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