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