Actual source code: baij.c
petsc-3.5.1 2014-07-24
2: /*
3: Defines the basic matrix operations for the BAIJ (compressed row)
4: matrix storage format.
5: */
6: #include <../src/mat/impls/baij/seq/baij.h> /*I "petscmat.h" I*/
7: #include <petscblaslapack.h>
8: #include <petsc-private/kernels/blockinvert.h>
9: #include <petsc-private/kernels/blockmatmult.h>
13: PetscErrorCode MatInvertBlockDiagonal_SeqBAIJ(Mat A,const PetscScalar **values)
14: {
15: Mat_SeqBAIJ *a = (Mat_SeqBAIJ*) A->data;
17: PetscInt *diag_offset,i,bs = A->rmap->bs,mbs = a->mbs,ipvt[5],bs2 = bs*bs,*v_pivots;
18: MatScalar *v = a->a,*odiag,*diag,*mdiag,work[25],*v_work;
19: PetscReal shift = 0.0;
22: if (a->idiagvalid) {
23: if (values) *values = a->idiag;
24: return(0);
25: }
26: MatMarkDiagonal_SeqBAIJ(A);
27: diag_offset = a->diag;
28: if (!a->idiag) {
29: PetscMalloc1(2*bs2*mbs,&a->idiag);
30: PetscLogObjectMemory((PetscObject)A,2*bs2*mbs*sizeof(PetscScalar));
31: }
32: diag = a->idiag;
33: mdiag = a->idiag+bs2*mbs;
34: if (values) *values = a->idiag;
35: /* factor and invert each block */
36: switch (bs) {
37: case 1:
38: for (i=0; i<mbs; i++) {
39: odiag = v + 1*diag_offset[i];
40: diag[0] = odiag[0];
41: mdiag[0] = odiag[0];
42: diag[0] = (PetscScalar)1.0 / (diag[0] + shift);
43: diag += 1;
44: mdiag += 1;
45: }
46: break;
47: case 2:
48: for (i=0; i<mbs; i++) {
49: odiag = v + 4*diag_offset[i];
50: diag[0] = odiag[0]; diag[1] = odiag[1]; diag[2] = odiag[2]; diag[3] = odiag[3];
51: mdiag[0] = odiag[0]; mdiag[1] = odiag[1]; mdiag[2] = odiag[2]; mdiag[3] = odiag[3];
52: PetscKernel_A_gets_inverse_A_2(diag,shift);
53: diag += 4;
54: mdiag += 4;
55: }
56: break;
57: case 3:
58: for (i=0; i<mbs; i++) {
59: odiag = v + 9*diag_offset[i];
60: diag[0] = odiag[0]; diag[1] = odiag[1]; diag[2] = odiag[2]; diag[3] = odiag[3];
61: diag[4] = odiag[4]; diag[5] = odiag[5]; diag[6] = odiag[6]; diag[7] = odiag[7];
62: diag[8] = odiag[8];
63: mdiag[0] = odiag[0]; mdiag[1] = odiag[1]; mdiag[2] = odiag[2]; mdiag[3] = odiag[3];
64: mdiag[4] = odiag[4]; mdiag[5] = odiag[5]; mdiag[6] = odiag[6]; mdiag[7] = odiag[7];
65: mdiag[8] = odiag[8];
66: PetscKernel_A_gets_inverse_A_3(diag,shift);
67: diag += 9;
68: mdiag += 9;
69: }
70: break;
71: case 4:
72: for (i=0; i<mbs; i++) {
73: odiag = v + 16*diag_offset[i];
74: PetscMemcpy(diag,odiag,16*sizeof(PetscScalar));
75: PetscMemcpy(mdiag,odiag,16*sizeof(PetscScalar));
76: PetscKernel_A_gets_inverse_A_4(diag,shift);
77: diag += 16;
78: mdiag += 16;
79: }
80: break;
81: case 5:
82: for (i=0; i<mbs; i++) {
83: odiag = v + 25*diag_offset[i];
84: PetscMemcpy(diag,odiag,25*sizeof(PetscScalar));
85: PetscMemcpy(mdiag,odiag,25*sizeof(PetscScalar));
86: PetscKernel_A_gets_inverse_A_5(diag,ipvt,work,shift);
87: diag += 25;
88: mdiag += 25;
89: }
90: break;
91: case 6:
92: for (i=0; i<mbs; i++) {
93: odiag = v + 36*diag_offset[i];
94: PetscMemcpy(diag,odiag,36*sizeof(PetscScalar));
95: PetscMemcpy(mdiag,odiag,36*sizeof(PetscScalar));
96: PetscKernel_A_gets_inverse_A_6(diag,shift);
97: diag += 36;
98: mdiag += 36;
99: }
100: break;
101: case 7:
102: for (i=0; i<mbs; i++) {
103: odiag = v + 49*diag_offset[i];
104: PetscMemcpy(diag,odiag,49*sizeof(PetscScalar));
105: PetscMemcpy(mdiag,odiag,49*sizeof(PetscScalar));
106: PetscKernel_A_gets_inverse_A_7(diag,shift);
107: diag += 49;
108: mdiag += 49;
109: }
110: break;
111: default:
112: PetscMalloc2(bs,&v_work,bs,&v_pivots);
113: for (i=0; i<mbs; i++) {
114: odiag = v + bs2*diag_offset[i];
115: PetscMemcpy(diag,odiag,bs2*sizeof(PetscScalar));
116: PetscMemcpy(mdiag,odiag,bs2*sizeof(PetscScalar));
117: PetscKernel_A_gets_inverse_A(bs,diag,v_pivots,v_work);
118: diag += bs2;
119: mdiag += bs2;
120: }
121: PetscFree2(v_work,v_pivots);
122: }
123: a->idiagvalid = PETSC_TRUE;
124: return(0);
125: }
129: PetscErrorCode MatSOR_SeqBAIJ(Mat A,Vec bb,PetscReal omega,MatSORType flag,PetscReal fshift,PetscInt its,PetscInt lits,Vec xx)
130: {
131: Mat_SeqBAIJ *a = (Mat_SeqBAIJ*)A->data;
132: PetscScalar *x,*work,*w,*workt,*t;
133: const MatScalar *v,*aa = a->a, *idiag;
134: const PetscScalar *b,*xb;
135: PetscScalar s[7], xw[7]={0}; /* avoid some compilers thinking xw is uninitialized */
136: PetscErrorCode ierr;
137: PetscInt m = a->mbs,i,i2,nz,bs = A->rmap->bs,bs2 = bs*bs,k,j,idx,it;
138: const PetscInt *diag,*ai = a->i,*aj = a->j,*vi;
141: its = its*lits;
142: if (flag & SOR_EISENSTAT) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"No support yet for Eisenstat");
143: if (its <= 0) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Relaxation requires global its %D and local its %D both positive",its,lits);
144: if (fshift) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Sorry, no support for diagonal shift");
145: if (omega != 1.0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Sorry, no support for non-trivial relaxation factor");
146: if ((flag & SOR_APPLY_UPPER) || (flag & SOR_APPLY_LOWER)) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Sorry, no support for applying upper or lower triangular parts");
148: if (!a->idiagvalid) {MatInvertBlockDiagonal(A,NULL);}
150: if (!m) return(0);
151: diag = a->diag;
152: idiag = a->idiag;
153: k = PetscMax(A->rmap->n,A->cmap->n);
154: if (!a->mult_work) {
155: PetscMalloc1((2*k+1),&a->mult_work);
156: }
157: work = a->mult_work;
158: t = work + k+1;
159: if (!a->sor_work) {
160: PetscMalloc1(bs,&a->sor_work);
161: }
162: w = a->sor_work;
164: VecGetArray(xx,&x);
165: VecGetArrayRead(bb,&b);
167: if (flag & SOR_ZERO_INITIAL_GUESS) {
168: if (flag & SOR_FORWARD_SWEEP || flag & SOR_LOCAL_FORWARD_SWEEP) {
169: switch (bs) {
170: case 1:
171: PetscKernel_v_gets_A_times_w_1(x,idiag,b);
172: t[0] = b[0];
173: i2 = 1;
174: idiag += 1;
175: for (i=1; i<m; i++) {
176: v = aa + ai[i];
177: vi = aj + ai[i];
178: nz = diag[i] - ai[i];
179: s[0] = b[i2];
180: for (j=0; j<nz; j++) {
181: xw[0] = x[vi[j]];
182: PetscKernel_v_gets_v_minus_A_times_w_1(s,(v+j),xw);
183: }
184: t[i2] = s[0];
185: PetscKernel_v_gets_A_times_w_1(xw,idiag,s);
186: x[i2] = xw[0];
187: idiag += 1;
188: i2 += 1;
189: }
190: break;
191: case 2:
192: PetscKernel_v_gets_A_times_w_2(x,idiag,b);
193: t[0] = b[0]; t[1] = b[1];
194: i2 = 2;
195: idiag += 4;
196: for (i=1; i<m; i++) {
197: v = aa + 4*ai[i];
198: vi = aj + ai[i];
199: nz = diag[i] - ai[i];
200: s[0] = b[i2]; s[1] = b[i2+1];
201: for (j=0; j<nz; j++) {
202: idx = 2*vi[j];
203: it = 4*j;
204: xw[0] = x[idx]; xw[1] = x[1+idx];
205: PetscKernel_v_gets_v_minus_A_times_w_2(s,(v+it),xw);
206: }
207: t[i2] = s[0]; t[i2+1] = s[1];
208: PetscKernel_v_gets_A_times_w_2(xw,idiag,s);
209: x[i2] = xw[0]; x[i2+1] = xw[1];
210: idiag += 4;
211: i2 += 2;
212: }
213: break;
214: case 3:
215: PetscKernel_v_gets_A_times_w_3(x,idiag,b);
216: t[0] = b[0]; t[1] = b[1]; t[2] = b[2];
217: i2 = 3;
218: idiag += 9;
219: for (i=1; i<m; i++) {
220: v = aa + 9*ai[i];
221: vi = aj + ai[i];
222: nz = diag[i] - ai[i];
223: s[0] = b[i2]; s[1] = b[i2+1]; s[2] = b[i2+2];
224: while (nz--) {
225: idx = 3*(*vi++);
226: xw[0] = x[idx]; xw[1] = x[1+idx]; xw[2] = x[2+idx];
227: PetscKernel_v_gets_v_minus_A_times_w_3(s,v,xw);
228: v += 9;
229: }
230: t[i2] = s[0]; t[i2+1] = s[1]; t[i2+2] = s[2];
231: PetscKernel_v_gets_A_times_w_3(xw,idiag,s);
232: x[i2] = xw[0]; x[i2+1] = xw[1]; x[i2+2] = xw[2];
233: idiag += 9;
234: i2 += 3;
235: }
236: break;
237: case 4:
238: PetscKernel_v_gets_A_times_w_4(x,idiag,b);
239: t[0] = b[0]; t[1] = b[1]; t[2] = b[2]; t[3] = b[3];
240: i2 = 4;
241: idiag += 16;
242: for (i=1; i<m; i++) {
243: v = aa + 16*ai[i];
244: vi = aj + ai[i];
245: nz = diag[i] - ai[i];
246: s[0] = b[i2]; s[1] = b[i2+1]; s[2] = b[i2+2]; s[3] = b[i2+3];
247: while (nz--) {
248: idx = 4*(*vi++);
249: xw[0] = x[idx]; xw[1] = x[1+idx]; xw[2] = x[2+idx]; xw[3] = x[3+idx];
250: PetscKernel_v_gets_v_minus_A_times_w_4(s,v,xw);
251: v += 16;
252: }
253: t[i2] = s[0]; t[i2+1] = s[1]; t[i2+2] = s[2]; t[i2 + 3] = s[3];
254: PetscKernel_v_gets_A_times_w_4(xw,idiag,s);
255: x[i2] = xw[0]; x[i2+1] = xw[1]; x[i2+2] = xw[2]; x[i2+3] = xw[3];
256: idiag += 16;
257: i2 += 4;
258: }
259: break;
260: case 5:
261: PetscKernel_v_gets_A_times_w_5(x,idiag,b);
262: t[0] = b[0]; t[1] = b[1]; t[2] = b[2]; t[3] = b[3]; t[4] = b[4];
263: i2 = 5;
264: idiag += 25;
265: for (i=1; i<m; i++) {
266: v = aa + 25*ai[i];
267: vi = aj + ai[i];
268: nz = diag[i] - ai[i];
269: s[0] = b[i2]; s[1] = b[i2+1]; s[2] = b[i2+2]; s[3] = b[i2+3]; s[4] = b[i2+4];
270: while (nz--) {
271: idx = 5*(*vi++);
272: xw[0] = x[idx]; xw[1] = x[1+idx]; xw[2] = x[2+idx]; xw[3] = x[3+idx]; xw[4] = x[4+idx];
273: PetscKernel_v_gets_v_minus_A_times_w_5(s,v,xw);
274: v += 25;
275: }
276: t[i2] = s[0]; t[i2+1] = s[1]; t[i2+2] = s[2]; t[i2+3] = s[3]; t[i2+4] = s[4];
277: PetscKernel_v_gets_A_times_w_5(xw,idiag,s);
278: x[i2] = xw[0]; x[i2+1] = xw[1]; x[i2+2] = xw[2]; x[i2+3] = xw[3]; x[i2+4] = xw[4];
279: idiag += 25;
280: i2 += 5;
281: }
282: break;
283: case 6:
284: PetscKernel_v_gets_A_times_w_6(x,idiag,b);
285: t[0] = b[0]; t[1] = b[1]; t[2] = b[2]; t[3] = b[3]; t[4] = b[4]; t[5] = b[5];
286: i2 = 6;
287: idiag += 36;
288: for (i=1; i<m; i++) {
289: v = aa + 36*ai[i];
290: vi = aj + ai[i];
291: nz = diag[i] - ai[i];
292: s[0] = b[i2]; s[1] = b[i2+1]; s[2] = b[i2+2]; s[3] = b[i2+3]; s[4] = b[i2+4]; s[5] = b[i2+5];
293: while (nz--) {
294: idx = 6*(*vi++);
295: xw[0] = x[idx]; xw[1] = x[1+idx]; xw[2] = x[2+idx];
296: xw[3] = x[3+idx]; xw[4] = x[4+idx]; xw[5] = x[5+idx];
297: PetscKernel_v_gets_v_minus_A_times_w_6(s,v,xw);
298: v += 36;
299: }
300: t[i2] = s[0]; t[i2+1] = s[1]; t[i2+2] = s[2];
301: t[i2+3] = s[3]; t[i2+4] = s[4]; t[i2+5] = s[5];
302: PetscKernel_v_gets_A_times_w_6(xw,idiag,s);
303: x[i2] = xw[0]; x[i2+1] = xw[1]; x[i2+2] = xw[2]; x[i2+3] = xw[3]; x[i2+4] = xw[4]; x[i2+5] = xw[5];
304: idiag += 36;
305: i2 += 6;
306: }
307: break;
308: case 7:
309: PetscKernel_v_gets_A_times_w_7(x,idiag,b);
310: t[0] = b[0]; t[1] = b[1]; t[2] = b[2];
311: t[3] = b[3]; t[4] = b[4]; t[5] = b[5]; t[6] = b[6];
312: i2 = 7;
313: idiag += 49;
314: for (i=1; i<m; i++) {
315: v = aa + 49*ai[i];
316: vi = aj + ai[i];
317: nz = diag[i] - ai[i];
318: s[0] = b[i2]; s[1] = b[i2+1]; s[2] = b[i2+2];
319: s[3] = b[i2+3]; s[4] = b[i2+4]; s[5] = b[i2+5]; s[6] = b[i2+6];
320: while (nz--) {
321: idx = 7*(*vi++);
322: xw[0] = x[idx]; xw[1] = x[1+idx]; xw[2] = x[2+idx];
323: xw[3] = x[3+idx]; xw[4] = x[4+idx]; xw[5] = x[5+idx]; xw[6] = x[6+idx];
324: PetscKernel_v_gets_v_minus_A_times_w_7(s,v,xw);
325: v += 49;
326: }
327: t[i2] = s[0]; t[i2+1] = s[1]; t[i2+2] = s[2];
328: t[i2+3] = s[3]; t[i2+4] = s[4]; t[i2+5] = s[5]; t[i2+6] = s[6];
329: PetscKernel_v_gets_A_times_w_7(xw,idiag,s);
330: x[i2] = xw[0]; x[i2+1] = xw[1]; x[i2+2] = xw[2];
331: x[i2+3] = xw[3]; x[i2+4] = xw[4]; x[i2+5] = xw[5]; x[i2+6] = xw[6];
332: idiag += 49;
333: i2 += 7;
334: }
335: break;
336: default:
337: PetscKernel_w_gets_Ar_times_v(bs,bs,b,idiag,x);
338: PetscMemcpy(t,b,bs*sizeof(PetscScalar));
339: i2 = bs;
340: idiag += bs2;
341: for (i=1; i<m; i++) {
342: v = aa + bs2*ai[i];
343: vi = aj + ai[i];
344: nz = diag[i] - ai[i];
346: PetscMemcpy(w,b+i2,bs*sizeof(PetscScalar));
347: /* copy all rows of x that are needed into contiguous space */
348: workt = work;
349: for (j=0; j<nz; j++) {
350: PetscMemcpy(workt,x + bs*(*vi++),bs*sizeof(PetscScalar));
351: workt += bs;
352: }
353: PetscKernel_w_gets_w_minus_Ar_times_v(bs,bs*nz,w,v,work);
354: PetscMemcpy(t+i2,w,bs*sizeof(PetscScalar));
355: PetscKernel_w_gets_Ar_times_v(bs,bs,w,idiag,x+i2);
357: idiag += bs2;
358: i2 += bs;
359: }
360: break;
361: }
362: /* for logging purposes assume number of nonzero in lower half is 1/2 of total */
363: PetscLogFlops(1.0*bs2*a->nz);
364: xb = t;
365: }
366: else xb = b;
367: if (flag & SOR_BACKWARD_SWEEP || flag & SOR_LOCAL_BACKWARD_SWEEP) {
368: idiag = a->idiag+bs2*(a->mbs-1);
369: i2 = bs * (m-1);
370: switch (bs) {
371: case 1:
372: s[0] = xb[i2];
373: PetscKernel_v_gets_A_times_w_1(xw,idiag,s);
374: x[i2] = xw[0];
375: i2 -= 1;
376: for (i=m-2; i>=0; i--) {
377: v = aa + (diag[i]+1);
378: vi = aj + diag[i] + 1;
379: nz = ai[i+1] - diag[i] - 1;
380: s[0] = xb[i2];
381: for (j=0; j<nz; j++) {
382: xw[0] = x[vi[j]];
383: PetscKernel_v_gets_v_minus_A_times_w_1(s,(v+j),xw);
384: }
385: PetscKernel_v_gets_A_times_w_1(xw,idiag,s);
386: x[i2] = xw[0];
387: idiag -= 1;
388: i2 -= 1;
389: }
390: break;
391: case 2:
392: s[0] = xb[i2]; s[1] = xb[i2+1];
393: PetscKernel_v_gets_A_times_w_2(xw,idiag,s);
394: x[i2] = xw[0]; x[i2+1] = xw[1];
395: i2 -= 2;
396: idiag -= 4;
397: for (i=m-2; i>=0; i--) {
398: v = aa + 4*(diag[i] + 1);
399: vi = aj + diag[i] + 1;
400: nz = ai[i+1] - diag[i] - 1;
401: s[0] = xb[i2]; s[1] = xb[i2+1];
402: for (j=0; j<nz; j++) {
403: idx = 2*vi[j];
404: it = 4*j;
405: xw[0] = x[idx]; xw[1] = x[1+idx];
406: PetscKernel_v_gets_v_minus_A_times_w_2(s,(v+it),xw);
407: }
408: PetscKernel_v_gets_A_times_w_2(xw,idiag,s);
409: x[i2] = xw[0]; x[i2+1] = xw[1];
410: idiag -= 4;
411: i2 -= 2;
412: }
413: break;
414: case 3:
415: s[0] = xb[i2]; s[1] = xb[i2+1]; s[2] = xb[i2+2];
416: PetscKernel_v_gets_A_times_w_3(xw,idiag,s);
417: x[i2] = xw[0]; x[i2+1] = xw[1]; x[i2+2] = xw[2];
418: i2 -= 3;
419: idiag -= 9;
420: for (i=m-2; i>=0; i--) {
421: v = aa + 9*(diag[i]+1);
422: vi = aj + diag[i] + 1;
423: nz = ai[i+1] - diag[i] - 1;
424: s[0] = xb[i2]; s[1] = xb[i2+1]; s[2] = xb[i2+2];
425: while (nz--) {
426: idx = 3*(*vi++);
427: xw[0] = x[idx]; xw[1] = x[1+idx]; xw[2] = x[2+idx];
428: PetscKernel_v_gets_v_minus_A_times_w_3(s,v,xw);
429: v += 9;
430: }
431: PetscKernel_v_gets_A_times_w_3(xw,idiag,s);
432: x[i2] = xw[0]; x[i2+1] = xw[1]; x[i2+2] = xw[2];
433: idiag -= 9;
434: i2 -= 3;
435: }
436: break;
437: case 4:
438: s[0] = xb[i2]; s[1] = xb[i2+1]; s[2] = xb[i2+2]; s[3] = xb[i2+3];
439: PetscKernel_v_gets_A_times_w_4(xw,idiag,s);
440: x[i2] = xw[0]; x[i2+1] = xw[1]; x[i2+2] = xw[2]; x[i2+3] = xw[3];
441: i2 -= 4;
442: idiag -= 16;
443: for (i=m-2; i>=0; i--) {
444: v = aa + 16*(diag[i]+1);
445: vi = aj + diag[i] + 1;
446: nz = ai[i+1] - diag[i] - 1;
447: s[0] = xb[i2]; s[1] = xb[i2+1]; s[2] = xb[i2+2]; s[3] = xb[i2+3];
448: while (nz--) {
449: idx = 4*(*vi++);
450: xw[0] = x[idx]; xw[1] = x[1+idx]; xw[2] = x[2+idx]; xw[3] = x[3+idx];
451: PetscKernel_v_gets_v_minus_A_times_w_4(s,v,xw);
452: v += 16;
453: }
454: PetscKernel_v_gets_A_times_w_4(xw,idiag,s);
455: x[i2] = xw[0]; x[i2+1] = xw[1]; x[i2+2] = xw[2]; x[i2+3] = xw[3];
456: idiag -= 16;
457: i2 -= 4;
458: }
459: break;
460: case 5:
461: s[0] = xb[i2]; s[1] = xb[i2+1]; s[2] = xb[i2+2]; s[3] = xb[i2+3]; s[4] = xb[i2+4];
462: PetscKernel_v_gets_A_times_w_5(xw,idiag,s);
463: x[i2] = xw[0]; x[i2+1] = xw[1]; x[i2+2] = xw[2]; x[i2+3] = xw[3]; x[i2+4] = xw[4];
464: i2 -= 5;
465: idiag -= 25;
466: for (i=m-2; i>=0; i--) {
467: v = aa + 25*(diag[i]+1);
468: vi = aj + diag[i] + 1;
469: nz = ai[i+1] - diag[i] - 1;
470: s[0] = xb[i2]; s[1] = xb[i2+1]; s[2] = xb[i2+2]; s[3] = xb[i2+3]; s[4] = xb[i2+4];
471: while (nz--) {
472: idx = 5*(*vi++);
473: xw[0] = x[idx]; xw[1] = x[1+idx]; xw[2] = x[2+idx]; xw[3] = x[3+idx]; xw[4] = x[4+idx];
474: PetscKernel_v_gets_v_minus_A_times_w_5(s,v,xw);
475: v += 25;
476: }
477: PetscKernel_v_gets_A_times_w_5(xw,idiag,s);
478: x[i2] = xw[0]; x[i2+1] = xw[1]; x[i2+2] = xw[2]; x[i2+3] = xw[3]; x[i2+4] = xw[4];
479: idiag -= 25;
480: i2 -= 5;
481: }
482: break;
483: case 6:
484: s[0] = xb[i2]; s[1] = xb[i2+1]; s[2] = xb[i2+2]; s[3] = xb[i2+3]; s[4] = xb[i2+4]; s[5] = xb[i2+5];
485: PetscKernel_v_gets_A_times_w_6(xw,idiag,s);
486: x[i2] = xw[0]; x[i2+1] = xw[1]; x[i2+2] = xw[2]; x[i2+3] = xw[3]; x[i2+4] = xw[4]; x[i2+5] = xw[5];
487: i2 -= 6;
488: idiag -= 36;
489: for (i=m-2; i>=0; i--) {
490: v = aa + 36*(diag[i]+1);
491: vi = aj + diag[i] + 1;
492: nz = ai[i+1] - diag[i] - 1;
493: s[0] = xb[i2]; s[1] = xb[i2+1]; s[2] = xb[i2+2]; s[3] = xb[i2+3]; s[4] = xb[i2+4]; s[5] = xb[i2+5];
494: while (nz--) {
495: idx = 6*(*vi++);
496: xw[0] = x[idx]; xw[1] = x[1+idx]; xw[2] = x[2+idx];
497: xw[3] = x[3+idx]; xw[4] = x[4+idx]; xw[5] = x[5+idx];
498: PetscKernel_v_gets_v_minus_A_times_w_6(s,v,xw);
499: v += 36;
500: }
501: PetscKernel_v_gets_A_times_w_6(xw,idiag,s);
502: x[i2] = xw[0]; x[i2+1] = xw[1]; x[i2+2] = xw[2]; x[i2+3] = xw[3]; x[i2+4] = xw[4]; x[i2+5] = xw[5];
503: idiag -= 36;
504: i2 -= 6;
505: }
506: break;
507: case 7:
508: s[0] = xb[i2]; s[1] = xb[i2+1]; s[2] = xb[i2+2];
509: s[3] = xb[i2+3]; s[4] = xb[i2+4]; s[5] = xb[i2+5]; s[6] = xb[i2+6];
510: PetscKernel_v_gets_A_times_w_7(x,idiag,b);
511: x[i2] = xw[0]; x[i2+1] = xw[1]; x[i2+2] = xw[2];
512: x[i2+3] = xw[3]; x[i2+4] = xw[4]; x[i2+5] = xw[5]; x[i2+6] = xw[6];
513: i2 -= 7;
514: idiag -= 49;
515: for (i=m-2; i>=0; i--) {
516: v = aa + 49*(diag[i]+1);
517: vi = aj + diag[i] + 1;
518: nz = ai[i+1] - diag[i] - 1;
519: s[0] = xb[i2]; s[1] = xb[i2+1]; s[2] = xb[i2+2];
520: s[3] = xb[i2+3]; s[4] = xb[i2+4]; s[5] = xb[i2+5]; s[6] = xb[i2+6];
521: while (nz--) {
522: idx = 7*(*vi++);
523: xw[0] = x[idx]; xw[1] = x[1+idx]; xw[2] = x[2+idx];
524: xw[3] = x[3+idx]; xw[4] = x[4+idx]; xw[5] = x[5+idx]; xw[6] = x[6+idx];
525: PetscKernel_v_gets_v_minus_A_times_w_7(s,v,xw);
526: v += 49;
527: }
528: PetscKernel_v_gets_A_times_w_7(xw,idiag,s);
529: x[i2] = xw[0]; x[i2+1] = xw[1]; x[i2+2] = xw[2];
530: x[i2+3] = xw[3]; x[i2+4] = xw[4]; x[i2+5] = xw[5]; x[i2+6] = xw[6];
531: idiag -= 49;
532: i2 -= 7;
533: }
534: break;
535: default:
536: PetscMemcpy(w,xb+i2,bs*sizeof(PetscScalar));
537: PetscKernel_w_gets_Ar_times_v(bs,bs,w,idiag,x+i2);
538: i2 -= bs;
539: idiag -= bs2;
540: for (i=m-2; i>=0; i--) {
541: v = aa + bs2*(diag[i]+1);
542: vi = aj + diag[i] + 1;
543: nz = ai[i+1] - diag[i] - 1;
545: PetscMemcpy(w,xb+i2,bs*sizeof(PetscScalar));
546: /* copy all rows of x that are needed into contiguous space */
547: workt = work;
548: for (j=0; j<nz; j++) {
549: PetscMemcpy(workt,x + bs*(*vi++),bs*sizeof(PetscScalar));
550: workt += bs;
551: }
552: PetscKernel_w_gets_w_minus_Ar_times_v(bs,bs*nz,w,v,work);
553: PetscKernel_w_gets_Ar_times_v(bs,bs,w,idiag,x+i2);
555: idiag -= bs2;
556: i2 -= bs;
557: }
558: break;
559: }
560: PetscLogFlops(1.0*bs2*(a->nz));
561: }
562: its--;
563: }
564: while (its--) {
565: if (flag & SOR_FORWARD_SWEEP || flag & SOR_LOCAL_FORWARD_SWEEP) {
566: idiag = a->idiag;
567: i2 = 0;
568: switch (bs) {
569: case 1:
570: for (i=0; i<m; i++) {
571: v = aa + ai[i];
572: vi = aj + ai[i];
573: nz = ai[i+1] - ai[i];
574: s[0] = b[i2];
575: for (j=0; j<nz; j++) {
576: xw[0] = x[vi[j]];
577: PetscKernel_v_gets_v_minus_A_times_w_1(s,(v+j),xw);
578: }
579: PetscKernel_v_gets_A_times_w_1(xw,idiag,s);
580: x[i2] += xw[0];
581: idiag += 1;
582: i2 += 1;
583: }
584: break;
585: case 2:
586: for (i=0; i<m; i++) {
587: v = aa + 4*ai[i];
588: vi = aj + ai[i];
589: nz = ai[i+1] - ai[i];
590: s[0] = b[i2]; s[1] = b[i2+1];
591: for (j=0; j<nz; j++) {
592: idx = 2*vi[j];
593: it = 4*j;
594: xw[0] = x[idx]; xw[1] = x[1+idx];
595: PetscKernel_v_gets_v_minus_A_times_w_2(s,(v+it),xw);
596: }
597: PetscKernel_v_gets_A_times_w_2(xw,idiag,s);
598: x[i2] += xw[0]; x[i2+1] += xw[1];
599: idiag += 4;
600: i2 += 2;
601: }
602: break;
603: case 3:
604: for (i=0; i<m; i++) {
605: v = aa + 9*ai[i];
606: vi = aj + ai[i];
607: nz = ai[i+1] - ai[i];
608: s[0] = b[i2]; s[1] = b[i2+1]; s[2] = b[i2+2];
609: while (nz--) {
610: idx = 3*(*vi++);
611: xw[0] = x[idx]; xw[1] = x[1+idx]; xw[2] = x[2+idx];
612: PetscKernel_v_gets_v_minus_A_times_w_3(s,v,xw);
613: v += 9;
614: }
615: PetscKernel_v_gets_A_times_w_3(xw,idiag,s);
616: x[i2] += xw[0]; x[i2+1] += xw[1]; x[i2+2] += xw[2];
617: idiag += 9;
618: i2 += 3;
619: }
620: break;
621: case 4:
622: for (i=0; i<m; i++) {
623: v = aa + 16*ai[i];
624: vi = aj + ai[i];
625: nz = ai[i+1] - ai[i];
626: s[0] = b[i2]; s[1] = b[i2+1]; s[2] = b[i2+2]; s[3] = b[i2+3];
627: while (nz--) {
628: idx = 4*(*vi++);
629: xw[0] = x[idx]; xw[1] = x[1+idx]; xw[2] = x[2+idx]; xw[3] = x[3+idx];
630: PetscKernel_v_gets_v_minus_A_times_w_4(s,v,xw);
631: v += 16;
632: }
633: PetscKernel_v_gets_A_times_w_4(xw,idiag,s);
634: x[i2] += xw[0]; x[i2+1] += xw[1]; x[i2+2] += xw[2]; x[i2+3] += xw[3];
635: idiag += 16;
636: i2 += 4;
637: }
638: break;
639: case 5:
640: for (i=0; i<m; i++) {
641: v = aa + 25*ai[i];
642: vi = aj + ai[i];
643: nz = ai[i+1] - ai[i];
644: s[0] = b[i2]; s[1] = b[i2+1]; s[2] = b[i2+2]; s[3] = b[i2+3]; s[4] = b[i2+4];
645: while (nz--) {
646: idx = 5*(*vi++);
647: xw[0] = x[idx]; xw[1] = x[1+idx]; xw[2] = x[2+idx]; xw[3] = x[3+idx]; xw[4] = x[4+idx];
648: PetscKernel_v_gets_v_minus_A_times_w_5(s,v,xw);
649: v += 25;
650: }
651: PetscKernel_v_gets_A_times_w_5(xw,idiag,s);
652: x[i2] += xw[0]; x[i2+1] += xw[1]; x[i2+2] += xw[2]; x[i2+3] += xw[3]; x[i2+4] += xw[4];
653: idiag += 25;
654: i2 += 5;
655: }
656: break;
657: case 6:
658: for (i=0; i<m; i++) {
659: v = aa + 36*ai[i];
660: vi = aj + ai[i];
661: nz = ai[i+1] - ai[i];
662: s[0] = b[i2]; s[1] = b[i2+1]; s[2] = b[i2+2]; s[3] = b[i2+3]; s[4] = b[i2+4]; s[5] = b[i2+5];
663: while (nz--) {
664: idx = 6*(*vi++);
665: xw[0] = x[idx]; xw[1] = x[1+idx]; xw[2] = x[2+idx];
666: xw[3] = x[3+idx]; xw[4] = x[4+idx]; xw[5] = x[5+idx];
667: PetscKernel_v_gets_v_minus_A_times_w_6(s,v,xw);
668: v += 36;
669: }
670: PetscKernel_v_gets_A_times_w_6(xw,idiag,s);
671: x[i2] += xw[0]; x[i2+1] += xw[1]; x[i2+2] += xw[2];
672: x[i2+3] += xw[3]; x[i2+4] += xw[4]; x[i2+5] += xw[5];
673: idiag += 36;
674: i2 += 6;
675: }
676: break;
677: case 7:
678: for (i=0; i<m; i++) {
679: v = aa + 49*ai[i];
680: vi = aj + ai[i];
681: nz = ai[i+1] - ai[i];
682: s[0] = b[i2]; s[1] = b[i2+1]; s[2] = b[i2+2];
683: s[3] = b[i2+3]; s[4] = b[i2+4]; s[5] = b[i2+5]; s[6] = b[i2+6];
684: while (nz--) {
685: idx = 7*(*vi++);
686: xw[0] = x[idx]; xw[1] = x[1+idx]; xw[2] = x[2+idx];
687: xw[3] = x[3+idx]; xw[4] = x[4+idx]; xw[5] = x[5+idx]; xw[6] = x[6+idx];
688: PetscKernel_v_gets_v_minus_A_times_w_7(s,v,xw);
689: v += 49;
690: }
691: PetscKernel_v_gets_A_times_w_7(xw,idiag,s);
692: x[i2] += xw[0]; x[i2+1] += xw[1]; x[i2+2] += xw[2];
693: x[i2+3] += xw[3]; x[i2+4] += xw[4]; x[i2+5] += xw[5]; x[i2+6] += xw[6];
694: idiag += 49;
695: i2 += 7;
696: }
697: break;
698: default:
699: for (i=0; i<m; i++) {
700: v = aa + bs2*ai[i];
701: vi = aj + ai[i];
702: nz = ai[i+1] - ai[i];
704: PetscMemcpy(w,b+i2,bs*sizeof(PetscScalar));
705: /* copy all rows of x that are needed into contiguous space */
706: workt = work;
707: for (j=0; j<nz; j++) {
708: PetscMemcpy(workt,x + bs*(*vi++),bs*sizeof(PetscScalar));
709: workt += bs;
710: }
711: PetscKernel_w_gets_w_minus_Ar_times_v(bs,bs*nz,w,v,work);
712: PetscKernel_w_gets_w_plus_Ar_times_v(bs,bs,w,idiag,x+i2);
714: idiag += bs2;
715: i2 += bs;
716: }
717: break;
718: }
719: PetscLogFlops(2.0*bs2*a->nz);
720: }
721: if (flag & SOR_BACKWARD_SWEEP || flag & SOR_LOCAL_BACKWARD_SWEEP) {
722: idiag = a->idiag+bs2*(a->mbs-1);
723: i2 = bs * (m-1);
724: switch (bs) {
725: case 1:
726: for (i=m-1; i>=0; i--) {
727: v = aa + ai[i];
728: vi = aj + ai[i];
729: nz = ai[i+1] - ai[i];
730: s[0] = b[i2];
731: for (j=0; j<nz; j++) {
732: xw[0] = x[vi[j]];
733: PetscKernel_v_gets_v_minus_A_times_w_1(s,(v+j),xw);
734: }
735: PetscKernel_v_gets_A_times_w_1(xw,idiag,s);
736: x[i2] += xw[0];
737: idiag -= 1;
738: i2 -= 1;
739: }
740: break;
741: case 2:
742: for (i=m-1; i>=0; i--) {
743: v = aa + 4*ai[i];
744: vi = aj + ai[i];
745: nz = ai[i+1] - ai[i];
746: s[0] = b[i2]; s[1] = b[i2+1];
747: for (j=0; j<nz; j++) {
748: idx = 2*vi[j];
749: it = 4*j;
750: xw[0] = x[idx]; xw[1] = x[1+idx];
751: PetscKernel_v_gets_v_minus_A_times_w_2(s,(v+it),xw);
752: }
753: PetscKernel_v_gets_A_times_w_2(xw,idiag,s);
754: x[i2] += xw[0]; x[i2+1] += xw[1];
755: idiag -= 4;
756: i2 -= 2;
757: }
758: break;
759: case 3:
760: for (i=m-1; i>=0; i--) {
761: v = aa + 9*ai[i];
762: vi = aj + ai[i];
763: nz = ai[i+1] - ai[i];
764: s[0] = b[i2]; s[1] = b[i2+1]; s[2] = b[i2+2];
765: while (nz--) {
766: idx = 3*(*vi++);
767: xw[0] = x[idx]; xw[1] = x[1+idx]; xw[2] = x[2+idx];
768: PetscKernel_v_gets_v_minus_A_times_w_3(s,v,xw);
769: v += 9;
770: }
771: PetscKernel_v_gets_A_times_w_3(xw,idiag,s);
772: x[i2] += xw[0]; x[i2+1] += xw[1]; x[i2+2] += xw[2];
773: idiag -= 9;
774: i2 -= 3;
775: }
776: break;
777: case 4:
778: for (i=m-1; i>=0; i--) {
779: v = aa + 16*ai[i];
780: vi = aj + ai[i];
781: nz = ai[i+1] - ai[i];
782: s[0] = b[i2]; s[1] = b[i2+1]; s[2] = b[i2+2]; s[3] = b[i2+3];
783: while (nz--) {
784: idx = 4*(*vi++);
785: xw[0] = x[idx]; xw[1] = x[1+idx]; xw[2] = x[2+idx]; xw[3] = x[3+idx];
786: PetscKernel_v_gets_v_minus_A_times_w_4(s,v,xw);
787: v += 16;
788: }
789: PetscKernel_v_gets_A_times_w_4(xw,idiag,s);
790: x[i2] += xw[0]; x[i2+1] += xw[1]; x[i2+2] += xw[2]; x[i2+3] += xw[3];
791: idiag -= 16;
792: i2 -= 4;
793: }
794: break;
795: case 5:
796: for (i=m-1; i>=0; i--) {
797: v = aa + 25*ai[i];
798: vi = aj + ai[i];
799: nz = ai[i+1] - ai[i];
800: s[0] = b[i2]; s[1] = b[i2+1]; s[2] = b[i2+2]; s[3] = b[i2+3]; s[4] = b[i2+4];
801: while (nz--) {
802: idx = 5*(*vi++);
803: xw[0] = x[idx]; xw[1] = x[1+idx]; xw[2] = x[2+idx]; xw[3] = x[3+idx]; xw[4] = x[4+idx];
804: PetscKernel_v_gets_v_minus_A_times_w_5(s,v,xw);
805: v += 25;
806: }
807: PetscKernel_v_gets_A_times_w_5(xw,idiag,s);
808: x[i2] += xw[0]; x[i2+1] += xw[1]; x[i2+2] += xw[2]; x[i2+3] += xw[3]; x[i2+4] += xw[4];
809: idiag -= 25;
810: i2 -= 5;
811: }
812: break;
813: case 6:
814: for (i=m-1; i>=0; i--) {
815: v = aa + 36*ai[i];
816: vi = aj + ai[i];
817: nz = ai[i+1] - ai[i];
818: s[0] = b[i2]; s[1] = b[i2+1]; s[2] = b[i2+2]; s[3] = b[i2+3]; s[4] = b[i2+4]; s[5] = b[i2+5];
819: while (nz--) {
820: idx = 6*(*vi++);
821: xw[0] = x[idx]; xw[1] = x[1+idx]; xw[2] = x[2+idx];
822: xw[3] = x[3+idx]; xw[4] = x[4+idx]; xw[5] = x[5+idx];
823: PetscKernel_v_gets_v_minus_A_times_w_6(s,v,xw);
824: v += 36;
825: }
826: PetscKernel_v_gets_A_times_w_6(xw,idiag,s);
827: x[i2] += xw[0]; x[i2+1] += xw[1]; x[i2+2] += xw[2];
828: x[i2+3] += xw[3]; x[i2+4] += xw[4]; x[i2+5] += xw[5];
829: idiag -= 36;
830: i2 -= 6;
831: }
832: break;
833: case 7:
834: for (i=m-1; i>=0; i--) {
835: v = aa + 49*ai[i];
836: vi = aj + ai[i];
837: nz = ai[i+1] - ai[i];
838: s[0] = b[i2]; s[1] = b[i2+1]; s[2] = b[i2+2];
839: s[3] = b[i2+3]; s[4] = b[i2+4]; s[5] = b[i2+5]; s[6] = b[i2+6];
840: while (nz--) {
841: idx = 7*(*vi++);
842: xw[0] = x[idx]; xw[1] = x[1+idx]; xw[2] = x[2+idx];
843: xw[3] = x[3+idx]; xw[4] = x[4+idx]; xw[5] = x[5+idx]; xw[6] = x[6+idx];
844: PetscKernel_v_gets_v_minus_A_times_w_7(s,v,xw);
845: v += 49;
846: }
847: PetscKernel_v_gets_A_times_w_7(xw,idiag,s);
848: x[i2] += xw[0]; x[i2+1] += xw[1]; x[i2+2] += xw[2];
849: x[i2+3] += xw[3]; x[i2+4] += xw[4]; x[i2+5] += xw[5]; x[i2+6] += xw[6];
850: idiag -= 49;
851: i2 -= 7;
852: }
853: break;
854: default:
855: for (i=m-1; i>=0; i--) {
856: v = aa + bs2*ai[i];
857: vi = aj + ai[i];
858: nz = ai[i+1] - ai[i];
860: PetscMemcpy(w,b+i2,bs*sizeof(PetscScalar));
861: /* copy all rows of x that are needed into contiguous space */
862: workt = work;
863: for (j=0; j<nz; j++) {
864: PetscMemcpy(workt,x + bs*(*vi++),bs*sizeof(PetscScalar));
865: workt += bs;
866: }
867: PetscKernel_w_gets_w_minus_Ar_times_v(bs,bs*nz,w,v,work);
868: PetscKernel_w_gets_w_plus_Ar_times_v(bs,bs,w,idiag,x+i2);
870: idiag -= bs2;
871: i2 -= bs;
872: }
873: break;
874: }
875: PetscLogFlops(2.0*bs2*(a->nz));
876: }
877: }
878: VecRestoreArray(xx,&x);
879: VecRestoreArrayRead(bb,&b);
880: return(0);
881: }
884: /*
885: Special version for direct calls from Fortran (Used in PETSc-fun3d)
886: */
887: #if defined(PETSC_HAVE_FORTRAN_CAPS)
888: #define matsetvaluesblocked4_ MATSETVALUESBLOCKED4
889: #elif !defined(PETSC_HAVE_FORTRAN_UNDERSCORE)
890: #define matsetvaluesblocked4_ matsetvaluesblocked4
891: #endif
895: PETSC_EXTERN void matsetvaluesblocked4_(Mat *AA,PetscInt *mm,const PetscInt im[],PetscInt *nn,const PetscInt in[],const PetscScalar v[])
896: {
897: Mat A = *AA;
898: Mat_SeqBAIJ *a = (Mat_SeqBAIJ*)A->data;
899: PetscInt *rp,k,low,high,t,ii,jj,row,nrow,i,col,l,N,m = *mm,n = *nn;
900: PetscInt *ai =a->i,*ailen=a->ilen;
901: PetscInt *aj =a->j,stepval,lastcol = -1;
902: const PetscScalar *value = v;
903: MatScalar *ap,*aa = a->a,*bap;
906: if (A->rmap->bs != 4) SETERRABORT(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONG,"Can only be called with a block size of 4");
907: stepval = (n-1)*4;
908: for (k=0; k<m; k++) { /* loop over added rows */
909: row = im[k];
910: rp = aj + ai[row];
911: ap = aa + 16*ai[row];
912: nrow = ailen[row];
913: low = 0;
914: high = nrow;
915: for (l=0; l<n; l++) { /* loop over added columns */
916: col = in[l];
917: if (col <= lastcol) low = 0;
918: else high = nrow;
919: lastcol = col;
920: value = v + k*(stepval+4 + l)*4;
921: while (high-low > 7) {
922: t = (low+high)/2;
923: if (rp[t] > col) high = t;
924: else low = t;
925: }
926: for (i=low; i<high; i++) {
927: if (rp[i] > col) break;
928: if (rp[i] == col) {
929: bap = ap + 16*i;
930: for (ii=0; ii<4; ii++,value+=stepval) {
931: for (jj=ii; jj<16; jj+=4) {
932: bap[jj] += *value++;
933: }
934: }
935: goto noinsert2;
936: }
937: }
938: N = nrow++ - 1;
939: high++; /* added new column index thus must search to one higher than before */
940: /* shift up all the later entries in this row */
941: for (ii=N; ii>=i; ii--) {
942: rp[ii+1] = rp[ii];
943: PetscMemcpy(ap+16*(ii+1),ap+16*(ii),16*sizeof(MatScalar));
944: }
945: if (N >= i) {
946: PetscMemzero(ap+16*i,16*sizeof(MatScalar));
947: }
948: rp[i] = col;
949: bap = ap + 16*i;
950: for (ii=0; ii<4; ii++,value+=stepval) {
951: for (jj=ii; jj<16; jj+=4) {
952: bap[jj] = *value++;
953: }
954: }
955: noinsert2:;
956: low = i;
957: }
958: ailen[row] = nrow;
959: }
960: PetscFunctionReturnVoid();
961: }
963: #if defined(PETSC_HAVE_FORTRAN_CAPS)
964: #define matsetvalues4_ MATSETVALUES4
965: #elif !defined(PETSC_HAVE_FORTRAN_UNDERSCORE)
966: #define matsetvalues4_ matsetvalues4
967: #endif
971: PETSC_EXTERN void matsetvalues4_(Mat *AA,PetscInt *mm,PetscInt *im,PetscInt *nn,PetscInt *in,PetscScalar *v)
972: {
973: Mat A = *AA;
974: Mat_SeqBAIJ *a = (Mat_SeqBAIJ*)A->data;
975: PetscInt *rp,k,low,high,t,ii,row,nrow,i,col,l,N,n = *nn,m = *mm;
976: PetscInt *ai=a->i,*ailen=a->ilen;
977: PetscInt *aj=a->j,brow,bcol;
978: PetscInt ridx,cidx,lastcol = -1;
979: MatScalar *ap,value,*aa=a->a,*bap;
982: for (k=0; k<m; k++) { /* loop over added rows */
983: row = im[k]; brow = row/4;
984: rp = aj + ai[brow];
985: ap = aa + 16*ai[brow];
986: nrow = ailen[brow];
987: low = 0;
988: high = nrow;
989: for (l=0; l<n; l++) { /* loop over added columns */
990: col = in[l]; bcol = col/4;
991: ridx = row % 4; cidx = col % 4;
992: value = v[l + k*n];
993: if (col <= lastcol) low = 0;
994: else high = nrow;
995: lastcol = col;
996: while (high-low > 7) {
997: t = (low+high)/2;
998: if (rp[t] > bcol) high = t;
999: else low = t;
1000: }
1001: for (i=low; i<high; i++) {
1002: if (rp[i] > bcol) break;
1003: if (rp[i] == bcol) {
1004: bap = ap + 16*i + 4*cidx + ridx;
1005: *bap += value;
1006: goto noinsert1;
1007: }
1008: }
1009: N = nrow++ - 1;
1010: high++; /* added new column thus must search to one higher than before */
1011: /* shift up all the later entries in this row */
1012: for (ii=N; ii>=i; ii--) {
1013: rp[ii+1] = rp[ii];
1014: PetscMemcpy(ap+16*(ii+1),ap+16*(ii),16*sizeof(MatScalar));
1015: }
1016: if (N>=i) {
1017: PetscMemzero(ap+16*i,16*sizeof(MatScalar));
1018: }
1019: rp[i] = bcol;
1020: ap[16*i + 4*cidx + ridx] = value;
1021: noinsert1:;
1022: low = i;
1023: }
1024: ailen[brow] = nrow;
1025: }
1026: PetscFunctionReturnVoid();
1027: }
1029: /*
1030: Checks for missing diagonals
1031: */
1034: PetscErrorCode MatMissingDiagonal_SeqBAIJ(Mat A,PetscBool *missing,PetscInt *d)
1035: {
1036: Mat_SeqBAIJ *a = (Mat_SeqBAIJ*)A->data;
1038: PetscInt *diag,*ii = a->i,i;
1041: MatMarkDiagonal_SeqBAIJ(A);
1042: *missing = PETSC_FALSE;
1043: if (A->rmap->n > 0 && !ii) {
1044: *missing = PETSC_TRUE;
1045: if (d) *d = 0;
1046: PetscInfo(A,"Matrix has no entries therefore is missing diagonal");
1047: } else {
1048: diag = a->diag;
1049: for (i=0; i<a->mbs; i++) {
1050: if (diag[i] >= ii[i+1]) {
1051: *missing = PETSC_TRUE;
1052: if (d) *d = i;
1053: PetscInfo1(A,"Matrix is missing block diagonal number %D",i);
1054: break;
1055: }
1056: }
1057: }
1058: return(0);
1059: }
1063: PetscErrorCode MatMarkDiagonal_SeqBAIJ(Mat A)
1064: {
1065: Mat_SeqBAIJ *a = (Mat_SeqBAIJ*)A->data;
1067: PetscInt i,j,m = a->mbs;
1070: if (!a->diag) {
1071: PetscMalloc1(m,&a->diag);
1072: PetscLogObjectMemory((PetscObject)A,m*sizeof(PetscInt));
1073: a->free_diag = PETSC_TRUE;
1074: }
1075: for (i=0; i<m; i++) {
1076: a->diag[i] = a->i[i+1];
1077: for (j=a->i[i]; j<a->i[i+1]; j++) {
1078: if (a->j[j] == i) {
1079: a->diag[i] = j;
1080: break;
1081: }
1082: }
1083: }
1084: return(0);
1085: }
1090: static PetscErrorCode MatGetRowIJ_SeqBAIJ(Mat A,PetscInt oshift,PetscBool symmetric,PetscBool blockcompressed,PetscInt *nn,const PetscInt *inia[],const PetscInt *inja[],PetscBool *done)
1091: {
1092: Mat_SeqBAIJ *a = (Mat_SeqBAIJ*)A->data;
1094: PetscInt i,j,n = a->mbs,nz = a->i[n],*tia,*tja,bs = A->rmap->bs,k,l,cnt;
1095: PetscInt **ia = (PetscInt**)inia,**ja = (PetscInt**)inja;
1098: *nn = n;
1099: if (!ia) return(0);
1100: if (symmetric) {
1101: MatToSymmetricIJ_SeqAIJ(n,a->i,a->j,0,0,&tia,&tja);
1102: nz = tia[n];
1103: } else {
1104: tia = a->i; tja = a->j;
1105: }
1107: if (!blockcompressed && bs > 1) {
1108: (*nn) *= bs;
1109: /* malloc & create the natural set of indices */
1110: PetscMalloc1((n+1)*bs,ia);
1111: if (n) {
1112: (*ia)[0] = 0;
1113: for (j=1; j<bs; j++) {
1114: (*ia)[j] = (tia[1]-tia[0])*bs+(*ia)[j-1];
1115: }
1116: }
1118: for (i=1; i<n; i++) {
1119: (*ia)[i*bs] = (tia[i]-tia[i-1])*bs + (*ia)[i*bs-1];
1120: for (j=1; j<bs; j++) {
1121: (*ia)[i*bs+j] = (tia[i+1]-tia[i])*bs + (*ia)[i*bs+j-1];
1122: }
1123: }
1124: if (n) {
1125: (*ia)[n*bs] = (tia[n]-tia[n-1])*bs + (*ia)[n*bs-1];
1126: }
1128: if (inja) {
1129: PetscMalloc1(nz*bs*bs,ja);
1130: cnt = 0;
1131: for (i=0; i<n; i++) {
1132: for (j=0; j<bs; j++) {
1133: for (k=tia[i]; k<tia[i+1]; k++) {
1134: for (l=0; l<bs; l++) {
1135: (*ja)[cnt++] = bs*tja[k] + l;
1136: }
1137: }
1138: }
1139: }
1140: }
1142: if (symmetric) { /* deallocate memory allocated in MatToSymmetricIJ_SeqAIJ() */
1143: PetscFree(tia);
1144: PetscFree(tja);
1145: }
1146: } else if (oshift == 1) {
1147: if (symmetric) {
1148: nz = tia[A->rmap->n/bs];
1149: /* add 1 to i and j indices */
1150: for (i=0; i<A->rmap->n/bs+1; i++) tia[i] = tia[i] + 1;
1151: *ia = tia;
1152: if (ja) {
1153: for (i=0; i<nz; i++) tja[i] = tja[i] + 1;
1154: *ja = tja;
1155: }
1156: } else {
1157: nz = a->i[A->rmap->n/bs];
1158: /* malloc space and add 1 to i and j indices */
1159: PetscMalloc1((A->rmap->n/bs+1),ia);
1160: for (i=0; i<A->rmap->n/bs+1; i++) (*ia)[i] = a->i[i] + 1;
1161: if (ja) {
1162: PetscMalloc1(nz,ja);
1163: for (i=0; i<nz; i++) (*ja)[i] = a->j[i] + 1;
1164: }
1165: }
1166: } else {
1167: *ia = tia;
1168: if (ja) *ja = tja;
1169: }
1170: return(0);
1171: }
1175: static PetscErrorCode MatRestoreRowIJ_SeqBAIJ(Mat A,PetscInt oshift,PetscBool symmetric,PetscBool blockcompressed,PetscInt *nn,const PetscInt *ia[],const PetscInt *ja[],PetscBool *done)
1176: {
1180: if (!ia) return(0);
1181: if ((!blockcompressed && A->rmap->bs > 1) || (symmetric || oshift == 1)) {
1182: PetscFree(*ia);
1183: if (ja) {PetscFree(*ja);}
1184: }
1185: return(0);
1186: }
1190: PetscErrorCode MatDestroy_SeqBAIJ(Mat A)
1191: {
1192: Mat_SeqBAIJ *a = (Mat_SeqBAIJ*)A->data;
1196: #if defined(PETSC_USE_LOG)
1197: PetscLogObjectState((PetscObject)A,"Rows=%D, Cols=%D, NZ=%D",A->rmap->N,A->cmap->n,a->nz);
1198: #endif
1199: MatSeqXAIJFreeAIJ(A,&a->a,&a->j,&a->i);
1200: ISDestroy(&a->row);
1201: ISDestroy(&a->col);
1202: if (a->free_diag) {PetscFree(a->diag);}
1203: PetscFree(a->idiag);
1204: if (a->free_imax_ilen) {PetscFree2(a->imax,a->ilen);}
1205: PetscFree(a->solve_work);
1206: PetscFree(a->mult_work);
1207: PetscFree(a->sor_work);
1208: ISDestroy(&a->icol);
1209: PetscFree(a->saved_values);
1210: PetscFree(a->xtoy);
1211: PetscFree2(a->compressedrow.i,a->compressedrow.rindex);
1213: MatDestroy(&a->sbaijMat);
1214: MatDestroy(&a->parent);
1215: PetscFree(A->data);
1217: PetscObjectChangeTypeName((PetscObject)A,0);
1218: PetscObjectComposeFunction((PetscObject)A,"MatInvertBlockDiagonal_C",NULL);
1219: PetscObjectComposeFunction((PetscObject)A,"MatStoreValues_C",NULL);
1220: PetscObjectComposeFunction((PetscObject)A,"MatRetrieveValues_C",NULL);
1221: PetscObjectComposeFunction((PetscObject)A,"MatSeqBAIJSetColumnIndices_C",NULL);
1222: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqbaij_seqaij_C",NULL);
1223: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqbaij_seqsbaij_C",NULL);
1224: PetscObjectComposeFunction((PetscObject)A,"MatSeqBAIJSetPreallocation_C",NULL);
1225: PetscObjectComposeFunction((PetscObject)A,"MatSeqBAIJSetPreallocationCSR_C",NULL);
1226: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqbaij_seqbstrm_C",NULL);
1227: PetscObjectComposeFunction((PetscObject)A,"MatIsTranspose_C",NULL);
1228: return(0);
1229: }
1233: PetscErrorCode MatSetOption_SeqBAIJ(Mat A,MatOption op,PetscBool flg)
1234: {
1235: Mat_SeqBAIJ *a = (Mat_SeqBAIJ*)A->data;
1239: switch (op) {
1240: case MAT_ROW_ORIENTED:
1241: a->roworiented = flg;
1242: break;
1243: case MAT_KEEP_NONZERO_PATTERN:
1244: a->keepnonzeropattern = flg;
1245: break;
1246: case MAT_NEW_NONZERO_LOCATIONS:
1247: a->nonew = (flg ? 0 : 1);
1248: break;
1249: case MAT_NEW_NONZERO_LOCATION_ERR:
1250: a->nonew = (flg ? -1 : 0);
1251: break;
1252: case MAT_NEW_NONZERO_ALLOCATION_ERR:
1253: a->nonew = (flg ? -2 : 0);
1254: break;
1255: case MAT_UNUSED_NONZERO_LOCATION_ERR:
1256: a->nounused = (flg ? -1 : 0);
1257: break;
1258: case MAT_NEW_DIAGONALS:
1259: case MAT_IGNORE_OFF_PROC_ENTRIES:
1260: case MAT_USE_HASH_TABLE:
1261: PetscInfo1(A,"Option %s ignored\n",MatOptions[op]);
1262: break;
1263: case MAT_SPD:
1264: case MAT_SYMMETRIC:
1265: case MAT_STRUCTURALLY_SYMMETRIC:
1266: case MAT_HERMITIAN:
1267: case MAT_SYMMETRY_ETERNAL:
1268: /* These options are handled directly by MatSetOption() */
1269: break;
1270: default:
1271: SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"unknown option %d",op);
1272: }
1273: return(0);
1274: }
1278: PetscErrorCode MatGetRow_SeqBAIJ(Mat A,PetscInt row,PetscInt *nz,PetscInt **idx,PetscScalar **v)
1279: {
1280: Mat_SeqBAIJ *a = (Mat_SeqBAIJ*)A->data;
1282: PetscInt itmp,i,j,k,M,*ai,*aj,bs,bn,bp,*idx_i,bs2;
1283: MatScalar *aa,*aa_i;
1284: PetscScalar *v_i;
1287: bs = A->rmap->bs;
1288: ai = a->i;
1289: aj = a->j;
1290: aa = a->a;
1291: bs2 = a->bs2;
1293: if (row < 0 || row >= A->rmap->N) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row %D out of range", row);
1295: bn = row/bs; /* Block number */
1296: bp = row % bs; /* Block Position */
1297: M = ai[bn+1] - ai[bn];
1298: *nz = bs*M;
1300: if (v) {
1301: *v = 0;
1302: if (*nz) {
1303: PetscMalloc1((*nz),v);
1304: for (i=0; i<M; i++) { /* for each block in the block row */
1305: v_i = *v + i*bs;
1306: aa_i = aa + bs2*(ai[bn] + i);
1307: for (j=bp,k=0; j<bs2; j+=bs,k++) v_i[k] = aa_i[j];
1308: }
1309: }
1310: }
1312: if (idx) {
1313: *idx = 0;
1314: if (*nz) {
1315: PetscMalloc1((*nz),idx);
1316: for (i=0; i<M; i++) { /* for each block in the block row */
1317: idx_i = *idx + i*bs;
1318: itmp = bs*aj[ai[bn] + i];
1319: for (j=0; j<bs; j++) idx_i[j] = itmp++;
1320: }
1321: }
1322: }
1323: return(0);
1324: }
1328: PetscErrorCode MatRestoreRow_SeqBAIJ(Mat A,PetscInt row,PetscInt *nz,PetscInt **idx,PetscScalar **v)
1329: {
1333: if (idx) {PetscFree(*idx);}
1334: if (v) {PetscFree(*v);}
1335: return(0);
1336: }
1338: extern PetscErrorCode MatSetValues_SeqBAIJ(Mat,PetscInt,const PetscInt[],PetscInt,const PetscInt[],const PetscScalar[],InsertMode);
1342: PetscErrorCode MatTranspose_SeqBAIJ(Mat A,MatReuse reuse,Mat *B)
1343: {
1344: Mat_SeqBAIJ *a=(Mat_SeqBAIJ*)A->data;
1345: Mat C;
1347: PetscInt i,j,k,*aj=a->j,*ai=a->i,bs=A->rmap->bs,mbs=a->mbs,nbs=a->nbs,len,*col;
1348: PetscInt *rows,*cols,bs2=a->bs2;
1349: MatScalar *array;
1352: if (reuse == MAT_REUSE_MATRIX && A == *B && mbs != nbs) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Square matrix only for in-place");
1353: if (reuse == MAT_INITIAL_MATRIX || A == *B) {
1354: PetscCalloc1((1+nbs),&col);
1356: for (i=0; i<ai[mbs]; i++) col[aj[i]] += 1;
1357: MatCreate(PetscObjectComm((PetscObject)A),&C);
1358: MatSetSizes(C,A->cmap->n,A->rmap->N,A->cmap->n,A->rmap->N);
1359: MatSetType(C,((PetscObject)A)->type_name);
1360: MatSeqBAIJSetPreallocation_SeqBAIJ(C,bs,0,col);
1361: PetscFree(col);
1362: } else {
1363: C = *B;
1364: }
1366: array = a->a;
1367: PetscMalloc2(bs,&rows,bs,&cols);
1368: for (i=0; i<mbs; i++) {
1369: cols[0] = i*bs;
1370: for (k=1; k<bs; k++) cols[k] = cols[k-1] + 1;
1371: len = ai[i+1] - ai[i];
1372: for (j=0; j<len; j++) {
1373: rows[0] = (*aj++)*bs;
1374: for (k=1; k<bs; k++) rows[k] = rows[k-1] + 1;
1375: MatSetValues_SeqBAIJ(C,bs,rows,bs,cols,array,INSERT_VALUES);
1376: array += bs2;
1377: }
1378: }
1379: PetscFree2(rows,cols);
1381: MatAssemblyBegin(C,MAT_FINAL_ASSEMBLY);
1382: MatAssemblyEnd(C,MAT_FINAL_ASSEMBLY);
1384: if (reuse == MAT_INITIAL_MATRIX || *B != A) {
1385: *B = C;
1386: } else {
1387: MatHeaderMerge(A,C);
1388: }
1389: return(0);
1390: }
1394: PetscErrorCode MatIsTranspose_SeqBAIJ(Mat A,Mat B,PetscReal tol,PetscBool *f)
1395: {
1397: Mat Btrans;
1400: *f = PETSC_FALSE;
1401: MatTranspose_SeqBAIJ(A,MAT_INITIAL_MATRIX,&Btrans);
1402: MatEqual_SeqBAIJ(B,Btrans,f);
1403: MatDestroy(&Btrans);
1404: return(0);
1405: }
1409: static PetscErrorCode MatView_SeqBAIJ_Binary(Mat A,PetscViewer viewer)
1410: {
1411: Mat_SeqBAIJ *a = (Mat_SeqBAIJ*)A->data;
1413: PetscInt i,*col_lens,bs = A->rmap->bs,count,*jj,j,k,l,bs2=a->bs2;
1414: int fd;
1415: PetscScalar *aa;
1416: FILE *file;
1419: PetscViewerBinaryGetDescriptor(viewer,&fd);
1420: PetscMalloc1((4+A->rmap->N),&col_lens);
1421: col_lens[0] = MAT_FILE_CLASSID;
1423: col_lens[1] = A->rmap->N;
1424: col_lens[2] = A->cmap->n;
1425: col_lens[3] = a->nz*bs2;
1427: /* store lengths of each row and write (including header) to file */
1428: count = 0;
1429: for (i=0; i<a->mbs; i++) {
1430: for (j=0; j<bs; j++) {
1431: col_lens[4+count++] = bs*(a->i[i+1] - a->i[i]);
1432: }
1433: }
1434: PetscBinaryWrite(fd,col_lens,4+A->rmap->N,PETSC_INT,PETSC_TRUE);
1435: PetscFree(col_lens);
1437: /* store column indices (zero start index) */
1438: PetscMalloc1((a->nz+1)*bs2,&jj);
1439: count = 0;
1440: for (i=0; i<a->mbs; i++) {
1441: for (j=0; j<bs; j++) {
1442: for (k=a->i[i]; k<a->i[i+1]; k++) {
1443: for (l=0; l<bs; l++) {
1444: jj[count++] = bs*a->j[k] + l;
1445: }
1446: }
1447: }
1448: }
1449: PetscBinaryWrite(fd,jj,bs2*a->nz,PETSC_INT,PETSC_FALSE);
1450: PetscFree(jj);
1452: /* store nonzero values */
1453: PetscMalloc1((a->nz+1)*bs2,&aa);
1454: count = 0;
1455: for (i=0; i<a->mbs; i++) {
1456: for (j=0; j<bs; j++) {
1457: for (k=a->i[i]; k<a->i[i+1]; k++) {
1458: for (l=0; l<bs; l++) {
1459: aa[count++] = a->a[bs2*k + l*bs + j];
1460: }
1461: }
1462: }
1463: }
1464: PetscBinaryWrite(fd,aa,bs2*a->nz,PETSC_SCALAR,PETSC_FALSE);
1465: PetscFree(aa);
1467: PetscViewerBinaryGetInfoPointer(viewer,&file);
1468: if (file) {
1469: fprintf(file,"-matload_block_size %d\n",(int)A->rmap->bs);
1470: }
1471: return(0);
1472: }
1476: static PetscErrorCode MatView_SeqBAIJ_ASCII(Mat A,PetscViewer viewer)
1477: {
1478: Mat_SeqBAIJ *a = (Mat_SeqBAIJ*)A->data;
1479: PetscErrorCode ierr;
1480: PetscInt i,j,bs = A->rmap->bs,k,l,bs2=a->bs2;
1481: PetscViewerFormat format;
1484: PetscViewerGetFormat(viewer,&format);
1485: if (format == PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL) {
1486: PetscViewerASCIIPrintf(viewer," block size is %D\n",bs);
1487: } else if (format == PETSC_VIEWER_ASCII_MATLAB) {
1488: Mat aij;
1489: MatConvert(A,MATSEQAIJ,MAT_INITIAL_MATRIX,&aij);
1490: MatView(aij,viewer);
1491: MatDestroy(&aij);
1492: } else if (format == PETSC_VIEWER_ASCII_FACTOR_INFO) {
1493: return(0);
1494: } else if (format == PETSC_VIEWER_ASCII_COMMON) {
1495: PetscViewerASCIIUseTabs(viewer,PETSC_FALSE);
1496: for (i=0; i<a->mbs; i++) {
1497: for (j=0; j<bs; j++) {
1498: PetscViewerASCIIPrintf(viewer,"row %D:",i*bs+j);
1499: for (k=a->i[i]; k<a->i[i+1]; k++) {
1500: for (l=0; l<bs; l++) {
1501: #if defined(PETSC_USE_COMPLEX)
1502: if (PetscImaginaryPart(a->a[bs2*k + l*bs + j]) > 0.0 && PetscRealPart(a->a[bs2*k + l*bs + j]) != 0.0) {
1503: PetscViewerASCIIPrintf(viewer," (%D, %g + %gi) ",bs*a->j[k]+l,
1504: (double)PetscRealPart(a->a[bs2*k + l*bs + j]),(double)PetscImaginaryPart(a->a[bs2*k + l*bs + j]));
1505: } else if (PetscImaginaryPart(a->a[bs2*k + l*bs + j]) < 0.0 && PetscRealPart(a->a[bs2*k + l*bs + j]) != 0.0) {
1506: PetscViewerASCIIPrintf(viewer," (%D, %g - %gi) ",bs*a->j[k]+l,
1507: (double)PetscRealPart(a->a[bs2*k + l*bs + j]),-(double)PetscImaginaryPart(a->a[bs2*k + l*bs + j]));
1508: } else if (PetscRealPart(a->a[bs2*k + l*bs + j]) != 0.0) {
1509: PetscViewerASCIIPrintf(viewer," (%D, %g) ",bs*a->j[k]+l,(double)PetscRealPart(a->a[bs2*k + l*bs + j]));
1510: }
1511: #else
1512: if (a->a[bs2*k + l*bs + j] != 0.0) {
1513: PetscViewerASCIIPrintf(viewer," (%D, %g) ",bs*a->j[k]+l,(double)a->a[bs2*k + l*bs + j]);
1514: }
1515: #endif
1516: }
1517: }
1518: PetscViewerASCIIPrintf(viewer,"\n");
1519: }
1520: }
1521: PetscViewerASCIIUseTabs(viewer,PETSC_TRUE);
1522: } else {
1523: PetscViewerASCIIUseTabs(viewer,PETSC_FALSE);
1524: for (i=0; i<a->mbs; i++) {
1525: for (j=0; j<bs; j++) {
1526: PetscViewerASCIIPrintf(viewer,"row %D:",i*bs+j);
1527: for (k=a->i[i]; k<a->i[i+1]; k++) {
1528: for (l=0; l<bs; l++) {
1529: #if defined(PETSC_USE_COMPLEX)
1530: if (PetscImaginaryPart(a->a[bs2*k + l*bs + j]) > 0.0) {
1531: PetscViewerASCIIPrintf(viewer," (%D, %g + %g i) ",bs*a->j[k]+l,
1532: (double)PetscRealPart(a->a[bs2*k + l*bs + j]),(double)PetscImaginaryPart(a->a[bs2*k + l*bs + j]));
1533: } else if (PetscImaginaryPart(a->a[bs2*k + l*bs + j]) < 0.0) {
1534: PetscViewerASCIIPrintf(viewer," (%D, %g - %g i) ",bs*a->j[k]+l,
1535: (double)PetscRealPart(a->a[bs2*k + l*bs + j]),-(double)PetscImaginaryPart(a->a[bs2*k + l*bs + j]));
1536: } else {
1537: PetscViewerASCIIPrintf(viewer," (%D, %g) ",bs*a->j[k]+l,(double)PetscRealPart(a->a[bs2*k + l*bs + j]));
1538: }
1539: #else
1540: PetscViewerASCIIPrintf(viewer," (%D, %g) ",bs*a->j[k]+l,(double)a->a[bs2*k + l*bs + j]);
1541: #endif
1542: }
1543: }
1544: PetscViewerASCIIPrintf(viewer,"\n");
1545: }
1546: }
1547: PetscViewerASCIIUseTabs(viewer,PETSC_TRUE);
1548: }
1549: PetscViewerFlush(viewer);
1550: return(0);
1551: }
1553: #include <petscdraw.h>
1556: static PetscErrorCode MatView_SeqBAIJ_Draw_Zoom(PetscDraw draw,void *Aa)
1557: {
1558: Mat A = (Mat) Aa;
1559: Mat_SeqBAIJ *a=(Mat_SeqBAIJ*)A->data;
1560: PetscErrorCode ierr;
1561: PetscInt row,i,j,k,l,mbs=a->mbs,color,bs=A->rmap->bs,bs2=a->bs2;
1562: PetscReal xl,yl,xr,yr,x_l,x_r,y_l,y_r;
1563: MatScalar *aa;
1564: PetscViewer viewer;
1565: PetscViewerFormat format;
1568: PetscObjectQuery((PetscObject)A,"Zoomviewer",(PetscObject*)&viewer);
1569: PetscViewerGetFormat(viewer,&format);
1571: PetscDrawGetCoordinates(draw,&xl,&yl,&xr,&yr);
1573: /* loop over matrix elements drawing boxes */
1575: if (format != PETSC_VIEWER_DRAW_CONTOUR) {
1576: color = PETSC_DRAW_BLUE;
1577: for (i=0,row=0; i<mbs; i++,row+=bs) {
1578: for (j=a->i[i]; j<a->i[i+1]; j++) {
1579: y_l = A->rmap->N - row - 1.0; y_r = y_l + 1.0;
1580: x_l = a->j[j]*bs; x_r = x_l + 1.0;
1581: aa = a->a + j*bs2;
1582: for (k=0; k<bs; k++) {
1583: for (l=0; l<bs; l++) {
1584: if (PetscRealPart(*aa++) >= 0.) continue;
1585: PetscDrawRectangle(draw,x_l+k,y_l-l,x_r+k,y_r-l,color,color,color,color);
1586: }
1587: }
1588: }
1589: }
1590: color = PETSC_DRAW_CYAN;
1591: for (i=0,row=0; i<mbs; i++,row+=bs) {
1592: for (j=a->i[i]; j<a->i[i+1]; j++) {
1593: y_l = A->rmap->N - row - 1.0; y_r = y_l + 1.0;
1594: x_l = a->j[j]*bs; x_r = x_l + 1.0;
1595: aa = a->a + j*bs2;
1596: for (k=0; k<bs; k++) {
1597: for (l=0; l<bs; l++) {
1598: if (PetscRealPart(*aa++) != 0.) continue;
1599: PetscDrawRectangle(draw,x_l+k,y_l-l,x_r+k,y_r-l,color,color,color,color);
1600: }
1601: }
1602: }
1603: }
1604: color = PETSC_DRAW_RED;
1605: for (i=0,row=0; i<mbs; i++,row+=bs) {
1606: for (j=a->i[i]; j<a->i[i+1]; j++) {
1607: y_l = A->rmap->N - row - 1.0; y_r = y_l + 1.0;
1608: x_l = a->j[j]*bs; x_r = x_l + 1.0;
1609: aa = a->a + j*bs2;
1610: for (k=0; k<bs; k++) {
1611: for (l=0; l<bs; l++) {
1612: if (PetscRealPart(*aa++) <= 0.) continue;
1613: PetscDrawRectangle(draw,x_l+k,y_l-l,x_r+k,y_r-l,color,color,color,color);
1614: }
1615: }
1616: }
1617: }
1618: } else {
1619: /* use contour shading to indicate magnitude of values */
1620: /* first determine max of all nonzero values */
1621: PetscDraw popup;
1622: PetscReal scale,maxv = 0.0;
1624: for (i=0; i<a->nz*a->bs2; i++) {
1625: if (PetscAbsScalar(a->a[i]) > maxv) maxv = PetscAbsScalar(a->a[i]);
1626: }
1627: scale = (245.0 - PETSC_DRAW_BASIC_COLORS)/maxv;
1628: PetscDrawGetPopup(draw,&popup);
1629: if (popup) {
1630: PetscDrawScalePopup(popup,0.0,maxv);
1631: }
1632: for (i=0,row=0; i<mbs; i++,row+=bs) {
1633: for (j=a->i[i]; j<a->i[i+1]; j++) {
1634: y_l = A->rmap->N - row - 1.0; y_r = y_l + 1.0;
1635: x_l = a->j[j]*bs; x_r = x_l + 1.0;
1636: aa = a->a + j*bs2;
1637: for (k=0; k<bs; k++) {
1638: for (l=0; l<bs; l++) {
1639: color = PETSC_DRAW_BASIC_COLORS + (PetscInt)(scale*PetscAbsScalar(*aa++));
1640: PetscDrawRectangle(draw,x_l+k,y_l-l,x_r+k,y_r-l,color,color,color,color);
1641: }
1642: }
1643: }
1644: }
1645: }
1646: return(0);
1647: }
1651: static PetscErrorCode MatView_SeqBAIJ_Draw(Mat A,PetscViewer viewer)
1652: {
1654: PetscReal xl,yl,xr,yr,w,h;
1655: PetscDraw draw;
1656: PetscBool isnull;
1659: PetscViewerDrawGetDraw(viewer,0,&draw);
1660: PetscDrawIsNull(draw,&isnull); if (isnull) return(0);
1662: PetscObjectCompose((PetscObject)A,"Zoomviewer",(PetscObject)viewer);
1663: xr = A->cmap->n; yr = A->rmap->N; h = yr/10.0; w = xr/10.0;
1664: xr += w; yr += h; xl = -w; yl = -h;
1665: PetscDrawSetCoordinates(draw,xl,yl,xr,yr);
1666: PetscDrawZoom(draw,MatView_SeqBAIJ_Draw_Zoom,A);
1667: PetscObjectCompose((PetscObject)A,"Zoomviewer",NULL);
1668: return(0);
1669: }
1673: PetscErrorCode MatView_SeqBAIJ(Mat A,PetscViewer viewer)
1674: {
1676: PetscBool iascii,isbinary,isdraw;
1679: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&iascii);
1680: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERBINARY,&isbinary);
1681: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERDRAW,&isdraw);
1682: if (iascii) {
1683: MatView_SeqBAIJ_ASCII(A,viewer);
1684: } else if (isbinary) {
1685: MatView_SeqBAIJ_Binary(A,viewer);
1686: } else if (isdraw) {
1687: MatView_SeqBAIJ_Draw(A,viewer);
1688: } else {
1689: Mat B;
1690: MatConvert(A,MATSEQAIJ,MAT_INITIAL_MATRIX,&B);
1691: MatView(B,viewer);
1692: MatDestroy(&B);
1693: }
1694: return(0);
1695: }
1700: PetscErrorCode MatGetValues_SeqBAIJ(Mat A,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],PetscScalar v[])
1701: {
1702: Mat_SeqBAIJ *a = (Mat_SeqBAIJ*)A->data;
1703: PetscInt *rp,k,low,high,t,row,nrow,i,col,l,*aj = a->j;
1704: PetscInt *ai = a->i,*ailen = a->ilen;
1705: PetscInt brow,bcol,ridx,cidx,bs=A->rmap->bs,bs2=a->bs2;
1706: MatScalar *ap,*aa = a->a;
1709: for (k=0; k<m; k++) { /* loop over rows */
1710: row = im[k]; brow = row/bs;
1711: if (row < 0) {v += n; continue;} /* SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative row"); */
1712: if (row >= A->rmap->N) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row %D too large", row);
1713: rp = aj + ai[brow]; ap = aa + bs2*ai[brow];
1714: nrow = ailen[brow];
1715: for (l=0; l<n; l++) { /* loop over columns */
1716: if (in[l] < 0) {v++; continue;} /* SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative column"); */
1717: if (in[l] >= A->cmap->n) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Column %D too large", in[l]);
1718: col = in[l];
1719: bcol = col/bs;
1720: cidx = col%bs;
1721: ridx = row%bs;
1722: high = nrow;
1723: low = 0; /* assume unsorted */
1724: while (high-low > 5) {
1725: t = (low+high)/2;
1726: if (rp[t] > bcol) high = t;
1727: else low = t;
1728: }
1729: for (i=low; i<high; i++) {
1730: if (rp[i] > bcol) break;
1731: if (rp[i] == bcol) {
1732: *v++ = ap[bs2*i+bs*cidx+ridx];
1733: goto finished;
1734: }
1735: }
1736: *v++ = 0.0;
1737: finished:;
1738: }
1739: }
1740: return(0);
1741: }
1745: PetscErrorCode MatSetValuesBlocked_SeqBAIJ(Mat A,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],const PetscScalar v[],InsertMode is)
1746: {
1747: Mat_SeqBAIJ *a = (Mat_SeqBAIJ*)A->data;
1748: PetscInt *rp,k,low,high,t,ii,jj,row,nrow,i,col,l,rmax,N,lastcol = -1;
1749: PetscInt *imax=a->imax,*ai=a->i,*ailen=a->ilen;
1750: PetscErrorCode ierr;
1751: PetscInt *aj =a->j,nonew=a->nonew,bs2=a->bs2,bs=A->rmap->bs,stepval;
1752: PetscBool roworiented=a->roworiented;
1753: const PetscScalar *value = v;
1754: MatScalar *ap,*aa = a->a,*bap;
1757: if (roworiented) {
1758: stepval = (n-1)*bs;
1759: } else {
1760: stepval = (m-1)*bs;
1761: }
1762: for (k=0; k<m; k++) { /* loop over added rows */
1763: row = im[k];
1764: if (row < 0) continue;
1765: #if defined(PETSC_USE_DEBUG)
1766: if (row >= a->mbs) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row too large: row %D max %D",row,a->mbs-1);
1767: #endif
1768: rp = aj + ai[row];
1769: ap = aa + bs2*ai[row];
1770: rmax = imax[row];
1771: nrow = ailen[row];
1772: low = 0;
1773: high = nrow;
1774: for (l=0; l<n; l++) { /* loop over added columns */
1775: if (in[l] < 0) continue;
1776: #if defined(PETSC_USE_DEBUG)
1777: if (in[l] >= a->nbs) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Column too large: col %D max %D",in[l],a->nbs-1);
1778: #endif
1779: col = in[l];
1780: if (roworiented) {
1781: value = v + (k*(stepval+bs) + l)*bs;
1782: } else {
1783: value = v + (l*(stepval+bs) + k)*bs;
1784: }
1785: if (col <= lastcol) low = 0;
1786: else high = nrow;
1787: lastcol = col;
1788: while (high-low > 7) {
1789: t = (low+high)/2;
1790: if (rp[t] > col) high = t;
1791: else low = t;
1792: }
1793: for (i=low; i<high; i++) {
1794: if (rp[i] > col) break;
1795: if (rp[i] == col) {
1796: bap = ap + bs2*i;
1797: if (roworiented) {
1798: if (is == ADD_VALUES) {
1799: for (ii=0; ii<bs; ii++,value+=stepval) {
1800: for (jj=ii; jj<bs2; jj+=bs) {
1801: bap[jj] += *value++;
1802: }
1803: }
1804: } else {
1805: for (ii=0; ii<bs; ii++,value+=stepval) {
1806: for (jj=ii; jj<bs2; jj+=bs) {
1807: bap[jj] = *value++;
1808: }
1809: }
1810: }
1811: } else {
1812: if (is == ADD_VALUES) {
1813: for (ii=0; ii<bs; ii++,value+=bs+stepval) {
1814: for (jj=0; jj<bs; jj++) {
1815: bap[jj] += value[jj];
1816: }
1817: bap += bs;
1818: }
1819: } else {
1820: for (ii=0; ii<bs; ii++,value+=bs+stepval) {
1821: for (jj=0; jj<bs; jj++) {
1822: bap[jj] = value[jj];
1823: }
1824: bap += bs;
1825: }
1826: }
1827: }
1828: goto noinsert2;
1829: }
1830: }
1831: if (nonew == 1) goto noinsert2;
1832: if (nonew == -1) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Inserting a new nonzero (%D, %D) in the matrix", row, col);
1833: MatSeqXAIJReallocateAIJ(A,a->mbs,bs2,nrow,row,col,rmax,aa,ai,aj,rp,ap,imax,nonew,MatScalar);
1834: N = nrow++ - 1; high++;
1835: /* shift up all the later entries in this row */
1836: for (ii=N; ii>=i; ii--) {
1837: rp[ii+1] = rp[ii];
1838: PetscMemcpy(ap+bs2*(ii+1),ap+bs2*(ii),bs2*sizeof(MatScalar));
1839: }
1840: if (N >= i) {
1841: PetscMemzero(ap+bs2*i,bs2*sizeof(MatScalar));
1842: }
1843: rp[i] = col;
1844: bap = ap + bs2*i;
1845: if (roworiented) {
1846: for (ii=0; ii<bs; ii++,value+=stepval) {
1847: for (jj=ii; jj<bs2; jj+=bs) {
1848: bap[jj] = *value++;
1849: }
1850: }
1851: } else {
1852: for (ii=0; ii<bs; ii++,value+=stepval) {
1853: for (jj=0; jj<bs; jj++) {
1854: *bap++ = *value++;
1855: }
1856: }
1857: }
1858: noinsert2:;
1859: low = i;
1860: }
1861: ailen[row] = nrow;
1862: }
1863: return(0);
1864: }
1868: PetscErrorCode MatAssemblyEnd_SeqBAIJ(Mat A,MatAssemblyType mode)
1869: {
1870: Mat_SeqBAIJ *a = (Mat_SeqBAIJ*)A->data;
1871: PetscInt fshift = 0,i,j,*ai = a->i,*aj = a->j,*imax = a->imax;
1872: PetscInt m = A->rmap->N,*ip,N,*ailen = a->ilen;
1874: PetscInt mbs = a->mbs,bs2 = a->bs2,rmax = 0;
1875: MatScalar *aa = a->a,*ap;
1876: PetscReal ratio=0.6;
1879: if (mode == MAT_FLUSH_ASSEMBLY) return(0);
1881: if (m) rmax = ailen[0];
1882: for (i=1; i<mbs; i++) {
1883: /* move each row back by the amount of empty slots (fshift) before it*/
1884: fshift += imax[i-1] - ailen[i-1];
1885: rmax = PetscMax(rmax,ailen[i]);
1886: if (fshift) {
1887: ip = aj + ai[i]; ap = aa + bs2*ai[i];
1888: N = ailen[i];
1889: for (j=0; j<N; j++) {
1890: ip[j-fshift] = ip[j];
1892: PetscMemcpy(ap+(j-fshift)*bs2,ap+j*bs2,bs2*sizeof(MatScalar));
1893: }
1894: }
1895: ai[i] = ai[i-1] + ailen[i-1];
1896: }
1897: if (mbs) {
1898: fshift += imax[mbs-1] - ailen[mbs-1];
1899: ai[mbs] = ai[mbs-1] + ailen[mbs-1];
1900: }
1902: /* reset ilen and imax for each row */
1903: a->nonzerorowcnt = 0;
1904: for (i=0; i<mbs; i++) {
1905: ailen[i] = imax[i] = ai[i+1] - ai[i];
1906: a->nonzerorowcnt += ((ai[i+1] - ai[i]) > 0);
1907: }
1908: a->nz = ai[mbs];
1910: /* diagonals may have moved, so kill the diagonal pointers */
1911: a->idiagvalid = PETSC_FALSE;
1912: if (fshift && a->diag) {
1913: PetscFree(a->diag);
1914: PetscLogObjectMemory((PetscObject)A,-(mbs+1)*sizeof(PetscInt));
1915: a->diag = 0;
1916: }
1917: if (fshift && a->nounused == -1) SETERRQ4(PETSC_COMM_SELF,PETSC_ERR_PLIB, "Unused space detected in matrix: %D X %D block size %D, %D unneeded", m, A->cmap->n, A->rmap->bs, fshift*bs2);
1918: PetscInfo5(A,"Matrix size: %D X %D, block size %D; storage space: %D unneeded, %D used\n",m,A->cmap->n,A->rmap->bs,fshift*bs2,a->nz*bs2);
1919: PetscInfo1(A,"Number of mallocs during MatSetValues is %D\n",a->reallocs);
1920: PetscInfo1(A,"Most nonzeros blocks in any row is %D\n",rmax);
1922: A->info.mallocs += a->reallocs;
1923: a->reallocs = 0;
1924: A->info.nz_unneeded = (PetscReal)fshift*bs2;
1926: MatCheckCompressedRow(A,a->nonzerorowcnt,&a->compressedrow,a->i,mbs,ratio);
1927: return(0);
1928: }
1930: /*
1931: This function returns an array of flags which indicate the locations of contiguous
1932: blocks that should be zeroed. for eg: if bs = 3 and is = [0,1,2,3,5,6,7,8,9]
1933: then the resulting sizes = [3,1,1,3,1] correspondig to sets [(0,1,2),(3),(5),(6,7,8),(9)]
1934: Assume: sizes should be long enough to hold all the values.
1935: */
1938: static PetscErrorCode MatZeroRows_SeqBAIJ_Check_Blocks(PetscInt idx[],PetscInt n,PetscInt bs,PetscInt sizes[], PetscInt *bs_max)
1939: {
1940: PetscInt i,j,k,row;
1941: PetscBool flg;
1944: for (i=0,j=0; i<n; j++) {
1945: row = idx[i];
1946: if (row%bs!=0) { /* Not the begining of a block */
1947: sizes[j] = 1;
1948: i++;
1949: } else if (i+bs > n) { /* complete block doesn't exist (at idx end) */
1950: sizes[j] = 1; /* Also makes sure atleast 'bs' values exist for next else */
1951: i++;
1952: } else { /* Begining of the block, so check if the complete block exists */
1953: flg = PETSC_TRUE;
1954: for (k=1; k<bs; k++) {
1955: if (row+k != idx[i+k]) { /* break in the block */
1956: flg = PETSC_FALSE;
1957: break;
1958: }
1959: }
1960: if (flg) { /* No break in the bs */
1961: sizes[j] = bs;
1962: i += bs;
1963: } else {
1964: sizes[j] = 1;
1965: i++;
1966: }
1967: }
1968: }
1969: *bs_max = j;
1970: return(0);
1971: }
1975: PetscErrorCode MatZeroRows_SeqBAIJ(Mat A,PetscInt is_n,const PetscInt is_idx[],PetscScalar diag,Vec x, Vec b)
1976: {
1977: Mat_SeqBAIJ *baij=(Mat_SeqBAIJ*)A->data;
1978: PetscErrorCode ierr;
1979: PetscInt i,j,k,count,*rows;
1980: PetscInt bs=A->rmap->bs,bs2=baij->bs2,*sizes,row,bs_max;
1981: PetscScalar zero = 0.0;
1982: MatScalar *aa;
1983: const PetscScalar *xx;
1984: PetscScalar *bb;
1987: /* fix right hand side if needed */
1988: if (x && b) {
1989: VecGetArrayRead(x,&xx);
1990: VecGetArray(b,&bb);
1991: for (i=0; i<is_n; i++) {
1992: bb[is_idx[i]] = diag*xx[is_idx[i]];
1993: }
1994: VecRestoreArrayRead(x,&xx);
1995: VecRestoreArray(b,&bb);
1996: }
1998: /* Make a copy of the IS and sort it */
1999: /* allocate memory for rows,sizes */
2000: PetscMalloc2(is_n,&rows,2*is_n,&sizes);
2002: /* copy IS values to rows, and sort them */
2003: for (i=0; i<is_n; i++) rows[i] = is_idx[i];
2004: PetscSortInt(is_n,rows);
2006: if (baij->keepnonzeropattern) {
2007: for (i=0; i<is_n; i++) sizes[i] = 1;
2008: bs_max = is_n;
2009: } else {
2010: MatZeroRows_SeqBAIJ_Check_Blocks(rows,is_n,bs,sizes,&bs_max);
2011: A->nonzerostate++;
2012: }
2014: for (i=0,j=0; i<bs_max; j+=sizes[i],i++) {
2015: row = rows[j];
2016: if (row < 0 || row > A->rmap->N) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"row %D out of range",row);
2017: count = (baij->i[row/bs +1] - baij->i[row/bs])*bs;
2018: aa = ((MatScalar*)(baij->a)) + baij->i[row/bs]*bs2 + (row%bs);
2019: if (sizes[i] == bs && !baij->keepnonzeropattern) {
2020: if (diag != (PetscScalar)0.0) {
2021: if (baij->ilen[row/bs] > 0) {
2022: baij->ilen[row/bs] = 1;
2023: baij->j[baij->i[row/bs]] = row/bs;
2025: PetscMemzero(aa,count*bs*sizeof(MatScalar));
2026: }
2027: /* Now insert all the diagonal values for this bs */
2028: for (k=0; k<bs; k++) {
2029: (*A->ops->setvalues)(A,1,rows+j+k,1,rows+j+k,&diag,INSERT_VALUES);
2030: }
2031: } else { /* (diag == 0.0) */
2032: baij->ilen[row/bs] = 0;
2033: } /* end (diag == 0.0) */
2034: } else { /* (sizes[i] != bs) */
2035: #if defined(PETSC_USE_DEBUG)
2036: if (sizes[i] != 1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Internal Error. Value should be 1");
2037: #endif
2038: for (k=0; k<count; k++) {
2039: aa[0] = zero;
2040: aa += bs;
2041: }
2042: if (diag != (PetscScalar)0.0) {
2043: (*A->ops->setvalues)(A,1,rows+j,1,rows+j,&diag,INSERT_VALUES);
2044: }
2045: }
2046: }
2048: PetscFree2(rows,sizes);
2049: MatAssemblyEnd_SeqBAIJ(A,MAT_FINAL_ASSEMBLY);
2050: return(0);
2051: }
2055: PetscErrorCode MatZeroRowsColumns_SeqBAIJ(Mat A,PetscInt is_n,const PetscInt is_idx[],PetscScalar diag,Vec x, Vec b)
2056: {
2057: Mat_SeqBAIJ *baij=(Mat_SeqBAIJ*)A->data;
2058: PetscErrorCode ierr;
2059: PetscInt i,j,k,count;
2060: PetscInt bs =A->rmap->bs,bs2=baij->bs2,row,col;
2061: PetscScalar zero = 0.0;
2062: MatScalar *aa;
2063: const PetscScalar *xx;
2064: PetscScalar *bb;
2065: PetscBool *zeroed,vecs = PETSC_FALSE;
2068: /* fix right hand side if needed */
2069: if (x && b) {
2070: VecGetArrayRead(x,&xx);
2071: VecGetArray(b,&bb);
2072: vecs = PETSC_TRUE;
2073: }
2075: /* zero the columns */
2076: PetscCalloc1(A->rmap->n,&zeroed);
2077: for (i=0; i<is_n; i++) {
2078: if (is_idx[i] < 0 || is_idx[i] >= A->rmap->N) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"row %D out of range",is_idx[i]);
2079: zeroed[is_idx[i]] = PETSC_TRUE;
2080: }
2081: for (i=0; i<A->rmap->N; i++) {
2082: if (!zeroed[i]) {
2083: row = i/bs;
2084: for (j=baij->i[row]; j<baij->i[row+1]; j++) {
2085: for (k=0; k<bs; k++) {
2086: col = bs*baij->j[j] + k;
2087: if (zeroed[col]) {
2088: aa = ((MatScalar*)(baij->a)) + j*bs2 + (i%bs) + bs*k;
2089: if (vecs) bb[i] -= aa[0]*xx[col];
2090: aa[0] = 0.0;
2091: }
2092: }
2093: }
2094: } else if (vecs) bb[i] = diag*xx[i];
2095: }
2096: PetscFree(zeroed);
2097: if (vecs) {
2098: VecRestoreArrayRead(x,&xx);
2099: VecRestoreArray(b,&bb);
2100: }
2102: /* zero the rows */
2103: for (i=0; i<is_n; i++) {
2104: row = is_idx[i];
2105: count = (baij->i[row/bs +1] - baij->i[row/bs])*bs;
2106: aa = ((MatScalar*)(baij->a)) + baij->i[row/bs]*bs2 + (row%bs);
2107: for (k=0; k<count; k++) {
2108: aa[0] = zero;
2109: aa += bs;
2110: }
2111: if (diag != (PetscScalar)0.0) {
2112: (*A->ops->setvalues)(A,1,&row,1,&row,&diag,INSERT_VALUES);
2113: }
2114: }
2115: MatAssemblyEnd_SeqBAIJ(A,MAT_FINAL_ASSEMBLY);
2116: return(0);
2117: }
2121: PetscErrorCode MatSetValues_SeqBAIJ(Mat A,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],const PetscScalar v[],InsertMode is)
2122: {
2123: Mat_SeqBAIJ *a = (Mat_SeqBAIJ*)A->data;
2124: PetscInt *rp,k,low,high,t,ii,row,nrow,i,col,l,rmax,N,lastcol = -1;
2125: PetscInt *imax=a->imax,*ai=a->i,*ailen=a->ilen;
2126: PetscInt *aj =a->j,nonew=a->nonew,bs=A->rmap->bs,brow,bcol;
2128: PetscInt ridx,cidx,bs2=a->bs2;
2129: PetscBool roworiented=a->roworiented;
2130: MatScalar *ap,value,*aa=a->a,*bap;
2133: for (k=0; k<m; k++) { /* loop over added rows */
2134: row = im[k];
2135: brow = row/bs;
2136: if (row < 0) continue;
2137: #if defined(PETSC_USE_DEBUG)
2138: if (row >= A->rmap->N) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row too large: row %D max %D",row,A->rmap->N-1);
2139: #endif
2140: rp = aj + ai[brow];
2141: ap = aa + bs2*ai[brow];
2142: rmax = imax[brow];
2143: nrow = ailen[brow];
2144: low = 0;
2145: high = nrow;
2146: for (l=0; l<n; l++) { /* loop over added columns */
2147: if (in[l] < 0) continue;
2148: #if defined(PETSC_USE_DEBUG)
2149: if (in[l] >= A->cmap->n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Column too large: col %D max %D",in[l],A->cmap->n-1);
2150: #endif
2151: col = in[l]; bcol = col/bs;
2152: ridx = row % bs; cidx = col % bs;
2153: if (roworiented) {
2154: value = v[l + k*n];
2155: } else {
2156: value = v[k + l*m];
2157: }
2158: if (col <= lastcol) low = 0; else high = nrow;
2159: lastcol = col;
2160: while (high-low > 7) {
2161: t = (low+high)/2;
2162: if (rp[t] > bcol) high = t;
2163: else low = t;
2164: }
2165: for (i=low; i<high; i++) {
2166: if (rp[i] > bcol) break;
2167: if (rp[i] == bcol) {
2168: bap = ap + bs2*i + bs*cidx + ridx;
2169: if (is == ADD_VALUES) *bap += value;
2170: else *bap = value;
2171: goto noinsert1;
2172: }
2173: }
2174: if (nonew == 1) goto noinsert1;
2175: if (nonew == -1) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Inserting a new nonzero (%D, %D) in the matrix", row, col);
2176: MatSeqXAIJReallocateAIJ(A,a->mbs,bs2,nrow,brow,bcol,rmax,aa,ai,aj,rp,ap,imax,nonew,MatScalar);
2177: N = nrow++ - 1; high++;
2178: /* shift up all the later entries in this row */
2179: for (ii=N; ii>=i; ii--) {
2180: rp[ii+1] = rp[ii];
2181: PetscMemcpy(ap+bs2*(ii+1),ap+bs2*(ii),bs2*sizeof(MatScalar));
2182: }
2183: if (N>=i) {
2184: PetscMemzero(ap+bs2*i,bs2*sizeof(MatScalar));
2185: }
2186: rp[i] = bcol;
2187: ap[bs2*i + bs*cidx + ridx] = value;
2188: a->nz++;
2189: A->nonzerostate++;
2190: noinsert1:;
2191: low = i;
2192: }
2193: ailen[brow] = nrow;
2194: }
2195: return(0);
2196: }
2200: PetscErrorCode MatILUFactor_SeqBAIJ(Mat inA,IS row,IS col,const MatFactorInfo *info)
2201: {
2202: Mat_SeqBAIJ *a = (Mat_SeqBAIJ*)inA->data;
2203: Mat outA;
2205: PetscBool row_identity,col_identity;
2208: if (info->levels != 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Only levels = 0 supported for in-place ILU");
2209: ISIdentity(row,&row_identity);
2210: ISIdentity(col,&col_identity);
2211: if (!row_identity || !col_identity) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Row and column permutations must be identity for in-place ILU");
2213: outA = inA;
2214: inA->factortype = MAT_FACTOR_LU;
2216: MatMarkDiagonal_SeqBAIJ(inA);
2218: PetscObjectReference((PetscObject)row);
2219: ISDestroy(&a->row);
2220: a->row = row;
2221: PetscObjectReference((PetscObject)col);
2222: ISDestroy(&a->col);
2223: a->col = col;
2225: /* Create the invert permutation so that it can be used in MatLUFactorNumeric() */
2226: ISDestroy(&a->icol);
2227: ISInvertPermutation(col,PETSC_DECIDE,&a->icol);
2228: PetscLogObjectParent((PetscObject)inA,(PetscObject)a->icol);
2230: MatSeqBAIJSetNumericFactorization_inplace(inA,(PetscBool)(row_identity && col_identity));
2231: if (!a->solve_work) {
2232: PetscMalloc1((inA->rmap->N+inA->rmap->bs),&a->solve_work);
2233: PetscLogObjectMemory((PetscObject)inA,(inA->rmap->N+inA->rmap->bs)*sizeof(PetscScalar));
2234: }
2235: MatLUFactorNumeric(outA,inA,info);
2236: return(0);
2237: }
2241: PetscErrorCode MatSeqBAIJSetColumnIndices_SeqBAIJ(Mat mat,PetscInt *indices)
2242: {
2243: Mat_SeqBAIJ *baij = (Mat_SeqBAIJ*)mat->data;
2244: PetscInt i,nz,mbs;
2247: nz = baij->maxnz;
2248: mbs = baij->mbs;
2249: for (i=0; i<nz; i++) {
2250: baij->j[i] = indices[i];
2251: }
2252: baij->nz = nz;
2253: for (i=0; i<mbs; i++) {
2254: baij->ilen[i] = baij->imax[i];
2255: }
2256: return(0);
2257: }
2261: /*@
2262: MatSeqBAIJSetColumnIndices - Set the column indices for all the rows
2263: in the matrix.
2265: Input Parameters:
2266: + mat - the SeqBAIJ matrix
2267: - indices - the column indices
2269: Level: advanced
2271: Notes:
2272: This can be called if you have precomputed the nonzero structure of the
2273: matrix and want to provide it to the matrix object to improve the performance
2274: of the MatSetValues() operation.
2276: You MUST have set the correct numbers of nonzeros per row in the call to
2277: MatCreateSeqBAIJ(), and the columns indices MUST be sorted.
2279: MUST be called before any calls to MatSetValues();
2281: @*/
2282: PetscErrorCode MatSeqBAIJSetColumnIndices(Mat mat,PetscInt *indices)
2283: {
2289: PetscUseMethod(mat,"MatSeqBAIJSetColumnIndices_C",(Mat,PetscInt*),(mat,indices));
2290: return(0);
2291: }
2295: PetscErrorCode MatGetRowMaxAbs_SeqBAIJ(Mat A,Vec v,PetscInt idx[])
2296: {
2297: Mat_SeqBAIJ *a = (Mat_SeqBAIJ*)A->data;
2299: PetscInt i,j,n,row,bs,*ai,*aj,mbs;
2300: PetscReal atmp;
2301: PetscScalar *x,zero = 0.0;
2302: MatScalar *aa;
2303: PetscInt ncols,brow,krow,kcol;
2306: if (A->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2307: bs = A->rmap->bs;
2308: aa = a->a;
2309: ai = a->i;
2310: aj = a->j;
2311: mbs = a->mbs;
2313: VecSet(v,zero);
2314: VecGetArray(v,&x);
2315: VecGetLocalSize(v,&n);
2316: if (n != A->rmap->N) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Nonconforming matrix and vector");
2317: for (i=0; i<mbs; i++) {
2318: ncols = ai[1] - ai[0]; ai++;
2319: brow = bs*i;
2320: for (j=0; j<ncols; j++) {
2321: for (kcol=0; kcol<bs; kcol++) {
2322: for (krow=0; krow<bs; krow++) {
2323: atmp = PetscAbsScalar(*aa);aa++;
2324: row = brow + krow; /* row index */
2325: if (PetscAbsScalar(x[row]) < atmp) {x[row] = atmp; if (idx) idx[row] = bs*(*aj) + kcol;}
2326: }
2327: }
2328: aj++;
2329: }
2330: }
2331: VecRestoreArray(v,&x);
2332: return(0);
2333: }
2337: PetscErrorCode MatCopy_SeqBAIJ(Mat A,Mat B,MatStructure str)
2338: {
2342: /* If the two matrices have the same copy implementation, use fast copy. */
2343: if (str == SAME_NONZERO_PATTERN && (A->ops->copy == B->ops->copy)) {
2344: Mat_SeqBAIJ *a = (Mat_SeqBAIJ*)A->data;
2345: Mat_SeqBAIJ *b = (Mat_SeqBAIJ*)B->data;
2346: PetscInt ambs=a->mbs,bmbs=b->mbs,abs=A->rmap->bs,bbs=B->rmap->bs,bs2=abs*abs;
2348: if (a->i[ambs] != b->i[bmbs]) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_INCOMP,"Number of nonzero blocks in matrices A %D and B %D are different",a->i[ambs],b->i[bmbs]);
2349: if (abs != bbs) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_INCOMP,"Block size A %D and B %D are different",abs,bbs);
2350: PetscMemcpy(b->a,a->a,(bs2*a->i[ambs])*sizeof(PetscScalar));
2351: } else {
2352: MatCopy_Basic(A,B,str);
2353: }
2354: return(0);
2355: }
2359: PetscErrorCode MatSetUp_SeqBAIJ(Mat A)
2360: {
2364: MatSeqBAIJSetPreallocation_SeqBAIJ(A,A->rmap->bs,PETSC_DEFAULT,0);
2365: return(0);
2366: }
2370: PetscErrorCode MatSeqBAIJGetArray_SeqBAIJ(Mat A,PetscScalar *array[])
2371: {
2372: Mat_SeqBAIJ *a = (Mat_SeqBAIJ*)A->data;
2375: *array = a->a;
2376: return(0);
2377: }
2381: PetscErrorCode MatSeqBAIJRestoreArray_SeqBAIJ(Mat A,PetscScalar *array[])
2382: {
2384: return(0);
2385: }
2389: PetscErrorCode MatAXPY_SeqBAIJ(Mat Y,PetscScalar a,Mat X,MatStructure str)
2390: {
2391: Mat_SeqBAIJ *x = (Mat_SeqBAIJ*)X->data,*y = (Mat_SeqBAIJ*)Y->data;
2393: PetscInt i,bs=Y->rmap->bs,j,bs2=bs*bs;
2394: PetscBLASInt one=1;
2397: if (str == SAME_NONZERO_PATTERN) {
2398: PetscScalar alpha = a;
2399: PetscBLASInt bnz;
2400: PetscBLASIntCast(x->nz*bs2,&bnz);
2401: PetscStackCallBLAS("BLASaxpy",BLASaxpy_(&bnz,&alpha,x->a,&one,y->a,&one));
2402: PetscObjectStateIncrease((PetscObject)Y);
2403: } else if (str == SUBSET_NONZERO_PATTERN) { /* nonzeros of X is a subset of Y's */
2404: if (y->xtoy && y->XtoY != X) {
2405: PetscFree(y->xtoy);
2406: MatDestroy(&y->XtoY);
2407: }
2408: if (!y->xtoy) { /* get xtoy */
2409: MatAXPYGetxtoy_Private(x->mbs,x->i,x->j,NULL, y->i,y->j,NULL, &y->xtoy);
2410: y->XtoY = X;
2411: PetscObjectReference((PetscObject)X);
2412: }
2413: for (i=0; i<x->nz; i++) {
2414: j = 0;
2415: while (j < bs2) {
2416: y->a[bs2*y->xtoy[i]+j] += a*(x->a[bs2*i+j]);
2417: j++;
2418: }
2419: }
2420: PetscObjectStateIncrease((PetscObject)Y);
2421: PetscInfo3(Y,"ratio of nnz(X)/nnz(Y): %D/%D = %g\n",bs2*x->nz,bs2*y->nz,(double)((PetscReal)(bs2*x->nz)/(bs2*y->nz)));
2422: } else {
2423: MatAXPY_Basic(Y,a,X,str);
2424: }
2425: return(0);
2426: }
2430: PetscErrorCode MatRealPart_SeqBAIJ(Mat A)
2431: {
2432: Mat_SeqBAIJ *a = (Mat_SeqBAIJ*)A->data;
2433: PetscInt i,nz = a->bs2*a->i[a->mbs];
2434: MatScalar *aa = a->a;
2437: for (i=0; i<nz; i++) aa[i] = PetscRealPart(aa[i]);
2438: return(0);
2439: }
2443: PetscErrorCode MatImaginaryPart_SeqBAIJ(Mat A)
2444: {
2445: Mat_SeqBAIJ *a = (Mat_SeqBAIJ*)A->data;
2446: PetscInt i,nz = a->bs2*a->i[a->mbs];
2447: MatScalar *aa = a->a;
2450: for (i=0; i<nz; i++) aa[i] = PetscImaginaryPart(aa[i]);
2451: return(0);
2452: }
2456: /*
2457: Code almost idential to MatGetColumnIJ_SeqAIJ() should share common code
2458: */
2459: PetscErrorCode MatGetColumnIJ_SeqBAIJ(Mat A,PetscInt oshift,PetscBool symmetric,PetscBool inodecompressed,PetscInt *nn,const PetscInt *ia[],const PetscInt *ja[],PetscBool *done)
2460: {
2461: Mat_SeqBAIJ *a = (Mat_SeqBAIJ*)A->data;
2463: PetscInt bs = A->rmap->bs,i,*collengths,*cia,*cja,n = A->cmap->n/bs,m = A->rmap->n/bs;
2464: PetscInt nz = a->i[m],row,*jj,mr,col;
2467: *nn = n;
2468: if (!ia) return(0);
2469: if (symmetric) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Not for BAIJ matrices");
2470: else {
2471: PetscCalloc1((n+1),&collengths);
2472: PetscMalloc1((n+1),&cia);
2473: PetscMalloc1((nz+1),&cja);
2474: jj = a->j;
2475: for (i=0; i<nz; i++) {
2476: collengths[jj[i]]++;
2477: }
2478: cia[0] = oshift;
2479: for (i=0; i<n; i++) {
2480: cia[i+1] = cia[i] + collengths[i];
2481: }
2482: PetscMemzero(collengths,n*sizeof(PetscInt));
2483: jj = a->j;
2484: for (row=0; row<m; row++) {
2485: mr = a->i[row+1] - a->i[row];
2486: for (i=0; i<mr; i++) {
2487: col = *jj++;
2489: cja[cia[col] + collengths[col]++ - oshift] = row + oshift;
2490: }
2491: }
2492: PetscFree(collengths);
2493: *ia = cia; *ja = cja;
2494: }
2495: return(0);
2496: }
2500: PetscErrorCode MatRestoreColumnIJ_SeqBAIJ(Mat A,PetscInt oshift,PetscBool symmetric,PetscBool inodecompressed,PetscInt *n,const PetscInt *ia[],const PetscInt *ja[],PetscBool *done)
2501: {
2505: if (!ia) return(0);
2506: PetscFree(*ia);
2507: PetscFree(*ja);
2508: return(0);
2509: }
2511: /*
2512: MatGetColumnIJ_SeqBAIJ_Color() and MatRestoreColumnIJ_SeqBAIJ_Color() are customized from
2513: MatGetColumnIJ_SeqBAIJ() and MatRestoreColumnIJ_SeqBAIJ() by adding an output
2514: spidx[], index of a->a, to be used in MatTransposeColoringCreate() and MatFDColoringCreate()
2515: */
2518: PetscErrorCode MatGetColumnIJ_SeqBAIJ_Color(Mat A,PetscInt oshift,PetscBool symmetric,PetscBool inodecompressed,PetscInt *nn,const PetscInt *ia[],const PetscInt *ja[],PetscInt *spidx[],PetscBool *done)
2519: {
2520: Mat_SeqBAIJ *a = (Mat_SeqBAIJ*)A->data;
2522: PetscInt i,*collengths,*cia,*cja,n=a->nbs,m=a->mbs;
2523: PetscInt nz = a->i[m],row,*jj,mr,col;
2524: PetscInt *cspidx;
2527: *nn = n;
2528: if (!ia) return(0);
2530: PetscCalloc1((n+1),&collengths);
2531: PetscMalloc1((n+1),&cia);
2532: PetscMalloc1((nz+1),&cja);
2533: PetscMalloc1((nz+1),&cspidx);
2534: jj = a->j;
2535: for (i=0; i<nz; i++) {
2536: collengths[jj[i]]++;
2537: }
2538: cia[0] = oshift;
2539: for (i=0; i<n; i++) {
2540: cia[i+1] = cia[i] + collengths[i];
2541: }
2542: PetscMemzero(collengths,n*sizeof(PetscInt));
2543: jj = a->j;
2544: for (row=0; row<m; row++) {
2545: mr = a->i[row+1] - a->i[row];
2546: for (i=0; i<mr; i++) {
2547: col = *jj++;
2548: cspidx[cia[col] + collengths[col] - oshift] = a->i[row] + i; /* index of a->j */
2549: cja[cia[col] + collengths[col]++ - oshift] = row + oshift;
2550: }
2551: }
2552: PetscFree(collengths);
2553: *ia = cia; *ja = cja;
2554: *spidx = cspidx;
2555: return(0);
2556: }
2560: PetscErrorCode MatRestoreColumnIJ_SeqBAIJ_Color(Mat A,PetscInt oshift,PetscBool symmetric,PetscBool inodecompressed,PetscInt *n,const PetscInt *ia[],const PetscInt *ja[],PetscInt *spidx[],PetscBool *done)
2561: {
2565: MatRestoreColumnIJ_SeqBAIJ(A,oshift,symmetric,inodecompressed,n,ia,ja,done);
2566: PetscFree(*spidx);
2567: return(0);
2568: }
2570: /* -------------------------------------------------------------------*/
2571: static struct _MatOps MatOps_Values = {MatSetValues_SeqBAIJ,
2572: MatGetRow_SeqBAIJ,
2573: MatRestoreRow_SeqBAIJ,
2574: MatMult_SeqBAIJ_N,
2575: /* 4*/ MatMultAdd_SeqBAIJ_N,
2576: MatMultTranspose_SeqBAIJ,
2577: MatMultTransposeAdd_SeqBAIJ,
2578: 0,
2579: 0,
2580: 0,
2581: /* 10*/ 0,
2582: MatLUFactor_SeqBAIJ,
2583: 0,
2584: 0,
2585: MatTranspose_SeqBAIJ,
2586: /* 15*/ MatGetInfo_SeqBAIJ,
2587: MatEqual_SeqBAIJ,
2588: MatGetDiagonal_SeqBAIJ,
2589: MatDiagonalScale_SeqBAIJ,
2590: MatNorm_SeqBAIJ,
2591: /* 20*/ 0,
2592: MatAssemblyEnd_SeqBAIJ,
2593: MatSetOption_SeqBAIJ,
2594: MatZeroEntries_SeqBAIJ,
2595: /* 24*/ MatZeroRows_SeqBAIJ,
2596: 0,
2597: 0,
2598: 0,
2599: 0,
2600: /* 29*/ MatSetUp_SeqBAIJ,
2601: 0,
2602: 0,
2603: 0,
2604: 0,
2605: /* 34*/ MatDuplicate_SeqBAIJ,
2606: 0,
2607: 0,
2608: MatILUFactor_SeqBAIJ,
2609: 0,
2610: /* 39*/ MatAXPY_SeqBAIJ,
2611: MatGetSubMatrices_SeqBAIJ,
2612: MatIncreaseOverlap_SeqBAIJ,
2613: MatGetValues_SeqBAIJ,
2614: MatCopy_SeqBAIJ,
2615: /* 44*/ 0,
2616: MatScale_SeqBAIJ,
2617: 0,
2618: 0,
2619: MatZeroRowsColumns_SeqBAIJ,
2620: /* 49*/ 0,
2621: MatGetRowIJ_SeqBAIJ,
2622: MatRestoreRowIJ_SeqBAIJ,
2623: MatGetColumnIJ_SeqBAIJ,
2624: MatRestoreColumnIJ_SeqBAIJ,
2625: /* 54*/ MatFDColoringCreate_SeqXAIJ,
2626: 0,
2627: 0,
2628: 0,
2629: MatSetValuesBlocked_SeqBAIJ,
2630: /* 59*/ MatGetSubMatrix_SeqBAIJ,
2631: MatDestroy_SeqBAIJ,
2632: MatView_SeqBAIJ,
2633: 0,
2634: 0,
2635: /* 64*/ 0,
2636: 0,
2637: 0,
2638: 0,
2639: 0,
2640: /* 69*/ MatGetRowMaxAbs_SeqBAIJ,
2641: 0,
2642: MatConvert_Basic,
2643: 0,
2644: 0,
2645: /* 74*/ 0,
2646: MatFDColoringApply_BAIJ,
2647: 0,
2648: 0,
2649: 0,
2650: /* 79*/ 0,
2651: 0,
2652: 0,
2653: 0,
2654: MatLoad_SeqBAIJ,
2655: /* 84*/ 0,
2656: 0,
2657: 0,
2658: 0,
2659: 0,
2660: /* 89*/ 0,
2661: 0,
2662: 0,
2663: 0,
2664: 0,
2665: /* 94*/ 0,
2666: 0,
2667: 0,
2668: 0,
2669: 0,
2670: /* 99*/ 0,
2671: 0,
2672: 0,
2673: 0,
2674: 0,
2675: /*104*/ 0,
2676: MatRealPart_SeqBAIJ,
2677: MatImaginaryPart_SeqBAIJ,
2678: 0,
2679: 0,
2680: /*109*/ 0,
2681: 0,
2682: 0,
2683: 0,
2684: MatMissingDiagonal_SeqBAIJ,
2685: /*114*/ 0,
2686: 0,
2687: 0,
2688: 0,
2689: 0,
2690: /*119*/ 0,
2691: 0,
2692: MatMultHermitianTranspose_SeqBAIJ,
2693: MatMultHermitianTransposeAdd_SeqBAIJ,
2694: 0,
2695: /*124*/ 0,
2696: 0,
2697: MatInvertBlockDiagonal_SeqBAIJ,
2698: 0,
2699: 0,
2700: /*129*/ 0,
2701: 0,
2702: 0,
2703: 0,
2704: 0,
2705: /*134*/ 0,
2706: 0,
2707: 0,
2708: 0,
2709: 0,
2710: /*139*/ 0,
2711: 0,
2712: 0,
2713: MatFDColoringSetUp_SeqXAIJ
2714: };
2718: PetscErrorCode MatStoreValues_SeqBAIJ(Mat mat)
2719: {
2720: Mat_SeqBAIJ *aij = (Mat_SeqBAIJ*)mat->data;
2721: PetscInt nz = aij->i[aij->mbs]*aij->bs2;
2725: if (aij->nonew != 1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Must call MatSetOption(A,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE);first");
2727: /* allocate space for values if not already there */
2728: if (!aij->saved_values) {
2729: PetscMalloc1((nz+1),&aij->saved_values);
2730: PetscLogObjectMemory((PetscObject)mat,(nz+1)*sizeof(PetscScalar));
2731: }
2733: /* copy values over */
2734: PetscMemcpy(aij->saved_values,aij->a,nz*sizeof(PetscScalar));
2735: return(0);
2736: }
2740: PetscErrorCode MatRetrieveValues_SeqBAIJ(Mat mat)
2741: {
2742: Mat_SeqBAIJ *aij = (Mat_SeqBAIJ*)mat->data;
2744: PetscInt nz = aij->i[aij->mbs]*aij->bs2;
2747: if (aij->nonew != 1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Must call MatSetOption(A,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE);first");
2748: if (!aij->saved_values) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Must call MatStoreValues(A);first");
2750: /* copy values over */
2751: PetscMemcpy(aij->a,aij->saved_values,nz*sizeof(PetscScalar));
2752: return(0);
2753: }
2755: PETSC_EXTERN PetscErrorCode MatConvert_SeqBAIJ_SeqAIJ(Mat, MatType,MatReuse,Mat*);
2756: PETSC_EXTERN PetscErrorCode MatConvert_SeqBAIJ_SeqSBAIJ(Mat, MatType,MatReuse,Mat*);
2760: PetscErrorCode MatSeqBAIJSetPreallocation_SeqBAIJ(Mat B,PetscInt bs,PetscInt nz,PetscInt *nnz)
2761: {
2762: Mat_SeqBAIJ *b;
2764: PetscInt i,mbs,nbs,bs2;
2765: PetscBool flg,skipallocation = PETSC_FALSE,realalloc = PETSC_FALSE;
2768: if (nz >= 0 || nnz) realalloc = PETSC_TRUE;
2769: if (nz == MAT_SKIP_ALLOCATION) {
2770: skipallocation = PETSC_TRUE;
2771: nz = 0;
2772: }
2774: MatSetBlockSize(B,PetscAbs(bs));
2775: PetscLayoutSetUp(B->rmap);
2776: PetscLayoutSetUp(B->cmap);
2777: PetscLayoutGetBlockSize(B->rmap,&bs);
2779: B->preallocated = PETSC_TRUE;
2781: mbs = B->rmap->n/bs;
2782: nbs = B->cmap->n/bs;
2783: bs2 = bs*bs;
2785: if (mbs*bs!=B->rmap->n || nbs*bs!=B->cmap->n) SETERRQ3(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Number rows %D, cols %D must be divisible by blocksize %D",B->rmap->N,B->cmap->n,bs);
2787: if (nz == PETSC_DEFAULT || nz == PETSC_DECIDE) nz = 5;
2788: if (nz < 0) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"nz cannot be less than 0: value %D",nz);
2789: if (nnz) {
2790: for (i=0; i<mbs; i++) {
2791: if (nnz[i] < 0) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"nnz cannot be less than 0: local row %D value %D",i,nnz[i]);
2792: if (nnz[i] > nbs) SETERRQ3(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"nnz cannot be greater than block row length: local row %D value %D rowlength %D",i,nnz[i],nbs);
2793: }
2794: }
2796: b = (Mat_SeqBAIJ*)B->data;
2797: PetscOptionsBegin(PetscObjectComm((PetscObject)B),NULL,"Optimize options for SEQBAIJ matrix 2 ","Mat");
2798: PetscOptionsBool("-mat_no_unroll","Do not optimize for block size (slow)",NULL,PETSC_FALSE,&flg,NULL);
2799: PetscOptionsEnd();
2801: if (!flg) {
2802: switch (bs) {
2803: case 1:
2804: B->ops->mult = MatMult_SeqBAIJ_1;
2805: B->ops->multadd = MatMultAdd_SeqBAIJ_1;
2806: break;
2807: case 2:
2808: B->ops->mult = MatMult_SeqBAIJ_2;
2809: B->ops->multadd = MatMultAdd_SeqBAIJ_2;
2810: break;
2811: case 3:
2812: B->ops->mult = MatMult_SeqBAIJ_3;
2813: B->ops->multadd = MatMultAdd_SeqBAIJ_3;
2814: break;
2815: case 4:
2816: B->ops->mult = MatMult_SeqBAIJ_4;
2817: B->ops->multadd = MatMultAdd_SeqBAIJ_4;
2818: break;
2819: case 5:
2820: B->ops->mult = MatMult_SeqBAIJ_5;
2821: B->ops->multadd = MatMultAdd_SeqBAIJ_5;
2822: break;
2823: case 6:
2824: B->ops->mult = MatMult_SeqBAIJ_6;
2825: B->ops->multadd = MatMultAdd_SeqBAIJ_6;
2826: break;
2827: case 7:
2828: B->ops->mult = MatMult_SeqBAIJ_7;
2829: B->ops->multadd = MatMultAdd_SeqBAIJ_7;
2830: break;
2831: case 15:
2832: B->ops->mult = MatMult_SeqBAIJ_15_ver1;
2833: B->ops->multadd = MatMultAdd_SeqBAIJ_N;
2834: break;
2835: default:
2836: B->ops->mult = MatMult_SeqBAIJ_N;
2837: B->ops->multadd = MatMultAdd_SeqBAIJ_N;
2838: break;
2839: }
2840: }
2841: B->ops->sor = MatSOR_SeqBAIJ;
2842: b->mbs = mbs;
2843: b->nbs = nbs;
2844: if (!skipallocation) {
2845: if (!b->imax) {
2846: PetscMalloc2(mbs,&b->imax,mbs,&b->ilen);
2847: PetscLogObjectMemory((PetscObject)B,2*mbs*sizeof(PetscInt));
2849: b->free_imax_ilen = PETSC_TRUE;
2850: }
2851: /* b->ilen will count nonzeros in each block row so far. */
2852: for (i=0; i<mbs; i++) b->ilen[i] = 0;
2853: if (!nnz) {
2854: if (nz == PETSC_DEFAULT || nz == PETSC_DECIDE) nz = 5;
2855: else if (nz < 0) nz = 1;
2856: for (i=0; i<mbs; i++) b->imax[i] = nz;
2857: nz = nz*mbs;
2858: } else {
2859: nz = 0;
2860: for (i=0; i<mbs; i++) {b->imax[i] = nnz[i]; nz += nnz[i];}
2861: }
2863: /* allocate the matrix space */
2864: MatSeqXAIJFreeAIJ(B,&b->a,&b->j,&b->i);
2865: PetscMalloc3(bs2*nz,&b->a,nz,&b->j,B->rmap->N+1,&b->i);
2866: PetscLogObjectMemory((PetscObject)B,(B->rmap->N+1)*sizeof(PetscInt)+nz*(bs2*sizeof(PetscScalar)+sizeof(PetscInt)));
2867: PetscMemzero(b->a,nz*bs2*sizeof(MatScalar));
2868: PetscMemzero(b->j,nz*sizeof(PetscInt));
2870: b->singlemalloc = PETSC_TRUE;
2871: b->i[0] = 0;
2872: for (i=1; i<mbs+1; i++) {
2873: b->i[i] = b->i[i-1] + b->imax[i-1];
2874: }
2875: b->free_a = PETSC_TRUE;
2876: b->free_ij = PETSC_TRUE;
2877: #if defined(PETSC_THREADCOMM_ACTIVE)
2878: MatZeroEntries_SeqBAIJ(B);
2879: #endif
2880: } else {
2881: b->free_a = PETSC_FALSE;
2882: b->free_ij = PETSC_FALSE;
2883: }
2885: b->bs2 = bs2;
2886: b->mbs = mbs;
2887: b->nz = 0;
2888: b->maxnz = nz;
2889: B->info.nz_unneeded = (PetscReal)b->maxnz*bs2;
2890: if (realalloc) {MatSetOption(B,MAT_NEW_NONZERO_ALLOCATION_ERR,PETSC_TRUE);}
2891: return(0);
2892: }
2896: PetscErrorCode MatSeqBAIJSetPreallocationCSR_SeqBAIJ(Mat B,PetscInt bs,const PetscInt ii[],const PetscInt jj[],const PetscScalar V[])
2897: {
2898: PetscInt i,m,nz,nz_max=0,*nnz;
2899: PetscScalar *values=0;
2900: PetscBool roworiented = ((Mat_SeqBAIJ*)B->data)->roworiented;
2904: if (bs < 1) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Invalid block size specified, must be positive but it is %D",bs);
2905: PetscLayoutSetBlockSize(B->rmap,bs);
2906: PetscLayoutSetBlockSize(B->cmap,bs);
2907: PetscLayoutSetUp(B->rmap);
2908: PetscLayoutSetUp(B->cmap);
2909: PetscLayoutGetBlockSize(B->rmap,&bs);
2910: m = B->rmap->n/bs;
2912: if (ii[0] != 0) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE, "ii[0] must be 0 but it is %D",ii[0]);
2913: PetscMalloc1((m+1), &nnz);
2914: for (i=0; i<m; i++) {
2915: nz = ii[i+1]- ii[i];
2916: if (nz < 0) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE, "Local row %D has a negative number of columns %D",i,nz);
2917: nz_max = PetscMax(nz_max, nz);
2918: nnz[i] = nz;
2919: }
2920: MatSeqBAIJSetPreallocation(B,bs,0,nnz);
2921: PetscFree(nnz);
2923: values = (PetscScalar*)V;
2924: if (!values) {
2925: PetscCalloc1(bs*bs*(nz_max+1),&values);
2926: }
2927: for (i=0; i<m; i++) {
2928: PetscInt ncols = ii[i+1] - ii[i];
2929: const PetscInt *icols = jj + ii[i];
2930: const PetscScalar *svals = values + (V ? (bs*bs*ii[i]) : 0);
2931: if (!roworiented) {
2932: MatSetValuesBlocked_SeqBAIJ(B,1,&i,ncols,icols,svals,INSERT_VALUES);
2933: } else {
2934: PetscInt j;
2935: for (j=0; j<ncols; j++) {
2936: const PetscScalar *svals = values + (V ? (bs*bs*(ii[i]+j)) : 0);
2937: MatSetValuesBlocked_SeqBAIJ(B,1,&i,1,&icols[j],svals,INSERT_VALUES);
2938: }
2939: }
2940: }
2941: if (!V) { PetscFree(values); }
2942: MatAssemblyBegin(B,MAT_FINAL_ASSEMBLY);
2943: MatAssemblyEnd(B,MAT_FINAL_ASSEMBLY);
2944: MatSetOption(B,MAT_NEW_NONZERO_LOCATION_ERR,PETSC_TRUE);
2945: return(0);
2946: }
2948: PETSC_EXTERN PetscErrorCode MatGetFactor_seqbaij_petsc(Mat,MatFactorType,Mat*);
2949: PETSC_EXTERN PetscErrorCode MatGetFactor_seqbaij_bstrm(Mat,MatFactorType,Mat*);
2950: #if defined(PETSC_HAVE_MUMPS)
2951: PETSC_EXTERN PetscErrorCode MatGetFactor_baij_mumps(Mat,MatFactorType,Mat*);
2952: #endif
2953: extern PetscErrorCode MatGetFactorAvailable_seqbaij_petsc(Mat,MatFactorType,PetscBool*);
2955: /*MC
2956: MATSEQBAIJ - MATSEQBAIJ = "seqbaij" - A matrix type to be used for sequential block sparse matrices, based on
2957: block sparse compressed row format.
2959: Options Database Keys:
2960: . -mat_type seqbaij - sets the matrix type to "seqbaij" during a call to MatSetFromOptions()
2962: Level: beginner
2964: .seealso: MatCreateSeqBAIJ()
2965: M*/
2967: PETSC_EXTERN PetscErrorCode MatConvert_SeqBAIJ_SeqBSTRM(Mat, MatType,MatReuse,Mat*);
2971: PETSC_EXTERN PetscErrorCode MatCreate_SeqBAIJ(Mat B)
2972: {
2974: PetscMPIInt size;
2975: Mat_SeqBAIJ *b;
2978: MPI_Comm_size(PetscObjectComm((PetscObject)B),&size);
2979: if (size > 1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Comm must be of size 1");
2981: PetscNewLog(B,&b);
2982: B->data = (void*)b;
2983: PetscMemcpy(B->ops,&MatOps_Values,sizeof(struct _MatOps));
2985: b->row = 0;
2986: b->col = 0;
2987: b->icol = 0;
2988: b->reallocs = 0;
2989: b->saved_values = 0;
2991: b->roworiented = PETSC_TRUE;
2992: b->nonew = 0;
2993: b->diag = 0;
2994: b->solve_work = 0;
2995: b->mult_work = 0;
2996: B->spptr = 0;
2997: B->info.nz_unneeded = (PetscReal)b->maxnz*b->bs2;
2998: b->keepnonzeropattern = PETSC_FALSE;
2999: b->xtoy = 0;
3000: b->XtoY = 0;
3002: PetscObjectComposeFunction((PetscObject)B,"MatGetFactorAvailable_petsc_C",MatGetFactorAvailable_seqbaij_petsc);
3003: PetscObjectComposeFunction((PetscObject)B,"MatGetFactor_petsc_C",MatGetFactor_seqbaij_petsc);
3004: PetscObjectComposeFunction((PetscObject)B,"MatGetFactor_bstrm_C",MatGetFactor_seqbaij_bstrm);
3005: #if defined(PETSC_HAVE_MUMPS)
3006: PetscObjectComposeFunction((PetscObject)B,"MatGetFactor_mumps_C", MatGetFactor_baij_mumps);
3007: #endif
3008: PetscObjectComposeFunction((PetscObject)B,"MatInvertBlockDiagonal_C",MatInvertBlockDiagonal_SeqBAIJ);
3009: PetscObjectComposeFunction((PetscObject)B,"MatStoreValues_C",MatStoreValues_SeqBAIJ);
3010: PetscObjectComposeFunction((PetscObject)B,"MatRetrieveValues_C",MatRetrieveValues_SeqBAIJ);
3011: PetscObjectComposeFunction((PetscObject)B,"MatSeqBAIJSetColumnIndices_C",MatSeqBAIJSetColumnIndices_SeqBAIJ);
3012: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqbaij_seqaij_C",MatConvert_SeqBAIJ_SeqAIJ);
3013: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqbaij_seqsbaij_C",MatConvert_SeqBAIJ_SeqSBAIJ);
3014: PetscObjectComposeFunction((PetscObject)B,"MatSeqBAIJSetPreallocation_C",MatSeqBAIJSetPreallocation_SeqBAIJ);
3015: PetscObjectComposeFunction((PetscObject)B,"MatSeqBAIJSetPreallocationCSR_C",MatSeqBAIJSetPreallocationCSR_SeqBAIJ);
3016: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqbaij_seqbstrm_C",MatConvert_SeqBAIJ_SeqBSTRM);
3017: PetscObjectComposeFunction((PetscObject)B,"MatIsTranspose_C",MatIsTranspose_SeqBAIJ);
3018: PetscObjectChangeTypeName((PetscObject)B,MATSEQBAIJ);
3019: return(0);
3020: }
3024: PetscErrorCode MatDuplicateNoCreate_SeqBAIJ(Mat C,Mat A,MatDuplicateOption cpvalues,PetscBool mallocmatspace)
3025: {
3026: Mat_SeqBAIJ *c = (Mat_SeqBAIJ*)C->data,*a = (Mat_SeqBAIJ*)A->data;
3028: PetscInt i,mbs = a->mbs,nz = a->nz,bs2 = a->bs2;
3031: if (a->i[mbs] != nz) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Corrupt matrix");
3033: if (cpvalues == MAT_SHARE_NONZERO_PATTERN) {
3034: c->imax = a->imax;
3035: c->ilen = a->ilen;
3036: c->free_imax_ilen = PETSC_FALSE;
3037: } else {
3038: PetscMalloc2(mbs,&c->imax,mbs,&c->ilen);
3039: PetscLogObjectMemory((PetscObject)C,2*mbs*sizeof(PetscInt));
3040: for (i=0; i<mbs; i++) {
3041: c->imax[i] = a->imax[i];
3042: c->ilen[i] = a->ilen[i];
3043: }
3044: c->free_imax_ilen = PETSC_TRUE;
3045: }
3047: /* allocate the matrix space */
3048: if (mallocmatspace) {
3049: if (cpvalues == MAT_SHARE_NONZERO_PATTERN) {
3050: PetscCalloc1(bs2*nz,&c->a);
3051: PetscLogObjectMemory((PetscObject)C,a->i[mbs]*bs2*sizeof(PetscScalar));
3053: c->i = a->i;
3054: c->j = a->j;
3055: c->singlemalloc = PETSC_FALSE;
3056: c->free_a = PETSC_TRUE;
3057: c->free_ij = PETSC_FALSE;
3058: c->parent = A;
3059: C->preallocated = PETSC_TRUE;
3060: C->assembled = PETSC_TRUE;
3062: PetscObjectReference((PetscObject)A);
3063: MatSetOption(A,MAT_NEW_NONZERO_LOCATION_ERR,PETSC_TRUE);
3064: MatSetOption(C,MAT_NEW_NONZERO_LOCATION_ERR,PETSC_TRUE);
3065: } else {
3066: PetscMalloc3(bs2*nz,&c->a,nz,&c->j,mbs+1,&c->i);
3067: PetscLogObjectMemory((PetscObject)C,a->i[mbs]*(bs2*sizeof(PetscScalar)+sizeof(PetscInt))+(mbs+1)*sizeof(PetscInt));
3069: c->singlemalloc = PETSC_TRUE;
3070: c->free_a = PETSC_TRUE;
3071: c->free_ij = PETSC_TRUE;
3073: PetscMemcpy(c->i,a->i,(mbs+1)*sizeof(PetscInt));
3074: if (mbs > 0) {
3075: PetscMemcpy(c->j,a->j,nz*sizeof(PetscInt));
3076: if (cpvalues == MAT_COPY_VALUES) {
3077: PetscMemcpy(c->a,a->a,bs2*nz*sizeof(MatScalar));
3078: } else {
3079: PetscMemzero(c->a,bs2*nz*sizeof(MatScalar));
3080: }
3081: }
3082: C->preallocated = PETSC_TRUE;
3083: C->assembled = PETSC_TRUE;
3084: }
3085: }
3087: c->roworiented = a->roworiented;
3088: c->nonew = a->nonew;
3090: PetscLayoutReference(A->rmap,&C->rmap);
3091: PetscLayoutReference(A->cmap,&C->cmap);
3093: c->bs2 = a->bs2;
3094: c->mbs = a->mbs;
3095: c->nbs = a->nbs;
3097: if (a->diag) {
3098: if (cpvalues == MAT_SHARE_NONZERO_PATTERN) {
3099: c->diag = a->diag;
3100: c->free_diag = PETSC_FALSE;
3101: } else {
3102: PetscMalloc1((mbs+1),&c->diag);
3103: PetscLogObjectMemory((PetscObject)C,(mbs+1)*sizeof(PetscInt));
3104: for (i=0; i<mbs; i++) c->diag[i] = a->diag[i];
3105: c->free_diag = PETSC_TRUE;
3106: }
3107: } else c->diag = 0;
3109: c->nz = a->nz;
3110: c->maxnz = a->nz; /* Since we allocate exactly the right amount */
3111: c->solve_work = 0;
3112: c->mult_work = 0;
3114: c->compressedrow.use = a->compressedrow.use;
3115: c->compressedrow.nrows = a->compressedrow.nrows;
3116: if (a->compressedrow.use) {
3117: i = a->compressedrow.nrows;
3118: PetscMalloc2(i+1,&c->compressedrow.i,i+1,&c->compressedrow.rindex);
3119: PetscLogObjectMemory((PetscObject)C,(2*i+1)*sizeof(PetscInt));
3120: PetscMemcpy(c->compressedrow.i,a->compressedrow.i,(i+1)*sizeof(PetscInt));
3121: PetscMemcpy(c->compressedrow.rindex,a->compressedrow.rindex,i*sizeof(PetscInt));
3122: } else {
3123: c->compressedrow.use = PETSC_FALSE;
3124: c->compressedrow.i = NULL;
3125: c->compressedrow.rindex = NULL;
3126: }
3127: C->nonzerostate = A->nonzerostate;
3129: PetscFunctionListDuplicate(((PetscObject)A)->qlist,&((PetscObject)C)->qlist);
3130: PetscMemcpy(C->ops,A->ops,sizeof(struct _MatOps));
3131: return(0);
3132: }
3136: PetscErrorCode MatDuplicate_SeqBAIJ(Mat A,MatDuplicateOption cpvalues,Mat *B)
3137: {
3141: MatCreate(PetscObjectComm((PetscObject)A),B);
3142: MatSetSizes(*B,A->rmap->N,A->cmap->n,A->rmap->N,A->cmap->n);
3143: MatSetType(*B,MATSEQBAIJ);
3144: MatDuplicateNoCreate_SeqBAIJ(*B,A,cpvalues,PETSC_TRUE);
3145: return(0);
3146: }
3150: PetscErrorCode MatLoad_SeqBAIJ(Mat newmat,PetscViewer viewer)
3151: {
3152: Mat_SeqBAIJ *a;
3154: PetscInt i,nz,header[4],*rowlengths=0,M,N,bs=1;
3155: PetscInt *mask,mbs,*jj,j,rowcount,nzcount,k,*browlengths,maskcount;
3156: PetscInt kmax,jcount,block,idx,point,nzcountb,extra_rows,rows,cols;
3157: PetscInt *masked,nmask,tmp,bs2,ishift;
3158: PetscMPIInt size;
3159: int fd;
3160: PetscScalar *aa;
3161: MPI_Comm comm;
3164: PetscObjectGetComm((PetscObject)viewer,&comm);
3165: PetscOptionsBegin(comm,NULL,"Options for loading SEQBAIJ matrix","Mat");
3166: PetscOptionsInt("-matload_block_size","Set the blocksize used to store the matrix","MatLoad",bs,&bs,NULL);
3167: PetscOptionsEnd();
3168: bs2 = bs*bs;
3170: MPI_Comm_size(comm,&size);
3171: if (size > 1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"view must have one processor");
3172: PetscViewerBinaryGetDescriptor(viewer,&fd);
3173: PetscBinaryRead(fd,header,4,PETSC_INT);
3174: if (header[0] != MAT_FILE_CLASSID) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"not Mat object");
3175: M = header[1]; N = header[2]; nz = header[3];
3177: if (header[3] < 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"Matrix stored in special format, cannot load as SeqBAIJ");
3178: if (M != N) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Can only do square matrices");
3180: /*
3181: This code adds extra rows to make sure the number of rows is
3182: divisible by the blocksize
3183: */
3184: mbs = M/bs;
3185: extra_rows = bs - M + bs*(mbs);
3186: if (extra_rows == bs) extra_rows = 0;
3187: else mbs++;
3188: if (extra_rows) {
3189: PetscInfo(viewer,"Padding loaded matrix to match blocksize\n");
3190: }
3192: /* Set global sizes if not already set */
3193: if (newmat->rmap->n < 0 && newmat->rmap->N < 0 && newmat->cmap->n < 0 && newmat->cmap->N < 0) {
3194: MatSetSizes(newmat,PETSC_DECIDE,PETSC_DECIDE,M+extra_rows,N+extra_rows);
3195: } else { /* Check if the matrix global sizes are correct */
3196: MatGetSize(newmat,&rows,&cols);
3197: if (rows < 0 && cols < 0) { /* user might provide local size instead of global size */
3198: MatGetLocalSize(newmat,&rows,&cols);
3199: }
3200: if (M != rows || N != cols) SETERRQ4(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"Matrix in file of different length (%d, %d) than the input matrix (%d, %d)",M,N,rows,cols);
3201: }
3203: /* read in row lengths */
3204: PetscMalloc1((M+extra_rows),&rowlengths);
3205: PetscBinaryRead(fd,rowlengths,M,PETSC_INT);
3206: for (i=0; i<extra_rows; i++) rowlengths[M+i] = 1;
3208: /* read in column indices */
3209: PetscMalloc1((nz+extra_rows),&jj);
3210: PetscBinaryRead(fd,jj,nz,PETSC_INT);
3211: for (i=0; i<extra_rows; i++) jj[nz+i] = M+i;
3213: /* loop over row lengths determining block row lengths */
3214: PetscCalloc1(mbs,&browlengths);
3215: PetscMalloc2(mbs,&mask,mbs,&masked);
3216: PetscMemzero(mask,mbs*sizeof(PetscInt));
3217: rowcount = 0;
3218: nzcount = 0;
3219: for (i=0; i<mbs; i++) {
3220: nmask = 0;
3221: for (j=0; j<bs; j++) {
3222: kmax = rowlengths[rowcount];
3223: for (k=0; k<kmax; k++) {
3224: tmp = jj[nzcount++]/bs;
3225: if (!mask[tmp]) {masked[nmask++] = tmp; mask[tmp] = 1;}
3226: }
3227: rowcount++;
3228: }
3229: browlengths[i] += nmask;
3230: /* zero out the mask elements we set */
3231: for (j=0; j<nmask; j++) mask[masked[j]] = 0;
3232: }
3234: /* Do preallocation */
3235: MatSeqBAIJSetPreallocation_SeqBAIJ(newmat,bs,0,browlengths);
3236: a = (Mat_SeqBAIJ*)newmat->data;
3238: /* set matrix "i" values */
3239: a->i[0] = 0;
3240: for (i=1; i<= mbs; i++) {
3241: a->i[i] = a->i[i-1] + browlengths[i-1];
3242: a->ilen[i-1] = browlengths[i-1];
3243: }
3244: a->nz = 0;
3245: for (i=0; i<mbs; i++) a->nz += browlengths[i];
3247: /* read in nonzero values */
3248: PetscMalloc1((nz+extra_rows),&aa);
3249: PetscBinaryRead(fd,aa,nz,PETSC_SCALAR);
3250: for (i=0; i<extra_rows; i++) aa[nz+i] = 1.0;
3252: /* set "a" and "j" values into matrix */
3253: nzcount = 0; jcount = 0;
3254: for (i=0; i<mbs; i++) {
3255: nzcountb = nzcount;
3256: nmask = 0;
3257: for (j=0; j<bs; j++) {
3258: kmax = rowlengths[i*bs+j];
3259: for (k=0; k<kmax; k++) {
3260: tmp = jj[nzcount++]/bs;
3261: if (!mask[tmp]) { masked[nmask++] = tmp; mask[tmp] = 1;}
3262: }
3263: }
3264: /* sort the masked values */
3265: PetscSortInt(nmask,masked);
3267: /* set "j" values into matrix */
3268: maskcount = 1;
3269: for (j=0; j<nmask; j++) {
3270: a->j[jcount++] = masked[j];
3271: mask[masked[j]] = maskcount++;
3272: }
3273: /* set "a" values into matrix */
3274: ishift = bs2*a->i[i];
3275: for (j=0; j<bs; j++) {
3276: kmax = rowlengths[i*bs+j];
3277: for (k=0; k<kmax; k++) {
3278: tmp = jj[nzcountb]/bs;
3279: block = mask[tmp] - 1;
3280: point = jj[nzcountb] - bs*tmp;
3281: idx = ishift + bs2*block + j + bs*point;
3282: a->a[idx] = (MatScalar)aa[nzcountb++];
3283: }
3284: }
3285: /* zero out the mask elements we set */
3286: for (j=0; j<nmask; j++) mask[masked[j]] = 0;
3287: }
3288: if (jcount != a->nz) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"Bad binary matrix");
3290: PetscFree(rowlengths);
3291: PetscFree(browlengths);
3292: PetscFree(aa);
3293: PetscFree(jj);
3294: PetscFree2(mask,masked);
3296: MatAssemblyBegin(newmat,MAT_FINAL_ASSEMBLY);
3297: MatAssemblyEnd(newmat,MAT_FINAL_ASSEMBLY);
3298: return(0);
3299: }
3303: /*@C
3304: MatCreateSeqBAIJ - Creates a sparse matrix in block AIJ (block
3305: compressed row) format. For good matrix assembly performance the
3306: user should preallocate the matrix storage by setting the parameter nz
3307: (or the array nnz). By setting these parameters accurately, performance
3308: during matrix assembly can be increased by more than a factor of 50.
3310: Collective on MPI_Comm
3312: Input Parameters:
3313: + comm - MPI communicator, set to PETSC_COMM_SELF
3314: . bs - size of block
3315: . m - number of rows
3316: . n - number of columns
3317: . nz - number of nonzero blocks per block row (same for all rows)
3318: - nnz - array containing the number of nonzero blocks in the various block rows
3319: (possibly different for each block row) or NULL
3321: Output Parameter:
3322: . A - the matrix
3324: It is recommended that one use the MatCreate(), MatSetType() and/or MatSetFromOptions(),
3325: MatXXXXSetPreallocation() paradgm instead of this routine directly.
3326: [MatXXXXSetPreallocation() is, for example, MatSeqAIJSetPreallocation]
3328: Options Database Keys:
3329: . -mat_no_unroll - uses code that does not unroll the loops in the
3330: block calculations (much slower)
3331: . -mat_block_size - size of the blocks to use
3333: Level: intermediate
3335: Notes:
3336: The number of rows and columns must be divisible by blocksize.
3338: If the nnz parameter is given then the nz parameter is ignored
3340: A nonzero block is any block that as 1 or more nonzeros in it
3342: The block AIJ format is fully compatible with standard Fortran 77
3343: storage. That is, the stored row and column indices can begin at
3344: either one (as in Fortran) or zero. See the users' manual for details.
3346: Specify the preallocated storage with either nz or nnz (not both).
3347: Set nz=PETSC_DEFAULT and nnz=NULL for PETSc to control dynamic memory
3348: allocation. See Users-Manual: ch_mat for details.
3349: matrices.
3351: .seealso: MatCreate(), MatCreateSeqAIJ(), MatSetValues(), MatCreateBAIJ()
3352: @*/
3353: PetscErrorCode MatCreateSeqBAIJ(MPI_Comm comm,PetscInt bs,PetscInt m,PetscInt n,PetscInt nz,const PetscInt nnz[],Mat *A)
3354: {
3358: MatCreate(comm,A);
3359: MatSetSizes(*A,m,n,m,n);
3360: MatSetType(*A,MATSEQBAIJ);
3361: MatSeqBAIJSetPreallocation_SeqBAIJ(*A,bs,nz,(PetscInt*)nnz);
3362: return(0);
3363: }
3367: /*@C
3368: MatSeqBAIJSetPreallocation - Sets the block size and expected nonzeros
3369: per row in the matrix. For good matrix assembly performance the
3370: user should preallocate the matrix storage by setting the parameter nz
3371: (or the array nnz). By setting these parameters accurately, performance
3372: during matrix assembly can be increased by more than a factor of 50.
3374: Collective on MPI_Comm
3376: Input Parameters:
3377: + B - the matrix
3378: . bs - size of block
3379: . nz - number of block nonzeros per block row (same for all rows)
3380: - nnz - array containing the number of block nonzeros in the various block rows
3381: (possibly different for each block row) or NULL
3383: Options Database Keys:
3384: . -mat_no_unroll - uses code that does not unroll the loops in the
3385: block calculations (much slower)
3386: . -mat_block_size - size of the blocks to use
3388: Level: intermediate
3390: Notes:
3391: If the nnz parameter is given then the nz parameter is ignored
3393: You can call MatGetInfo() to get information on how effective the preallocation was;
3394: for example the fields mallocs,nz_allocated,nz_used,nz_unneeded;
3395: You can also run with the option -info and look for messages with the string
3396: malloc in them to see if additional memory allocation was needed.
3398: The block AIJ format is fully compatible with standard Fortran 77
3399: storage. That is, the stored row and column indices can begin at
3400: either one (as in Fortran) or zero. See the users' manual for details.
3402: Specify the preallocated storage with either nz or nnz (not both).
3403: Set nz=PETSC_DEFAULT and nnz=NULL for PETSc to control dynamic memory
3404: allocation. See Users-Manual: ch_mat for details.
3406: .seealso: MatCreate(), MatCreateSeqAIJ(), MatSetValues(), MatCreateBAIJ(), MatGetInfo()
3407: @*/
3408: PetscErrorCode MatSeqBAIJSetPreallocation(Mat B,PetscInt bs,PetscInt nz,const PetscInt nnz[])
3409: {
3416: PetscTryMethod(B,"MatSeqBAIJSetPreallocation_C",(Mat,PetscInt,PetscInt,const PetscInt[]),(B,bs,nz,nnz));
3417: return(0);
3418: }
3422: /*@C
3423: MatSeqBAIJSetPreallocationCSR - Allocates memory for a sparse sequential matrix in AIJ format
3424: (the default sequential PETSc format).
3426: Collective on MPI_Comm
3428: Input Parameters:
3429: + B - the matrix
3430: . i - the indices into j for the start of each local row (starts with zero)
3431: . j - the column indices for each local row (starts with zero) these must be sorted for each row
3432: - v - optional values in the matrix
3434: Level: developer
3436: Notes:
3437: The order of the entries in values is specified by the MatOption MAT_ROW_ORIENTED. For example, C programs
3438: may want to use the default MAT_ROW_ORIENTED=PETSC_TRUE and use an array v[nnz][bs][bs] where the second index is
3439: over rows within a block and the last index is over columns within a block row. Fortran programs will likely set
3440: MAT_ROW_ORIENTED=PETSC_FALSE and use a Fortran array v(bs,bs,nnz) in which the first index is over rows within a
3441: block column and the second index is over columns within a block.
3443: .keywords: matrix, aij, compressed row, sparse
3445: .seealso: MatCreate(), MatCreateSeqBAIJ(), MatSetValues(), MatSeqBAIJSetPreallocation(), MATSEQBAIJ
3446: @*/
3447: PetscErrorCode MatSeqBAIJSetPreallocationCSR(Mat B,PetscInt bs,const PetscInt i[],const PetscInt j[], const PetscScalar v[])
3448: {
3455: PetscTryMethod(B,"MatSeqBAIJSetPreallocationCSR_C",(Mat,PetscInt,const PetscInt[],const PetscInt[],const PetscScalar[]),(B,bs,i,j,v));
3456: return(0);
3457: }
3462: /*@
3463: MatCreateSeqBAIJWithArrays - Creates an sequential BAIJ matrix using matrix elements provided by the user.
3465: Collective on MPI_Comm
3467: Input Parameters:
3468: + comm - must be an MPI communicator of size 1
3469: . bs - size of block
3470: . m - number of rows
3471: . n - number of columns
3472: . i - row indices
3473: . j - column indices
3474: - a - matrix values
3476: Output Parameter:
3477: . mat - the matrix
3479: Level: advanced
3481: Notes:
3482: The i, j, and a arrays are not copied by this routine, the user must free these arrays
3483: once the matrix is destroyed
3485: You cannot set new nonzero locations into this matrix, that will generate an error.
3487: The i and j indices are 0 based
3489: When block size is greater than 1 the matrix values must be stored using the BAIJ storage format (see the BAIJ code to determine this).
3491: The order of the entries in values is the same as the block compressed sparse row storage format; that is, it is
3492: the same as a three dimensional array in Fortran values(bs,bs,nnz) that contains the first column of the first
3493: block, followed by the second column of the first block etc etc. That is, the blocks are contiguous in memory
3494: with column-major ordering within blocks.
3496: .seealso: MatCreate(), MatCreateBAIJ(), MatCreateSeqBAIJ()
3498: @*/
3499: PetscErrorCode MatCreateSeqBAIJWithArrays(MPI_Comm comm,PetscInt bs,PetscInt m,PetscInt n,PetscInt *i,PetscInt *j,PetscScalar *a,Mat *mat)
3500: {
3502: PetscInt ii;
3503: Mat_SeqBAIJ *baij;
3506: if (bs != 1) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"block size %D > 1 is not supported yet",bs);
3507: if (i[0]) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"i (row indices) must start with 0");
3509: MatCreate(comm,mat);
3510: MatSetSizes(*mat,m,n,m,n);
3511: MatSetType(*mat,MATSEQBAIJ);
3512: MatSeqBAIJSetPreallocation_SeqBAIJ(*mat,bs,MAT_SKIP_ALLOCATION,0);
3513: baij = (Mat_SeqBAIJ*)(*mat)->data;
3514: PetscMalloc2(m,&baij->imax,m,&baij->ilen);
3515: PetscLogObjectMemory((PetscObject)*mat,2*m*sizeof(PetscInt));
3517: baij->i = i;
3518: baij->j = j;
3519: baij->a = a;
3521: baij->singlemalloc = PETSC_FALSE;
3522: baij->nonew = -1; /*this indicates that inserting a new value in the matrix that generates a new nonzero is an error*/
3523: baij->free_a = PETSC_FALSE;
3524: baij->free_ij = PETSC_FALSE;
3526: for (ii=0; ii<m; ii++) {
3527: baij->ilen[ii] = baij->imax[ii] = i[ii+1] - i[ii];
3528: #if defined(PETSC_USE_DEBUG)
3529: if (i[ii+1] - i[ii] < 0) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative row length in i (row indices) row = %d length = %d",ii,i[ii+1] - i[ii]);
3530: #endif
3531: }
3532: #if defined(PETSC_USE_DEBUG)
3533: for (ii=0; ii<baij->i[m]; ii++) {
3534: if (j[ii] < 0) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative column index at location = %d index = %d",ii,j[ii]);
3535: if (j[ii] > n - 1) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Column index to large at location = %d index = %d",ii,j[ii]);
3536: }
3537: #endif
3539: MatAssemblyBegin(*mat,MAT_FINAL_ASSEMBLY);
3540: MatAssemblyEnd(*mat,MAT_FINAL_ASSEMBLY);
3541: return(0);
3542: }