = MatTransposeMatMultSymbolic_MPIAIJ_MPIAIJ_nonscalable(P,A,fill,C);CHKERRQ(ierr); 1188 break; 1189 case 2: 1190 { 1191 Mat Pt; 1192 Mat_PtAPMPI *ptap; 1193 Mat_MPIAIJ *c; 1194 ierr = MatTranspose(P,MAT_INITIAL_MATRIX,&Pt);CHKERRQ(ierr); 1195 ierr = MatMatMult(Pt,A,MAT_INITIAL_MATRIX,fill,C);CHKERRQ(ierr); 1196 c = (Mat_MPIAIJ*)(*C)->data; 1197 ptap = c->ptap; 1198 ptap->Pt = Pt; 1199 (*C)->ops->mattransposemultnumeric = MatTransposeMatMultNumeric_MPIAIJ_MPIAIJ_matmatmult; 1200 PetscFunctionReturn(0); 1201 } 1202 break; 1203 default: /* scalable algorithm */ 1204 ierr = MatTransposeMatMultSymbolic_MPIAIJ_MPIAIJ(P,A,fill,C);CHKERRQ(ierr); 1205 break; 1206 } 1207 ierr = PetscLogEventEnd(MAT_TransposeMatMultSymbolic,P,A,0,0);CHKERRQ(ierr); 1208 } 1209 ierr = PetscLogEventBegin(MAT_TransposeMatMultNumeric,P,A,0,0);CHKERRQ(ierr); 1210 ierr = (*(*C)->ops->mattransposemultnumeric)(P,A,*C);CHKERRQ(ierr); 1211 ierr = PetscLogEventEnd(MAT_TransposeMatMultNumeric,P,A,0,0);CHKERRQ(ierr); 1212 PetscFunctionReturn(0); 1213 } 1214 1215 /* This routine only works when scall=MAT_REUSE_MATRIX! */ 1216 PetscErrorCode MatTransposeMatMultNumeric_MPIAIJ_MPIAIJ_matmatmult(Mat P,Mat A,Mat C) 1217 { 1218 PetscErrorCode ierr; 1219 Mat_MPIAIJ *c=(Mat_MPIAIJ*)C->data; 1220 Mat_PtAPMPI *ptap= c->ptap; 1221 Mat Pt=ptap->Pt; 1222 1223 PetscFunctionBegin; 1224 ierr = MatTranspose(P,MAT_REUSE_MATRIX,&Pt);CHKERRQ(ierr); 1225 ierr = MatMatMultNumeric(Pt,A,C);CHKERRQ(ierr); 1226 PetscFunctionReturn(0); 1227 } 1228 1229 PetscErrorCode MatDuplicate_MPIAIJ_MatPtAP(Mat,MatDuplicateOption,Mat*); 1230 1231 /* This routine is modified from MatPtAPSymbolic_MPIAIJ_MPIAIJ() */ 1232 PetscErrorCode MatTransposeMatMultSymbolic_MPIAIJ_MPIAIJ_nonscalable(Mat P,Mat A,PetscReal fill,Mat *C) 1233 { 1234 PetscErrorCode ierr; 1235 Mat_PtAPMPI *ptap; 1236 Mat_MPIAIJ *p=(Mat_MPIAIJ*)P->data,*c; 1237 MPI_Comm comm; 1238 PetscMPIInt size,rank; 1239 Mat Cmpi; 1240 PetscFreeSpaceList free_space=NULL,current_space=NULL; 1241 PetscInt pn=P->cmap->n,aN=A->cmap->N,an=A->cmap->n; 1242 PetscInt *lnk,i,k,nsend; 1243 PetscBT lnkbt; 1244 PetscMPIInt tagi,tagj,*len_si,*len_s,*len_ri,icompleted=0,nrecv; 1245 PetscInt **buf_rj,**buf_ri,**buf_ri_k; 1246 PetscInt len,proc,*dnz,*onz,*owners,nzi; 1247 PetscInt nrows,*buf_s,*buf_si,*buf_si_i,**nextrow,**nextci; 1248 MPI_Request *swaits,*rwaits; 1249 MPI_Status *sstatus,rstatus; 1250 PetscLayout rowmap; 1251 PetscInt *owners_co,*coi,*coj; /* i and j array of (p->B)^T*A*P - used in the communication */ 1252 PetscMPIInt *len_r,*id_r; /* array of length of comm->size, store send/recv matrix values */ 1253 PetscInt *Jptr,*prmap=p->garray,con,j,Crmax; 1254 Mat_SeqAIJ *a_loc,*c_loc,*c_oth; 1255 PetscTable ta; 1256 1257 PetscFunctionBegin; 1258 ierr = PetscObjectGetComm((PetscObject)A,&comm);CHKERRQ(ierr); 1259 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr); 1260 ierr = MPI_Comm_rank(comm,&rank);CHKERRQ(ierr); 1261 1262 /* create symbolic parallel matrix Cmpi */ 1263 ierr = MatCreate(comm,&Cmpi);CHKERRQ(ierr); 1264 ierr = MatSetType(Cmpi,MATMPIAIJ);CHKERRQ(ierr); 1265 1266 Cmpi->ops->mattransposemultnumeric = MatTransposeMatMultNumeric_MPIAIJ_MPIAIJ_nonscalable; 1267 1268 /* create struct Mat_PtAPMPI and attached it to C later */ 1269 ierr = PetscNew(&ptap);CHKERRQ(ierr); 1270 ptap->reuse = MAT_INITIAL_MATRIX; 1271 1272 /* (0) compute Rd = Pd^T, Ro = Po^T */ 1273 /* --------------------------------- */ 1274 ierr = MatTranspose_SeqAIJ(p->A,MAT_INITIAL_MATRIX,&ptap->Rd);CHKERRQ(ierr); 1275 ierr = MatTranspose_SeqAIJ(p->B,MAT_INITIAL_MATRIX,&ptap->Ro);CHKERRQ(ierr); 1276 1277 /* (1) compute symbolic A_loc */ 1278 /* ---------------------------*/ 1279 ierr = MatMPIAIJGetLocalMat(A,MAT_INITIAL_MATRIX,&ptap->A_loc);CHKERRQ(ierr); 1280 1281 /* (2-1) compute symbolic C_oth = Ro*A_loc */ 1282 /* ------------------------------------ */ 1283 ierr = MatMatMultSymbolic_SeqAIJ_SeqAIJ(ptap->Ro,ptap->A_loc,fill,&ptap->C_oth);CHKERRQ(ierr); 1284 1285 /* (3) send coj of C_oth to other processors */ 1286 /* ------------------------------------------ */ 1287 /* determine row ownership */ 1288 ierr = PetscLayoutCreate(comm,&rowmap);CHKERRQ(ierr); 1289 rowmap->n = pn; 1290 rowmap->bs = 1; 1291 ierr = PetscLayoutSetUp(rowmap);CHKERRQ(ierr); 1292 owners = rowmap->range; 1293 1294 /* determine the number of messages to send, their lengths */ 1295 ierr = PetscMalloc4(size,&len_s,size,&len_si,size,&sstatus,size+2,&owners_co);CHKERRQ(ierr); 1296 ierr = PetscMemzero(len_s,size*sizeof(PetscMPIInt));CHKERRQ(ierr); 1297 ierr = PetscMemzero(len_si,size*sizeof(PetscMPIInt));CHKERRQ(ierr); 1298 1299 c_oth = (Mat_SeqAIJ*)ptap->C_oth->data; 1300 coi = c_oth->i; coj = c_oth->j; 1301 con = ptap->C_oth->rmap->n; 1302 proc = 0; 1303 for (i=0; i<con; i++) { 1304 while (prmap[i] >= owners[proc+1]) proc++; 1305 len_si[proc]++; /* num of rows in Co(=Pt*A) to be sent to [proc] */ 1306 len_s[proc] += coi[i+1] - coi[i]; /* num of nonzeros in Co to be sent to [proc] */ 1307 } 1308 1309 len = 0; /* max length of buf_si[], see (4) */ 1310 owners_co[0] = 0; 1311 nsend = 0; 1312 for (proc=0; proc<size; proc++) { 1313 owners_co[proc+1] = owners_co[proc] + len_si[proc]; 1314 if (len_s[proc]) { 1315 nsend++; 1316 len_si[proc] = 2*(len_si[proc] + 1); /* length of buf_si to be sent to [proc] */ 1317 len += len_si[proc]; 1318 } 1319 } 1320 1321 /* determine the number and length of messages to receive for coi and coj */ 1322 ierr = PetscGatherNumberOfMessages(comm,NULL,len_s,&nrecv);CHKERRQ(ierr); 1323 ierr = PetscGatherMessageLengths2(comm,nsend,nrecv,len_s,len_si,&id_r,&len_r,&len_ri);CHKERRQ(ierr); 1324 1325 /* post the Irecv and Isend of coj */ 1326 ierr = PetscCommGetNewTag(comm,&tagj);CHKERRQ(ierr); 1327 ierr = PetscPostIrecvInt(comm,tagj,nrecv,id_r,len_r,&buf_rj,&rwaits);CHKERRQ(ierr); 1328 ierr = PetscMalloc1(nsend+1,&swaits);CHKERRQ(ierr); 1329 for (proc=0, k=0; proc<size; proc++) { 1330 if (!len_s[proc]) continue; 1331 i = owners_co[proc]; 1332 ierr = MPI_Isend(coj+coi[i],len_s[proc],MPIU_INT,proc,tagj,comm,swaits+k);CHKERRQ(ierr); 1333 k++; 1334 } 1335 1336 /* (2-2) compute symbolic C_loc = Rd*A_loc */ 1337 /* ---------------------------------------- */ 1338 ierr = MatMatMultSymbolic_SeqAIJ_SeqAIJ(ptap->Rd,ptap->A_loc,fill,&ptap->C_loc);CHKERRQ(ierr); 1339 c_loc = (Mat_SeqAIJ*)ptap->C_loc->data; 1340 1341 /* receives coj are complete */ 1342 for (i=0; i<nrecv; i++) { 1343 ierr = MPI_Waitany(nrecv,rwaits,&icompleted,&rstatus);CHKERRQ(ierr); 1344 } 1345 ierr = PetscFree(rwaits);CHKERRQ(ierr); 1346 if (nsend) {ierr = MPI_Waitall(nsend,swaits,sstatus);CHKERRQ(ierr);} 1347 1348 /* add received column indices into ta to update Crmax */ 1349 a_loc = (Mat_SeqAIJ*)(ptap->A_loc)->data; 1350 1351 /* create and initialize a linked list */ 1352 ierr = PetscTableCreate(an,aN,&ta);CHKERRQ(ierr); /* for compute Crmax */ 1353 MatRowMergeMax_SeqAIJ(a_loc,ptap->A_loc->rmap->N,ta); 1354 1355 for (k=0; k<nrecv; k++) {/* k-th received message */ 1356 Jptr = buf_rj[k]; 1357 for (j=0; j<len_r[k]; j++) { 1358 ierr = PetscTableAdd(ta,*(Jptr+j)+1,1,INSERT_VALUES);CHKERRQ(ierr); 1359 } 1360 } 1361 ierr = PetscTableGetCount(ta,&Crmax);CHKERRQ(ierr); 1362 ierr = PetscTableDestroy(&ta);CHKERRQ(ierr); 1363 1364 /* (4) send and recv coi */ 1365 /*-----------------------*/ 1366 ierr = PetscCommGetNewTag(comm,&tagi);CHKERRQ(ierr); 1367 ierr = PetscPostIrecvInt(comm,tagi,nrecv,id_r,len_ri,&buf_ri,&rwaits);CHKERRQ(ierr); 1368 ierr = PetscMalloc1(len+1,&buf_s);CHKERRQ(ierr); 1369 buf_si = buf_s; /* points to the beginning of k-th msg to be sent */ 1370 for (proc=0,k=0; proc<size; proc++) { 1371 if (!len_s[proc]) continue; 1372 /* form outgoing message for i-structure: 1373 buf_si[0]: nrows to be sent 1374 [1:nrows]: row index (global) 1375 [nrows+1:2*nrows+1]: i-structure index 1376 */ 1377 /*-------------------------------------------*/ 1378 nrows = len_si[proc]/2 - 1; /* num of rows in Co to be sent to [proc] */ 1379 buf_si_i = buf_si + nrows+1; 1380 buf_si[0] = nrows; 1381 buf_si_i[0] = 0; 1382 nrows = 0; 1383 for (i=owners_co[proc]; i<owners_co[proc+1]; i++) { 1384 nzi = coi[i+1] - coi[i]; 1385 buf_si_i[nrows+1] = buf_si_i[nrows] + nzi; /* i-structure */ 1386 buf_si[nrows+1] = prmap[i] -owners[proc]; /* local row index */ 1387 nrows++; 1388 } 1389 ierr = MPI_Isend(buf_si,len_si[proc],MPIU_INT,proc,tagi,comm,swaits+k);CHKERRQ(ierr); 1390 k++; 1391 buf_si += len_si[proc]; 1392 } 1393 for (i=0; i<nrecv; i++) { 1394 ierr = MPI_Waitany(nrecv,rwaits,&icompleted,&rstatus);CHKERRQ(ierr); 1395 } 1396 ierr = PetscFree(rwaits);CHKERRQ(ierr); 1397 if (nsend) {ierr = MPI_Waitall(nsend,swaits,sstatus);CHKERRQ(ierr);} 1398 1399 ierr = PetscFree4(len_s,len_si,sstatus,owners_co);CHKERRQ(ierr); 1400 ierr = PetscFree(len_ri);CHKERRQ(ierr); 1401 ierr = PetscFree(swaits);CHKERRQ(ierr); 1402 ierr = PetscFree(buf_s);CHKERRQ(ierr); 1403 1404 /* (5) compute the local portion of Cmpi */ 1405 /* ------------------------------------------ */ 1406 /* set initial free space to be Crmax, sufficient for holding nozeros in each row of Cmpi */ 1407 ierr = PetscFreeSpaceGet(Crmax,&free_space);CHKERRQ(ierr); 1408 current_space = free_space; 1409 1410 ierr = PetscMalloc3(nrecv,&buf_ri_k,nrecv,&nextrow,nrecv,&nextci);CHKERRQ(ierr); 1411 for (k=0; k<nrecv; k++) { 1412 buf_ri_k[k] = buf_ri[k]; /* beginning of k-th recved i-structure */ 1413 nrows = *buf_ri_k[k]; 1414 nextrow[k] = buf_ri_k[k] + 1; /* next row number of k-th recved i-structure */ 1415 nextci[k] = buf_ri_k[k] + (nrows + 1); /* poins to the next i-structure of k-th recved i-structure */ 1416 } 1417 1418 ierr = MatPreallocateInitialize(comm,pn,an,dnz,onz);CHKERRQ(ierr); 1419 ierr = PetscLLCondensedCreate(Crmax,aN,&lnk,&lnkbt);CHKERRQ(ierr); 1420 for (i=0; i<pn; i++) { 1421 /* add C_loc into Cmpi */ 1422 nzi = c_loc->i[i+1] - c_loc->i[i]; 1423 Jptr = c_loc->j + c_loc->i[i]; 1424 ierr = PetscLLCondensedAddSorted(nzi,Jptr,lnk,lnkbt);CHKERRQ(ierr); 1425 1426 /* add received col data into lnk */ 1427 for (k=0; k<nrecv; k++) { /* k-th received message */ 1428 if (i == *nextrow[k]) { /* i-th row */ 1429 nzi = *(nextci[k]+1) - *nextci[k]; 1430 Jptr = buf_rj[k] + *nextci[k]; 1431 ierr = PetscLLCondensedAddSorted(nzi,Jptr,lnk,lnkbt);CHKERRQ(ierr); 1432 nextrow[k]++; nextci[k]++; 1433 } 1434 } 1435 nzi = lnk[0]; 1436 1437 /* copy data into free space, then initialize lnk */ 1438 ierr = PetscLLCondensedClean(aN,nzi,current_space->array,lnk,lnkbt);CHKERRQ(ierr); 1439 ierr = MatPreallocateSet(i+owners[rank],nzi,current_space->array,dnz,onz);CHKERRQ(ierr); 1440 } 1441 ierr = PetscFree3(buf_ri_k,nextrow,nextci);CHKERRQ(ierr); 1442 ierr = PetscLLDestroy(lnk,lnkbt);CHKERRQ(ierr); 1443 ierr = PetscFreeSpaceDestroy(free_space);CHKERRQ(ierr); 1444 1445 /* local sizes and preallocation */ 1446 ierr = MatSetSizes(Cmpi,pn,an,PETSC_DETERMINE,PETSC_DETERMINE);CHKERRQ(ierr); 1447 ierr = MatSetBlockSizes(Cmpi,PetscAbs(P->cmap->bs),PetscAbs(P->cmap->bs));CHKERRQ(ierr); 1448 ierr = MatMPIAIJSetPreallocation(Cmpi,0,dnz,0,onz);CHKERRQ(ierr); 1449 ierr = MatPreallocateFinalize(dnz,onz);CHKERRQ(ierr); 1450 1451 /* members in merge */ 1452 ierr = PetscFree(id_r);CHKERRQ(ierr); 1453 ierr = PetscFree(len_r);CHKERRQ(ierr); 1454 ierr = PetscFree(buf_ri[0]);CHKERRQ(ierr); 1455 ierr = PetscFree(buf_ri);CHKERRQ(ierr); 1456 ierr = PetscFree(buf_rj[0]);CHKERRQ(ierr); 1457 ierr = PetscFree(buf_rj);CHKERRQ(ierr); 1458 ierr = PetscLayoutDestroy(&rowmap);CHKERRQ(ierr); 1459 1460 /* attach the supporting struct to Cmpi for reuse */ 1461 c = (Mat_MPIAIJ*)Cmpi->data; 1462 c->ptap = ptap; 1463 ptap->duplicate = Cmpi->ops->duplicate; 1464 ptap->destroy = Cmpi->ops->destroy; 1465 1466 /* Cmpi is not ready for use - assembly will be done by MatPtAPNumeric() */ 1467 Cmpi->assembled = PETSC_FALSE; 1468 Cmpi->ops->destroy = MatDestroy_MPIAIJ_PtAP; 1469 Cmpi->ops->duplicate = MatDuplicate_MPIAIJ_MatPtAP; 1470 *C = Cmpi; 1471 PetscFunctionReturn(0); 1472 } 1473 1474 PetscErrorCode MatTransposeMatMultNumeric_MPIAIJ_MPIAIJ_nonscalable(Mat P,Mat A,Mat C) 1475 { 1476 PetscErrorCode ierr; 1477 Mat_MPIAIJ *p=(Mat_MPIAIJ*)P->data,*c=(Mat_MPIAIJ*)C->data; 1478 Mat_SeqAIJ *c_seq; 1479 Mat_PtAPMPI *ptap = c->ptap; 1480 Mat A_loc,C_loc,C_oth; 1481 PetscInt i,rstart,rend,cm,ncols,row; 1482 const PetscInt *cols; 1483 const PetscScalar *vals; 1484 1485 PetscFunctionBegin; 1486 ierr = MatZeroEntries(C);CHKERRQ(ierr); 1487 1488 if (ptap->reuse == MAT_REUSE_MATRIX) { 1489 /* These matrices are obtained in MatTransposeMatMultSymbolic() */ 1490 /* 1) get R = Pd^T, Ro = Po^T */ 1491 /*----------------------------*/ 1492 ierr = MatTranspose_SeqAIJ(p->A,MAT_REUSE_MATRIX,&ptap->Rd);CHKERRQ(ierr); 1493 ierr = MatTranspose_SeqAIJ(p->B,MAT_REUSE_MATRIX,&ptap->Ro);CHKERRQ(ierr); 1494 1495 /* 2) compute numeric A_loc */ 1496 /*--------------------------*/ 1497 ierr = MatMPIAIJGetLocalMat(A,MAT_REUSE_MATRIX,&ptap->A_loc);CHKERRQ(ierr); 1498 } 1499 1500 /* 3) C_loc = Rd*A_loc, C_oth = Ro*A_loc */ 1501 A_loc = ptap->A_loc; 1502 ierr = ((ptap->C_loc)->ops->matmultnumeric)(ptap->Rd,A_loc,ptap->C_loc);CHKERRQ(ierr); 1503 ierr = ((ptap->C_oth)->ops->matmultnumeric)(ptap->Ro,A_loc,ptap->C_oth);CHKERRQ(ierr); 1504 C_loc = ptap->C_loc; 1505 C_oth = ptap->C_oth; 1506 1507 /* add C_loc and Co to to C */ 1508 ierr = MatGetOwnershipRange(C,&rstart,&rend);CHKERRQ(ierr); 1509 1510 /* C_loc -> C */ 1511 cm = C_loc->rmap->N; 1512 c_seq = (Mat_SeqAIJ*)C_loc->data; 1513 cols = c_seq->j; 1514 vals = c_seq->a; 1515 for (i=0; i<cm; i++) { 1516 ncols = c_seq->i[i+1] - c_seq->i[i]; 1517 row = rstart + i; 1518 ierr = MatSetValues(C,1,&row,ncols,cols,vals,ADD_VALUES);CHKERRQ(ierr); 1519 cols += ncols; vals += ncols; 1520 } 1521 1522 /* Co -> C, off-processor part */ 1523 cm = C_oth->rmap->N; 1524 c_seq = (Mat_SeqAIJ*)C_oth->data; 1525 cols = c_seq->j; 1526 vals = c_seq->a; 1527 for (i=0; i<cm; i++) { 1528 ncols = c_seq->i[i+1] - c_seq->i[i]; 1529 row = p->garray[i]; 1530 ierr = MatSetValues(C,1,&row,ncols,cols,vals,ADD_VALUES);CHKERRQ(ierr); 1531 cols += ncols; vals += ncols; 1532 } 1533 ierr = MatAssemblyBegin(C,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1534 ierr = MatAssemblyEnd(C,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1535 1536 ptap->reuse = MAT_REUSE_MATRIX; 1537 PetscFunctionReturn(0); 1538 } 1539 1540 PetscErrorCode MatTransposeMatMultNumeric_MPIAIJ_MPIAIJ(Mat P,Mat A,Mat C) 1541 { 1542 PetscErrorCode ierr; 1543 Mat_Merge_SeqsToMPI *merge; 1544 Mat_MPIAIJ *p =(Mat_MPIAIJ*)P->data,*c=(Mat_MPIAIJ*)C->data; 1545 Mat_SeqAIJ *pd=(Mat_SeqAIJ*)(p->A)->data,*po=(Mat_SeqAIJ*)(p->B)->data; 1546 Mat_PtAPMPI *ptap; 1547 PetscInt *adj; 1548 PetscInt i,j,k,anz,pnz,row,*cj,nexta; 1549 MatScalar *ada,*ca,valtmp; 1550 PetscInt am =A->rmap->n,cm=C->rmap->n,pon=(p->B)->cmap->n; 1551 MPI_Comm comm; 1552 PetscMPIInt size,rank,taga,*len_s; 1553 PetscInt *owners,proc,nrows,**buf_ri_k,**nextrow,**nextci; 1554 PetscInt **buf_ri,**buf_rj; 1555 PetscInt cnz=0,*bj_i,*bi,*bj,bnz,nextcj; /* bi,bj,ba: local array of C(mpi mat) */ 1556 MPI_Request *s_waits,*r_waits; 1557 MPI_Status *status; 1558 MatScalar **abuf_r,*ba_i,*pA,*coa,*ba; 1559 PetscInt *ai,*aj,*coi,*coj,*poJ,*pdJ; 1560 Mat A_loc; 1561 Mat_SeqAIJ *a_loc; 1562 1563 PetscFunctionBegin; 1564 ierr = PetscObjectGetComm((PetscObject)C,&comm);CHKERRQ(ierr); 1565 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr); 1566 ierr = MPI_Comm_rank(comm,&rank);CHKERRQ(ierr); 1567 1568 ptap = c->ptap; 1569 merge = ptap->merge; 1570 1571 /* 2) compute numeric C_seq = P_loc^T*A_loc */ 1572 /*------------------------------------------*/ 1573 /* get data from symbolic products */ 1574 coi = merge->coi; coj = merge->coj; 1575 ierr = PetscCalloc1(coi[pon]+1,&coa);CHKERRQ(ierr); 1576 bi = merge->bi; bj = merge->bj; 1577 owners = merge->rowmap->range; 1578 ierr = PetscCalloc1(bi[cm]+1,&ba);CHKERRQ(ierr); 1579 1580 /* get A_loc by taking all local rows of A */ 1581 A_loc = ptap->A_loc; 1582 ierr = MatMPIAIJGetLocalMat(A,MAT_REUSE_MATRIX,&A_loc);CHKERRQ(ierr); 1583 a_loc = (Mat_SeqAIJ*)(A_loc)->data; 1584 ai = a_loc->i; 1585 aj = a_loc->j; 1586 1587 for (i=0; i<am; i++) { 1588 anz = ai[i+1] - ai[i]; 1589 adj = aj + ai[i]; 1590 ada = a_loc->a + ai[i]; 1591 1592 /* 2-b) Compute Cseq = P_loc[i,:]^T*A[i,:] using outer product */ 1593 /*-------------------------------------------------------------*/ 1594 /* put the value into Co=(p->B)^T*A (off-diagonal part, send to others) */ 1595 pnz = po->i[i+1] - po->i[i]; 1596 poJ = po->j + po->i[i]; 1597 pA = po->a + po->i[i]; 1598 for (j=0; j<pnz; j++) { 1599 row = poJ[j]; 1600 cj = coj + coi[row]; 1601 ca = coa + coi[row]; 1602 /* perform sparse axpy */ 1603 nexta = 0; 1604 valtmp = pA[j]; 1605 for (k=0; nexta<anz; k++) { 1606 if (cj[k] == adj[nexta]) { 1607 ca[k] += valtmp*ada[nexta]; 1608 nexta++; 1609 } 1610 } 1611 ierr = PetscLogFlops(2.0*anz);CHKERRQ(ierr); 1612 } 1613 1614 /* put the value into Cd (diagonal part) */ 1615 pnz = pd->i[i+1] - pd->i[i]; 1616 pdJ = pd->j + pd->i[i]; 1617 pA = pd->a + pd->i[i]; 1618 for (j=0; j<pnz; j++) { 1619 row = pdJ[j]; 1620 cj = bj + bi[row]; 1621 ca = ba + bi[row]; 1622 /* perform sparse axpy */ 1623 nexta = 0; 1624 valtmp = pA[j]; 1625 for (k=0; nexta<anz; k++) { 1626 if (cj[k] == adj[nexta]) { 1627 ca[k] += valtmp*ada[nexta]; 1628 nexta++; 1629 } 1630 } 1631 ierr = PetscLogFlops(2.0*anz);CHKERRQ(ierr); 1632 } 1633 } 1634 1635 /* 3) send and recv matrix values coa */ 1636 /*------------------------------------*/ 1637 buf_ri = merge->buf_ri; 1638 buf_rj = merge->buf_rj; 1639 len_s = merge->len_s; 1640 ierr = PetscCommGetNewTag(comm,&taga);CHKERRQ(ierr); 1641 ierr = PetscPostIrecvScalar(comm,taga,merge->nrecv,merge->id_r,merge->len_r,&abuf_r,&r_waits);CHKERRQ(ierr); 1642 1643 ierr = PetscMalloc2(merge->nsend+1,&s_waits,size,&status);CHKERRQ(ierr); 1644 for (proc=0,k=0; proc<size; proc++) { 1645 if (!len_s[proc]) continue; 1646 i = merge->owners_co[proc]; 1647 ierr = MPI_Isend(coa+coi[i],len_s[proc],MPIU_MATSCALAR,proc,taga,comm,s_waits+k);CHKERRQ(ierr); 1648 k++; 1649 } 1650 if (merge->nrecv) {ierr = MPI_Waitall(merge->nrecv,r_waits,status);CHKERRQ(ierr);} 1651 if (merge->nsend) {ierr = MPI_Waitall(merge->nsend,s_waits,status);CHKERRQ(ierr);} 1652 1653 ierr = PetscFree2(s_waits,status);CHKERRQ(ierr); 1654 ierr = PetscFree(r_waits);CHKERRQ(ierr); 1655 ierr = PetscFree(coa);CHKERRQ(ierr); 1656 1657 /* 4) insert local Cseq and received values into Cmpi */ 1658 /*----------------------------------------------------*/ 1659 ierr = PetscMalloc3(merge->nrecv,&buf_ri_k,merge->nrecv,&nextrow,merge->nrecv,&nextci);CHKERRQ(ierr); 1660 for (k=0; k<merge->nrecv; k++) { 1661 buf_ri_k[k] = buf_ri[k]; /* beginning of k-th recved i-structure */ 1662 nrows = *(buf_ri_k[k]); 1663 nextrow[k] = buf_ri_k[k]+1; /* next row number of k-th recved i-structure */ 1664 nextci[k] = buf_ri_k[k] + (nrows + 1); /* poins to the next i-structure of k-th recved i-structure */ 1665 } 1666 1667 for (i=0; i<cm; i++) { 1668 row = owners[rank] + i; /* global row index of C_seq */ 1669 bj_i = bj + bi[i]; /* col indices of the i-th row of C */ 1670 ba_i = ba + bi[i]; 1671 bnz = bi[i+1] - bi[i]; 1672 /* add received vals into ba */ 1673 for (k=0; k<merge->nrecv; k++) { /* k-th received message */ 1674 /* i-th row */ 1675 if (i == *nextrow[k]) { 1676 cnz = *(nextci[k]+1) - *nextci[k]; 1677 cj = buf_rj[k] + *(nextci[k]); 1678 ca = abuf_r[k] + *(nextci[k]); 1679 nextcj = 0; 1680 for (j=0; nextcj<cnz; j++) { 1681 if (bj_i[j] == cj[nextcj]) { /* bcol == ccol */ 1682 ba_i[j] += ca[nextcj++]; 1683 } 1684 } 1685 nextrow[k]++; nextci[k]++; 1686 ierr = PetscLogFlops(2.0*cnz);CHKERRQ(ierr); 1687 } 1688 } 1689 ierr = MatSetValues(C,1,&row,bnz,bj_i,ba_i,INSERT_VALUES);CHKERRQ(ierr); 1690 } 1691 ierr = MatAssemblyBegin(C,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1692 ierr = MatAssemblyEnd(C,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1693 1694 ierr = PetscFree(ba);CHKERRQ(ierr); 1695 ierr = PetscFree(abuf_r[0]);CHKERRQ(ierr); 1696 ierr = PetscFree(abuf_r);CHKERRQ(ierr); 1697 ierr = PetscFree3(buf_ri_k,nextrow,nextci);CHKERRQ(ierr); 1698 PetscFunctionReturn(0); 1699 } 1700 1701 PetscErrorCode MatTransposeMatMultSymbolic_MPIAIJ_MPIAIJ(Mat P,Mat A,PetscReal fill,Mat *C) 1702 { 1703 PetscErrorCode ierr; 1704 Mat Cmpi,A_loc,POt,PDt; 1705 Mat_PtAPMPI *ptap; 1706 PetscFreeSpaceList free_space=NULL,current_space=NULL; 1707 Mat_MPIAIJ *p=(Mat_MPIAIJ*)P->data,*a=(Mat_MPIAIJ*)A->data,*c; 1708 PetscInt *pdti,*pdtj,*poti,*potj,*ptJ; 1709 PetscInt nnz; 1710 PetscInt *lnk,*owners_co,*coi,*coj,i,k,pnz,row; 1711 PetscInt am =A->rmap->n,pn=P->cmap->n; 1712 MPI_Comm comm; 1713 PetscMPIInt size,rank,tagi,tagj,*len_si,*len_s,*len_ri; 1714 PetscInt **buf_rj,**buf_ri,**buf_ri_k; 1715 PetscInt len,proc,*dnz,*onz,*owners; 1716 PetscInt nzi,*bi,*bj; 1717 PetscInt nrows,*buf_s,*buf_si,*buf_si_i,**nextrow,**nextci; 1718 MPI_Request *swaits,*rwaits; 1719 MPI_Status *sstatus,rstatus; 1720 Mat_Merge_SeqsToMPI *merge; 1721 PetscInt *ai,*aj,*Jptr,anz,*prmap=p->garray,pon,nspacedouble=0,j; 1722 PetscReal afill =1.0,afill_tmp; 1723 PetscInt rstart = P->cmap->rstart,rmax,aN=A->cmap->N,Armax; 1724 PetscScalar *vals; 1725 Mat_SeqAIJ *a_loc,*pdt,*pot; 1726 PetscTable ta; 1727 1728 PetscFunctionBegin; 1729 ierr = PetscObjectGetComm((PetscObject)A,&comm);CHKERRQ(ierr); 1730 /* check if matrix local sizes are compatible */ 1731 if (A->rmap->rstart != P->rmap->rstart || A->rmap->rend != P->rmap->rend) SETERRQ4(comm,PETSC_ERR_ARG_SIZ,"Matrix local dimensions are incompatible, A (%D, %D) != P (%D,%D)",A->rmap->rstart,A->rmap->rend,P->rmap->rstart,P->rmap->rend); 1732 1733 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr); 1734 ierr = MPI_Comm_rank(comm,&rank);CHKERRQ(ierr); 1735 1736 /* create struct Mat_PtAPMPI and attached it to C later */ 1737 ierr = PetscNew(&ptap);CHKERRQ(ierr); 1738 1739 /* get A_loc by taking all local rows of A */ 1740 ierr = MatMPIAIJGetLocalMat(A,MAT_INITIAL_MATRIX,&A_loc);CHKERRQ(ierr); 1741 1742 ptap->A_loc = A_loc; 1743 a_loc = (Mat_SeqAIJ*)(A_loc)->data; 1744 ai = a_loc->i; 1745 aj = a_loc->j; 1746 1747 /* determine symbolic Co=(p->B)^T*A - send to others */ 1748 /*----------------------------------------------------*/ 1749 ierr = MatTransposeSymbolic_SeqAIJ(p->A,&PDt);CHKERRQ(ierr); 1750 pdt = (Mat_SeqAIJ*)PDt->data; 1751 pdti = pdt->i; pdtj = pdt->j; 1752 1753 ierr = MatTransposeSymbolic_SeqAIJ(p->B,&POt);CHKERRQ(ierr); 1754 pot = (Mat_SeqAIJ*)POt->data; 1755 poti = pot->i; potj = pot->j; 1756 1757 /* then, compute symbolic Co = (p->B)^T*A */ 1758 pon = (p->B)->cmap->n; /* total num of rows to be sent to other processors 1759 >= (num of nonzero rows of C_seq) - pn */ 1760 ierr = PetscMalloc1(pon+1,&coi);CHKERRQ(ierr); 1761 coi[0] = 0; 1762 1763 /* set initial free space to be fill*(nnz(p->B) + nnz(A)) */ 1764 nnz = PetscRealIntMultTruncate(fill,PetscIntSumTruncate(poti[pon],ai[am])); 1765 ierr = PetscFreeSpaceGet(nnz,&free_space);CHKERRQ(ierr); 1766 current_space = free_space; 1767 1768 /* create and initialize a linked list */ 1769 ierr = PetscTableCreate(A->cmap->n + a->B->cmap->N,aN,&ta);CHKERRQ(ierr); 1770 MatRowMergeMax_SeqAIJ(a_loc,am,ta); 1771 ierr = PetscTableGetCount(ta,&Armax);CHKERRQ(ierr); 1772 1773 ierr = PetscLLCondensedCreate_Scalable(Armax,&lnk);CHKERRQ(ierr); 1774 1775 for (i=0; i<pon; i++) { 1776 pnz = poti[i+1] - poti[i]; 1777 ptJ = potj + poti[i]; 1778 for (j=0; j<pnz; j++) { 1779 row = ptJ[j]; /* row of A_loc == col of Pot */ 1780 anz = ai[row+1] - ai[row]; 1781 Jptr = aj + ai[row]; 1782 /* add non-zero cols of AP into the sorted linked list lnk */ 1783 ierr = PetscLLCondensedAddSorted_Scalable(anz,Jptr,lnk);CHKERRQ(ierr); 1784 } 1785 nnz = lnk[0]; 1786 1787 /* If free space is not available, double the total space in the list */ 1788 if (current_space->local_remaining<nnz) { 1789 ierr = PetscFreeSpaceGet(PetscIntSumTruncate(nnz,current_space->total_array_size),¤t_space);CHKERRQ(ierr); 1790 nspacedouble++; 1791 } 1792 1793 /* Copy data into free space, and zero out denserows */ 1794 ierr = PetscLLCondensedClean_Scalable(nnz,current_space->array,lnk);CHKERRQ(ierr); 1795 1796 current_space->array += nnz; 1797 current_space->local_used += nnz; 1798 current_space->local_remaining -= nnz; 1799 1800 coi[i+1] = coi[i] + nnz; 1801 } 1802 1803 ierr = PetscMalloc1(coi[pon]+1,&coj);CHKERRQ(ierr); 1804 ierr = PetscFreeSpaceContiguous(&free_space,coj);CHKERRQ(ierr); 1805 ierr = PetscLLCondensedDestroy_Scalable(lnk);CHKERRQ(ierr); /* must destroy to get a new one for C */ 1806 1807 afill_tmp = (PetscReal)coi[pon]/(poti[pon] + ai[am]+1); 1808 if (afill_tmp > afill) afill = afill_tmp; 1809 1810 /* send j-array (coj) of Co to other processors */ 1811 /*----------------------------------------------*/ 1812 /* determine row ownership */ 1813 ierr = PetscNew(&merge);CHKERRQ(ierr); 1814 ierr = PetscLayoutCreate(comm,&merge->rowmap);CHKERRQ(ierr); 1815 1816 merge->rowmap->n = pn; 1817 merge->rowmap->bs = 1; 1818 1819 ierr = PetscLayoutSetUp(merge->rowmap);CHKERRQ(ierr); 1820 owners = merge->rowmap->range; 1821 1822 /* determine the number of messages to send, their lengths */ 1823 ierr = PetscCalloc1(size,&len_si);CHKERRQ(ierr); 1824 ierr = PetscMalloc1(size,&merge->len_s);CHKERRQ(ierr); 1825 1826 len_s = merge->len_s; 1827 merge->nsend = 0; 1828 1829 ierr = PetscMalloc1(size+2,&owners_co);CHKERRQ(ierr); 1830 ierr = PetscMemzero(len_s,size*sizeof(PetscMPIInt));CHKERRQ(ierr); 1831 1832 proc = 0; 1833 for (i=0; i<pon; i++) { 1834 while (prmap[i] >= owners[proc+1]) proc++; 1835 len_si[proc]++; /* num of rows in Co to be sent to [proc] */ 1836 len_s[proc] += coi[i+1] - coi[i]; 1837 } 1838 1839 len = 0; /* max length of buf_si[] */ 1840 owners_co[0] = 0; 1841 for (proc=0; proc<size; proc++) { 1842 owners_co[proc+1] = owners_co[proc] + len_si[proc]; 1843 if (len_si[proc]) { 1844 merge->nsend++; 1845 len_si[proc] = 2*(len_si[proc] + 1); 1846 len += len_si[proc]; 1847 } 1848 } 1849 1850 /* determine the number and length of messages to receive for coi and coj */ 1851 ierr = PetscGatherNumberOfMessages(comm,NULL,len_s,&merge->nrecv);CHKERRQ(ierr); 1852 ierr = PetscGatherMessageLengths2(comm,merge->nsend,merge->nrecv,len_s,len_si,&merge->id_r,&merge->len_r,&len_ri);CHKERRQ(ierr); 1853 1854 /* post the Irecv and Isend of coj */ 1855 ierr = PetscCommGetNewTag(comm,&tagj);CHKERRQ(ierr); 1856 ierr = PetscPostIrecvInt(comm,tagj,merge->nrecv,merge->id_r,merge->len_r,&buf_rj,&rwaits);CHKERRQ(ierr); 1857 ierr = PetscMalloc1(merge->nsend+1,&swaits);CHKERRQ(ierr); 1858 for (proc=0, k=0; proc<size; proc++) { 1859 if (!len_s[proc]) continue; 1860 i = owners_co[proc]; 1861 ierr = MPI_Isend(coj+coi[i],len_s[proc],MPIU_INT,proc,tagj,comm,swaits+k);CHKERRQ(ierr); 1862 k++; 1863 } 1864 1865 /* receives and sends of coj are complete */ 1866 ierr = PetscMalloc1(size,&sstatus);CHKERRQ(ierr); 1867 for (i=0; i<merge->nrecv; i++) { 1868 PetscMPIInt icompleted; 1869 ierr = MPI_Waitany(merge->nrecv,rwaits,&icompleted,&rstatus);CHKERRQ(ierr); 1870 } 1871 ierr = PetscFree(rwaits);CHKERRQ(ierr); 1872 if (merge->nsend) {ierr = MPI_Waitall(merge->nsend,swaits,sstatus);CHKERRQ(ierr);} 1873 1874 /* add received column indices into table to update Armax */ 1875 /* Armax can be as large as aN if a P[row,:] is dense, see src/ksp/ksp/examples/tutorials/ex56.c! */ 1876 for (k=0; k<merge->nrecv; k++) {/* k-th received message */ 1877 Jptr = buf_rj[k]; 1878 for (j=0; j<merge->len_r[k]; j++) { 1879 ierr = PetscTableAdd(ta,*(Jptr+j)+1,1,INSERT_VALUES);CHKERRQ(ierr); 1880 } 1881 } 1882 ierr = PetscTableGetCount(ta,&Armax);CHKERRQ(ierr); 1883 /* printf("Armax %d, an %d + Bn %d = %d, aN %d\n",Armax,A->cmap->n,a->B->cmap->N,A->cmap->n+a->B->cmap->N,aN); */ 1884 1885 /* send and recv coi */ 1886 /*-------------------*/ 1887 ierr = PetscCommGetNewTag(comm,&tagi);CHKERRQ(ierr); 1888 ierr = PetscPostIrecvInt(comm,tagi,merge->nrecv,merge->id_r,len_ri,&buf_ri,&rwaits);CHKERRQ(ierr); 1889 ierr = PetscMalloc1(len+1,&buf_s);CHKERRQ(ierr); 1890 buf_si = buf_s; /* points to the beginning of k-th msg to be sent */ 1891 for (proc=0,k=0; proc<size; proc++) { 1892 if (!len_s[proc]) continue; 1893 /* form outgoing message for i-structure: 1894 buf_si[0]: nrows to be sent 1895 [1:nrows]: row index (global) 1896 [nrows+1:2*nrows+1]: i-structure index 1897 */ 1898 /*-------------------------------------------*/ 1899 nrows = len_si[proc]/2 - 1; 1900 buf_si_i = buf_si + nrows+1; 1901 buf_si[0] = nrows; 1902 buf_si_i[0] = 0; 1903 nrows = 0; 1904 for (i=owners_co[proc]; i<owners_co[proc+1]; i++) { 1905 nzi = coi[i+1] - coi[i]; 1906 buf_si_i[nrows+1] = buf_si_i[nrows] + nzi; /* i-structure */ 1907 buf_si[nrows+1] = prmap[i] -owners[proc]; /* local row index */ 1908 nrows++; 1909 } 1910 ierr = MPI_Isend(buf_si,len_si[proc],MPIU_INT,proc,tagi,comm,swaits+k);CHKERRQ(ierr); 1911 k++; 1912 buf_si += len_si[proc]; 1913 } 1914 i = merge->nrecv; 1915 while (i--) { 1916 PetscMPIInt icompleted; 1917 ierr = MPI_Waitany(merge->nrecv,rwaits,&icompleted,&rstatus);CHKERRQ(ierr); 1918 } 1919 ierr = PetscFree(rwaits);CHKERRQ(ierr); 1920 if (merge->nsend) {ierr = MPI_Waitall(merge->nsend,swaits,sstatus);CHKERRQ(ierr);} 1921 ierr = PetscFree(len_si);CHKERRQ(ierr); 1922 ierr = PetscFree(len_ri);CHKERRQ(ierr); 1923 ierr = PetscFree(swaits);CHKERRQ(ierr); 1924 ierr = PetscFree(sstatus);CHKERRQ(ierr); 1925 ierr = PetscFree(buf_s);CHKERRQ(ierr); 1926 1927 /* compute the local portion of C (mpi mat) */ 1928 /*------------------------------------------*/ 1929 /* allocate bi array and free space for accumulating nonzero column info */ 1930 ierr = PetscMalloc1(pn+1,&bi);CHKERRQ(ierr); 1931 bi[0] = 0; 1932 1933 /* set initial free space to be fill*(nnz(P) + nnz(AP)) */ 1934 nnz = PetscRealIntMultTruncate(fill,PetscIntSumTruncate(pdti[pn],PetscIntSumTruncate(poti[pon],ai[am]))); 1935 ierr = PetscFreeSpaceGet(nnz,&free_space);CHKERRQ(ierr); 1936 current_space = free_space; 1937 1938 ierr = PetscMalloc3(merge->nrecv,&buf_ri_k,merge->nrecv,&nextrow,merge->nrecv,&nextci);CHKERRQ(ierr); 1939 for (k=0; k<merge->nrecv; k++) { 1940 buf_ri_k[k] = buf_ri[k]; /* beginning of k-th recved i-structure */ 1941 nrows = *buf_ri_k[k]; 1942 nextrow[k] = buf_ri_k[k] + 1; /* next row number of k-th recved i-structure */ 1943 nextci[k] = buf_ri_k[k] + (nrows + 1); /* points to the next i-structure of k-th recieved i-structure */ 1944 } 1945 1946 ierr = PetscLLCondensedCreate_Scalable(Armax,&lnk);CHKERRQ(ierr); 1947 ierr = MatPreallocateInitialize(comm,pn,A->cmap->n,dnz,onz);CHKERRQ(ierr); 1948 rmax = 0; 1949 for (i=0; i<pn; i++) { 1950 /* add pdt[i,:]*AP into lnk */ 1951 pnz = pdti[i+1] - pdti[i]; 1952 ptJ = pdtj + pdti[i]; 1953 for (j=0; j<pnz; j++) { 1954 row = ptJ[j]; /* row of AP == col of Pt */ 1955 anz = ai[row+1] - ai[row]; 1956 Jptr = aj + ai[row]; 1957 /* add non-zero cols of AP into the sorted linked list lnk */ 1958 ierr = PetscLLCondensedAddSorted_Scalable(anz,Jptr,lnk);CHKERRQ(ierr); 1959 } 1960 1961 /* add received col data into lnk */ 1962 for (k=0; k<merge->nrecv; k++) { /* k-th received message */ 1963 if (i == *nextrow[k]) { /* i-th row */ 1964 nzi = *(nextci[k]+1) - *nextci[k]; 1965 Jptr = buf_rj[k] + *nextci[k]; 1966 ierr = PetscLLCondensedAddSorted_Scalable(nzi,Jptr,lnk);CHKERRQ(ierr); 1967 nextrow[k]++; nextci[k]++; 1968 } 1969 } 1970 nnz = lnk[0]; 1971 1972 /* if free space is not available, make more free space */ 1973 if (current_space->local_remaining<nnz) { 1974 ierr = PetscFreeSpaceGet(PetscIntSumTruncate(nnz,current_space->total_array_size),¤t_space);CHKERRQ(ierr); 1975 nspacedouble++; 1976 } 1977 /* copy data into free space, then initialize lnk */ 1978 ierr = PetscLLCondensedClean_Scalable(nnz,current_space->array,lnk);CHKERRQ(ierr); 1979 ierr = MatPreallocateSet(i+owners[rank],nnz,current_space->array,dnz,onz);CHKERRQ(ierr); 1980 1981 current_space->array += nnz; 1982 current_space->local_used += nnz; 1983 current_space->local_remaining -= nnz; 1984 1985 bi[i+1] = bi[i] + nnz; 1986 if (nnz > rmax) rmax = nnz; 1987 } 1988 ierr = PetscFree3(buf_ri_k,nextrow,nextci);CHKERRQ(ierr); 1989 1990 ierr = PetscMalloc1(bi[pn]+1,&bj);CHKERRQ(ierr); 1991 ierr = PetscFreeSpaceContiguous(&free_space,bj);CHKERRQ(ierr); 1992 afill_tmp = (PetscReal)bi[pn]/(pdti[pn] + poti[pon] + ai[am]+1); 1993 if (afill_tmp > afill) afill = afill_tmp; 1994 ierr = PetscLLCondensedDestroy_Scalable(lnk);CHKERRQ(ierr); 1995 ierr = PetscTableDestroy(&ta);CHKERRQ(ierr); 1996 1997 ierr = MatDestroy(&POt);CHKERRQ(ierr); 1998 ierr = MatDestroy(&PDt);CHKERRQ(ierr); 1999 2000 /* create symbolic parallel matrix Cmpi - why cannot be assembled in Numeric part */ 2001 /*----------------------------------------------------------------------------------*/ 2002 ierr = PetscCalloc1(rmax+1,&vals);CHKERRQ(ierr); 2003 2004 ierr = MatCreate(comm,&Cmpi);CHKERRQ(ierr); 2005 ierr = MatSetSizes(Cmpi,pn,A->cmap->n,PETSC_DETERMINE,PETSC_DETERMINE);CHKERRQ(ierr); 2006 ierr = MatSetBlockSizes(Cmpi,PetscAbs(P->cmap->bs),PetscAbs(A->cmap->bs));CHKERRQ(ierr); 2007 ierr = MatSetType(Cmpi,MATMPIAIJ);CHKERRQ(ierr); 2008 ierr = MatMPIAIJSetPreallocation(Cmpi,0,dnz,0,onz);CHKERRQ(ierr); 2009 ierr = MatPreallocateFinalize(dnz,onz);CHKERRQ(ierr); 2010 ierr = MatSetBlockSize(Cmpi,1);CHKERRQ(ierr); 2011 for (i=0; i<pn; i++) { 2012 row = i + rstart; 2013 nnz = bi[i+1] - bi[i]; 2014 Jptr = bj + bi[i]; 2015 ierr = MatSetValues(Cmpi,1,&row,nnz,Jptr,vals,INSERT_VALUES);CHKERRQ(ierr); 2016 } 2017 ierr = MatAssemblyBegin(Cmpi,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 2018 ierr = MatAssemblyEnd(Cmpi,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 2019 ierr = PetscFree(vals);CHKERRQ(ierr); 2020 2021 merge->bi = bi; 2022 merge->bj = bj; 2023 merge->coi = coi; 2024 merge->coj = coj; 2025 merge->buf_ri = buf_ri; 2026 merge->buf_rj = buf_rj; 2027 merge->owners_co = owners_co; 2028 2029 /* attach the supporting struct to Cmpi for reuse */ 2030 c = (Mat_MPIAIJ*)Cmpi->data; 2031 2032 c->ptap = ptap; 2033 ptap->api = NULL; 2034 ptap->apj = NULL; 2035 ptap->merge = merge; 2036 ptap->apa = NULL; 2037 ptap->destroy = Cmpi->ops->destroy; 2038 ptap->duplicate = Cmpi->ops->duplicate; 2039 2040 Cmpi->ops->mattransposemultnumeric = MatTransposeMatMultNumeric_MPIAIJ_MPIAIJ; 2041 Cmpi->ops->destroy = MatDestroy_MPIAIJ_PtAP; 2042 Cmpi->ops->duplicate = MatDuplicate_MPIAIJ_MatPtAP; 2043 2044 *C = Cmpi; 2045 #if defined(PETSC_USE_INFO) 2046 if (bi[pn] != 0) { 2047 ierr = PetscInfo3(Cmpi,"Reallocs %D; Fill ratio: given %g needed %g.\n",nspacedouble,(double)fill,(double)afill);CHKERRQ(ierr); 2048 ierr = PetscInfo1(Cmpi,"Use MatTransposeMatMult(A,B,MatReuse,%g,&C) for best performance.\n",(double)afill);CHKERRQ(ierr); 2049 } else { 2050 ierr = PetscInfo(Cmpi,"Empty matrix product\n");CHKERRQ(ierr); 2051 } 2052 #endif 2053 PetscFunctionReturn(0); 2054 } 2055