Actual source code: crl.c
1: #define PETSCMAT_DLL
3: /*
4: Defines a matrix-vector product for the MATSEQAIJCRL matrix class.
5: This class is derived from the MATSEQAIJ class and retains the
6: compressed row storage (aka Yale sparse matrix format) but augments
7: it with a column oriented storage that is more efficient for
8: matrix vector products on Vector machines.
10: CRL stands for constant row length (that is the same number of columns
11: is kept (padded with zeros) for each row of the sparse matrix.
12: */
13: #include ../src/mat/impls/aij/seq/crl/crl.h
17: PetscErrorCode MatDestroy_SeqCRL(Mat A)
18: {
20: Mat_CRL *crl = (Mat_CRL *) A->spptr;
22: /* We are going to convert A back into a SEQAIJ matrix, since we are
23: * eventually going to use MatDestroy() to destroy everything
24: * that is not specific to CRL.
25: * In preparation for this, reset the operations pointers in A to
26: * their SeqAIJ versions. */
27: A->ops->assemblyend = crl->AssemblyEnd;
28: A->ops->destroy = crl->MatDestroy;
29: A->ops->duplicate = crl->MatDuplicate;
31: /* Free everything in the Mat_CRL data structure. */
32: PetscFree2(crl->acols,crl->icols);
34: /* Change the type of A back to SEQAIJ and use MatDestroy()
35: * to destroy everything that remains. */
36: PetscObjectChangeTypeName( (PetscObject)A, MATSEQAIJ);
37: /* Note that I don't call MatSetType(). I believe this is because that
38: * is only to be called when *building* a matrix. */
39: (*A->ops->destroy)(A);
40: return(0);
41: }
43: PetscErrorCode MatDuplicate_CRL(Mat A, MatDuplicateOption op, Mat *M)
44: {
46: Mat_CRL *crl = (Mat_CRL *) A->spptr;
49: (*crl->MatDuplicate)(A,op,M);
50: SETERRQ(PETSC_ERR_SUP,"Cannot duplicate CRL matrices yet");
51: return(0);
52: }
56: PetscErrorCode SeqCRL_create_crl(Mat A)
57: {
58: Mat_SeqAIJ *a = (Mat_SeqAIJ *)(A)->data;
59: Mat_CRL *crl = (Mat_CRL*) A->spptr;
60: PetscInt m = A->rmap->n; /* Number of rows in the matrix. */
61: PetscInt *aj = a->j; /* From the CSR representation; points to the beginning of each row. */
62: PetscInt i, j,rmax = a->rmax,*icols, *ilen = a->ilen;
63: MatScalar *aa = a->a;
64: PetscScalar *acols;
68: crl->nz = a->nz;
69: crl->m = A->rmap->n;
70: crl->rmax = rmax;
71: PetscMalloc2(rmax*m,PetscScalar,&crl->acols,rmax*m,PetscInt,&crl->icols);
72: acols = crl->acols;
73: icols = crl->icols;
74: for (i=0; i<m; i++) {
75: for (j=0; j<ilen[i]; j++) {
76: acols[j*m+i] = *aa++;
77: icols[j*m+i] = *aj++;
78: }
79: for (;j<rmax; j++) { /* empty column entries */
80: acols[j*m+i] = 0.0;
81: icols[j*m+i] = (j) ? icols[(j-1)*m+i] : 0; /* handle case where row is EMPTY */
82: }
83: }
84: PetscInfo2(A,"Percentage of 0's introduced for vectorized multiply %G. Rmax= %D\n",1.0-((double)a->nz)/((double)(rmax*m)),rmax);
85: return(0);
86: }
90: PetscErrorCode MatAssemblyEnd_SeqCRL(Mat A, MatAssemblyType mode)
91: {
93: Mat_CRL *crl = (Mat_CRL*) A->spptr;
94: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
97: if (mode == MAT_FLUSH_ASSEMBLY) return(0);
98:
99: /* Since a MATSEQCRL matrix is really just a MATSEQAIJ with some
100: * extra information, call the AssemblyEnd routine for a MATSEQAIJ.
101: * I'm not sure if this is the best way to do this, but it avoids
102: * a lot of code duplication.
103: * I also note that currently MATSEQCRL doesn't know anything about
104: * the Mat_CompressedRow data structure that SeqAIJ now uses when there
105: * are many zero rows. If the SeqAIJ assembly end routine decides to use
106: * this, this may break things. (Don't know... haven't looked at it.) */
107: a->inode.use = PETSC_FALSE;
108: (*crl->AssemblyEnd)(A, mode);
110: /* Now calculate the permutation and grouping information. */
111: SeqCRL_create_crl(A);
112: return(0);
113: }
115: #include ../src/inline/dot.h
119: PetscErrorCode MatMult_CRL(Mat A,Vec xx,Vec yy)
120: {
121: Mat_CRL *crl = (Mat_CRL*) A->spptr;
122: PetscInt m = crl->m; /* Number of rows in the matrix. */
123: PetscInt rmax = crl->rmax,*icols = crl->icols;
124: PetscScalar *acols = crl->acols;
126: PetscScalar *x,*y;
127: #if !defined(PETSC_USE_FORTRAN_KERNEL_MULTCRL)
128: PetscInt i,j,ii;
129: #endif
132: #if defined(PETSC_HAVE_PRAGMA_DISJOINT)
133: #pragma disjoint(*x,*y,*aa)
134: #endif
137: if (crl->xscat) {
138: VecCopy(xx,crl->xwork);
139: /* get remote values needed for local part of multiply */
140: VecScatterBegin(crl->xscat,xx,crl->fwork,INSERT_VALUES,SCATTER_FORWARD);
141: VecScatterEnd(crl->xscat,xx,crl->fwork,INSERT_VALUES,SCATTER_FORWARD);
142: xx = crl->xwork;
143: };
145: VecGetArray(xx,&x);
146: VecGetArray(yy,&y);
148: #if defined(PETSC_USE_FORTRAN_KERNEL_MULTCRL)
149: fortranmultcrl_(&m,&rmax,x,y,icols,acols);
150: #else
152: /* first column */
153: for (j=0; j<m; j++) {
154: y[j] = acols[j]*x[icols[j]];
155: }
157: /* other columns */
158: #if defined(PETSC_HAVE_CRAYC)
159: #pragma _CRI preferstream
160: #endif
161: for (i=1; i<rmax; i++) {
162: ii = i*m;
163: #if defined(PETSC_HAVE_CRAYC)
164: #pragma _CRI prefervector
165: #endif
166: for (j=0; j<m; j++) {
167: y[j] = y[j] + acols[ii+j]*x[icols[ii+j]];
168: }
169: }
170: #if defined(PETSC_HAVE_CRAYC)
171: #pragma _CRI ivdep
172: #endif
174: #endif
175: PetscLogFlops(2*crl->nz - m);
176: VecRestoreArray(xx,&x);
177: VecRestoreArray(yy,&y);
178: return(0);
179: }
182: /* MatConvert_SeqAIJ_SeqCRL converts a SeqAIJ matrix into a
183: * SeqCRL matrix. This routine is called by the MatCreate_SeqCRL()
184: * routine, but can also be used to convert an assembled SeqAIJ matrix
185: * into a SeqCRL one. */
189: PetscErrorCode MatConvert_SeqAIJ_SeqCRL(Mat A,const MatType type,MatReuse reuse,Mat *newmat)
190: {
192: Mat B = *newmat;
193: Mat_CRL *crl;
196: if (reuse == MAT_INITIAL_MATRIX) {
197: MatDuplicate(A,MAT_COPY_VALUES,&B);
198: }
200: PetscNewLog(B,Mat_CRL,&crl);
201: B->spptr = (void *) crl;
203: crl->AssemblyEnd = A->ops->assemblyend;
204: crl->MatDestroy = A->ops->destroy;
205: crl->MatDuplicate = A->ops->duplicate;
207: /* Set function pointers for methods that we inherit from AIJ but override. */
208: B->ops->duplicate = MatDuplicate_CRL;
209: B->ops->assemblyend = MatAssemblyEnd_SeqCRL;
210: B->ops->destroy = MatDestroy_SeqCRL;
211: B->ops->mult = MatMult_CRL;
213: /* If A has already been assembled, compute the permutation. */
214: if (A->assembled == PETSC_TRUE) {
215: SeqCRL_create_crl(B);
216: }
217: PetscObjectChangeTypeName((PetscObject)B,MATSEQCRL);
218: *newmat = B;
219: return(0);
220: }
226: /*@C
227: MatCreateSeqCRL - Creates a sparse matrix of type SEQCRL.
228: This type inherits from AIJ, but stores some additional
229: information that is used to allow better vectorization of
230: the matrix-vector product. At the cost of increased storage, the AIJ formatted
231: matrix can be copied to a format in which pieces of the matrix are
232: stored in ELLPACK format, allowing the vectorized matrix multiply
233: routine to use stride-1 memory accesses. As with the AIJ type, it is
234: important to preallocate matrix storage in order to get good assembly
235: performance.
236:
237: Collective on MPI_Comm
239: Input Parameters:
240: + comm - MPI communicator, set to PETSC_COMM_SELF
241: . m - number of rows
242: . n - number of columns
243: . nz - number of nonzeros per row (same for all rows)
244: - nnz - array containing the number of nonzeros in the various rows
245: (possibly different for each row) or PETSC_NULL
247: Output Parameter:
248: . A - the matrix
250: Notes:
251: If nnz is given then nz is ignored
253: Level: intermediate
255: .keywords: matrix, cray, sparse, parallel
257: .seealso: MatCreate(), MatCreateMPICSRPERM(), MatSetValues()
258: @*/
259: PetscErrorCode MatCreateSeqCRL(MPI_Comm comm,PetscInt m,PetscInt n,PetscInt nz,const PetscInt nnz[],Mat *A)
260: {
264: MatCreate(comm,A);
265: MatSetSizes(*A,m,n,m,n);
266: MatSetType(*A,MATSEQCRL);
267: MatSeqAIJSetPreallocation_SeqAIJ(*A,nz,(PetscInt*)nnz);
268: return(0);
269: }
275: PetscErrorCode MatCreate_SeqCRL(Mat A)
276: {
280: MatSetType(A,MATSEQAIJ);
281: MatConvert_SeqAIJ_SeqCRL(A,MATSEQCRL,MAT_REUSE_MATRIX,&A);
282: return(0);
283: }