Actual source code: baijsolvtran4.c
1: #include <../src/mat/impls/baij/seq/baij.h>
2: #include <petsc/private/kernels/blockinvert.h>
4: PetscErrorCode MatSolveTranspose_SeqBAIJ_4_inplace(Mat A, Vec bb, Vec xx)
5: {
6: Mat_SeqBAIJ *a = (Mat_SeqBAIJ *)A->data;
7: IS iscol = a->col, isrow = a->row;
8: const PetscInt *r, *c, *rout, *cout;
9: const PetscInt *diag = a->diag, n = a->mbs, *vi, *ai = a->i, *aj = a->j;
10: PetscInt i, nz, idx, idt, ii, ic, ir, oidx;
11: const MatScalar *aa = a->a, *v;
12: PetscScalar s1, s2, s3, s4, x1, x2, x3, x4, *x, *t;
13: const PetscScalar *b;
15: VecGetArrayRead(bb, &b);
16: VecGetArray(xx, &x);
17: t = a->solve_work;
19: ISGetIndices(isrow, &rout);
20: r = rout;
21: ISGetIndices(iscol, &cout);
22: c = cout;
24: /* copy the b into temp work space according to permutation */
25: ii = 0;
26: for (i = 0; i < n; i++) {
27: ic = 4 * c[i];
28: t[ii] = b[ic];
29: t[ii + 1] = b[ic + 1];
30: t[ii + 2] = b[ic + 2];
31: t[ii + 3] = b[ic + 3];
32: ii += 4;
33: }
35: /* forward solve the U^T */
36: idx = 0;
37: for (i = 0; i < n; i++) {
38: v = aa + 16 * diag[i];
39: /* multiply by the inverse of the block diagonal */
40: x1 = t[idx];
41: x2 = t[1 + idx];
42: x3 = t[2 + idx];
43: x4 = t[3 + idx];
44: s1 = v[0] * x1 + v[1] * x2 + v[2] * x3 + v[3] * x4;
45: s2 = v[4] * x1 + v[5] * x2 + v[6] * x3 + v[7] * x4;
46: s3 = v[8] * x1 + v[9] * x2 + v[10] * x3 + v[11] * x4;
47: s4 = v[12] * x1 + v[13] * x2 + v[14] * x3 + v[15] * x4;
48: v += 16;
50: vi = aj + diag[i] + 1;
51: nz = ai[i + 1] - diag[i] - 1;
52: while (nz--) {
53: oidx = 4 * (*vi++);
54: t[oidx] -= v[0] * s1 + v[1] * s2 + v[2] * s3 + v[3] * s4;
55: t[oidx + 1] -= v[4] * s1 + v[5] * s2 + v[6] * s3 + v[7] * s4;
56: t[oidx + 2] -= v[8] * s1 + v[9] * s2 + v[10] * s3 + v[11] * s4;
57: t[oidx + 3] -= v[12] * s1 + v[13] * s2 + v[14] * s3 + v[15] * s4;
58: v += 16;
59: }
60: t[idx] = s1;
61: t[1 + idx] = s2;
62: t[2 + idx] = s3;
63: t[3 + idx] = s4;
64: idx += 4;
65: }
66: /* backward solve the L^T */
67: for (i = n - 1; i >= 0; i--) {
68: v = aa + 16 * diag[i] - 16;
69: vi = aj + diag[i] - 1;
70: nz = diag[i] - ai[i];
71: idt = 4 * i;
72: s1 = t[idt];
73: s2 = t[1 + idt];
74: s3 = t[2 + idt];
75: s4 = t[3 + idt];
76: while (nz--) {
77: idx = 4 * (*vi--);
78: t[idx] -= v[0] * s1 + v[1] * s2 + v[2] * s3 + v[3] * s4;
79: t[idx + 1] -= v[4] * s1 + v[5] * s2 + v[6] * s3 + v[7] * s4;
80: t[idx + 2] -= v[8] * s1 + v[9] * s2 + v[10] * s3 + v[11] * s4;
81: t[idx + 3] -= v[12] * s1 + v[13] * s2 + v[14] * s3 + v[15] * s4;
82: v -= 16;
83: }
84: }
86: /* copy t into x according to permutation */
87: ii = 0;
88: for (i = 0; i < n; i++) {
89: ir = 4 * r[i];
90: x[ir] = t[ii];
91: x[ir + 1] = t[ii + 1];
92: x[ir + 2] = t[ii + 2];
93: x[ir + 3] = t[ii + 3];
94: ii += 4;
95: }
97: ISRestoreIndices(isrow, &rout);
98: ISRestoreIndices(iscol, &cout);
99: VecRestoreArrayRead(bb, &b);
100: VecRestoreArray(xx, &x);
101: PetscLogFlops(2.0 * 16 * (a->nz) - 4.0 * A->cmap->n);
102: return 0;
103: }
105: PetscErrorCode MatSolveTranspose_SeqBAIJ_4(Mat A, Vec bb, Vec xx)
106: {
107: Mat_SeqBAIJ *a = (Mat_SeqBAIJ *)A->data;
108: IS iscol = a->col, isrow = a->row;
109: const PetscInt n = a->mbs, *vi, *ai = a->i, *aj = a->j, *diag = a->diag;
110: const PetscInt *r, *c, *rout, *cout;
111: PetscInt nz, idx, idt, j, i, oidx, ii, ic, ir;
112: const PetscInt bs = A->rmap->bs, bs2 = a->bs2;
113: const MatScalar *aa = a->a, *v;
114: PetscScalar s1, s2, s3, s4, x1, x2, x3, x4, *x, *t;
115: const PetscScalar *b;
117: VecGetArrayRead(bb, &b);
118: VecGetArray(xx, &x);
119: t = a->solve_work;
121: ISGetIndices(isrow, &rout);
122: r = rout;
123: ISGetIndices(iscol, &cout);
124: c = cout;
126: /* copy b into temp work space according to permutation */
127: for (i = 0; i < n; i++) {
128: ii = bs * i;
129: ic = bs * c[i];
130: t[ii] = b[ic];
131: t[ii + 1] = b[ic + 1];
132: t[ii + 2] = b[ic + 2];
133: t[ii + 3] = b[ic + 3];
134: }
136: /* forward solve the U^T */
137: idx = 0;
138: for (i = 0; i < n; i++) {
139: v = aa + bs2 * diag[i];
140: /* multiply by the inverse of the block diagonal */
141: x1 = t[idx];
142: x2 = t[1 + idx];
143: x3 = t[2 + idx];
144: x4 = t[3 + idx];
145: s1 = v[0] * x1 + v[1] * x2 + v[2] * x3 + v[3] * x4;
146: s2 = v[4] * x1 + v[5] * x2 + v[6] * x3 + v[7] * x4;
147: s3 = v[8] * x1 + v[9] * x2 + v[10] * x3 + v[11] * x4;
148: s4 = v[12] * x1 + v[13] * x2 + v[14] * x3 + v[15] * x4;
149: v -= bs2;
151: vi = aj + diag[i] - 1;
152: nz = diag[i] - diag[i + 1] - 1;
153: for (j = 0; j > -nz; j--) {
154: oidx = bs * vi[j];
155: t[oidx] -= v[0] * s1 + v[1] * s2 + v[2] * s3 + v[3] * s4;
156: t[oidx + 1] -= v[4] * s1 + v[5] * s2 + v[6] * s3 + v[7] * s4;
157: t[oidx + 2] -= v[8] * s1 + v[9] * s2 + v[10] * s3 + v[11] * s4;
158: t[oidx + 3] -= v[12] * s1 + v[13] * s2 + v[14] * s3 + v[15] * s4;
159: v -= bs2;
160: }
161: t[idx] = s1;
162: t[1 + idx] = s2;
163: t[2 + idx] = s3;
164: t[3 + idx] = s4;
165: idx += bs;
166: }
167: /* backward solve the L^T */
168: for (i = n - 1; i >= 0; i--) {
169: v = aa + bs2 * ai[i];
170: vi = aj + ai[i];
171: nz = ai[i + 1] - ai[i];
172: idt = bs * i;
173: s1 = t[idt];
174: s2 = t[1 + idt];
175: s3 = t[2 + idt];
176: s4 = t[3 + idt];
177: for (j = 0; j < nz; j++) {
178: idx = bs * vi[j];
179: t[idx] -= v[0] * s1 + v[1] * s2 + v[2] * s3 + v[3] * s4;
180: t[idx + 1] -= v[4] * s1 + v[5] * s2 + v[6] * s3 + v[7] * s4;
181: t[idx + 2] -= v[8] * s1 + v[9] * s2 + v[10] * s3 + v[11] * s4;
182: t[idx + 3] -= v[12] * s1 + v[13] * s2 + v[14] * s3 + v[15] * s4;
183: v += bs2;
184: }
185: }
187: /* copy t into x according to permutation */
188: for (i = 0; i < n; i++) {
189: ii = bs * i;
190: ir = bs * r[i];
191: x[ir] = t[ii];
192: x[ir + 1] = t[ii + 1];
193: x[ir + 2] = t[ii + 2];
194: x[ir + 3] = t[ii + 3];
195: }
197: ISRestoreIndices(isrow, &rout);
198: ISRestoreIndices(iscol, &cout);
199: VecRestoreArrayRead(bb, &b);
200: VecRestoreArray(xx, &x);
201: PetscLogFlops(2.0 * bs2 * (a->nz) - bs * A->cmap->n);
202: return 0;
203: }