1 /* 2 Factorization code for BAIJ format. 3 */ 4 #include "src/mat/impls/baij/seq/baij.h" 5 #include "src/inline/ilu.h" 6 7 /* ----------------------------------------------------------- */ 8 #undef __FUNCT__ 9 #define __FUNCT__ "MatLUFactorNumeric_SeqBAIJ_N" 10 PetscErrorCode MatLUFactorNumeric_SeqBAIJ_N(Mat A,Mat *B) 11 { 12 Mat C = *B; 13 Mat_SeqBAIJ *a = (Mat_SeqBAIJ*)A->data,*b = (Mat_SeqBAIJ *)C->data; 14 IS isrow = b->row,isicol = b->icol; 15 int *r,*ic,ierr,i,j,n = a->mbs,*bi = b->i,*bj = b->j; 16 int *ajtmpold,*ajtmp,nz,row,bslog,*ai=a->i,*aj=a->j,k,flg; 17 int *diag_offset=b->diag,diag,bs=a->bs,bs2 = a->bs2,*v_pivots,*pj; 18 MatScalar *ba = b->a,*aa = a->a,*pv,*v,*rtmp,*multiplier,*v_work,*pc,*w; 19 20 PetscFunctionBegin; 21 ierr = ISGetIndices(isrow,&r);CHKERRQ(ierr); 22 ierr = ISGetIndices(isicol,&ic);CHKERRQ(ierr); 23 ierr = PetscMalloc(bs2*(n+1)*sizeof(MatScalar),&rtmp);CHKERRQ(ierr); 24 ierr = PetscMemzero(rtmp,bs2*(n+1)*sizeof(MatScalar));CHKERRQ(ierr); 25 /* generate work space needed by dense LU factorization */ 26 ierr = PetscMalloc(bs*sizeof(int) + (bs+bs2)*sizeof(MatScalar),&v_work);CHKERRQ(ierr); 27 multiplier = v_work + bs; 28 v_pivots = (int*)(multiplier + bs2); 29 30 /* flops in while loop */ 31 bslog = 2*bs*bs2; 32 33 for (i=0; i<n; i++) { 34 nz = bi[i+1] - bi[i]; 35 ajtmp = bj + bi[i]; 36 for (j=0; j<nz; j++) { 37 ierr = PetscMemzero(rtmp+bs2*ajtmp[j],bs2*sizeof(MatScalar));CHKERRQ(ierr); 38 } 39 /* load in initial (unfactored row) */ 40 nz = ai[r[i]+1] - ai[r[i]]; 41 ajtmpold = aj + ai[r[i]]; 42 v = aa + bs2*ai[r[i]]; 43 for (j=0; j<nz; j++) { 44 ierr = PetscMemcpy(rtmp+bs2*ic[ajtmpold[j]],v+bs2*j,bs2*sizeof(MatScalar));CHKERRQ(ierr); 45 } 46 row = *ajtmp++; 47 while (row < i) { 48 pc = rtmp + bs2*row; 49 /* if (*pc) { */ 50 for (flg=0,k=0; k<bs2; k++) { if (pc[k]!=0.0) { flg = 1; break; }} 51 if (flg) { 52 pv = ba + bs2*diag_offset[row]; 53 pj = bj + diag_offset[row] + 1; 54 Kernel_A_gets_A_times_B(bs,pc,pv,multiplier); 55 nz = bi[row+1] - diag_offset[row] - 1; 56 pv += bs2; 57 for (j=0; j<nz; j++) { 58 Kernel_A_gets_A_minus_B_times_C(bs,rtmp+bs2*pj[j],pc,pv+bs2*j); 59 } 60 PetscLogFlops(bslog*(nz+1)-bs); 61 } 62 row = *ajtmp++; 63 } 64 /* finished row so stick it into b->a */ 65 pv = ba + bs2*bi[i]; 66 pj = bj + bi[i]; 67 nz = bi[i+1] - bi[i]; 68 for (j=0; j<nz; j++) { 69 ierr = PetscMemcpy(pv+bs2*j,rtmp+bs2*pj[j],bs2*sizeof(MatScalar));CHKERRQ(ierr); 70 } 71 diag = diag_offset[i] - bi[i]; 72 /* invert diagonal block */ 73 w = pv + bs2*diag; 74 Kernel_A_gets_inverse_A(bs,w,v_pivots,v_work); 75 } 76 77 ierr = PetscFree(rtmp);CHKERRQ(ierr); 78 ierr = PetscFree(v_work);CHKERRQ(ierr); 79 ierr = ISRestoreIndices(isicol,&ic);CHKERRQ(ierr); 80 ierr = ISRestoreIndices(isrow,&r);CHKERRQ(ierr); 81 C->factor = FACTOR_LU; 82 C->assembled = PETSC_TRUE; 83 PetscLogFlops(1.3333*bs*bs2*b->mbs); /* from inverting diagonal blocks */ 84 PetscFunctionReturn(0); 85 } 86