Actual source code: test1.c

slepc-3.22.1 2024-10-28
Report Typos and Errors
  1: /*
  2:    - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
  3:    SLEPc - Scalable Library for Eigenvalue Problem Computations
  4:    Copyright (c) 2002-, Universitat Politecnica de Valencia, Spain

  6:    This file is part of SLEPc.
  7:    SLEPc is distributed under a 2-clause BSD license (see LICENSE).
  8:    - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
  9: */

 11: static char help[] = "Tests B-orthonormality of eigenvectors in a GHEP problem.\n\n";

 13: #include <slepceps.h>

 15: int main(int argc,char **argv)
 16: {
 17:   Mat               A,B;        /* matrices */
 18:   EPS               eps;        /* eigenproblem solver context */
 19:   ST                st;
 20:   Vec               *X,v;
 21:   PetscReal         lev=0.0,tol=PETSC_SMALL;
 22:   PetscInt          N,n=45,m,Istart,Iend,II,i,j,nconv;
 23:   PetscBool         flag,skiporth=PETSC_FALSE;
 24:   EPSPowerShiftType variant;

 26:   PetscFunctionBeginUser;
 27:   PetscCall(SlepcInitialize(&argc,&argv,NULL,help));
 28:   PetscCall(PetscOptionsGetInt(NULL,NULL,"-n",&n,NULL));
 29:   PetscCall(PetscOptionsGetInt(NULL,NULL,"-m",&m,&flag));
 30:   if (!flag) m=n;
 31:   N = n*m;
 32:   PetscCall(PetscPrintf(PETSC_COMM_WORLD,"\nGeneralized Symmetric Eigenproblem, N=%" PetscInt_FMT " (%" PetscInt_FMT "x%" PetscInt_FMT " grid)\n\n",N,n,m));
 33:   PetscCall(PetscOptionsGetBool(NULL,NULL,"-skiporth",&skiporth,NULL));

 35:   /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
 36:      Compute the matrices that define the eigensystem, Ax=kBx
 37:      - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */

 39:   PetscCall(MatCreate(PETSC_COMM_WORLD,&A));
 40:   PetscCall(MatSetSizes(A,PETSC_DECIDE,PETSC_DECIDE,N,N));
 41:   PetscCall(MatSetFromOptions(A));

 43:   PetscCall(MatCreate(PETSC_COMM_WORLD,&B));
 44:   PetscCall(MatSetSizes(B,PETSC_DECIDE,PETSC_DECIDE,N,N));
 45:   PetscCall(MatSetFromOptions(B));

 47:   PetscCall(MatGetOwnershipRange(A,&Istart,&Iend));
 48:   for (II=Istart;II<Iend;II++) {
 49:     i = II/n; j = II-i*n;
 50:     if (i>0) PetscCall(MatSetValue(A,II,II-n,-1.0,INSERT_VALUES));
 51:     if (i<m-1) PetscCall(MatSetValue(A,II,II+n,-1.0,INSERT_VALUES));
 52:     if (j>0) PetscCall(MatSetValue(A,II,II-1,-1.0,INSERT_VALUES));
 53:     if (j<n-1) PetscCall(MatSetValue(A,II,II+1,-1.0,INSERT_VALUES));
 54:     PetscCall(MatSetValue(A,II,II,4.0,INSERT_VALUES));
 55:     PetscCall(MatSetValue(B,II,II,2.0/PetscLogScalar(II+2),INSERT_VALUES));
 56:   }

 58:   PetscCall(MatAssemblyBegin(A,MAT_FINAL_ASSEMBLY));
 59:   PetscCall(MatAssemblyEnd(A,MAT_FINAL_ASSEMBLY));
 60:   PetscCall(MatAssemblyBegin(B,MAT_FINAL_ASSEMBLY));
 61:   PetscCall(MatAssemblyEnd(B,MAT_FINAL_ASSEMBLY));
 62:   PetscCall(MatCreateVecs(B,&v,NULL));

 64:   /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
 65:                 Create the eigensolver and set various options
 66:      - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */

 68:   PetscCall(EPSCreate(PETSC_COMM_WORLD,&eps));
 69:   PetscCall(EPSSetOperators(eps,A,B));
 70:   PetscCall(EPSSetProblemType(eps,EPS_GHEP));
 71:   PetscCall(EPSSetTolerances(eps,tol,PETSC_CURRENT));
 72:   PetscCall(EPSSetConvergenceTest(eps,EPS_CONV_NORM));
 73:   PetscCall(EPSSetFromOptions(eps));

 75:   /* illustrate how to extract parameters from specific solver types */
 76:   PetscCall(PetscObjectTypeCompare((PetscObject)eps,EPSPOWER,&flag));
 77:   if (flag) {
 78:     PetscCall(EPSGetST(eps,&st));
 79:     PetscCall(PetscObjectTypeCompare((PetscObject)st,STSHIFT,&flag));
 80:     if (flag) {
 81:       PetscCall(EPSPowerGetShiftType(eps,&variant));
 82:       PetscCall(PetscPrintf(PETSC_COMM_WORLD,"Type of shifts used during power iteration: %s\n",EPSPowerShiftTypes[variant]));
 83:     }
 84:   }

 86:   /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
 87:                       Solve the eigensystem
 88:      - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */

 90:   PetscCall(EPSSolve(eps));

 92:   /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
 93:                     Display solution and clean up
 94:      - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */

 96:   PetscCall(EPSGetTolerances(eps,&tol,NULL));
 97:   PetscCall(EPSErrorView(eps,EPS_ERROR_BACKWARD,NULL));
 98:   PetscCall(EPSGetConverged(eps,&nconv));
 99:   if (nconv>1) {
100:     PetscCall(VecDuplicateVecs(v,nconv,&X));
101:     for (i=0;i<nconv;i++) PetscCall(EPSGetEigenvector(eps,i,X[i],NULL));
102:     if (!skiporth) PetscCall(VecCheckOrthonormality(X,nconv,NULL,nconv,B,NULL,&lev));
103:     if (lev<10*tol) PetscCall(PetscPrintf(PETSC_COMM_WORLD,"Level of orthogonality below the tolerance\n"));
104:     else PetscCall(PetscPrintf(PETSC_COMM_WORLD,"Level of orthogonality: %g\n",(double)lev));
105:     PetscCall(VecDestroyVecs(nconv,&X));
106:   }

108:   PetscCall(EPSDestroy(&eps));
109:   PetscCall(MatDestroy(&A));
110:   PetscCall(MatDestroy(&B));
111:   PetscCall(VecDestroy(&v));
112:   PetscCall(SlepcFinalize());
113:   return 0;
114: }

116: /*TEST

118:    testset:
119:       args: -n 18 -eps_nev 4 -eps_max_it 1500
120:       requires: !single
121:       output_file: output/test1_1.out
122:       test:
123:          suffix: 1
124:          args: -eps_type {{krylovschur arnoldi gd jd lapack}}
125:       test:
126:          suffix: 1_subspace
127:          args: -eps_type subspace -eps_conv_rel
128:       test:
129:          suffix: 1_ks_nopurify
130:          args: -eps_purify 0
131:       test:
132:          suffix: 1_ks_trueres
133:          args: -eps_true_residual
134:       test:
135:          suffix: 1_ks_sinvert
136:          args: -st_type sinvert -eps_target 22
137:       test:
138:          suffix: 1_ks_cayley
139:          args: -st_type cayley -eps_target 22
140:       test:
141:          suffix: 1_lanczos
142:          args: -eps_type lanczos -eps_lanczos_reorthog full
143:       test:
144:          suffix: 1_gd2
145:          args: -eps_type gd -eps_gd_double_expansion
146:       test:
147:          suffix: 1_gd_borth
148:          args: -eps_type gd -eps_gd_borth
149:       test:
150:          suffix: 1_jd_borth
151:          args: -eps_type jd -eps_jd_borth
152:       test:
153:          suffix: 1_lobpcg
154:          args: -eps_type lobpcg -st_shift 22 -eps_largest_real
155:       test:
156:          suffix: 1_hpddm
157:          requires: hpddm
158:          args: -eps_type lobpcg -st_shift 22 -eps_largest_real -st_pc_type lu -st_ksp_type hpddm
159:       test:
160:          suffix: 1_cholesky
161:          args: -mat_type sbaij
162:       test:
163:          suffix: 1_scalapack
164:          nsize: {{1 2 3}}
165:          requires: scalapack
166:          args: -eps_type scalapack
167:       test:
168:          suffix: 1_elpa
169:          nsize: {{1 2 3}}
170:          requires: elpa
171:          args: -eps_type elpa
172:          filter: grep -v "Buffering level"
173:       test:
174:          suffix: 1_elemental
175:          nsize: {{1 2}}
176:          requires: elemental
177:          args: -eps_type elemental

179:    testset:
180:       args: -n 18 -eps_type ciss -rg_interval_endpoints 20.8,22
181:       requires: !single
182:       output_file: output/test1_1_ciss.out
183:       test:
184:          suffix: 1_ciss
185:          args: -eps_ciss_extraction {{ritz hankel}}
186:       test:
187:          suffix: 1_ciss_ksps
188:          args: -eps_ciss_usest 0 -eps_ciss_integration_points 12
189:       test:
190:          suffix: 1_ciss_gnhep
191:          args: -eps_gen_non_hermitian -skiporth
192:       test:
193:          suffix: 1_ciss_trapezoidal
194:          args: -eps_ciss_quadrule trapezoidal -eps_ciss_integration_points 24 -eps_ciss_extraction hankel -eps_ciss_delta 1e-10 -eps_tol 5e-11 -skiporth
195:       test:
196:          suffix: 1_ciss_cuda
197:          args: -mat_type aijcusparse -st_pc_factor_mat_solver_type cusparse
198:          requires: cuda
199:       test:
200:          suffix: 1_ciss_hip
201:          args: -mat_type aijhipsparse -st_pc_factor_mat_solver_type hipsparse
202:          requires: hip

204:    testset:
205:       requires: !single
206:       args: -eps_tol 1e-10 -st_type sinvert -st_ksp_type preonly -st_pc_type cholesky
207:       test:
208:          suffix: 2
209:          args: -eps_interval .1,1.1
210:       test:
211:          suffix: 2_open
212:          args: -eps_interval -inf,1.1
213:       test:
214:          suffix: 2_parallel
215:          requires: mumps !complex
216:          nsize: 3
217:          args: -eps_interval .1,1.1 -eps_krylovschur_partitions 2 -st_pc_factor_mat_solver_type mumps -st_mat_mumps_icntl_13 1
218:          output_file: output/test1_2.out

220:    test:
221:       suffix: 3
222:       requires: !single
223:       args: -n 18 -eps_type power -eps_conv_rel -eps_nev 3

225:    test:
226:       suffix: 4
227:       requires: !single
228:       args: -n 18 -eps_type power -eps_conv_rel -eps_nev 3 -st_type sinvert -eps_target 1.149 -eps_power_shift_type {{constant rayleigh wilkinson}}

230:    testset:
231:       args: -n 18 -eps_nev 3 -eps_smallest_real -eps_max_it 500 -st_pc_type icc
232:       output_file: output/test1_5.out
233:       test:
234:          suffix: 5_rqcg
235:          args: -eps_type rqcg
236:       test:
237:          suffix: 5_lobpcg
238:          args: -eps_type lobpcg -eps_lobpcg_blocksize 3
239:       test:
240:          suffix: 5_hpddm
241:          args: -eps_type lobpcg -eps_lobpcg_blocksize 3 -st_pc_type lu -st_ksp_type hpddm
242:          requires: hpddm
243:       test:
244:          suffix: 5_blopex
245:          args: -eps_type blopex -eps_conv_abs -st_shift 0.1
246:          requires: blopex

248:    testset:
249:       args: -n 18 -eps_nev 12 -eps_mpd 8 -eps_max_it 3000
250:       requires: !single
251:       output_file: output/test1_6.out
252:       test:
253:          suffix: 6
254:          args: -eps_type {{krylovschur arnoldi gd}}
255:       test:
256:          suffix: 6_lanczos
257:          args: -eps_type lanczos -eps_lanczos_reorthog full
258:       test:
259:          suffix: 6_subspace
260:          args: -eps_type subspace -eps_conv_rel

262:    testset:
263:       args: -n 18 -eps_nev 4 -eps_max_it 1500 -mat_type aijcusparse
264:       requires: cuda !single
265:       output_file: output/test1_1.out
266:       test:
267:          suffix: 7
268:          args: -eps_type {{krylovschur arnoldi gd jd}}
269:       test:
270:          suffix: 7_subspace
271:          args: -eps_type subspace -eps_conv_rel
272:       test:
273:          suffix: 7_ks_sinvert
274:          args: -st_type sinvert -eps_target 22
275:       test:
276:          suffix: 7_lanczos
277:          args: -eps_type lanczos -eps_lanczos_reorthog full
278:       test:
279:          suffix: 7_ciss
280:          args: -eps_type ciss -rg_interval_endpoints 20.8,22 -st_pc_factor_mat_solver_type cusparse
281:          output_file: output/test1_1_ciss.out

283:    testset:
284:       args: -n 18 -eps_nev 3 -eps_smallest_real -eps_max_it 500 -st_pc_type sor -mat_type aijcusparse
285:       requires: cuda
286:       output_file: output/test1_5.out
287:       test:
288:          suffix: 8_rqcg
289:          args: -eps_type rqcg
290:       test:
291:          suffix: 8_lobpcg
292:          args: -eps_type lobpcg -eps_lobpcg_blocksize 3

294:    testset:
295:       nsize: 2
296:       args: -n 18 -eps_nev 7 -eps_ncv 32 -ds_parallel synchronized
297:       filter: grep -v "orthogonality" | sed -e "s/[+-]0\.0*i//g" | sed -e "s/0.61338/0.61339/g"
298:       output_file: output/test1_9.out
299:       test:
300:          suffix: 9_ks_ghep
301:          args: -eps_gen_hermitian -st_pc_type redundant -st_type sinvert
302:       test:
303:          suffix: 9_ks_gnhep
304:          args: -eps_gen_non_hermitian -st_pc_type redundant -st_type sinvert
305:       test:
306:          suffix: 9_ks_ghiep
307:          args: -eps_gen_indefinite -st_pc_type redundant -st_type sinvert
308:          requires: !single
309:       test:
310:          suffix: 9_lobpcg_ghep
311:          args: -eps_gen_hermitian -eps_type lobpcg -eps_max_it 200 -eps_lobpcg_blocksize 6
312:          requires: !single
313:          timeoutfactor: 2
314:       test:
315:          suffix: 9_jd_gnhep
316:          args: -eps_gen_non_hermitian -eps_type jd -eps_target 0 -eps_ncv 64
317:          requires: !single
318:          timeoutfactor: 2

320:    test:
321:       suffix: 10_feast
322:       args: -n 25 -eps_type feast -eps_interval .95,1.1 -eps_conv_rel -eps_tol 1e-6
323:       requires: feast

325:    testset:
326:       args: -n 18 -eps_nev 4 -eps_max_it 1500 -mat_type aijhipsparse
327:       requires: hip !single
328:       output_file: output/test1_1.out
329:       test:
330:          suffix: 11
331:          args: -eps_type {{krylovschur arnoldi gd jd}}
332:       test:
333:          suffix: 11_subspace
334:          args: -eps_type subspace -eps_conv_rel
335:       test:
336:          suffix: 11_ks_sinvert
337:          args: -st_type sinvert -eps_target 22
338:       test:
339:          suffix: 11_lanczos
340:          args: -eps_type lanczos -eps_lanczos_reorthog full
341:       test:
342:          suffix: 11_ciss
343:          args: -eps_type ciss -rg_interval_endpoints 20.8,22 -st_pc_factor_mat_solver_type hipsparse
344:          output_file: output/test1_1_ciss.out

346:    testset:
347:       args: -n 18 -eps_nev 3 -eps_smallest_real -eps_max_it 500 -st_pc_type sor -mat_type aijhipsparse
348:       requires: hip
349:       output_file: output/test1_5.out
350:       test:
351:          suffix: 12_rqcg
352:          args: -eps_type rqcg
353:       test:
354:          suffix: 12_lobpcg
355:          args: -eps_type lobpcg -eps_lobpcg_blocksize 3

357: TEST*/