1 #define PETSCMAT_DLL 2 3 /* 4 Factorization code for BAIJ format. 5 */ 6 #include "../src/mat/impls/baij/seq/baij.h" 7 #include "../src/mat/blockinvert.h" 8 9 #undef __FUNCT__ 10 #define __FUNCT__ "MatLUFactorNumeric_SeqBAIJ_2" 11 PetscErrorCode MatLUFactorNumeric_SeqBAIJ_2(Mat B,Mat A,const MatFactorInfo *info) 12 { 13 Mat C=B; 14 Mat_SeqBAIJ *a=(Mat_SeqBAIJ*)A->data,*b=(Mat_SeqBAIJ *)C->data; 15 IS isrow = b->row,isicol = b->icol; 16 PetscErrorCode ierr; 17 const PetscInt *r,*ic,*ics; 18 PetscInt i,j,k,nz,nzL,row,*pj; 19 const PetscInt n=a->mbs,*ai=a->i,*aj=a->j,*bi=b->i,*bj=b->j,bs2=a->bs2; 20 const PetscInt *ajtmp,*bjtmp,*bdiag=b->diag; 21 MatScalar *rtmp,*pc,*mwork,*pv; 22 MatScalar *aa=a->a,*v; 23 PetscInt flg; 24 PetscReal shift = info->shiftinblocks; 25 26 PetscFunctionBegin; 27 ierr = ISGetIndices(isrow,&r);CHKERRQ(ierr); 28 ierr = ISGetIndices(isicol,&ic);CHKERRQ(ierr); 29 30 /* generate work space needed by the factorization */ 31 ierr = PetscMalloc2(bs2*n,MatScalar,&rtmp,bs2,MatScalar,&mwork);CHKERRQ(ierr); 32 ierr = PetscMemzero(rtmp,bs2*n*sizeof(MatScalar));CHKERRQ(ierr); 33 ics = ic; 34 35 for (i=0; i<n; i++){ 36 /* zero rtmp */ 37 /* L part */ 38 nz = bi[i+1] - bi[i]; 39 bjtmp = bj + bi[i]; 40 for (j=0; j<nz; j++){ 41 ierr = PetscMemzero(rtmp+bs2*bjtmp[j],bs2*sizeof(MatScalar));CHKERRQ(ierr); 42 } 43 44 /* U part */ 45 nz = bdiag[i] - bdiag[i+1]; 46 bjtmp = bj + bdiag[i+1]+1; 47 for (j=0; j<nz; j++){ 48 ierr = PetscMemzero(rtmp+bs2*bjtmp[j],bs2*sizeof(MatScalar));CHKERRQ(ierr); 49 } 50 51 /* load in initial (unfactored row) */ 52 nz = ai[r[i]+1] - ai[r[i]]; 53 ajtmp = aj + ai[r[i]]; 54 v = aa + bs2*ai[r[i]]; 55 for (j=0; j<nz; j++) { 56 ierr = PetscMemcpy(rtmp+bs2*ic[ajtmp[j]],v+bs2*j,bs2*sizeof(MatScalar));CHKERRQ(ierr); 57 } 58 59 /* elimination */ 60 bjtmp = bj + bi[i]; 61 nzL = bi[i+1] - bi[i]; 62 for(k=0;k < nzL;k++) { 63 row = bjtmp[k]; 64 pc = rtmp + bs2*row; 65 for (flg=0,j=0; j<bs2; j++) { if (pc[j]!=0.0) { flg = 1; break; }} 66 if (flg) { 67 pv = b->a + bs2*bdiag[row]; 68 /* Kernel_A_gets_A_times_B(bs,pc,pv,mwork); *pc = *pc * (*pv); */ 69 ierr = Kernel_A_gets_A_times_B_2(pc,pv,mwork);CHKERRQ(ierr); 70 71 pj = b->j + bdiag[row+1]+1; /* begining of U(row,:) */ 72 pv = b->a + bs2*(bdiag[row+1]+1); 73 nz = bdiag[row] - bdiag[row+1] - 1; /* num of entries inU(row,:), excluding diag */ 74 for (j=0; j<nz; j++) { 75 /* Kernel_A_gets_A_minus_B_times_C(bs,rtmp+bs2*pj[j],pc,pv+bs2*j); */ 76 /* rtmp+bs2*pj[j] = rtmp+bs2*pj[j] - (*pc)*(pv+bs2*j) */ 77 v = rtmp + 4*pj[j]; 78 ierr = Kernel_A_gets_A_minus_B_times_C_2(v,pc,pv);CHKERRQ(ierr); 79 pv += 4; 80 } 81 ierr = PetscLogFlops(16*nz+12);CHKERRQ(ierr); /* flops = 2*bs^3*nz + 2*bs^3 - bs2) */ 82 } 83 } 84 85 /* finished row so stick it into b->a */ 86 /* L part */ 87 pv = b->a + bs2*bi[i] ; 88 pj = b->j + bi[i] ; 89 nz = bi[i+1] - bi[i]; 90 for (j=0; j<nz; j++) { 91 ierr = PetscMemcpy(pv+bs2*j,rtmp+bs2*pj[j],bs2*sizeof(MatScalar));CHKERRQ(ierr); 92 } 93 94 /* Mark diagonal and invert diagonal for simplier triangular solves */ 95 pv = b->a + bs2*bdiag[i]; 96 pj = b->j + bdiag[i]; 97 ierr = PetscMemcpy(pv,rtmp+bs2*pj[0],bs2*sizeof(MatScalar));CHKERRQ(ierr); 98 /* ierr = Kernel_A_gets_inverse_A(bs,pv,v_pivots,v_work);CHKERRQ(ierr); */ 99 ierr = Kernel_A_gets_inverse_A_2(pv,shift);CHKERRQ(ierr); 100 101 /* U part */ 102 pv = b->a + bs2*(bdiag[i+1]+1); 103 pj = b->j + bdiag[i+1]+1; 104 nz = bdiag[i] - bdiag[i+1] - 1; 105 for (j=0; j<nz; j++){ 106 ierr = PetscMemcpy(pv+bs2*j,rtmp+bs2*pj[j],bs2*sizeof(MatScalar));CHKERRQ(ierr); 107 } 108 } 109 110 ierr = PetscFree2(rtmp,mwork);CHKERRQ(ierr); 111 ierr = ISRestoreIndices(isicol,&ic);CHKERRQ(ierr); 112 ierr = ISRestoreIndices(isrow,&r);CHKERRQ(ierr); 113 C->ops->solve = MatSolve_SeqBAIJ_2; 114 C->ops->solvetranspose = MatSolveTranspose_SeqBAIJ_2; 115 116 C->assembled = PETSC_TRUE; 117 ierr = PetscLogFlops(1.333333333333*2*2*2*n);CHKERRQ(ierr); /* from inverting diagonal blocks */ 118 PetscFunctionReturn(0); 119 } 120 121 #undef __FUNCT__ 122 #define __FUNCT__ "MatLUFactorNumeric_SeqBAIJ_2_NaturalOrdering" 123 PetscErrorCode MatLUFactorNumeric_SeqBAIJ_2_NaturalOrdering(Mat B,Mat A,const MatFactorInfo *info) 124 { 125 Mat C=B; 126 Mat_SeqBAIJ *a=(Mat_SeqBAIJ*)A->data,*b=(Mat_SeqBAIJ *)C->data; 127 PetscErrorCode ierr; 128 PetscInt i,j,k,nz,nzL,row,*pj; 129 const PetscInt n=a->mbs,*ai=a->i,*aj=a->j,*bi=b->i,*bj=b->j,bs2=a->bs2; 130 const PetscInt *ajtmp,*bjtmp,*bdiag=b->diag; 131 MatScalar *rtmp,*pc,*mwork,*pv; 132 MatScalar *aa=a->a,*v; 133 PetscInt flg; 134 PetscReal shift = info->shiftinblocks; 135 136 PetscFunctionBegin; 137 /* generate work space needed by the factorization */ 138 ierr = PetscMalloc2(bs2*n,MatScalar,&rtmp,bs2,MatScalar,&mwork);CHKERRQ(ierr); 139 ierr = PetscMemzero(rtmp,bs2*n*sizeof(MatScalar));CHKERRQ(ierr); 140 141 for (i=0; i<n; i++){ 142 /* zero rtmp */ 143 /* L part */ 144 nz = bi[i+1] - bi[i]; 145 bjtmp = bj + bi[i]; 146 for (j=0; j<nz; j++){ 147 ierr = PetscMemzero(rtmp+bs2*bjtmp[j],bs2*sizeof(MatScalar));CHKERRQ(ierr); 148 } 149 150 /* U part */ 151 nz = bdiag[i] - bdiag[i+1]; 152 bjtmp = bj + bdiag[i+1]+1; 153 for (j=0; j<nz; j++){ 154 ierr = PetscMemzero(rtmp+bs2*bjtmp[j],bs2*sizeof(MatScalar));CHKERRQ(ierr); 155 } 156 157 /* load in initial (unfactored row) */ 158 nz = ai[i+1] - ai[i]; 159 ajtmp = aj + ai[i]; 160 v = aa + bs2*ai[i]; 161 for (j=0; j<nz; j++) { 162 ierr = PetscMemcpy(rtmp+bs2*ajtmp[j],v+bs2*j,bs2*sizeof(MatScalar));CHKERRQ(ierr); 163 } 164 165 /* elimination */ 166 bjtmp = bj + bi[i]; 167 nzL = bi[i+1] - bi[i]; 168 for(k=0;k < nzL;k++) { 169 row = bjtmp[k]; 170 pc = rtmp + bs2*row; 171 for (flg=0,j=0; j<bs2; j++) { if (pc[j]!=0.0) { flg = 1; break; }} 172 if (flg) { 173 pv = b->a + bs2*bdiag[row]; 174 /* Kernel_A_gets_A_times_B(bs,pc,pv,mwork); *pc = *pc * (*pv); */ 175 ierr = Kernel_A_gets_A_times_B_2(pc,pv,mwork);CHKERRQ(ierr); 176 177 pj = b->j + bdiag[row+1]+1; /* beginning of U(row,:) */ 178 pv = b->a + bs2*(bdiag[row+1]+1); 179 nz = bdiag[row]-bdiag[row+1] - 1; /* num of entries in U(row,:) excluding diag */ 180 for (j=0; j<nz; j++) { 181 /* Kernel_A_gets_A_minus_B_times_C(bs,rtmp+bs2*pj[j],pc,pv+bs2*j); */ 182 /* rtmp+bs2*pj[j] = rtmp+bs2*pj[j] - (*pc)*(pv+bs2*j) */ 183 v = rtmp + 4*pj[j]; 184 ierr = Kernel_A_gets_A_minus_B_times_C_2(v,pc,pv);CHKERRQ(ierr); 185 pv += 4; 186 } 187 ierr = PetscLogFlops(16*nz+12);CHKERRQ(ierr); /* flops = 2*bs^3*nz + 2*bs^3 - bs2) */ 188 } 189 } 190 191 /* finished row so stick it into b->a */ 192 /* L part */ 193 pv = b->a + bs2*bi[i] ; 194 pj = b->j + bi[i] ; 195 nz = bi[i+1] - bi[i]; 196 for (j=0; j<nz; j++) { 197 ierr = PetscMemcpy(pv+bs2*j,rtmp+bs2*pj[j],bs2*sizeof(MatScalar));CHKERRQ(ierr); 198 } 199 200 /* Mark diagonal and invert diagonal for simplier triangular solves */ 201 pv = b->a + bs2*bdiag[i]; 202 pj = b->j + bdiag[i]; 203 ierr = PetscMemcpy(pv,rtmp+bs2*pj[0],bs2*sizeof(MatScalar));CHKERRQ(ierr); 204 /* ierr = Kernel_A_gets_inverse_A(bs,pv,v_pivots,v_work);CHKERRQ(ierr); */ 205 ierr = Kernel_A_gets_inverse_A_2(pv,shift);CHKERRQ(ierr); 206 207 /* U part */ 208 /* 209 pv = b->a + bs2*bi[2*n-i]; 210 pj = b->j + bi[2*n-i]; 211 nz = bi[2*n-i+1] - bi[2*n-i] - 1; 212 */ 213 pv = b->a + bs2*(bdiag[i+1]+1); 214 pj = b->j + bdiag[i+1]+1; 215 nz = bdiag[i] - bdiag[i+1] - 1; 216 for (j=0; j<nz; j++){ 217 ierr = PetscMemcpy(pv+bs2*j,rtmp+bs2*pj[j],bs2*sizeof(MatScalar));CHKERRQ(ierr); 218 } 219 } 220 ierr = PetscFree2(rtmp,mwork);CHKERRQ(ierr); 221 222 C->ops->solve = MatSolve_SeqBAIJ_2_NaturalOrdering; 223 C->ops->solvetranspose = MatSolveTranspose_SeqBAIJ_2_NaturalOrdering; 224 C->assembled = PETSC_TRUE; 225 ierr = PetscLogFlops(1.333333333333*2*2*2*n);CHKERRQ(ierr); /* from inverting diagonal blocks */ 226 PetscFunctionReturn(0); 227 } 228 229 #undef __FUNCT__ 230 #define __FUNCT__ "MatLUFactorNumeric_SeqBAIJ_2_inplace" 231 PetscErrorCode MatLUFactorNumeric_SeqBAIJ_2_inplace(Mat B,Mat A,const MatFactorInfo *info) 232 { 233 Mat C = B; 234 Mat_SeqBAIJ *a = (Mat_SeqBAIJ*)A->data,*b = (Mat_SeqBAIJ *)C->data; 235 IS isrow = b->row,isicol = b->icol; 236 PetscErrorCode ierr; 237 const PetscInt *r,*ic; 238 PetscInt i,j,n = a->mbs,*bi = b->i,*bj = b->j; 239 PetscInt *ajtmpold,*ajtmp,nz,row; 240 PetscInt *diag_offset=b->diag,idx,*ai=a->i,*aj=a->j,*pj; 241 MatScalar *pv,*v,*rtmp,m1,m2,m3,m4,*pc,*w,*x,x1,x2,x3,x4; 242 MatScalar p1,p2,p3,p4; 243 MatScalar *ba = b->a,*aa = a->a; 244 PetscReal shift = info->shiftinblocks; 245 246 PetscFunctionBegin; 247 ierr = ISGetIndices(isrow,&r);CHKERRQ(ierr); 248 ierr = ISGetIndices(isicol,&ic);CHKERRQ(ierr); 249 ierr = PetscMalloc(4*(n+1)*sizeof(MatScalar),&rtmp);CHKERRQ(ierr); 250 251 for (i=0; i<n; i++) { 252 nz = bi[i+1] - bi[i]; 253 ajtmp = bj + bi[i]; 254 for (j=0; j<nz; j++) { 255 x = rtmp+4*ajtmp[j]; x[0] = x[1] = x[2] = x[3] = 0.0; 256 } 257 /* load in initial (unfactored row) */ 258 idx = r[i]; 259 nz = ai[idx+1] - ai[idx]; 260 ajtmpold = aj + ai[idx]; 261 v = aa + 4*ai[idx]; 262 for (j=0; j<nz; j++) { 263 x = rtmp+4*ic[ajtmpold[j]]; 264 x[0] = v[0]; x[1] = v[1]; x[2] = v[2]; x[3] = v[3]; 265 v += 4; 266 } 267 row = *ajtmp++; 268 while (row < i) { 269 pc = rtmp + 4*row; 270 p1 = pc[0]; p2 = pc[1]; p3 = pc[2]; p4 = pc[3]; 271 if (p1 != 0.0 || p2 != 0.0 || p3 != 0.0 || p4 != 0.0) { 272 pv = ba + 4*diag_offset[row]; 273 pj = bj + diag_offset[row] + 1; 274 x1 = pv[0]; x2 = pv[1]; x3 = pv[2]; x4 = pv[3]; 275 pc[0] = m1 = p1*x1 + p3*x2; 276 pc[1] = m2 = p2*x1 + p4*x2; 277 pc[2] = m3 = p1*x3 + p3*x4; 278 pc[3] = m4 = p2*x3 + p4*x4; 279 nz = bi[row+1] - diag_offset[row] - 1; 280 pv += 4; 281 for (j=0; j<nz; j++) { 282 x1 = pv[0]; x2 = pv[1]; x3 = pv[2]; x4 = pv[3]; 283 x = rtmp + 4*pj[j]; 284 x[0] -= m1*x1 + m3*x2; 285 x[1] -= m2*x1 + m4*x2; 286 x[2] -= m1*x3 + m3*x4; 287 x[3] -= m2*x3 + m4*x4; 288 pv += 4; 289 } 290 ierr = PetscLogFlops(16.0*nz+12.0);CHKERRQ(ierr); 291 } 292 row = *ajtmp++; 293 } 294 /* finished row so stick it into b->a */ 295 pv = ba + 4*bi[i]; 296 pj = bj + bi[i]; 297 nz = bi[i+1] - bi[i]; 298 for (j=0; j<nz; j++) { 299 x = rtmp+4*pj[j]; 300 pv[0] = x[0]; pv[1] = x[1]; pv[2] = x[2]; pv[3] = x[3]; 301 pv += 4; 302 } 303 /* invert diagonal block */ 304 w = ba + 4*diag_offset[i]; 305 ierr = Kernel_A_gets_inverse_A_2(w,shift);CHKERRQ(ierr); 306 } 307 308 ierr = PetscFree(rtmp);CHKERRQ(ierr); 309 ierr = ISRestoreIndices(isicol,&ic);CHKERRQ(ierr); 310 ierr = ISRestoreIndices(isrow,&r);CHKERRQ(ierr); 311 C->ops->solve = MatSolve_SeqBAIJ_2_inplace; 312 C->ops->solvetranspose = MatSolveTranspose_SeqBAIJ_2_inplace; 313 C->assembled = PETSC_TRUE; 314 ierr = PetscLogFlops(1.333333333333*8*b->mbs);CHKERRQ(ierr); /* from inverting diagonal blocks */ 315 PetscFunctionReturn(0); 316 } 317 /* 318 Version for when blocks are 2 by 2 Using natural ordering 319 */ 320 #undef __FUNCT__ 321 #define __FUNCT__ "MatLUFactorNumeric_SeqBAIJ_2_NaturalOrdering_inplace" 322 PetscErrorCode MatLUFactorNumeric_SeqBAIJ_2_NaturalOrdering_inplace(Mat C,Mat A,const MatFactorInfo *info) 323 { 324 Mat_SeqBAIJ *a = (Mat_SeqBAIJ*)A->data,*b = (Mat_SeqBAIJ *)C->data; 325 PetscErrorCode ierr; 326 PetscInt i,j,n = a->mbs,*bi = b->i,*bj = b->j; 327 PetscInt *ajtmpold,*ajtmp,nz,row; 328 PetscInt *diag_offset = b->diag,*ai=a->i,*aj=a->j,*pj; 329 MatScalar *pv,*v,*rtmp,*pc,*w,*x; 330 MatScalar p1,p2,p3,p4,m1,m2,m3,m4,x1,x2,x3,x4; 331 MatScalar *ba = b->a,*aa = a->a; 332 PetscReal shift = info->shiftinblocks; 333 334 PetscFunctionBegin; 335 ierr = PetscMalloc(4*(n+1)*sizeof(MatScalar),&rtmp);CHKERRQ(ierr); 336 for (i=0; i<n; i++) { 337 nz = bi[i+1] - bi[i]; 338 ajtmp = bj + bi[i]; 339 for (j=0; j<nz; j++) { 340 x = rtmp+4*ajtmp[j]; 341 x[0] = x[1] = x[2] = x[3] = 0.0; 342 } 343 /* load in initial (unfactored row) */ 344 nz = ai[i+1] - ai[i]; 345 ajtmpold = aj + ai[i]; 346 v = aa + 4*ai[i]; 347 for (j=0; j<nz; j++) { 348 x = rtmp+4*ajtmpold[j]; 349 x[0] = v[0]; x[1] = v[1]; x[2] = v[2]; x[3] = v[3]; 350 v += 4; 351 } 352 row = *ajtmp++; 353 while (row < i) { 354 pc = rtmp + 4*row; 355 p1 = pc[0]; p2 = pc[1]; p3 = pc[2]; p4 = pc[3]; 356 if (p1 != 0.0 || p2 != 0.0 || p3 != 0.0 || p4 != 0.0) { 357 pv = ba + 4*diag_offset[row]; 358 pj = bj + diag_offset[row] + 1; 359 x1 = pv[0]; x2 = pv[1]; x3 = pv[2]; x4 = pv[3]; 360 pc[0] = m1 = p1*x1 + p3*x2; 361 pc[1] = m2 = p2*x1 + p4*x2; 362 pc[2] = m3 = p1*x3 + p3*x4; 363 pc[3] = m4 = p2*x3 + p4*x4; 364 nz = bi[row+1] - diag_offset[row] - 1; 365 pv += 4; 366 for (j=0; j<nz; j++) { 367 x1 = pv[0]; x2 = pv[1]; x3 = pv[2]; x4 = pv[3]; 368 x = rtmp + 4*pj[j]; 369 x[0] -= m1*x1 + m3*x2; 370 x[1] -= m2*x1 + m4*x2; 371 x[2] -= m1*x3 + m3*x4; 372 x[3] -= m2*x3 + m4*x4; 373 pv += 4; 374 } 375 ierr = PetscLogFlops(16.0*nz+12.0);CHKERRQ(ierr); 376 } 377 row = *ajtmp++; 378 } 379 /* finished row so stick it into b->a */ 380 pv = ba + 4*bi[i]; 381 pj = bj + bi[i]; 382 nz = bi[i+1] - bi[i]; 383 for (j=0; j<nz; j++) { 384 x = rtmp+4*pj[j]; 385 pv[0] = x[0]; pv[1] = x[1]; pv[2] = x[2]; pv[3] = x[3]; 386 /* 387 printf(" col %d:",pj[j]); 388 PetscInt j1; 389 for (j1=0; j1<4; j1++) printf(" %g,",*(pv+j1)); 390 printf("\n"); 391 */ 392 pv += 4; 393 } 394 /* invert diagonal block */ 395 w = ba + 4*diag_offset[i]; 396 /* 397 printf(" \n%d -th: diag: ",i); 398 for (j=0; j<4; j++){ 399 printf(" %g,",w[j]); 400 } 401 printf("\n----------------------------\n"); 402 */ 403 ierr = Kernel_A_gets_inverse_A_2(w,shift);CHKERRQ(ierr); 404 } 405 406 ierr = PetscFree(rtmp);CHKERRQ(ierr); 407 C->ops->solve = MatSolve_SeqBAIJ_2_NaturalOrdering_inplace; 408 C->ops->solvetranspose = MatSolveTranspose_SeqBAIJ_2_NaturalOrdering_inplace; 409 C->assembled = PETSC_TRUE; 410 ierr = PetscLogFlops(1.333333333333*8*b->mbs);CHKERRQ(ierr); /* from inverting diagonal blocks */ 411 PetscFunctionReturn(0); 412 } 413 414 /* ----------------------------------------------------------- */ 415 /* 416 Version for when blocks are 1 by 1. 417 */ 418 #undef __FUNCT__ 419 #define __FUNCT__ "MatLUFactorNumeric_SeqBAIJ_1_inplace" 420 PetscErrorCode MatLUFactorNumeric_SeqBAIJ_1_inplace(Mat C,Mat A,const MatFactorInfo *info) 421 { 422 Mat_SeqBAIJ *a = (Mat_SeqBAIJ*)A->data,*b = (Mat_SeqBAIJ *)C->data; 423 IS isrow = b->row,isicol = b->icol; 424 PetscErrorCode ierr; 425 const PetscInt *r,*ic; 426 PetscInt i,j,n = a->mbs,*bi = b->i,*bj = b->j; 427 PetscInt *ajtmpold,*ajtmp,nz,row,*ai = a->i,*aj = a->j; 428 PetscInt *diag_offset = b->diag,diag,*pj; 429 MatScalar *pv,*v,*rtmp,multiplier,*pc; 430 MatScalar *ba = b->a,*aa = a->a; 431 PetscTruth row_identity, col_identity; 432 433 PetscFunctionBegin; 434 ierr = ISGetIndices(isrow,&r);CHKERRQ(ierr); 435 ierr = ISGetIndices(isicol,&ic);CHKERRQ(ierr); 436 ierr = PetscMalloc((n+1)*sizeof(MatScalar),&rtmp);CHKERRQ(ierr); 437 438 for (i=0; i<n; i++) { 439 nz = bi[i+1] - bi[i]; 440 ajtmp = bj + bi[i]; 441 for (j=0; j<nz; j++) rtmp[ajtmp[j]] = 0.0; 442 443 /* load in initial (unfactored row) */ 444 nz = ai[r[i]+1] - ai[r[i]]; 445 ajtmpold = aj + ai[r[i]]; 446 v = aa + ai[r[i]]; 447 for (j=0; j<nz; j++) rtmp[ic[ajtmpold[j]]] = v[j]; 448 449 row = *ajtmp++; 450 while (row < i) { 451 pc = rtmp + row; 452 if (*pc != 0.0) { 453 pv = ba + diag_offset[row]; 454 pj = bj + diag_offset[row] + 1; 455 multiplier = *pc * *pv++; 456 *pc = multiplier; 457 nz = bi[row+1] - diag_offset[row] - 1; 458 for (j=0; j<nz; j++) rtmp[pj[j]] -= multiplier * pv[j]; 459 ierr = PetscLogFlops(1.0+2.0*nz);CHKERRQ(ierr); 460 } 461 row = *ajtmp++; 462 } 463 /* finished row so stick it into b->a */ 464 pv = ba + bi[i]; 465 pj = bj + bi[i]; 466 nz = bi[i+1] - bi[i]; 467 for (j=0; j<nz; j++) {pv[j] = rtmp[pj[j]];} 468 diag = diag_offset[i] - bi[i]; 469 /* check pivot entry for current row */ 470 if (pv[diag] == 0.0) { 471 SETERRQ2(PETSC_ERR_MAT_LU_ZRPVT,"Zero pivot: row in original ordering %D in permuted ordering %D",r[i],i); 472 } 473 pv[diag] = 1.0/pv[diag]; 474 } 475 476 ierr = PetscFree(rtmp);CHKERRQ(ierr); 477 ierr = ISRestoreIndices(isicol,&ic);CHKERRQ(ierr); 478 ierr = ISRestoreIndices(isrow,&r);CHKERRQ(ierr); 479 ierr = ISIdentity(isrow,&row_identity);CHKERRQ(ierr); 480 ierr = ISIdentity(isicol,&col_identity);CHKERRQ(ierr); 481 if (row_identity && col_identity) { 482 C->ops->solve = MatSolve_SeqBAIJ_1_NaturalOrdering_inplace; 483 C->ops->solvetranspose = MatSolveTranspose_SeqBAIJ_1_NaturalOrdering_inplace; 484 } else { 485 C->ops->solve = MatSolve_SeqBAIJ_1_inplace; 486 C->ops->solvetranspose = MatSolveTranspose_SeqBAIJ_1_inplace; 487 } 488 C->assembled = PETSC_TRUE; 489 ierr = PetscLogFlops(C->cmap->n);CHKERRQ(ierr); 490 PetscFunctionReturn(0); 491 } 492 493 EXTERN_C_BEGIN 494 #undef __FUNCT__ 495 #define __FUNCT__ "MatGetFactor_seqbaij_petsc" 496 PetscErrorCode MatGetFactor_seqbaij_petsc(Mat A,MatFactorType ftype,Mat *B) 497 { 498 PetscInt n = A->rmap->n; 499 PetscErrorCode ierr; 500 501 PetscFunctionBegin; 502 ierr = MatCreate(((PetscObject)A)->comm,B);CHKERRQ(ierr); 503 ierr = MatSetSizes(*B,n,n,n,n);CHKERRQ(ierr); 504 if (ftype == MAT_FACTOR_LU || ftype == MAT_FACTOR_ILU || ftype == MAT_FACTOR_ILUDT) { 505 ierr = MatSetType(*B,MATSEQBAIJ);CHKERRQ(ierr); 506 (*B)->ops->lufactorsymbolic = MatLUFactorSymbolic_SeqBAIJ; 507 (*B)->ops->ilufactorsymbolic = MatILUFactorSymbolic_SeqBAIJ; 508 } else if (ftype == MAT_FACTOR_CHOLESKY || ftype == MAT_FACTOR_ICC) { 509 ierr = MatSetType(*B,MATSEQSBAIJ);CHKERRQ(ierr); 510 ierr = MatSeqSBAIJSetPreallocation(*B,1,MAT_SKIP_ALLOCATION,PETSC_NULL);CHKERRQ(ierr); 511 (*B)->ops->iccfactorsymbolic = MatICCFactorSymbolic_SeqBAIJ; 512 (*B)->ops->choleskyfactorsymbolic = MatCholeskyFactorSymbolic_SeqBAIJ; 513 } else SETERRQ(PETSC_ERR_SUP,"Factor type not supported"); 514 (*B)->factor = ftype; 515 PetscFunctionReturn(0); 516 } 517 EXTERN_C_END 518 519 EXTERN_C_BEGIN 520 #undef __FUNCT__ 521 #define __FUNCT__ "MatGetFactorAvailable_seqaij_petsc" 522 PetscErrorCode MatGetFactorAvailable_seqbaij_petsc(Mat A,MatFactorType ftype,PetscTruth *flg) 523 { 524 PetscFunctionBegin; 525 *flg = PETSC_TRUE; 526 PetscFunctionReturn(0); 527 } 528 EXTERN_C_END 529 530 /* ----------------------------------------------------------- */ 531 #undef __FUNCT__ 532 #define __FUNCT__ "MatLUFactor_SeqBAIJ" 533 PetscErrorCode MatLUFactor_SeqBAIJ(Mat A,IS row,IS col,const MatFactorInfo *info) 534 { 535 PetscErrorCode ierr; 536 Mat C; 537 538 PetscFunctionBegin; 539 ierr = MatGetFactor(A,MAT_SOLVER_PETSC,MAT_FACTOR_LU,&C);CHKERRQ(ierr); 540 ierr = MatLUFactorSymbolic(C,A,row,col,info);CHKERRQ(ierr); 541 ierr = MatLUFactorNumeric(C,A,info);CHKERRQ(ierr); 542 A->ops->solve = C->ops->solve; 543 A->ops->solvetranspose = C->ops->solvetranspose; 544 ierr = MatHeaderCopy(A,C);CHKERRQ(ierr); 545 ierr = PetscLogObjectParent(A,((Mat_SeqBAIJ*)(A->data))->icol);CHKERRQ(ierr); 546 PetscFunctionReturn(0); 547 } 548 549 #include "../src/mat/impls/sbaij/seq/sbaij.h" 550 #undef __FUNCT__ 551 #define __FUNCT__ "MatCholeskyFactorNumeric_SeqBAIJ_N" 552 PetscErrorCode MatCholeskyFactorNumeric_SeqBAIJ_N(Mat C,Mat A,const MatFactorInfo *info) 553 { 554 PetscErrorCode ierr; 555 Mat_SeqBAIJ *a=(Mat_SeqBAIJ*)A->data; 556 Mat_SeqSBAIJ *b=(Mat_SeqSBAIJ*)C->data; 557 IS ip=b->row; 558 const PetscInt *rip; 559 PetscInt i,j,mbs=a->mbs,bs=A->rmap->bs,*bi=b->i,*bj=b->j,*bcol; 560 PetscInt *ai=a->i,*aj=a->j; 561 PetscInt k,jmin,jmax,*jl,*il,col,nexti,ili,nz; 562 MatScalar *rtmp,*ba=b->a,*bval,*aa=a->a,dk,uikdi; 563 PetscReal zeropivot,rs,shiftnz; 564 PetscReal shiftpd; 565 ChShift_Ctx sctx; 566 PetscInt newshift; 567 568 PetscFunctionBegin; 569 if (bs > 1) { 570 if (!a->sbaijMat){ 571 ierr = MatConvert(A,MATSEQSBAIJ,MAT_INITIAL_MATRIX,&a->sbaijMat);CHKERRQ(ierr); 572 } 573 ierr = (a->sbaijMat)->ops->choleskyfactornumeric(C,a->sbaijMat,info);CHKERRQ(ierr); 574 ierr = MatDestroy(a->sbaijMat);CHKERRQ(ierr); 575 a->sbaijMat = PETSC_NULL; 576 PetscFunctionReturn(0); 577 } 578 579 /* initialization */ 580 shiftnz = info->shiftnz; 581 shiftpd = info->shiftpd; 582 zeropivot = info->zeropivot; 583 584 ierr = ISGetIndices(ip,&rip);CHKERRQ(ierr); 585 ierr = PetscMalloc3(mbs,MatScalar,&rtmp,mbs,PetscInt,&il,mbs,PetscInt,&jl);CHKERRQ(ierr); 586 587 sctx.shift_amount = 0.; 588 sctx.nshift = 0; 589 do { 590 sctx.chshift = PETSC_FALSE; 591 for (i=0; i<mbs; i++) { 592 rtmp[i] = 0.0; jl[i] = mbs; il[0] = 0; 593 } 594 595 for (k = 0; k<mbs; k++){ 596 bval = ba + bi[k]; 597 /* initialize k-th row by the perm[k]-th row of A */ 598 jmin = ai[rip[k]]; jmax = ai[rip[k]+1]; 599 for (j = jmin; j < jmax; j++){ 600 col = rip[aj[j]]; 601 if (col >= k){ /* only take upper triangular entry */ 602 rtmp[col] = aa[j]; 603 *bval++ = 0.0; /* for in-place factorization */ 604 } 605 } 606 607 /* shift the diagonal of the matrix */ 608 if (sctx.nshift) rtmp[k] += sctx.shift_amount; 609 610 /* modify k-th row by adding in those rows i with U(i,k)!=0 */ 611 dk = rtmp[k]; 612 i = jl[k]; /* first row to be added to k_th row */ 613 614 while (i < k){ 615 nexti = jl[i]; /* next row to be added to k_th row */ 616 617 /* compute multiplier, update diag(k) and U(i,k) */ 618 ili = il[i]; /* index of first nonzero element in U(i,k:bms-1) */ 619 uikdi = - ba[ili]*ba[bi[i]]; /* diagonal(k) */ 620 dk += uikdi*ba[ili]; 621 ba[ili] = uikdi; /* -U(i,k) */ 622 623 /* add multiple of row i to k-th row */ 624 jmin = ili + 1; jmax = bi[i+1]; 625 if (jmin < jmax){ 626 for (j=jmin; j<jmax; j++) rtmp[bj[j]] += uikdi*ba[j]; 627 /* update il and jl for row i */ 628 il[i] = jmin; 629 j = bj[jmin]; jl[i] = jl[j]; jl[j] = i; 630 } 631 i = nexti; 632 } 633 634 /* shift the diagonals when zero pivot is detected */ 635 /* compute rs=sum of abs(off-diagonal) */ 636 rs = 0.0; 637 jmin = bi[k]+1; 638 nz = bi[k+1] - jmin; 639 if (nz){ 640 bcol = bj + jmin; 641 while (nz--){ 642 rs += PetscAbsScalar(rtmp[*bcol]); 643 bcol++; 644 } 645 } 646 647 sctx.rs = rs; 648 sctx.pv = dk; 649 ierr = MatCholeskyCheckShift_inline(info,sctx,k,newshift);CHKERRQ(ierr); 650 if (newshift == 1) break; 651 652 /* copy data into U(k,:) */ 653 ba[bi[k]] = 1.0/dk; /* U(k,k) */ 654 jmin = bi[k]+1; jmax = bi[k+1]; 655 if (jmin < jmax) { 656 for (j=jmin; j<jmax; j++){ 657 col = bj[j]; ba[j] = rtmp[col]; rtmp[col] = 0.0; 658 } 659 /* add the k-th row into il and jl */ 660 il[k] = jmin; 661 i = bj[jmin]; jl[k] = jl[i]; jl[i] = k; 662 } 663 } 664 } while (sctx.chshift); 665 ierr = PetscFree3(rtmp,il,jl);CHKERRQ(ierr); 666 667 ierr = ISRestoreIndices(ip,&rip);CHKERRQ(ierr); 668 C->assembled = PETSC_TRUE; 669 C->preallocated = PETSC_TRUE; 670 ierr = PetscLogFlops(C->rmap->N);CHKERRQ(ierr); 671 if (sctx.nshift){ 672 if (shiftpd) { 673 ierr = PetscInfo2(A,"number of shiftpd tries %D, shift_amount %G\n",sctx.nshift,sctx.shift_amount);CHKERRQ(ierr); 674 } else if (shiftnz) { 675 ierr = PetscInfo2(A,"number of shiftnz tries %D, shift_amount %G\n",sctx.nshift,sctx.shift_amount);CHKERRQ(ierr); 676 } 677 } 678 PetscFunctionReturn(0); 679 } 680 681 #undef __FUNCT__ 682 #define __FUNCT__ "MatCholeskyFactorNumeric_SeqBAIJ_N_NaturalOrdering" 683 PetscErrorCode MatCholeskyFactorNumeric_SeqBAIJ_N_NaturalOrdering(Mat C,Mat A,const MatFactorInfo *info) 684 { 685 Mat_SeqBAIJ *a=(Mat_SeqBAIJ*)A->data; 686 Mat_SeqSBAIJ *b=(Mat_SeqSBAIJ*)C->data; 687 PetscErrorCode ierr; 688 PetscInt i,j,am=a->mbs; 689 PetscInt *ai=a->i,*aj=a->j,*bi=b->i,*bj=b->j; 690 PetscInt k,jmin,*jl,*il,nexti,ili,*acol,*bcol,nz; 691 MatScalar *rtmp,*ba=b->a,*aa=a->a,dk,uikdi,*aval,*bval; 692 PetscReal zeropivot,rs,shiftnz; 693 PetscReal shiftpd; 694 ChShift_Ctx sctx; 695 PetscInt newshift; 696 697 PetscFunctionBegin; 698 /* initialization */ 699 shiftnz = info->shiftnz; 700 shiftpd = info->shiftpd; 701 zeropivot = info->zeropivot; 702 703 ierr = PetscMalloc3(am,MatScalar,&rtmp,am,PetscInt,&il,am,PetscInt,&jl);CHKERRQ(ierr); 704 705 sctx.shift_amount = 0.; 706 sctx.nshift = 0; 707 do { 708 sctx.chshift = PETSC_FALSE; 709 for (i=0; i<am; i++) { 710 rtmp[i] = 0.0; jl[i] = am; il[0] = 0; 711 } 712 713 for (k = 0; k<am; k++){ 714 /* initialize k-th row with elements nonzero in row perm(k) of A */ 715 nz = ai[k+1] - ai[k]; 716 acol = aj + ai[k]; 717 aval = aa + ai[k]; 718 bval = ba + bi[k]; 719 while (nz -- ){ 720 if (*acol < k) { /* skip lower triangular entries */ 721 acol++; aval++; 722 } else { 723 rtmp[*acol++] = *aval++; 724 *bval++ = 0.0; /* for in-place factorization */ 725 } 726 } 727 728 /* shift the diagonal of the matrix */ 729 if (sctx.nshift) rtmp[k] += sctx.shift_amount; 730 731 /* modify k-th row by adding in those rows i with U(i,k)!=0 */ 732 dk = rtmp[k]; 733 i = jl[k]; /* first row to be added to k_th row */ 734 735 while (i < k){ 736 nexti = jl[i]; /* next row to be added to k_th row */ 737 /* compute multiplier, update D(k) and U(i,k) */ 738 ili = il[i]; /* index of first nonzero element in U(i,k:bms-1) */ 739 uikdi = - ba[ili]*ba[bi[i]]; 740 dk += uikdi*ba[ili]; 741 ba[ili] = uikdi; /* -U(i,k) */ 742 743 /* add multiple of row i to k-th row ... */ 744 jmin = ili + 1; 745 nz = bi[i+1] - jmin; 746 if (nz > 0){ 747 bcol = bj + jmin; 748 bval = ba + jmin; 749 while (nz --) rtmp[*bcol++] += uikdi*(*bval++); 750 /* update il and jl for i-th row */ 751 il[i] = jmin; 752 j = bj[jmin]; jl[i] = jl[j]; jl[j] = i; 753 } 754 i = nexti; 755 } 756 757 /* shift the diagonals when zero pivot is detected */ 758 /* compute rs=sum of abs(off-diagonal) */ 759 rs = 0.0; 760 jmin = bi[k]+1; 761 nz = bi[k+1] - jmin; 762 if (nz){ 763 bcol = bj + jmin; 764 while (nz--){ 765 rs += PetscAbsScalar(rtmp[*bcol]); 766 bcol++; 767 } 768 } 769 770 sctx.rs = rs; 771 sctx.pv = dk; 772 ierr = MatCholeskyCheckShift_inline(info,sctx,k,newshift);CHKERRQ(ierr); 773 if (newshift == 1) break; /* sctx.shift_amount is updated */ 774 775 /* copy data into U(k,:) */ 776 ba[bi[k]] = 1.0/dk; 777 jmin = bi[k]+1; 778 nz = bi[k+1] - jmin; 779 if (nz){ 780 bcol = bj + jmin; 781 bval = ba + jmin; 782 while (nz--){ 783 *bval++ = rtmp[*bcol]; 784 rtmp[*bcol++] = 0.0; 785 } 786 /* add k-th row into il and jl */ 787 il[k] = jmin; 788 i = bj[jmin]; jl[k] = jl[i]; jl[i] = k; 789 } 790 } 791 } while (sctx.chshift); 792 ierr = PetscFree3(rtmp,il,jl);CHKERRQ(ierr); 793 794 C->ops->solve = MatSolve_SeqSBAIJ_1_NaturalOrdering_inplace; 795 C->ops->solvetranspose = MatSolve_SeqSBAIJ_1_NaturalOrdering_inplace; 796 C->assembled = PETSC_TRUE; 797 C->preallocated = PETSC_TRUE; 798 ierr = PetscLogFlops(C->rmap->N);CHKERRQ(ierr); 799 if (sctx.nshift){ 800 if (shiftnz) { 801 ierr = PetscInfo2(A,"number of shiftnz tries %D, shift_amount %G\n",sctx.nshift,sctx.shift_amount);CHKERRQ(ierr); 802 } else if (shiftpd) { 803 ierr = PetscInfo2(A,"number of shiftpd tries %D, shift_amount %G\n",sctx.nshift,sctx.shift_amount);CHKERRQ(ierr); 804 } 805 } 806 PetscFunctionReturn(0); 807 } 808 809 #include "petscbt.h" 810 #include "../src/mat/utils/freespace.h" 811 #undef __FUNCT__ 812 #define __FUNCT__ "MatICCFactorSymbolic_SeqBAIJ" 813 PetscErrorCode MatICCFactorSymbolic_SeqBAIJ(Mat fact,Mat A,IS perm,const MatFactorInfo *info) 814 { 815 Mat_SeqBAIJ *a = (Mat_SeqBAIJ*)A->data; 816 Mat_SeqSBAIJ *b; 817 Mat B; 818 PetscErrorCode ierr; 819 PetscTruth perm_identity; 820 PetscInt reallocs=0,i,*ai=a->i,*aj=a->j,am=a->mbs,bs=A->rmap->bs,*ui; 821 const PetscInt *rip; 822 PetscInt jmin,jmax,nzk,k,j,*jl,prow,*il,nextprow; 823 PetscInt nlnk,*lnk,*lnk_lvl=PETSC_NULL,ncols,ncols_upper,*cols,*cols_lvl,*uj,**uj_ptr,**uj_lvl_ptr; 824 PetscReal fill=info->fill,levels=info->levels; 825 PetscFreeSpaceList free_space=PETSC_NULL,current_space=PETSC_NULL; 826 PetscFreeSpaceList free_space_lvl=PETSC_NULL,current_space_lvl=PETSC_NULL; 827 PetscBT lnkbt; 828 829 PetscFunctionBegin; 830 if (bs > 1){ 831 if (!a->sbaijMat){ 832 ierr = MatConvert(A,MATSEQSBAIJ,MAT_INITIAL_MATRIX,&a->sbaijMat);CHKERRQ(ierr); 833 } 834 (fact)->ops->iccfactorsymbolic = MatICCFactorSymbolic_SeqSBAIJ; /* undue the change made in MatGetFactor_seqbaij_petsc */ 835 ierr = MatICCFactorSymbolic(fact,a->sbaijMat,perm,info);CHKERRQ(ierr); 836 PetscFunctionReturn(0); 837 } 838 839 ierr = ISIdentity(perm,&perm_identity);CHKERRQ(ierr); 840 ierr = ISGetIndices(perm,&rip);CHKERRQ(ierr); 841 842 /* special case that simply copies fill pattern */ 843 if (!levels && perm_identity) { 844 ierr = MatMarkDiagonal_SeqBAIJ(A);CHKERRQ(ierr); 845 ierr = PetscMalloc((am+1)*sizeof(PetscInt),&ui);CHKERRQ(ierr); 846 for (i=0; i<am; i++) { 847 ui[i] = ai[i+1] - a->diag[i]; /* ui: rowlengths - changes when !perm_identity */ 848 } 849 B = fact; 850 ierr = MatSeqSBAIJSetPreallocation(B,1,0,ui);CHKERRQ(ierr); 851 852 853 b = (Mat_SeqSBAIJ*)B->data; 854 uj = b->j; 855 for (i=0; i<am; i++) { 856 aj = a->j + a->diag[i]; 857 for (j=0; j<ui[i]; j++){ 858 *uj++ = *aj++; 859 } 860 b->ilen[i] = ui[i]; 861 } 862 ierr = PetscFree(ui);CHKERRQ(ierr); 863 B->factor = MAT_FACTOR_NONE; 864 ierr = MatAssemblyBegin(B,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 865 ierr = MatAssemblyEnd(B,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 866 B->factor = MAT_FACTOR_ICC; 867 868 B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqBAIJ_N_NaturalOrdering; 869 PetscFunctionReturn(0); 870 } 871 872 /* initialization */ 873 ierr = PetscMalloc((am+1)*sizeof(PetscInt),&ui);CHKERRQ(ierr); 874 ui[0] = 0; 875 ierr = PetscMalloc((2*am+1)*sizeof(PetscInt),&cols_lvl);CHKERRQ(ierr); 876 877 /* jl: linked list for storing indices of the pivot rows 878 il: il[i] points to the 1st nonzero entry of U(i,k:am-1) */ 879 ierr = PetscMalloc4(am,PetscInt*,&uj_ptr,am,PetscInt*,&uj_lvl_ptr,am,PetscInt,&il,am,PetscInt,&jl);CHKERRQ(ierr); 880 for (i=0; i<am; i++){ 881 jl[i] = am; il[i] = 0; 882 } 883 884 /* create and initialize a linked list for storing column indices of the active row k */ 885 nlnk = am + 1; 886 ierr = PetscIncompleteLLCreate(am,am,nlnk,lnk,lnk_lvl,lnkbt);CHKERRQ(ierr); 887 888 /* initial FreeSpace size is fill*(ai[am]+1) */ 889 ierr = PetscFreeSpaceGet((PetscInt)(fill*(ai[am]+1)),&free_space);CHKERRQ(ierr); 890 current_space = free_space; 891 ierr = PetscFreeSpaceGet((PetscInt)(fill*(ai[am]+1)),&free_space_lvl);CHKERRQ(ierr); 892 current_space_lvl = free_space_lvl; 893 894 for (k=0; k<am; k++){ /* for each active row k */ 895 /* initialize lnk by the column indices of row rip[k] of A */ 896 nzk = 0; 897 ncols = ai[rip[k]+1] - ai[rip[k]]; 898 ncols_upper = 0; 899 cols = cols_lvl + am; 900 for (j=0; j<ncols; j++){ 901 i = rip[*(aj + ai[rip[k]] + j)]; 902 if (i >= k){ /* only take upper triangular entry */ 903 cols[ncols_upper] = i; 904 cols_lvl[ncols_upper] = -1; /* initialize level for nonzero entries */ 905 ncols_upper++; 906 } 907 } 908 ierr = PetscIncompleteLLAdd(ncols_upper,cols,levels,cols_lvl,am,nlnk,lnk,lnk_lvl,lnkbt);CHKERRQ(ierr); 909 nzk += nlnk; 910 911 /* update lnk by computing fill-in for each pivot row to be merged in */ 912 prow = jl[k]; /* 1st pivot row */ 913 914 while (prow < k){ 915 nextprow = jl[prow]; 916 917 /* merge prow into k-th row */ 918 jmin = il[prow] + 1; /* index of the 2nd nzero entry in U(prow,k:am-1) */ 919 jmax = ui[prow+1]; 920 ncols = jmax-jmin; 921 i = jmin - ui[prow]; 922 cols = uj_ptr[prow] + i; /* points to the 2nd nzero entry in U(prow,k:am-1) */ 923 for (j=0; j<ncols; j++) cols_lvl[j] = *(uj_lvl_ptr[prow] + i + j); 924 ierr = PetscIncompleteLLAddSorted(ncols,cols,levels,cols_lvl,am,nlnk,lnk,lnk_lvl,lnkbt);CHKERRQ(ierr); 925 nzk += nlnk; 926 927 /* update il and jl for prow */ 928 if (jmin < jmax){ 929 il[prow] = jmin; 930 j = *cols; jl[prow] = jl[j]; jl[j] = prow; 931 } 932 prow = nextprow; 933 } 934 935 /* if free space is not available, make more free space */ 936 if (current_space->local_remaining<nzk) { 937 i = am - k + 1; /* num of unfactored rows */ 938 i = PetscMin(i*nzk, i*(i-1)); /* i*nzk, i*(i-1): estimated and max additional space needed */ 939 ierr = PetscFreeSpaceGet(i,¤t_space);CHKERRQ(ierr); 940 ierr = PetscFreeSpaceGet(i,¤t_space_lvl);CHKERRQ(ierr); 941 reallocs++; 942 } 943 944 /* copy data into free_space and free_space_lvl, then initialize lnk */ 945 ierr = PetscIncompleteLLClean(am,am,nzk,lnk,lnk_lvl,current_space->array,current_space_lvl->array,lnkbt);CHKERRQ(ierr); 946 947 /* add the k-th row into il and jl */ 948 if (nzk-1 > 0){ 949 i = current_space->array[1]; /* col value of the first nonzero element in U(k, k+1:am-1) */ 950 jl[k] = jl[i]; jl[i] = k; 951 il[k] = ui[k] + 1; 952 } 953 uj_ptr[k] = current_space->array; 954 uj_lvl_ptr[k] = current_space_lvl->array; 955 956 current_space->array += nzk; 957 current_space->local_used += nzk; 958 current_space->local_remaining -= nzk; 959 960 current_space_lvl->array += nzk; 961 current_space_lvl->local_used += nzk; 962 current_space_lvl->local_remaining -= nzk; 963 964 ui[k+1] = ui[k] + nzk; 965 } 966 967 #if defined(PETSC_USE_INFO) 968 if (ai[am] != 0) { 969 PetscReal af = ((PetscReal)(2*ui[am]-am))/((PetscReal)ai[am]); 970 ierr = PetscInfo3(A,"Reallocs %D Fill ratio:given %G needed %G\n",reallocs,fill,af);CHKERRQ(ierr); 971 ierr = PetscInfo1(A,"Run with -pc_factor_fill %G or use \n",af);CHKERRQ(ierr); 972 ierr = PetscInfo1(A,"PCFactorSetFill(pc,%G) for best performance.\n",af);CHKERRQ(ierr); 973 } else { 974 ierr = PetscInfo(A,"Empty matrix.\n");CHKERRQ(ierr); 975 } 976 #endif 977 978 ierr = ISRestoreIndices(perm,&rip);CHKERRQ(ierr); 979 ierr = PetscFree4(uj_ptr,uj_lvl_ptr,il,jl);CHKERRQ(ierr); 980 ierr = PetscFree(cols_lvl);CHKERRQ(ierr); 981 982 /* destroy list of free space and other temporary array(s) */ 983 ierr = PetscMalloc((ui[am]+1)*sizeof(PetscInt),&uj);CHKERRQ(ierr); 984 ierr = PetscFreeSpaceContiguous(&free_space,uj);CHKERRQ(ierr); 985 ierr = PetscIncompleteLLDestroy(lnk,lnkbt);CHKERRQ(ierr); 986 ierr = PetscFreeSpaceDestroy(free_space_lvl);CHKERRQ(ierr); 987 988 /* put together the new matrix in MATSEQSBAIJ format */ 989 B = fact; 990 ierr = MatSeqSBAIJSetPreallocation(B,1,MAT_SKIP_ALLOCATION,PETSC_NULL);CHKERRQ(ierr); 991 992 b = (Mat_SeqSBAIJ*)B->data; 993 b->singlemalloc = PETSC_FALSE; 994 b->free_a = PETSC_TRUE; 995 b->free_ij = PETSC_TRUE; 996 ierr = PetscMalloc((ui[am]+1)*sizeof(MatScalar),&b->a);CHKERRQ(ierr); 997 b->j = uj; 998 b->i = ui; 999 b->diag = 0; 1000 b->ilen = 0; 1001 b->imax = 0; 1002 b->row = perm; 1003 b->pivotinblocks = PETSC_FALSE; /* need to get from MatFactorInfo */ 1004 ierr = PetscObjectReference((PetscObject)perm);CHKERRQ(ierr); 1005 b->icol = perm; 1006 ierr = PetscObjectReference((PetscObject)perm);CHKERRQ(ierr); 1007 ierr = PetscMalloc((am+1)*sizeof(PetscScalar),&b->solve_work);CHKERRQ(ierr); 1008 ierr = PetscLogObjectMemory(B,(ui[am]-am)*(sizeof(PetscInt)+sizeof(MatScalar)));CHKERRQ(ierr); 1009 b->maxnz = b->nz = ui[am]; 1010 1011 B->info.factor_mallocs = reallocs; 1012 B->info.fill_ratio_given = fill; 1013 if (ai[am] != 0.) { 1014 B->info.fill_ratio_needed = ((PetscReal)ui[am])/((PetscReal)ai[am]); 1015 } else { 1016 B->info.fill_ratio_needed = 0.0; 1017 } 1018 if (perm_identity){ 1019 B->ops->solve = MatSolve_SeqSBAIJ_1_NaturalOrdering_inplace; 1020 B->ops->solvetranspose = MatSolve_SeqSBAIJ_1_NaturalOrdering_inplace; 1021 B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqBAIJ_N_NaturalOrdering; 1022 } else { 1023 (fact)->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqBAIJ_N; 1024 } 1025 PetscFunctionReturn(0); 1026 } 1027 1028 #undef __FUNCT__ 1029 #define __FUNCT__ "MatCholeskyFactorSymbolic_SeqBAIJ" 1030 PetscErrorCode MatCholeskyFactorSymbolic_SeqBAIJ(Mat fact,Mat A,IS perm,const MatFactorInfo *info) 1031 { 1032 Mat_SeqBAIJ *a = (Mat_SeqBAIJ*)A->data; 1033 Mat_SeqSBAIJ *b; 1034 Mat B; 1035 PetscErrorCode ierr; 1036 PetscTruth perm_identity; 1037 PetscReal fill = info->fill; 1038 const PetscInt *rip; 1039 PetscInt i,mbs=a->mbs,bs=A->rmap->bs,*ai=a->i,*aj=a->j,reallocs=0,prow; 1040 PetscInt *jl,jmin,jmax,nzk,*ui,k,j,*il,nextprow; 1041 PetscInt nlnk,*lnk,ncols,ncols_upper,*cols,*uj,**ui_ptr,*uj_ptr; 1042 PetscFreeSpaceList free_space=PETSC_NULL,current_space=PETSC_NULL; 1043 PetscBT lnkbt; 1044 1045 PetscFunctionBegin; 1046 if (bs > 1) { /* convert to seqsbaij */ 1047 if (!a->sbaijMat){ 1048 ierr = MatConvert(A,MATSEQSBAIJ,MAT_INITIAL_MATRIX,&a->sbaijMat);CHKERRQ(ierr); 1049 } 1050 (fact)->ops->choleskyfactorsymbolic = MatCholeskyFactorSymbolic_SeqSBAIJ; /* undue the change made in MatGetFactor_seqbaij_petsc */ 1051 ierr = MatCholeskyFactorSymbolic(fact,a->sbaijMat,perm,info);CHKERRQ(ierr); 1052 PetscFunctionReturn(0); 1053 } 1054 1055 /* check whether perm is the identity mapping */ 1056 ierr = ISIdentity(perm,&perm_identity);CHKERRQ(ierr); 1057 if (!perm_identity) SETERRQ(PETSC_ERR_SUP,"Matrix reordering is not supported"); 1058 ierr = ISGetIndices(perm,&rip);CHKERRQ(ierr); 1059 1060 /* initialization */ 1061 ierr = PetscMalloc((mbs+1)*sizeof(PetscInt),&ui);CHKERRQ(ierr); 1062 ui[0] = 0; 1063 1064 /* jl: linked list for storing indices of the pivot rows 1065 il: il[i] points to the 1st nonzero entry of U(i,k:mbs-1) */ 1066 ierr = PetscMalloc4(mbs,PetscInt*,&ui_ptr,mbs,PetscInt,&il,mbs,PetscInt,&jl,mbs,PetscInt,&cols);CHKERRQ(ierr); 1067 for (i=0; i<mbs; i++){ 1068 jl[i] = mbs; il[i] = 0; 1069 } 1070 1071 /* create and initialize a linked list for storing column indices of the active row k */ 1072 nlnk = mbs + 1; 1073 ierr = PetscLLCreate(mbs,mbs,nlnk,lnk,lnkbt);CHKERRQ(ierr); 1074 1075 /* initial FreeSpace size is fill*(ai[mbs]+1) */ 1076 ierr = PetscFreeSpaceGet((PetscInt)(fill*(ai[mbs]+1)),&free_space);CHKERRQ(ierr); 1077 current_space = free_space; 1078 1079 for (k=0; k<mbs; k++){ /* for each active row k */ 1080 /* initialize lnk by the column indices of row rip[k] of A */ 1081 nzk = 0; 1082 ncols = ai[rip[k]+1] - ai[rip[k]]; 1083 ncols_upper = 0; 1084 for (j=0; j<ncols; j++){ 1085 i = rip[*(aj + ai[rip[k]] + j)]; 1086 if (i >= k){ /* only take upper triangular entry */ 1087 cols[ncols_upper] = i; 1088 ncols_upper++; 1089 } 1090 } 1091 ierr = PetscLLAdd(ncols_upper,cols,mbs,nlnk,lnk,lnkbt);CHKERRQ(ierr); 1092 nzk += nlnk; 1093 1094 /* update lnk by computing fill-in for each pivot row to be merged in */ 1095 prow = jl[k]; /* 1st pivot row */ 1096 1097 while (prow < k){ 1098 nextprow = jl[prow]; 1099 /* merge prow into k-th row */ 1100 jmin = il[prow] + 1; /* index of the 2nd nzero entry in U(prow,k:mbs-1) */ 1101 jmax = ui[prow+1]; 1102 ncols = jmax-jmin; 1103 uj_ptr = ui_ptr[prow] + jmin - ui[prow]; /* points to the 2nd nzero entry in U(prow,k:mbs-1) */ 1104 ierr = PetscLLAddSorted(ncols,uj_ptr,mbs,nlnk,lnk,lnkbt);CHKERRQ(ierr); 1105 nzk += nlnk; 1106 1107 /* update il and jl for prow */ 1108 if (jmin < jmax){ 1109 il[prow] = jmin; 1110 j = *uj_ptr; jl[prow] = jl[j]; jl[j] = prow; 1111 } 1112 prow = nextprow; 1113 } 1114 1115 /* if free space is not available, make more free space */ 1116 if (current_space->local_remaining<nzk) { 1117 i = mbs - k + 1; /* num of unfactored rows */ 1118 i = PetscMin(i*nzk, i*(i-1)); /* i*nzk, i*(i-1): estimated and max additional space needed */ 1119 ierr = PetscFreeSpaceGet(i,¤t_space);CHKERRQ(ierr); 1120 reallocs++; 1121 } 1122 1123 /* copy data into free space, then initialize lnk */ 1124 ierr = PetscLLClean(mbs,mbs,nzk,lnk,current_space->array,lnkbt);CHKERRQ(ierr); 1125 1126 /* add the k-th row into il and jl */ 1127 if (nzk-1 > 0){ 1128 i = current_space->array[1]; /* col value of the first nonzero element in U(k, k+1:mbs-1) */ 1129 jl[k] = jl[i]; jl[i] = k; 1130 il[k] = ui[k] + 1; 1131 } 1132 ui_ptr[k] = current_space->array; 1133 current_space->array += nzk; 1134 current_space->local_used += nzk; 1135 current_space->local_remaining -= nzk; 1136 1137 ui[k+1] = ui[k] + nzk; 1138 } 1139 1140 #if defined(PETSC_USE_INFO) 1141 if (ai[mbs] != 0.) { 1142 PetscReal af = ((PetscReal)ui[mbs])/((PetscReal)ai[mbs]); 1143 ierr = PetscInfo3(A,"Reallocs %D Fill ratio:given %G needed %G\n",reallocs,fill,af);CHKERRQ(ierr); 1144 ierr = PetscInfo1(A,"Run with -pc_factor_fill %G or use \n",af);CHKERRQ(ierr); 1145 ierr = PetscInfo1(A,"PCFactorSetFill(pc,%G) for best performance.\n",af);CHKERRQ(ierr); 1146 } else { 1147 ierr = PetscInfo(A,"Empty matrix.\n");CHKERRQ(ierr); 1148 } 1149 #endif 1150 1151 ierr = ISRestoreIndices(perm,&rip);CHKERRQ(ierr); 1152 ierr = PetscFree4(ui_ptr,il,jl,cols);CHKERRQ(ierr); 1153 1154 /* destroy list of free space and other temporary array(s) */ 1155 ierr = PetscMalloc((ui[mbs]+1)*sizeof(PetscInt),&uj);CHKERRQ(ierr); 1156 ierr = PetscFreeSpaceContiguous(&free_space,uj);CHKERRQ(ierr); 1157 ierr = PetscLLDestroy(lnk,lnkbt);CHKERRQ(ierr); 1158 1159 /* put together the new matrix in MATSEQSBAIJ format */ 1160 B = fact; 1161 ierr = MatSeqSBAIJSetPreallocation(B,bs,MAT_SKIP_ALLOCATION,PETSC_NULL);CHKERRQ(ierr); 1162 1163 b = (Mat_SeqSBAIJ*)B->data; 1164 b->singlemalloc = PETSC_FALSE; 1165 b->free_a = PETSC_TRUE; 1166 b->free_ij = PETSC_TRUE; 1167 ierr = PetscMalloc((ui[mbs]+1)*sizeof(MatScalar),&b->a);CHKERRQ(ierr); 1168 b->j = uj; 1169 b->i = ui; 1170 b->diag = 0; 1171 b->ilen = 0; 1172 b->imax = 0; 1173 b->row = perm; 1174 b->pivotinblocks = PETSC_FALSE; /* need to get from MatFactorInfo */ 1175 ierr = PetscObjectReference((PetscObject)perm);CHKERRQ(ierr); 1176 b->icol = perm; 1177 ierr = PetscObjectReference((PetscObject)perm);CHKERRQ(ierr); 1178 ierr = PetscMalloc((mbs+1)*sizeof(PetscScalar),&b->solve_work);CHKERRQ(ierr); 1179 ierr = PetscLogObjectMemory(B,(ui[mbs]-mbs)*(sizeof(PetscInt)+sizeof(MatScalar)));CHKERRQ(ierr); 1180 b->maxnz = b->nz = ui[mbs]; 1181 1182 B->info.factor_mallocs = reallocs; 1183 B->info.fill_ratio_given = fill; 1184 if (ai[mbs] != 0.) { 1185 B->info.fill_ratio_needed = ((PetscReal)ui[mbs])/((PetscReal)ai[mbs]); 1186 } else { 1187 B->info.fill_ratio_needed = 0.0; 1188 } 1189 if (perm_identity){ 1190 B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqBAIJ_N_NaturalOrdering; 1191 } else { 1192 B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqBAIJ_N; 1193 } 1194 PetscFunctionReturn(0); 1195 } 1196 1197 #undef __FUNCT__ 1198 #define __FUNCT__ "MatSolve_SeqBAIJ_N_NaturalOrdering" 1199 PetscErrorCode MatSolve_SeqBAIJ_N_NaturalOrdering(Mat A,Vec bb,Vec xx) 1200 { 1201 Mat_SeqBAIJ *a=(Mat_SeqBAIJ *)A->data; 1202 PetscErrorCode ierr; 1203 const PetscInt *ai=a->i,*aj=a->j,*adiag=a->diag,*vi; 1204 PetscInt i,k,n=a->mbs; 1205 PetscInt nz,bs=A->rmap->bs,bs2=a->bs2; 1206 MatScalar *aa=a->a,*v; 1207 PetscScalar *x,*b,*s,*t,*ls; 1208 1209 PetscFunctionBegin; 1210 ierr = VecGetArray(bb,&b);CHKERRQ(ierr); 1211 ierr = VecGetArray(xx,&x);CHKERRQ(ierr); 1212 t = a->solve_work; 1213 1214 /* forward solve the lower triangular */ 1215 ierr = PetscMemcpy(t,b,bs*sizeof(PetscScalar));CHKERRQ(ierr); /* copy 1st block of b to t */ 1216 1217 for (i=1; i<n; i++) { 1218 v = aa + bs2*ai[i]; 1219 vi = aj + ai[i]; 1220 nz = ai[i+1] - ai[i]; 1221 s = t + bs*i; 1222 ierr = PetscMemcpy(s,b+bs*i,bs*sizeof(PetscScalar));CHKERRQ(ierr); /* copy i_th block of b to t */ 1223 for(k=0;k<nz;k++){ 1224 Kernel_v_gets_v_minus_A_times_w(bs,s,v,t+bs*vi[k]); 1225 v += bs2; 1226 } 1227 } 1228 1229 /* backward solve the upper triangular */ 1230 ls = a->solve_work + A->cmap->n; 1231 for (i=n-1; i>=0; i--){ 1232 v = aa + bs2*(adiag[i+1]+1); 1233 vi = aj + adiag[i+1]+1; 1234 nz = adiag[i] - adiag[i+1]-1; 1235 ierr = PetscMemcpy(ls,t+i*bs,bs*sizeof(PetscScalar));CHKERRQ(ierr); 1236 for(k=0;k<nz;k++){ 1237 Kernel_v_gets_v_minus_A_times_w(bs,ls,v,t+bs*vi[k]); 1238 v += bs2; 1239 } 1240 Kernel_w_gets_A_times_v(bs,ls,aa+bs2*adiag[i],t+i*bs); /* *inv(diagonal[i]) */ 1241 ierr = PetscMemcpy(x+i*bs,t+i*bs,bs*sizeof(PetscScalar));CHKERRQ(ierr); 1242 } 1243 1244 ierr = VecRestoreArray(bb,&b);CHKERRQ(ierr); 1245 ierr = VecRestoreArray(xx,&x);CHKERRQ(ierr); 1246 ierr = PetscLogFlops(2.0*(a->bs2)*(a->nz) - A->rmap->bs*A->cmap->n);CHKERRQ(ierr); 1247 PetscFunctionReturn(0); 1248 } 1249 1250 #undef __FUNCT__ 1251 #define __FUNCT__ "MatSolve_SeqBAIJ_N" 1252 PetscErrorCode MatSolve_SeqBAIJ_N(Mat A,Vec bb,Vec xx) 1253 { 1254 Mat_SeqBAIJ *a=(Mat_SeqBAIJ *)A->data; 1255 IS iscol=a->col,isrow=a->row; 1256 PetscErrorCode ierr; 1257 const PetscInt *r,*c,*rout,*cout,*ai=a->i,*aj=a->j,*adiag=a->diag,*vi; 1258 PetscInt i,m,n=a->mbs; 1259 PetscInt nz,bs=A->rmap->bs,bs2=a->bs2; 1260 MatScalar *aa=a->a,*v; 1261 PetscScalar *x,*b,*s,*t,*ls; 1262 1263 PetscFunctionBegin; 1264 ierr = VecGetArray(bb,&b);CHKERRQ(ierr); 1265 ierr = VecGetArray(xx,&x);CHKERRQ(ierr); 1266 t = a->solve_work; 1267 1268 ierr = ISGetIndices(isrow,&rout);CHKERRQ(ierr); r = rout; 1269 ierr = ISGetIndices(iscol,&cout);CHKERRQ(ierr); c = cout; 1270 1271 /* forward solve the lower triangular */ 1272 ierr = PetscMemcpy(t,b+bs*r[0],bs*sizeof(PetscScalar));CHKERRQ(ierr); 1273 for (i=1; i<n; i++) { 1274 v = aa + bs2*ai[i]; 1275 vi = aj + ai[i]; 1276 nz = ai[i+1] - ai[i]; 1277 s = t + bs*i; 1278 ierr = PetscMemcpy(s,b+bs*r[i],bs*sizeof(PetscScalar));CHKERRQ(ierr); 1279 for(m=0;m<nz;m++){ 1280 Kernel_v_gets_v_minus_A_times_w(bs,s,v,t+bs*vi[m]); 1281 v += bs2; 1282 } 1283 } 1284 1285 /* backward solve the upper triangular */ 1286 ls = a->solve_work + A->cmap->n; 1287 for (i=n-1; i>=0; i--){ 1288 v = aa + bs2*(adiag[i+1]+1); 1289 vi = aj + adiag[i+1]+1; 1290 nz = adiag[i] - adiag[i+1] - 1; 1291 ierr = PetscMemcpy(ls,t+i*bs,bs*sizeof(PetscScalar));CHKERRQ(ierr); 1292 for(m=0;m<nz;m++){ 1293 Kernel_v_gets_v_minus_A_times_w(bs,ls,v,t+bs*vi[m]); 1294 v += bs2; 1295 } 1296 Kernel_w_gets_A_times_v(bs,ls,v,t+i*bs); /* *inv(diagonal[i]) */ 1297 ierr = PetscMemcpy(x + bs*c[i],t+i*bs,bs*sizeof(PetscScalar));CHKERRQ(ierr); 1298 } 1299 ierr = ISRestoreIndices(isrow,&rout);CHKERRQ(ierr); 1300 ierr = ISRestoreIndices(iscol,&cout);CHKERRQ(ierr); 1301 ierr = VecRestoreArray(bb,&b);CHKERRQ(ierr); 1302 ierr = VecRestoreArray(xx,&x);CHKERRQ(ierr); 1303 ierr = PetscLogFlops(2.0*(a->bs2)*(a->nz) - A->rmap->bs*A->cmap->n);CHKERRQ(ierr); 1304 PetscFunctionReturn(0); 1305 } 1306 1307 #undef __FUNCT__ 1308 #define __FUNCT__ "BlockAbs_privat" 1309 PetscErrorCode BlockAbs_private(PetscInt nbs,PetscInt bs2,PetscScalar *blockarray,PetscReal *absarray) 1310 { 1311 PetscErrorCode ierr; 1312 PetscInt i,j; 1313 PetscFunctionBegin; 1314 ierr = PetscMemzero(absarray,(nbs+1)*sizeof(PetscReal));CHKERRQ(ierr); 1315 for (i=0; i<nbs; i++){ 1316 for (j=0; j<bs2; j++){ 1317 if (absarray[i] < PetscAbsScalar(blockarray[i*nbs+j])) absarray[i] = PetscAbsScalar(blockarray[i*nbs+j]); 1318 } 1319 } 1320 PetscFunctionReturn(0); 1321 } 1322 1323 #undef __FUNCT__ 1324 #define __FUNCT__ "MatILUDTFactor_SeqBAIJ" 1325 /* 1326 This needs to be renamed and called by the regular MatILUFactor_SeqBAIJ when drop tolerance is used 1327 */ 1328 PetscErrorCode MatILUDTFactor_SeqBAIJ(Mat A,IS isrow,IS iscol,const MatFactorInfo *info,Mat *fact) 1329 { 1330 Mat B = *fact; 1331 Mat_SeqBAIJ *a=(Mat_SeqBAIJ*)A->data,*b; 1332 IS isicol; 1333 PetscErrorCode ierr; 1334 const PetscInt *r,*ic; 1335 PetscInt i,mbs=a->mbs,bs=A->rmap->bs,bs2=a->bs2,*ai=a->i,*aj=a->j,*ajtmp,*adiag; 1336 PetscInt *bi,*bj,*bdiag; 1337 1338 PetscInt row,nzi,nzi_bl,nzi_bu,*im,dtcount,nzi_al,nzi_au; 1339 PetscInt nlnk,*lnk; 1340 PetscBT lnkbt; 1341 PetscTruth row_identity,icol_identity,both_identity; 1342 MatScalar *aatmp,*pv,*batmp,*ba,*rtmp,*pc,*multiplier,*vtmp; 1343 const PetscInt *ics; 1344 PetscInt j,nz,*pj,*bjtmp,k,ncut,*jtmp; 1345 1346 PetscReal dt=info->dt; /* shift=info->shiftinblocks; */ 1347 PetscInt nnz_max; 1348 PetscTruth missing; 1349 PetscReal *vtmp_abs; 1350 MatScalar *v_work; 1351 PetscInt *v_pivots; 1352 1353 PetscFunctionBegin; 1354 /* ------- symbolic factorization, can be reused ---------*/ 1355 ierr = MatMissingDiagonal(A,&missing,&i);CHKERRQ(ierr); 1356 if (missing) SETERRQ1(PETSC_ERR_ARG_WRONGSTATE,"Matrix is missing diagonal entry %D",i); 1357 adiag=a->diag; 1358 1359 ierr = ISInvertPermutation(iscol,PETSC_DECIDE,&isicol);CHKERRQ(ierr); 1360 1361 /* bdiag is location of diagonal in factor */ 1362 ierr = PetscMalloc((mbs+1)*sizeof(PetscInt),&bdiag);CHKERRQ(ierr); 1363 1364 /* allocate row pointers bi */ 1365 ierr = PetscMalloc((2*mbs+2)*sizeof(PetscInt),&bi);CHKERRQ(ierr); 1366 1367 /* allocate bj and ba; max num of nonzero entries is (ai[n]+2*n*dtcount+2) */ 1368 dtcount = (PetscInt)info->dtcount; 1369 if (dtcount > mbs-1) dtcount = mbs-1; 1370 nnz_max = ai[mbs]+2*mbs*dtcount +2; 1371 /* printf("MatILUDTFactor_SeqBAIJ, bs %d, ai[mbs] %d, nnz_max %d, dtcount %d\n",bs,ai[mbs],nnz_max,dtcount); */ 1372 ierr = PetscMalloc(nnz_max*sizeof(PetscInt),&bj);CHKERRQ(ierr); 1373 nnz_max = nnz_max*bs2; 1374 ierr = PetscMalloc(nnz_max*sizeof(MatScalar),&ba);CHKERRQ(ierr); 1375 1376 /* put together the new matrix */ 1377 ierr = MatSeqBAIJSetPreallocation_SeqBAIJ(B,bs,MAT_SKIP_ALLOCATION,PETSC_NULL);CHKERRQ(ierr); 1378 ierr = PetscLogObjectParent(B,isicol);CHKERRQ(ierr); 1379 b = (Mat_SeqBAIJ*)(B)->data; 1380 b->free_a = PETSC_TRUE; 1381 b->free_ij = PETSC_TRUE; 1382 b->singlemalloc = PETSC_FALSE; 1383 b->a = ba; 1384 b->j = bj; 1385 b->i = bi; 1386 b->diag = bdiag; 1387 b->ilen = 0; 1388 b->imax = 0; 1389 b->row = isrow; 1390 b->col = iscol; 1391 ierr = PetscObjectReference((PetscObject)isrow);CHKERRQ(ierr); 1392 ierr = PetscObjectReference((PetscObject)iscol);CHKERRQ(ierr); 1393 b->icol = isicol; 1394 ierr = PetscMalloc((bs*(mbs+1))*sizeof(PetscScalar),&b->solve_work);CHKERRQ(ierr); 1395 1396 ierr = PetscLogObjectMemory(B,nnz_max*(sizeof(PetscInt)+sizeof(MatScalar)));CHKERRQ(ierr); 1397 b->maxnz = nnz_max/bs2; 1398 1399 (B)->factor = MAT_FACTOR_ILUDT; 1400 (B)->info.factor_mallocs = 0; 1401 (B)->info.fill_ratio_given = ((PetscReal)nnz_max)/((PetscReal)(ai[mbs]*bs2)); 1402 CHKMEMQ; 1403 /* ------- end of symbolic factorization ---------*/ 1404 ierr = ISGetIndices(isrow,&r);CHKERRQ(ierr); 1405 ierr = ISGetIndices(isicol,&ic);CHKERRQ(ierr); 1406 ics = ic; 1407 1408 /* linked list for storing column indices of the active row */ 1409 nlnk = mbs + 1; 1410 ierr = PetscLLCreate(mbs,mbs,nlnk,lnk,lnkbt);CHKERRQ(ierr); 1411 1412 /* im: used by PetscLLAddSortedLU(); jtmp: working array for column indices of active row */ 1413 ierr = PetscMalloc2(mbs,PetscInt,&im,mbs,PetscInt,&jtmp);CHKERRQ(ierr); 1414 /* rtmp, vtmp: working arrays for sparse and contiguous row entries of active row */ 1415 ierr = PetscMalloc2(mbs*bs2,MatScalar,&rtmp,mbs*bs2,MatScalar,&vtmp);CHKERRQ(ierr); 1416 ierr = PetscMalloc((mbs+1)*sizeof(PetscReal),&vtmp_abs);CHKERRQ(ierr); 1417 ierr = PetscMalloc3(bs,MatScalar,&v_work,bs2,MatScalar,&multiplier,bs,PetscInt,&v_pivots);CHKERRQ(ierr); 1418 1419 bi[0] = 0; 1420 bdiag[0] = (nnz_max/bs2)-1; /* location of diagonal in factor B */ 1421 bi[2*mbs+1] = bdiag[0]+1; /* endof bj and ba array */ 1422 for (i=0; i<mbs; i++) { 1423 /* copy initial fill into linked list */ 1424 nzi = 0; /* nonzeros for active row i */ 1425 nzi = ai[r[i]+1] - ai[r[i]]; 1426 if (!nzi) SETERRQ2(PETSC_ERR_MAT_LU_ZRPVT,"Empty row in matrix: row in original ordering %D in permuted ordering %D",r[i],i); 1427 nzi_al = adiag[r[i]] - ai[r[i]]; 1428 nzi_au = ai[r[i]+1] - adiag[r[i]] -1; 1429 /* printf("row %d, nzi_al/au %d %d\n",i,nzi_al,nzi_au); */ 1430 1431 /* load in initial unfactored row */ 1432 ajtmp = aj + ai[r[i]]; 1433 ierr = PetscLLAddPerm(nzi,ajtmp,ic,mbs,nlnk,lnk,lnkbt);CHKERRQ(ierr); 1434 ierr = PetscMemzero(rtmp,mbs*bs2*sizeof(PetscScalar));CHKERRQ(ierr); 1435 aatmp = a->a + bs2*ai[r[i]]; 1436 for (j=0; j<nzi; j++) { 1437 ierr = PetscMemcpy(rtmp+bs2*ic[ajtmp[j]],aatmp+bs2*j,bs2*sizeof(MatScalar));CHKERRQ(ierr); 1438 } 1439 1440 /* add pivot rows into linked list */ 1441 row = lnk[mbs]; 1442 while (row < i) { 1443 nzi_bl = bi[row+1] - bi[row] + 1; 1444 bjtmp = bj + bdiag[row+1]+1; /* points to 1st column next to the diagonal in U */ 1445 ierr = PetscLLAddSortedLU(bjtmp,row,nlnk,lnk,lnkbt,i,nzi_bl,im);CHKERRQ(ierr); 1446 nzi += nlnk; 1447 row = lnk[row]; 1448 } 1449 1450 /* copy data from lnk into jtmp, then initialize lnk */ 1451 ierr = PetscLLClean(mbs,mbs,nzi,lnk,jtmp,lnkbt);CHKERRQ(ierr); 1452 1453 /* numerical factorization */ 1454 bjtmp = jtmp; 1455 row = *bjtmp++; /* 1st pivot row */ 1456 1457 while (row < i) { 1458 pc = rtmp + bs2*row; 1459 pv = ba + bs2*bdiag[row]; /* inv(diag) of the pivot row */ 1460 Kernel_A_gets_A_times_B(bs,pc,pv,multiplier); /* pc= multiplier = pc*inv(diag[row]) */ 1461 ierr = BlockAbs_private(1,bs2,pc,vtmp_abs);CHKERRQ(ierr); 1462 if (vtmp_abs[0] > dt){ /* apply tolerance dropping rule */ 1463 pj = bj + bdiag[row+1] + 1; /* point to 1st entry of U(row,:) */ 1464 pv = ba + bs2*(bdiag[row+1] + 1); 1465 nz = bdiag[row] - bdiag[row+1] - 1; /* num of entries in U(row,:), excluding diagonal */ 1466 for (j=0; j<nz; j++){ 1467 Kernel_A_gets_A_minus_B_times_C(bs,rtmp+bs2*pj[j],pc,pv+bs2*j); 1468 } 1469 /* ierr = PetscLogFlops(bslog*(nz+1.0)-bs);CHKERRQ(ierr); */ 1470 } 1471 row = *bjtmp++; 1472 } 1473 1474 /* copy sparse rtmp into contiguous vtmp; separate L and U part */ 1475 nzi_bl = 0; j = 0; 1476 while (jtmp[j] < i){ /* L-part. Note: jtmp is sorted */ 1477 ierr = PetscMemcpy(vtmp+bs2*j,rtmp+bs2*jtmp[j],bs2*sizeof(MatScalar));CHKERRQ(ierr); 1478 nzi_bl++; j++; 1479 } 1480 nzi_bu = nzi - nzi_bl -1; 1481 /* printf("nzi %d, nzi_bl %d, nzi_bu %d\n",nzi,nzi_bl,nzi_bu); */ 1482 1483 while (j < nzi){ /* U-part */ 1484 ierr = PetscMemcpy(vtmp+bs2*j,rtmp+bs2*jtmp[j],bs2*sizeof(MatScalar));CHKERRQ(ierr); 1485 /* 1486 printf(" col %d: ",jtmp[j]); 1487 for (j1=0; j1<bs2; j1++) printf(" %g",*(vtmp+bs2*j+j1)); 1488 printf(" \n"); 1489 */ 1490 j++; 1491 } 1492 1493 ierr = BlockAbs_private(nzi,bs2,vtmp,vtmp_abs);CHKERRQ(ierr); 1494 /* 1495 printf(" row %d, nzi %d, vtmp_abs\n",i,nzi); 1496 for (j1=0; j1<nzi; j1++) printf(" (%d %g),",jtmp[j1],vtmp_abs[j1]); 1497 printf(" \n"); 1498 */ 1499 bjtmp = bj + bi[i]; 1500 batmp = ba + bs2*bi[i]; 1501 /* apply level dropping rule to L part */ 1502 ncut = nzi_al + dtcount; 1503 if (ncut < nzi_bl){ 1504 ierr = PetscSortSplitReal(ncut,nzi_bl,vtmp_abs,jtmp);CHKERRQ(ierr); 1505 ierr = PetscSortIntWithScalarArray(ncut,jtmp,vtmp);CHKERRQ(ierr); 1506 } else { 1507 ncut = nzi_bl; 1508 } 1509 for (j=0; j<ncut; j++){ 1510 bjtmp[j] = jtmp[j]; 1511 ierr = PetscMemcpy(batmp+bs2*j,rtmp+bs2*bjtmp[j],bs2*sizeof(MatScalar));CHKERRQ(ierr); 1512 /* 1513 printf(" col %d: ",bjtmp[j]); 1514 for (j1=0; j1<bs2; j1++) printf(" %g,",*(batmp+bs2*j+j1)); 1515 printf("\n"); 1516 */ 1517 } 1518 bi[i+1] = bi[i] + ncut; 1519 nzi = ncut + 1; 1520 1521 /* apply level dropping rule to U part */ 1522 ncut = nzi_au + dtcount; 1523 if (ncut < nzi_bu){ 1524 ierr = PetscSortSplitReal(ncut,nzi_bu,vtmp_abs+nzi_bl+1,jtmp+nzi_bl+1);CHKERRQ(ierr); 1525 ierr = PetscSortIntWithScalarArray(ncut,jtmp+nzi_bl+1,vtmp+nzi_bl+1);CHKERRQ(ierr); 1526 } else { 1527 ncut = nzi_bu; 1528 } 1529 nzi += ncut; 1530 1531 /* mark bdiagonal */ 1532 bdiag[i+1] = bdiag[i] - (ncut + 1); 1533 bi[2*mbs - i] = bi[2*mbs - i +1] - (ncut + 1); 1534 1535 bjtmp = bj + bdiag[i]; 1536 batmp = ba + bs2*bdiag[i]; 1537 ierr = PetscMemcpy(batmp,rtmp+bs2*i,bs2*sizeof(MatScalar));CHKERRQ(ierr); 1538 *bjtmp = i; 1539 /* 1540 printf(" diag %d: ",*bjtmp); 1541 for (j=0; j<bs2; j++){ 1542 printf(" %g,",batmp[j]); 1543 } 1544 printf("\n"); 1545 */ 1546 bjtmp = bj + bdiag[i+1]+1; 1547 batmp = ba + (bdiag[i+1]+1)*bs2; 1548 1549 for (k=0; k<ncut; k++){ 1550 bjtmp[k] = jtmp[nzi_bl+1+k]; 1551 ierr = PetscMemcpy(batmp+bs2*k,rtmp+bs2*bjtmp[k],bs2*sizeof(MatScalar));CHKERRQ(ierr); 1552 /* 1553 printf(" col %d:",bjtmp[k]); 1554 for (j1=0; j1<bs2; j1++) printf(" %g,",*(batmp+bs2*k+j1)); 1555 printf("\n"); 1556 */ 1557 } 1558 1559 im[i] = nzi; /* used by PetscLLAddSortedLU() */ 1560 1561 /* invert diagonal block for simplier triangular solves - add shift??? */ 1562 batmp = ba + bs2*bdiag[i]; 1563 ierr = Kernel_A_gets_inverse_A(bs,batmp,v_pivots,v_work);CHKERRQ(ierr); 1564 } /* for (i=0; i<mbs; i++) */ 1565 ierr = PetscFree3(v_work,multiplier,v_pivots);CHKERRQ(ierr); 1566 1567 /* printf("end of L %d, beginning of U %d\n",bi[mbs],bdiag[mbs]); */ 1568 if (bi[mbs] >= bdiag[mbs]) SETERRQ2(PETSC_ERR_ARG_SIZ,"end of L array %d cannot >= the beginning of U array %d",bi[mbs],bdiag[mbs]); 1569 1570 ierr = ISRestoreIndices(isrow,&r);CHKERRQ(ierr); 1571 ierr = ISRestoreIndices(isicol,&ic);CHKERRQ(ierr); 1572 1573 ierr = PetscLLDestroy(lnk,lnkbt);CHKERRQ(ierr); 1574 1575 ierr = PetscFree2(im,jtmp);CHKERRQ(ierr); 1576 ierr = PetscFree2(rtmp,vtmp);CHKERRQ(ierr); 1577 1578 ierr = PetscLogFlops(bs2*B->cmap->n);CHKERRQ(ierr); 1579 b->maxnz = b->nz = bi[mbs] + bdiag[0] - bdiag[mbs]; 1580 1581 ierr = ISIdentity(isrow,&row_identity);CHKERRQ(ierr); 1582 ierr = ISIdentity(isicol,&icol_identity);CHKERRQ(ierr); 1583 both_identity = (PetscTruth) (row_identity && icol_identity); 1584 if (row_identity && icol_identity) { 1585 B->ops->solve = MatSolve_SeqBAIJ_N_NaturalOrdering; 1586 } else { 1587 B->ops->solve = MatSolve_SeqBAIJ_N; 1588 } 1589 1590 B->ops->solveadd = 0; 1591 B->ops->solvetranspose = 0; 1592 B->ops->solvetransposeadd = 0; 1593 B->ops->matsolve = 0; 1594 B->assembled = PETSC_TRUE; 1595 B->preallocated = PETSC_TRUE; 1596 PetscFunctionReturn(0); 1597 } 1598