Actual source code: mmaij.c
1: /*$Id: mmaij.c,v 1.59 2001/08/07 03:02:49 balay Exp $*/
3: /*
4: Support for the parallel AIJ matrix vector multiply
5: */
6: #include src/mat/impls/aij/mpi/mpiaij.h
10: int MatSetUpMultiply_MPIAIJ(Mat mat)
11: {
12: Mat_MPIAIJ *aij = (Mat_MPIAIJ*)mat->data;
13: Mat_SeqAIJ *B = (Mat_SeqAIJ*)(aij->B->data);
14: int i,j,*aj = B->j,ierr,ec = 0,*garray;
15: IS from,to;
16: Vec gvec;
17: #if defined (PETSC_USE_CTABLE)
18: PetscTable gid1_lid1;
19: PetscTablePosition tpos;
20: int gid,lid;
21: #else
22: int N = mat->N,*indices;
24: #endif
28: #if defined (PETSC_USE_CTABLE)
29: /* use a table - Mark Adams (this has not been tested with "shift") */
30: PetscTableCreate(aij->B->m,&gid1_lid1);
31: for (i=0; i<aij->B->m; i++) {
32: for (j=0; j<B->ilen[i]; j++) {
33: int data,gid1 = aj[B->i[i] + j] + 1;
34: PetscTableFind(gid1_lid1,gid1,&data);
35: if (!data) {
36: /* one based table */
37: PetscTableAdd(gid1_lid1,gid1,++ec);
38: }
39: }
40: }
41: /* form array of columns we need */
42: PetscMalloc((ec+1)*sizeof(int),&garray);
43: PetscTableGetHeadPosition(gid1_lid1,&tpos);
44: while (tpos) {
45: PetscTableGetNext(gid1_lid1,&tpos,&gid,&lid);
46: gid--;
47: lid--;
48: garray[lid] = gid;
49: }
50: PetscSortInt(ec,garray); /* sort, and rebuild */
51: PetscTableRemoveAll(gid1_lid1);
52: for (i=0; i<ec; i++) {
53: PetscTableAdd(gid1_lid1,garray[i]+1,i+1);
54: }
55: /* compact out the extra columns in B */
56: for (i=0; i<aij->B->m; i++) {
57: for (j=0; j<B->ilen[i]; j++) {
58: int gid1 = aj[B->i[i] + j] + 1;
59: PetscTableFind(gid1_lid1,gid1,&lid);
60: lid --;
61: aj[B->i[i] + j] = lid;
62: }
63: }
64: aij->B->n = aij->B->N = ec;
65: PetscTableDelete(gid1_lid1);
66: /* Mark Adams */
67: #else
68: /* For the first stab we make an array as long as the number of columns */
69: /* mark those columns that are in aij->B */
70: PetscMalloc((N+1)*sizeof(int),&indices);
71: PetscMemzero(indices,N*sizeof(int));
72: for (i=0; i<aij->B->m; i++) {
73: for (j=0; j<B->ilen[i]; j++) {
74: if (!indices[aj[B->i[i] + j] ]) ec++;
75: indices[aj[B->i[i] + j] ] = 1;
76: }
77: }
79: /* form array of columns we need */
80: PetscMalloc((ec+1)*sizeof(int),&garray);
81: ec = 0;
82: for (i=0; i<N; i++) {
83: if (indices[i]) garray[ec++] = i;
84: }
86: /* make indices now point into garray */
87: for (i=0; i<ec; i++) {
88: indices[garray[i]] = i;
89: }
91: /* compact out the extra columns in B */
92: for (i=0; i<aij->B->m; i++) {
93: for (j=0; j<B->ilen[i]; j++) {
94: aj[B->i[i] + j] = indices[aj[B->i[i] + j]];
95: }
96: }
97: aij->B->n = aij->B->N = ec;
98: PetscFree(indices);
99: #endif
100: /* create local vector that is used to scatter into */
101: VecCreateSeq(PETSC_COMM_SELF,ec,&aij->lvec);
103: /* create two temporary Index sets for build scatter gather */
104: ISCreateGeneral(mat->comm,ec,garray,&from);
105: ISCreateStride(PETSC_COMM_SELF,ec,0,1,&to);
107: /* create temporary global vector to generate scatter context */
108: /* this is inefficient, but otherwise we must do either
109: 1) save garray until the first actual scatter when the vector is known or
110: 2) have another way of generating a scatter context without a vector.*/
111: VecCreateMPI(mat->comm,mat->n,mat->N,&gvec);
113: /* generate the scatter context */
114: VecScatterCreate(gvec,from,aij->lvec,to,&aij->Mvctx);
115: PetscLogObjectParent(mat,aij->Mvctx);
116: PetscLogObjectParent(mat,aij->lvec);
117: PetscLogObjectParent(mat,from);
118: PetscLogObjectParent(mat,to);
119: aij->garray = garray;
120: PetscLogObjectMemory(mat,(ec+1)*sizeof(int));
121: ISDestroy(from);
122: ISDestroy(to);
123: VecDestroy(gvec);
124: return(0);
125: }
130: /*
131: Takes the local part of an already assembled MPIAIJ matrix
132: and disassembles it. This is to allow new nonzeros into the matrix
133: that require more communication in the matrix vector multiply.
134: Thus certain data-structures must be rebuilt.
136: Kind of slow! But that's what application programmers get when
137: they are sloppy.
138: */
139: int DisAssemble_MPIAIJ(Mat A)
140: {
141: Mat_MPIAIJ *aij = (Mat_MPIAIJ*)A->data;
142: Mat B = aij->B,Bnew;
143: Mat_SeqAIJ *Baij = (Mat_SeqAIJ*)B->data;
144: int ierr,i,j,m = B->m,n = A->N,col,ct = 0,*garray = aij->garray;
145: int *nz,ec;
146: PetscScalar v;
149: /* free stuff related to matrix-vec multiply */
150: VecGetSize(aij->lvec,&ec); /* needed for PetscLogObjectMemory below */
151: VecDestroy(aij->lvec); aij->lvec = 0;
152: VecScatterDestroy(aij->Mvctx); aij->Mvctx = 0;
153: if (aij->colmap) {
154: #if defined (PETSC_USE_CTABLE)
155: PetscTableDelete(aij->colmap);
156: aij->colmap = 0;
157: #else
158: PetscFree(aij->colmap);
159: aij->colmap = 0;
160: PetscLogObjectMemory(A,-aij->B->n*sizeof(int));
161: #endif
162: }
164: /* make sure that B is assembled so we can access its values */
165: MatAssemblyBegin(B,MAT_FINAL_ASSEMBLY);
166: MatAssemblyEnd(B,MAT_FINAL_ASSEMBLY);
168: /* invent new B and copy stuff over */
169: PetscMalloc((m+1)*sizeof(int),&nz);
170: for (i=0; i<m; i++) {
171: nz[i] = Baij->i[i+1] - Baij->i[i];
172: }
173: MatCreate(PETSC_COMM_SELF,m,n,m,n,&Bnew);
174: MatSetType(Bnew,B->type_name);
175: MatSeqAIJSetPreallocation(Bnew,0,nz);
176: PetscFree(nz);
177: for (i=0; i<m; i++) {
178: for (j=Baij->i[i]; j<Baij->i[i+1]; j++) {
179: col = garray[Baij->j[ct]];
180: v = Baij->a[ct++];
181: MatSetValues(Bnew,1,&i,1,&col,&v,B->insertmode);
182: }
183: }
184: PetscFree(aij->garray);
185: aij->garray = 0;
186: PetscLogObjectMemory(A,-ec*sizeof(int));
187: MatDestroy(B);
188: PetscLogObjectParent(A,Bnew);
189: aij->B = Bnew;
190: A->was_assembled = PETSC_FALSE;
191: return(0);
192: }
194: /* ugly stuff added for Glenn someday we should fix this up */
196: static int *auglyrmapd = 0,*auglyrmapo = 0; /* mapping from the local ordering to the "diagonal" and "off-diagonal"
197: parts of the local matrix */
198: static Vec auglydd = 0,auglyoo = 0; /* work vectors used to scale the two parts of the local matrix */
203: int MatMPIAIJDiagonalScaleLocalSetUp(Mat inA,Vec scale)
204: {
205: Mat_MPIAIJ *ina = (Mat_MPIAIJ*) inA->data; /*access private part of matrix */
206: int ierr,i,n,nt,cstart,cend,no,*garray = ina->garray,*lindices;
207: int *r_rmapd,*r_rmapo;
208:
210: MatGetOwnershipRange(inA,&cstart,&cend);
211: MatGetSize(ina->A,PETSC_NULL,&n);
212: PetscMalloc((inA->mapping->n+1)*sizeof(int),&r_rmapd);
213: PetscMemzero(r_rmapd,inA->mapping->n*sizeof(int));
214: nt = 0;
215: for (i=0; i<inA->mapping->n; i++) {
216: if (inA->mapping->indices[i] >= cstart && inA->mapping->indices[i] < cend) {
217: nt++;
218: r_rmapd[i] = inA->mapping->indices[i] + 1;
219: }
220: }
221: if (nt != n) SETERRQ2(1,"Hmm nt %d n %d",nt,n);
222: PetscMalloc((n+1)*sizeof(int),&auglyrmapd);
223: for (i=0; i<inA->mapping->n; i++) {
224: if (r_rmapd[i]){
225: auglyrmapd[(r_rmapd[i]-1)-cstart] = i;
226: }
227: }
228: PetscFree(r_rmapd);
229: VecCreateSeq(PETSC_COMM_SELF,n,&auglydd);
231: PetscMalloc((inA->N+1)*sizeof(int),&lindices);
232: PetscMemzero(lindices,inA->N*sizeof(int));
233: for (i=0; i<ina->B->n; i++) {
234: lindices[garray[i]] = i+1;
235: }
236: no = inA->mapping->n - nt;
237: PetscMalloc((inA->mapping->n+1)*sizeof(int),&r_rmapo);
238: PetscMemzero(r_rmapo,inA->mapping->n*sizeof(int));
239: nt = 0;
240: for (i=0; i<inA->mapping->n; i++) {
241: if (lindices[inA->mapping->indices[i]]) {
242: nt++;
243: r_rmapo[i] = lindices[inA->mapping->indices[i]];
244: }
245: }
246: if (nt > no) SETERRQ2(1,"Hmm nt %d no %d",nt,n);
247: PetscFree(lindices);
248: PetscMalloc((nt+1)*sizeof(int),&auglyrmapo);
249: for (i=0; i<inA->mapping->n; i++) {
250: if (r_rmapo[i]){
251: auglyrmapo[(r_rmapo[i]-1)] = i;
252: }
253: }
254: PetscFree(r_rmapo);
255: VecCreateSeq(PETSC_COMM_SELF,nt,&auglyoo);
257: return(0);
258: }
262: int MatMPIAIJDiagonalScaleLocal(Mat A,Vec scale)
263: {
264: /* This routine should really be abandoned as it duplicates MatDiagonalScaleLocal */
265: int ierr,(*f)(Mat,Vec);
268: PetscObjectQueryFunction((PetscObject)A,"MatDiagonalScaleLocal_C",(void (**)(void))&f);
269: if (f) {
270: (*f)(A,scale);
271: }
272: return(0);
273: }
275: EXTERN_C_BEGIN
278: int MatDiagonalScaleLocal_MPIAIJ(Mat A,Vec scale)
279: {
280: Mat_MPIAIJ *a = (Mat_MPIAIJ*) A->data; /*access private part of matrix */
281: int ierr,n,i;
282: PetscScalar *d,*o,*s;
283:
285: if (!auglyrmapd) {
286: MatMPIAIJDiagonalScaleLocalSetUp(A,scale);
287: }
289: VecGetArray(scale,&s);
290:
291: VecGetLocalSize(auglydd,&n);
292: VecGetArray(auglydd,&d);
293: for (i=0; i<n; i++) {
294: d[i] = s[auglyrmapd[i]]; /* copy "diagonal" (true local) portion of scale into dd vector */
295: }
296: VecRestoreArray(auglydd,&d);
297: /* column scale "diagonal" portion of local matrix */
298: MatDiagonalScale(a->A,PETSC_NULL,auglydd);
300: VecGetLocalSize(auglyoo,&n);
301: VecGetArray(auglyoo,&o);
302: for (i=0; i<n; i++) {
303: o[i] = s[auglyrmapo[i]]; /* copy "off-diagonal" portion of scale into oo vector */
304: }
305: VecRestoreArray(scale,&s);
306: VecRestoreArray(auglyoo,&o);
307: /* column scale "off-diagonal" portion of local matrix */
308: MatDiagonalScale(a->B,PETSC_NULL,auglyoo);
310: return(0);
311: }
312: EXTERN_C_END