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" 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" 655 PetscErrorCode MatMatMultSymbolic_MPIAIJ_MPIAIJ(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 const char *algTypes[3] = {"scalable","nonscalable","matmatmult"}; 812 PetscInt alg=0; /* set default algorithm */ 813 814 PetscFunctionBegin; 815 if (scall == MAT_INITIAL_MATRIX) { 816 ierr = PetscObjectOptionsBegin((PetscObject)A);CHKERRQ(ierr); 817 ierr = PetscOptionsEList("-mattransposematmult_via","Algorithmic approach","MatTransposeMatMult",algTypes,3,algTypes[0],&alg,NULL);CHKERRQ(ierr); 818 ierr = PetscOptionsEnd();CHKERRQ(ierr); 819 820 ierr = PetscLogEventBegin(MAT_TransposeMatMultSymbolic,P,A,0,0);CHKERRQ(ierr); 821 switch (alg) { 822 case 1: 823 ierr = MatTransposeMatMultSymbolic_MPIAIJ_MPIAIJ_nonscalable(P,A,fill,C);CHKERRQ(ierr); 824 break; 825 case 2: 826 Mat Pt; 827 Mat_PtAPMPI *ptap; 828 Mat_MPIAIJ *c; 829 ierr = MatTranspose(P,MAT_INITIAL_MATRIX,&Pt);CHKERRQ(ierr); 830 ierr = MatMatMult(Pt,A,MAT_INITIAL_MATRIX,fill,C);CHKERRQ(ierr); 831 c = (Mat_MPIAIJ*)(*C)->data; 832 ptap = c->ptap; 833 ptap->Pt = Pt; 834 (*C)->ops->mattransposemultnumeric = MatTransposeMatMultNumeric_MPIAIJ_MPIAIJ_matmatmult; 835 PetscFunctionReturn(0); 836 break; 837 default: 838 ierr = MatTransposeMatMultSymbolic_MPIAIJ_MPIAIJ(P,A,fill,C);CHKERRQ(ierr); 839 break; 840 } 841 ierr = PetscLogEventEnd(MAT_TransposeMatMultSymbolic,P,A,0,0);CHKERRQ(ierr); 842 } 843 ierr = PetscLogEventBegin(MAT_TransposeMatMultNumeric,P,A,0,0);CHKERRQ(ierr); 844 ierr = (*(*C)->ops->mattransposemultnumeric)(P,A,*C);CHKERRQ(ierr); 845 ierr = PetscLogEventEnd(MAT_TransposeMatMultNumeric,P,A,0,0);CHKERRQ(ierr); 846 PetscFunctionReturn(0); 847 } 848 849 /* This routine only works when scall=MAT_REUSE_MATRIX! */ 850 #undef __FUNCT__ 851 #define __FUNCT__ "MatTransposeMatMultNumeric_MPIAIJ_MPIAIJ_matmatmult" 852 PetscErrorCode MatTransposeMatMultNumeric_MPIAIJ_MPIAIJ_matmatmult(Mat P,Mat A,Mat C) 853 { 854 PetscErrorCode ierr; 855 Mat_MPIAIJ *c=(Mat_MPIAIJ*)C->data; 856 Mat_PtAPMPI *ptap= c->ptap; 857 Mat Pt=ptap->Pt; 858 859 PetscFunctionBegin; 860 ierr = MatTranspose(P,MAT_REUSE_MATRIX,&Pt);CHKERRQ(ierr); 861 ierr = MatMatMultNumeric(Pt,A,C);CHKERRQ(ierr); 862 PetscFunctionReturn(0); 863 } 864 865 /* Non-scalable version, use dense axpy */ 866 #undef __FUNCT__ 867 #define __FUNCT__ "MatTransposeMatMultNumeric_MPIAIJ_MPIAIJ_nonscalable" 868 PetscErrorCode MatTransposeMatMultNumeric_MPIAIJ_MPIAIJ_nonscalable(Mat P,Mat A,Mat C) 869 { 870 PetscErrorCode ierr; 871 Mat_Merge_SeqsToMPI *merge; 872 Mat_MPIAIJ *p =(Mat_MPIAIJ*)P->data,*c=(Mat_MPIAIJ*)C->data; 873 Mat_SeqAIJ *pd=(Mat_SeqAIJ*)(p->A)->data,*po=(Mat_SeqAIJ*)(p->B)->data; 874 Mat_PtAPMPI *ptap; 875 PetscInt *adj,*aJ; 876 PetscInt i,j,k,anz,pnz,row,*cj; 877 MatScalar *ada,*aval,*ca,valtmp; 878 PetscInt am =A->rmap->n,cm=C->rmap->n,pon=(p->B)->cmap->n; 879 MPI_Comm comm; 880 PetscMPIInt size,rank,taga,*len_s; 881 PetscInt *owners,proc,nrows,**buf_ri_k,**nextrow,**nextci; 882 PetscInt **buf_ri,**buf_rj; 883 PetscInt cnz=0,*bj_i,*bi,*bj,bnz,nextcj; /* bi,bj,ba: local array of C(mpi mat) */ 884 MPI_Request *s_waits,*r_waits; 885 MPI_Status *status; 886 MatScalar **abuf_r,*ba_i,*pA,*coa,*ba; 887 PetscInt *ai,*aj,*coi,*coj; 888 PetscInt *poJ,*pdJ; 889 Mat A_loc; 890 Mat_SeqAIJ *a_loc; 891 892 PetscFunctionBegin; 893 ierr = PetscObjectGetComm((PetscObject)C,&comm);CHKERRQ(ierr); 894 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr); 895 ierr = MPI_Comm_rank(comm,&rank);CHKERRQ(ierr); 896 897 ptap = c->ptap; 898 merge = ptap->merge; 899 900 /* 2) compute numeric C_seq = P_loc^T*A_loc*P - dominating part */ 901 /*--------------------------------------------------------------*/ 902 /* get data from symbolic products */ 903 coi = merge->coi; coj = merge->coj; 904 ierr = PetscMalloc((coi[pon]+1)*sizeof(MatScalar),&coa);CHKERRQ(ierr); 905 ierr = PetscMemzero(coa,coi[pon]*sizeof(MatScalar));CHKERRQ(ierr); 906 907 bi = merge->bi; bj = merge->bj; 908 owners = merge->rowmap->range; 909 ierr = PetscMalloc((bi[cm]+1)*sizeof(MatScalar),&ba);CHKERRQ(ierr); 910 ierr = PetscMemzero(ba,bi[cm]*sizeof(MatScalar));CHKERRQ(ierr); 911 912 /* get A_loc by taking all local rows of A */ 913 A_loc = ptap->A_loc; 914 ierr = MatMPIAIJGetLocalMat(A,MAT_REUSE_MATRIX,&A_loc);CHKERRQ(ierr); 915 a_loc = (Mat_SeqAIJ*)(A_loc)->data; 916 ai = a_loc->i; 917 aj = a_loc->j; 918 919 ierr = PetscMalloc((A->cmap->N)*sizeof(PetscScalar),&aval);CHKERRQ(ierr); /* non-scalable!!! */ 920 ierr = PetscMemzero(aval,A->cmap->N*sizeof(PetscScalar));CHKERRQ(ierr); 921 922 for (i=0; i<am; i++) { 923 /* 2-a) put A[i,:] to dense array aval */ 924 anz = ai[i+1] - ai[i]; 925 adj = aj + ai[i]; 926 ada = a_loc->a + ai[i]; 927 for (j=0; j<anz; j++) { 928 aval[adj[j]] = ada[j]; 929 } 930 931 /* 2-b) Compute Cseq = P_loc[i,:]^T*A[i,:] using outer product */ 932 /*--------------------------------------------------------------*/ 933 /* put the value into Co=(p->B)^T*A (off-diagonal part, send to others) */ 934 pnz = po->i[i+1] - po->i[i]; 935 poJ = po->j + po->i[i]; 936 pA = po->a + po->i[i]; 937 for (j=0; j<pnz; j++) { 938 row = poJ[j]; 939 cnz = coi[row+1] - coi[row]; 940 cj = coj + coi[row]; 941 ca = coa + coi[row]; 942 /* perform dense axpy */ 943 valtmp = pA[j]; 944 for (k=0; k<cnz; k++) { 945 ca[k] += valtmp*aval[cj[k]]; 946 } 947 ierr = PetscLogFlops(2.0*cnz);CHKERRQ(ierr); 948 } 949 950 /* put the value into Cd (diagonal part) */ 951 pnz = pd->i[i+1] - pd->i[i]; 952 pdJ = pd->j + pd->i[i]; 953 pA = pd->a + pd->i[i]; 954 for (j=0; j<pnz; j++) { 955 row = pdJ[j]; 956 cnz = bi[row+1] - bi[row]; 957 cj = bj + bi[row]; 958 ca = ba + bi[row]; 959 /* perform dense axpy */ 960 valtmp = pA[j]; 961 for (k=0; k<cnz; k++) { 962 ca[k] += valtmp*aval[cj[k]]; 963 } 964 ierr = PetscLogFlops(2.0*cnz);CHKERRQ(ierr); 965 } 966 967 /* zero the current row of Pt*A */ 968 aJ = aj + ai[i]; 969 for (k=0; k<anz; k++) aval[aJ[k]] = 0.0; 970 } 971 972 /* 3) send and recv matrix values coa */ 973 /*------------------------------------*/ 974 buf_ri = merge->buf_ri; 975 buf_rj = merge->buf_rj; 976 len_s = merge->len_s; 977 ierr = PetscCommGetNewTag(comm,&taga);CHKERRQ(ierr); 978 ierr = PetscPostIrecvScalar(comm,taga,merge->nrecv,merge->id_r,merge->len_r,&abuf_r,&r_waits);CHKERRQ(ierr); 979 980 ierr = PetscMalloc2(merge->nsend+1,MPI_Request,&s_waits,size,MPI_Status,&status);CHKERRQ(ierr); 981 for (proc=0,k=0; proc<size; proc++) { 982 if (!len_s[proc]) continue; 983 i = merge->owners_co[proc]; 984 ierr = MPI_Isend(coa+coi[i],len_s[proc],MPIU_MATSCALAR,proc,taga,comm,s_waits+k);CHKERRQ(ierr); 985 k++; 986 } 987 if (merge->nrecv) {ierr = MPI_Waitall(merge->nrecv,r_waits,status);CHKERRQ(ierr);} 988 if (merge->nsend) {ierr = MPI_Waitall(merge->nsend,s_waits,status);CHKERRQ(ierr);} 989 990 ierr = PetscFree2(s_waits,status);CHKERRQ(ierr); 991 ierr = PetscFree(r_waits);CHKERRQ(ierr); 992 ierr = PetscFree(coa);CHKERRQ(ierr); 993 994 /* 4) insert local Cseq and received values into Cmpi */ 995 /*----------------------------------------------------*/ 996 ierr = PetscMalloc3(merge->nrecv,PetscInt**,&buf_ri_k,merge->nrecv,PetscInt*,&nextrow,merge->nrecv,PetscInt*,&nextci);CHKERRQ(ierr); 997 for (k=0; k<merge->nrecv; k++) { 998 buf_ri_k[k] = buf_ri[k]; /* beginning of k-th recved i-structure */ 999 nrows = *(buf_ri_k[k]); 1000 nextrow[k] = buf_ri_k[k]+1; /* next row number of k-th recved i-structure */ 1001 nextci[k] = buf_ri_k[k] + (nrows + 1); /* poins to the next i-structure of k-th recved i-structure */ 1002 } 1003 1004 for (i=0; i<cm; i++) { 1005 row = owners[rank] + i; /* global row index of C_seq */ 1006 bj_i = bj + bi[i]; /* col indices of the i-th row of C */ 1007 ba_i = ba + bi[i]; 1008 bnz = bi[i+1] - bi[i]; 1009 /* add received vals into ba */ 1010 for (k=0; k<merge->nrecv; k++) { /* k-th received message */ 1011 /* i-th row */ 1012 if (i == *nextrow[k]) { 1013 cnz = *(nextci[k]+1) - *nextci[k]; 1014 cj = buf_rj[k] + *(nextci[k]); 1015 ca = abuf_r[k] + *(nextci[k]); 1016 nextcj = 0; 1017 for (j=0; nextcj<cnz; j++) { 1018 if (bj_i[j] == cj[nextcj]) { /* bcol == ccol */ 1019 ba_i[j] += ca[nextcj++]; 1020 } 1021 } 1022 nextrow[k]++; nextci[k]++; 1023 ierr = PetscLogFlops(2.0*cnz);CHKERRQ(ierr); 1024 } 1025 } 1026 ierr = MatSetValues(C,1,&row,bnz,bj_i,ba_i,INSERT_VALUES);CHKERRQ(ierr); 1027 } 1028 ierr = MatAssemblyBegin(C,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1029 ierr = MatAssemblyEnd(C,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1030 1031 ierr = PetscFree(ba);CHKERRQ(ierr); 1032 ierr = PetscFree(abuf_r[0]);CHKERRQ(ierr); 1033 ierr = PetscFree(abuf_r);CHKERRQ(ierr); 1034 ierr = PetscFree3(buf_ri_k,nextrow,nextci);CHKERRQ(ierr); 1035 ierr = PetscFree(aval);CHKERRQ(ierr); 1036 PetscFunctionReturn(0); 1037 } 1038 1039 /* This routine is modified from MatPtAPSymbolic_MPIAIJ_MPIAIJ() */ 1040 #undef __FUNCT__ 1041 #define __FUNCT__ "MatTransposeMatMultSymbolic_MPIAIJ_MPIAIJ_nonscalable" 1042 PetscErrorCode MatTransposeMatMultSymbolic_MPIAIJ_MPIAIJ_nonscalable(Mat P,Mat A,PetscReal fill,Mat *C) 1043 { 1044 PetscErrorCode ierr; 1045 Mat Cmpi,A_loc,POt,PDt; 1046 Mat_PtAPMPI *ptap; 1047 PetscFreeSpaceList free_space=NULL,current_space=NULL; 1048 Mat_MPIAIJ *p =(Mat_MPIAIJ*)P->data,*c; 1049 PetscInt *pdti,*pdtj,*poti,*potj,*ptJ; 1050 PetscInt nnz; 1051 PetscInt *lnk,*owners_co,*coi,*coj,i,k,pnz,row; 1052 PetscInt am=A->rmap->n,pn=P->cmap->n; 1053 PetscBT lnkbt; 1054 MPI_Comm comm; 1055 PetscMPIInt size,rank,tagi,tagj,*len_si,*len_s,*len_ri; 1056 PetscInt **buf_rj,**buf_ri,**buf_ri_k; 1057 PetscInt len,proc,*dnz,*onz,*owners; 1058 PetscInt nzi,*bi,*bj; 1059 PetscInt nrows,*buf_s,*buf_si,*buf_si_i,**nextrow,**nextci; 1060 MPI_Request *swaits,*rwaits; 1061 MPI_Status *sstatus,rstatus; 1062 Mat_Merge_SeqsToMPI *merge; 1063 PetscInt *ai,*aj,*Jptr,anz,*prmap=p->garray,pon,nspacedouble=0,j; 1064 PetscReal afill =1.0,afill_tmp; 1065 PetscInt rstart = P->cmap->rstart,rmax,aN=A->cmap->N,Crmax; 1066 PetscScalar *vals; 1067 Mat_SeqAIJ *a_loc, *pdt,*pot; 1068 1069 PetscFunctionBegin; 1070 ierr = PetscObjectGetComm((PetscObject)A,&comm);CHKERRQ(ierr); 1071 /* check if matrix local sizes are compatible */ 1072 if (A->rmap->rstart != P->rmap->rstart || A->rmap->rend != P->rmap->rend) { 1073 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); 1074 } 1075 1076 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr); 1077 ierr = MPI_Comm_rank(comm,&rank);CHKERRQ(ierr); 1078 1079 /* create struct Mat_PtAPMPI and attached it to C later */ 1080 ierr = PetscNew(Mat_PtAPMPI,&ptap);CHKERRQ(ierr); 1081 1082 /* get A_loc by taking all local rows of A */ 1083 ierr = MatMPIAIJGetLocalMat(A,MAT_INITIAL_MATRIX,&A_loc);CHKERRQ(ierr); 1084 1085 ptap->A_loc = A_loc; 1086 1087 a_loc = (Mat_SeqAIJ*)(A_loc)->data; 1088 ai = a_loc->i; 1089 aj = a_loc->j; 1090 1091 /* determine symbolic Co=(p->B)^T*A - send to others */ 1092 /*----------------------------------------------------*/ 1093 ierr = MatTransposeSymbolic_SeqAIJ(p->A,&PDt);CHKERRQ(ierr); 1094 pdt = (Mat_SeqAIJ*)PDt->data; 1095 pdti = pdt->i; pdtj = pdt->j; 1096 1097 ierr = MatTransposeSymbolic_SeqAIJ(p->B,&POt);CHKERRQ(ierr); 1098 pot = (Mat_SeqAIJ*)POt->data; 1099 poti = pot->i; potj = pot->j; 1100 1101 /* then, compute symbolic Co = (p->B)^T*A */ 1102 pon = (p->B)->cmap->n; /* total num of rows to be sent to other processors >= (num of nonzero rows of C_seq) - pn */ 1103 ierr = PetscMalloc((pon+1)*sizeof(PetscInt),&coi);CHKERRQ(ierr); 1104 coi[0] = 0; 1105 1106 /* set initial free space to be fill*(nnz(p->B) + nnz(A)) */ 1107 nnz = fill*(poti[pon] + ai[am]); 1108 ierr = PetscFreeSpaceGet(nnz,&free_space);CHKERRQ(ierr); 1109 current_space = free_space; 1110 1111 /* create and initialize a linked list */ 1112 i = PetscMax(pdt->rmax,pot->rmax); 1113 Crmax = i*a_loc->rmax*size; 1114 if (!Crmax || Crmax > aN) Crmax = aN; 1115 ierr = PetscLLCondensedCreate(Crmax,aN,&lnk,&lnkbt);CHKERRQ(ierr); 1116 1117 for (i=0; i<pon; i++) { 1118 pnz = poti[i+1] - poti[i]; 1119 ptJ = potj + poti[i]; 1120 for (j=0; j<pnz; j++) { 1121 row = ptJ[j]; /* row of A_loc == col of Pot */ 1122 anz = ai[row+1] - ai[row]; 1123 Jptr = aj + ai[row]; 1124 /* add non-zero cols of AP into the sorted linked list lnk */ 1125 ierr = PetscLLCondensedAddSorted(anz,Jptr,lnk,lnkbt);CHKERRQ(ierr); 1126 } 1127 nnz = lnk[0]; 1128 1129 /* If free space is not available, double the total space in the list */ 1130 if (current_space->local_remaining<nnz) { 1131 ierr = PetscFreeSpaceGet(nnz+current_space->total_array_size,¤t_space);CHKERRQ(ierr); 1132 nspacedouble++; 1133 } 1134 1135 /* Copy data into free space, and zero out denserows */ 1136 ierr = PetscLLCondensedClean(aN,nnz,current_space->array,lnk,lnkbt);CHKERRQ(ierr); 1137 1138 current_space->array += nnz; 1139 current_space->local_used += nnz; 1140 current_space->local_remaining -= nnz; 1141 1142 coi[i+1] = coi[i] + nnz; 1143 } 1144 1145 ierr = PetscMalloc((coi[pon]+1)*sizeof(PetscInt),&coj);CHKERRQ(ierr); 1146 ierr = PetscFreeSpaceContiguous(&free_space,coj);CHKERRQ(ierr); 1147 1148 afill_tmp = (PetscReal)coi[pon]/(poti[pon] + ai[am]+1); 1149 if (afill_tmp > afill) afill = afill_tmp; 1150 1151 /* send j-array (coj) of Co to other processors */ 1152 /*----------------------------------------------*/ 1153 /* determine row ownership */ 1154 ierr = PetscNew(Mat_Merge_SeqsToMPI,&merge);CHKERRQ(ierr); 1155 ierr = PetscLayoutCreate(comm,&merge->rowmap);CHKERRQ(ierr); 1156 1157 merge->rowmap->n = pn; 1158 merge->rowmap->bs = 1; 1159 1160 ierr = PetscLayoutSetUp(merge->rowmap);CHKERRQ(ierr); 1161 owners = merge->rowmap->range; 1162 1163 /* determine the number of messages to send, their lengths */ 1164 ierr = PetscMalloc(size*sizeof(PetscMPIInt),&len_si);CHKERRQ(ierr); 1165 ierr = PetscMemzero(len_si,size*sizeof(PetscMPIInt));CHKERRQ(ierr); 1166 ierr = PetscMalloc(size*sizeof(PetscMPIInt),&merge->len_s);CHKERRQ(ierr); 1167 1168 len_s = merge->len_s; 1169 merge->nsend = 0; 1170 1171 ierr = PetscMalloc((size+2)*sizeof(PetscInt),&owners_co);CHKERRQ(ierr); 1172 ierr = PetscMemzero(len_s,size*sizeof(PetscMPIInt));CHKERRQ(ierr); 1173 1174 proc = 0; 1175 for (i=0; i<pon; i++) { 1176 while (prmap[i] >= owners[proc+1]) proc++; 1177 len_si[proc]++; /* num of rows in Co to be sent to [proc] */ 1178 len_s[proc] += coi[i+1] - coi[i]; 1179 } 1180 1181 len = 0; /* max length of buf_si[] */ 1182 owners_co[0] = 0; 1183 for (proc=0; proc<size; proc++) { 1184 owners_co[proc+1] = owners_co[proc] + len_si[proc]; 1185 if (len_si[proc]) { 1186 merge->nsend++; 1187 len_si[proc] = 2*(len_si[proc] + 1); 1188 len += len_si[proc]; 1189 } 1190 } 1191 1192 /* determine the number and length of messages to receive for coi and coj */ 1193 ierr = PetscGatherNumberOfMessages(comm,NULL,len_s,&merge->nrecv);CHKERRQ(ierr); 1194 ierr = PetscGatherMessageLengths2(comm,merge->nsend,merge->nrecv,len_s,len_si,&merge->id_r,&merge->len_r,&len_ri);CHKERRQ(ierr); 1195 1196 /* post the Irecv and Isend of coj */ 1197 ierr = PetscCommGetNewTag(comm,&tagj);CHKERRQ(ierr); 1198 ierr = PetscPostIrecvInt(comm,tagj,merge->nrecv,merge->id_r,merge->len_r,&buf_rj,&rwaits);CHKERRQ(ierr); 1199 ierr = PetscMalloc((merge->nsend+1)*sizeof(MPI_Request),&swaits);CHKERRQ(ierr); 1200 for (proc=0, k=0; proc<size; proc++) { 1201 if (!len_s[proc]) continue; 1202 i = owners_co[proc]; 1203 ierr = MPI_Isend(coj+coi[i],len_s[proc],MPIU_INT,proc,tagj,comm,swaits+k);CHKERRQ(ierr); 1204 k++; 1205 } 1206 1207 /* receives and sends of coj are complete */ 1208 ierr = PetscMalloc(size*sizeof(MPI_Status),&sstatus);CHKERRQ(ierr); 1209 for (i=0; i<merge->nrecv; i++) { 1210 PetscMPIInt icompleted; 1211 ierr = MPI_Waitany(merge->nrecv,rwaits,&icompleted,&rstatus);CHKERRQ(ierr); 1212 } 1213 ierr = PetscFree(rwaits);CHKERRQ(ierr); 1214 if (merge->nsend) {ierr = MPI_Waitall(merge->nsend,swaits,sstatus);CHKERRQ(ierr);} 1215 1216 /* send and recv coi */ 1217 /*-------------------*/ 1218 ierr = PetscCommGetNewTag(comm,&tagi);CHKERRQ(ierr); 1219 ierr = PetscPostIrecvInt(comm,tagi,merge->nrecv,merge->id_r,len_ri,&buf_ri,&rwaits);CHKERRQ(ierr); 1220 ierr = PetscMalloc((len+1)*sizeof(PetscInt),&buf_s);CHKERRQ(ierr); 1221 buf_si = buf_s; /* points to the beginning of k-th msg to be sent */ 1222 for (proc=0,k=0; proc<size; proc++) { 1223 if (!len_s[proc]) continue; 1224 /* form outgoing message for i-structure: 1225 buf_si[0]: nrows to be sent 1226 [1:nrows]: row index (global) 1227 [nrows+1:2*nrows+1]: i-structure index 1228 */ 1229 /*-------------------------------------------*/ 1230 nrows = len_si[proc]/2 - 1; 1231 buf_si_i = buf_si + nrows+1; 1232 buf_si[0] = nrows; 1233 buf_si_i[0] = 0; 1234 nrows = 0; 1235 for (i=owners_co[proc]; i<owners_co[proc+1]; i++) { 1236 nzi = coi[i+1] - coi[i]; 1237 buf_si_i[nrows+1] = buf_si_i[nrows] + nzi; /* i-structure */ 1238 buf_si[nrows+1] = prmap[i] -owners[proc]; /* local row index */ 1239 nrows++; 1240 } 1241 ierr = MPI_Isend(buf_si,len_si[proc],MPIU_INT,proc,tagi,comm,swaits+k);CHKERRQ(ierr); 1242 k++; 1243 buf_si += len_si[proc]; 1244 } 1245 i = merge->nrecv; 1246 while (i--) { 1247 PetscMPIInt icompleted; 1248 ierr = MPI_Waitany(merge->nrecv,rwaits,&icompleted,&rstatus);CHKERRQ(ierr); 1249 } 1250 ierr = PetscFree(rwaits);CHKERRQ(ierr); 1251 if (merge->nsend) {ierr = MPI_Waitall(merge->nsend,swaits,sstatus);CHKERRQ(ierr);} 1252 ierr = PetscFree(len_si);CHKERRQ(ierr); 1253 ierr = PetscFree(len_ri);CHKERRQ(ierr); 1254 ierr = PetscFree(swaits);CHKERRQ(ierr); 1255 ierr = PetscFree(sstatus);CHKERRQ(ierr); 1256 ierr = PetscFree(buf_s);CHKERRQ(ierr); 1257 1258 /* compute the local portion of C (mpi mat) */ 1259 /*------------------------------------------*/ 1260 /* allocate bi array and free space for accumulating nonzero column info */ 1261 ierr = PetscMalloc((pn+1)*sizeof(PetscInt),&bi);CHKERRQ(ierr); 1262 bi[0] = 0; 1263 1264 /* set initial free space to be fill*(nnz(P) + nnz(A)) */ 1265 nnz = fill*(pdti[pn] + poti[pon] + ai[am]); 1266 ierr = PetscFreeSpaceGet(nnz,&free_space);CHKERRQ(ierr); 1267 current_space = free_space; 1268 1269 ierr = PetscMalloc3(merge->nrecv,PetscInt**,&buf_ri_k,merge->nrecv,PetscInt*,&nextrow,merge->nrecv,PetscInt*,&nextci);CHKERRQ(ierr); 1270 for (k=0; k<merge->nrecv; k++) { 1271 buf_ri_k[k] = buf_ri[k]; /* beginning of k-th recved i-structure */ 1272 nrows = *buf_ri_k[k]; 1273 nextrow[k] = buf_ri_k[k] + 1; /* next row number of k-th recved i-structure */ 1274 nextci[k] = buf_ri_k[k] + (nrows + 1); /* poins to the next i-structure of k-th recved i-structure */ 1275 } 1276 1277 ierr = MatPreallocateInitialize(comm,pn,A->cmap->n,dnz,onz);CHKERRQ(ierr); 1278 rmax = 0; 1279 for (i=0; i<pn; i++) { 1280 /* add pdt[i,:]*AP into lnk */ 1281 pnz = pdti[i+1] - pdti[i]; 1282 ptJ = pdtj + pdti[i]; 1283 for (j=0; j<pnz; j++) { 1284 row = ptJ[j]; /* row of AP == col of Pt */ 1285 anz = ai[row+1] - ai[row]; 1286 Jptr = aj + ai[row]; 1287 /* add non-zero cols of AP into the sorted linked list lnk */ 1288 ierr = PetscLLCondensedAddSorted(anz,Jptr,lnk,lnkbt);CHKERRQ(ierr); 1289 } 1290 1291 /* add received col data into lnk */ 1292 for (k=0; k<merge->nrecv; k++) { /* k-th received message */ 1293 if (i == *nextrow[k]) { /* i-th row */ 1294 nzi = *(nextci[k]+1) - *nextci[k]; 1295 Jptr = buf_rj[k] + *nextci[k]; 1296 ierr = PetscLLCondensedAddSorted(nzi,Jptr,lnk,lnkbt);CHKERRQ(ierr); 1297 nextrow[k]++; nextci[k]++; 1298 } 1299 } 1300 nnz = lnk[0]; 1301 1302 /* if free space is not available, make more free space */ 1303 if (current_space->local_remaining<nnz) { 1304 ierr = PetscFreeSpaceGet(nnz+current_space->total_array_size,¤t_space);CHKERRQ(ierr); 1305 nspacedouble++; 1306 } 1307 /* copy data into free space, then initialize lnk */ 1308 ierr = PetscLLCondensedClean(aN,nnz,current_space->array,lnk,lnkbt);CHKERRQ(ierr); 1309 ierr = MatPreallocateSet(i+owners[rank],nnz,current_space->array,dnz,onz);CHKERRQ(ierr); 1310 1311 current_space->array += nnz; 1312 current_space->local_used += nnz; 1313 current_space->local_remaining -= nnz; 1314 1315 bi[i+1] = bi[i] + nnz; 1316 if (nnz > rmax) rmax = nnz; 1317 } 1318 ierr = PetscFree3(buf_ri_k,nextrow,nextci);CHKERRQ(ierr); 1319 1320 ierr = PetscMalloc((bi[pn]+1)*sizeof(PetscInt),&bj);CHKERRQ(ierr); 1321 ierr = PetscFreeSpaceContiguous(&free_space,bj);CHKERRQ(ierr); 1322 1323 afill_tmp = (PetscReal)bi[pn]/(pdti[pn] + poti[pon] + ai[am]+1); 1324 if (afill_tmp > afill) afill = afill_tmp; 1325 ierr = PetscLLCondensedDestroy(lnk,lnkbt);CHKERRQ(ierr); 1326 ierr = MatDestroy(&POt);CHKERRQ(ierr); 1327 ierr = MatDestroy(&PDt);CHKERRQ(ierr); 1328 1329 /* create symbolic parallel matrix Cmpi - why cannot be assembled in Numeric part */ 1330 /*----------------------------------------------------------------------------------*/ 1331 ierr = PetscMalloc((rmax+1)*sizeof(PetscScalar),&vals);CHKERRQ(ierr); 1332 ierr = PetscMemzero(vals,rmax*sizeof(PetscScalar));CHKERRQ(ierr); 1333 1334 ierr = MatCreate(comm,&Cmpi);CHKERRQ(ierr); 1335 ierr = MatSetSizes(Cmpi,pn,A->cmap->n,PETSC_DETERMINE,PETSC_DETERMINE);CHKERRQ(ierr); 1336 ierr = MatSetBlockSizes(Cmpi,P->cmap->bs,A->cmap->bs);CHKERRQ(ierr); 1337 ierr = MatSetType(Cmpi,MATMPIAIJ);CHKERRQ(ierr); 1338 ierr = MatMPIAIJSetPreallocation(Cmpi,0,dnz,0,onz);CHKERRQ(ierr); 1339 ierr = MatPreallocateFinalize(dnz,onz);CHKERRQ(ierr); 1340 ierr = MatSetBlockSize(Cmpi,1);CHKERRQ(ierr); 1341 for (i=0; i<pn; i++) { 1342 row = i + rstart; 1343 nnz = bi[i+1] - bi[i]; 1344 Jptr = bj + bi[i]; 1345 ierr = MatSetValues(Cmpi,1,&row,nnz,Jptr,vals,INSERT_VALUES);CHKERRQ(ierr); 1346 } 1347 ierr = MatAssemblyBegin(Cmpi,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1348 ierr = MatAssemblyEnd(Cmpi,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1349 ierr = PetscFree(vals);CHKERRQ(ierr); 1350 1351 merge->bi = bi; 1352 merge->bj = bj; 1353 merge->coi = coi; 1354 merge->coj = coj; 1355 merge->buf_ri = buf_ri; 1356 merge->buf_rj = buf_rj; 1357 merge->owners_co = owners_co; 1358 merge->destroy = Cmpi->ops->destroy; 1359 merge->duplicate = Cmpi->ops->duplicate; 1360 1361 Cmpi->ops->mattransposemultnumeric = MatTransposeMatMultNumeric_MPIAIJ_MPIAIJ_nonscalable; 1362 Cmpi->ops->destroy = MatDestroy_MPIAIJ_PtAP; 1363 1364 /* attach the supporting struct to Cmpi for reuse */ 1365 c = (Mat_MPIAIJ*)Cmpi->data; 1366 c->ptap = ptap; 1367 ptap->api = NULL; 1368 ptap->apj = NULL; 1369 ptap->merge = merge; 1370 ptap->rmax = rmax; 1371 1372 *C = Cmpi; 1373 #if defined(PETSC_USE_INFO) 1374 if (bi[pn] != 0) { 1375 ierr = PetscInfo3(Cmpi,"Reallocs %D; Fill ratio: given %G needed %G.\n",nspacedouble,fill,afill);CHKERRQ(ierr); 1376 ierr = PetscInfo1(Cmpi,"Use MatTransposeMatMult(A,B,MatReuse,%G,&C) for best performance.\n",afill);CHKERRQ(ierr); 1377 } else { 1378 ierr = PetscInfo(Cmpi,"Empty matrix product\n");CHKERRQ(ierr); 1379 } 1380 #endif 1381 PetscFunctionReturn(0); 1382 } 1383 1384 #undef __FUNCT__ 1385 #define __FUNCT__ "MatTransposeMatMultNumeric_MPIAIJ_MPIAIJ" 1386 PetscErrorCode MatTransposeMatMultNumeric_MPIAIJ_MPIAIJ(Mat P,Mat A,Mat C) 1387 { 1388 PetscErrorCode ierr; 1389 Mat_Merge_SeqsToMPI *merge; 1390 Mat_MPIAIJ *p =(Mat_MPIAIJ*)P->data,*c=(Mat_MPIAIJ*)C->data; 1391 Mat_SeqAIJ *pd=(Mat_SeqAIJ*)(p->A)->data,*po=(Mat_SeqAIJ*)(p->B)->data; 1392 Mat_PtAPMPI *ptap; 1393 PetscInt *adj; 1394 PetscInt i,j,k,anz,pnz,row,*cj,nexta; 1395 MatScalar *ada,*ca,valtmp; 1396 PetscInt am =A->rmap->n,cm=C->rmap->n,pon=(p->B)->cmap->n; 1397 MPI_Comm comm; 1398 PetscMPIInt size,rank,taga,*len_s; 1399 PetscInt *owners,proc,nrows,**buf_ri_k,**nextrow,**nextci; 1400 PetscInt **buf_ri,**buf_rj; 1401 PetscInt cnz=0,*bj_i,*bi,*bj,bnz,nextcj; /* bi,bj,ba: local array of C(mpi mat) */ 1402 MPI_Request *s_waits,*r_waits; 1403 MPI_Status *status; 1404 MatScalar **abuf_r,*ba_i,*pA,*coa,*ba; 1405 PetscInt *ai,*aj,*coi,*coj; 1406 PetscInt *poJ,*pdJ; 1407 Mat A_loc; 1408 Mat_SeqAIJ *a_loc; 1409 1410 PetscFunctionBegin; 1411 ierr = PetscObjectGetComm((PetscObject)C,&comm);CHKERRQ(ierr); 1412 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr); 1413 ierr = MPI_Comm_rank(comm,&rank);CHKERRQ(ierr); 1414 1415 ptap = c->ptap; 1416 merge = ptap->merge; 1417 1418 /* 2) compute numeric C_seq = P_loc^T*A_loc */ 1419 /*------------------------------------------*/ 1420 /* get data from symbolic products */ 1421 coi = merge->coi; coj = merge->coj; 1422 ierr = PetscMalloc((coi[pon]+1)*sizeof(MatScalar),&coa);CHKERRQ(ierr); 1423 ierr = PetscMemzero(coa,coi[pon]*sizeof(MatScalar));CHKERRQ(ierr); 1424 bi = merge->bi; bj = merge->bj; 1425 owners = merge->rowmap->range; 1426 ierr = PetscMalloc((bi[cm]+1)*sizeof(MatScalar),&ba);CHKERRQ(ierr); 1427 ierr = PetscMemzero(ba,bi[cm]*sizeof(MatScalar));CHKERRQ(ierr); 1428 1429 /* get A_loc by taking all local rows of A */ 1430 A_loc = ptap->A_loc; 1431 ierr = MatMPIAIJGetLocalMat(A,MAT_REUSE_MATRIX,&A_loc);CHKERRQ(ierr); 1432 a_loc = (Mat_SeqAIJ*)(A_loc)->data; 1433 ai = a_loc->i; 1434 aj = a_loc->j; 1435 1436 for (i=0; i<am; i++) { 1437 anz = ai[i+1] - ai[i]; 1438 adj = aj + ai[i]; 1439 ada = a_loc->a + ai[i]; 1440 1441 /* 2-b) Compute Cseq = P_loc[i,:]^T*A[i,:] using outer product */ 1442 /*-------------------------------------------------------------*/ 1443 /* put the value into Co=(p->B)^T*A (off-diagonal part, send to others) */ 1444 pnz = po->i[i+1] - po->i[i]; 1445 poJ = po->j + po->i[i]; 1446 pA = po->a + po->i[i]; 1447 for (j=0; j<pnz; j++) { 1448 row = poJ[j]; 1449 cj = coj + coi[row]; 1450 ca = coa + coi[row]; 1451 /* perform sparse axpy */ 1452 nexta = 0; 1453 valtmp = pA[j]; 1454 for (k=0; nexta<anz; k++) { 1455 if (cj[k] == adj[nexta]) { 1456 ca[k] += valtmp*ada[nexta]; 1457 nexta++; 1458 } 1459 } 1460 ierr = PetscLogFlops(2.0*anz);CHKERRQ(ierr); 1461 } 1462 1463 /* put the value into Cd (diagonal part) */ 1464 pnz = pd->i[i+1] - pd->i[i]; 1465 pdJ = pd->j + pd->i[i]; 1466 pA = pd->a + pd->i[i]; 1467 for (j=0; j<pnz; j++) { 1468 row = pdJ[j]; 1469 cj = bj + bi[row]; 1470 ca = ba + bi[row]; 1471 /* perform sparse axpy */ 1472 nexta = 0; 1473 valtmp = pA[j]; 1474 for (k=0; nexta<anz; k++) { 1475 if (cj[k] == adj[nexta]) { 1476 ca[k] += valtmp*ada[nexta]; 1477 nexta++; 1478 } 1479 } 1480 ierr = PetscLogFlops(2.0*anz);CHKERRQ(ierr); 1481 } 1482 } 1483 1484 /* 3) send and recv matrix values coa */ 1485 /*------------------------------------*/ 1486 buf_ri = merge->buf_ri; 1487 buf_rj = merge->buf_rj; 1488 len_s = merge->len_s; 1489 ierr = PetscCommGetNewTag(comm,&taga);CHKERRQ(ierr); 1490 ierr = PetscPostIrecvScalar(comm,taga,merge->nrecv,merge->id_r,merge->len_r,&abuf_r,&r_waits);CHKERRQ(ierr); 1491 1492 ierr = PetscMalloc2(merge->nsend+1,MPI_Request,&s_waits,size,MPI_Status,&status);CHKERRQ(ierr); 1493 for (proc=0,k=0; proc<size; proc++) { 1494 if (!len_s[proc]) continue; 1495 i = merge->owners_co[proc]; 1496 ierr = MPI_Isend(coa+coi[i],len_s[proc],MPIU_MATSCALAR,proc,taga,comm,s_waits+k);CHKERRQ(ierr); 1497 k++; 1498 } 1499 if (merge->nrecv) {ierr = MPI_Waitall(merge->nrecv,r_waits,status);CHKERRQ(ierr);} 1500 if (merge->nsend) {ierr = MPI_Waitall(merge->nsend,s_waits,status);CHKERRQ(ierr);} 1501 1502 ierr = PetscFree2(s_waits,status);CHKERRQ(ierr); 1503 ierr = PetscFree(r_waits);CHKERRQ(ierr); 1504 ierr = PetscFree(coa);CHKERRQ(ierr); 1505 1506 /* 4) insert local Cseq and received values into Cmpi */ 1507 /*----------------------------------------------------*/ 1508 ierr = PetscMalloc3(merge->nrecv,PetscInt**,&buf_ri_k,merge->nrecv,PetscInt*,&nextrow,merge->nrecv,PetscInt*,&nextci);CHKERRQ(ierr); 1509 for (k=0; k<merge->nrecv; k++) { 1510 buf_ri_k[k] = buf_ri[k]; /* beginning of k-th recved i-structure */ 1511 nrows = *(buf_ri_k[k]); 1512 nextrow[k] = buf_ri_k[k]+1; /* next row number of k-th recved i-structure */ 1513 nextci[k] = buf_ri_k[k] + (nrows + 1); /* poins to the next i-structure of k-th recved i-structure */ 1514 } 1515 1516 for (i=0; i<cm; i++) { 1517 row = owners[rank] + i; /* global row index of C_seq */ 1518 bj_i = bj + bi[i]; /* col indices of the i-th row of C */ 1519 ba_i = ba + bi[i]; 1520 bnz = bi[i+1] - bi[i]; 1521 /* add received vals into ba */ 1522 for (k=0; k<merge->nrecv; k++) { /* k-th received message */ 1523 /* i-th row */ 1524 if (i == *nextrow[k]) { 1525 cnz = *(nextci[k]+1) - *nextci[k]; 1526 cj = buf_rj[k] + *(nextci[k]); 1527 ca = abuf_r[k] + *(nextci[k]); 1528 nextcj = 0; 1529 for (j=0; nextcj<cnz; j++) { 1530 if (bj_i[j] == cj[nextcj]) { /* bcol == ccol */ 1531 ba_i[j] += ca[nextcj++]; 1532 } 1533 } 1534 nextrow[k]++; nextci[k]++; 1535 ierr = PetscLogFlops(2.0*cnz);CHKERRQ(ierr); 1536 } 1537 } 1538 ierr = MatSetValues(C,1,&row,bnz,bj_i,ba_i,INSERT_VALUES);CHKERRQ(ierr); 1539 } 1540 ierr = MatAssemblyBegin(C,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1541 ierr = MatAssemblyEnd(C,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1542 1543 ierr = PetscFree(ba);CHKERRQ(ierr); 1544 ierr = PetscFree(abuf_r[0]);CHKERRQ(ierr); 1545 ierr = PetscFree(abuf_r);CHKERRQ(ierr); 1546 ierr = PetscFree3(buf_ri_k,nextrow,nextci);CHKERRQ(ierr); 1547 PetscFunctionReturn(0); 1548 } 1549 1550 /* This routine is modified from MatPtAPSymbolic_MPIAIJ_MPIAIJ(); 1551 differ from MatTransposeMatMultSymbolic_MPIAIJ_MPIAIJ_nonscalable in using LLCondensedCreate_Scalable() */ 1552 #undef __FUNCT__ 1553 #define __FUNCT__ "MatTransposeMatMultSymbolic_MPIAIJ_MPIAIJ" 1554 PetscErrorCode MatTransposeMatMultSymbolic_MPIAIJ_MPIAIJ(Mat P,Mat A,PetscReal fill,Mat *C) 1555 { 1556 PetscErrorCode ierr; 1557 Mat Cmpi,A_loc,POt,PDt; 1558 Mat_PtAPMPI *ptap; 1559 PetscFreeSpaceList free_space=NULL,current_space=NULL; 1560 Mat_MPIAIJ *p =(Mat_MPIAIJ*)P->data,*c; 1561 PetscInt *pdti,*pdtj,*poti,*potj,*ptJ; 1562 PetscInt nnz; 1563 PetscInt *lnk,*owners_co,*coi,*coj,i,k,pnz,row; 1564 PetscInt am =A->rmap->n,pn=P->cmap->n; 1565 MPI_Comm comm; 1566 PetscMPIInt size,rank,tagi,tagj,*len_si,*len_s,*len_ri; 1567 PetscInt **buf_rj,**buf_ri,**buf_ri_k; 1568 PetscInt len,proc,*dnz,*onz,*owners; 1569 PetscInt nzi,*bi,*bj; 1570 PetscInt nrows,*buf_s,*buf_si,*buf_si_i,**nextrow,**nextci; 1571 MPI_Request *swaits,*rwaits; 1572 MPI_Status *sstatus,rstatus; 1573 Mat_Merge_SeqsToMPI *merge; 1574 PetscInt *ai,*aj,*Jptr,anz,*prmap=p->garray,pon,nspacedouble=0,j; 1575 PetscReal afill =1.0,afill_tmp; 1576 PetscInt rstart = P->cmap->rstart,rmax,aN=A->cmap->N,Crmax; 1577 PetscScalar *vals; 1578 Mat_SeqAIJ *a_loc, *pdt,*pot; 1579 1580 PetscFunctionBegin; 1581 ierr = PetscObjectGetComm((PetscObject)A,&comm);CHKERRQ(ierr); 1582 /* check if matrix local sizes are compatible */ 1583 if (A->rmap->rstart != P->rmap->rstart || A->rmap->rend != P->rmap->rend) { 1584 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); 1585 } 1586 1587 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr); 1588 ierr = MPI_Comm_rank(comm,&rank);CHKERRQ(ierr); 1589 1590 /* create struct Mat_PtAPMPI and attached it to C later */ 1591 ierr = PetscNew(Mat_PtAPMPI,&ptap);CHKERRQ(ierr); 1592 1593 /* get A_loc by taking all local rows of A */ 1594 ierr = MatMPIAIJGetLocalMat(A,MAT_INITIAL_MATRIX,&A_loc);CHKERRQ(ierr); 1595 1596 ptap->A_loc = A_loc; 1597 a_loc = (Mat_SeqAIJ*)(A_loc)->data; 1598 ai = a_loc->i; 1599 aj = a_loc->j; 1600 1601 /* determine symbolic Co=(p->B)^T*A - send to others */ 1602 /*----------------------------------------------------*/ 1603 ierr = MatTransposeSymbolic_SeqAIJ(p->A,&PDt);CHKERRQ(ierr); 1604 pdt = (Mat_SeqAIJ*)PDt->data; 1605 pdti = pdt->i; pdtj = pdt->j; 1606 1607 ierr = MatTransposeSymbolic_SeqAIJ(p->B,&POt);CHKERRQ(ierr); 1608 pot = (Mat_SeqAIJ*)POt->data; 1609 poti = pot->i; potj = pot->j; 1610 1611 /* then, compute symbolic Co = (p->B)^T*A */ 1612 pon = (p->B)->cmap->n; /* total num of rows to be sent to other processors 1613 >= (num of nonzero rows of C_seq) - pn */ 1614 ierr = PetscMalloc((pon+1)*sizeof(PetscInt),&coi);CHKERRQ(ierr); 1615 coi[0] = 0; 1616 1617 /* set initial free space to be fill*(nnz(p->B) + nnz(A)) */ 1618 nnz = fill*(poti[pon] + ai[am]); 1619 ierr = PetscFreeSpaceGet(nnz,&free_space);CHKERRQ(ierr); 1620 current_space = free_space; 1621 1622 /* create and initialize a linked list */ 1623 i = PetscMax(pdt->rmax,pot->rmax); 1624 Crmax = i*a_loc->rmax*size; /* non-scalable! */ 1625 if (!Crmax || Crmax > aN) Crmax = aN; 1626 ierr = PetscLLCondensedCreate_Scalable(Crmax,&lnk);CHKERRQ(ierr); 1627 1628 for (i=0; i<pon; i++) { 1629 pnz = poti[i+1] - poti[i]; 1630 ptJ = potj + poti[i]; 1631 for (j=0; j<pnz; j++) { 1632 row = ptJ[j]; /* row of A_loc == col of Pot */ 1633 anz = ai[row+1] - ai[row]; 1634 Jptr = aj + ai[row]; 1635 /* add non-zero cols of AP into the sorted linked list lnk */ 1636 ierr = PetscLLCondensedAddSorted_Scalable(anz,Jptr,lnk);CHKERRQ(ierr); 1637 } 1638 nnz = lnk[0]; 1639 1640 /* If free space is not available, double the total space in the list */ 1641 if (current_space->local_remaining<nnz) { 1642 ierr = PetscFreeSpaceGet(nnz+current_space->total_array_size,¤t_space);CHKERRQ(ierr); 1643 nspacedouble++; 1644 } 1645 1646 /* Copy data into free space, and zero out denserows */ 1647 ierr = PetscLLCondensedClean_Scalable(nnz,current_space->array,lnk);CHKERRQ(ierr); 1648 1649 current_space->array += nnz; 1650 current_space->local_used += nnz; 1651 current_space->local_remaining -= nnz; 1652 1653 coi[i+1] = coi[i] + nnz; 1654 } 1655 1656 ierr = PetscMalloc((coi[pon]+1)*sizeof(PetscInt),&coj);CHKERRQ(ierr); 1657 ierr = PetscFreeSpaceContiguous(&free_space,coj);CHKERRQ(ierr); 1658 1659 afill_tmp = (PetscReal)coi[pon]/(poti[pon] + ai[am]+1); 1660 if (afill_tmp > afill) afill = afill_tmp; 1661 1662 /* send j-array (coj) of Co to other processors */ 1663 /*----------------------------------------------*/ 1664 /* determine row ownership */ 1665 ierr = PetscNew(Mat_Merge_SeqsToMPI,&merge);CHKERRQ(ierr); 1666 ierr = PetscLayoutCreate(comm,&merge->rowmap);CHKERRQ(ierr); 1667 1668 merge->rowmap->n = pn; 1669 merge->rowmap->bs = 1; 1670 1671 ierr = PetscLayoutSetUp(merge->rowmap);CHKERRQ(ierr); 1672 owners = merge->rowmap->range; 1673 1674 /* determine the number of messages to send, their lengths */ 1675 ierr = PetscMalloc(size*sizeof(PetscMPIInt),&len_si);CHKERRQ(ierr); 1676 ierr = PetscMemzero(len_si,size*sizeof(PetscMPIInt));CHKERRQ(ierr); 1677 ierr = PetscMalloc(size*sizeof(PetscMPIInt),&merge->len_s);CHKERRQ(ierr); 1678 1679 len_s = merge->len_s; 1680 merge->nsend = 0; 1681 1682 ierr = PetscMalloc((size+2)*sizeof(PetscInt),&owners_co);CHKERRQ(ierr); 1683 ierr = PetscMemzero(len_s,size*sizeof(PetscMPIInt));CHKERRQ(ierr); 1684 1685 proc = 0; 1686 for (i=0; i<pon; i++) { 1687 while (prmap[i] >= owners[proc+1]) proc++; 1688 len_si[proc]++; /* num of rows in Co to be sent to [proc] */ 1689 len_s[proc] += coi[i+1] - coi[i]; 1690 } 1691 1692 len = 0; /* max length of buf_si[] */ 1693 owners_co[0] = 0; 1694 for (proc=0; proc<size; proc++) { 1695 owners_co[proc+1] = owners_co[proc] + len_si[proc]; 1696 if (len_si[proc]) { 1697 merge->nsend++; 1698 len_si[proc] = 2*(len_si[proc] + 1); 1699 len += len_si[proc]; 1700 } 1701 } 1702 1703 /* determine the number and length of messages to receive for coi and coj */ 1704 ierr = PetscGatherNumberOfMessages(comm,NULL,len_s,&merge->nrecv);CHKERRQ(ierr); 1705 ierr = PetscGatherMessageLengths2(comm,merge->nsend,merge->nrecv,len_s,len_si,&merge->id_r,&merge->len_r,&len_ri);CHKERRQ(ierr); 1706 1707 /* post the Irecv and Isend of coj */ 1708 ierr = PetscCommGetNewTag(comm,&tagj);CHKERRQ(ierr); 1709 ierr = PetscPostIrecvInt(comm,tagj,merge->nrecv,merge->id_r,merge->len_r,&buf_rj,&rwaits);CHKERRQ(ierr); 1710 ierr = PetscMalloc((merge->nsend+1)*sizeof(MPI_Request),&swaits);CHKERRQ(ierr); 1711 for (proc=0, k=0; proc<size; proc++) { 1712 if (!len_s[proc]) continue; 1713 i = owners_co[proc]; 1714 ierr = MPI_Isend(coj+coi[i],len_s[proc],MPIU_INT,proc,tagj,comm,swaits+k);CHKERRQ(ierr); 1715 k++; 1716 } 1717 1718 /* receives and sends of coj are complete */ 1719 ierr = PetscMalloc(size*sizeof(MPI_Status),&sstatus);CHKERRQ(ierr); 1720 for (i=0; i<merge->nrecv; i++) { 1721 PetscMPIInt icompleted; 1722 ierr = MPI_Waitany(merge->nrecv,rwaits,&icompleted,&rstatus);CHKERRQ(ierr); 1723 } 1724 ierr = PetscFree(rwaits);CHKERRQ(ierr); 1725 if (merge->nsend) {ierr = MPI_Waitall(merge->nsend,swaits,sstatus);CHKERRQ(ierr);} 1726 1727 /* send and recv coi */ 1728 /*-------------------*/ 1729 ierr = PetscCommGetNewTag(comm,&tagi);CHKERRQ(ierr); 1730 ierr = PetscPostIrecvInt(comm,tagi,merge->nrecv,merge->id_r,len_ri,&buf_ri,&rwaits);CHKERRQ(ierr); 1731 ierr = PetscMalloc((len+1)*sizeof(PetscInt),&buf_s);CHKERRQ(ierr); 1732 buf_si = buf_s; /* points to the beginning of k-th msg to be sent */ 1733 for (proc=0,k=0; proc<size; proc++) { 1734 if (!len_s[proc]) continue; 1735 /* form outgoing message for i-structure: 1736 buf_si[0]: nrows to be sent 1737 [1:nrows]: row index (global) 1738 [nrows+1:2*nrows+1]: i-structure index 1739 */ 1740 /*-------------------------------------------*/ 1741 nrows = len_si[proc]/2 - 1; 1742 buf_si_i = buf_si + nrows+1; 1743 buf_si[0] = nrows; 1744 buf_si_i[0] = 0; 1745 nrows = 0; 1746 for (i=owners_co[proc]; i<owners_co[proc+1]; i++) { 1747 nzi = coi[i+1] - coi[i]; 1748 buf_si_i[nrows+1] = buf_si_i[nrows] + nzi; /* i-structure */ 1749 buf_si[nrows+1] = prmap[i] -owners[proc]; /* local row index */ 1750 nrows++; 1751 } 1752 ierr = MPI_Isend(buf_si,len_si[proc],MPIU_INT,proc,tagi,comm,swaits+k);CHKERRQ(ierr); 1753 k++; 1754 buf_si += len_si[proc]; 1755 } 1756 i = merge->nrecv; 1757 while (i--) { 1758 PetscMPIInt icompleted; 1759 ierr = MPI_Waitany(merge->nrecv,rwaits,&icompleted,&rstatus);CHKERRQ(ierr); 1760 } 1761 ierr = PetscFree(rwaits);CHKERRQ(ierr); 1762 if (merge->nsend) {ierr = MPI_Waitall(merge->nsend,swaits,sstatus);CHKERRQ(ierr);} 1763 ierr = PetscFree(len_si);CHKERRQ(ierr); 1764 ierr = PetscFree(len_ri);CHKERRQ(ierr); 1765 ierr = PetscFree(swaits);CHKERRQ(ierr); 1766 ierr = PetscFree(sstatus);CHKERRQ(ierr); 1767 ierr = PetscFree(buf_s);CHKERRQ(ierr); 1768 1769 /* compute the local portion of C (mpi mat) */ 1770 /*------------------------------------------*/ 1771 /* allocate bi array and free space for accumulating nonzero column info */ 1772 ierr = PetscMalloc((pn+1)*sizeof(PetscInt),&bi);CHKERRQ(ierr); 1773 bi[0] = 0; 1774 1775 /* set initial free space to be fill*(nnz(P) + nnz(AP)) */ 1776 nnz = fill*(pdti[pn] + poti[pon] + ai[am]); 1777 ierr = PetscFreeSpaceGet(nnz,&free_space);CHKERRQ(ierr); 1778 current_space = free_space; 1779 1780 ierr = PetscMalloc3(merge->nrecv,PetscInt**,&buf_ri_k,merge->nrecv,PetscInt*,&nextrow,merge->nrecv,PetscInt*,&nextci);CHKERRQ(ierr); 1781 for (k=0; k<merge->nrecv; k++) { 1782 buf_ri_k[k] = buf_ri[k]; /* beginning of k-th recved i-structure */ 1783 nrows = *buf_ri_k[k]; 1784 nextrow[k] = buf_ri_k[k] + 1; /* next row number of k-th recved i-structure */ 1785 nextci[k] = buf_ri_k[k] + (nrows + 1); /* points to the next i-structure of k-th recieved i-structure */ 1786 } 1787 1788 ierr = MatPreallocateInitialize(comm,pn,A->cmap->n,dnz,onz);CHKERRQ(ierr); 1789 rmax = 0; 1790 for (i=0; i<pn; i++) { 1791 /* add pdt[i,:]*AP into lnk */ 1792 pnz = pdti[i+1] - pdti[i]; 1793 ptJ = pdtj + pdti[i]; 1794 for (j=0; j<pnz; j++) { 1795 row = ptJ[j]; /* row of AP == col of Pt */ 1796 anz = ai[row+1] - ai[row]; 1797 Jptr = aj + ai[row]; 1798 /* add non-zero cols of AP into the sorted linked list lnk */ 1799 ierr = PetscLLCondensedAddSorted_Scalable(anz,Jptr,lnk);CHKERRQ(ierr); 1800 } 1801 1802 /* add received col data into lnk */ 1803 for (k=0; k<merge->nrecv; k++) { /* k-th received message */ 1804 if (i == *nextrow[k]) { /* i-th row */ 1805 nzi = *(nextci[k]+1) - *nextci[k]; 1806 Jptr = buf_rj[k] + *nextci[k]; 1807 ierr = PetscLLCondensedAddSorted_Scalable(nzi,Jptr,lnk);CHKERRQ(ierr); 1808 nextrow[k]++; nextci[k]++; 1809 } 1810 } 1811 nnz = lnk[0]; 1812 1813 /* if free space is not available, make more free space */ 1814 if (current_space->local_remaining<nnz) { 1815 ierr = PetscFreeSpaceGet(nnz+current_space->total_array_size,¤t_space);CHKERRQ(ierr); 1816 nspacedouble++; 1817 } 1818 /* copy data into free space, then initialize lnk */ 1819 ierr = PetscLLCondensedClean_Scalable(nnz,current_space->array,lnk);CHKERRQ(ierr); 1820 ierr = MatPreallocateSet(i+owners[rank],nnz,current_space->array,dnz,onz);CHKERRQ(ierr); 1821 1822 current_space->array += nnz; 1823 current_space->local_used += nnz; 1824 current_space->local_remaining -= nnz; 1825 1826 bi[i+1] = bi[i] + nnz; 1827 if (nnz > rmax) rmax = nnz; 1828 } 1829 ierr = PetscFree3(buf_ri_k,nextrow,nextci);CHKERRQ(ierr); 1830 1831 ierr = PetscMalloc((bi[pn]+1)*sizeof(PetscInt),&bj);CHKERRQ(ierr); 1832 ierr = PetscFreeSpaceContiguous(&free_space,bj);CHKERRQ(ierr); 1833 afill_tmp = (PetscReal)bi[pn]/(pdti[pn] + poti[pon] + ai[am]+1); 1834 if (afill_tmp > afill) afill = afill_tmp; 1835 ierr = PetscLLCondensedDestroy_Scalable(lnk);CHKERRQ(ierr); 1836 ierr = MatDestroy(&POt);CHKERRQ(ierr); 1837 ierr = MatDestroy(&PDt);CHKERRQ(ierr); 1838 1839 /* create symbolic parallel matrix Cmpi - why cannot be assembled in Numeric part */ 1840 /*----------------------------------------------------------------------------------*/ 1841 ierr = PetscMalloc((rmax+1)*sizeof(PetscScalar),&vals);CHKERRQ(ierr); 1842 ierr = PetscMemzero(vals,rmax*sizeof(PetscScalar));CHKERRQ(ierr); 1843 1844 ierr = MatCreate(comm,&Cmpi);CHKERRQ(ierr); 1845 ierr = MatSetSizes(Cmpi,pn,A->cmap->n,PETSC_DETERMINE,PETSC_DETERMINE);CHKERRQ(ierr); 1846 ierr = MatSetBlockSizes(Cmpi,P->cmap->bs,A->cmap->bs);CHKERRQ(ierr); 1847 ierr = MatSetType(Cmpi,MATMPIAIJ);CHKERRQ(ierr); 1848 ierr = MatMPIAIJSetPreallocation(Cmpi,0,dnz,0,onz);CHKERRQ(ierr); 1849 ierr = MatPreallocateFinalize(dnz,onz);CHKERRQ(ierr); 1850 ierr = MatSetBlockSize(Cmpi,1);CHKERRQ(ierr); 1851 for (i=0; i<pn; i++) { 1852 row = i + rstart; 1853 nnz = bi[i+1] - bi[i]; 1854 Jptr = bj + bi[i]; 1855 ierr = MatSetValues(Cmpi,1,&row,nnz,Jptr,vals,INSERT_VALUES);CHKERRQ(ierr); 1856 } 1857 ierr = MatAssemblyBegin(Cmpi,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1858 ierr = MatAssemblyEnd(Cmpi,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1859 ierr = PetscFree(vals);CHKERRQ(ierr); 1860 1861 merge->bi = bi; 1862 merge->bj = bj; 1863 merge->coi = coi; 1864 merge->coj = coj; 1865 merge->buf_ri = buf_ri; 1866 merge->buf_rj = buf_rj; 1867 merge->owners_co = owners_co; 1868 merge->destroy = Cmpi->ops->destroy; 1869 merge->duplicate = Cmpi->ops->duplicate; 1870 1871 Cmpi->ops->mattransposemultnumeric = MatTransposeMatMultNumeric_MPIAIJ_MPIAIJ; 1872 Cmpi->ops->destroy = MatDestroy_MPIAIJ_PtAP; 1873 1874 /* attach the supporting struct to Cmpi for reuse */ 1875 c = (Mat_MPIAIJ*)Cmpi->data; 1876 1877 c->ptap = ptap; 1878 ptap->api = NULL; 1879 ptap->apj = NULL; 1880 ptap->merge = merge; 1881 ptap->rmax = rmax; 1882 ptap->apa = NULL; 1883 1884 *C = Cmpi; 1885 #if defined(PETSC_USE_INFO) 1886 if (bi[pn] != 0) { 1887 ierr = PetscInfo3(Cmpi,"Reallocs %D; Fill ratio: given %G needed %G.\n",nspacedouble,fill,afill);CHKERRQ(ierr); 1888 ierr = PetscInfo1(Cmpi,"Use MatTransposeMatMult(A,B,MatReuse,%G,&C) for best performance.\n",afill);CHKERRQ(ierr); 1889 } else { 1890 ierr = PetscInfo(Cmpi,"Empty matrix product\n");CHKERRQ(ierr); 1891 } 1892 #endif 1893 PetscFunctionReturn(0); 1894 } 1895