xref: /petsc/src/mat/tests/ex214.c (revision 030f984af8d8bb4c203755d35bded3c05b3d83ce)
1 
2 static char help[] = "Tests MatMatSolve() and MatMatTransposeSolve() for computing inv(A) with MUMPS.\n\
3 Example: mpiexec -n <np> ./ex214 -displ \n\n";
4 
5 #include <petscmat.h>
6 
7 int main(int argc,char **args)
8 {
9   PetscErrorCode ierr;
10   PetscMPIInt    size,rank;
11 #if defined(PETSC_HAVE_MUMPS)
12   Mat            A,RHS,C,F,X,AX,spRHST;
13   PetscInt       m,n,nrhs,M,N,i,Istart,Iend,Ii,j,J,test;
14   PetscScalar    v;
15   PetscReal      norm,tol=PETSC_SQRT_MACHINE_EPSILON;
16   PetscRandom    rand;
17   PetscBool      displ=PETSC_FALSE;
18   char           solver[256];
19 #endif
20 
21   ierr = PetscInitialize(&argc,&args,(char*)0,help);if (ierr) return ierr;
22   ierr = MPI_Comm_size(PETSC_COMM_WORLD,&size);CHKERRMPI(ierr);
23   ierr = MPI_Comm_rank(PETSC_COMM_WORLD,&rank);CHKERRMPI(ierr);
24 
25 #if !defined(PETSC_HAVE_MUMPS)
26   if (!rank) {ierr = PetscPrintf(PETSC_COMM_SELF,"This example requires MUMPS, exit...\n");CHKERRQ(ierr);}
27   ierr = PetscFinalize();
28   return ierr;
29 #else
30 
31   ierr = PetscOptionsGetBool(NULL,NULL,"-displ",&displ,NULL);CHKERRQ(ierr);
32 
33   /* Create matrix A */
34   m = 4; n = 4;
35   ierr = PetscOptionsGetInt(NULL,NULL,"-m",&m,NULL);CHKERRQ(ierr);
36   ierr = PetscOptionsGetInt(NULL,NULL,"-n",&n,NULL);CHKERRQ(ierr);
37 
38   ierr = MatCreate(PETSC_COMM_WORLD,&A);CHKERRQ(ierr);
39   ierr = MatSetSizes(A,PETSC_DECIDE,PETSC_DECIDE,m*n,m*n);CHKERRQ(ierr);
40   ierr = MatSetFromOptions(A);CHKERRQ(ierr);
41   ierr = MatMPIAIJSetPreallocation(A,5,NULL,5,NULL);CHKERRQ(ierr);
42   ierr = MatSeqAIJSetPreallocation(A,5,NULL);CHKERRQ(ierr);
43 
44   ierr = MatGetOwnershipRange(A,&Istart,&Iend);CHKERRQ(ierr);
45   for (Ii=Istart; Ii<Iend; Ii++) {
46     v = -1.0; i = Ii/n; j = Ii - i*n;
47     if (i>0)   {J = Ii - n; ierr = MatSetValues(A,1,&Ii,1,&J,&v,ADD_VALUES);CHKERRQ(ierr);}
48     if (i<m-1) {J = Ii + n; ierr = MatSetValues(A,1,&Ii,1,&J,&v,ADD_VALUES);CHKERRQ(ierr);}
49     if (j>0)   {J = Ii - 1; ierr = MatSetValues(A,1,&Ii,1,&J,&v,ADD_VALUES);CHKERRQ(ierr);}
50     if (j<n-1) {J = Ii + 1; ierr = MatSetValues(A,1,&Ii,1,&J,&v,ADD_VALUES);CHKERRQ(ierr);}
51     v = 4.0; ierr = MatSetValues(A,1,&Ii,1,&Ii,&v,ADD_VALUES);CHKERRQ(ierr);
52   }
53   ierr = MatAssemblyBegin(A,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
54   ierr = MatAssemblyEnd(A,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
55 
56   ierr = MatGetLocalSize(A,&m,&n);CHKERRQ(ierr);
57   ierr = MatGetSize(A,&M,&N);CHKERRQ(ierr);
58   if (m != n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ, "This example is not intended for rectangular matrices (%d, %d)", m, n);
59 
60   /* Create dense matrix C and X; C holds true solution with identical colums */
61   nrhs = N;
62   ierr = PetscOptionsGetInt(NULL,NULL,"-nrhs",&nrhs,NULL);CHKERRQ(ierr);
63   ierr = MatCreate(PETSC_COMM_WORLD,&C);CHKERRQ(ierr);
64   ierr = MatSetSizes(C,m,PETSC_DECIDE,PETSC_DECIDE,nrhs);CHKERRQ(ierr);
65   ierr = MatSetType(C,MATDENSE);CHKERRQ(ierr);
66   ierr = MatSetFromOptions(C);CHKERRQ(ierr);
67   ierr = MatSetUp(C);CHKERRQ(ierr);
68 
69   ierr = PetscRandomCreate(PETSC_COMM_WORLD,&rand);CHKERRQ(ierr);
70   ierr = PetscRandomSetFromOptions(rand);CHKERRQ(ierr);
71   ierr = MatSetRandom(C,rand);CHKERRQ(ierr);
72   ierr = MatDuplicate(C,MAT_DO_NOT_COPY_VALUES,&X);CHKERRQ(ierr);
73 
74   ierr = PetscStrcpy(solver,MATSOLVERMUMPS);CHKERRQ(ierr);
75   if (!rank && displ) {ierr = PetscPrintf(PETSC_COMM_SELF,"Solving with %s: nrhs %D, size mat %D x %D\n",solver,nrhs,M,N);CHKERRQ(ierr);}
76 
77   for (test=0; test<2; test++) {
78     if (test == 0) {
79       /* Test LU Factorization */
80       ierr = PetscPrintf(PETSC_COMM_WORLD,"test LU factorization\n");CHKERRQ(ierr);
81       ierr = MatGetFactor(A,solver,MAT_FACTOR_LU,&F);CHKERRQ(ierr);
82       ierr = MatLUFactorSymbolic(F,A,NULL,NULL,NULL);CHKERRQ(ierr);
83       ierr = MatLUFactorNumeric(F,A,NULL);CHKERRQ(ierr);
84     } else {
85       /* Test Cholesky Factorization */
86       PetscBool flg;
87       ierr = MatIsSymmetric(A,0.0,&flg);CHKERRQ(ierr);
88       if (!flg) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"A must be symmetric for Cholesky factorization");
89 
90       ierr = PetscPrintf(PETSC_COMM_WORLD,"test Cholesky factorization\n");CHKERRQ(ierr);
91       ierr = MatGetFactor(A,solver,MAT_FACTOR_CHOLESKY,&F);CHKERRQ(ierr);
92       ierr = MatCholeskyFactorSymbolic(F,A,NULL,NULL);CHKERRQ(ierr);
93       ierr = MatCholeskyFactorNumeric(F,A,NULL);CHKERRQ(ierr);
94     }
95 
96     /* (1) Test MatMatSolve(): dense RHS = A*C, C: true solutions */
97     /* ---------------------------------------------------------- */
98     ierr = MatMatMult(A,C,MAT_INITIAL_MATRIX,2.0,&RHS);CHKERRQ(ierr);
99     ierr = MatMatSolve(F,RHS,X);CHKERRQ(ierr);
100     /* Check the error */
101     ierr = MatAXPY(X,-1.0,C,SAME_NONZERO_PATTERN);CHKERRQ(ierr);
102     ierr = MatNorm(X,NORM_FROBENIUS,&norm);CHKERRQ(ierr);
103     if (norm > tol) {
104       ierr = PetscPrintf(PETSC_COMM_SELF,"(1) MatMatSolve: Norm of error %g\n",norm);CHKERRQ(ierr);
105     }
106 
107     /* Test X=RHS */
108     ierr = MatMatSolve(F,RHS,RHS);CHKERRQ(ierr);
109     /* Check the error */
110     ierr = MatAXPY(RHS,-1.0,C,SAME_NONZERO_PATTERN);CHKERRQ(ierr);
111     ierr = MatNorm(RHS,NORM_FROBENIUS,&norm);CHKERRQ(ierr);
112     if (norm > tol) {
113       ierr = PetscPrintf(PETSC_COMM_SELF,"(1.1) MatMatSolve: Norm of error %g\n",norm);CHKERRQ(ierr);
114     }
115 
116     /* (2) Test MatMatSolve() for inv(A) with dense RHS:
117      RHS = [e[0],...,e[nrhs-1]], dense X holds first nrhs columns of inv(A) */
118     /* -------------------------------------------------------------------- */
119     ierr = MatZeroEntries(RHS);CHKERRQ(ierr);
120     for (i=0; i<nrhs; i++) {
121       v = 1.0;
122       ierr = MatSetValues(RHS,1,&i,1,&i,&v,INSERT_VALUES);CHKERRQ(ierr);
123     }
124     ierr = MatAssemblyBegin(RHS,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
125     ierr = MatAssemblyEnd(RHS,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
126 
127     ierr = MatMatSolve(F,RHS,X);CHKERRQ(ierr);
128     if (displ) {
129       if (!rank) {ierr = PetscPrintf(PETSC_COMM_SELF," \n(2) first %D columns of inv(A) with dense RHS:\n",nrhs);CHKERRQ(ierr);}
130       ierr = MatView(X,PETSC_VIEWER_STDOUT_WORLD);CHKERRQ(ierr);
131     }
132 
133     /* Check the residual */
134     ierr = MatMatMult(A,X,MAT_INITIAL_MATRIX,2.0,&AX);CHKERRQ(ierr);
135     ierr = MatAXPY(AX,-1.0,RHS,SAME_NONZERO_PATTERN);CHKERRQ(ierr);
136     ierr = MatNorm(AX,NORM_INFINITY,&norm);CHKERRQ(ierr);
137     if (norm > tol) {
138       ierr = PetscPrintf(PETSC_COMM_SELF,"(2) MatMatSolve: Norm of residual %g\n",norm);CHKERRQ(ierr);
139     }
140     ierr = MatZeroEntries(X);CHKERRQ(ierr);
141 
142     /* (3) Test MatMatTransposeSolve() for inv(A) with sparse RHS stored in the host:
143      spRHST = [e[0],...,e[nrhs-1]]^T, dense X holds first nrhs columns of inv(A) */
144     /* --------------------------------------------------------------------------- */
145     /* Create spRHST: PETSc does not support compressed column format which is required by MUMPS for sparse RHS matrix,
146      thus user must create spRHST=spRHS^T and call MatMatTransposeSolve() */
147     ierr = MatCreate(PETSC_COMM_WORLD,&spRHST);CHKERRQ(ierr);
148     if (!rank) {
149       /* MUMPS requirs RHS be centralized on the host! */
150       ierr = MatSetSizes(spRHST,nrhs,M,PETSC_DECIDE,PETSC_DECIDE);CHKERRQ(ierr);
151     } else {
152       ierr = MatSetSizes(spRHST,0,0,PETSC_DECIDE,PETSC_DECIDE);CHKERRQ(ierr);
153     }
154     ierr = MatSetType(spRHST,MATAIJ);CHKERRQ(ierr);
155     ierr = MatSetFromOptions(spRHST);CHKERRQ(ierr);
156     ierr = MatSetUp(spRHST);CHKERRQ(ierr);
157     if (!rank) {
158       v = 1.0;
159       for (i=0; i<nrhs; i++) {
160         ierr = MatSetValues(spRHST,1,&i,1,&i,&v,INSERT_VALUES);CHKERRQ(ierr);
161       }
162     }
163     ierr = MatAssemblyBegin(spRHST,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
164     ierr = MatAssemblyEnd(spRHST,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
165 
166     ierr = MatMatTransposeSolve(F,spRHST,X);CHKERRQ(ierr);
167 
168     if (displ) {
169       if (!rank) {ierr = PetscPrintf(PETSC_COMM_SELF," \n(3) first %D columns of inv(A) with sparse RHS:\n",nrhs);CHKERRQ(ierr);}
170       ierr = MatView(X,PETSC_VIEWER_STDOUT_WORLD);CHKERRQ(ierr);
171     }
172 
173     /* Check the residual: R = A*X - RHS */
174     ierr = MatMatMult(A,X,MAT_REUSE_MATRIX,2.0,&AX);CHKERRQ(ierr);
175 
176     ierr = MatAXPY(AX,-1.0,RHS,SAME_NONZERO_PATTERN);CHKERRQ(ierr);
177     ierr = MatNorm(AX,NORM_INFINITY,&norm);CHKERRQ(ierr);
178     if (norm > tol) {
179       ierr = PetscPrintf(PETSC_COMM_SELF,"(3) MatMatSolve: Norm of residual %g\n",norm);CHKERRQ(ierr);
180     }
181 
182     /* (4) Test MatMatSolve() for inv(A) with selected entries:
183      input: spRHS gives selected indices; output: spRHS holds selected entries of inv(A) */
184     /* --------------------------------------------------------------------------------- */
185     if (nrhs == N) { /* mumps requires nrhs = n */
186       /* Create spRHS on proc[0] */
187       Mat spRHS = NULL;
188 
189       /* Create spRHS = spRHST^T. Two matrices share internal matrix data structure */
190       ierr = MatCreateTranspose(spRHST,&spRHS);CHKERRQ(ierr);
191       ierr = MatMumpsGetInverse(F,spRHS);CHKERRQ(ierr);
192       ierr = MatDestroy(&spRHS);CHKERRQ(ierr);
193 
194       ierr = MatMumpsGetInverseTranspose(F,spRHST);CHKERRQ(ierr);
195       if (displ) {
196         ierr = PetscPrintf(PETSC_COMM_WORLD,"\nSelected entries of inv(A^T):\n");CHKERRQ(ierr);
197         ierr = MatView(spRHST,PETSC_VIEWER_STDOUT_WORLD);CHKERRQ(ierr);
198       }
199       ierr = MatDestroy(&spRHS);CHKERRQ(ierr);
200     }
201     ierr = MatDestroy(&AX);CHKERRQ(ierr);
202     ierr = MatDestroy(&F);CHKERRQ(ierr);
203     ierr = MatDestroy(&RHS);CHKERRQ(ierr);
204     ierr = MatDestroy(&spRHST);CHKERRQ(ierr);
205   }
206 
207   /* Free data structures */
208   ierr = MatDestroy(&A);CHKERRQ(ierr);
209   ierr = MatDestroy(&C);CHKERRQ(ierr);
210   ierr = MatDestroy(&X);CHKERRQ(ierr);
211   ierr = PetscRandomDestroy(&rand);CHKERRQ(ierr);
212   ierr = PetscFinalize();
213   return ierr;
214 #endif
215 }
216 
217 /*TEST
218 
219    test:
220      requires: mumps double !complex
221 
222    test:
223      suffix: 2
224      requires: mumps double !complex
225      nsize: 2
226 
227 TEST*/
228