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 /* 12 #define DEBUG_MATMATMULT 13 */ 14 /* 15 #define DEBUG_MATTrMatMult 16 */ 17 18 #undef __FUNCT__ 19 #define __FUNCT__ "MatMatMult_MPIAIJ_MPIAIJ" 20 PetscErrorCode MatMatMult_MPIAIJ_MPIAIJ(Mat A,Mat B,MatReuse scall,PetscReal fill, Mat *C) 21 { 22 PetscErrorCode ierr; 23 24 PetscFunctionBegin; 25 if (scall == MAT_INITIAL_MATRIX){ 26 ierr = PetscLogEventBegin(MAT_MatMultSymbolic,A,B,0,0);CHKERRQ(ierr); 27 ierr = MatMatMultSymbolic_MPIAIJ_MPIAIJ(A,B,fill,C);CHKERRQ(ierr); 28 ierr = PetscLogEventEnd(MAT_MatMultSymbolic,A,B,0,0);CHKERRQ(ierr); 29 } 30 ierr = PetscLogEventBegin(MAT_MatMultNumeric,A,B,0,0);CHKERRQ(ierr); 31 ierr = (*(*C)->ops->matmultnumeric)(A,B,*C);CHKERRQ(ierr); 32 ierr = PetscLogEventEnd(MAT_MatMultNumeric,A,B,0,0);CHKERRQ(ierr); 33 PetscFunctionReturn(0); 34 } 35 36 #undef __FUNCT__ 37 #define __FUNCT__ "PetscContainerDestroy_Mat_MatMatMultMPI" 38 PetscErrorCode PetscContainerDestroy_Mat_MatMatMultMPI(void *ptr) 39 { 40 PetscErrorCode ierr; 41 Mat_MatMatMultMPI *mult=(Mat_MatMatMultMPI*)ptr; 42 43 PetscFunctionBegin; 44 ierr = ISDestroy(&mult->isrowa);CHKERRQ(ierr); 45 ierr = ISDestroy(&mult->isrowb);CHKERRQ(ierr); 46 ierr = ISDestroy(&mult->iscolb);CHKERRQ(ierr); 47 ierr = MatDestroy(&mult->C_seq);CHKERRQ(ierr); 48 ierr = MatDestroy(&mult->A_loc);CHKERRQ(ierr); 49 ierr = MatDestroy(&mult->B_seq);CHKERRQ(ierr); 50 ierr = PetscFree(mult);CHKERRQ(ierr); 51 PetscFunctionReturn(0); 52 } 53 54 #undef __FUNCT__ 55 #define __FUNCT__ "MatDestroy_MPIAIJ_MatMatMult" 56 PetscErrorCode MatDestroy_MPIAIJ_MatMatMult(Mat A) 57 { 58 PetscErrorCode ierr; 59 Mat_MPIAIJ *a=(Mat_MPIAIJ*)A->data; 60 Mat_PtAPMPI *ptap=a->ptap; 61 62 PetscFunctionBegin; 63 ierr = PetscFree2(ptap->startsj_s,ptap->startsj_r);CHKERRQ(ierr); 64 ierr = PetscFree(ptap->bufa);CHKERRQ(ierr); 65 ierr = MatDestroy(&ptap->P_loc);CHKERRQ(ierr); 66 ierr = MatDestroy(&ptap->P_oth);CHKERRQ(ierr); 67 ierr = PetscFree(ptap->api);CHKERRQ(ierr); 68 ierr = PetscFree(ptap->apj);CHKERRQ(ierr); 69 ierr = PetscFree(ptap->apa);CHKERRQ(ierr); 70 ierr = ptap->destroy(A);CHKERRQ(ierr); 71 ierr = PetscFree(ptap);CHKERRQ(ierr); 72 PetscFunctionReturn(0); 73 } 74 75 #undef __FUNCT__ 76 #define __FUNCT__ "MatDestroy_MPIAIJ_MatMatMult_32" 77 PetscErrorCode MatDestroy_MPIAIJ_MatMatMult_32(Mat A) 78 { 79 PetscErrorCode ierr; 80 PetscContainer container; 81 Mat_MatMatMultMPI *mult=PETSC_NULL; 82 83 PetscFunctionBegin; 84 ierr = PetscObjectQuery((PetscObject)A,"Mat_MatMatMultMPI",(PetscObject *)&container);CHKERRQ(ierr); 85 if (!container) SETERRQ(((PetscObject)A)->comm,PETSC_ERR_PLIB,"Container does not exist"); 86 ierr = PetscContainerGetPointer(container,(void **)&mult);CHKERRQ(ierr); 87 A->ops->destroy = mult->destroy; 88 A->ops->duplicate = mult->duplicate; 89 if (A->ops->destroy) { 90 ierr = (*A->ops->destroy)(A);CHKERRQ(ierr); 91 } 92 ierr = PetscObjectCompose((PetscObject)A,"Mat_MatMatMultMPI",0);CHKERRQ(ierr); 93 PetscFunctionReturn(0); 94 } 95 96 #undef __FUNCT__ 97 #define __FUNCT__ "MatDuplicate_MPIAIJ_MatMatMult_32" 98 PetscErrorCode MatDuplicate_MPIAIJ_MatMatMult_32(Mat A, MatDuplicateOption op, Mat *M) 99 { 100 PetscErrorCode ierr; 101 Mat_MatMatMultMPI *mult; 102 PetscContainer container; 103 104 PetscFunctionBegin; 105 ierr = PetscObjectQuery((PetscObject)A,"Mat_MatMatMultMPI",(PetscObject *)&container);CHKERRQ(ierr); 106 if (!container) SETERRQ(((PetscObject)A)->comm,PETSC_ERR_PLIB,"Container does not exist"); 107 ierr = PetscContainerGetPointer(container,(void **)&mult);CHKERRQ(ierr); 108 /* Note: the container is not duplicated, because it requires deep copying of 109 several large data sets (see PetscContainerDestroy_Mat_MatMatMultMPI()). 110 These data sets are only used for repeated calling of MatMatMultNumeric(). 111 *M is unlikely being used in this way. Thus we create *M with pure mpiaij format */ 112 ierr = (*mult->duplicate)(A,op,M);CHKERRQ(ierr); 113 (*M)->ops->destroy = mult->destroy; /* = MatDestroy_MPIAIJ, *M doesn't duplicate A's container! */ 114 (*M)->ops->duplicate = mult->duplicate; /* = MatDuplicate_MPIAIJ */ 115 PetscFunctionReturn(0); 116 } 117 118 #undef __FUNCT__ 119 #define __FUNCT__ "MatDuplicate_MPIAIJ_MatMatMult" 120 PetscErrorCode MatDuplicate_MPIAIJ_MatMatMult(Mat A, MatDuplicateOption op, Mat *M) 121 { 122 PetscErrorCode ierr; 123 Mat_MPIAIJ *a=(Mat_MPIAIJ*)A->data; 124 Mat_PtAPMPI *ptap=a->ptap; 125 126 PetscFunctionBegin; 127 ierr = (*ptap->duplicate)(A,op,M);CHKERRQ(ierr); 128 (*M)->ops->destroy = ptap->destroy; /* = MatDestroy_MPIAIJ, *M doesn't duplicate A's special structure! */ 129 (*M)->ops->duplicate = ptap->duplicate; /* = MatDuplicate_MPIAIJ */ 130 PetscFunctionReturn(0); 131 } 132 133 #undef __FUNCT__ 134 #define __FUNCT__ "MatMatMultNumeric_MPIAIJ_MPIAIJ" 135 PetscErrorCode MatMatMultNumeric_MPIAIJ_MPIAIJ(Mat A,Mat P,Mat C) 136 { 137 PetscErrorCode ierr; 138 Mat_MPIAIJ *a=(Mat_MPIAIJ*)A->data,*c=(Mat_MPIAIJ*)C->data; 139 Mat_SeqAIJ *ad=(Mat_SeqAIJ*)(a->A)->data,*ao=(Mat_SeqAIJ*)(a->B)->data; 140 Mat_SeqAIJ *cd=(Mat_SeqAIJ*)(c->A)->data,*co=(Mat_SeqAIJ*)(c->B)->data; 141 PetscInt *adi=ad->i,*adj,*aoi=ao->i,*aoj; 142 PetscScalar *ada,*aoa,*cda=cd->a,*coa=co->a; 143 Mat_SeqAIJ *p_loc,*p_oth; 144 PetscInt *pi_loc,*pj_loc,*pi_oth,*pj_oth,*pj; 145 PetscScalar *pa_loc,*pa_oth,*pa,*apa,valtmp,*ca; 146 PetscInt cm=C->rmap->n,anz,pnz; 147 Mat_PtAPMPI *ptap=c->ptap; 148 PetscInt *api,*apj,*apJ,i,j,k,row; 149 PetscInt cstart=C->cmap->rstart; 150 PetscInt cdnz,conz,k0,k1; 151 #if defined(DEBUG_MATMATMULT) 152 PetscMPIInt rank=a->rank; 153 #endif 154 155 PetscFunctionBegin; 156 #if defined(DEBUG_MATMATMULT) 157 if (!rank) ierr = PetscPrintf(PETSC_COMM_SELF,"[%d] call MatMatMultNumeric_MPIAIJ_MPIAIJ()...\n",rank); 158 #endif 159 160 /* 1) get P_oth = ptap->P_oth and P_loc = ptap->P_loc */ 161 /*-----------------------------------------------------*/ 162 /* update numerical values of P_oth and P_loc */ 163 ierr = MatGetBrowsOfAoCols_MPIAIJ(A,P,MAT_REUSE_MATRIX,&ptap->startsj_s,&ptap->startsj_r,&ptap->bufa,&ptap->P_oth);CHKERRQ(ierr); 164 ierr = MatMPIAIJGetLocalMat(P,MAT_REUSE_MATRIX,&ptap->P_loc);CHKERRQ(ierr); 165 #if defined(DEBUG_MATMATMULT) 166 if (!rank) ierr = PetscPrintf(PETSC_COMM_SELF,"[%d] got P_oth and P_loc...\n",rank); 167 #endif 168 169 /* 2) compute numeric C_loc = A_loc*P = Ad*P_loc + Ao*P_oth */ 170 /*----------------------------------------------------------*/ 171 /* get data from symbolic products */ 172 p_loc = (Mat_SeqAIJ*)(ptap->P_loc)->data; 173 p_oth = (Mat_SeqAIJ*)(ptap->P_oth)->data; 174 pi_loc=p_loc->i; pj_loc=p_loc->j; pa_loc=p_loc->a; 175 pi_oth=p_oth->i; pj_oth=p_oth->j; pa_oth=p_oth->a; 176 177 /* get apa for storing dense row A[i,:]*P */ 178 apa = ptap->apa; 179 180 api = ptap->api; 181 apj = ptap->apj; 182 for (i=0; i<cm; i++) { 183 /* diagonal portion of A */ 184 anz = adi[i+1] - adi[i]; 185 adj = ad->j + adi[i]; 186 ada = ad->a + adi[i]; 187 for (j=0; j<anz; j++) { 188 row = adj[j]; 189 pnz = pi_loc[row+1] - pi_loc[row]; 190 pj = pj_loc + pi_loc[row]; 191 pa = pa_loc + pi_loc[row]; 192 193 /* perform dense axpy */ 194 valtmp = ada[j]; 195 for (k=0; k<pnz; k++){ 196 apa[pj[k]] += valtmp*pa[k]; 197 } 198 ierr = PetscLogFlops(2.0*pnz);CHKERRQ(ierr); 199 } 200 201 /* off-diagonal portion of A */ 202 anz = aoi[i+1] - aoi[i]; 203 aoj = ao->j + aoi[i]; 204 aoa = ao->a + aoi[i]; 205 for (j=0; j<anz; j++) { 206 row = aoj[j]; 207 pnz = pi_oth[row+1] - pi_oth[row]; 208 pj = pj_oth + pi_oth[row]; 209 pa = pa_oth + pi_oth[row]; 210 211 /* perform dense axpy */ 212 valtmp = aoa[j]; 213 for (k=0; k<pnz; k++){ 214 apa[pj[k]] += valtmp*pa[k]; 215 } 216 ierr = PetscLogFlops(2.0*pnz);CHKERRQ(ierr); 217 } 218 219 /* set values in C */ 220 apJ = apj + api[i]; 221 cdnz = cd->i[i+1] - cd->i[i]; 222 conz = co->i[i+1] - co->i[i]; 223 224 /* 1st off-diagoanl part of C */ 225 ca = coa + co->i[i]; 226 k = 0; 227 for (k0=0; k0<conz; k0++){ 228 if (apJ[k] >= cstart) break; 229 ca[k0] = apa[apJ[k]]; 230 apa[apJ[k]] = 0.0; 231 k++; 232 } 233 234 /* diagonal part of C */ 235 ca = cda + cd->i[i]; 236 for (k1=0; k1<cdnz; k1++){ 237 ca[k1] = apa[apJ[k]]; 238 apa[apJ[k]] = 0.0; 239 k++; 240 } 241 242 /* 2nd off-diagoanl part of C */ 243 ca = coa + co->i[i]; 244 for (; k0<conz; k0++){ 245 ca[k0] = apa[apJ[k]]; 246 apa[apJ[k]] = 0.0; 247 k++; 248 } 249 } 250 ierr = MatAssemblyBegin(C,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 251 ierr = MatAssemblyEnd(C,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 252 PetscFunctionReturn(0); 253 } 254 255 #undef __FUNCT__ 256 #define __FUNCT__ "MatMatMultSymbolic_MPIAIJ_MPIAIJ" 257 PetscErrorCode MatMatMultSymbolic_MPIAIJ_MPIAIJ(Mat A,Mat P,PetscReal fill,Mat *C) 258 { 259 PetscErrorCode ierr; 260 MPI_Comm comm=((PetscObject)A)->comm; 261 Mat Cmpi; 262 Mat_PtAPMPI *ptap; 263 PetscFreeSpaceList free_space=PETSC_NULL,current_space=PETSC_NULL; 264 Mat_MPIAIJ *a=(Mat_MPIAIJ*)A->data,*c; 265 Mat_SeqAIJ *ad=(Mat_SeqAIJ*)(a->A)->data,*ao=(Mat_SeqAIJ*)(a->B)->data,*p_loc,*p_oth; 266 PetscInt *pi_loc,*pj_loc,*pi_oth,*pj_oth,*dnz,*onz; 267 PetscInt *adi=ad->i,*adj=ad->j,*aoi=ao->i,*aoj=ao->j,rstart=A->rmap->rstart; 268 PetscInt *lnk,i,pnz,row,*api,*apj,*Jptr,apnz,nspacedouble=0,j,nzi; 269 PetscInt am=A->rmap->n,pN=P->cmap->N,pn=P->cmap->n,pm=P->rmap->n; 270 PetscBT lnkbt; 271 PetscScalar *apa; 272 PetscReal afill; 273 PetscBool scalable=PETSC_FALSE; 274 PetscInt nlnk_max,armax,prmax; 275 #if defined(DEBUG_MATMATMULT) 276 PetscMPIInt rank=a->rank; 277 #endif 278 279 PetscFunctionBegin; 280 if (A->cmap->rstart != P->rmap->rstart || A->cmap->rend != P->rmap->rend){ 281 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); 282 } 283 284 ierr = PetscObjectOptionsBegin((PetscObject)A);CHKERRQ(ierr); 285 286 ierr = PetscOptionsBool("-matmatmult_32","Use a scalable but slower C=A*B","",scalable,&scalable,PETSC_NULL);CHKERRQ(ierr); 287 if (scalable){ 288 ierr = MatMatMultSymbolic_MPIAIJ_MPIAIJ_Scalable_32(A,P,fill,C);;CHKERRQ(ierr); 289 PetscFunctionReturn(0); 290 } 291 ierr = PetscOptionsBool("-matmatmult_scalable","Use a scalable but slower C=A*B","",scalable,&scalable,PETSC_NULL);CHKERRQ(ierr); 292 if (scalable){ 293 ierr = MatMatMultSymbolic_MPIAIJ_MPIAIJ_Scalable(A,P,fill,C);;CHKERRQ(ierr); 294 PetscFunctionReturn(0); 295 } 296 ierr = PetscOptionsEnd();CHKERRQ(ierr); 297 298 #if defined(DEBUG_MATMATMULT) 299 if (!rank) ierr = PetscPrintf(PETSC_COMM_SELF,"[%d] call MatMatMultSymbolic_MPIAIJ_MPIAIJ()...\n",rank); 300 #endif 301 302 /* create struct Mat_PtAPMPI and attached it to C later */ 303 ierr = PetscNew(Mat_PtAPMPI,&ptap);CHKERRQ(ierr); 304 305 /* get P_oth by taking rows of P (= non-zero cols of local A) from other processors */ 306 ierr = MatGetBrowsOfAoCols_MPIAIJ(A,P,MAT_INITIAL_MATRIX,&ptap->startsj_s,&ptap->startsj_r,&ptap->bufa,&ptap->P_oth);CHKERRQ(ierr); 307 #if defined(DEBUG_MATMATMULT) 308 if (!rank) ierr = PetscPrintf(PETSC_COMM_SELF,"[%d] P_oth is done...\n",rank); 309 #endif 310 /* get P_loc by taking all local rows of P */ 311 ierr = MatMPIAIJGetLocalMat(P,MAT_INITIAL_MATRIX,&ptap->P_loc);CHKERRQ(ierr); 312 313 p_loc = (Mat_SeqAIJ*)(ptap->P_loc)->data; 314 p_oth = (Mat_SeqAIJ*)(ptap->P_oth)->data; 315 pi_loc = p_loc->i; pj_loc = p_loc->j; 316 pi_oth = p_oth->i; pj_oth = p_oth->j; 317 318 #if defined(DEBUG_MATMATMULT) 319 if (!rank) ierr = PetscPrintf(PETSC_COMM_SELF,"[%d] P_loc is done, Annz %D * P_locnnz %D = %D...\n",rank,ad->i[am]+ao->i[am],p_loc->rmax,(ad->i[am]+ao->i[am])*p_loc->rmax); 320 #endif 321 322 /* first, compute symbolic AP = A_loc*P = A_diag*P_loc + A_off*P_oth */ 323 /*-------------------------------------------------------------------*/ 324 ierr = PetscMalloc((am+2)*sizeof(PetscInt),&api);CHKERRQ(ierr); 325 ptap->api = api; 326 api[0] = 0; 327 328 /* create and initialize a linked list */ 329 armax = ad->rmax+ao->rmax; 330 prmax = PetscMax(p_loc->rmax,p_oth->rmax); 331 nlnk_max = armax*prmax; 332 if (!nlnk_max || nlnk_max > pN) nlnk_max = pN; 333 #if defined(DEBUG_MATMATMULT) 334 if (!rank) ierr = PetscPrintf(PETSC_COMM_SELF,"[%d] pN %d; nlnk_max %d; Armax %d+%d=%d; Prmax Max(%d,%d)=%d\n",rank,pN,nlnk_max,ad->rmax,ao->rmax,armax,p_loc->rmax,p_oth->rmax,prmax); 335 #endif 336 ierr = PetscLLCondensedCreate(nlnk_max,pN,&lnk,&lnkbt);CHKERRQ(ierr); 337 338 /* Initial FreeSpace size is fill*(nnz(A)+nnz(P)) */ 339 ierr = PetscFreeSpaceGet((PetscInt)(fill*(adi[am]+aoi[am]+pi_loc[pm])),&free_space);CHKERRQ(ierr); 340 current_space = free_space; 341 342 ierr = MatPreallocateInitialize(comm,am,pn,dnz,onz);CHKERRQ(ierr); 343 for (i=0; i<am; i++) { 344 apnz = 0; 345 /* diagonal portion of A */ 346 nzi = adi[i+1] - adi[i]; 347 for (j=0; j<nzi; j++){ 348 row = *adj++; 349 pnz = pi_loc[row+1] - pi_loc[row]; 350 Jptr = pj_loc + pi_loc[row]; 351 /* add non-zero cols of P into the sorted linked list lnk */ 352 ierr = PetscLLCondensedAddSorted(pnz,Jptr,lnk,lnkbt);CHKERRQ(ierr); 353 } 354 /* off-diagonal portion of A */ 355 nzi = aoi[i+1] - aoi[i]; 356 for (j=0; j<nzi; j++){ 357 row = *aoj++; 358 pnz = pi_oth[row+1] - pi_oth[row]; 359 Jptr = pj_oth + pi_oth[row]; 360 ierr = PetscLLCondensedAddSorted(pnz,Jptr,lnk,lnkbt);CHKERRQ(ierr); 361 } 362 363 apnz = lnk[0]; 364 api[i+1] = api[i] + apnz; 365 366 /* if free space is not available, double the total space in the list */ 367 if (current_space->local_remaining<apnz) { 368 ierr = PetscFreeSpaceGet(apnz+current_space->total_array_size,¤t_space);CHKERRQ(ierr); 369 nspacedouble++; 370 } 371 372 /* Copy data into free space, then initialize lnk */ 373 ierr = PetscLLCondensedClean(pN,apnz,current_space->array,lnk,lnkbt);CHKERRQ(ierr); 374 ierr = MatPreallocateSet(i+rstart,apnz,current_space->array,dnz,onz);CHKERRQ(ierr); 375 current_space->array += apnz; 376 current_space->local_used += apnz; 377 current_space->local_remaining -= apnz; 378 } 379 380 /* Allocate space for apj, initialize apj, and */ 381 /* destroy list of free space and other temporary array(s) */ 382 ierr = PetscMalloc((api[am]+1)*sizeof(PetscInt),&ptap->apj);CHKERRQ(ierr); 383 apj = ptap->apj; 384 ierr = PetscFreeSpaceContiguous(&free_space,ptap->apj);CHKERRQ(ierr); 385 ierr = PetscLLDestroy(lnk,lnkbt);CHKERRQ(ierr); 386 #if defined(DEBUG_MATMATMULT) 387 if (!rank) ierr = PetscPrintf(PETSC_COMM_SELF,"[%d] AP is done...\n",rank); 388 #endif 389 390 /* malloc apa to store dense row A[i,:]*P */ 391 ierr = PetscMalloc(pN*sizeof(PetscScalar),&apa);CHKERRQ(ierr); 392 ierr = PetscMemzero(apa,pN*sizeof(PetscScalar));CHKERRQ(ierr); 393 ptap->apa = apa; 394 #if defined(DEBUG_MATMATMULT) 395 if (!rank) ierr = PetscPrintf(PETSC_COMM_SELF,"[%d] Malloc apa pN %D is done...\n",rank,pN); 396 #endif 397 398 /* create and assemble symbolic parallel matrix Cmpi */ 399 /*----------------------------------------------------*/ 400 ierr = MatCreate(comm,&Cmpi);CHKERRQ(ierr); 401 ierr = MatSetSizes(Cmpi,am,pn,PETSC_DETERMINE,PETSC_DETERMINE);CHKERRQ(ierr); 402 ierr = MatSetType(Cmpi,MATMPIAIJ);CHKERRQ(ierr); 403 ierr = MatMPIAIJSetPreallocation(Cmpi,0,dnz,0,onz);CHKERRQ(ierr); 404 ierr = MatPreallocateFinalize(dnz,onz);CHKERRQ(ierr); 405 for (i=0; i<am; i++){ 406 row = i + rstart; 407 apnz = api[i+1] - api[i]; 408 ierr = MatSetValues(Cmpi,1,&row,apnz,apj,apa,INSERT_VALUES);CHKERRQ(ierr); 409 apj += apnz; 410 } 411 ierr = MatAssemblyBegin(Cmpi,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 412 ierr = MatAssemblyEnd(Cmpi,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 413 414 ptap->destroy = Cmpi->ops->destroy; 415 ptap->duplicate = Cmpi->ops->duplicate; 416 Cmpi->ops->matmultnumeric = MatMatMultNumeric_MPIAIJ_MPIAIJ; 417 Cmpi->ops->destroy = MatDestroy_MPIAIJ_MatMatMult; 418 Cmpi->ops->duplicate = MatDuplicate_MPIAIJ_MatMatMult; 419 420 /* attach the supporting struct to Cmpi for reuse */ 421 c = (Mat_MPIAIJ*)Cmpi->data; 422 c->ptap = ptap; 423 424 *C = Cmpi; 425 426 /* set MatInfo */ 427 afill = (PetscReal)api[am]/(adi[am]+aoi[am]+pi_loc[pm]) + 1.e-5; 428 if (afill < 1.0) afill = 1.0; 429 Cmpi->info.mallocs = nspacedouble; 430 Cmpi->info.fill_ratio_given = fill; 431 Cmpi->info.fill_ratio_needed = afill; 432 433 #if defined(PETSC_USE_INFO) 434 if (api[am]) { 435 ierr = PetscInfo3(Cmpi,"Reallocs %D; Fill ratio: given %G needed %G.\n",nspacedouble,fill,afill);CHKERRQ(ierr); 436 ierr = PetscInfo1(Cmpi,"Use MatMatMult(A,B,MatReuse,%G,&C) for best performance.;\n",afill);CHKERRQ(ierr); 437 } else { 438 ierr = PetscInfo(Cmpi,"Empty matrix product\n");CHKERRQ(ierr); 439 } 440 #endif 441 PetscFunctionReturn(0); 442 } 443 444 /* implementation used in PETSc-3.2 */ 445 /* This routine is called ONLY in the case of reusing previously computed symbolic C */ 446 #undef __FUNCT__ 447 #define __FUNCT__ "MatMatMultNumeric_MPIAIJ_MPIAIJ_Scalable_32" 448 PetscErrorCode MatMatMultNumeric_MPIAIJ_MPIAIJ_Scalable_32(Mat A,Mat B,Mat C) 449 { 450 PetscErrorCode ierr; 451 Mat *seq; 452 Mat_MatMatMultMPI *mult; 453 PetscContainer container; 454 455 PetscFunctionBegin; 456 ierr = PetscObjectQuery((PetscObject)C,"Mat_MatMatMultMPI",(PetscObject *)&container);CHKERRQ(ierr); 457 if (!container) SETERRQ(((PetscObject)A)->comm,PETSC_ERR_PLIB,"Container does not exist"); 458 ierr = PetscContainerGetPointer(container,(void **)&mult);CHKERRQ(ierr); 459 460 if (mult->skipNumeric){ /* first numeric product is done during symbolic product */ 461 mult->skipNumeric = PETSC_FALSE; 462 PetscFunctionReturn(0); 463 } 464 #if defined(DEBUG_MATMATMULT) 465 PetscMPIInt rank; 466 MPI_Comm comm = ((PetscObject)C)->comm; 467 ierr = MPI_Comm_rank(comm,&rank);CHKERRQ(ierr); 468 ierr = MPI_Barrier(comm);CHKERRQ(ierr); 469 if (!rank) ierr = PetscPrintf(PETSC_COMM_SELF,"[%d] call MatMatMultNumeric_MPIAIJ_MPIAIJ_Scalable_32()...\n",rank); 470 #endif 471 472 seq = &mult->B_seq; 473 ierr = MatGetSubMatrices(B,1,&mult->isrowb,&mult->iscolb,MAT_REUSE_MATRIX,&seq);CHKERRQ(ierr); 474 mult->B_seq = *seq; 475 476 seq = &mult->A_loc; 477 ierr = MatGetSubMatrices(A,1,&mult->isrowa,&mult->isrowb,MAT_REUSE_MATRIX,&seq);CHKERRQ(ierr); 478 mult->A_loc = *seq; 479 480 ierr = MatMatMultNumeric_SeqAIJ_SeqAIJ_Scalable(mult->A_loc,mult->B_seq,mult->C_seq);CHKERRQ(ierr); 481 482 ierr = PetscObjectReference((PetscObject)mult->C_seq);CHKERRQ(ierr); 483 ierr = MatMerge(((PetscObject)A)->comm,mult->C_seq,B->cmap->n,MAT_REUSE_MATRIX,&C);CHKERRQ(ierr); 484 PetscFunctionReturn(0); 485 } 486 487 /* numeric product is computed as well */ 488 #undef __FUNCT__ 489 #define __FUNCT__ "MatMatMultSymbolic_MPIAIJ_MPIAIJ_Scalable_32" 490 PetscErrorCode MatMatMultSymbolic_MPIAIJ_MPIAIJ_Scalable_32(Mat A,Mat B,PetscReal fill,Mat *C) 491 { 492 PetscErrorCode ierr; 493 Mat_MatMatMultMPI *mult; 494 PetscContainer container; 495 Mat AB,*seq; 496 Mat_MPIAIJ *a=(Mat_MPIAIJ*)A->data; 497 PetscInt *idx,i,start,ncols,nzA,nzB,*cmap,imark; 498 #if defined(DEBUG_MATMATMULT) 499 MPI_Comm comm = ((PetscObject)A)->comm; 500 PetscMPIInt rank=a->rank; 501 #endif 502 503 PetscFunctionBegin; 504 #if defined(DEBUG_MATMATMULT) 505 if (!rank) ierr = PetscPrintf(PETSC_COMM_SELF,"[%d] call MatMatMultSymbolic_MPIAIJ_MPIAIJ_Scalable_32()...\n",rank); 506 #endif 507 if (A->cmap->rstart != B->rmap->rstart || A->cmap->rend != B->rmap->rend){ 508 SETERRQ4(((PetscObject)A)->comm,PETSC_ERR_ARG_SIZ,"Matrix local dimensions are incompatible, (%D, %D) != (%D,%D)",A->cmap->rstart,A->cmap->rend,B->rmap->rstart,B->rmap->rend); 509 } 510 511 ierr = PetscNew(Mat_MatMatMultMPI,&mult);CHKERRQ(ierr); 512 513 /* get isrowb: nonzero col of A */ 514 start = A->cmap->rstart; 515 cmap = a->garray; 516 nzA = a->A->cmap->n; 517 nzB = a->B->cmap->n; 518 ierr = PetscMalloc((nzA+nzB)*sizeof(PetscInt), &idx);CHKERRQ(ierr); 519 ncols = 0; 520 for (i=0; i<nzB; i++) { /* row < local row index */ 521 if (cmap[i] < start) idx[ncols++] = cmap[i]; 522 else break; 523 } 524 imark = i; 525 for (i=0; i<nzA; i++) idx[ncols++] = start + i; /* local rows */ 526 for (i=imark; i<nzB; i++) idx[ncols++] = cmap[i]; /* row > local row index */ 527 ierr = ISCreateGeneral(PETSC_COMM_SELF,ncols,idx,PETSC_OWN_POINTER,&mult->isrowb);CHKERRQ(ierr); 528 ierr = ISCreateStride(PETSC_COMM_SELF,B->cmap->N,0,1,&mult->iscolb);CHKERRQ(ierr); 529 530 /* get isrowa: all local rows of A */ 531 ierr = ISCreateStride(PETSC_COMM_SELF,A->rmap->n,A->rmap->rstart,1,&mult->isrowa);CHKERRQ(ierr); 532 533 /* Below should go to MatMatMultNumeric_MPIAIJ_MPIAIJ() - How to generate C there? */ 534 /* create a seq matrix B_seq = submatrix of B by taking rows of B that equal to nonzero col of A */ 535 ierr = MatGetSubMatrices(B,1,&mult->isrowb,&mult->iscolb,MAT_INITIAL_MATRIX,&seq);CHKERRQ(ierr); 536 mult->B_seq = *seq; 537 ierr = PetscFree(seq);CHKERRQ(ierr); 538 #if defined(DEBUG_MATMATMULT) 539 if (!rank) ierr = PetscPrintf(PETSC_COMM_SELF,"[%d] B_seq is done...\n",rank); 540 #endif 541 542 /* create a seq matrix A_seq = submatrix of A by taking all local rows of A */ 543 ierr = MatGetSubMatrices(A,1,&mult->isrowa,&mult->isrowb,MAT_INITIAL_MATRIX,&seq);CHKERRQ(ierr); 544 mult->A_loc = *seq; 545 ierr = PetscFree(seq);CHKERRQ(ierr); 546 #if defined(DEBUG_MATMATMULT) 547 if (!rank) ierr = PetscPrintf(PETSC_COMM_SELF,"[%d] A_loc is done...\n",rank); 548 #endif 549 550 /* compute C_seq = A_seq * B_seq */ 551 ierr = MatMatMultSymbolic_SeqAIJ_SeqAIJ_Scalable(mult->A_loc,mult->B_seq,fill,&mult->C_seq);CHKERRQ(ierr); 552 #if defined(DEBUG_MATMATMULT) 553 ierr = MPI_Barrier(comm);CHKERRQ(ierr); 554 if (!rank) ierr = PetscPrintf(PETSC_COMM_SELF,"[%d] C_seq Symbolic is done...\n",rank); 555 #endif 556 ierr = MatMatMultNumeric_SeqAIJ_SeqAIJ_Scalable(mult->A_loc,mult->B_seq,mult->C_seq);CHKERRQ(ierr); 557 #if defined(DEBUG_MATMATMULT) 558 if (!rank) ierr = PetscPrintf(PETSC_COMM_SELF,"[%d] C_seq Numeric is done...\n",rank); 559 #endif 560 561 /* create mpi matrix C by concatinating C_seq */ 562 ierr = PetscObjectReference((PetscObject)mult->C_seq);CHKERRQ(ierr); /* prevent C_seq being destroyed by MatMerge() */ 563 ierr = MatMergeSymbolic(((PetscObject)A)->comm,mult->C_seq,B->cmap->n,&AB);CHKERRQ(ierr); 564 ierr = MatMergeNumeric(((PetscObject)A)->comm,mult->C_seq,B->cmap->n,AB);CHKERRQ(ierr); 565 #if defined(DEBUG_MATMATMULT) 566 if (!rank) ierr = PetscPrintf(PETSC_COMM_SELF,"[%d] Merge is done...\n",rank); 567 #endif 568 569 /* attach the supporting struct to C for reuse of symbolic C */ 570 ierr = PetscContainerCreate(PETSC_COMM_SELF,&container);CHKERRQ(ierr); 571 ierr = PetscContainerSetPointer(container,mult);CHKERRQ(ierr); 572 ierr = PetscContainerSetUserDestroy(container,PetscContainerDestroy_Mat_MatMatMultMPI);CHKERRQ(ierr); 573 ierr = PetscObjectCompose((PetscObject)AB,"Mat_MatMatMultMPI",(PetscObject)container);CHKERRQ(ierr); 574 ierr = PetscContainerDestroy(&container);CHKERRQ(ierr); 575 mult->skipNumeric = PETSC_TRUE; /* a numeric product is done here */ 576 mult->destroy = AB->ops->destroy; 577 mult->duplicate = AB->ops->duplicate; 578 AB->ops->matmultnumeric = MatMatMultNumeric_MPIAIJ_MPIAIJ_Scalable_32; 579 AB->ops->destroy = MatDestroy_MPIAIJ_MatMatMult_32; 580 AB->ops->duplicate = MatDuplicate_MPIAIJ_MatMatMult_32; 581 AB->ops->matmult = MatMatMult_MPIAIJ_MPIAIJ; 582 *C = AB; 583 PetscFunctionReturn(0); 584 } 585 586 #undef __FUNCT__ 587 #define __FUNCT__ "MatMatMult_MPIAIJ_MPIDense" 588 PetscErrorCode MatMatMult_MPIAIJ_MPIDense(Mat A,Mat B,MatReuse scall,PetscReal fill,Mat *C) 589 { 590 PetscErrorCode ierr; 591 592 PetscFunctionBegin; 593 if (scall == MAT_INITIAL_MATRIX){ 594 ierr = MatMatMultSymbolic_MPIAIJ_MPIDense(A,B,fill,C);CHKERRQ(ierr); 595 } 596 ierr = MatMatMultNumeric_MPIAIJ_MPIDense(A,B,*C);CHKERRQ(ierr); 597 PetscFunctionReturn(0); 598 } 599 600 typedef struct { 601 Mat workB; 602 PetscScalar *rvalues,*svalues; 603 MPI_Request *rwaits,*swaits; 604 } MPIAIJ_MPIDense; 605 606 #undef __FUNCT__ 607 #define __FUNCT__ "MPIAIJ_MPIDenseDestroy" 608 PetscErrorCode MPIAIJ_MPIDenseDestroy(void *ctx) 609 { 610 MPIAIJ_MPIDense *contents = (MPIAIJ_MPIDense*) ctx; 611 PetscErrorCode ierr; 612 613 PetscFunctionBegin; 614 ierr = MatDestroy(&contents->workB);CHKERRQ(ierr); 615 ierr = PetscFree4(contents->rvalues,contents->svalues,contents->rwaits,contents->swaits);CHKERRQ(ierr); 616 ierr = PetscFree(contents);CHKERRQ(ierr); 617 PetscFunctionReturn(0); 618 } 619 620 #undef __FUNCT__ 621 #define __FUNCT__ "MatMatMultSymbolic_MPIAIJ_MPIDense" 622 PetscErrorCode MatMatMultSymbolic_MPIAIJ_MPIDense(Mat A,Mat B,PetscReal fill,Mat *C) 623 { 624 PetscErrorCode ierr; 625 Mat_MPIAIJ *aij = (Mat_MPIAIJ*) A->data; 626 PetscInt nz = aij->B->cmap->n; 627 PetscContainer container; 628 MPIAIJ_MPIDense *contents; 629 VecScatter ctx = aij->Mvctx; 630 VecScatter_MPI_General *from = (VecScatter_MPI_General*) ctx->fromdata; 631 VecScatter_MPI_General *to = ( VecScatter_MPI_General*) ctx->todata; 632 PetscInt m=A->rmap->n,n=B->cmap->n; 633 634 PetscFunctionBegin; 635 ierr = MatCreate(((PetscObject)B)->comm,C);CHKERRQ(ierr); 636 ierr = MatSetSizes(*C,m,n,A->rmap->N,B->cmap->N);CHKERRQ(ierr); 637 ierr = MatSetType(*C,MATMPIDENSE);CHKERRQ(ierr); 638 ierr = MatMPIDenseSetPreallocation(*C,PETSC_NULL);CHKERRQ(ierr); 639 ierr = MatAssemblyBegin(*C,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 640 ierr = MatAssemblyEnd(*C,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 641 (*C)->ops->matmult = MatMatMult_MPIAIJ_MPIDense; 642 643 ierr = PetscNew(MPIAIJ_MPIDense,&contents);CHKERRQ(ierr); 644 /* Create work matrix used to store off processor rows of B needed for local product */ 645 ierr = MatCreateSeqDense(PETSC_COMM_SELF,nz,B->cmap->N,PETSC_NULL,&contents->workB);CHKERRQ(ierr); 646 /* Create work arrays needed */ 647 ierr = PetscMalloc4(B->cmap->N*from->starts[from->n],PetscScalar,&contents->rvalues, 648 B->cmap->N*to->starts[to->n],PetscScalar,&contents->svalues, 649 from->n,MPI_Request,&contents->rwaits, 650 to->n,MPI_Request,&contents->swaits);CHKERRQ(ierr); 651 652 ierr = PetscContainerCreate(((PetscObject)A)->comm,&container);CHKERRQ(ierr); 653 ierr = PetscContainerSetPointer(container,contents);CHKERRQ(ierr); 654 ierr = PetscContainerSetUserDestroy(container,MPIAIJ_MPIDenseDestroy);CHKERRQ(ierr); 655 ierr = PetscObjectCompose((PetscObject)(*C),"workB",(PetscObject)container);CHKERRQ(ierr); 656 ierr = PetscContainerDestroy(&container);CHKERRQ(ierr); 657 PetscFunctionReturn(0); 658 } 659 660 #undef __FUNCT__ 661 #define __FUNCT__ "MatMPIDenseScatter" 662 /* 663 Performs an efficient scatter on the rows of B needed by this process; this is 664 a modification of the VecScatterBegin_() routines. 665 */ 666 PetscErrorCode MatMPIDenseScatter(Mat A,Mat B,Mat C,Mat *outworkB) 667 { 668 Mat_MPIAIJ *aij = (Mat_MPIAIJ*)A->data; 669 PetscErrorCode ierr; 670 PetscScalar *b,*w,*svalues,*rvalues; 671 VecScatter ctx = aij->Mvctx; 672 VecScatter_MPI_General *from = (VecScatter_MPI_General*) ctx->fromdata; 673 VecScatter_MPI_General *to = ( VecScatter_MPI_General*) ctx->todata; 674 PetscInt i,j,k; 675 PetscInt *sindices,*sstarts,*rindices,*rstarts; 676 PetscMPIInt *sprocs,*rprocs,nrecvs; 677 MPI_Request *swaits,*rwaits; 678 MPI_Comm comm = ((PetscObject)A)->comm; 679 PetscMPIInt tag = ((PetscObject)ctx)->tag,ncols = B->cmap->N, nrows = aij->B->cmap->n,imdex,nrowsB = B->rmap->n; 680 MPI_Status status; 681 MPIAIJ_MPIDense *contents; 682 PetscContainer container; 683 Mat workB; 684 685 PetscFunctionBegin; 686 ierr = PetscObjectQuery((PetscObject)C,"workB",(PetscObject*)&container);CHKERRQ(ierr); 687 if (!container) SETERRQ(comm,PETSC_ERR_PLIB,"Container does not exist"); 688 ierr = PetscContainerGetPointer(container,(void**)&contents);CHKERRQ(ierr); 689 690 workB = *outworkB = contents->workB; 691 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); 692 sindices = to->indices; 693 sstarts = to->starts; 694 sprocs = to->procs; 695 swaits = contents->swaits; 696 svalues = contents->svalues; 697 698 rindices = from->indices; 699 rstarts = from->starts; 700 rprocs = from->procs; 701 rwaits = contents->rwaits; 702 rvalues = contents->rvalues; 703 704 ierr = MatGetArray(B,&b);CHKERRQ(ierr); 705 ierr = MatGetArray(workB,&w);CHKERRQ(ierr); 706 707 for (i=0; i<from->n; i++) { 708 ierr = MPI_Irecv(rvalues+ncols*rstarts[i],ncols*(rstarts[i+1]-rstarts[i]),MPIU_SCALAR,rprocs[i],tag,comm,rwaits+i);CHKERRQ(ierr); 709 } 710 711 for (i=0; i<to->n; i++) { 712 /* pack a message at a time */ 713 CHKMEMQ; 714 for (j=0; j<sstarts[i+1]-sstarts[i]; j++){ 715 for (k=0; k<ncols; k++) { 716 svalues[ncols*(sstarts[i] + j) + k] = b[sindices[sstarts[i]+j] + nrowsB*k]; 717 } 718 } 719 CHKMEMQ; 720 ierr = MPI_Isend(svalues+ncols*sstarts[i],ncols*(sstarts[i+1]-sstarts[i]),MPIU_SCALAR,sprocs[i],tag,comm,swaits+i);CHKERRQ(ierr); 721 } 722 723 nrecvs = from->n; 724 while (nrecvs) { 725 ierr = MPI_Waitany(from->n,rwaits,&imdex,&status);CHKERRQ(ierr); 726 nrecvs--; 727 /* unpack a message at a time */ 728 CHKMEMQ; 729 for (j=0; j<rstarts[imdex+1]-rstarts[imdex]; j++){ 730 for (k=0; k<ncols; k++) { 731 w[rindices[rstarts[imdex]+j] + nrows*k] = rvalues[ncols*(rstarts[imdex] + j) + k]; 732 } 733 } 734 CHKMEMQ; 735 } 736 if (to->n) {ierr = MPI_Waitall(to->n,swaits,to->sstatus);CHKERRQ(ierr);} 737 738 ierr = MatRestoreArray(B,&b);CHKERRQ(ierr); 739 ierr = MatRestoreArray(workB,&w);CHKERRQ(ierr); 740 ierr = MatAssemblyBegin(workB,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 741 ierr = MatAssemblyEnd(workB,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 742 PetscFunctionReturn(0); 743 } 744 extern PetscErrorCode MatMatMultNumericAdd_SeqAIJ_SeqDense(Mat,Mat,Mat); 745 746 #undef __FUNCT__ 747 #define __FUNCT__ "MatMatMultNumeric_MPIAIJ_MPIDense" 748 PetscErrorCode MatMatMultNumeric_MPIAIJ_MPIDense(Mat A,Mat B,Mat C) 749 { 750 PetscErrorCode ierr; 751 Mat_MPIAIJ *aij = (Mat_MPIAIJ*)A->data; 752 Mat_MPIDense *bdense = (Mat_MPIDense*)B->data; 753 Mat_MPIDense *cdense = (Mat_MPIDense*)C->data; 754 Mat workB; 755 756 PetscFunctionBegin; 757 758 /* diagonal block of A times all local rows of B*/ 759 ierr = MatMatMultNumeric_SeqAIJ_SeqDense(aij->A,bdense->A,cdense->A);CHKERRQ(ierr); 760 761 /* get off processor parts of B needed to complete the product */ 762 ierr = MatMPIDenseScatter(A,B,C,&workB);CHKERRQ(ierr); 763 764 /* off-diagonal block of A times nonlocal rows of B */ 765 ierr = MatMatMultNumericAdd_SeqAIJ_SeqDense(aij->B,workB,cdense->A);CHKERRQ(ierr); 766 ierr = MatAssemblyBegin(C,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 767 ierr = MatAssemblyEnd(C,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 768 PetscFunctionReturn(0); 769 } 770 771 #undef __FUNCT__ 772 #define __FUNCT__ "MatMatMultNumeric_MPIAIJ_MPIAIJ_Scalable" 773 PetscErrorCode MatMatMultNumeric_MPIAIJ_MPIAIJ_Scalable(Mat A,Mat P,Mat C) 774 { 775 PetscErrorCode ierr; 776 Mat_MPIAIJ *a=(Mat_MPIAIJ*)A->data,*c=(Mat_MPIAIJ*)C->data; 777 Mat_SeqAIJ *ad=(Mat_SeqAIJ*)(a->A)->data,*ao=(Mat_SeqAIJ*)(a->B)->data; 778 Mat_SeqAIJ *cd=(Mat_SeqAIJ*)(c->A)->data,*co=(Mat_SeqAIJ*)(c->B)->data; 779 PetscInt *adi=ad->i,*adj,*aoi=ao->i,*aoj; 780 PetscScalar *ada,*aoa,*cda=cd->a,*coa=co->a; 781 Mat_SeqAIJ *p_loc,*p_oth; 782 PetscInt *pi_loc,*pj_loc,*pi_oth,*pj_oth,*pj; 783 PetscScalar *pa_loc,*pa_oth,*pa,valtmp,*ca; 784 PetscInt cm=C->rmap->n,anz,pnz; 785 Mat_PtAPMPI *ptap=c->ptap; 786 PetscScalar *apa_sparse=ptap->apa; 787 PetscInt *api,*apj,*apJ,i,j,k,row; 788 PetscInt cstart=C->cmap->rstart; 789 PetscInt cdnz,conz,k0,k1,nextp; 790 #if defined(DEBUG_MATMATMULT) 791 PetscMPIInt rank=a->rank; 792 #endif 793 794 PetscFunctionBegin; 795 #if defined(DEBUG_MATMATMULT) 796 if (!rank) ierr = PetscPrintf(PETSC_COMM_SELF,"[%d] call MatMatMultNumeric_MPIAIJ_MPIAIJ_Scalable()...\n",rank); 797 #endif 798 799 /* 1) get P_oth = ptap->P_oth and P_loc = ptap->P_loc */ 800 /*-----------------------------------------------------*/ 801 /* update numerical values of P_oth and P_loc */ 802 ierr = MatGetBrowsOfAoCols_MPIAIJ(A,P,MAT_REUSE_MATRIX,&ptap->startsj_s,&ptap->startsj_r,&ptap->bufa,&ptap->P_oth);CHKERRQ(ierr); 803 ierr = MatMPIAIJGetLocalMat(P,MAT_REUSE_MATRIX,&ptap->P_loc);CHKERRQ(ierr); 804 #if defined(DEBUG_MATMATMULT) 805 if (!rank) ierr = PetscPrintf(PETSC_COMM_SELF,"[%d] got P_oth and P_loc...\n",rank); 806 #endif 807 808 /* 2) compute numeric C_loc = A_loc*P = Ad*P_loc + Ao*P_oth */ 809 /*----------------------------------------------------------*/ 810 /* get data from symbolic products */ 811 p_loc = (Mat_SeqAIJ*)(ptap->P_loc)->data; 812 p_oth = (Mat_SeqAIJ*)(ptap->P_oth)->data; 813 pi_loc=p_loc->i; pj_loc=p_loc->j; pa_loc=p_loc->a; 814 pi_oth=p_oth->i; pj_oth=p_oth->j; pa_oth=p_oth->a; 815 816 api = ptap->api; 817 apj = ptap->apj; 818 for (i=0; i<cm; i++) { 819 apJ = apj + api[i]; 820 821 /* diagonal portion of A */ 822 anz = adi[i+1] - adi[i]; 823 adj = ad->j + adi[i]; 824 ada = ad->a + adi[i]; 825 for (j=0; j<anz; j++) { 826 row = adj[j]; 827 pnz = pi_loc[row+1] - pi_loc[row]; 828 pj = pj_loc + pi_loc[row]; 829 pa = pa_loc + pi_loc[row]; 830 /* perform sparse axpy */ 831 valtmp = ada[j]; 832 nextp = 0; 833 for (k=0; nextp<pnz; k++) { 834 if (apJ[k] == pj[nextp]) { /* column of AP == column of P */ 835 apa_sparse[k] += valtmp*pa[nextp++]; 836 } 837 } 838 ierr = PetscLogFlops(2.0*pnz);CHKERRQ(ierr); 839 } 840 841 /* off-diagonal portion of A */ 842 anz = aoi[i+1] - aoi[i]; 843 aoj = ao->j + aoi[i]; 844 aoa = ao->a + aoi[i]; 845 for (j=0; j<anz; j++) { 846 row = aoj[j]; 847 pnz = pi_oth[row+1] - pi_oth[row]; 848 pj = pj_oth + pi_oth[row]; 849 pa = pa_oth + pi_oth[row]; 850 /* perform sparse axpy */ 851 valtmp = aoa[j]; 852 nextp = 0; 853 for (k=0; nextp<pnz; k++) { 854 if (apJ[k] == pj[nextp]) { /* column of AP == column of P */ 855 apa_sparse[k] += valtmp*pa[nextp++]; 856 } 857 } 858 ierr = PetscLogFlops(2.0*pnz);CHKERRQ(ierr); 859 } 860 861 /* set values in C */ 862 cdnz = cd->i[i+1] - cd->i[i]; 863 conz = co->i[i+1] - co->i[i]; 864 865 /* 1st off-diagoanl part of C */ 866 ca = coa + co->i[i]; 867 k = 0; 868 for (k0=0; k0<conz; k0++){ 869 if (apJ[k] >= cstart) break; 870 ca[k0] = apa_sparse[k]; 871 apa_sparse[k] = 0.0; 872 k++; 873 } 874 875 /* diagonal part of C */ 876 ca = cda + cd->i[i]; 877 for (k1=0; k1<cdnz; k1++){ 878 ca[k1] = apa_sparse[k]; 879 apa_sparse[k] = 0.0; 880 k++; 881 } 882 883 /* 2nd off-diagoanl part of C */ 884 ca = coa + co->i[i]; 885 for (; k0<conz; k0++){ 886 ca[k0] = apa_sparse[k]; 887 apa_sparse[k] = 0.0; 888 k++; 889 } 890 } 891 ierr = MatAssemblyBegin(C,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 892 ierr = MatAssemblyEnd(C,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 893 PetscFunctionReturn(0); 894 } 895 896 /* same as MatMatMultSymbolic_MPIAIJ_MPIAIJ(), except using LLCondensed to avoid O(BN) memory requirement */ 897 #undef __FUNCT__ 898 #define __FUNCT__ "MatMatMultSymbolic_MPIAIJ_MPIAIJ_Scalable" 899 PetscErrorCode MatMatMultSymbolic_MPIAIJ_MPIAIJ_Scalable(Mat A,Mat P,PetscReal fill,Mat *C) 900 { 901 PetscErrorCode ierr; 902 MPI_Comm comm=((PetscObject)A)->comm; 903 Mat Cmpi; 904 Mat_PtAPMPI *ptap; 905 PetscFreeSpaceList free_space=PETSC_NULL,current_space=PETSC_NULL; 906 Mat_MPIAIJ *a=(Mat_MPIAIJ*)A->data,*c; 907 Mat_SeqAIJ *ad=(Mat_SeqAIJ*)(a->A)->data,*ao=(Mat_SeqAIJ*)(a->B)->data,*p_loc,*p_oth; 908 PetscInt *pi_loc,*pj_loc,*pi_oth,*pj_oth,*dnz,*onz; 909 PetscInt *adi=ad->i,*adj=ad->j,*aoi=ao->i,*aoj=ao->j,rstart=A->rmap->rstart; 910 PetscInt i,pnz,row,*api,*apj,*Jptr,apnz,nspacedouble=0,j,nzi,*lnk,apnz_max=0; 911 PetscInt am=A->rmap->n,pN=P->cmap->N,pn=P->cmap->n,pm=P->rmap->n; 912 PetscInt nlnk_max,armax,prmax; 913 PetscReal afill; 914 PetscScalar *apa; 915 #if defined(DEBUG_MATMATMULT) 916 PetscMPIInt rank=a->rank; 917 #endif 918 919 PetscFunctionBegin; 920 #if defined(DEBUG_MATMATMULT) 921 ierr = MPI_Comm_rank(comm,&rank);CHKERRQ(ierr); 922 if (!rank) ierr = PetscPrintf(PETSC_COMM_SELF,"[%d] call MatMatMultSymbolic_MPIAIJ_MPIAIJ_Scalable()...\n",rank); 923 #endif 924 925 /* create struct Mat_PtAPMPI and attached it to C later */ 926 ierr = PetscNew(Mat_PtAPMPI,&ptap);CHKERRQ(ierr); 927 928 /* get P_oth by taking rows of P (= non-zero cols of local A) from other processors */ 929 ierr = MatGetBrowsOfAoCols_MPIAIJ(A,P,MAT_INITIAL_MATRIX,&ptap->startsj_s,&ptap->startsj_r,&ptap->bufa,&ptap->P_oth);CHKERRQ(ierr); 930 #if defined(DEBUG_MATMATMULT) 931 if (!rank) ierr = PetscPrintf(PETSC_COMM_SELF,"[%d] P_oth is done...\n",rank); 932 #endif 933 /* get P_loc by taking all local rows of P */ 934 ierr = MatMPIAIJGetLocalMat(P,MAT_INITIAL_MATRIX,&ptap->P_loc);CHKERRQ(ierr); 935 936 p_loc = (Mat_SeqAIJ*)(ptap->P_loc)->data; 937 p_oth = (Mat_SeqAIJ*)(ptap->P_oth)->data; 938 pi_loc = p_loc->i; pj_loc = p_loc->j; 939 pi_oth = p_oth->i; pj_oth = p_oth->j; 940 941 #if defined(DEBUG_MATMATMULT) 942 if (!rank) ierr = PetscPrintf(PETSC_COMM_SELF,"[%d] P_loc is done, Annz %D * P_locnnz %D = %D...\n",rank,ad->i[am]+ao->i[am],p_loc->rmax,(ad->i[am]+ao->i[am])*p_loc->rmax); 943 #endif 944 945 /* first, compute symbolic AP = A_loc*P = A_diag*P_loc + A_off*P_oth */ 946 /*-------------------------------------------------------------------*/ 947 ierr = PetscMalloc((am+2)*sizeof(PetscInt),&api);CHKERRQ(ierr); 948 ptap->api = api; 949 api[0] = 0; 950 951 /* create and initialize a linked list */ 952 armax = ad->rmax+ao->rmax; 953 prmax = PetscMax(p_loc->rmax,p_oth->rmax); 954 nlnk_max = armax*prmax; 955 if (!nlnk_max || nlnk_max > pN) nlnk_max = pN; 956 #if defined(DEBUG_MATMATMULT) 957 if (!rank) ierr = PetscPrintf(PETSC_COMM_SELF,"[%d] pN %d; nlnk_max %d; Armax %d+%d=%d; Prmax Max(%d,%d)=%d\n",rank,pN,nlnk_max,ad->rmax,ao->rmax,armax,p_loc->rmax,p_oth->rmax,prmax); 958 #endif 959 ierr = PetscLLCondensedCreate_Scalable(nlnk_max,&lnk);CHKERRQ(ierr); 960 961 /* Initial FreeSpace size is fill*(nnz(A)+nnz(P)) */ 962 ierr = PetscFreeSpaceGet((PetscInt)(fill*(adi[am]+aoi[am]+pi_loc[pm])),&free_space);CHKERRQ(ierr); 963 current_space = free_space; 964 965 ierr = MatPreallocateInitialize(comm,am,pn,dnz,onz);CHKERRQ(ierr); 966 for (i=0; i<am; i++) { 967 apnz = 0; 968 /* diagonal portion of A */ 969 nzi = adi[i+1] - adi[i]; 970 for (j=0; j<nzi; j++){ 971 row = *adj++; 972 pnz = pi_loc[row+1] - pi_loc[row]; 973 Jptr = pj_loc + pi_loc[row]; 974 /* add non-zero cols of P into the sorted linked list lnk */ 975 ierr = PetscLLCondensedAddSorted_Scalable(pnz,Jptr,lnk);CHKERRQ(ierr); 976 } 977 /* off-diagonal portion of A */ 978 nzi = aoi[i+1] - aoi[i]; 979 for (j=0; j<nzi; j++){ 980 row = *aoj++; 981 pnz = pi_oth[row+1] - pi_oth[row]; 982 Jptr = pj_oth + pi_oth[row]; 983 ierr = PetscLLCondensedAddSorted_Scalable(pnz,Jptr,lnk);CHKERRQ(ierr); 984 } 985 986 apnz = *lnk; 987 api[i+1] = api[i] + apnz; 988 if (apnz > apnz_max) apnz_max = apnz; 989 990 /* if free space is not available, double the total space in the list */ 991 if (current_space->local_remaining<apnz) { 992 ierr = PetscFreeSpaceGet(apnz+current_space->total_array_size,¤t_space);CHKERRQ(ierr); 993 nspacedouble++; 994 } 995 996 /* Copy data into free space, then initialize lnk */ 997 ierr = PetscLLCondensedClean_Scalable(apnz,current_space->array,lnk);CHKERRQ(ierr); 998 ierr = MatPreallocateSet(i+rstart,apnz,current_space->array,dnz,onz);CHKERRQ(ierr); 999 current_space->array += apnz; 1000 current_space->local_used += apnz; 1001 current_space->local_remaining -= apnz; 1002 } 1003 1004 /* Allocate space for apj, initialize apj, and */ 1005 /* destroy list of free space and other temporary array(s) */ 1006 ierr = PetscMalloc((api[am]+1)*sizeof(PetscInt),&ptap->apj);CHKERRQ(ierr); 1007 apj = ptap->apj; 1008 ierr = PetscFreeSpaceContiguous(&free_space,ptap->apj);CHKERRQ(ierr); 1009 ierr = PetscLLCondensedDestroy_Scalable(lnk);CHKERRQ(ierr); 1010 #if defined(DEBUG_MATMATMULT) 1011 if (!rank) ierr = PetscPrintf(PETSC_COMM_SELF,"[%d] AP is done..., apnz_max %d\n",rank,apnz_max); 1012 #endif 1013 1014 /* create and assemble symbolic parallel matrix Cmpi */ 1015 /*----------------------------------------------------*/ 1016 ierr = MatCreate(comm,&Cmpi);CHKERRQ(ierr); 1017 ierr = MatSetSizes(Cmpi,am,pn,PETSC_DETERMINE,PETSC_DETERMINE);CHKERRQ(ierr); 1018 ierr = MatSetType(Cmpi,MATMPIAIJ);CHKERRQ(ierr); 1019 ierr = MatMPIAIJSetPreallocation(Cmpi,0,dnz,0,onz);CHKERRQ(ierr); 1020 ierr = MatPreallocateFinalize(dnz,onz);CHKERRQ(ierr); 1021 ierr = MatSetBlockSize(Cmpi,1);CHKERRQ(ierr); 1022 1023 /* malloc apa for assembly Cmpi */ 1024 ierr = PetscMalloc(apnz_max*sizeof(PetscScalar),&apa);CHKERRQ(ierr); 1025 ierr = PetscMemzero(apa,apnz_max*sizeof(PetscScalar));CHKERRQ(ierr); 1026 ptap->apa = apa; 1027 for (i=0; i<am; i++){ 1028 row = i + rstart; 1029 apnz = api[i+1] - api[i]; 1030 ierr = MatSetValues(Cmpi,1,&row,apnz,apj,apa,INSERT_VALUES);CHKERRQ(ierr); 1031 apj += apnz; 1032 } 1033 ierr = MatAssemblyBegin(Cmpi,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1034 ierr = MatAssemblyEnd(Cmpi,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1035 1036 ptap->destroy = Cmpi->ops->destroy; 1037 ptap->duplicate = Cmpi->ops->duplicate; 1038 Cmpi->ops->matmultnumeric = MatMatMultNumeric_MPIAIJ_MPIAIJ_Scalable; 1039 Cmpi->ops->destroy = MatDestroy_MPIAIJ_MatMatMult; 1040 Cmpi->ops->duplicate = MatDuplicate_MPIAIJ_MatMatMult; 1041 1042 /* attach the supporting struct to Cmpi for reuse */ 1043 c = (Mat_MPIAIJ*)Cmpi->data; 1044 c->ptap = ptap; 1045 1046 *C = Cmpi; 1047 1048 /* set MatInfo */ 1049 afill = (PetscReal)api[am]/(adi[am]+aoi[am]+pi_loc[pm]) + 1.e-5; 1050 if (afill < 1.0) afill = 1.0; 1051 Cmpi->info.mallocs = nspacedouble; 1052 Cmpi->info.fill_ratio_given = fill; 1053 Cmpi->info.fill_ratio_needed = afill; 1054 1055 #if defined(PETSC_USE_INFO) 1056 if (api[am]) { 1057 ierr = PetscInfo3(Cmpi,"Reallocs %D; Fill ratio: given %G needed %G.\n",nspacedouble,fill,afill);CHKERRQ(ierr); 1058 ierr = PetscInfo1(Cmpi,"Use MatMatMult(A,B,MatReuse,%G,&C) for best performance.;\n",afill);CHKERRQ(ierr); 1059 } else { 1060 ierr = PetscInfo(Cmpi,"Empty matrix product\n");CHKERRQ(ierr); 1061 } 1062 #endif 1063 PetscFunctionReturn(0); 1064 } 1065 1066 /*-------------------------------------------------------------------------*/ 1067 #undef __FUNCT__ 1068 #define __FUNCT__ "MatTransposeMatMult_MPIAIJ_MPIAIJ" 1069 PetscErrorCode MatTransposeMatMult_MPIAIJ_MPIAIJ(Mat P,Mat A,MatReuse scall,PetscReal fill,Mat *C) 1070 { 1071 PetscErrorCode ierr; 1072 PetscBool scalable=PETSC_FALSE; 1073 1074 PetscFunctionBegin; 1075 if (scall == MAT_INITIAL_MATRIX){ 1076 ierr = PetscObjectOptionsBegin((PetscObject)A);CHKERRQ(ierr); 1077 ierr = PetscOptionsBool("-mattransposematmult_scalable","Use a scalable but slower C=Pt*A","",scalable,&scalable,PETSC_NULL);CHKERRQ(ierr); 1078 if (scalable){ 1079 ierr = MatTransposeMatMultSymbolic_MPIAIJ_MPIAIJ_Scalable(P,A,fill,C);CHKERRQ(ierr); 1080 } else { 1081 ierr = MatTransposeMatMultSymbolic_MPIAIJ_MPIAIJ(P,A,fill,C);CHKERRQ(ierr); 1082 } 1083 ierr = PetscOptionsEnd();CHKERRQ(ierr); 1084 } 1085 ierr = (*(*C)->ops->mattransposemultnumeric)(P,A,*C);CHKERRQ(ierr); 1086 PetscFunctionReturn(0); 1087 } 1088 1089 #undef __FUNCT__ 1090 #define __FUNCT__ "MatTransposeMatMultNumeric_MPIAIJ_MPIAIJ" 1091 PetscErrorCode MatTransposeMatMultNumeric_MPIAIJ_MPIAIJ(Mat P,Mat A,Mat C) 1092 { 1093 PetscErrorCode ierr; 1094 Mat_Merge_SeqsToMPI *merge; 1095 Mat_MPIAIJ *p=(Mat_MPIAIJ*)P->data,*c=(Mat_MPIAIJ*)C->data; 1096 Mat_SeqAIJ *pd=(Mat_SeqAIJ*)(p->A)->data,*po=(Mat_SeqAIJ*)(p->B)->data; 1097 Mat_PtAPMPI *ptap; 1098 PetscInt *adj,*aJ; 1099 PetscInt i,j,k,anz,pnz,row,*cj; 1100 MatScalar *ada,*aval,*ca,valtmp; 1101 PetscInt am=A->rmap->n,cm=C->rmap->n,pon=(p->B)->cmap->n; 1102 MPI_Comm comm=((PetscObject)C)->comm; 1103 PetscMPIInt size,rank,taga,*len_s; 1104 PetscInt *owners,proc,nrows,**buf_ri_k,**nextrow,**nextci; 1105 PetscInt **buf_ri,**buf_rj; 1106 PetscInt cnz=0,*bj_i,*bi,*bj,bnz,nextcj; /* bi,bj,ba: local array of C(mpi mat) */ 1107 MPI_Request *s_waits,*r_waits; 1108 MPI_Status *status; 1109 MatScalar **abuf_r,*ba_i,*pA,*coa,*ba; 1110 PetscInt *ai,*aj,*coi,*coj; 1111 PetscInt *poJ=po->j,*pdJ=pd->j; 1112 Mat A_loc; 1113 Mat_SeqAIJ *a_loc; 1114 1115 PetscFunctionBegin; 1116 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr); 1117 ierr = MPI_Comm_rank(comm,&rank);CHKERRQ(ierr); 1118 1119 ptap = c->ptap; 1120 merge = ptap->merge; 1121 1122 #if defined(DEBUG_MATTrMatMult) 1123 ierr = PetscSynchronizedPrintf(comm,"[%d] TransposeMatMultNumeric: Crmax %d \n",rank,ptap->rmax); 1124 ierr = PetscSynchronizedFlush(comm);CHKERRQ(ierr); 1125 #endif 1126 1127 /* 2) compute numeric C_seq = P_loc^T*A_loc*P - dominating part */ 1128 /*--------------------------------------------------------------*/ 1129 /* get data from symbolic products */ 1130 coi = merge->coi; coj = merge->coj; 1131 ierr = PetscMalloc((coi[pon]+1)*sizeof(MatScalar),&coa);CHKERRQ(ierr); 1132 ierr = PetscMemzero(coa,coi[pon]*sizeof(MatScalar));CHKERRQ(ierr); 1133 1134 bi = merge->bi; bj = merge->bj; 1135 owners = merge->rowmap->range; 1136 ierr = PetscMalloc((bi[cm]+1)*sizeof(MatScalar),&ba);CHKERRQ(ierr); 1137 ierr = PetscMemzero(ba,bi[cm]*sizeof(MatScalar));CHKERRQ(ierr); 1138 1139 /* get A_loc by taking all local rows of A */ 1140 A_loc = ptap->A_loc; 1141 ierr = MatMPIAIJGetLocalMat(A,MAT_REUSE_MATRIX,&A_loc);CHKERRQ(ierr); 1142 a_loc = (Mat_SeqAIJ*)(A_loc)->data; 1143 ai = a_loc->i; 1144 aj = a_loc->j; 1145 1146 ierr = PetscMalloc((A->cmap->N)*sizeof(PetscScalar),&aval);CHKERRQ(ierr); /* non-scalable!!! */ 1147 ierr = PetscMemzero(aval,A->cmap->N*sizeof(PetscScalar));CHKERRQ(ierr); 1148 1149 for (i=0; i<am; i++) { 1150 /* 2-a) put A[i,:] to dense array aval */ 1151 anz = ai[i+1] - ai[i]; 1152 adj = aj + ai[i]; 1153 ada = a_loc->a + ai[i]; 1154 for (j=0; j<anz; j++){ 1155 aval[adj[j]] = ada[j]; 1156 } 1157 1158 /* 2-b) Compute Cseq = P_loc[i,:]^T*A[i,:] using outer product */ 1159 /*--------------------------------------------------------------*/ 1160 /* put the value into Co=(p->B)^T*A (off-diagonal part, send to others) */ 1161 pnz = po->i[i+1] - po->i[i]; 1162 poJ = po->j + po->i[i]; 1163 pA = po->a + po->i[i]; 1164 for (j=0; j<pnz; j++){ 1165 row = poJ[j]; 1166 cnz = coi[row+1] - coi[row]; 1167 cj = coj + coi[row]; 1168 ca = coa + coi[row]; 1169 /* perform dense axpy */ 1170 valtmp = pA[j]; 1171 for (k=0; k<cnz; k++) { 1172 ca[k] += valtmp*aval[cj[k]]; 1173 } 1174 ierr = PetscLogFlops(2.0*cnz);CHKERRQ(ierr); 1175 } 1176 1177 /* put the value into Cd (diagonal part) */ 1178 pnz = pd->i[i+1] - pd->i[i]; 1179 pdJ = pd->j + pd->i[i]; 1180 pA = pd->a + pd->i[i]; 1181 for (j=0; j<pnz; j++){ 1182 row = pdJ[j]; 1183 cnz = bi[row+1] - bi[row]; 1184 cj = bj + bi[row]; 1185 ca = ba + bi[row]; 1186 /* perform dense axpy */ 1187 valtmp = pA[j]; 1188 for (k=0; k<cnz; k++) { 1189 ca[k] += valtmp*aval[cj[k]]; 1190 } 1191 ierr = PetscLogFlops(2.0*cnz);CHKERRQ(ierr); 1192 } 1193 1194 /* zero the current row of Pt*A */ 1195 aJ = aj + ai[i]; 1196 for (k=0; k<anz; k++) aval[aJ[k]] = 0.0; 1197 } 1198 1199 /* 3) send and recv matrix values coa */ 1200 /*------------------------------------*/ 1201 buf_ri = merge->buf_ri; 1202 buf_rj = merge->buf_rj; 1203 len_s = merge->len_s; 1204 ierr = PetscCommGetNewTag(comm,&taga);CHKERRQ(ierr); 1205 ierr = PetscPostIrecvScalar(comm,taga,merge->nrecv,merge->id_r,merge->len_r,&abuf_r,&r_waits);CHKERRQ(ierr); 1206 1207 ierr = PetscMalloc2(merge->nsend+1,MPI_Request,&s_waits,size,MPI_Status,&status);CHKERRQ(ierr); 1208 for (proc=0,k=0; proc<size; proc++){ 1209 if (!len_s[proc]) continue; 1210 i = merge->owners_co[proc]; 1211 ierr = MPI_Isend(coa+coi[i],len_s[proc],MPIU_MATSCALAR,proc,taga,comm,s_waits+k);CHKERRQ(ierr); 1212 k++; 1213 } 1214 if (merge->nrecv) {ierr = MPI_Waitall(merge->nrecv,r_waits,status);CHKERRQ(ierr);} 1215 if (merge->nsend) {ierr = MPI_Waitall(merge->nsend,s_waits,status);CHKERRQ(ierr);} 1216 1217 ierr = PetscFree2(s_waits,status);CHKERRQ(ierr); 1218 ierr = PetscFree(r_waits);CHKERRQ(ierr); 1219 ierr = PetscFree(coa);CHKERRQ(ierr); 1220 1221 /* 4) insert local Cseq and received values into Cmpi */ 1222 /*----------------------------------------------------*/ 1223 ierr = PetscMalloc3(merge->nrecv,PetscInt**,&buf_ri_k,merge->nrecv,PetscInt*,&nextrow,merge->nrecv,PetscInt*,&nextci);CHKERRQ(ierr); 1224 for (k=0; k<merge->nrecv; k++){ 1225 buf_ri_k[k] = buf_ri[k]; /* beginning of k-th recved i-structure */ 1226 nrows = *(buf_ri_k[k]); 1227 nextrow[k] = buf_ri_k[k]+1; /* next row number of k-th recved i-structure */ 1228 nextci[k] = buf_ri_k[k] + (nrows + 1);/* poins to the next i-structure of k-th recved i-structure */ 1229 } 1230 1231 for (i=0; i<cm; i++) { 1232 row = owners[rank] + i; /* global row index of C_seq */ 1233 bj_i = bj + bi[i]; /* col indices of the i-th row of C */ 1234 ba_i = ba + bi[i]; 1235 bnz = bi[i+1] - bi[i]; 1236 /* add received vals into ba */ 1237 for (k=0; k<merge->nrecv; k++){ /* k-th received message */ 1238 /* i-th row */ 1239 if (i == *nextrow[k]) { 1240 cnz = *(nextci[k]+1) - *nextci[k]; 1241 cj = buf_rj[k] + *(nextci[k]); 1242 ca = abuf_r[k] + *(nextci[k]); 1243 nextcj = 0; 1244 for (j=0; nextcj<cnz; j++){ 1245 if (bj_i[j] == cj[nextcj]){ /* bcol == ccol */ 1246 ba_i[j] += ca[nextcj++]; 1247 } 1248 } 1249 nextrow[k]++; nextci[k]++; 1250 ierr = PetscLogFlops(2.0*cnz);CHKERRQ(ierr); 1251 } 1252 } 1253 ierr = MatSetValues(C,1,&row,bnz,bj_i,ba_i,INSERT_VALUES);CHKERRQ(ierr); 1254 } 1255 ierr = MatAssemblyBegin(C,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1256 ierr = MatAssemblyEnd(C,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1257 1258 ierr = PetscFree(ba);CHKERRQ(ierr); 1259 ierr = PetscFree(abuf_r[0]);CHKERRQ(ierr); 1260 ierr = PetscFree(abuf_r);CHKERRQ(ierr); 1261 ierr = PetscFree3(buf_ri_k,nextrow,nextci);CHKERRQ(ierr); 1262 ierr = PetscFree(aval);CHKERRQ(ierr); 1263 PetscFunctionReturn(0); 1264 } 1265 1266 /* This routine is modified from MatPtAPSymbolic_MPIAIJ_MPIAIJ() */ 1267 #undef __FUNCT__ 1268 #define __FUNCT__ "MatTransposeMatMultSymbolic_MPIAIJ_MPIAIJ" 1269 PetscErrorCode MatTransposeMatMultSymbolic_MPIAIJ_MPIAIJ(Mat P,Mat A,PetscReal fill,Mat *C) 1270 { 1271 PetscErrorCode ierr; 1272 Mat Cmpi,A_loc,POt,PDt; 1273 Mat_PtAPMPI *ptap; 1274 PetscFreeSpaceList free_space=PETSC_NULL,current_space=PETSC_NULL; 1275 Mat_MPIAIJ *p=(Mat_MPIAIJ*)P->data,*c; 1276 PetscInt *pdti,*pdtj,*poti,*potj,*ptJ; 1277 PetscInt nnz; 1278 PetscInt *lnk,*owners_co,*coi,*coj,i,k,pnz,row; 1279 PetscInt am=A->rmap->n,pn=P->cmap->n; 1280 PetscBT lnkbt; 1281 MPI_Comm comm=((PetscObject)A)->comm; 1282 PetscMPIInt size,rank,tagi,tagj,*len_si,*len_s,*len_ri; 1283 PetscInt **buf_rj,**buf_ri,**buf_ri_k; 1284 PetscInt len,proc,*dnz,*onz,*owners; 1285 PetscInt nzi,*bi,*bj; 1286 PetscInt nrows,*buf_s,*buf_si,*buf_si_i,**nextrow,**nextci; 1287 MPI_Request *swaits,*rwaits; 1288 MPI_Status *sstatus,rstatus; 1289 Mat_Merge_SeqsToMPI *merge; 1290 PetscInt *ai,*aj,*Jptr,anz,*prmap=p->garray,pon,nspacedouble=0,j; 1291 PetscReal afill=1.0,afill_tmp; 1292 PetscInt rstart = P->cmap->rstart,rmax,aN=A->cmap->N,Crmax; 1293 PetscScalar *vals; 1294 Mat_SeqAIJ *a_loc, *pdt,*pot; 1295 1296 PetscFunctionBegin; 1297 /* check if matrix local sizes are compatible */ 1298 if (A->rmap->rstart != P->rmap->rstart || A->rmap->rend != P->rmap->rend){ 1299 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); 1300 } 1301 1302 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr); 1303 ierr = MPI_Comm_rank(comm,&rank);CHKERRQ(ierr); 1304 #if defined(DEBUG_MATTrMatMult) 1305 ierr = PetscSynchronizedPrintf(comm,"[%d] TransposeMatMultSymbolic P: %d %d, %d %d; A %d %d, %d %d\n",rank,P->rmap->N,P->cmap->N,P->rmap->n,P->cmap->n,A->rmap->N,aN,A->rmap->n,A->cmap->n); 1306 ierr = PetscSynchronizedFlush(comm);CHKERRQ(ierr); 1307 #endif 1308 1309 /* create struct Mat_PtAPMPI and attached it to C later */ 1310 ierr = PetscNew(Mat_PtAPMPI,&ptap);CHKERRQ(ierr); 1311 1312 /* get A_loc by taking all local rows of A */ 1313 ierr = MatMPIAIJGetLocalMat(A,MAT_INITIAL_MATRIX,&A_loc);CHKERRQ(ierr); 1314 ptap->A_loc = A_loc; 1315 a_loc = (Mat_SeqAIJ*)(A_loc)->data; 1316 ai = a_loc->i; 1317 aj = a_loc->j; 1318 1319 /* determine symbolic Co=(p->B)^T*A - send to others */ 1320 /*----------------------------------------------------*/ 1321 ierr = MatTransposeSymbolic_SeqAIJ(p->A,&PDt);CHKERRQ(ierr); 1322 pdt = (Mat_SeqAIJ*)PDt->data; 1323 pdti = pdt->i; pdtj = pdt->j; 1324 1325 ierr = MatTransposeSymbolic_SeqAIJ(p->B,&POt);CHKERRQ(ierr); 1326 pot = (Mat_SeqAIJ*)POt->data; 1327 poti = pot->i; potj = pot->j; 1328 1329 /* then, compute symbolic Co = (p->B)^T*A */ 1330 pon = (p->B)->cmap->n; /* total num of rows to be sent to other processors 1331 >= (num of nonzero rows of C_seq) - pn */ 1332 ierr = PetscMalloc((pon+1)*sizeof(PetscInt),&coi);CHKERRQ(ierr); 1333 coi[0] = 0; 1334 1335 /* set initial free space to be fill*(nnz(p->B) + nnz(A)) */ 1336 nnz = fill*(poti[pon] + ai[am]); 1337 ierr = PetscFreeSpaceGet(nnz,&free_space); 1338 current_space = free_space; 1339 #if defined(DEBUG_MATTrMatMult) 1340 ierr = PetscSynchronizedPrintf(comm, "[%d] nnz = fill %g *(%d + %d)\n",rank,fill,poti[pon],ai[am]);CHKERRQ(ierr); 1341 ierr = PetscSynchronizedFlush(comm);CHKERRQ(ierr); 1342 #endif 1343 /* create and initialize a linked list */ 1344 i = PetscMax(pdt->rmax,pot->rmax); 1345 Crmax = i*a_loc->rmax*size; 1346 if (!Crmax || Crmax > aN) Crmax = aN; 1347 #if defined(DEBUG_MATTrMatMult) 1348 printf("[%d] rmax A_loc %d * max(PD %d, PO %d)=%d, Crmax %d\n",rank,a_loc->rmax,pdt->rmax,pot->rmax,i*a_loc->rmax,Crmax); 1349 #endif 1350 ierr = PetscLLCondensedCreate(Crmax,aN,&lnk,&lnkbt);CHKERRQ(ierr); 1351 1352 for (i=0; i<pon; i++) { 1353 pnz = poti[i+1] - poti[i]; 1354 ptJ = potj + poti[i]; 1355 for (j=0; j<pnz; j++){ 1356 row = ptJ[j]; /* row of A_loc == col of Pot */ 1357 anz = ai[row+1] - ai[row]; 1358 Jptr = aj + ai[row]; 1359 /* add non-zero cols of AP into the sorted linked list lnk */ 1360 ierr = PetscLLCondensedAddSorted(anz,Jptr,lnk,lnkbt);CHKERRQ(ierr); 1361 } 1362 nnz = lnk[0]; 1363 1364 /* If free space is not available, double the total space in the list */ 1365 if (current_space->local_remaining<nnz) { 1366 ierr = PetscFreeSpaceGet(nnz+current_space->total_array_size,¤t_space);CHKERRQ(ierr); 1367 nspacedouble++; 1368 } 1369 1370 /* Copy data into free space, and zero out denserows */ 1371 ierr = PetscLLCondensedClean(aN,nnz,current_space->array,lnk,lnkbt);CHKERRQ(ierr); 1372 current_space->array += nnz; 1373 current_space->local_used += nnz; 1374 current_space->local_remaining -= nnz; 1375 coi[i+1] = coi[i] + nnz; 1376 } 1377 1378 ierr = PetscMalloc((coi[pon]+1)*sizeof(PetscInt),&coj);CHKERRQ(ierr); 1379 ierr = PetscFreeSpaceContiguous(&free_space,coj);CHKERRQ(ierr); 1380 afill_tmp = (PetscReal)coi[pon]/(poti[pon] + ai[am]); 1381 if (afill_tmp > afill) afill = afill_tmp; 1382 1383 /* send j-array (coj) of Co to other processors */ 1384 /*----------------------------------------------*/ 1385 /* determine row ownership */ 1386 ierr = PetscNew(Mat_Merge_SeqsToMPI,&merge);CHKERRQ(ierr); 1387 ierr = PetscLayoutCreate(comm,&merge->rowmap);CHKERRQ(ierr); 1388 merge->rowmap->n = pn; 1389 merge->rowmap->bs = 1; 1390 ierr = PetscLayoutSetUp(merge->rowmap);CHKERRQ(ierr); 1391 owners = merge->rowmap->range; 1392 1393 /* determine the number of messages to send, their lengths */ 1394 ierr = PetscMalloc(size*sizeof(PetscMPIInt),&len_si);CHKERRQ(ierr); 1395 ierr = PetscMemzero(len_si,size*sizeof(PetscMPIInt));CHKERRQ(ierr); 1396 ierr = PetscMalloc(size*sizeof(PetscMPIInt),&merge->len_s);CHKERRQ(ierr); 1397 len_s = merge->len_s; 1398 merge->nsend = 0; 1399 1400 ierr = PetscMalloc((size+2)*sizeof(PetscInt),&owners_co);CHKERRQ(ierr); 1401 ierr = PetscMemzero(len_s,size*sizeof(PetscMPIInt));CHKERRQ(ierr); 1402 1403 proc = 0; 1404 for (i=0; i<pon; i++){ 1405 while (prmap[i] >= owners[proc+1]) proc++; 1406 len_si[proc]++; /* num of rows in Co to be sent to [proc] */ 1407 len_s[proc] += coi[i+1] - coi[i]; 1408 } 1409 1410 len = 0; /* max length of buf_si[] */ 1411 owners_co[0] = 0; 1412 for (proc=0; proc<size; proc++){ 1413 owners_co[proc+1] = owners_co[proc] + len_si[proc]; 1414 if (len_si[proc]){ 1415 merge->nsend++; 1416 len_si[proc] = 2*(len_si[proc] + 1); 1417 len += len_si[proc]; 1418 } 1419 } 1420 1421 /* determine the number and length of messages to receive for coi and coj */ 1422 ierr = PetscGatherNumberOfMessages(comm,PETSC_NULL,len_s,&merge->nrecv);CHKERRQ(ierr); 1423 ierr = PetscGatherMessageLengths2(comm,merge->nsend,merge->nrecv,len_s,len_si,&merge->id_r,&merge->len_r,&len_ri);CHKERRQ(ierr); 1424 1425 /* post the Irecv and Isend of coj */ 1426 ierr = PetscCommGetNewTag(comm,&tagj);CHKERRQ(ierr); 1427 ierr = PetscPostIrecvInt(comm,tagj,merge->nrecv,merge->id_r,merge->len_r,&buf_rj,&rwaits);CHKERRQ(ierr); 1428 ierr = PetscMalloc((merge->nsend+1)*sizeof(MPI_Request),&swaits);CHKERRQ(ierr); 1429 for (proc=0, k=0; proc<size; proc++){ 1430 if (!len_s[proc]) continue; 1431 i = owners_co[proc]; 1432 ierr = MPI_Isend(coj+coi[i],len_s[proc],MPIU_INT,proc,tagj,comm,swaits+k);CHKERRQ(ierr); 1433 k++; 1434 } 1435 1436 /* receives and sends of coj are complete */ 1437 ierr = PetscMalloc(size*sizeof(MPI_Status),&sstatus);CHKERRQ(ierr); 1438 for (i=0; i<merge->nrecv; i++){ 1439 PetscMPIInt icompleted; 1440 ierr = MPI_Waitany(merge->nrecv,rwaits,&icompleted,&rstatus);CHKERRQ(ierr); 1441 } 1442 ierr = PetscFree(rwaits);CHKERRQ(ierr); 1443 if (merge->nsend) {ierr = MPI_Waitall(merge->nsend,swaits,sstatus);CHKERRQ(ierr);} 1444 1445 /* send and recv coi */ 1446 /*-------------------*/ 1447 ierr = PetscCommGetNewTag(comm,&tagi);CHKERRQ(ierr); 1448 ierr = PetscPostIrecvInt(comm,tagi,merge->nrecv,merge->id_r,len_ri,&buf_ri,&rwaits);CHKERRQ(ierr); 1449 ierr = PetscMalloc((len+1)*sizeof(PetscInt),&buf_s);CHKERRQ(ierr); 1450 buf_si = buf_s; /* points to the beginning of k-th msg to be sent */ 1451 for (proc=0,k=0; proc<size; proc++){ 1452 if (!len_s[proc]) continue; 1453 /* form outgoing message for i-structure: 1454 buf_si[0]: nrows to be sent 1455 [1:nrows]: row index (global) 1456 [nrows+1:2*nrows+1]: i-structure index 1457 */ 1458 /*-------------------------------------------*/ 1459 nrows = len_si[proc]/2 - 1; 1460 buf_si_i = buf_si + nrows+1; 1461 buf_si[0] = nrows; 1462 buf_si_i[0] = 0; 1463 nrows = 0; 1464 for (i=owners_co[proc]; i<owners_co[proc+1]; i++){ 1465 nzi = coi[i+1] - coi[i]; 1466 buf_si_i[nrows+1] = buf_si_i[nrows] + nzi; /* i-structure */ 1467 buf_si[nrows+1] =prmap[i] -owners[proc]; /* local row index */ 1468 nrows++; 1469 } 1470 ierr = MPI_Isend(buf_si,len_si[proc],MPIU_INT,proc,tagi,comm,swaits+k);CHKERRQ(ierr); 1471 k++; 1472 buf_si += len_si[proc]; 1473 } 1474 i = merge->nrecv; 1475 while (i--) { 1476 PetscMPIInt icompleted; 1477 ierr = MPI_Waitany(merge->nrecv,rwaits,&icompleted,&rstatus);CHKERRQ(ierr); 1478 } 1479 ierr = PetscFree(rwaits);CHKERRQ(ierr); 1480 if (merge->nsend) {ierr = MPI_Waitall(merge->nsend,swaits,sstatus);CHKERRQ(ierr);} 1481 ierr = PetscFree(len_si);CHKERRQ(ierr); 1482 ierr = PetscFree(len_ri);CHKERRQ(ierr); 1483 ierr = PetscFree(swaits);CHKERRQ(ierr); 1484 ierr = PetscFree(sstatus);CHKERRQ(ierr); 1485 ierr = PetscFree(buf_s);CHKERRQ(ierr); 1486 1487 /* compute the local portion of C (mpi mat) */ 1488 /*------------------------------------------*/ 1489 /* allocate bi array and free space for accumulating nonzero column info */ 1490 ierr = PetscMalloc((pn+1)*sizeof(PetscInt),&bi);CHKERRQ(ierr); 1491 bi[0] = 0; 1492 1493 /* set initial free space to be fill*(nnz(P) + nnz(A)) */ 1494 nnz = fill*(pdti[pn] + poti[pon] + ai[am]); 1495 ierr = PetscFreeSpaceGet(nnz,&free_space); 1496 current_space = free_space; 1497 1498 #if defined(DEBUG_MATTrMatMult) 1499 ierr = PetscPrintf(PETSC_COMM_SELF,"[%d] nnz=fill %g*(%d + %d + %d)\n",rank,fill,pdti[pn],poti[pon],ai[am]); 1500 #endif 1501 1502 ierr = PetscMalloc3(merge->nrecv,PetscInt**,&buf_ri_k,merge->nrecv,PetscInt*,&nextrow,merge->nrecv,PetscInt*,&nextci);CHKERRQ(ierr); 1503 for (k=0; k<merge->nrecv; k++){ 1504 buf_ri_k[k] = buf_ri[k]; /* beginning of k-th recved i-structure */ 1505 nrows = *buf_ri_k[k]; 1506 nextrow[k] = buf_ri_k[k] + 1; /* next row number of k-th recved i-structure */ 1507 nextci[k] = buf_ri_k[k] + (nrows + 1);/* poins to the next i-structure of k-th recved i-structure */ 1508 } 1509 1510 ierr = MatPreallocateInitialize(comm,pn,A->cmap->n,dnz,onz);CHKERRQ(ierr); 1511 rmax = 0; 1512 for (i=0; i<pn; i++) { 1513 /* add pdt[i,:]*AP into lnk */ 1514 pnz = pdti[i+1] - pdti[i]; 1515 ptJ = pdtj + pdti[i]; 1516 for (j=0; j<pnz; j++){ 1517 row = ptJ[j]; /* row of AP == col of Pt */ 1518 anz = ai[row+1] - ai[row]; 1519 Jptr = aj + ai[row]; 1520 /* add non-zero cols of AP into the sorted linked list lnk */ 1521 ierr = PetscLLCondensedAddSorted(anz,Jptr,lnk,lnkbt);CHKERRQ(ierr); 1522 } 1523 1524 /* add received col data into lnk */ 1525 for (k=0; k<merge->nrecv; k++){ /* k-th received message */ 1526 if (i == *nextrow[k]) { /* i-th row */ 1527 nzi = *(nextci[k]+1) - *nextci[k]; 1528 Jptr = buf_rj[k] + *nextci[k]; 1529 ierr = PetscLLCondensedAddSorted(nzi,Jptr,lnk,lnkbt);CHKERRQ(ierr); 1530 nextrow[k]++; nextci[k]++; 1531 } 1532 } 1533 nnz = lnk[0]; 1534 1535 /* if free space is not available, make more free space */ 1536 if (current_space->local_remaining<nnz) { 1537 ierr = PetscFreeSpaceGet(nnz+current_space->total_array_size,¤t_space);CHKERRQ(ierr); 1538 nspacedouble++; 1539 } 1540 /* copy data into free space, then initialize lnk */ 1541 ierr = PetscLLCondensedClean(aN,nnz,current_space->array,lnk,lnkbt);CHKERRQ(ierr); 1542 ierr = MatPreallocateSet(i+owners[rank],nnz,current_space->array,dnz,onz);CHKERRQ(ierr); 1543 current_space->array += nnz; 1544 current_space->local_used += nnz; 1545 current_space->local_remaining -= nnz; 1546 bi[i+1] = bi[i] + nnz; 1547 if (nnz > rmax) rmax = nnz; 1548 } 1549 ierr = PetscFree3(buf_ri_k,nextrow,nextci);CHKERRQ(ierr); 1550 1551 ierr = PetscMalloc((bi[pn]+1)*sizeof(PetscInt),&bj);CHKERRQ(ierr); 1552 ierr = PetscFreeSpaceContiguous(&free_space,bj);CHKERRQ(ierr); 1553 afill_tmp = (PetscReal)bi[pn]/(pdti[pn] + poti[pon] + ai[am]); 1554 if (afill_tmp > afill) afill = afill_tmp; 1555 ierr = PetscLLCondensedDestroy(lnk,lnkbt);CHKERRQ(ierr); 1556 ierr = MatDestroy(&POt);CHKERRQ(ierr); 1557 ierr = MatDestroy(&PDt);CHKERRQ(ierr); 1558 1559 /* create symbolic parallel matrix Cmpi - why cannot be assembled in Numeric part */ 1560 /*----------------------------------------------------------------------------------*/ 1561 ierr = PetscMalloc((rmax+1)*sizeof(PetscScalar),&vals);CHKERRQ(ierr); 1562 ierr = PetscMemzero(vals,rmax*sizeof(PetscScalar));CHKERRQ(ierr); 1563 1564 ierr = MatCreate(comm,&Cmpi);CHKERRQ(ierr); 1565 ierr = MatSetSizes(Cmpi,pn,A->cmap->n,PETSC_DETERMINE,PETSC_DETERMINE);CHKERRQ(ierr); 1566 ierr = MatSetType(Cmpi,MATMPIAIJ);CHKERRQ(ierr); 1567 ierr = MatMPIAIJSetPreallocation(Cmpi,0,dnz,0,onz);CHKERRQ(ierr); 1568 ierr = MatPreallocateFinalize(dnz,onz);CHKERRQ(ierr); 1569 ierr = MatSetBlockSize(Cmpi,1);CHKERRQ(ierr); 1570 for (i=0; i<pn; i++){ 1571 row = i + rstart; 1572 nnz = bi[i+1] - bi[i]; 1573 Jptr = bj + bi[i]; 1574 ierr = MatSetValues(Cmpi,1,&row,nnz,Jptr,vals,INSERT_VALUES);CHKERRQ(ierr); 1575 } 1576 ierr = MatAssemblyBegin(Cmpi,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1577 ierr = MatAssemblyEnd(Cmpi,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1578 ierr = PetscFree(vals);CHKERRQ(ierr); 1579 1580 merge->bi = bi; 1581 merge->bj = bj; 1582 merge->coi = coi; 1583 merge->coj = coj; 1584 merge->buf_ri = buf_ri; 1585 merge->buf_rj = buf_rj; 1586 merge->owners_co = owners_co; 1587 merge->destroy = Cmpi->ops->destroy; 1588 merge->duplicate = Cmpi->ops->duplicate; 1589 1590 Cmpi->ops->mattransposemultnumeric = MatTransposeMatMultNumeric_MPIAIJ_MPIAIJ; 1591 Cmpi->ops->destroy = MatDestroy_MPIAIJ_PtAP; 1592 1593 /* attach the supporting struct to Cmpi for reuse */ 1594 c = (Mat_MPIAIJ*)Cmpi->data; 1595 c->ptap = ptap; 1596 ptap->api = PETSC_NULL; 1597 ptap->apj = PETSC_NULL; 1598 ptap->merge = merge; 1599 ptap->rmax = rmax; 1600 1601 *C = Cmpi; 1602 #if defined(PETSC_USE_INFO) 1603 if (bi[pn] != 0) { 1604 ierr = PetscInfo3(Cmpi,"Reallocs %D; Fill ratio: given %G needed %G.\n",nspacedouble,fill,afill);CHKERRQ(ierr); 1605 ierr = PetscInfo1(Cmpi,"Use MatTransposeMatMult(A,B,MatReuse,%G,&C) for best performance.\n",afill);CHKERRQ(ierr); 1606 } else { 1607 ierr = PetscInfo(Cmpi,"Empty matrix product\n");CHKERRQ(ierr); 1608 } 1609 #endif 1610 PetscFunctionReturn(0); 1611 } 1612 1613 #undef __FUNCT__ 1614 #define __FUNCT__ "MatTransposeMatMultNumeric_MPIAIJ_MPIAIJ_Scalable" 1615 PetscErrorCode MatTransposeMatMultNumeric_MPIAIJ_MPIAIJ_Scalable(Mat P,Mat A,Mat C) 1616 { 1617 PetscErrorCode ierr; 1618 Mat_Merge_SeqsToMPI *merge; 1619 Mat_MPIAIJ *p=(Mat_MPIAIJ*)P->data,*c=(Mat_MPIAIJ*)C->data; 1620 Mat_SeqAIJ *pd=(Mat_SeqAIJ*)(p->A)->data,*po=(Mat_SeqAIJ*)(p->B)->data; 1621 Mat_PtAPMPI *ptap; 1622 PetscInt *adj; 1623 PetscInt i,j,k,anz,pnz,row,*cj,nexta; 1624 MatScalar *ada,*ca,valtmp; 1625 PetscInt am=A->rmap->n,cm=C->rmap->n,pon=(p->B)->cmap->n; 1626 MPI_Comm comm=((PetscObject)C)->comm; 1627 PetscMPIInt size,rank,taga,*len_s; 1628 PetscInt *owners,proc,nrows,**buf_ri_k,**nextrow,**nextci; 1629 PetscInt **buf_ri,**buf_rj; 1630 PetscInt cnz=0,*bj_i,*bi,*bj,bnz,nextcj; /* bi,bj,ba: local array of C(mpi mat) */ 1631 MPI_Request *s_waits,*r_waits; 1632 MPI_Status *status; 1633 MatScalar **abuf_r,*ba_i,*pA,*coa,*ba; 1634 PetscInt *ai,*aj,*coi,*coj; 1635 PetscInt *poJ=po->j,*pdJ=pd->j; 1636 Mat A_loc; 1637 Mat_SeqAIJ *a_loc; 1638 1639 PetscFunctionBegin; 1640 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr); 1641 ierr = MPI_Comm_rank(comm,&rank);CHKERRQ(ierr); 1642 1643 ptap = c->ptap; 1644 merge = ptap->merge; 1645 1646 #if defined(DEBUG_MATTrMatMult) 1647 ierr = PetscSynchronizedPrintf(comm,"[%d] TransposeMatMultNumeric_Scalable: Crmax %d \n",rank,ptap->rmax); 1648 ierr = PetscSynchronizedFlush(comm);CHKERRQ(ierr); 1649 #endif 1650 1651 /* 2) compute numeric C_seq = P_loc^T*A_loc */ 1652 /*------------------------------------------*/ 1653 /* get data from symbolic products */ 1654 coi = merge->coi; coj = merge->coj; 1655 ierr = PetscMalloc((coi[pon]+1)*sizeof(MatScalar),&coa);CHKERRQ(ierr); 1656 ierr = PetscMemzero(coa,coi[pon]*sizeof(MatScalar));CHKERRQ(ierr); 1657 1658 bi = merge->bi; bj = merge->bj; 1659 owners = merge->rowmap->range; 1660 ierr = PetscMalloc((bi[cm]+1)*sizeof(MatScalar),&ba);CHKERRQ(ierr); 1661 ierr = PetscMemzero(ba,bi[cm]*sizeof(MatScalar));CHKERRQ(ierr); 1662 1663 /* get A_loc by taking all local rows of A */ 1664 A_loc = ptap->A_loc; 1665 ierr = MatMPIAIJGetLocalMat(A,MAT_REUSE_MATRIX,&A_loc);CHKERRQ(ierr); 1666 a_loc = (Mat_SeqAIJ*)(A_loc)->data; 1667 ai = a_loc->i; 1668 aj = a_loc->j; 1669 1670 for (i=0; i<am; i++) { 1671 /* 2-a) put A[i,:] to dense array aval */ 1672 anz = ai[i+1] - ai[i]; 1673 adj = aj + ai[i]; 1674 ada = a_loc->a + ai[i]; 1675 1676 /* 2-b) Compute Cseq = P_loc[i,:]^T*A[i,:] using outer product */ 1677 /*-------------------------------------------------------------*/ 1678 /* put the value into Co=(p->B)^T*A (off-diagonal part, send to others) */ 1679 pnz = po->i[i+1] - po->i[i]; 1680 poJ = po->j + po->i[i]; 1681 pA = po->a + po->i[i]; 1682 for (j=0; j<pnz; j++){ 1683 row = poJ[j]; 1684 cnz = coi[row+1] - coi[row]; 1685 cj = coj + coi[row]; 1686 ca = coa + coi[row]; 1687 /* perform sparse axpy */ 1688 nexta = 0; 1689 valtmp = pA[j]; 1690 for (k=0; nexta<anz; k++) { 1691 if (cj[k] == adj[nexta]){ 1692 ca[k] += valtmp*ada[nexta]; 1693 nexta++; 1694 } 1695 } 1696 ierr = PetscLogFlops(2.0*anz);CHKERRQ(ierr); 1697 } 1698 1699 /* put the value into Cd (diagonal part) */ 1700 pnz = pd->i[i+1] - pd->i[i]; 1701 pdJ = pd->j + pd->i[i]; 1702 pA = pd->a + pd->i[i]; 1703 for (j=0; j<pnz; j++){ 1704 row = pdJ[j]; 1705 cnz = bi[row+1] - bi[row]; 1706 cj = bj + bi[row]; 1707 ca = ba + bi[row]; 1708 /* perform sparse axpy */ 1709 nexta = 0; 1710 valtmp = pA[j]; 1711 for (k=0; nexta<anz; k++) { 1712 if (cj[k] == adj[nexta]){ 1713 ca[k] += valtmp*ada[nexta]; 1714 nexta++; 1715 } 1716 } 1717 ierr = PetscLogFlops(2.0*anz);CHKERRQ(ierr); 1718 } 1719 1720 } 1721 1722 /* 3) send and recv matrix values coa */ 1723 /*------------------------------------*/ 1724 buf_ri = merge->buf_ri; 1725 buf_rj = merge->buf_rj; 1726 len_s = merge->len_s; 1727 ierr = PetscCommGetNewTag(comm,&taga);CHKERRQ(ierr); 1728 ierr = PetscPostIrecvScalar(comm,taga,merge->nrecv,merge->id_r,merge->len_r,&abuf_r,&r_waits);CHKERRQ(ierr); 1729 1730 ierr = PetscMalloc2(merge->nsend+1,MPI_Request,&s_waits,size,MPI_Status,&status);CHKERRQ(ierr); 1731 for (proc=0,k=0; proc<size; proc++){ 1732 if (!len_s[proc]) continue; 1733 i = merge->owners_co[proc]; 1734 ierr = MPI_Isend(coa+coi[i],len_s[proc],MPIU_MATSCALAR,proc,taga,comm,s_waits+k);CHKERRQ(ierr); 1735 k++; 1736 } 1737 if (merge->nrecv) {ierr = MPI_Waitall(merge->nrecv,r_waits,status);CHKERRQ(ierr);} 1738 if (merge->nsend) {ierr = MPI_Waitall(merge->nsend,s_waits,status);CHKERRQ(ierr);} 1739 1740 ierr = PetscFree2(s_waits,status);CHKERRQ(ierr); 1741 ierr = PetscFree(r_waits);CHKERRQ(ierr); 1742 ierr = PetscFree(coa);CHKERRQ(ierr); 1743 1744 /* 4) insert local Cseq and received values into Cmpi */ 1745 /*----------------------------------------------------*/ 1746 ierr = PetscMalloc3(merge->nrecv,PetscInt**,&buf_ri_k,merge->nrecv,PetscInt*,&nextrow,merge->nrecv,PetscInt*,&nextci);CHKERRQ(ierr); 1747 for (k=0; k<merge->nrecv; k++){ 1748 buf_ri_k[k] = buf_ri[k]; /* beginning of k-th recved i-structure */ 1749 nrows = *(buf_ri_k[k]); 1750 nextrow[k] = buf_ri_k[k]+1; /* next row number of k-th recved i-structure */ 1751 nextci[k] = buf_ri_k[k] + (nrows + 1);/* poins to the next i-structure of k-th recved i-structure */ 1752 } 1753 1754 for (i=0; i<cm; i++) { 1755 row = owners[rank] + i; /* global row index of C_seq */ 1756 bj_i = bj + bi[i]; /* col indices of the i-th row of C */ 1757 ba_i = ba + bi[i]; 1758 bnz = bi[i+1] - bi[i]; 1759 /* add received vals into ba */ 1760 for (k=0; k<merge->nrecv; k++){ /* k-th received message */ 1761 /* i-th row */ 1762 if (i == *nextrow[k]) { 1763 cnz = *(nextci[k]+1) - *nextci[k]; 1764 cj = buf_rj[k] + *(nextci[k]); 1765 ca = abuf_r[k] + *(nextci[k]); 1766 nextcj = 0; 1767 for (j=0; nextcj<cnz; j++){ 1768 if (bj_i[j] == cj[nextcj]){ /* bcol == ccol */ 1769 ba_i[j] += ca[nextcj++]; 1770 } 1771 } 1772 nextrow[k]++; nextci[k]++; 1773 ierr = PetscLogFlops(2.0*cnz);CHKERRQ(ierr); 1774 } 1775 } 1776 ierr = MatSetValues(C,1,&row,bnz,bj_i,ba_i,INSERT_VALUES);CHKERRQ(ierr); 1777 } 1778 ierr = MatAssemblyBegin(C,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1779 ierr = MatAssemblyEnd(C,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 1780 1781 ierr = PetscFree(ba);CHKERRQ(ierr); 1782 ierr = PetscFree(abuf_r[0]);CHKERRQ(ierr); 1783 ierr = PetscFree(abuf_r);CHKERRQ(ierr); 1784 ierr = PetscFree3(buf_ri_k,nextrow,nextci);CHKERRQ(ierr); 1785 PetscFunctionReturn(0); 1786 } 1787 1788 /* This routine is modified from MatPtAPSymbolic_MPIAIJ_MPIAIJ() */ 1789 #undef __FUNCT__ 1790 #define __FUNCT__ "MatTransposeMatMultSymbolic_MPIAIJ_MPIAIJ_Scalable" 1791 PetscErrorCode MatTransposeMatMultSymbolic_MPIAIJ_MPIAIJ_Scalable(Mat P,Mat A,PetscReal fill,Mat *C) 1792 { 1793 PetscErrorCode ierr; 1794 Mat Cmpi,A_loc,POt,PDt; 1795 Mat_PtAPMPI *ptap; 1796 PetscFreeSpaceList free_space=PETSC_NULL,current_space=PETSC_NULL; 1797 Mat_MPIAIJ *p=(Mat_MPIAIJ*)P->data,*c; 1798 PetscInt *pdti,*pdtj,*poti,*potj,*ptJ; 1799 PetscInt nnz; 1800 PetscInt *lnk,*owners_co,*coi,*coj,i,k,pnz,row; 1801 PetscInt am=A->rmap->n,pn=P->cmap->n; 1802 MPI_Comm comm=((PetscObject)A)->comm; 1803 PetscMPIInt size,rank,tagi,tagj,*len_si,*len_s,*len_ri; 1804 PetscInt **buf_rj,**buf_ri,**buf_ri_k; 1805 PetscInt len,proc,*dnz,*onz,*owners; 1806 PetscInt nzi,*bi,*bj; 1807 PetscInt nrows,*buf_s,*buf_si,*buf_si_i,**nextrow,**nextci; 1808 MPI_Request *swaits,*rwaits; 1809 MPI_Status *sstatus,rstatus; 1810 Mat_Merge_SeqsToMPI *merge; 1811 PetscInt *ai,*aj,*Jptr,anz,*prmap=p->garray,pon,nspacedouble=0,j; 1812 PetscReal afill=1.0,afill_tmp; 1813 PetscInt rstart = P->cmap->rstart,rmax,aN=A->cmap->N,Crmax; 1814 PetscScalar *vals; 1815 Mat_SeqAIJ *a_loc, *pdt,*pot; 1816 1817 PetscFunctionBegin; 1818 /* check if matrix local sizes are compatible */ 1819 if (A->rmap->rstart != P->rmap->rstart || A->rmap->rend != P->rmap->rend){ 1820 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); 1821 } 1822 1823 ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr); 1824 ierr = MPI_Comm_rank(comm,&rank);CHKERRQ(ierr); 1825 #if defined(DEBUG_MATTrMatMult) 1826 ierr = PetscSynchronizedPrintf(comm,"[%d] TransposeMatMultSymbolic_Scalable P: %d %d, %d %d; A %d %d, %d %d\n",rank,P->rmap->N,P->cmap->N,P->rmap->n,P->cmap->n,A->rmap->N,aN,A->rmap->n,A->cmap->n); 1827 ierr = PetscSynchronizedFlush(comm);CHKERRQ(ierr); 1828 #endif 1829 1830 /* create struct Mat_PtAPMPI and attached it to C later */ 1831 ierr = PetscNew(Mat_PtAPMPI,&ptap);CHKERRQ(ierr); 1832 1833 /* get A_loc by taking all local rows of A */ 1834 ierr = MatMPIAIJGetLocalMat(A,MAT_INITIAL_MATRIX,&A_loc);CHKERRQ(ierr); 1835 ptap->A_loc = A_loc; 1836 a_loc = (Mat_SeqAIJ*)(A_loc)->data; 1837 ai = a_loc->i; 1838 aj = a_loc->j; 1839 1840 /* determine symbolic Co=(p->B)^T*A - send to others */ 1841 /*----------------------------------------------------*/ 1842 ierr = MatTransposeSymbolic_SeqAIJ(p->A,&PDt);CHKERRQ(ierr); 1843 pdt = (Mat_SeqAIJ*)PDt->data; 1844 pdti = pdt->i; pdtj = pdt->j; 1845 1846 ierr = MatTransposeSymbolic_SeqAIJ(p->B,&POt);CHKERRQ(ierr); 1847 pot = (Mat_SeqAIJ*)POt->data; 1848 poti = pot->i; potj = pot->j; 1849 1850 /* then, compute symbolic Co = (p->B)^T*A */ 1851 pon = (p->B)->cmap->n; /* total num of rows to be sent to other processors 1852 >= (num of nonzero rows of C_seq) - pn */ 1853 ierr = PetscMalloc((pon+1)*sizeof(PetscInt),&coi);CHKERRQ(ierr); 1854 coi[0] = 0; 1855 1856 /* set initial free space to be fill*(nnz(p->B) + nnz(A)) */ 1857 nnz = fill*(poti[pon] + ai[am]); 1858 ierr = PetscFreeSpaceGet(nnz,&free_space); 1859 current_space = free_space; 1860 #if defined(DEBUG_MATTrMatMult) 1861 ierr = PetscSynchronizedPrintf(comm, "[%d] nnz = fill %g *(%d + %d)\n",rank,fill,poti[pon],ai[am]);CHKERRQ(ierr); 1862 ierr = PetscSynchronizedFlush(comm);CHKERRQ(ierr); 1863 #endif 1864 /* create and initialize a linked list */ 1865 i = PetscMax(pdt->rmax,pot->rmax); 1866 Crmax = i*a_loc->rmax*size; /* non-scalable! */ 1867 if (!Crmax || Crmax > aN) Crmax = aN; 1868 #if defined(DEBUG_MATTrMatMult) 1869 printf("[%d] rmax A_loc %d * max(PD %d, PO %d)=%d, Crmax %d\n",rank,a_loc->rmax,pdt->rmax,pot->rmax,i*a_loc->rmax,Crmax); 1870 #endif 1871 ierr = PetscLLCondensedCreate_Scalable(Crmax,&lnk);CHKERRQ(ierr); 1872 1873 for (i=0; i<pon; i++) { 1874 nnz = 0; 1875 pnz = poti[i+1] - poti[i]; 1876 ptJ = potj + poti[i]; 1877 for (j=0; j<pnz; j++){ 1878 row = ptJ[j]; /* row of A_loc == col of Pot */ 1879 anz = ai[row+1] - ai[row]; 1880 Jptr = aj + ai[row]; 1881 /* add non-zero cols of AP into the sorted linked list lnk */ 1882 ierr = PetscLLCondensedAddSorted_Scalable(anz,Jptr,lnk);CHKERRQ(ierr); 1883 } 1884 nnz = lnk[0]; 1885 1886 /* If free space is not available, double the total space in the list */ 1887 if (current_space->local_remaining<nnz) { 1888 ierr = PetscFreeSpaceGet(nnz+current_space->total_array_size,¤t_space);CHKERRQ(ierr); 1889 nspacedouble++; 1890 } 1891 1892 /* Copy data into free space, and zero out denserows */ 1893 ierr = PetscLLCondensedClean_Scalable(nnz,current_space->array,lnk);CHKERRQ(ierr); 1894 current_space->array += nnz; 1895 current_space->local_used += nnz; 1896 current_space->local_remaining -= nnz; 1897 coi[i+1] = coi[i] + nnz; 1898 } 1899 1900 ierr = PetscMalloc((coi[pon]+1)*sizeof(PetscInt),&coj);CHKERRQ(ierr); 1901 ierr = PetscFreeSpaceContiguous(&free_space,coj);CHKERRQ(ierr); 1902 afill_tmp = (PetscReal)coi[pon]/(poti[pon] + ai[am]); 1903 if (afill_tmp > afill) afill = afill_tmp; 1904 1905 /* send j-array (coj) of Co to other processors */ 1906 /*----------------------------------------------*/ 1907 /* determine row ownership */ 1908 ierr = PetscNew(Mat_Merge_SeqsToMPI,&merge);CHKERRQ(ierr); 1909 ierr = PetscLayoutCreate(comm,&merge->rowmap);CHKERRQ(ierr); 1910 merge->rowmap->n = pn; 1911 merge->rowmap->bs = 1; 1912 ierr = PetscLayoutSetUp(merge->rowmap);CHKERRQ(ierr); 1913 owners = merge->rowmap->range; 1914 1915 /* determine the number of messages to send, their lengths */ 1916 ierr = PetscMalloc(size*sizeof(PetscMPIInt),&len_si);CHKERRQ(ierr); 1917 ierr = PetscMemzero(len_si,size*sizeof(PetscMPIInt));CHKERRQ(ierr); 1918 ierr = PetscMalloc(size*sizeof(PetscMPIInt),&merge->len_s);CHKERRQ(ierr); 1919 len_s = merge->len_s; 1920 merge->nsend = 0; 1921 1922 ierr = PetscMalloc((size+2)*sizeof(PetscInt),&owners_co);CHKERRQ(ierr); 1923 ierr = PetscMemzero(len_s,size*sizeof(PetscMPIInt));CHKERRQ(ierr); 1924 1925 proc = 0; 1926 for (i=0; i<pon; i++){ 1927 while (prmap[i] >= owners[proc+1]) proc++; 1928 len_si[proc]++; /* num of rows in Co to be sent to [proc] */ 1929 len_s[proc] += coi[i+1] - coi[i]; 1930 } 1931 1932 len = 0; /* max length of buf_si[] */ 1933 owners_co[0] = 0; 1934 for (proc=0; proc<size; proc++){ 1935 owners_co[proc+1] = owners_co[proc] + len_si[proc]; 1936 if (len_si[proc]){ 1937 merge->nsend++; 1938 len_si[proc] = 2*(len_si[proc] + 1); 1939 len += len_si[proc]; 1940 } 1941 } 1942 1943 /* determine the number and length of messages to receive for coi and coj */ 1944 ierr = PetscGatherNumberOfMessages(comm,PETSC_NULL,len_s,&merge->nrecv);CHKERRQ(ierr); 1945 ierr = PetscGatherMessageLengths2(comm,merge->nsend,merge->nrecv,len_s,len_si,&merge->id_r,&merge->len_r,&len_ri);CHKERRQ(ierr); 1946 1947 /* post the Irecv and Isend of coj */ 1948 ierr = PetscCommGetNewTag(comm,&tagj);CHKERRQ(ierr); 1949 ierr = PetscPostIrecvInt(comm,tagj,merge->nrecv,merge->id_r,merge->len_r,&buf_rj,&rwaits);CHKERRQ(ierr); 1950 ierr = PetscMalloc((merge->nsend+1)*sizeof(MPI_Request),&swaits);CHKERRQ(ierr); 1951 for (proc=0, k=0; proc<size; proc++){ 1952 if (!len_s[proc]) continue; 1953 i = owners_co[proc]; 1954 ierr = MPI_Isend(coj+coi[i],len_s[proc],MPIU_INT,proc,tagj,comm,swaits+k);CHKERRQ(ierr); 1955 k++; 1956 } 1957 1958 /* receives and sends of coj are complete */ 1959 ierr = PetscMalloc(size*sizeof(MPI_Status),&sstatus);CHKERRQ(ierr); 1960 for (i=0; i<merge->nrecv; i++){ 1961 PetscMPIInt icompleted; 1962 ierr = MPI_Waitany(merge->nrecv,rwaits,&icompleted,&rstatus);CHKERRQ(ierr); 1963 } 1964 ierr = PetscFree(rwaits);CHKERRQ(ierr); 1965 if (merge->nsend) {ierr = MPI_Waitall(merge->nsend,swaits,sstatus);CHKERRQ(ierr);} 1966 1967 /* send and recv coi */ 1968 /*-------------------*/ 1969 ierr = PetscCommGetNewTag(comm,&tagi);CHKERRQ(ierr); 1970 ierr = PetscPostIrecvInt(comm,tagi,merge->nrecv,merge->id_r,len_ri,&buf_ri,&rwaits);CHKERRQ(ierr); 1971 ierr = PetscMalloc((len+1)*sizeof(PetscInt),&buf_s);CHKERRQ(ierr); 1972 buf_si = buf_s; /* points to the beginning of k-th msg to be sent */ 1973 for (proc=0,k=0; proc<size; proc++){ 1974 if (!len_s[proc]) continue; 1975 /* form outgoing message for i-structure: 1976 buf_si[0]: nrows to be sent 1977 [1:nrows]: row index (global) 1978 [nrows+1:2*nrows+1]: i-structure index 1979 */ 1980 /*-------------------------------------------*/ 1981 nrows = len_si[proc]/2 - 1; 1982 buf_si_i = buf_si + nrows+1; 1983 buf_si[0] = nrows; 1984 buf_si_i[0] = 0; 1985 nrows = 0; 1986 for (i=owners_co[proc]; i<owners_co[proc+1]; i++){ 1987 nzi = coi[i+1] - coi[i]; 1988 buf_si_i[nrows+1] = buf_si_i[nrows] + nzi; /* i-structure */ 1989 buf_si[nrows+1] =prmap[i] -owners[proc]; /* local row index */ 1990 nrows++; 1991 } 1992 ierr = MPI_Isend(buf_si,len_si[proc],MPIU_INT,proc,tagi,comm,swaits+k);CHKERRQ(ierr); 1993 k++; 1994 buf_si += len_si[proc]; 1995 } 1996 i = merge->nrecv; 1997 while (i--) { 1998 PetscMPIInt icompleted; 1999 ierr = MPI_Waitany(merge->nrecv,rwaits,&icompleted,&rstatus);CHKERRQ(ierr); 2000 } 2001 ierr = PetscFree(rwaits);CHKERRQ(ierr); 2002 if (merge->nsend) {ierr = MPI_Waitall(merge->nsend,swaits,sstatus);CHKERRQ(ierr);} 2003 ierr = PetscFree(len_si);CHKERRQ(ierr); 2004 ierr = PetscFree(len_ri);CHKERRQ(ierr); 2005 ierr = PetscFree(swaits);CHKERRQ(ierr); 2006 ierr = PetscFree(sstatus);CHKERRQ(ierr); 2007 ierr = PetscFree(buf_s);CHKERRQ(ierr); 2008 2009 /* compute the local portion of C (mpi mat) */ 2010 /*------------------------------------------*/ 2011 /* allocate bi array and free space for accumulating nonzero column info */ 2012 ierr = PetscMalloc((pn+1)*sizeof(PetscInt),&bi);CHKERRQ(ierr); 2013 bi[0] = 0; 2014 2015 /* set initial free space to be fill*(nnz(P) + nnz(AP)) */ 2016 nnz = fill*(pdti[pn] + poti[pon] + ai[am]); 2017 ierr = PetscFreeSpaceGet(nnz,&free_space); 2018 current_space = free_space; 2019 2020 #if defined(DEBUG_MATTrMatMult) 2021 ierr = PetscPrintf(PETSC_COMM_SELF,"[%d] nnz=%d + %d + %d\n",rank,pdti[pn],poti[pon],ai[am]); 2022 #endif 2023 2024 ierr = PetscMalloc3(merge->nrecv,PetscInt**,&buf_ri_k,merge->nrecv,PetscInt*,&nextrow,merge->nrecv,PetscInt*,&nextci);CHKERRQ(ierr); 2025 for (k=0; k<merge->nrecv; k++){ 2026 buf_ri_k[k] = buf_ri[k]; /* beginning of k-th recved i-structure */ 2027 nrows = *buf_ri_k[k]; 2028 nextrow[k] = buf_ri_k[k] + 1; /* next row number of k-th recved i-structure */ 2029 nextci[k] = buf_ri_k[k] + (nrows + 1);/* poins to the next i-structure of k-th recved i-structure */ 2030 } 2031 2032 ierr = MatPreallocateInitialize(comm,pn,A->cmap->n,dnz,onz);CHKERRQ(ierr); 2033 rmax = 0; 2034 for (i=0; i<pn; i++) { 2035 /* add pdt[i,:]*AP into lnk */ 2036 pnz = pdti[i+1] - pdti[i]; 2037 ptJ = pdtj + pdti[i]; 2038 for (j=0; j<pnz; j++){ 2039 row = ptJ[j]; /* row of AP == col of Pt */ 2040 anz = ai[row+1] - ai[row]; 2041 Jptr = aj + ai[row]; 2042 /* add non-zero cols of AP into the sorted linked list lnk */ 2043 ierr = PetscLLCondensedAddSorted_Scalable(anz,Jptr,lnk);CHKERRQ(ierr); 2044 } 2045 2046 /* add received col data into lnk */ 2047 for (k=0; k<merge->nrecv; k++){ /* k-th received message */ 2048 if (i == *nextrow[k]) { /* i-th row */ 2049 nzi = *(nextci[k]+1) - *nextci[k]; 2050 Jptr = buf_rj[k] + *nextci[k]; 2051 ierr = PetscLLCondensedAddSorted_Scalable(nzi,Jptr,lnk);CHKERRQ(ierr); 2052 nextrow[k]++; nextci[k]++; 2053 } 2054 } 2055 nnz = lnk[0]; 2056 2057 /* if free space is not available, make more free space */ 2058 if (current_space->local_remaining<nnz) { 2059 ierr = PetscFreeSpaceGet(nnz+current_space->total_array_size,¤t_space);CHKERRQ(ierr); 2060 nspacedouble++; 2061 } 2062 /* copy data into free space, then initialize lnk */ 2063 ierr = PetscLLCondensedClean_Scalable(nnz,current_space->array,lnk);CHKERRQ(ierr); 2064 ierr = MatPreallocateSet(i+owners[rank],nnz,current_space->array,dnz,onz);CHKERRQ(ierr); 2065 current_space->array += nnz; 2066 current_space->local_used += nnz; 2067 current_space->local_remaining -= nnz; 2068 bi[i+1] = bi[i] + nnz; 2069 if (nnz > rmax) rmax = nnz; 2070 } 2071 ierr = PetscFree3(buf_ri_k,nextrow,nextci);CHKERRQ(ierr); 2072 2073 ierr = PetscMalloc((bi[pn]+1)*sizeof(PetscInt),&bj);CHKERRQ(ierr); 2074 ierr = PetscFreeSpaceContiguous(&free_space,bj);CHKERRQ(ierr); 2075 afill_tmp = (PetscReal)bi[pn]/(pdti[pn] + poti[pon] + ai[am]); 2076 if (afill_tmp > afill) afill = afill_tmp; 2077 ierr = PetscLLCondensedDestroy_Scalable(lnk);CHKERRQ(ierr); 2078 ierr = MatDestroy(&POt);CHKERRQ(ierr); 2079 ierr = MatDestroy(&PDt);CHKERRQ(ierr); 2080 2081 /* create symbolic parallel matrix Cmpi - why cannot be assembled in Numeric part */ 2082 /*----------------------------------------------------------------------------------*/ 2083 ierr = PetscMalloc((rmax+1)*sizeof(PetscScalar),&vals);CHKERRQ(ierr); 2084 ierr = PetscMemzero(vals,rmax*sizeof(PetscScalar));CHKERRQ(ierr); 2085 2086 ierr = MatCreate(comm,&Cmpi);CHKERRQ(ierr); 2087 ierr = MatSetSizes(Cmpi,pn,A->cmap->n,PETSC_DETERMINE,PETSC_DETERMINE);CHKERRQ(ierr); 2088 ierr = MatSetType(Cmpi,MATMPIAIJ);CHKERRQ(ierr); 2089 ierr = MatMPIAIJSetPreallocation(Cmpi,0,dnz,0,onz);CHKERRQ(ierr); 2090 ierr = MatPreallocateFinalize(dnz,onz);CHKERRQ(ierr); 2091 ierr = MatSetBlockSize(Cmpi,1);CHKERRQ(ierr); 2092 for (i=0; i<pn; i++){ 2093 row = i + rstart; 2094 nnz = bi[i+1] - bi[i]; 2095 Jptr = bj + bi[i]; 2096 ierr = MatSetValues(Cmpi,1,&row,nnz,Jptr,vals,INSERT_VALUES);CHKERRQ(ierr); 2097 } 2098 ierr = MatAssemblyBegin(Cmpi,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 2099 ierr = MatAssemblyEnd(Cmpi,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 2100 ierr = PetscFree(vals);CHKERRQ(ierr); 2101 2102 merge->bi = bi; 2103 merge->bj = bj; 2104 merge->coi = coi; 2105 merge->coj = coj; 2106 merge->buf_ri = buf_ri; 2107 merge->buf_rj = buf_rj; 2108 merge->owners_co = owners_co; 2109 merge->destroy = Cmpi->ops->destroy; 2110 merge->duplicate = Cmpi->ops->duplicate; 2111 2112 Cmpi->ops->mattransposemultnumeric = MatTransposeMatMultNumeric_MPIAIJ_MPIAIJ_Scalable; 2113 Cmpi->ops->destroy = MatDestroy_MPIAIJ_PtAP; 2114 2115 /* attach the supporting struct to Cmpi for reuse */ 2116 c = (Mat_MPIAIJ*)Cmpi->data; 2117 c->ptap = ptap; 2118 ptap->api = PETSC_NULL; 2119 ptap->apj = PETSC_NULL; 2120 ptap->merge = merge; 2121 ptap->rmax = rmax; 2122 ptap->apa = PETSC_NULL; 2123 2124 *C = Cmpi; 2125 #if defined(PETSC_USE_INFO) 2126 if (bi[pn] != 0) { 2127 ierr = PetscInfo3(Cmpi,"Reallocs %D; Fill ratio: given %G needed %G.\n",nspacedouble,fill,afill);CHKERRQ(ierr); 2128 ierr = PetscInfo1(Cmpi,"Use MatTransposeMatMult(A,B,MatReuse,%G,&C) for best performance.\n",afill);CHKERRQ(ierr); 2129 } else { 2130 ierr = PetscInfo(Cmpi,"Empty matrix product\n");CHKERRQ(ierr); 2131 } 2132 #endif 2133 PetscFunctionReturn(0); 2134 } 2135