Line data Source code
1 : /*
2 : - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
3 : SLEPc - Scalable Library for Eigenvalue Problem Computations
4 : Copyright (c) 2002-, Universitat Politecnica de Valencia, Spain
5 :
6 : This file is part of SLEPc.
7 : SLEPc is distributed under a 2-clause BSD license (see LICENSE).
8 : - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
9 : */
10 : /*
11 : Example based on spring problem in NLEVP collection [1]. See the parameters
12 : meaning at Example 2 in [2].
13 :
14 : [1] T. Betcke, N. J. Higham, V. Mehrmann, C. Schroder, and F. Tisseur,
15 : NLEVP: A Collection of Nonlinear Eigenvalue Problems, MIMS EPrint
16 : 2010.98, November 2010.
17 : [2] F. Tisseur, Backward error and condition of polynomial eigenvalue
18 : problems, Linear Algebra and its Applications, 309 (2000), pp. 339--361,
19 : April 2000.
20 : */
21 :
22 : static char help[] = "Test the solution of a PEP from a finite element model of "
23 : "damped mass-spring system (problem from NLEVP collection).\n\n"
24 : "The command line options are:\n"
25 : " -n <n> ... number of grid subdivisions.\n"
26 : " -mu <value> ... mass (default 1).\n"
27 : " -tau <value> ... damping constant of the dampers (default 10).\n"
28 : " -kappa <value> ... damping constant of the springs (default 5).\n"
29 : " -initv ... set an initial vector.\n\n";
30 :
31 : #include <slepcpep.h>
32 :
33 : /*
34 : Check if computed eigenvectors have unit norm
35 : */
36 53 : PetscErrorCode CheckNormalizedVectors(PEP pep)
37 : {
38 53 : PetscInt i,nconv;
39 53 : Mat A;
40 53 : Vec xr,xi;
41 53 : PetscReal error=0.0,normr;
42 : #if !defined(PETSC_USE_COMPLEX)
43 53 : PetscReal normi;
44 : #endif
45 :
46 53 : PetscFunctionBeginUser;
47 53 : PetscCall(PEPGetConverged(pep,&nconv));
48 53 : if (nconv>0) {
49 53 : PetscCall(PEPGetOperators(pep,0,&A));
50 53 : PetscCall(MatCreateVecs(A,&xr,&xi));
51 456 : for (i=0;i<nconv;i++) {
52 403 : PetscCall(PEPGetEigenpair(pep,i,NULL,NULL,xr,xi));
53 : #if defined(PETSC_USE_COMPLEX)
54 : PetscCall(VecNorm(xr,NORM_2,&normr));
55 : error = PetscMax(error,PetscAbsReal(normr-PetscRealConstant(1.0)));
56 : #else
57 403 : PetscCall(VecNormBegin(xr,NORM_2,&normr));
58 403 : PetscCall(VecNormBegin(xi,NORM_2,&normi));
59 403 : PetscCall(VecNormEnd(xr,NORM_2,&normr));
60 403 : PetscCall(VecNormEnd(xi,NORM_2,&normi));
61 403 : error = PetscMax(error,PetscAbsReal(SlepcAbsEigenvalue(normr,normi)-PetscRealConstant(1.0)));
62 : #endif
63 : }
64 53 : PetscCall(VecDestroy(&xr));
65 53 : PetscCall(VecDestroy(&xi));
66 53 : if (error>100*PETSC_MACHINE_EPSILON) PetscCall(PetscPrintf(PETSC_COMM_WORLD,"Vectors are not normalized. Error=%g\n",(double)error));
67 : }
68 53 : PetscFunctionReturn(PETSC_SUCCESS);
69 : }
70 :
71 53 : int main(int argc,char **argv)
72 : {
73 53 : Mat M,C,K,A[3]; /* problem matrices */
74 53 : PEP pep; /* polynomial eigenproblem solver context */
75 53 : PetscInt n=30,Istart,Iend,i,nev;
76 53 : PetscReal mu=1.0,tau=10.0,kappa=5.0;
77 53 : PetscBool initv=PETSC_FALSE,skipnorm=PETSC_FALSE;
78 53 : Vec IV[2];
79 :
80 53 : PetscFunctionBeginUser;
81 53 : PetscCall(SlepcInitialize(&argc,&argv,NULL,help));
82 :
83 53 : PetscCall(PetscOptionsGetInt(NULL,NULL,"-n",&n,NULL));
84 53 : PetscCall(PetscOptionsGetReal(NULL,NULL,"-mu",&mu,NULL));
85 53 : PetscCall(PetscOptionsGetReal(NULL,NULL,"-tau",&tau,NULL));
86 53 : PetscCall(PetscOptionsGetReal(NULL,NULL,"-kappa",&kappa,NULL));
87 53 : PetscCall(PetscOptionsGetBool(NULL,NULL,"-initv",&initv,NULL));
88 53 : PetscCall(PetscOptionsGetBool(NULL,NULL,"-skipnorm",&skipnorm,NULL));
89 :
90 : /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
91 : Compute the matrices that define the eigensystem, (k^2*M+k*C+K)x=0
92 : - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
93 :
94 : /* K is a tridiagonal */
95 53 : PetscCall(MatCreate(PETSC_COMM_WORLD,&K));
96 53 : PetscCall(MatSetSizes(K,PETSC_DECIDE,PETSC_DECIDE,n,n));
97 53 : PetscCall(MatSetFromOptions(K));
98 :
99 53 : PetscCall(MatGetOwnershipRange(K,&Istart,&Iend));
100 1373 : for (i=Istart;i<Iend;i++) {
101 1320 : if (i>0) PetscCall(MatSetValue(K,i,i-1,-kappa,INSERT_VALUES));
102 1320 : PetscCall(MatSetValue(K,i,i,kappa*3.0,INSERT_VALUES));
103 1320 : if (i<n-1) PetscCall(MatSetValue(K,i,i+1,-kappa,INSERT_VALUES));
104 : }
105 :
106 53 : PetscCall(MatAssemblyBegin(K,MAT_FINAL_ASSEMBLY));
107 53 : PetscCall(MatAssemblyEnd(K,MAT_FINAL_ASSEMBLY));
108 :
109 : /* C is a tridiagonal */
110 53 : PetscCall(MatCreate(PETSC_COMM_WORLD,&C));
111 53 : PetscCall(MatSetSizes(C,PETSC_DECIDE,PETSC_DECIDE,n,n));
112 53 : PetscCall(MatSetFromOptions(C));
113 :
114 53 : PetscCall(MatGetOwnershipRange(C,&Istart,&Iend));
115 1373 : for (i=Istart;i<Iend;i++) {
116 1320 : if (i>0) PetscCall(MatSetValue(C,i,i-1,-tau,INSERT_VALUES));
117 1320 : PetscCall(MatSetValue(C,i,i,tau*3.0,INSERT_VALUES));
118 1320 : if (i<n-1) PetscCall(MatSetValue(C,i,i+1,-tau,INSERT_VALUES));
119 : }
120 :
121 53 : PetscCall(MatAssemblyBegin(C,MAT_FINAL_ASSEMBLY));
122 53 : PetscCall(MatAssemblyEnd(C,MAT_FINAL_ASSEMBLY));
123 :
124 : /* M is a diagonal matrix */
125 53 : PetscCall(MatCreate(PETSC_COMM_WORLD,&M));
126 53 : PetscCall(MatSetSizes(M,PETSC_DECIDE,PETSC_DECIDE,n,n));
127 53 : PetscCall(MatSetFromOptions(M));
128 53 : PetscCall(MatGetOwnershipRange(M,&Istart,&Iend));
129 1373 : for (i=Istart;i<Iend;i++) PetscCall(MatSetValue(M,i,i,mu,INSERT_VALUES));
130 53 : PetscCall(MatAssemblyBegin(M,MAT_FINAL_ASSEMBLY));
131 53 : PetscCall(MatAssemblyEnd(M,MAT_FINAL_ASSEMBLY));
132 :
133 : /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
134 : Create the eigensolver and set various options
135 : - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
136 :
137 53 : PetscCall(PEPCreate(PETSC_COMM_WORLD,&pep));
138 53 : A[0] = K; A[1] = C; A[2] = M;
139 53 : PetscCall(PEPSetOperators(pep,3,A));
140 53 : PetscCall(PEPSetProblemType(pep,PEP_GENERAL));
141 53 : PetscCall(PEPSetTolerances(pep,PETSC_SMALL,PETSC_CURRENT));
142 53 : if (initv) { /* initial vector */
143 5 : PetscCall(MatCreateVecs(K,&IV[0],NULL));
144 5 : PetscCall(VecSetValue(IV[0],0,-1.0,INSERT_VALUES));
145 5 : PetscCall(VecSetValue(IV[0],1,0.5,INSERT_VALUES));
146 5 : PetscCall(VecAssemblyBegin(IV[0]));
147 5 : PetscCall(VecAssemblyEnd(IV[0]));
148 5 : PetscCall(MatCreateVecs(K,&IV[1],NULL));
149 5 : PetscCall(VecSetValue(IV[1],0,4.0,INSERT_VALUES));
150 5 : PetscCall(VecSetValue(IV[1],2,1.5,INSERT_VALUES));
151 5 : PetscCall(VecAssemblyBegin(IV[1]));
152 5 : PetscCall(VecAssemblyEnd(IV[1]));
153 5 : PetscCall(PEPSetInitialSpace(pep,2,IV));
154 5 : PetscCall(VecDestroy(&IV[0]));
155 5 : PetscCall(VecDestroy(&IV[1]));
156 : }
157 53 : PetscCall(PEPSetFromOptions(pep));
158 :
159 : /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
160 : Solve the eigensystem
161 : - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
162 :
163 53 : PetscCall(PEPSolve(pep));
164 53 : PetscCall(PEPGetDimensions(pep,&nev,NULL,NULL));
165 53 : PetscCall(PetscPrintf(PETSC_COMM_WORLD," Number of requested eigenvalues: %" PetscInt_FMT "\n",nev));
166 :
167 : /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
168 : Display solution and clean up
169 : - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
170 :
171 53 : PetscCall(PEPErrorView(pep,PEP_ERROR_BACKWARD,NULL));
172 53 : if (!skipnorm) PetscCall(CheckNormalizedVectors(pep));
173 53 : PetscCall(PEPDestroy(&pep));
174 53 : PetscCall(MatDestroy(&M));
175 53 : PetscCall(MatDestroy(&C));
176 53 : PetscCall(MatDestroy(&K));
177 53 : PetscCall(SlepcFinalize());
178 : return 0;
179 : }
180 :
181 : /*TEST
182 :
183 : testset:
184 : args: -pep_nev 4 -initv
185 : requires: !single
186 : output_file: output/test2_1.out
187 : test:
188 : suffix: 1
189 : args: -pep_type {{toar linear}}
190 : test:
191 : suffix: 1_toar_mgs
192 : args: -pep_type toar -bv_orthog_type mgs
193 : test:
194 : suffix: 1_qarnoldi
195 : args: -pep_type qarnoldi -bv_orthog_refine never
196 : test:
197 : suffix: 1_linear_gd
198 : args: -pep_type linear -pep_linear_eps_type gd -pep_linear_explicitmatrix
199 :
200 : testset:
201 : args: -pep_target -0.43 -pep_nev 4 -pep_ncv 20 -st_type sinvert
202 : output_file: output/test2_2.out
203 : test:
204 : suffix: 2
205 : args: -pep_type {{toar linear}}
206 : test:
207 : suffix: 2_toar_scaleboth
208 : args: -pep_type toar -pep_scale both
209 : test:
210 : suffix: 2_toar_transform
211 : args: -pep_type toar -st_transform
212 : test:
213 : suffix: 2_qarnoldi
214 : args: -pep_type qarnoldi -bv_orthog_refine always
215 : test:
216 : suffix: 2_linear_explicit
217 : args: -pep_type linear -pep_linear_explicitmatrix -pep_linear_linearization 0,1
218 : test:
219 : suffix: 2_linear_explicit_her
220 : args: -pep_type linear -pep_linear_explicitmatrix -pep_hermitian -pep_linear_linearization 0,1
221 : test:
222 : suffix: 2_stoar
223 : args: -pep_type stoar -pep_hermitian
224 : test:
225 : suffix: 2_jd
226 : args: -pep_type jd -st_type precond -pep_max_it 200 -pep_ncv 24
227 : requires: !single
228 :
229 : test:
230 : suffix: 3
231 : args: -pep_nev 12 -pep_extract {{none norm residual structured}} -pep_monitor_cancel
232 : requires: !single
233 :
234 : testset:
235 : args: -st_type sinvert -pep_target -0.43 -pep_nev 4
236 : output_file: output/test2_2.out
237 : test:
238 : suffix: 4_schur
239 : args: -pep_refine simple -pep_refine_scheme schur
240 : test:
241 : suffix: 4_mbe
242 : args: -pep_refine simple -pep_refine_scheme mbe -pep_refine_ksp_type preonly -pep_refine_pc_type lu
243 : test:
244 : suffix: 4_explicit
245 : args: -pep_refine simple -pep_refine_scheme explicit
246 : test:
247 : suffix: 4_multiple_schur
248 : args: -pep_refine multiple -pep_refine_scheme schur
249 : requires: !single
250 : test:
251 : suffix: 4_multiple_mbe
252 : args: -pep_refine multiple -pep_refine_scheme mbe -pep_refine_ksp_type preonly -pep_refine_pc_type lu -pep_refine_pc_factor_shift_type nonzero
253 : test:
254 : suffix: 4_multiple_explicit
255 : args: -pep_refine multiple -pep_refine_scheme explicit
256 : requires: !single
257 :
258 : test:
259 : suffix: 5
260 : nsize: 2
261 : args: -pep_type linear -pep_linear_explicitmatrix -pep_general -pep_target -0.43 -pep_nev 4 -pep_ncv 20 -st_type sinvert -pep_linear_st_ksp_type bcgs -pep_linear_st_pc_type bjacobi
262 : output_file: output/test2_2.out
263 :
264 : test:
265 : suffix: 6
266 : args: -pep_type linear -pep_nev 12 -pep_extract {{none norm residual}}
267 : requires: !single
268 : output_file: output/test2_3.out
269 :
270 : test:
271 : suffix: 7
272 : args: -pep_nev 12 -pep_extract {{none norm residual structured}} -pep_refine multiple
273 : requires: !single
274 : output_file: output/test2_3.out
275 :
276 : testset:
277 : args: -st_type sinvert -pep_target -0.43 -pep_nev 4 -st_transform
278 : output_file: output/test2_2.out
279 : test:
280 : suffix: 8_schur
281 : args: -pep_refine simple -pep_refine_scheme schur
282 : test:
283 : suffix: 8_mbe
284 : args: -pep_refine simple -pep_refine_scheme mbe -pep_refine_ksp_type preonly -pep_refine_pc_type lu
285 : test:
286 : suffix: 8_explicit
287 : args: -pep_refine simple -pep_refine_scheme explicit
288 : test:
289 : suffix: 8_multiple_schur
290 : args: -pep_refine multiple -pep_refine_scheme schur
291 : test:
292 : suffix: 8_multiple_mbe
293 : args: -pep_refine multiple -pep_refine_scheme mbe -pep_refine_ksp_type preonly -pep_refine_pc_type lu
294 : test:
295 : suffix: 8_multiple_explicit
296 : args: -pep_refine multiple -pep_refine_scheme explicit
297 :
298 : testset:
299 : nsize: 2
300 : args: -st_type sinvert -pep_target -0.49 -pep_nev 4 -pep_refine_partitions 2 -st_ksp_type bcgs -st_pc_type bjacobi -pep_scale diagonal -pep_scale_its 4
301 : output_file: output/test2_2.out
302 : test:
303 : suffix: 9_mbe
304 : args: -pep_refine simple -pep_refine_scheme mbe -pep_refine_ksp_type preonly -pep_refine_pc_type lu
305 : test:
306 : suffix: 9_explicit
307 : args: -pep_refine simple -pep_refine_scheme explicit
308 : test:
309 : suffix: 9_multiple_mbe
310 : args: -pep_refine multiple -pep_refine_scheme mbe -pep_refine_ksp_type preonly -pep_refine_pc_type lu
311 : requires: !single
312 : test:
313 : suffix: 9_multiple_explicit
314 : args: -pep_refine multiple -pep_refine_scheme explicit
315 : requires: !single
316 :
317 : test:
318 : suffix: 10
319 : nsize: 4
320 : args: -st_type sinvert -pep_target -0.43 -pep_nev 4 -pep_refine simple -pep_refine_scheme explicit -pep_refine_partitions 2 -st_ksp_type bcgs -st_pc_type bjacobi -pep_scale diagonal -pep_scale_its 4
321 : output_file: output/test2_2.out
322 :
323 : testset:
324 : args: -pep_nev 4 -initv -mat_type aijcusparse
325 : output_file: output/test2_1.out
326 : requires: cuda !single
327 : test:
328 : suffix: 11_cuda
329 : args: -pep_type {{toar linear}}
330 : test:
331 : suffix: 11_cuda_qarnoldi
332 : args: -pep_type qarnoldi -bv_orthog_refine never
333 : test:
334 : suffix: 11_cuda_linear_gd
335 : args: -pep_type linear -pep_linear_eps_type gd -pep_linear_explicitmatrix
336 :
337 : test:
338 : suffix: 12
339 : nsize: 2
340 : args: -pep_type jd -ds_parallel synchronized -pep_target -0.43 -pep_nev 4 -pep_ncv 20
341 : requires: !single
342 :
343 : test:
344 : suffix: 13
345 : args: -pep_nev 12 -pep_view_values draw -pep_monitor draw::draw_lg
346 : requires: x !single
347 : output_file: output/test2_3.out
348 :
349 : test:
350 : suffix: 14
351 : requires: complex double
352 : args: -pep_type ciss -rg_type ellipse -rg_ellipse_center -48.5 -rg_ellipse_radius 1.5 -pep_ciss_delta 1e-10
353 :
354 : testset:
355 : args: -pep_nev 4 -initv -mat_type aijhipsparse
356 : output_file: output/test2_1.out
357 : requires: hip !single
358 : test:
359 : suffix: 15_hip
360 : args: -pep_type {{toar linear}}
361 : test:
362 : suffix: 15_hip_qarnoldi
363 : args: -pep_type qarnoldi -bv_orthog_refine never
364 : test:
365 : suffix: 15_hip_linear_gd
366 : args: -pep_type linear -pep_linear_eps_type gd -pep_linear_explicitmatrix
367 :
368 : TEST*/
|
