1 #ifndef lint 2 static char vcid[] = "$Id: baijfact.c,v 1.28 1996/08/08 14:43:31 bsmith Exp curfman $"; 3 #endif 4 /* 5 Factorization code for BAIJ format. 6 */ 7 8 #include "src/mat/impls/baij/seq/baij.h" 9 #include "src/vec/vecimpl.h" 10 #include "src/inline/ilu.h" 11 12 13 /* 14 The symbolic factorization code is identical to that for AIJ format, 15 except for very small changes since this is now a SeqBAIJ datastructure. 16 NOT good code reuse. 17 */ 18 int MatLUFactorSymbolic_SeqBAIJ(Mat A,IS isrow,IS iscol,double f,Mat *B) 19 { 20 Mat_SeqBAIJ *a = (Mat_SeqBAIJ *) A->data, *b; 21 IS isicol; 22 int *r,*ic, ierr, i, n = a->mbs, *ai = a->i, *aj = a->j; 23 int *ainew,*ajnew, jmax,*fill, *ajtmp, nz, bs = a->bs, bs2=a->bs2; 24 int *idnew, idx, row,m,fm, nnz, nzi,len, realloc = 0,nzbd,*im; 25 26 PetscValidHeaderSpecific(isrow,IS_COOKIE); 27 PetscValidHeaderSpecific(iscol,IS_COOKIE); 28 ierr = ISInvertPermutation(iscol,&isicol); CHKERRQ(ierr); 29 ISGetIndices(isrow,&r); ISGetIndices(isicol,&ic); 30 31 /* get new row pointers */ 32 ainew = (int *) PetscMalloc( (n+1)*sizeof(int) ); CHKPTRQ(ainew); 33 ainew[0] = 0; 34 /* don't know how many column pointers are needed so estimate */ 35 jmax = (int) (f*ai[n] + 1); 36 ajnew = (int *) PetscMalloc( (jmax)*sizeof(int) ); CHKPTRQ(ajnew); 37 /* fill is a linked list of nonzeros in active row */ 38 fill = (int *) PetscMalloc( (2*n+1)*sizeof(int)); CHKPTRQ(fill); 39 im = fill + n + 1; 40 /* idnew is location of diagonal in factor */ 41 idnew = (int *) PetscMalloc( (n+1)*sizeof(int)); CHKPTRQ(idnew); 42 idnew[0] = 0; 43 44 for ( i=0; i<n; i++ ) { 45 /* first copy previous fill into linked list */ 46 nnz = nz = ai[r[i]+1] - ai[r[i]]; 47 ajtmp = aj + ai[r[i]]; 48 fill[n] = n; 49 while (nz--) { 50 fm = n; 51 idx = ic[*ajtmp++]; 52 do { 53 m = fm; 54 fm = fill[m]; 55 } while (fm < idx); 56 fill[m] = idx; 57 fill[idx] = fm; 58 } 59 row = fill[n]; 60 while ( row < i ) { 61 ajtmp = ajnew + idnew[row] + 1; 62 nzbd = 1 + idnew[row] - ainew[row]; 63 nz = im[row] - nzbd; 64 fm = row; 65 while (nz-- > 0) { 66 idx = *ajtmp++; 67 nzbd++; 68 if (idx == i) im[row] = nzbd; 69 do { 70 m = fm; 71 fm = fill[m]; 72 } while (fm < idx); 73 if (fm != idx) { 74 fill[m] = idx; 75 fill[idx] = fm; 76 fm = idx; 77 nnz++; 78 } 79 } 80 row = fill[row]; 81 } 82 /* copy new filled row into permanent storage */ 83 ainew[i+1] = ainew[i] + nnz; 84 if (ainew[i+1] > jmax) { 85 /* allocate a longer ajnew */ 86 int maxadd; 87 maxadd = (int) ((f*(ai[n]+1)*(n-i+5))/n); 88 if (maxadd < nnz) maxadd = (n-i)*(nnz+1); 89 jmax += maxadd; 90 ajtmp = (int *) PetscMalloc( jmax*sizeof(int) );CHKPTRQ(ajtmp); 91 PetscMemcpy(ajtmp,ajnew,ainew[i]*sizeof(int)); 92 PetscFree(ajnew); 93 ajnew = ajtmp; 94 realloc++; /* count how many times we realloc */ 95 } 96 ajtmp = ajnew + ainew[i]; 97 fm = fill[n]; 98 nzi = 0; 99 im[i] = nnz; 100 while (nnz--) { 101 if (fm < i) nzi++; 102 *ajtmp++ = fm; 103 fm = fill[fm]; 104 } 105 idnew[i] = ainew[i] + nzi; 106 } 107 108 PLogInfo(A, 109 "Info:MatLUFactorSymbolic_SeqBAIJ:Reallocs %d Fill ratio:given %g needed %g\n", 110 realloc,f,((double)ainew[n])/((double)ai[i])); 111 112 ierr = ISRestoreIndices(isrow,&r); CHKERRQ(ierr); 113 ierr = ISRestoreIndices(isicol,&ic); CHKERRQ(ierr); 114 115 PetscFree(fill); 116 117 /* put together the new matrix */ 118 ierr = MatCreateSeqBAIJ(A->comm,bs,bs*n,bs*n,0,PETSC_NULL,B); CHKERRQ(ierr); 119 PLogObjectParent(*B,isicol); 120 ierr = ISDestroy(isicol); CHKERRQ(ierr); 121 b = (Mat_SeqBAIJ *) (*B)->data; 122 PetscFree(b->imax); 123 b->singlemalloc = 0; 124 len = ainew[n]*sizeof(Scalar); 125 /* the next line frees the default space generated by the Create() */ 126 PetscFree(b->a); PetscFree(b->ilen); 127 b->a = (Scalar *) PetscMalloc( len*bs2 ); CHKPTRQ(b->a); 128 b->j = ajnew; 129 b->i = ainew; 130 b->diag = idnew; 131 b->ilen = 0; 132 b->imax = 0; 133 b->row = isrow; 134 b->col = iscol; 135 b->solve_work = (Scalar *) PetscMalloc( (bs*n+bs)*sizeof(Scalar)); 136 CHKPTRQ(b->solve_work); 137 /* In b structure: Free imax, ilen, old a, old j. 138 Allocate idnew, solve_work, new a, new j */ 139 PLogObjectMemory(*B,(ainew[n]-n)*(sizeof(int)+sizeof(Scalar))); 140 b->maxnz = b->nz = ainew[n]; 141 142 (*B)->info.factor_mallocs = realloc; 143 (*B)->info.fill_ratio_given = f; 144 (*B)->info.fill_ratio_needed = ((double)ainew[n])/((double)ai[i]); 145 146 return 0; 147 } 148 149 /* ----------------------------------------------------------- */ 150 int MatLUFactorNumeric_SeqBAIJ_N(Mat A,Mat *B) 151 { 152 Mat C = *B; 153 Mat_SeqBAIJ *a = (Mat_SeqBAIJ *) A->data,*b = (Mat_SeqBAIJ *)C->data; 154 IS iscol = b->col, isrow = b->row, isicol; 155 int *r,*ic, ierr, i, j, n = a->mbs, *bi = b->i, *bj = b->j; 156 int *ajtmpold, *ajtmp, nz, row, bslog,*ai=a->i,*aj=a->j; 157 int *diag_offset=b->diag,diag,bs=a->bs,bs2 = bs*bs,*v_pivots; 158 Scalar *ba = b->a,*aa = a->a; 159 register Scalar *pv,*v,*rtmp,*multiplier,*v_work,*pc,*w; 160 register int *pj; 161 162 ierr = ISInvertPermutation(iscol,&isicol); CHKERRQ(ierr); 163 PLogObjectParent(*B,isicol); 164 ierr = ISGetIndices(isrow,&r); CHKERRQ(ierr); 165 ierr = ISGetIndices(isicol,&ic); CHKERRQ(ierr); 166 rtmp = (Scalar *) PetscMalloc(bs2*(n+1)*sizeof(Scalar));CHKPTRQ(rtmp); 167 PetscMemzero(rtmp,bs2*(n+1)*sizeof(Scalar)); 168 /* generate work space needed by dense LU factorization */ 169 v_work = (Scalar *) PetscMalloc(bs*sizeof(int) + (bs+bs2)*sizeof(Scalar)); 170 CHKPTRQ(v_work); 171 multiplier = v_work + bs; 172 v_pivots = (int *) (multiplier + bs2); 173 174 /* flops in while loop */ 175 bslog = 2*bs*bs2; 176 177 for ( i=0; i<n; i++ ) { 178 nz = bi[i+1] - bi[i]; 179 ajtmp = bj + bi[i]; 180 for ( j=0; j<nz; j++ ) { 181 PetscMemzero(rtmp+bs2*ajtmp[j],bs2*sizeof(Scalar)); 182 } 183 /* load in initial (unfactored row) */ 184 nz = ai[r[i]+1] - ai[r[i]]; 185 ajtmpold = aj + ai[r[i]]; 186 v = aa + bs2*ai[r[i]]; 187 for ( j=0; j<nz; j++ ) { 188 PetscMemcpy(rtmp+bs2*ic[ajtmpold[j]],v+bs2*j,bs2*sizeof(Scalar)); 189 } 190 row = *ajtmp++; 191 while (row < i) { 192 pc = rtmp + bs2*row; 193 /* if (*pc) { */ 194 pv = ba + bs2*diag_offset[row]; 195 pj = bj + diag_offset[row] + 1; 196 Kernel_A_gets_A_times_B(bs,pc,pv,multiplier); 197 nz = bi[row+1] - diag_offset[row] - 1; 198 pv += bs2; 199 for (j=0; j<nz; j++) { 200 Kernel_A_gets_A_minus_B_times_C(bs,rtmp+bs2*pj[j],pc,pv+bs2*j); 201 } 202 PLogFlops(bslog*(nz+1)-bs); 203 /* } */ 204 row = *ajtmp++; 205 } 206 /* finished row so stick it into b->a */ 207 pv = ba + bs2*bi[i]; 208 pj = bj + bi[i]; 209 nz = bi[i+1] - bi[i]; 210 for ( j=0; j<nz; j++ ) { 211 PetscMemcpy(pv+bs2*j,rtmp+bs2*pj[j],bs2*sizeof(Scalar)); 212 } 213 diag = diag_offset[i] - bi[i]; 214 /* invert diagonal block */ 215 w = pv + bs2*diag; 216 Kernel_A_gets_inverse_A(bs,w,v_pivots,v_work); 217 } 218 219 PetscFree(rtmp); PetscFree(v_work); 220 ierr = ISRestoreIndices(isicol,&ic); CHKERRQ(ierr); 221 ierr = ISRestoreIndices(isrow,&r); CHKERRQ(ierr); 222 ierr = ISDestroy(isicol); CHKERRQ(ierr); 223 C->factor = FACTOR_LU; 224 C->assembled = PETSC_TRUE; 225 PLogFlops(1.3333*bs*bs2*b->mbs); /* from inverting diagonal blocks */ 226 return 0; 227 } 228 /* ------------------------------------------------------------*/ 229 /* 230 Version for when blocks are 5 by 5 231 */ 232 int MatLUFactorNumeric_SeqBAIJ_5(Mat A,Mat *B) 233 { 234 Mat C = *B; 235 Mat_SeqBAIJ *a = (Mat_SeqBAIJ *) A->data,*b = (Mat_SeqBAIJ *)C->data; 236 IS iscol = b->col, isrow = b->row, isicol; 237 int *r,*ic, ierr, i, j, n = a->mbs, *bi = b->i, *bj = b->j; 238 int *ajtmpold, *ajtmp, nz, row, v_pivots[5]; 239 int *diag_offset = b->diag,bs = 5,idx,*ai=a->i,*aj=a->j; 240 register Scalar *pv,*v,*rtmp,*pc,*w,*x; 241 Scalar p1,p2,p3,p4,v_work[5],m1,m2,m3,m4,m5,m6,m7,m8,m9,x1,x2,x3,x4; 242 Scalar p5,p6,p7,p8,p9,x5,x6,x7,x8,x9,x10,x11,x12,x13,x14,x15,x16; 243 Scalar x17,x18,x19,x20,x21,x22,x23,x24,x25,p10,p11,p12,p13,p14; 244 Scalar p15,p16,p17,p18,p19,p20,p21,p22,p23,p24,p25,m10,m11,m12; 245 Scalar m13,m14,m15,m16,m17,m18,m19,m20,m21,m22,m23,m24,m25; 246 Scalar *ba = b->a,*aa = a->a; 247 register int *pj; 248 249 ierr = ISInvertPermutation(iscol,&isicol); CHKERRQ(ierr); 250 PLogObjectParent(*B,isicol); 251 ierr = ISGetIndices(isrow,&r); CHKERRQ(ierr); 252 ierr = ISGetIndices(isicol,&ic); CHKERRQ(ierr); 253 rtmp = (Scalar *) PetscMalloc(25*(n+1)*sizeof(Scalar));CHKPTRQ(rtmp); 254 255 for ( i=0; i<n; i++ ) { 256 nz = bi[i+1] - bi[i]; 257 ajtmp = bj + bi[i]; 258 for ( j=0; j<nz; j++ ) { 259 x = rtmp+25*ajtmp[j]; 260 x[0] = x[1] = x[2] = x[3] = x[4] = x[5] = x[6] = x[7] = x[8] = x[9] = 0.0; 261 x[10] = x[11] = x[12] = x[13] = x[14] = x[15] = x[16] = x[17] = 0.0; 262 x[18] = x[19] = x[20] = x[21] = x[22] = x[23] = x[24] = 0.0; 263 } 264 /* load in initial (unfactored row) */ 265 idx = r[i]; 266 nz = ai[idx+1] - ai[idx]; 267 ajtmpold = aj + ai[idx]; 268 v = aa + 25*ai[idx]; 269 for ( j=0; j<nz; j++ ) { 270 x = rtmp+25*ic[ajtmpold[j]]; 271 x[0] = v[0]; x[1] = v[1]; x[2] = v[2]; x[3] = v[3]; 272 x[4] = v[4]; x[5] = v[5]; x[6] = v[6]; x[7] = v[7]; x[8] = v[8]; 273 x[9] = v[9]; x[10] = v[10]; x[11] = v[11]; x[12] = v[12]; x[13] = v[13]; 274 x[14] = v[14]; x[15] = v[15]; x[16] = v[16]; x[17] = v[17]; 275 x[18] = v[18]; x[19] = v[19]; x[20] = v[20]; x[21] = v[21]; 276 x[22] = v[22]; x[23] = v[23]; x[24] = v[24]; 277 v += 25; 278 } 279 row = *ajtmp++; 280 while (row < i) { 281 pc = rtmp + 25*row; 282 p1 = pc[0]; p2 = pc[1]; p3 = pc[2]; p4 = pc[3]; 283 p5 = pc[4]; p6 = pc[5]; p7 = pc[6]; p8 = pc[7]; p9 = pc[8]; 284 p10 = pc[9]; p11 = pc[10]; p12 = pc[11]; p13 = pc[12]; p14 = pc[13]; 285 p15 = pc[14]; p16 = pc[15]; p17 = pc[16]; p18 = pc[17]; p19 = pc[18]; 286 p20 = pc[19]; p21 = pc[20]; p22 = pc[21]; p23 = pc[22]; p24 = pc[23]; 287 p25 = pc[24]; 288 if (p1 != 0.0 || p2 != 0.0 || p3 != 0.0 || p4 != 0.0 || p5 != 0.0 || 289 p6 != 0.0 || p7 != 0.0 || p8 != 0.0 || p9 != 0.0 || p10 != 0.0 || 290 p11 != 0.0 || p12 != 0.0 || p13 != 0.0 || p14 != 0.0 || p15 != 0.0 291 || p16 != 0.0 || p17 != 0.0 || p18 != 0.0 || p19 != 0.0 || 292 p20 != 0.0 || p21 != 0.0 || p22 != 0.0 || p23 != 0.0 || 293 p24 != 0.0 || p25 != 0.0) { 294 pv = ba + 25*diag_offset[row]; 295 pj = bj + diag_offset[row] + 1; 296 x1 = pv[0]; x2 = pv[1]; x3 = pv[2]; x4 = pv[3]; 297 x5 = pv[4]; x6 = pv[5]; x7 = pv[6]; x8 = pv[7]; x9 = pv[8]; 298 x10 = pv[9]; x11 = pv[10]; x12 = pv[11]; x13 = pv[12]; x14 = pv[13]; 299 x15 = pv[14]; x16 = pv[15]; x17 = pv[16]; x18 = pv[17]; 300 x19 = pv[18]; x20 = pv[19]; x21 = pv[20]; x22 = pv[21]; 301 x23 = pv[22]; x24 = pv[23]; x25 = pv[24]; 302 pc[0] = m1 = p1*x1 + p6*x2 + p11*x3 + p16*x4 + p21*x5; 303 pc[1] = m2 = p2*x1 + p7*x2 + p12*x3 + p17*x4 + p22*x5; 304 pc[2] = m3 = p3*x1 + p8*x2 + p13*x3 + p18*x4 + p23*x5; 305 pc[3] = m4 = p4*x1 + p9*x2 + p14*x3 + p19*x4 + p24*x5; 306 pc[4] = m5 = p5*x1 + p10*x2 + p15*x3 + p20*x4 + p25*x5; 307 308 pc[5] = m6 = p1*x6 + p6*x7 + p11*x8 + p16*x9 + p21*x10; 309 pc[6] = m7 = p2*x6 + p7*x7 + p12*x8 + p17*x9 + p22*x10; 310 pc[7] = m8 = p3*x6 + p8*x7 + p13*x8 + p18*x9 + p23*x10; 311 pc[8] = m9 = p4*x6 + p9*x7 + p14*x8 + p19*x9 + p24*x10; 312 pc[9] = m10 = p5*x6 + p10*x7 + p15*x8 + p20*x9 + p25*x10; 313 314 pc[10] = m11 = p1*x11 + p6*x12 + p11*x13 + p16*x14 + p21*x15; 315 pc[11] = m12 = p2*x11 + p7*x12 + p12*x13 + p17*x14 + p22*x15; 316 pc[12] = m13 = p3*x11 + p8*x12 + p13*x13 + p18*x14 + p23*x15; 317 pc[13] = m14 = p4*x11 + p9*x12 + p14*x13 + p19*x14 + p24*x15; 318 pc[14] = m15 = p5*x11 + p10*x12 + p15*x13 + p20*x14 + p25*x15; 319 320 pc[15] = m16 = p1*x16 + p6*x17 + p11*x18 + p16*x19 + p21*x20; 321 pc[16] = m17 = p2*x16 + p7*x17 + p12*x18 + p17*x19 + p22*x20; 322 pc[17] = m18 = p3*x16 + p8*x17 + p13*x18 + p18*x19 + p23*x20; 323 pc[18] = m19 = p4*x16 + p9*x17 + p14*x18 + p19*x19 + p24*x20; 324 pc[19] = m20 = p5*x16 + p10*x17 + p15*x18 + p20*x19 + p25*x20; 325 326 pc[20] = m21 = p1*x21 + p6*x22 + p11*x23 + p16*x24 + p21*x25; 327 pc[21] = m22 = p2*x21 + p7*x22 + p12*x23 + p17*x24 + p22*x25; 328 pc[22] = m23 = p3*x21 + p8*x22 + p13*x23 + p18*x24 + p23*x25; 329 pc[23] = m24 = p4*x21 + p9*x22 + p14*x23 + p19*x24 + p24*x25; 330 pc[24] = m25 = p5*x21 + p10*x22 + p15*x23 + p20*x24 + p25*x25; 331 332 nz = bi[row+1] - diag_offset[row] - 1; 333 pv += 25; 334 for (j=0; j<nz; j++) { 335 x1 = pv[0]; x2 = pv[1]; x3 = pv[2]; x4 = pv[3]; 336 x5 = pv[4]; x6 = pv[5]; x7 = pv[6]; x8 = pv[7]; x9 = pv[8]; 337 x10 = pv[9]; x11 = pv[10]; x12 = pv[11]; x13 = pv[12]; 338 x14 = pv[13]; x15 = pv[14]; x16 = pv[15]; x17 = pv[16]; 339 x18 = pv[17]; x19 = pv[18]; x20 = pv[19]; x21 = pv[20]; 340 x22 = pv[21]; x23 = pv[22]; x24 = pv[23]; x25 = pv[24]; 341 x = rtmp + 25*pj[j]; 342 x[0] -= m1*x1 + m6*x2 + m11*x3 + m16*x4 + m21*x5; 343 x[1] -= m2*x1 + m7*x2 + m12*x3 + m17*x4 + m22*x5; 344 x[2] -= m3*x1 + m8*x2 + m13*x3 + m18*x4 + m23*x5; 345 x[3] -= m4*x1 + m9*x2 + m14*x3 + m19*x4 + m24*x5; 346 x[4] -= m5*x1 + m10*x2 + m15*x3 + m20*x4 + m25*x5; 347 348 x[5] -= m1*x6 + m6*x7 + m11*x8 + m16*x9 + m21*x10; 349 x[6] -= m2*x6 + m7*x7 + m12*x8 + m17*x9 + m22*x10; 350 x[7] -= m3*x6 + m8*x7 + m13*x8 + m18*x9 + m23*x10; 351 x[8] -= m4*x6 + m9*x7 + m14*x8 + m19*x9 + m24*x10; 352 x[9] -= m5*x6 + m10*x7 + m15*x8 + m20*x9 + m25*x10; 353 354 x[10] -= m1*x11 + m6*x12 + m11*x13 + m16*x14 + m21*x15; 355 x[11] -= m2*x11 + m7*x12 + m12*x13 + m17*x14 + m22*x15; 356 x[12] -= m3*x11 + m8*x12 + m13*x13 + m18*x14 + m23*x15; 357 x[13] -= m4*x11 + m9*x12 + m14*x13 + m19*x14 + m24*x15; 358 x[14] -= m5*x11 + m10*x12 + m15*x13 + m20*x14 + m25*x15; 359 360 x[15] -= m1*x16 + m6*x17 + m11*x18 + m16*x19 + m21*x20; 361 x[16] -= m2*x16 + m7*x17 + m12*x18 + m17*x19 + m22*x20; 362 x[17] -= m3*x16 + m8*x17 + m13*x18 + m18*x19 + m23*x20; 363 x[18] -= m4*x16 + m9*x17 + m14*x18 + m19*x19 + m24*x20; 364 x[19] -= m5*x16 + m10*x17 + m15*x18 + m20*x19 + m25*x20; 365 366 x[20] -= m1*x21 + m6*x22 + m11*x23 + m16*x24 + m21*x25; 367 x[21] -= m2*x21 + m7*x22 + m12*x23 + m17*x24 + m22*x25; 368 x[22] -= m3*x21 + m8*x22 + m13*x23 + m18*x24 + m23*x25; 369 x[23] -= m4*x21 + m9*x22 + m14*x23 + m19*x24 + m24*x25; 370 x[24] -= m5*x21 + m10*x22 + m15*x23 + m20*x24 + m25*x25; 371 372 pv += 25; 373 } 374 PLogFlops(250*nz+225); 375 } 376 row = *ajtmp++; 377 } 378 /* finished row so stick it into b->a */ 379 pv = ba + 25*bi[i]; 380 pj = bj + bi[i]; 381 nz = bi[i+1] - bi[i]; 382 for ( j=0; j<nz; j++ ) { 383 x = rtmp+25*pj[j]; 384 pv[0] = x[0]; pv[1] = x[1]; pv[2] = x[2]; pv[3] = x[3]; 385 pv[4] = x[4]; pv[5] = x[5]; pv[6] = x[6]; pv[7] = x[7]; pv[8] = x[8]; 386 pv[9] = x[9]; pv[10] = x[10]; pv[11] = x[11]; pv[12] = x[12]; 387 pv[13] = x[13]; pv[14] = x[14]; pv[15] = x[15]; pv[16] = x[16]; 388 pv[17] = x[17]; pv[18] = x[18]; pv[19] = x[19]; pv[20] = x[20]; 389 pv[21] = x[21]; pv[22] = x[22]; pv[23] = x[23]; pv[24] = x[24]; 390 pv += 25; 391 } 392 /* invert diagonal block */ 393 w = ba + 25*diag_offset[i]; 394 Kernel_A_gets_inverse_A(bs,w,v_pivots,v_work); 395 } 396 397 PetscFree(rtmp); 398 ierr = ISRestoreIndices(isicol,&ic); CHKERRQ(ierr); 399 ierr = ISRestoreIndices(isrow,&r); CHKERRQ(ierr); 400 ierr = ISDestroy(isicol); CHKERRQ(ierr); 401 C->factor = FACTOR_LU; 402 C->assembled = PETSC_TRUE; 403 PLogFlops(1.3333*125*b->mbs); /* from inverting diagonal blocks */ 404 return 0; 405 } 406 407 /* ------------------------------------------------------------*/ 408 /* 409 Version for when blocks are 4 by 4 410 */ 411 int MatLUFactorNumeric_SeqBAIJ_4(Mat A,Mat *B) 412 { 413 Mat C = *B; 414 Mat_SeqBAIJ *a = (Mat_SeqBAIJ *) A->data,*b = (Mat_SeqBAIJ *)C->data; 415 IS iscol = b->col, isrow = b->row, isicol; 416 int *r,*ic, ierr, i, j, n = a->mbs, *bi = b->i, *bj = b->j; 417 int *ajtmpold, *ajtmp, nz, row, v_pivots[4]; 418 int *diag_offset = b->diag,bs = 4,idx,*ai=a->i,*aj=a->j; 419 register Scalar *pv,*v,*rtmp,*pc,*w,*x; 420 Scalar p1,p2,p3,p4,v_work[4],m1,m2,m3,m4,m5,m6,m7,m8,m9,x1,x2,x3,x4; 421 Scalar p5,p6,p7,p8,p9,x5,x6,x7,x8,x9,x10,x11,x12,x13,x14,x15,x16; 422 Scalar p10,p11,p12,p13,p14,p15,p16,m10,m11,m12; 423 Scalar m13,m14,m15,m16; 424 Scalar *ba = b->a,*aa = a->a; 425 register int *pj; 426 427 ierr = ISInvertPermutation(iscol,&isicol); CHKERRQ(ierr); 428 PLogObjectParent(*B,isicol); 429 ierr = ISGetIndices(isrow,&r); CHKERRQ(ierr); 430 ierr = ISGetIndices(isicol,&ic); CHKERRQ(ierr); 431 rtmp = (Scalar *) PetscMalloc(16*(n+1)*sizeof(Scalar));CHKPTRQ(rtmp); 432 433 for ( i=0; i<n; i++ ) { 434 nz = bi[i+1] - bi[i]; 435 ajtmp = bj + bi[i]; 436 for ( j=0; j<nz; j++ ) { 437 x = rtmp+16*ajtmp[j]; 438 x[0] = x[1] = x[2] = x[3] = x[4] = x[5] = x[6] = x[7] = x[8] = x[9] = 0.0; 439 x[10] = x[11] = x[12] = x[13] = x[14] = x[15] = 0.0; 440 } 441 /* load in initial (unfactored row) */ 442 idx = r[i]; 443 nz = ai[idx+1] - ai[idx]; 444 ajtmpold = aj + ai[idx]; 445 v = aa + 16*ai[idx]; 446 for ( j=0; j<nz; j++ ) { 447 x = rtmp+16*ic[ajtmpold[j]]; 448 x[0] = v[0]; x[1] = v[1]; x[2] = v[2]; x[3] = v[3]; 449 x[4] = v[4]; x[5] = v[5]; x[6] = v[6]; x[7] = v[7]; x[8] = v[8]; 450 x[9] = v[9]; x[10] = v[10]; x[11] = v[11]; x[12] = v[12]; x[13] = v[13]; 451 x[14] = v[14]; x[15] = v[15]; 452 v += 16; 453 } 454 row = *ajtmp++; 455 while (row < i) { 456 pc = rtmp + 16*row; 457 p1 = pc[0]; p2 = pc[1]; p3 = pc[2]; p4 = pc[3]; 458 p5 = pc[4]; p6 = pc[5]; p7 = pc[6]; p8 = pc[7]; p9 = pc[8]; 459 p10 = pc[9]; p11 = pc[10]; p12 = pc[11]; p13 = pc[12]; p14 = pc[13]; 460 p15 = pc[14]; p16 = pc[15]; 461 if (p1 != 0.0 || p2 != 0.0 || p3 != 0.0 || p4 != 0.0 || p5 != 0.0 || 462 p6 != 0.0 || p7 != 0.0 || p8 != 0.0 || p9 != 0.0 || p10 != 0.0 || 463 p11 != 0.0 || p12 != 0.0 || p13 != 0.0 || p14 != 0.0 || p15 != 0.0 464 || p16 != 0.0) { 465 pv = ba + 16*diag_offset[row]; 466 pj = bj + diag_offset[row] + 1; 467 x1 = pv[0]; x2 = pv[1]; x3 = pv[2]; x4 = pv[3]; 468 x5 = pv[4]; x6 = pv[5]; x7 = pv[6]; x8 = pv[7]; x9 = pv[8]; 469 x10 = pv[9]; x11 = pv[10]; x12 = pv[11]; x13 = pv[12]; x14 = pv[13]; 470 x15 = pv[14]; x16 = pv[15]; 471 pc[0] = m1 = p1*x1 + p5*x2 + p9*x3 + p13*x4; 472 pc[1] = m2 = p2*x1 + p6*x2 + p10*x3 + p14*x4; 473 pc[2] = m3 = p3*x1 + p7*x2 + p11*x3 + p15*x4; 474 pc[3] = m4 = p4*x1 + p8*x2 + p12*x3 + p16*x4; 475 476 pc[4] = m5 = p1*x5 + p5*x6 + p9*x7 + p13*x8; 477 pc[5] = m6 = p2*x5 + p6*x6 + p10*x7 + p14*x8; 478 pc[6] = m7 = p3*x5 + p7*x6 + p11*x7 + p15*x8; 479 pc[7] = m8 = p4*x5 + p8*x6 + p12*x7 + p16*x8; 480 481 pc[8] = m9 = p1*x9 + p5*x10 + p9*x11 + p13*x12; 482 pc[9] = m10 = p2*x9 + p6*x10 + p10*x11 + p14*x12; 483 pc[10] = m11 = p3*x9 + p7*x10 + p11*x11 + p15*x12; 484 pc[11] = m12 = p4*x9 + p8*x10 + p12*x11 + p16*x12; 485 486 pc[12] = m13 = p1*x13 + p5*x14 + p9*x15 + p13*x16; 487 pc[13] = m14 = p2*x13 + p6*x14 + p10*x15 + p14*x16; 488 pc[14] = m15 = p3*x13 + p7*x14 + p11*x15 + p15*x16; 489 pc[15] = m16 = p4*x13 + p8*x14 + p12*x15 + p16*x16; 490 491 nz = bi[row+1] - diag_offset[row] - 1; 492 pv += 16; 493 for (j=0; j<nz; j++) { 494 x1 = pv[0]; x2 = pv[1]; x3 = pv[2]; x4 = pv[3]; 495 x5 = pv[4]; x6 = pv[5]; x7 = pv[6]; x8 = pv[7]; x9 = pv[8]; 496 x10 = pv[9]; x11 = pv[10]; x12 = pv[11]; x13 = pv[12]; 497 x14 = pv[13]; x15 = pv[14]; x16 = pv[15]; 498 x = rtmp + 16*pj[j]; 499 x[0] -= m1*x1 + m5*x2 + m9*x3 + m13*x4; 500 x[1] -= m2*x1 + m6*x2 + m10*x3 + m14*x4; 501 x[2] -= m3*x1 + m7*x2 + m11*x3 + m15*x4; 502 x[3] -= m4*x1 + m8*x2 + m12*x3 + m16*x4; 503 504 x[4] -= m1*x5 + m5*x6 + m9*x7 + m13*x8; 505 x[5] -= m2*x5 + m6*x6 + m10*x7 + m14*x8; 506 x[6] -= m3*x5 + m7*x6 + m11*x7 + m15*x8; 507 x[7] -= m4*x5 + m8*x6 + m12*x7 + m16*x8; 508 509 x[8] -= m1*x9 + m5*x10 + m9*x11 + m13*x12; 510 x[9] -= m2*x9 + m6*x10 + m10*x11 + m14*x12; 511 x[10] -= m3*x9 + m7*x10 + m11*x11 + m15*x12; 512 x[11] -= m4*x9 + m8*x10 + m12*x11 + m16*x12; 513 514 x[12] -= m1*x13 + m5*x14 + m9*x15 + m13*x16; 515 x[13] -= m2*x13 + m6*x14 + m10*x15 + m14*x16; 516 x[14] -= m3*x13 + m7*x14 + m11*x15 + m15*x16; 517 x[15] -= m4*x13 + m8*x14 + m12*x15 + m16*x16; 518 519 pv += 16; 520 } 521 PLogFlops(128*nz+112); 522 } 523 row = *ajtmp++; 524 } 525 /* finished row so stick it into b->a */ 526 pv = ba + 16*bi[i]; 527 pj = bj + bi[i]; 528 nz = bi[i+1] - bi[i]; 529 for ( j=0; j<nz; j++ ) { 530 x = rtmp+16*pj[j]; 531 pv[0] = x[0]; pv[1] = x[1]; pv[2] = x[2]; pv[3] = x[3]; 532 pv[4] = x[4]; pv[5] = x[5]; pv[6] = x[6]; pv[7] = x[7]; pv[8] = x[8]; 533 pv[9] = x[9]; pv[10] = x[10]; pv[11] = x[11]; pv[12] = x[12]; 534 pv[13] = x[13]; pv[14] = x[14]; pv[15] = x[15]; 535 pv += 16; 536 } 537 /* invert diagonal block */ 538 w = ba + 16*diag_offset[i]; 539 Kernel_A_gets_inverse_A(bs,w,v_pivots,v_work); 540 } 541 542 PetscFree(rtmp); 543 ierr = ISRestoreIndices(isicol,&ic); CHKERRQ(ierr); 544 ierr = ISRestoreIndices(isrow,&r); CHKERRQ(ierr); 545 ierr = ISDestroy(isicol); CHKERRQ(ierr); 546 C->factor = FACTOR_LU; 547 C->assembled = PETSC_TRUE; 548 PLogFlops(1.3333*64*b->mbs); /* from inverting diagonal blocks */ 549 return 0; 550 } 551 /* ------------------------------------------------------------*/ 552 /* 553 Version for when blocks are 3 by 3 554 */ 555 int MatLUFactorNumeric_SeqBAIJ_3(Mat A,Mat *B) 556 { 557 Mat C = *B; 558 Mat_SeqBAIJ *a = (Mat_SeqBAIJ *) A->data,*b = (Mat_SeqBAIJ *)C->data; 559 IS iscol = b->col, isrow = b->row, isicol; 560 int *r,*ic, ierr, i, j, n = a->mbs, *bi = b->i, *bj = b->j; 561 int *ajtmpold, *ajtmp, nz, row, *ai=a->i,*aj=a->j; 562 int *diag_offset = b->diag,idx; 563 register Scalar *pv,*v,*rtmp,*pc,*w,*x; 564 Scalar p1,p2,p3,p4,m1,m2,m3,m4,m5,m6,m7,m8,m9,x1,x2,x3,x4; 565 Scalar p5,p6,p7,p8,p9,x5,x6,x7,x8,x9; 566 Scalar *ba = b->a,*aa = a->a; 567 register int *pj; 568 569 ierr = ISInvertPermutation(iscol,&isicol); CHKERRQ(ierr); 570 PLogObjectParent(*B,isicol); 571 ierr = ISGetIndices(isrow,&r); CHKERRQ(ierr); 572 ierr = ISGetIndices(isicol,&ic); CHKERRQ(ierr); 573 rtmp = (Scalar *) PetscMalloc(9*(n+1)*sizeof(Scalar));CHKPTRQ(rtmp); 574 575 for ( i=0; i<n; i++ ) { 576 nz = bi[i+1] - bi[i]; 577 ajtmp = bj + bi[i]; 578 for ( j=0; j<nz; j++ ) { 579 x = rtmp + 9*ajtmp[j]; 580 x[0] = x[1] = x[2] = x[3] = x[4] = x[5] = x[6] = x[7] = x[8] = x[9] = 0.0; 581 } 582 /* load in initial (unfactored row) */ 583 idx = r[i]; 584 nz = ai[idx+1] - ai[idx]; 585 ajtmpold = aj + ai[idx]; 586 v = aa + 9*ai[idx]; 587 for ( j=0; j<nz; j++ ) { 588 x = rtmp + 9*ic[ajtmpold[j]]; 589 x[0] = v[0]; x[1] = v[1]; x[2] = v[2]; x[3] = v[3]; 590 x[4] = v[4]; x[5] = v[5]; x[6] = v[6]; x[7] = v[7]; x[8] = v[8]; 591 v += 9; 592 } 593 row = *ajtmp++; 594 while (row < i) { 595 pc = rtmp + 9*row; 596 p1 = pc[0]; p2 = pc[1]; p3 = pc[2]; p4 = pc[3]; 597 p5 = pc[4]; p6 = pc[5]; p7 = pc[6]; p8 = pc[7]; p9 = pc[8]; 598 if (p1 != 0.0 || p2 != 0.0 || p3 != 0.0 || p4 != 0.0 || p5 != 0.0 || 599 p6 != 0.0 || p7 != 0.0 || p8 != 0.0 || p9 != 0.0) { 600 pv = ba + 9*diag_offset[row]; 601 pj = bj + diag_offset[row] + 1; 602 x1 = pv[0]; x2 = pv[1]; x3 = pv[2]; x4 = pv[3]; 603 x5 = pv[4]; x6 = pv[5]; x7 = pv[6]; x8 = pv[7]; x9 = pv[8]; 604 pc[0] = m1 = p1*x1 + p4*x2 + p7*x3; 605 pc[1] = m2 = p2*x1 + p5*x2 + p8*x3; 606 pc[2] = m3 = p3*x1 + p6*x2 + p9*x3; 607 608 pc[3] = m4 = p1*x4 + p4*x5 + p7*x6; 609 pc[4] = m5 = p2*x4 + p5*x5 + p8*x6; 610 pc[5] = m6 = p3*x4 + p6*x5 + p9*x6; 611 612 pc[6] = m7 = p1*x7 + p4*x8 + p7*x9; 613 pc[7] = m8 = p2*x7 + p5*x8 + p8*x9; 614 pc[8] = m9 = p3*x7 + p6*x8 + p9*x9; 615 nz = bi[row+1] - diag_offset[row] - 1; 616 pv += 9; 617 for (j=0; j<nz; j++) { 618 x1 = pv[0]; x2 = pv[1]; x3 = pv[2]; x4 = pv[3]; 619 x5 = pv[4]; x6 = pv[5]; x7 = pv[6]; x8 = pv[7]; x9 = pv[8]; 620 x = rtmp + 9*pj[j]; 621 x[0] -= m1*x1 + m4*x2 + m7*x3; 622 x[1] -= m2*x1 + m5*x2 + m8*x3; 623 x[2] -= m3*x1 + m6*x2 + m9*x3; 624 625 x[3] -= m1*x4 + m4*x5 + m7*x6; 626 x[4] -= m2*x4 + m5*x5 + m8*x6; 627 x[5] -= m3*x4 + m6*x5 + m9*x6; 628 629 x[6] -= m1*x7 + m4*x8 + m7*x9; 630 x[7] -= m2*x7 + m5*x8 + m8*x9; 631 x[8] -= m3*x7 + m6*x8 + m9*x9; 632 pv += 9; 633 } 634 PLogFlops(54*nz+36); 635 } 636 row = *ajtmp++; 637 } 638 /* finished row so stick it into b->a */ 639 pv = ba + 9*bi[i]; 640 pj = bj + bi[i]; 641 nz = bi[i+1] - bi[i]; 642 for ( j=0; j<nz; j++ ) { 643 x = rtmp + 9*pj[j]; 644 pv[0] = x[0]; pv[1] = x[1]; pv[2] = x[2]; pv[3] = x[3]; 645 pv[4] = x[4]; pv[5] = x[5]; pv[6] = x[6]; pv[7] = x[7]; pv[8] = x[8]; 646 pv += 9; 647 } 648 /* invert diagonal block */ 649 w = ba + 9*diag_offset[i]; 650 ierr = Kernel_A_gets_inverse_A_3(w); CHKERRQ(ierr); 651 } 652 653 PetscFree(rtmp); 654 ierr = ISRestoreIndices(isicol,&ic); CHKERRQ(ierr); 655 ierr = ISRestoreIndices(isrow,&r); CHKERRQ(ierr); 656 ierr = ISDestroy(isicol); CHKERRQ(ierr); 657 C->factor = FACTOR_LU; 658 C->assembled = PETSC_TRUE; 659 PLogFlops(1.3333*27*b->mbs); /* from inverting diagonal blocks */ 660 return 0; 661 } 662 663 /* ------------------------------------------------------------*/ 664 /* 665 Version for when blocks are 2 by 2 666 */ 667 int MatLUFactorNumeric_SeqBAIJ_2(Mat A,Mat *B) 668 { 669 Mat C = *B; 670 Mat_SeqBAIJ *a = (Mat_SeqBAIJ *) A->data,*b = (Mat_SeqBAIJ *)C->data; 671 IS iscol = b->col, isrow = b->row, isicol; 672 int *r,*ic, ierr, i, j, n = a->mbs, *bi = b->i, *bj = b->j; 673 int *ajtmpold, *ajtmp, nz, row, v_pivots[2]; 674 int *diag_offset=b->diag,bs = 2,idx,*ai=a->i,*aj=a->j; 675 register Scalar *pv,*v,*rtmp,m1,m2,m3,m4,*pc,*w,*x,x1,x2,x3,x4; 676 Scalar p1,p2,p3,p4,v_work[2]; 677 Scalar *ba = b->a,*aa = a->a; 678 register int *pj; 679 680 ierr = ISInvertPermutation(iscol,&isicol); CHKERRQ(ierr); 681 PLogObjectParent(*B,isicol); 682 ierr = ISGetIndices(isrow,&r); CHKERRQ(ierr); 683 ierr = ISGetIndices(isicol,&ic); CHKERRQ(ierr); 684 rtmp = (Scalar *) PetscMalloc(4*(n+1)*sizeof(Scalar));CHKPTRQ(rtmp); 685 686 for ( i=0; i<n; i++ ) { 687 nz = bi[i+1] - bi[i]; 688 ajtmp = bj + bi[i]; 689 for ( j=0; j<nz; j++ ) { 690 x = rtmp+4*ajtmp[j]; x[0] = x[1] = x[2] = x[3] = 0.0; 691 } 692 /* load in initial (unfactored row) */ 693 idx = r[i]; 694 nz = ai[idx+1] - ai[idx]; 695 ajtmpold = aj + ai[idx]; 696 v = aa + 4*ai[idx]; 697 for ( j=0; j<nz; j++ ) { 698 x = rtmp+4*ic[ajtmpold[j]]; 699 x[0] = v[0]; x[1] = v[1]; x[2] = v[2]; x[3] = v[3]; 700 v += 4; 701 } 702 row = *ajtmp++; 703 while (row < i) { 704 pc = rtmp + 4*row; 705 p1 = pc[0]; p2 = pc[1]; p3 = pc[2]; p4 = pc[3]; 706 if (p1 != 0.0 || p2 != 0.0 || p3 != 0.0 || p4 != 0.0) { 707 pv = ba + 4*diag_offset[row]; 708 pj = bj + diag_offset[row] + 1; 709 x1 = pv[0]; x2 = pv[1]; x3 = pv[2]; x4 = pv[3]; 710 pc[0] = m1 = p1*x1 + p3*x2; 711 pc[1] = m2 = p2*x1 + p4*x2; 712 pc[2] = m3 = p1*x3 + p3*x4; 713 pc[3] = m4 = p2*x3 + p4*x4; 714 nz = bi[row+1] - diag_offset[row] - 1; 715 pv += 4; 716 for (j=0; j<nz; j++) { 717 x1 = pv[0]; x2 = pv[1]; x3 = pv[2]; x4 = pv[3]; 718 x = rtmp + 4*pj[j]; 719 x[0] -= m1*x1 + m3*x2; 720 x[1] -= m2*x1 + m4*x2; 721 x[2] -= m1*x3 + m3*x4; 722 x[3] -= m2*x3 + m4*x4; 723 pv += 4; 724 } 725 PLogFlops(16*nz+12); 726 } 727 row = *ajtmp++; 728 } 729 /* finished row so stick it into b->a */ 730 pv = ba + 4*bi[i]; 731 pj = bj + bi[i]; 732 nz = bi[i+1] - bi[i]; 733 for ( j=0; j<nz; j++ ) { 734 x = rtmp+4*pj[j]; 735 pv[0] = x[0]; pv[1] = x[1]; pv[2] = x[2]; pv[3] = x[3]; 736 pv += 4; 737 } 738 /* invert diagonal block */ 739 w = ba + 4*diag_offset[i]; 740 Kernel_A_gets_inverse_A(bs,w,v_pivots,v_work); 741 } 742 743 PetscFree(rtmp); 744 ierr = ISRestoreIndices(isicol,&ic); CHKERRQ(ierr); 745 ierr = ISRestoreIndices(isrow,&r); CHKERRQ(ierr); 746 ierr = ISDestroy(isicol); CHKERRQ(ierr); 747 C->factor = FACTOR_LU; 748 C->assembled = PETSC_TRUE; 749 PLogFlops(1.3333*8*b->mbs); /* from inverting diagonal blocks */ 750 return 0; 751 } 752 753 /* ----------------------------------------------------------- */ 754 /* 755 Version for when blocks are 1 by 1. 756 */ 757 int MatLUFactorNumeric_SeqBAIJ_1(Mat A,Mat *B) 758 { 759 Mat C = *B; 760 Mat_SeqBAIJ *a = (Mat_SeqBAIJ *) A->data, *b = (Mat_SeqBAIJ *)C->data; 761 IS iscol = b->col, isrow = b->row, isicol; 762 int *r,*ic, ierr, i, j, n = a->mbs, *bi = b->i, *bj = b->j; 763 int *ajtmpold, *ajtmp, nz, row,*ai = a->i,*aj = a->j; 764 int *diag_offset = b->diag,diag; 765 register Scalar *pv,*v,*rtmp,multiplier,*pc; 766 Scalar *ba = b->a,*aa = a->a; 767 register int *pj; 768 769 ierr = ISInvertPermutation(iscol,&isicol); CHKERRQ(ierr); 770 PLogObjectParent(*B,isicol); 771 ierr = ISGetIndices(isrow,&r); CHKERRQ(ierr); 772 ierr = ISGetIndices(isicol,&ic); CHKERRQ(ierr); 773 rtmp = (Scalar *) PetscMalloc((n+1)*sizeof(Scalar));CHKPTRQ(rtmp); 774 775 for ( i=0; i<n; i++ ) { 776 nz = bi[i+1] - bi[i]; 777 ajtmp = bj + bi[i]; 778 for ( j=0; j<nz; j++ ) rtmp[ajtmp[j]] = 0.0; 779 780 /* load in initial (unfactored row) */ 781 nz = ai[r[i]+1] - ai[r[i]]; 782 ajtmpold = aj + ai[r[i]]; 783 v = aa + ai[r[i]]; 784 for ( j=0; j<nz; j++ ) rtmp[ic[ajtmpold[j]]] = v[j]; 785 786 row = *ajtmp++; 787 while (row < i) { 788 pc = rtmp + row; 789 if (*pc != 0.0) { 790 pv = ba + diag_offset[row]; 791 pj = bj + diag_offset[row] + 1; 792 multiplier = *pc * *pv++; 793 *pc = multiplier; 794 nz = bi[row+1] - diag_offset[row] - 1; 795 for (j=0; j<nz; j++) rtmp[pj[j]] -= multiplier * pv[j]; 796 PLogFlops(1+2*nz); 797 } 798 row = *ajtmp++; 799 } 800 /* finished row so stick it into b->a */ 801 pv = ba + bi[i]; 802 pj = bj + bi[i]; 803 nz = bi[i+1] - bi[i]; 804 for ( j=0; j<nz; j++ ) {pv[j] = rtmp[pj[j]];} 805 diag = diag_offset[i] - bi[i]; 806 /* check pivot entry for current row */ 807 if (pv[diag] == 0.0) { 808 SETERRQ(1,"MatLUFactorNumeric_SeqBAIJ_1:Zero pivot"); 809 } 810 pv[diag] = 1.0/pv[diag]; 811 } 812 813 PetscFree(rtmp); 814 ierr = ISRestoreIndices(isicol,&ic); CHKERRQ(ierr); 815 ierr = ISRestoreIndices(isrow,&r); CHKERRQ(ierr); 816 ierr = ISDestroy(isicol); CHKERRQ(ierr); 817 C->factor = FACTOR_LU; 818 C->assembled = PETSC_TRUE; 819 PLogFlops(b->n); 820 return 0; 821 } 822 823 /* ----------------------------------------------------------- */ 824 int MatLUFactor_SeqBAIJ(Mat A,IS row,IS col,double f) 825 { 826 Mat_SeqBAIJ *mat = (Mat_SeqBAIJ *) A->data; 827 int ierr; 828 Mat C; 829 830 ierr = MatLUFactorSymbolic_SeqBAIJ(A,row,col,f,&C); CHKERRQ(ierr); 831 ierr = MatLUFactorNumeric(A,&C); CHKERRQ(ierr); 832 833 /* free all the data structures from mat */ 834 PetscFree(mat->a); 835 if (!mat->singlemalloc) {PetscFree(mat->i); PetscFree(mat->j);} 836 if (mat->diag) PetscFree(mat->diag); 837 if (mat->ilen) PetscFree(mat->ilen); 838 if (mat->imax) PetscFree(mat->imax); 839 if (mat->solve_work) PetscFree(mat->solve_work); 840 if (mat->mult_work) PetscFree(mat->mult_work); 841 PetscFree(mat); 842 843 PetscMemcpy(A,C,sizeof(struct _Mat)); 844 PetscHeaderDestroy(C); 845 return 0; 846 } 847 /* ----------------------------------------------------------- */ 848 int MatSolve_SeqBAIJ_N(Mat A,Vec bb,Vec xx) 849 { 850 Mat_SeqBAIJ *a=(Mat_SeqBAIJ *)A->data; 851 IS iscol=a->col,isrow=a->row; 852 int *r,*c,ierr,i,n=a->mbs,*vi,*ai=a->i,*aj=a->j; 853 int nz,bs=a->bs,bs2=a->bs2; 854 Scalar *aa=a->a,*sum; 855 register Scalar *x,*b,*lsum,*tmp,*v; 856 857 VecGetArray_Fast(bb,b); 858 VecGetArray_Fast(xx,x); 859 tmp = a->solve_work; 860 861 ierr = ISGetIndices(isrow,&r);CHKERRQ(ierr); 862 ierr = ISGetIndices(iscol,&c);CHKERRQ(ierr); c = c + (n-1); 863 864 /* forward solve the lower triangular */ 865 PetscMemcpy(tmp,b + bs*(*r++), bs*sizeof(Scalar)); 866 for ( i=1; i<n; i++ ) { 867 v = aa + bs2*ai[i]; 868 vi = aj + ai[i]; 869 nz = a->diag[i] - ai[i]; 870 sum = tmp + bs*i; 871 PetscMemcpy(sum,b+bs*(*r++),bs*sizeof(Scalar)); 872 while (nz--) { 873 Kernel_v_gets_v_minus_A_times_w(bs,sum,v,tmp+bs*(*vi++)); 874 v += bs2; 875 } 876 } 877 /* backward solve the upper triangular */ 878 lsum = a->solve_work + a->n; 879 for ( i=n-1; i>=0; i-- ){ 880 v = aa + bs2*(a->diag[i] + 1); 881 vi = aj + a->diag[i] + 1; 882 nz = ai[i+1] - a->diag[i] - 1; 883 PetscMemcpy(lsum,tmp+i*bs,bs*sizeof(Scalar)); 884 while (nz--) { 885 Kernel_v_gets_v_minus_A_times_w(bs,lsum,v,tmp+bs*(*vi++)); 886 v += bs2; 887 } 888 Kernel_w_gets_A_times_v(bs,lsum,aa+bs2*a->diag[i],tmp+i*bs); 889 PetscMemcpy(x + bs*(*c--),tmp+i*bs,bs*sizeof(Scalar)); 890 } 891 892 ierr = ISRestoreIndices(isrow,&r); CHKERRQ(ierr); 893 ierr = ISRestoreIndices(iscol,&c); CHKERRQ(ierr); 894 VecRestoreArray_Fast(bb,b); 895 VecRestoreArray_Fast(xx,x); 896 PLogFlops(2*(a->bs2)*(a->nz) - a->n); 897 return 0; 898 } 899 900 int MatSolve_SeqBAIJ_5(Mat A,Vec bb,Vec xx) 901 { 902 Mat_SeqBAIJ *a=(Mat_SeqBAIJ *)A->data; 903 IS iscol=a->col,isrow=a->row; 904 int *r,*c,ierr,i,n=a->mbs,*vi,*ai=a->i,*aj=a->j,nz,idx,idt,idc; 905 Scalar *aa=a->a,sum1,sum2,sum3,sum4,sum5,x1,x2,x3,x4,x5; 906 register Scalar *x,*b,*tmp,*v; 907 908 VecGetArray_Fast(bb,b); 909 VecGetArray_Fast(xx,x); 910 tmp = a->solve_work; 911 912 ierr = ISGetIndices(isrow,&r);CHKERRQ(ierr); 913 ierr = ISGetIndices(iscol,&c);CHKERRQ(ierr); c = c + (n-1); 914 915 /* forward solve the lower triangular */ 916 idx = 5*(*r++); 917 tmp[0] = b[idx]; tmp[1] = b[1+idx]; 918 tmp[2] = b[2+idx]; tmp[3] = b[3+idx]; tmp[4] = b[4+idx]; 919 for ( i=1; i<n; i++ ) { 920 v = aa + 25*ai[i]; 921 vi = aj + ai[i]; 922 nz = a->diag[i] - ai[i]; 923 idx = 5*(*r++); 924 sum1 = b[idx];sum2 = b[1+idx];sum3 = b[2+idx];sum4 = b[3+idx]; 925 sum5 = b[4+idx]; 926 while (nz--) { 927 idx = 5*(*vi++); 928 x1 = tmp[idx]; x2 = tmp[1+idx];x3 = tmp[2+idx]; 929 x4 = tmp[3+idx];x5 = tmp[4+idx]; 930 sum1 -= v[0]*x1 + v[5]*x2 + v[10]*x3 + v[15]*x4 + v[20]*x5; 931 sum2 -= v[1]*x1 + v[6]*x2 + v[11]*x3 + v[16]*x4 + v[21]*x5; 932 sum3 -= v[2]*x1 + v[7]*x2 + v[12]*x3 + v[17]*x4 + v[22]*x5; 933 sum4 -= v[3]*x1 + v[8]*x2 + v[13]*x3 + v[18]*x4 + v[23]*x5; 934 sum5 -= v[4]*x1 + v[9]*x2 + v[14]*x3 + v[19]*x4 + v[24]*x5; 935 v += 25; 936 } 937 idx = 5*i; 938 tmp[idx] = sum1;tmp[1+idx] = sum2; 939 tmp[2+idx] = sum3;tmp[3+idx] = sum4; tmp[4+idx] = sum5; 940 } 941 /* backward solve the upper triangular */ 942 for ( i=n-1; i>=0; i-- ){ 943 v = aa + 25*a->diag[i] + 25; 944 vi = aj + a->diag[i] + 1; 945 nz = ai[i+1] - a->diag[i] - 1; 946 idt = 5*i; 947 sum1 = tmp[idt]; sum2 = tmp[1+idt]; 948 sum3 = tmp[2+idt];sum4 = tmp[3+idt]; sum5 = tmp[4+idt]; 949 while (nz--) { 950 idx = 5*(*vi++); 951 x1 = tmp[idx]; x2 = tmp[1+idx]; 952 x3 = tmp[2+idx]; x4 = tmp[3+idx]; x5 = tmp[4+idx]; 953 sum1 -= v[0]*x1 + v[5]*x2 + v[10]*x3 + v[15]*x4 + v[20]*x5; 954 sum2 -= v[1]*x1 + v[6]*x2 + v[11]*x3 + v[16]*x4 + v[21]*x5; 955 sum3 -= v[2]*x1 + v[7]*x2 + v[12]*x3 + v[17]*x4 + v[22]*x5; 956 sum4 -= v[3]*x1 + v[8]*x2 + v[13]*x3 + v[18]*x4 + v[23]*x5; 957 sum5 -= v[4]*x1 + v[9]*x2 + v[14]*x3 + v[19]*x4 + v[24]*x5; 958 v += 25; 959 } 960 idc = 5*(*c--); 961 v = aa + 25*a->diag[i]; 962 x[idc] = tmp[idt] = v[0]*sum1+v[5]*sum2+v[10]*sum3+ 963 v[15]*sum4+v[20]*sum5; 964 x[1+idc] = tmp[1+idt] = v[1]*sum1+v[6]*sum2+v[11]*sum3+ 965 v[16]*sum4+v[21]*sum5; 966 x[2+idc] = tmp[2+idt] = v[2]*sum1+v[7]*sum2+v[12]*sum3+ 967 v[17]*sum4+v[22]*sum5; 968 x[3+idc] = tmp[3+idt] = v[3]*sum1+v[8]*sum2+v[13]*sum3+ 969 v[18]*sum4+v[23]*sum5; 970 x[4+idc] = tmp[4+idt] = v[4]*sum1+v[9]*sum2+v[14]*sum3+ 971 v[19]*sum4+v[24]*sum5; 972 } 973 974 ierr = ISRestoreIndices(isrow,&r); CHKERRQ(ierr); 975 ierr = ISRestoreIndices(iscol,&c); CHKERRQ(ierr); 976 VecRestoreArray_Fast(bb,b); 977 VecRestoreArray_Fast(xx,x); 978 PLogFlops(2*25*(a->nz) - a->n); 979 return 0; 980 } 981 982 int MatSolve_SeqBAIJ_4(Mat A,Vec bb,Vec xx) 983 { 984 Mat_SeqBAIJ *a=(Mat_SeqBAIJ *)A->data; 985 IS iscol=a->col,isrow=a->row; 986 int *r,*c,ierr,i,n=a->mbs,*vi,*ai=a->i,*aj=a->j,nz,idx,idt,idc; 987 Scalar *aa=a->a,sum1,sum2,sum3,sum4,x1,x2,x3,x4; 988 register Scalar *x,*b,*tmp,*v; 989 990 VecGetArray_Fast(bb,b); 991 VecGetArray_Fast(xx,x); 992 tmp = a->solve_work; 993 994 ierr = ISGetIndices(isrow,&r);CHKERRQ(ierr); 995 ierr = ISGetIndices(iscol,&c);CHKERRQ(ierr); c = c + (n-1); 996 997 /* forward solve the lower triangular */ 998 idx = 4*(*r++); 999 tmp[0] = b[idx]; tmp[1] = b[1+idx]; 1000 tmp[2] = b[2+idx]; tmp[3] = b[3+idx]; 1001 for ( i=1; i<n; i++ ) { 1002 v = aa + 16*ai[i]; 1003 vi = aj + ai[i]; 1004 nz = a->diag[i] - ai[i]; 1005 idx = 4*(*r++); 1006 sum1 = b[idx];sum2 = b[1+idx];sum3 = b[2+idx];sum4 = b[3+idx]; 1007 while (nz--) { 1008 idx = 4*(*vi++); 1009 x1 = tmp[idx];x2 = tmp[1+idx];x3 = tmp[2+idx];x4 = tmp[3+idx]; 1010 sum1 -= v[0]*x1 + v[4]*x2 + v[8]*x3 + v[12]*x4; 1011 sum2 -= v[1]*x1 + v[5]*x2 + v[9]*x3 + v[13]*x4; 1012 sum3 -= v[2]*x1 + v[6]*x2 + v[10]*x3 + v[14]*x4; 1013 sum4 -= v[3]*x1 + v[7]*x2 + v[11]*x3 + v[15]*x4; 1014 v += 16; 1015 } 1016 idx = 4*i; 1017 tmp[idx] = sum1;tmp[1+idx] = sum2; 1018 tmp[2+idx] = sum3;tmp[3+idx] = sum4; 1019 } 1020 /* backward solve the upper triangular */ 1021 for ( i=n-1; i>=0; i-- ){ 1022 v = aa + 16*a->diag[i] + 16; 1023 vi = aj + a->diag[i] + 1; 1024 nz = ai[i+1] - a->diag[i] - 1; 1025 idt = 4*i; 1026 sum1 = tmp[idt]; sum2 = tmp[1+idt]; 1027 sum3 = tmp[2+idt];sum4 = tmp[3+idt]; 1028 while (nz--) { 1029 idx = 4*(*vi++); 1030 x1 = tmp[idx]; x2 = tmp[1+idx]; 1031 x3 = tmp[2+idx]; x4 = tmp[3+idx]; 1032 sum1 -= v[0]*x1 + v[4]*x2 + v[8]*x3 + v[12]*x4; 1033 sum2 -= v[1]*x1 + v[5]*x2 + v[9]*x3 + v[13]*x4; 1034 sum3 -= v[2]*x1 + v[6]*x2 + v[10]*x3 + v[14]*x4; 1035 sum4 -= v[3]*x1 + v[7]*x2 + v[11]*x3 + v[15]*x4; 1036 v += 16; 1037 } 1038 idc = 4*(*c--); 1039 v = aa + 16*a->diag[i]; 1040 x[idc] = tmp[idt] = v[0]*sum1+v[4]*sum2+v[8]*sum3+v[12]*sum4; 1041 x[1+idc] = tmp[1+idt] = v[1]*sum1+v[5]*sum2+v[9]*sum3+v[13]*sum4; 1042 x[2+idc] = tmp[2+idt] = v[2]*sum1+v[6]*sum2+v[10]*sum3+v[14]*sum4; 1043 x[3+idc] = tmp[3+idt] = v[3]*sum1+v[7]*sum2+v[11]*sum3+v[15]*sum4; 1044 } 1045 1046 ierr = ISRestoreIndices(isrow,&r); CHKERRQ(ierr); 1047 ierr = ISRestoreIndices(iscol,&c); CHKERRQ(ierr); 1048 VecRestoreArray_Fast(bb,b); 1049 VecRestoreArray_Fast(xx,x); 1050 PLogFlops(2*16*(a->nz) - a->n); 1051 return 0; 1052 } 1053 1054 1055 int MatSolve_SeqBAIJ_3(Mat A,Vec bb,Vec xx) 1056 { 1057 Mat_SeqBAIJ *a=(Mat_SeqBAIJ *)A->data; 1058 IS iscol=a->col,isrow=a->row; 1059 int *r,*c,ierr,i,n=a->mbs,*vi,*ai=a->i,*aj=a->j,nz,idx,idt,idc; 1060 Scalar *aa=a->a,sum1,sum2,sum3,x1,x2,x3; 1061 register Scalar *x,*b,*tmp,*v; 1062 1063 VecGetArray_Fast(bb,b); 1064 VecGetArray_Fast(xx,x); 1065 tmp = a->solve_work; 1066 1067 ierr = ISGetIndices(isrow,&r);CHKERRQ(ierr); 1068 ierr = ISGetIndices(iscol,&c);CHKERRQ(ierr); c = c + (n-1); 1069 1070 /* forward solve the lower triangular */ 1071 idx = 3*(*r++); 1072 tmp[0] = b[idx]; tmp[1] = b[1+idx]; tmp[2] = b[2+idx]; 1073 for ( i=1; i<n; i++ ) { 1074 v = aa + 9*ai[i]; 1075 vi = aj + ai[i]; 1076 nz = a->diag[i] - ai[i]; 1077 idx = 3*(*r++); 1078 sum1 = b[idx]; sum2 = b[1+idx]; sum3 = b[2+idx]; 1079 while (nz--) { 1080 idx = 3*(*vi++); 1081 x1 = tmp[idx]; x2 = tmp[1+idx]; x3 = tmp[2+idx]; 1082 sum1 -= v[0]*x1 + v[3]*x2 + v[6]*x3; 1083 sum2 -= v[1]*x1 + v[4]*x2 + v[7]*x3; 1084 sum3 -= v[2]*x1 + v[5]*x2 + v[8]*x3; 1085 v += 9; 1086 } 1087 idx = 3*i; 1088 tmp[idx] = sum1; tmp[1+idx] = sum2; tmp[2+idx] = sum3; 1089 } 1090 /* backward solve the upper triangular */ 1091 for ( i=n-1; i>=0; i-- ){ 1092 v = aa + 9*a->diag[i] + 9; 1093 vi = aj + a->diag[i] + 1; 1094 nz = ai[i+1] - a->diag[i] - 1; 1095 idt = 3*i; 1096 sum1 = tmp[idt]; sum2 = tmp[1+idt]; sum3 = tmp[2+idt]; 1097 while (nz--) { 1098 idx = 3*(*vi++); 1099 x1 = tmp[idx]; x2 = tmp[1+idx]; x3 = tmp[2+idx]; 1100 sum1 -= v[0]*x1 + v[3]*x2 + v[6]*x3; 1101 sum2 -= v[1]*x1 + v[4]*x2 + v[7]*x3; 1102 sum3 -= v[2]*x1 + v[5]*x2 + v[8]*x3; 1103 v += 9; 1104 } 1105 idc = 3*(*c--); 1106 v = aa + 9*a->diag[i]; 1107 x[idc] = tmp[idt] = v[0]*sum1 + v[3]*sum2 + v[6]*sum3; 1108 x[1+idc] = tmp[1+idt] = v[1]*sum1 + v[4]*sum2 + v[7]*sum3; 1109 x[2+idc] = tmp[2+idt] = v[2]*sum1 + v[5]*sum2 + v[8]*sum3; 1110 } 1111 ierr = ISRestoreIndices(isrow,&r); CHKERRQ(ierr); 1112 ierr = ISRestoreIndices(iscol,&c); CHKERRQ(ierr); 1113 VecRestoreArray_Fast(bb,b); 1114 VecRestoreArray_Fast(xx,x); 1115 PLogFlops(2*9*(a->nz) - a->n); 1116 return 0; 1117 } 1118 1119 int MatSolve_SeqBAIJ_2(Mat A,Vec bb,Vec xx) 1120 { 1121 Mat_SeqBAIJ *a=(Mat_SeqBAIJ *)A->data; 1122 IS iscol=a->col,isrow=a->row; 1123 int *r,*c,ierr,i,n=a->mbs,*vi,*ai=a->i,*aj=a->j,nz,idx,idt,idc; 1124 Scalar *aa=a->a,sum1,sum2,x1,x2; 1125 register Scalar *x,*b,*tmp,*v; 1126 1127 VecGetArray_Fast(bb,b); 1128 VecGetArray_Fast(xx,x); 1129 tmp = a->solve_work; 1130 1131 ierr = ISGetIndices(isrow,&r);CHKERRQ(ierr); 1132 ierr = ISGetIndices(iscol,&c);CHKERRQ(ierr); c = c + (n-1); 1133 1134 /* forward solve the lower triangular */ 1135 idx = 2*(*r++); 1136 tmp[0] = b[idx]; tmp[1] = b[1+idx]; 1137 for ( i=1; i<n; i++ ) { 1138 v = aa + 4*ai[i]; 1139 vi = aj + ai[i]; 1140 nz = a->diag[i] - ai[i]; 1141 idx = 2*(*r++); 1142 sum1 = b[idx]; sum2 = b[1+idx]; 1143 while (nz--) { 1144 idx = 2*(*vi++); 1145 x1 = tmp[idx]; x2 = tmp[1+idx]; 1146 sum1 -= v[0]*x1 + v[2]*x2; 1147 sum2 -= v[1]*x1 + v[3]*x2; 1148 v += 4; 1149 } 1150 idx = 2*i; 1151 tmp[idx] = sum1; tmp[1+idx] = sum2; 1152 } 1153 /* backward solve the upper triangular */ 1154 for ( i=n-1; i>=0; i-- ){ 1155 v = aa + 4*a->diag[i] + 4; 1156 vi = aj + a->diag[i] + 1; 1157 nz = ai[i+1] - a->diag[i] - 1; 1158 idt = 2*i; 1159 sum1 = tmp[idt]; sum2 = tmp[1+idt]; 1160 while (nz--) { 1161 idx = 2*(*vi++); 1162 x1 = tmp[idx]; x2 = tmp[1+idx]; 1163 sum1 -= v[0]*x1 + v[2]*x2; 1164 sum2 -= v[1]*x1 + v[3]*x2; 1165 v += 4; 1166 } 1167 idc = 2*(*c--); 1168 v = aa + 4*a->diag[i]; 1169 x[idc] = tmp[idt] = v[0]*sum1 + v[2]*sum2; 1170 x[1+idc] = tmp[1+idt] = v[1]*sum1 + v[3]*sum2; 1171 } 1172 ierr = ISRestoreIndices(isrow,&r); CHKERRQ(ierr); 1173 ierr = ISRestoreIndices(iscol,&c); CHKERRQ(ierr); 1174 VecRestoreArray_Fast(bb,b); 1175 VecRestoreArray_Fast(xx,x); 1176 PLogFlops(2*4*(a->nz) - a->n); 1177 return 0; 1178 } 1179 1180 1181 int MatSolve_SeqBAIJ_1(Mat A,Vec bb,Vec xx) 1182 { 1183 Mat_SeqBAIJ *a=(Mat_SeqBAIJ *)A->data; 1184 IS iscol=a->col,isrow=a->row; 1185 int *r,*c,ierr,i,n=a->mbs,*vi,*ai=a->i,*aj=a->j,nz; 1186 Scalar *aa=a->a,sum1; 1187 register Scalar *x,*b,*tmp,*v; 1188 1189 VecGetArray_Fast(bb,b); 1190 VecGetArray_Fast(xx,x); 1191 tmp = a->solve_work; 1192 1193 ierr = ISGetIndices(isrow,&r);CHKERRQ(ierr); 1194 ierr = ISGetIndices(iscol,&c);CHKERRQ(ierr); c = c + (n-1); 1195 1196 /* forward solve the lower triangular */ 1197 tmp[0] = b[*r++]; 1198 for ( i=1; i<n; i++ ) { 1199 v = aa + ai[i]; 1200 vi = aj + ai[i]; 1201 nz = a->diag[i] - ai[i]; 1202 sum1 = b[*r++]; 1203 while (nz--) { 1204 sum1 -= (*v++)*tmp[*vi++]; 1205 } 1206 tmp[i] = sum1; 1207 } 1208 /* backward solve the upper triangular */ 1209 for ( i=n-1; i>=0; i-- ){ 1210 v = aa + a->diag[i] + 1; 1211 vi = aj + a->diag[i] + 1; 1212 nz = ai[i+1] - a->diag[i] - 1; 1213 sum1 = tmp[i]; 1214 while (nz--) { 1215 sum1 -= (*v++)*tmp[*vi++]; 1216 } 1217 x[*c--] = tmp[i] = aa[a->diag[i]]*sum1; 1218 } 1219 1220 ierr = ISRestoreIndices(isrow,&r); CHKERRQ(ierr); 1221 ierr = ISRestoreIndices(iscol,&c); CHKERRQ(ierr); 1222 VecRestoreArray_Fast(bb,b); 1223 VecRestoreArray_Fast(xx,x); 1224 PLogFlops(2*1*(a->nz) - a->n); 1225 return 0; 1226 } 1227 1228 /* ----------------------------------------------------------------*/ 1229 /* 1230 This code is virtually identical to MatILUFactorSymbolic_SeqAIJ 1231 except that the data structure of Mat_SeqAIJ is slightly different. 1232 Not a good example of code reuse. 1233 */ 1234 int MatILUFactorSymbolic_SeqBAIJ(Mat A,IS isrow,IS iscol,double f,int levels, 1235 Mat *fact) 1236 { 1237 Mat_SeqBAIJ *a = (Mat_SeqBAIJ *) A->data, *b; 1238 IS isicol; 1239 int *r,*ic, ierr, prow, n = a->mbs, *ai = a->i, *aj = a->j; 1240 int *ainew,*ajnew, jmax,*fill, *xi, nz, *im,*ajfill,*flev; 1241 int *dloc, idx, row,m,fm, nzf, nzi,len, realloc = 0; 1242 int incrlev,nnz,i,bs = a->bs,bs2 = a->bs2; 1243 PetscTruth col_identity, row_identity; 1244 1245 /* special case that simply copies fill pattern */ 1246 PetscValidHeaderSpecific(isrow,IS_COOKIE); 1247 PetscValidHeaderSpecific(iscol,IS_COOKIE); 1248 ISIdentity(isrow,&row_identity); ISIdentity(iscol,&col_identity); 1249 if (levels == 0 && row_identity && col_identity) { 1250 ierr = MatConvertSameType_SeqBAIJ(A,fact,DO_NOT_COPY_VALUES); CHKERRQ(ierr); 1251 (*fact)->factor = FACTOR_LU; 1252 b = (Mat_SeqBAIJ *) (*fact)->data; 1253 if (!b->diag) { 1254 ierr = MatMarkDiag_SeqBAIJ(*fact); CHKERRQ(ierr); 1255 } 1256 b->row = isrow; 1257 b->col = iscol; 1258 b->solve_work = (Scalar *) PetscMalloc((b->m+1+b->bs)*sizeof(Scalar));CHKPTRQ(b->solve_work); 1259 return 0; 1260 } 1261 1262 ierr = ISInvertPermutation(iscol,&isicol); CHKERRQ(ierr); 1263 ierr = ISGetIndices(isrow,&r); CHKERRQ(ierr); 1264 ierr = ISGetIndices(isicol,&ic); CHKERRQ(ierr); 1265 1266 /* get new row pointers */ 1267 ainew = (int *) PetscMalloc( (n+1)*sizeof(int) ); CHKPTRQ(ainew); 1268 ainew[0] = 0; 1269 /* don't know how many column pointers are needed so estimate */ 1270 jmax = (int) (f*ai[n] + 1); 1271 ajnew = (int *) PetscMalloc( (jmax)*sizeof(int) ); CHKPTRQ(ajnew); 1272 /* ajfill is level of fill for each fill entry */ 1273 ajfill = (int *) PetscMalloc( (jmax)*sizeof(int) ); CHKPTRQ(ajfill); 1274 /* fill is a linked list of nonzeros in active row */ 1275 fill = (int *) PetscMalloc( (n+1)*sizeof(int)); CHKPTRQ(fill); 1276 /* im is level for each filled value */ 1277 im = (int *) PetscMalloc( (n+1)*sizeof(int)); CHKPTRQ(im); 1278 /* dloc is location of diagonal in factor */ 1279 dloc = (int *) PetscMalloc( (n+1)*sizeof(int)); CHKPTRQ(dloc); 1280 dloc[0] = 0; 1281 for ( prow=0; prow<n; prow++ ) { 1282 /* first copy previous fill into linked list */ 1283 nzf = nz = ai[r[prow]+1] - ai[r[prow]]; 1284 xi = aj + ai[r[prow]]; 1285 fill[n] = n; 1286 while (nz--) { 1287 fm = n; 1288 idx = ic[*xi++]; 1289 do { 1290 m = fm; 1291 fm = fill[m]; 1292 } while (fm < idx); 1293 fill[m] = idx; 1294 fill[idx] = fm; 1295 im[idx] = 0; 1296 } 1297 nzi = 0; 1298 row = fill[n]; 1299 while ( row < prow ) { 1300 incrlev = im[row] + 1; 1301 nz = dloc[row]; 1302 xi = ajnew + ainew[row] + nz; 1303 flev = ajfill + ainew[row] + nz + 1; 1304 nnz = ainew[row+1] - ainew[row] - nz - 1; 1305 if (*xi++ != row) { 1306 SETERRQ(1,"MatILUFactorSymbolic_SeqBAIJ:zero pivot"); 1307 } 1308 fm = row; 1309 while (nnz-- > 0) { 1310 idx = *xi++; 1311 if (*flev + incrlev > levels) { 1312 flev++; 1313 continue; 1314 } 1315 do { 1316 m = fm; 1317 fm = fill[m]; 1318 } while (fm < idx); 1319 if (fm != idx) { 1320 im[idx] = *flev + incrlev; 1321 fill[m] = idx; 1322 fill[idx] = fm; 1323 fm = idx; 1324 nzf++; 1325 } 1326 else { 1327 if (im[idx] > *flev + incrlev) im[idx] = *flev+incrlev; 1328 } 1329 flev++; 1330 } 1331 row = fill[row]; 1332 nzi++; 1333 } 1334 /* copy new filled row into permanent storage */ 1335 ainew[prow+1] = ainew[prow] + nzf; 1336 if (ainew[prow+1] > jmax) { 1337 /* allocate a longer ajnew */ 1338 int maxadd; 1339 maxadd = (int) (((f*ai[n]+1)*(n-prow+5))/n); 1340 if (maxadd < nzf) maxadd = (n-prow)*(nzf+1); 1341 jmax += maxadd; 1342 xi = (int *) PetscMalloc( jmax*sizeof(int) );CHKPTRQ(xi); 1343 PetscMemcpy(xi,ajnew,ainew[prow]*sizeof(int)); 1344 PetscFree(ajnew); 1345 ajnew = xi; 1346 /* allocate a longer ajfill */ 1347 xi = (int *) PetscMalloc( jmax*sizeof(int) );CHKPTRQ(xi); 1348 PetscMemcpy(xi,ajfill,ainew[prow]*sizeof(int)); 1349 PetscFree(ajfill); 1350 ajfill = xi; 1351 realloc++; 1352 } 1353 xi = ajnew + ainew[prow]; 1354 flev = ajfill + ainew[prow]; 1355 dloc[prow] = nzi; 1356 fm = fill[n]; 1357 while (nzf--) { 1358 *xi++ = fm; 1359 *flev++ = im[fm]; 1360 fm = fill[fm]; 1361 } 1362 } 1363 PetscFree(ajfill); 1364 ierr = ISRestoreIndices(isrow,&r); CHKERRQ(ierr); 1365 ierr = ISRestoreIndices(isicol,&ic); CHKERRQ(ierr); 1366 ierr = ISDestroy(isicol); CHKERRQ(ierr); 1367 PetscFree(fill); PetscFree(im); 1368 1369 PLogInfo(A, 1370 "Info:MatILUFactorSymbolic_SeqBAIJ:Realloc %d Fill ratio:given %g needed %g\n", 1371 realloc,f,((double)ainew[n])/((double)ai[prow])); 1372 1373 /* put together the new matrix */ 1374 ierr = MatCreateSeqBAIJ(A->comm,bs,bs*n,bs*n,0,PETSC_NULL,fact);CHKERRQ(ierr); 1375 b = (Mat_SeqBAIJ *) (*fact)->data; 1376 PetscFree(b->imax); 1377 b->singlemalloc = 0; 1378 len = bs2*ainew[n]*sizeof(Scalar); 1379 /* the next line frees the default space generated by the Create() */ 1380 PetscFree(b->a); PetscFree(b->ilen); 1381 b->a = (Scalar *) PetscMalloc( len ); CHKPTRQ(b->a); 1382 b->j = ajnew; 1383 b->i = ainew; 1384 for ( i=0; i<n; i++ ) dloc[i] += ainew[i]; 1385 b->diag = dloc; 1386 b->ilen = 0; 1387 b->imax = 0; 1388 b->row = isrow; 1389 b->col = iscol; 1390 b->solve_work = (Scalar *) PetscMalloc( (bs*n+bs)*sizeof(Scalar)); 1391 CHKPTRQ(b->solve_work); 1392 /* In b structure: Free imax, ilen, old a, old j. 1393 Allocate dloc, solve_work, new a, new j */ 1394 PLogObjectMemory(*fact,(ainew[n]-n)*(sizeof(int))+bs2*ainew[n]*sizeof(Scalar)); 1395 b->maxnz = b->nz = ainew[n]; 1396 (*fact)->factor = FACTOR_LU; 1397 1398 (*fact)->info.factor_mallocs = realloc; 1399 (*fact)->info.fill_ratio_given = f; 1400 (*fact)->info.fill_ratio_needed = ((double)ainew[n])/((double)ai[prow]); 1401 1402 return 0; 1403 } 1404 1405 1406 1407 1408