(git:103fa69)
Loading...
Searching...
No Matches
library_tests.F
Go to the documentation of this file.
1!--------------------------------------------------------------------------------------------------!
2! CP2K: A general program to perform molecular dynamics simulations !
3! Copyright 2000-2026 CP2K developers group <https://cp2k.org> !
4! !
5! SPDX-License-Identifier: GPL-2.0-or-later !
6!--------------------------------------------------------------------------------------------------!
7
8! **************************************************************************************************
9!> \brief Performance tests for basic tasks like matrix multiplies, copy, fft.
10!> \par History
11!> 30-Nov-2000 (JGH) added input
12!> 02-Jan-2001 (JGH) Parallel FFT
13!> 28-Feb-2002 (JGH) Clebsch-Gordon Coefficients
14!> 06-Jun-2003 (JGH) Real space grid test
15!> Eigensolver test (29.08.05,MK)
16!> \author JGH 6-NOV-2000
17! **************************************************************************************************
18MODULE library_tests
19
20 USE ai_coulomb_test, ONLY: eri_test
21 USE cell_methods, ONLY: cell_create,&
22 init_cell
23 USE cell_types, ONLY: cell_release,&
24 cell_type
25 USE cg_test, ONLY: gaunt_coefficient_test
26 USE cp_blacs_env, ONLY: cp_blacs_env_create,&
27 cp_blacs_env_release,&
28 cp_blacs_env_type
29 USE cp_dbcsr_api, ONLY: dbcsr_reset_randmat_seed,&
30 dbcsr_run_tests
31 USE cp_eri_mme_interface, ONLY: cp_eri_mme_perf_acc_test
32 USE cp_files, ONLY: close_file,&
33 open_file
34 USE cp_fm_basic_linalg, ONLY: cp_fm_gemm
35 USE cp_fm_diag, ONLY: cp_fm_syevd,&
36 cp_fm_syevx
37 USE cp_fm_struct, ONLY: cp_fm_struct_create,&
38 cp_fm_struct_get,&
39 cp_fm_struct_release,&
40 cp_fm_struct_type
41 USE cp_fm_types, ONLY: cp_fm_create,&
42 cp_fm_pilaenv,&
43 cp_fm_release,&
44 cp_fm_set_all,&
45 cp_fm_set_submatrix,&
46 cp_fm_to_fm,&
47 cp_fm_type
48 USE cp_log_handling, ONLY: cp_get_default_logger,&
49 cp_logger_type
50 USE cp_output_handling, ONLY: cp_print_key_finished_output,&
51 cp_print_key_unit_nr
52 USE cp_realspace_grid_init, ONLY: init_input_type
53 USE dbm_tests, ONLY: dbm_run_tests
54 USE fft_tools, ONLY: bwfft,&
55 fft_radix_closest,&
56 fwfft,&
57 fft3d,&
58 fft_radix_operations,&
59 finalize_fft,&
60 init_fft
61 USE global_types, ONLY: global_environment_type
62 USE input_constants, ONLY: do_diag_syevd,&
63 do_diag_syevx,&
64 do_mat_random,&
65 do_mat_read,&
66 do_pwgrid_ns_fullspace,&
67 do_pwgrid_ns_halfspace,&
68 do_pwgrid_spherical
69 USE input_section_types, ONLY: section_vals_get,&
70 section_vals_get_subs_vals,&
71 section_vals_type,&
72 section_vals_val_get
73 USE kinds, ONLY: dp
74 USE machine, ONLY: m_flush,&
75 m_walltime
76 USE mathconstants, ONLY: gaussi
77 USE message_passing, ONLY: mp_para_env_type
78 USE minimax_exp, ONLY: validate_exp_minimax
79 USE mp_perf_test, ONLY: mpi_perf_test
80 USE parallel_gemm_api, ONLY: parallel_gemm
81 USE parallel_rng_types, ONLY: uniform,&
82 rng_stream_type
83 USE pw_grid_types, ONLY: fullspace,&
84 halfspace,&
85 pw_grid_type
86 USE pw_grids, ONLY: pw_grid_create,&
87 pw_grid_release
88 USE pw_methods, ONLY: pw_transfer,&
89 pw_zero
90 USE pw_types, ONLY: pw_c1d_gs_type,&
91 pw_c3d_rs_type,&
92 pw_r3d_rs_type
93 USE realspace_grid_types, ONLY: &
94 realspace_grid_desc_type, realspace_grid_input_type, realspace_grid_type, rs_grid_create, &
95 rs_grid_create_descriptor, rs_grid_print, rs_grid_release, rs_grid_release_descriptor, &
96 rs_grid_zero, transfer_pw2rs, transfer_rs2pw
97 USE shg_integrals_test, ONLY: shg_integrals_perf_acc_test
98 USE time_frequency_grids, ONLY: test_least_square_ft
99#include "./base/base_uses.f90"
100
101 IMPLICIT NONE
102
103 PRIVATE
104 PUBLIC :: lib_test
105
106 INTEGER :: runtest(100)
107 REAL(KIND=dp) :: max_memory
108 REAL(KIND=dp), PARAMETER :: threshold = 1.0e-8_dp
109 CHARACTER(len=*), PARAMETER, PRIVATE :: moduleN = 'library_tests'
110
111CONTAINS
112
113! **************************************************************************************************
114!> \brief Master routine for tests
115!> \param root_section ...
116!> \param para_env ...
117!> \param globenv ...
118!> \par History
119!> none
120!> \author JGH 6-NOV-2000
121! **************************************************************************************************
122 SUBROUTINE lib_test(root_section, para_env, globenv)
123
124 TYPE(section_vals_type), POINTER :: root_section
125 TYPE(mp_para_env_type), POINTER :: para_env
126 TYPE(global_environment_type), POINTER :: globenv
127
128 CHARACTER(LEN=*), PARAMETER :: routinen = 'lib_test'
129
130 INTEGER :: handle, iw
131 LOGICAL :: explicit
132 TYPE(cp_logger_type), POINTER :: logger
133 TYPE(section_vals_type), POINTER :: cp_dbcsr_test_section, cp_fm_gemm_test_section, &
134 dbm_test_section, eigensolver_section, eri_mme_test_section, pw_transfer_section, &
135 rs_pw_transfer_section, shg_integrals_test_section
136
137 CALL timeset(routinen, handle)
138
139 logger => cp_get_default_logger()
140 iw = cp_print_key_unit_nr(logger, root_section, "TEST%PROGRAM_RUN_INFO", extension=".log")
141
142 IF (iw > 0) THEN
143 WRITE (iw, '(T2,79("*"))')
144 WRITE (iw, '(A,T31,A,T80,A)') ' *', ' PERFORMANCE TESTS ', '*'
145 WRITE (iw, '(T2,79("*"))')
146 END IF
147 !
148 CALL test_input(root_section, para_env)
149 !
150 IF (runtest(1) /= 0) CALL copy_test(para_env, iw)
151 !
152 IF (runtest(2) /= 0) CALL matmul_test(para_env, test_matmul=.true., test_dgemm=.false., iw=iw)
153 IF (runtest(5) /= 0) CALL matmul_test(para_env, test_matmul=.false., test_dgemm=.true., iw=iw)
154 !
155 IF (runtest(3) /= 0) CALL fft_test(para_env, iw, globenv%fftw_plan_type, &
156 globenv%fftw_wisdom_file_name, &
157 globenv%fftw_wisdom_file_explicit)
158 !
159 IF (runtest(4) /= 0) CALL eri_test(iw)
160 !
161 IF (runtest(6) /= 0) CALL gaunt_coefficient_test()
162 !
163 ! runtest 7 has been deleted and can be recycled
164 !
165 IF (runtest(8) /= 0) CALL mpi_perf_test(para_env, runtest(8), iw)
166 !
167 IF (runtest(10) /= 0) CALL validate_exp_minimax(runtest(10), iw)
168 !
169 IF (runtest(11) /= 0) CALL test_least_square_ft(runtest(11), iw)
170 !
171
172 rs_pw_transfer_section => section_vals_get_subs_vals(root_section, "TEST%RS_PW_TRANSFER")
173 CALL section_vals_get(rs_pw_transfer_section, explicit=explicit)
174 IF (explicit) THEN
175 CALL rs_pw_transfer_test(para_env, iw, globenv, rs_pw_transfer_section)
176 END IF
177
178 pw_transfer_section => section_vals_get_subs_vals(root_section, "TEST%PW_TRANSFER")
179 CALL section_vals_get(pw_transfer_section, explicit=explicit)
180 IF (explicit) THEN
181 CALL pw_fft_test(para_env, iw, globenv, pw_transfer_section)
182 END IF
183
184 cp_fm_gemm_test_section => section_vals_get_subs_vals(root_section, "TEST%CP_FM_GEMM")
185 CALL section_vals_get(cp_fm_gemm_test_section, explicit=explicit)
186 IF (explicit) THEN
187 CALL cp_fm_gemm_test(para_env, iw, cp_fm_gemm_test_section)
188 END IF
189
190 eigensolver_section => section_vals_get_subs_vals(root_section, "TEST%EIGENSOLVER")
191 CALL section_vals_get(eigensolver_section, explicit=explicit)
192 IF (explicit) THEN
193 CALL eigensolver_test(para_env, iw, eigensolver_section)
194 END IF
195
196 eri_mme_test_section => section_vals_get_subs_vals(root_section, "TEST%ERI_MME_TEST")
197 CALL section_vals_get(eri_mme_test_section, explicit=explicit)
198 IF (explicit) THEN
199 CALL cp_eri_mme_perf_acc_test(para_env, iw, eri_mme_test_section)
200 END IF
201
202 shg_integrals_test_section => section_vals_get_subs_vals(root_section, "TEST%SHG_INTEGRALS_TEST")
203 CALL section_vals_get(shg_integrals_test_section, explicit=explicit)
204 IF (explicit) THEN
205 CALL shg_integrals_perf_acc_test(iw, shg_integrals_test_section)
206 END IF
207
208 ! DBCSR tests
209 cp_dbcsr_test_section => section_vals_get_subs_vals(root_section, "TEST%CP_DBCSR")
210 CALL section_vals_get(cp_dbcsr_test_section, explicit=explicit)
211 IF (explicit) THEN
212 CALL cp_dbcsr_tests(para_env, iw, cp_dbcsr_test_section)
213 END IF
214
215 ! DBM tests
216 dbm_test_section => section_vals_get_subs_vals(root_section, "TEST%DBM")
217 CALL section_vals_get(dbm_test_section, explicit=explicit)
218 IF (explicit) THEN
219 CALL run_dbm_tests(para_env, iw, dbm_test_section)
220 END IF
221
222 CALL cp_print_key_finished_output(iw, logger, root_section, "TEST%PROGRAM_RUN_INFO")
223
224 CALL timestop(handle)
225
226 END SUBROUTINE lib_test
227
228! **************************************************************************************************
229!> \brief Reads input section &TEST ... &END
230!> \param root_section ...
231!> \param para_env ...
232!> \author JGH 30-NOV-2000
233!> \note
234!> I---------------------------------------------------------------------------I
235!> I SECTION: &TEST ... &END I
236!> I I
237!> I MEMORY max_memory I
238!> I COPY n I
239!> I MATMUL n I
240!> I FFT n I
241!> I ERI n I
242!> I PW_FFT n I
243!> I Clebsch-Gordon n -> Gaunt coefficients I
244!> I RS_GRIDS n I
245!> I MPI n I
246!> I RNG n -> Parallel random number generator I
247!> I---------------------------------------------------------------------------I
248! **************************************************************************************************
249 SUBROUTINE test_input(root_section, para_env)
250 TYPE(section_vals_type), POINTER :: root_section
251 TYPE(mp_para_env_type), POINTER :: para_env
252
253 TYPE(section_vals_type), POINTER :: test_section
254
255!
256!..defaults
257! using this style is not recommended, introduce sections instead (see e.g. cp_fm_gemm)
258
259 runtest = 0
260 test_section => section_vals_get_subs_vals(root_section, "TEST")
261 CALL section_vals_val_get(test_section, "MEMORY", r_val=max_memory)
262 CALL section_vals_val_get(test_section, 'COPY', i_val=runtest(1))
263 CALL section_vals_val_get(test_section, 'MATMUL', i_val=runtest(2))
264 CALL section_vals_val_get(test_section, 'DGEMM', i_val=runtest(5))
265 CALL section_vals_val_get(test_section, 'FFT', i_val=runtest(3))
266 CALL section_vals_val_get(test_section, 'ERI', i_val=runtest(4))
267 CALL section_vals_val_get(test_section, 'CLEBSCH_GORDON', i_val=runtest(6))
268 CALL section_vals_val_get(test_section, 'MPI', i_val=runtest(8))
269 CALL section_vals_val_get(test_section, 'MINIMAX', i_val=runtest(10))
270 CALL section_vals_val_get(test_section, 'LEAST_SQ_FT', i_val=runtest(11))
271
272 CALL para_env%sync()
273 END SUBROUTINE test_input
274
275! **************************************************************************************************
276!> \brief Tests the performance to copy two vectors.
277!> \param para_env ...
278!> \param iw ...
279!> \par History
280!> none
281!> \author JGH 6-NOV-2000
282!> \note
283!> The results of these tests allow to determine the size of the cache
284!> of the CPU. This can be used to optimize the performance of the
285!> FFTSG library.
286! **************************************************************************************************
287 SUBROUTINE copy_test(para_env, iw)
288 TYPE(mp_para_env_type), POINTER :: para_env
289 INTEGER :: iw
290
291 INTEGER :: i, j, len, ntim, siz
292 REAL(kind=dp) :: perf, t, tend, tstart
293 REAL(kind=dp), ALLOCATABLE, DIMENSION(:) :: ca, cb
294
295! test for copy --> Cache size
296
297 siz = abs(runtest(1))
298 IF (para_env%is_source()) WRITE (iw, '(//,A,/)') " Test of copy ( F95 ) "
299 DO i = 6, 24
300 len = 2**i
301 IF (8.0_dp*real(len, kind=dp) > max_memory*0.5_dp) EXIT
302 ALLOCATE (ca(len))
303 ALLOCATE (cb(len))
304
305 CALL random_number(ca)
306 ntim = nint(1.e7_dp/real(len, kind=dp))
307 ntim = max(ntim, 1)
308 ntim = min(ntim, siz*10000)
309
310 tstart = m_walltime()
311 DO j = 1, ntim
312 cb(:) = ca(:)
313 ca(1) = real(j, kind=dp)
314 END DO
315 tend = m_walltime()
316 t = tend - tstart + threshold
317 IF (t > 0.0_dp) THEN
318 perf = real(ntim, kind=dp)*real(len, kind=dp)*1.e-6_dp/t
319 ELSE
320 perf = 0.0_dp
321 END IF
322
323 IF (para_env%is_source()) THEN
324 WRITE (iw, '(A,i2,i10,A,T59,F14.4,A)') " Copy test: Size = 2^", i, &
325 len/1024, " Kwords", perf, " Mcopy/s"
326 END IF
327
328 DEALLOCATE (ca)
329 DEALLOCATE (cb)
330 END DO
331 CALL para_env%sync()
332 END SUBROUTINE copy_test
333
334! **************************************************************************************************
335!> \brief Tests the performance of different kinds of matrix matrix multiply
336!> kernels for the BLAS and F95 intrinsic matmul.
337!> \param para_env ...
338!> \param test_matmul ...
339!> \param test_dgemm ...
340!> \param iw ...
341!> \par History
342!> none
343!> \author JGH 6-NOV-2000
344! **************************************************************************************************
345 SUBROUTINE matmul_test(para_env, test_matmul, test_dgemm, iw)
346 TYPE(mp_para_env_type), POINTER :: para_env
347 LOGICAL :: test_matmul, test_dgemm
348 INTEGER :: iw
349
350 INTEGER :: i, j, len, ntim, siz
351 REAL(kind=dp) :: perf, t, tend, tstart, xdum
352 REAL(kind=dp), ALLOCATABLE, DIMENSION(:, :) :: ma, mb, mc
353
354! test for matrix multpies
355
356 IF (test_matmul) THEN
357 siz = abs(runtest(2))
358 IF (para_env%is_source()) WRITE (iw, '(//,A,/)') " Test of matmul ( F95 ) "
359 DO i = 5, siz, 2
360 len = 2**i + 1
361 IF (8.0_dp*real(len*len, kind=dp) > max_memory*0.3_dp) EXIT
362 ALLOCATE (ma(len, len))
363 ALLOCATE (mb(len, len))
364 ALLOCATE (mc(len, len))
365 mc = 0.0_dp
366
367 CALL random_number(xdum)
368 ma = xdum
369 CALL random_number(xdum)
370 mb = xdum
371 ntim = nint(1.e8_dp/(2.0_dp*real(len, kind=dp)**3))
372 ntim = max(ntim, 1)
373 ntim = min(ntim, siz*200)
374 tstart = m_walltime()
375 DO j = 1, ntim
376 mc(:, :) = matmul(ma, mb)
377 ma(1, 1) = real(j, kind=dp)
378 END DO
379 tend = m_walltime()
380 t = tend - tstart + threshold
381 perf = real(ntim, kind=dp)*2.0_dp*real(len, kind=dp)**3*1.e-6_dp/t
382 IF (para_env%is_source()) THEN
383 WRITE (iw, '(A,i6,T59,F14.4,A)') &
384 " Matrix multiply test: c = a * b Size = ", len, perf, " Mflop/s"
385 END IF
386 tstart = m_walltime()
387 DO j = 1, ntim
388 mc(:, :) = mc + matmul(ma, mb)
389 ma(1, 1) = real(j, kind=dp)
390 END DO
391 tend = m_walltime()
392 t = tend - tstart + threshold
393 IF (t > 0.0_dp) THEN
394 perf = real(ntim, kind=dp)*2.0_dp*real(len, kind=dp)**3*1.e-6_dp/t
395 ELSE
396 perf = 0.0_dp
397 END IF
398
399 IF (para_env%is_source()) THEN
400 WRITE (iw, '(A,i6,T59,F14.4,A)') &
401 " Matrix multiply test: a = a * b Size = ", len, perf, " Mflop/s"
402 END IF
403
404 tstart = m_walltime()
405 DO j = 1, ntim
406 mc(:, :) = mc + matmul(ma, transpose(mb))
407 ma(1, 1) = real(j, kind=dp)
408 END DO
409 tend = m_walltime()
410 t = tend - tstart + threshold
411 IF (t > 0.0_dp) THEN
412 perf = real(ntim, kind=dp)*2.0_dp*real(len, kind=dp)**3*1.e-6_dp/t
413 ELSE
414 perf = 0.0_dp
415 END IF
416
417 IF (para_env%is_source()) THEN
418 WRITE (iw, '(A,i6,T59,F14.4,A)') &
419 " Matrix multiply test: c = a * b(T) Size = ", len, perf, " Mflop/s"
420 END IF
421
422 tstart = m_walltime()
423 DO j = 1, ntim
424 mc(:, :) = mc + matmul(transpose(ma), mb)
425 ma(1, 1) = real(j, kind=dp)
426 END DO
427 tend = m_walltime()
428 t = tend - tstart + threshold
429 IF (t > 0.0_dp) THEN
430 perf = real(ntim, kind=dp)*2.0_dp*real(len, kind=dp)**3*1.e-6_dp/t
431 ELSE
432 perf = 0.0_dp
433 END IF
434
435 IF (para_env%is_source()) THEN
436 WRITE (iw, '(A,i6,T59,F14.4,A)') &
437 " Matrix multiply test: c = a(T) * b Size = ", len, perf, " Mflop/s"
438 END IF
439
440 DEALLOCATE (ma)
441 DEALLOCATE (mb)
442 DEALLOCATE (mc)
443 END DO
444 END IF
445
446 ! test for matrix multpies
447 IF (test_dgemm) THEN
448 siz = abs(runtest(5))
449 IF (para_env%is_source()) WRITE (iw, '(//,A,/)') " Test of matmul ( BLAS ) "
450 DO i = 5, siz, 2
451 len = 2**i + 1
452 IF (8.0_dp*real(len*len, kind=dp) > max_memory*0.3_dp) EXIT
453 ALLOCATE (ma(len, len))
454 ALLOCATE (mb(len, len))
455 ALLOCATE (mc(len, len))
456 mc = 0.0_dp
457
458 CALL random_number(xdum)
459 ma = xdum
460 CALL random_number(xdum)
461 mb = xdum
462 ntim = nint(1.e8_dp/(2.0_dp*real(len, kind=dp)**3))
463 ntim = max(ntim, 1)
464 ntim = min(ntim, 1000)
465
466 tstart = m_walltime()
467 DO j = 1, ntim
468 CALL dgemm("N", "N", len, len, len, 1.0_dp, ma, len, mb, len, 1.0_dp, mc, len)
469 END DO
470 tend = m_walltime()
471 t = tend - tstart + threshold
472 IF (t > 0.0_dp) THEN
473 perf = real(ntim, kind=dp)*2.0_dp*real(len, kind=dp)**3*1.e-6_dp/t
474 ELSE
475 perf = 0.0_dp
476 END IF
477
478 IF (para_env%is_source()) THEN
479 WRITE (iw, '(A,i6,T59,F14.4,A)') &
480 " Matrix multiply test: c = a * b Size = ", len, perf, " Mflop/s"
481 END IF
482
483 tstart = m_walltime()
484 DO j = 1, ntim
485 CALL dgemm("N", "N", len, len, len, 1.0_dp, ma, len, mb, len, 1.0_dp, mc, len)
486 END DO
487 tend = m_walltime()
488 t = tend - tstart + threshold
489 IF (t > 0.0_dp) THEN
490 perf = real(ntim, kind=dp)*2.0_dp*real(len, kind=dp)**3*1.e-6_dp/t
491 ELSE
492 perf = 0.0_dp
493 END IF
494
495 IF (para_env%is_source()) THEN
496 WRITE (iw, '(A,i6,T59,F14.4,A)') &
497 " Matrix multiply test: a = a * b Size = ", len, perf, " Mflop/s"
498 END IF
499
500 tstart = m_walltime()
501 DO j = 1, ntim
502 CALL dgemm("N", "T", len, len, len, 1.0_dp, ma, len, mb, len, 1.0_dp, mc, len)
503 END DO
504 tend = m_walltime()
505 t = tend - tstart + threshold
506 IF (t > 0.0_dp) THEN
507 perf = real(ntim, kind=dp)*2.0_dp*real(len, kind=dp)**3*1.e-6_dp/t
508 ELSE
509 perf = 0.0_dp
510 END IF
511
512 IF (para_env%is_source()) THEN
513 WRITE (iw, '(A,i6,T59,F14.4,A)') &
514 " Matrix multiply test: c = a * b(T) Size = ", len, perf, " Mflop/s"
515 END IF
516
517 tstart = m_walltime()
518 DO j = 1, ntim
519 CALL dgemm("T", "N", len, len, len, 1.0_dp, ma, len, mb, len, 1.0_dp, mc, len)
520 END DO
521 tend = m_walltime()
522 t = tend - tstart + threshold
523 IF (t > 0.0_dp) THEN
524 perf = real(ntim, kind=dp)*2.0_dp*real(len, kind=dp)**3*1.e-6_dp/t
525 ELSE
526 perf = 0.0_dp
527 END IF
528
529 IF (para_env%is_source()) THEN
530 WRITE (iw, '(A,i6,T59,F14.4,A)') &
531 " Matrix multiply test: c = a(T) * b Size = ", len, perf, " Mflop/s"
532 END IF
533
534 DEALLOCATE (ma)
535 DEALLOCATE (mb)
536 DEALLOCATE (mc)
537 END DO
538 END IF
539
540 CALL para_env%sync()
541
542 END SUBROUTINE matmul_test
543
544! **************************************************************************************************
545!> \brief Tests the performance of all available FFT libraries for 3D FFTs
546!> \param para_env ...
547!> \param iw ...
548!> \param fftw_plan_type ...
549!> \param wisdom_file where FFTW3 should look to save/load wisdom
550!> \param export_wisdom whether the FFTW3 wisdom is saved to wisdom_file
551!> \par History
552!> none
553!> \author JGH 6-NOV-2000
554! **************************************************************************************************
555 SUBROUTINE fft_test(para_env, iw, fftw_plan_type, wisdom_file, export_wisdom)
556
557 TYPE(mp_para_env_type), POINTER :: para_env
558 INTEGER :: iw, fftw_plan_type
559 CHARACTER(LEN=*), INTENT(IN) :: wisdom_file
560 LOGICAL, INTENT(IN) :: export_wisdom
561
562 INTEGER, PARAMETER :: ndate(3) = [12, 48, 96]
563
564 INTEGER :: iall, it, j, len, n(3), ntim, radix_in, &
565 radix_out, siz, stat
566 COMPLEX(KIND=dp), DIMENSION(4, 4, 4) :: zz
567 COMPLEX(KIND=dp), ALLOCATABLE, DIMENSION(:, :, :) :: ca, cb, cc
568 CHARACTER(LEN=7) :: method
569 REAL(kind=dp) :: flops, perf, scale, t, tdiff, tend, &
570 tstart
571 REAL(kind=dp), ALLOCATABLE, DIMENSION(:, :, :) :: ra
572
573! test for 3d FFT
574
575 IF (para_env%is_source()) WRITE (iw, '(//,A,/)') " Test of 3D-FFT "
576 siz = abs(runtest(3))
577
578 DO iall = 1, 100
579 SELECT CASE (iall)
580 CASE DEFAULT
581 EXIT
582 CASE (1)
583 CALL init_fft("FFTSG", alltoall=.false., wisdom_file=wisdom_file, &
584 pool_limit=10, plan_style=fftw_plan_type)
585 method = "FFTSG "
586 CASE (2)
587 cycle
588 CASE (3)
589 CALL init_fft("FFTW3", alltoall=.false., wisdom_file=wisdom_file, &
590 pool_limit=10, plan_style=fftw_plan_type)
591 method = "FFTW3 "
592 END SELECT
593 n = 4
594 zz = 0.0_dp
595 CALL fft3d(fwfft, n, zz, status=stat)
596 IF (stat == 0) THEN
597 DO it = 1, 3
598 radix_in = ndate(it)
599 CALL fft_radix_operations(radix_in, radix_out, fft_radix_closest)
600 len = radix_out
601 n = len
602 IF (16.0_dp*real(len*len*len, kind=dp) > max_memory*0.5_dp) EXIT
603 ALLOCATE (ra(len, len, len))
604 ALLOCATE (ca(len, len, len))
605 CALL random_number(ra)
606 ca(:, :, :) = ra
607 CALL random_number(ra)
608 ca(:, :, :) = ca + gaussi*ra
609 flops = real(len**3, kind=dp)*15.0_dp*log(real(len, kind=dp))
610 ntim = nint(siz*1.e7_dp/flops)
611 ntim = max(ntim, 1)
612 ntim = min(ntim, 200)
613 scale = 1.0_dp/real(len**3, kind=dp)
614 tstart = m_walltime()
615 DO j = 1, ntim
616 CALL fft3d(fwfft, n, ca)
617 CALL fft3d(bwfft, n, ca)
618 END DO
619 tend = m_walltime()
620 t = tend - tstart + threshold
621 IF (t > 0.0_dp) THEN
622 perf = real(ntim, kind=dp)*2.0_dp*flops*1.e-6_dp/t
623 ELSE
624 perf = 0.0_dp
625 END IF
626
627 IF (para_env%is_source()) THEN
628 WRITE (iw, '(T2,A,A,i6,T59,F14.4,A)') &
629 adjustr(method), " test (in-place) Size = ", len, perf, " Mflop/s"
630 END IF
631 DEALLOCATE (ca)
632 DEALLOCATE (ra)
633 END DO
634 IF (para_env%is_source()) WRITE (iw, *)
635 ! test if input data is preserved
636 len = 24
637 n = len
638 ALLOCATE (ra(len, len, len))
639 ALLOCATE (ca(len, len, len))
640 ALLOCATE (cb(len, len, len))
641 ALLOCATE (cc(len, len, len))
642 CALL random_number(ra)
643 ca(:, :, :) = ra
644 CALL random_number(ra)
645 ca(:, :, :) = ca + gaussi*ra
646 cc(:, :, :) = ca
647 CALL fft3d(fwfft, n, ca, cb)
648 tdiff = maxval(abs(ca - cc))
649 IF (tdiff > 1.0e-12_dp) THEN
650 IF (para_env%is_source()) THEN
651 WRITE (iw, '(T2,A,A,A)') adjustr(method), " FWFFT ", &
652 " Input array is changed in out-of-place FFT !"
653 END IF
654 ELSE
655 IF (para_env%is_source()) THEN
656 WRITE (iw, '(T2,A,A,A)') adjustr(method), " FWFFT ", &
657 " Input array is not changed in out-of-place FFT !"
658 END IF
659 END IF
660 ca(:, :, :) = cc
661 CALL fft3d(bwfft, n, ca, cb)
662 tdiff = maxval(abs(ca - cc))
663 IF (tdiff > 1.0e-12_dp) THEN
664 IF (para_env%is_source()) THEN
665 WRITE (iw, '(T2,A,A,A)') adjustr(method), " BWFFT ", &
666 " Input array is changed in out-of-place FFT !"
667 END IF
668 ELSE
669 IF (para_env%is_source()) THEN
670 WRITE (iw, '(T2,A,A,A)') adjustr(method), " BWFFT ", &
671 " Input array is not changed in out-of-place FFT !"
672 END IF
673 END IF
674 IF (para_env%is_source()) WRITE (iw, *)
675
676 DEALLOCATE (ra)
677 DEALLOCATE (ca)
678 DEALLOCATE (cb)
679 DEALLOCATE (cc)
680 END IF
681 CALL finalize_fft(para_env, wisdom_file=wisdom_file, export_wisdom=export_wisdom)
682 END DO
683
684 END SUBROUTINE fft_test
685
686! **************************************************************************************************
687!> \brief test rs_pw_transfer performance
688!> \param para_env ...
689!> \param iw ...
690!> \param globenv ...
691!> \param rs_pw_transfer_section ...
692!> \author Joost VandeVondele
693!> 9.2008 Randomise rs grid [Iain Bethune]
694!> (c) The Numerical Algorithms Group (NAG) Ltd, 2008 on behalf of the HECToR project
695! **************************************************************************************************
696 SUBROUTINE rs_pw_transfer_test(para_env, iw, globenv, rs_pw_transfer_section)
697
698 TYPE(mp_para_env_type), POINTER :: para_env
699 INTEGER :: iw
700 TYPE(global_environment_type), POINTER :: globenv
701 TYPE(section_vals_type), POINTER :: rs_pw_transfer_section
702
703 CHARACTER(LEN=*), PARAMETER :: routinen = 'rs_pw_transfer_test'
704
705 INTEGER :: halo_size, handle, i_loop, n_loop, ns_max
706 INTEGER, DIMENSION(3) :: no, np
707 INTEGER, DIMENSION(:), POINTER :: i_vals
708 LOGICAL :: do_rs2pw
709 REAL(kind=dp) :: tend, tstart
710 TYPE(cell_type), POINTER :: box
711 TYPE(pw_grid_type), POINTER :: grid
712 TYPE(pw_r3d_rs_type) :: ca
713 TYPE(realspace_grid_desc_type), POINTER :: rs_desc
714 TYPE(realspace_grid_input_type) :: input_settings
715 TYPE(realspace_grid_type) :: rs_grid
716 TYPE(section_vals_type), POINTER :: rs_grid_section
717
718 CALL timeset(routinen, handle)
719
720 !..set fft lib
721 CALL init_fft(globenv%default_fft_library, alltoall=.false., &
722 pool_limit=globenv%fft_pool_scratch_limit, &
723 wisdom_file=globenv%fftw_wisdom_file_name, &
724 plan_style=globenv%fftw_plan_type)
725
726 ! .. set cell (should otherwise be irrelevant)
727 NULLIFY (box)
728 CALL cell_create(box)
729 box%hmat = reshape([20.0_dp, 0.0_dp, 0.0_dp, 0.0_dp, 20.0_dp, 0.0_dp, &
730 0.0_dp, 0.0_dp, 20.0_dp], [3, 3])
731 CALL init_cell(box)
732
733 ! .. grid type and pw_grid
734 NULLIFY (grid)
735 CALL section_vals_val_get(rs_pw_transfer_section, "GRID", i_vals=i_vals)
736 np = i_vals
737 CALL pw_grid_create(grid, para_env, box%hmat, grid_span=fullspace, npts=np, fft_usage=.true., iounit=iw)
738 no = grid%npts
739
740 CALL ca%create(grid)
741 CALL pw_zero(ca)
742
743 ! .. rs input setting type
744 CALL section_vals_val_get(rs_pw_transfer_section, "HALO_SIZE", i_val=halo_size)
745 rs_grid_section => section_vals_get_subs_vals(rs_pw_transfer_section, "RS_GRID")
746 ns_max = 2*halo_size + 1
747 CALL init_input_type(input_settings, ns_max, rs_grid_section, 1, [-1, -1, -1])
748
749 ! .. rs type
750 NULLIFY (rs_desc)
751 CALL rs_grid_create_descriptor(rs_desc, pw_grid=grid, input_settings=input_settings)
752 CALL rs_grid_create(rs_grid, rs_desc)
753 CALL rs_grid_print(rs_grid, iw)
754 CALL rs_grid_zero(rs_grid)
755
756 ! Put random values on the grid, so summation check will pick up errors
757 CALL random_number(rs_grid%r)
758
759 CALL section_vals_val_get(rs_pw_transfer_section, "N_loop", i_val=n_loop)
760 CALL section_vals_val_get(rs_pw_transfer_section, "RS2PW", l_val=do_rs2pw)
761
762 ! go for the real loops, sync to get max timings
763 IF (para_env%is_source()) THEN
764 WRITE (iw, '(T2,A)') ""
765 WRITE (iw, '(T2,A)') "Timing rs_pw_transfer routine"
766 WRITE (iw, '(T2,A)') ""
767 WRITE (iw, '(T2,A)') "iteration time[s]"
768 END IF
769 DO i_loop = 1, n_loop
770 CALL para_env%sync()
771 tstart = m_walltime()
772 IF (do_rs2pw) THEN
773 CALL transfer_rs2pw(rs_grid, ca)
774 ELSE
775 CALL transfer_pw2rs(rs_grid, ca)
776 END IF
777 CALL para_env%sync()
778 tend = m_walltime()
779 IF (para_env%is_source()) THEN
780 WRITE (iw, '(T2,I9,1X,F12.6)') i_loop, tend - tstart
781 END IF
782 END DO
783
784 !cleanup
785 CALL rs_grid_release(rs_grid)
786 CALL rs_grid_release_descriptor(rs_desc)
787 CALL ca%release()
788 CALL pw_grid_release(grid)
789 CALL cell_release(box)
790 CALL finalize_fft(para_env, wisdom_file=globenv%fftw_wisdom_file_name, &
791 export_wisdom=globenv%fftw_wisdom_file_explicit)
792
793 CALL timestop(handle)
794
795 END SUBROUTINE rs_pw_transfer_test
796
797! **************************************************************************************************
798!> \brief Tests the performance of PW calls to FFT routines
799!> \param para_env ...
800!> \param iw ...
801!> \param globenv ...
802!> \param pw_transfer_section ...
803!> \par History
804!> JGH 6-Feb-2001 : Test and performance code
805!> Made input sensitive [Joost VandeVondele]
806!> \author JGH 1-JAN-2001
807! **************************************************************************************************
808 SUBROUTINE pw_fft_test(para_env, iw, globenv, pw_transfer_section)
809
810 TYPE(mp_para_env_type), POINTER :: para_env
811 INTEGER :: iw
812 TYPE(global_environment_type), POINTER :: globenv
813 TYPE(section_vals_type), POINTER :: pw_transfer_section
814
815 REAL(kind=dp), PARAMETER :: toler = 1.e-11_dp
816
817 INTEGER :: blocked_id, grid_span, i_layout, i_rep, &
818 ig, ip, itmp, n_loop, n_rep, nn, p, q
819 INTEGER, ALLOCATABLE, DIMENSION(:, :) :: layouts
820 INTEGER, DIMENSION(2) :: distribution_layout
821 INTEGER, DIMENSION(3) :: no, np
822 INTEGER, DIMENSION(:), POINTER :: i_vals
823 LOGICAL :: debug, is_fullspace, odd, &
824 pw_grid_layout_all, spherical
825 REAL(kind=dp) :: em, et, flops, gsq, perf, t, t_max, &
826 t_min, tend, tstart
827 REAL(kind=dp), ALLOCATABLE, DIMENSION(:) :: t_end, t_start
828 TYPE(cell_type), POINTER :: box
829 TYPE(pw_c1d_gs_type) :: ca, cc
830 TYPE(pw_c3d_rs_type) :: cb
831 TYPE(pw_grid_type), POINTER :: grid
832
833!..set fft lib
834
835 CALL init_fft(globenv%default_fft_library, alltoall=.false., &
836 pool_limit=globenv%fft_pool_scratch_limit, &
837 wisdom_file=globenv%fftw_wisdom_file_name, &
838 plan_style=globenv%fftw_plan_type)
839
840 !..the unit cell (should not really matter, the number of grid points do)
841 NULLIFY (box, grid)
842 CALL cell_create(box)
843 box%hmat = reshape([10.0_dp, 0.0_dp, 0.0_dp, 0.0_dp, 8.0_dp, 0.0_dp, &
844 0.0_dp, 0.0_dp, 7.0_dp], [3, 3])
845 CALL init_cell(box)
846
847 CALL section_vals_get(pw_transfer_section, n_repetition=n_rep)
848 DO i_rep = 1, n_rep
849
850 ! how often should we do the transfer
851 CALL section_vals_val_get(pw_transfer_section, "N_loop", i_rep_section=i_rep, i_val=n_loop)
852 ALLOCATE (t_start(n_loop))
853 ALLOCATE (t_end(n_loop))
854
855 ! setup of the grids
856 CALL section_vals_val_get(pw_transfer_section, "GRID", i_rep_section=i_rep, i_vals=i_vals)
857 np = i_vals
858
859 CALL section_vals_val_get(pw_transfer_section, "PW_GRID_BLOCKED", i_rep_section=i_rep, i_val=blocked_id)
860 CALL section_vals_val_get(pw_transfer_section, "DEBUG", i_rep_section=i_rep, l_val=debug)
861
862 CALL section_vals_val_get(pw_transfer_section, "PW_GRID_LAYOUT_ALL", i_rep_section=i_rep, &
863 l_val=pw_grid_layout_all)
864
865 ! prepare to loop over all or a specific layout
866 IF (pw_grid_layout_all) THEN
867 ! count layouts that fit
868 itmp = 0
869 ! start from 2, (/1,para_env%num_pe/) is not supported
870 DO p = 2, para_env%num_pe
871 q = para_env%num_pe/p
872 IF (p*q == para_env%num_pe) THEN
873 itmp = itmp + 1
874 END IF
875 END DO
876 ! build list
877 ALLOCATE (layouts(2, itmp))
878 itmp = 0
879 DO p = 2, para_env%num_pe
880 q = para_env%num_pe/p
881 IF (p*q == para_env%num_pe) THEN
882 itmp = itmp + 1
883 layouts(:, itmp) = [p, q]
884 END IF
885 END DO
886 ELSE
887 CALL section_vals_val_get(pw_transfer_section, "PW_GRID_LAYOUT", i_rep_section=i_rep, i_vals=i_vals)
888 ALLOCATE (layouts(2, 1))
889 layouts(:, 1) = i_vals
890 END IF
891
892 DO i_layout = 1, SIZE(layouts, 2)
893
894 distribution_layout = layouts(:, i_layout)
895
896 CALL section_vals_val_get(pw_transfer_section, "PW_GRID", i_rep_section=i_rep, i_val=itmp)
897
898 ! from cp_control_utils
899 SELECT CASE (itmp)
900 CASE (do_pwgrid_spherical)
901 spherical = .true.
902 is_fullspace = .false.
903 CASE (do_pwgrid_ns_fullspace)
904 spherical = .false.
905 is_fullspace = .true.
906 CASE (do_pwgrid_ns_halfspace)
907 spherical = .false.
908 is_fullspace = .false.
909 END SELECT
910
911 ! from pw_env_methods
912 IF (spherical) THEN
913 grid_span = halfspace
914 spherical = .true.
915 odd = .true.
916 ELSE IF (is_fullspace) THEN
917 grid_span = fullspace
918 spherical = .false.
919 odd = .false.
920 ELSE
921 grid_span = halfspace
922 spherical = .false.
923 odd = .true.
924 END IF
925
926 ! actual setup
927 CALL pw_grid_create(grid, para_env, box%hmat, grid_span=grid_span, odd=odd, spherical=spherical, &
928 blocked=blocked_id, npts=np, fft_usage=.true., &
929 rs_dims=distribution_layout, iounit=iw)
930
931 IF (iw > 0) CALL m_flush(iw)
932
933 ! note that the number of grid points might be different from what the user requested (fft-able needed)
934 no = grid%npts
935
936 CALL ca%create(grid)
937 CALL cb%create(grid)
938 CALL cc%create(grid)
939
940 ! initialize data
941 CALL pw_zero(ca)
942 CALL pw_zero(cb)
943 CALL pw_zero(cc)
944 nn = SIZE(ca%array)
945 DO ig = 1, nn
946 gsq = grid%gsq(ig)
947 ca%array(ig) = exp(-gsq)
948 END DO
949
950 flops = product(no)*30.0_dp*log(real(maxval(no), kind=dp))
951 tstart = m_walltime()
952 DO ip = 1, n_loop
953 CALL para_env%sync()
954 t_start(ip) = m_walltime()
955 CALL pw_transfer(ca, cb, debug)
956 CALL pw_transfer(cb, cc, debug)
957 CALL para_env%sync()
958 t_end(ip) = m_walltime()
959 END DO
960 tend = m_walltime()
961 t = tend - tstart + threshold
962 IF (t > 0.0_dp) THEN
963 perf = real(n_loop, kind=dp)*2.0_dp*flops*1.e-6_dp/t
964 ELSE
965 perf = 0.0_dp
966 END IF
967
968 em = maxval(abs(ca%array(:) - cc%array(:)))
969 CALL para_env%max(em)
970 et = sum(abs(ca%array(:) - cc%array(:)))
971 CALL para_env%sum(et)
972 t_min = minval(t_end - t_start)
973 t_max = maxval(t_end - t_start)
974
975 IF (para_env%is_source()) THEN
976 WRITE (iw, *)
977 WRITE (iw, '(A,T67,E14.6)') " Parallel FFT Tests: Maximal Error ", em
978 WRITE (iw, '(A,T67,E14.6)') " Parallel FFT Tests: Total Error ", et
979 WRITE (iw, '(A,T67,F14.0)') &
980 " Parallel FFT Tests: Performance [Mflops] ", perf
981 WRITE (iw, '(A,T67,F14.6)') " Best time : ", t_min
982 WRITE (iw, '(A,T67,F14.6)') " Worst time: ", t_max
983 IF (iw > 0) CALL m_flush(iw)
984 END IF
985
986 ! need debugging ???
987 IF (em > toler .OR. et > toler) THEN
988 cpwarn("The FFT results are not accurate ... starting debug pw_transfer")
989 CALL pw_transfer(ca, cb, .true.)
990 CALL pw_transfer(cb, cc, .true.)
991 END IF
992
993 ! done with these grids
994 CALL ca%release()
995 CALL cb%release()
996 CALL cc%release()
997 CALL pw_grid_release(grid)
998
999 END DO
1000
1001 ! local arrays
1002 DEALLOCATE (layouts)
1003 DEALLOCATE (t_start)
1004 DEALLOCATE (t_end)
1005
1006 END DO
1007
1008 ! cleanup
1009 CALL cell_release(box)
1010 CALL finalize_fft(para_env, wisdom_file=globenv%fftw_wisdom_file_name, &
1011 export_wisdom=globenv%fftw_wisdom_file_explicit)
1012
1013 END SUBROUTINE pw_fft_test
1014
1015! **************************************************************************************************
1016!> \brief Tests the eigensolver library routines
1017!> \param para_env ...
1018!> \param iw ...
1019!> \param eigensolver_section ...
1020!> \par History
1021!> JGH 6-Feb-2001 : Test and performance code
1022!> \author JGH 1-JAN-2001
1023! **************************************************************************************************
1024 SUBROUTINE eigensolver_test(para_env, iw, eigensolver_section)
1025
1026 TYPE(mp_para_env_type), POINTER :: para_env
1027 INTEGER :: iw
1028 TYPE(section_vals_type), POINTER :: eigensolver_section
1029
1030 INTEGER :: diag_method, i, i_loop, i_rep, &
1031 init_method, j, n, n_loop, n_rep, &
1032 neig, unit_number
1033 REAL(kind=dp) :: t1, t2
1034 REAL(kind=dp), ALLOCATABLE, DIMENSION(:) :: eigenvalues
1035 REAL(kind=dp), ALLOCATABLE, DIMENSION(:, :) :: buffer
1036 TYPE(cp_blacs_env_type), POINTER :: blacs_env
1037 TYPE(cp_fm_struct_type), POINTER :: fmstruct
1038 TYPE(cp_fm_type) :: eigenvectors, matrix, work
1039 TYPE(rng_stream_type), ALLOCATABLE :: rng_stream
1040
1041 IF (iw > 0) THEN
1042 WRITE (unit=iw, fmt="(/,/,T2,A,/)") "EIGENSOLVER TEST"
1043 END IF
1044
1045 ! create blacs env corresponding to para_env
1046 NULLIFY (blacs_env)
1047 CALL cp_blacs_env_create(blacs_env=blacs_env, &
1048 para_env=para_env)
1049
1050 ! loop over all tests
1051 CALL section_vals_get(eigensolver_section, n_repetition=n_rep)
1052 DO i_rep = 1, n_rep
1053
1054 ! parse section
1055 CALL section_vals_val_get(eigensolver_section, "N", i_rep_section=i_rep, i_val=n)
1056 CALL section_vals_val_get(eigensolver_section, "EIGENVALUES", i_rep_section=i_rep, i_val=neig)
1057 CALL section_vals_val_get(eigensolver_section, "DIAG_METHOD", i_rep_section=i_rep, i_val=diag_method)
1058 CALL section_vals_val_get(eigensolver_section, "INIT_METHOD", i_rep_section=i_rep, i_val=init_method)
1059 CALL section_vals_val_get(eigensolver_section, "N_loop", i_rep_section=i_rep, i_val=n_loop)
1060
1061 ! proper number of eigs
1062 IF (neig < 0) neig = n
1063 neig = min(neig, n)
1064
1065 ! report
1066 IF (iw > 0) THEN
1067 WRITE (iw, *) "Matrix size", n
1068 WRITE (iw, *) "Number of eigenvalues", neig
1069 WRITE (iw, *) "Timing loops", n_loop
1070 SELECT CASE (diag_method)
1071 CASE (do_diag_syevd)
1072 WRITE (iw, *) "Diag using syevd"
1073 CASE (do_diag_syevx)
1074 WRITE (iw, *) "Diag using syevx"
1075 CASE DEFAULT
1076 ! stop
1077 END SELECT
1078
1079 SELECT CASE (init_method)
1080 CASE (do_mat_random)
1081 WRITE (iw, *) "using random matrix"
1082 CASE (do_mat_read)
1083 WRITE (iw, *) "reading from file"
1084 CASE DEFAULT
1085 ! stop
1086 END SELECT
1087 END IF
1088
1089 ! create matrix struct type
1090 NULLIFY (fmstruct)
1091 CALL cp_fm_struct_create(fmstruct=fmstruct, &
1092 para_env=para_env, &
1093 context=blacs_env, &
1094 nrow_global=n, &
1095 ncol_global=n)
1096
1097 ! create all needed matrices, and buffers for the eigenvalues
1098 CALL cp_fm_create(matrix=matrix, &
1099 matrix_struct=fmstruct, &
1100 name="MATRIX")
1101 CALL cp_fm_set_all(matrix, 0.0_dp)
1102
1103 CALL cp_fm_create(matrix=eigenvectors, &
1104 matrix_struct=fmstruct, &
1105 name="EIGENVECTORS")
1106 CALL cp_fm_set_all(eigenvectors, 0.0_dp)
1107
1108 CALL cp_fm_create(matrix=work, &
1109 matrix_struct=fmstruct, &
1110 name="WORK")
1111 CALL cp_fm_set_all(matrix, 0.0_dp)
1112
1113 ALLOCATE (eigenvalues(n))
1114 eigenvalues = 0.0_dp
1115 ALLOCATE (buffer(1, n))
1116
1117 ! generate initial matrix, either by reading a file, or using random numbers
1118 IF (para_env%is_source()) THEN
1119 SELECT CASE (init_method)
1120 CASE (do_mat_random)
1121 rng_stream = rng_stream_type( &
1122 name="rng_stream", &
1123 distribution_type=uniform, &
1124 extended_precision=.true.)
1125 CASE (do_mat_read)
1126 CALL open_file(file_name="MATRIX", &
1127 file_action="READ", &
1128 file_form="FORMATTED", &
1129 file_status="OLD", &
1130 unit_number=unit_number)
1131 END SELECT
1132 END IF
1133
1134 DO i = 1, n
1135 IF (para_env%is_source()) THEN
1136 SELECT CASE (init_method)
1137 CASE (do_mat_random)
1138 DO j = i, n
1139 buffer(1, j) = rng_stream%next() - 0.5_dp
1140 END DO
1141 !MK activate/modify for a diagonal dominant symmetric matrix:
1142 !MK buffer(1,i) = 10.0_dp*buffer(1,i)
1143 CASE (do_mat_read)
1144 READ (unit=unit_number, fmt=*) buffer(1, 1:n)
1145 END SELECT
1146 END IF
1147 CALL para_env%bcast(buffer)
1148 SELECT CASE (init_method)
1149 CASE (do_mat_random)
1150 CALL cp_fm_set_submatrix(fm=matrix, &
1151 new_values=buffer, &
1152 start_row=i, &
1153 start_col=i, &
1154 n_rows=1, &
1155 n_cols=n - i + 1, &
1156 alpha=1.0_dp, &
1157 beta=0.0_dp, &
1158 transpose=.false.)
1159 CALL cp_fm_set_submatrix(fm=matrix, &
1160 new_values=buffer, &
1161 start_row=i, &
1162 start_col=i, &
1163 n_rows=n - i + 1, &
1164 n_cols=1, &
1165 alpha=1.0_dp, &
1166 beta=0.0_dp, &
1167 transpose=.true.)
1168 CASE (do_mat_read)
1169 CALL cp_fm_set_submatrix(fm=matrix, &
1170 new_values=buffer, &
1171 start_row=i, &
1172 start_col=1, &
1173 n_rows=1, &
1174 n_cols=n, &
1175 alpha=1.0_dp, &
1176 beta=0.0_dp, &
1177 transpose=.false.)
1178 END SELECT
1179 END DO
1180
1181 DEALLOCATE (buffer)
1182
1183 IF (para_env%is_source()) THEN
1184 SELECT CASE (init_method)
1185 CASE (do_mat_read)
1186 CALL close_file(unit_number=unit_number)
1187 END SELECT
1188 END IF
1189
1190 DO i_loop = 1, n_loop
1191 eigenvalues = 0.0_dp
1192 CALL cp_fm_set_all(eigenvectors, 0.0_dp)
1193 CALL cp_fm_to_fm(source=matrix, &
1194 destination=work)
1195
1196 ! DONE, now testing
1197 t1 = m_walltime()
1198 SELECT CASE (diag_method)
1199 CASE (do_diag_syevd)
1200 CALL cp_fm_syevd(matrix=work, &
1201 eigenvectors=eigenvectors, &
1202 eigenvalues=eigenvalues)
1203 CASE (do_diag_syevx)
1204 CALL cp_fm_syevx(matrix=work, &
1205 eigenvectors=eigenvectors, &
1206 eigenvalues=eigenvalues, &
1207 neig=neig, &
1208 work_syevx=1.0_dp)
1209 END SELECT
1210 t2 = m_walltime()
1211 IF (iw > 0) WRITE (iw, *) "Timing for loop ", i_loop, " : ", t2 - t1
1212 END DO
1213
1214 IF (iw > 0) THEN
1215 WRITE (iw, *) "Eigenvalues: "
1216 WRITE (unit=iw, fmt="(T3,5F14.6)") eigenvalues(1:neig)
1217 WRITE (unit=iw, fmt="(T3,A4,F16.6)") "Sum:", sum(eigenvalues(1:neig))
1218 WRITE (iw, *) ""
1219 END IF
1220
1221 ! Clean up
1222 DEALLOCATE (eigenvalues)
1223 CALL cp_fm_release(matrix=work)
1224 CALL cp_fm_release(matrix=eigenvectors)
1225 CALL cp_fm_release(matrix=matrix)
1226 CALL cp_fm_struct_release(fmstruct=fmstruct)
1227
1228 END DO
1229
1230 CALL cp_blacs_env_release(blacs_env=blacs_env)
1231
1232 END SUBROUTINE eigensolver_test
1233
1234! **************************************************************************************************
1235!> \brief Tests the parallel matrix multiply
1236!> \param para_env ...
1237!> \param iw ...
1238!> \param cp_fm_gemm_test_section ...
1239! **************************************************************************************************
1240 SUBROUTINE cp_fm_gemm_test(para_env, iw, cp_fm_gemm_test_section)
1241
1242 TYPE(mp_para_env_type), POINTER :: para_env
1243 INTEGER :: iw
1244 TYPE(section_vals_type), POINTER :: cp_fm_gemm_test_section
1245
1246 CHARACTER(LEN=1) :: transa, transb
1247 INTEGER :: i_loop, i_rep, k, m, n, n_loop, n_rep, ncol_block, ncol_block_actual, &
1248 ncol_global, np, nrow_block, nrow_block_actual, nrow_global
1249 INTEGER, DIMENSION(:), POINTER :: grid_2d
1250 LOGICAL :: force_blocksize, row_major, transa_p, &
1251 transb_p
1252 REAL(kind=dp) :: t1, t2, t3, t4
1253 TYPE(cp_blacs_env_type), POINTER :: blacs_env
1254 TYPE(cp_fm_struct_type), POINTER :: fmstruct_a, fmstruct_b, fmstruct_c
1255 TYPE(cp_fm_type) :: matrix_a, matrix_b, matrix_c
1256
1257 CALL section_vals_get(cp_fm_gemm_test_section, n_repetition=n_rep)
1258 DO i_rep = 1, n_rep
1259
1260 ! how often should we do the multiply
1261 CALL section_vals_val_get(cp_fm_gemm_test_section, "N_loop", i_rep_section=i_rep, i_val=n_loop)
1262
1263 ! matrices def.
1264 CALL section_vals_val_get(cp_fm_gemm_test_section, "K", i_rep_section=i_rep, i_val=k)
1265 CALL section_vals_val_get(cp_fm_gemm_test_section, "N", i_rep_section=i_rep, i_val=n)
1266 CALL section_vals_val_get(cp_fm_gemm_test_section, "M", i_rep_section=i_rep, i_val=m)
1267 CALL section_vals_val_get(cp_fm_gemm_test_section, "transa", i_rep_section=i_rep, l_val=transa_p)
1268 CALL section_vals_val_get(cp_fm_gemm_test_section, "transb", i_rep_section=i_rep, l_val=transb_p)
1269 CALL section_vals_val_get(cp_fm_gemm_test_section, "nrow_block", i_rep_section=i_rep, i_val=nrow_block)
1270 CALL section_vals_val_get(cp_fm_gemm_test_section, "ncol_block", i_rep_section=i_rep, i_val=ncol_block)
1271 CALL section_vals_val_get(cp_fm_gemm_test_section, "ROW_MAJOR", i_rep_section=i_rep, l_val=row_major)
1272 CALL section_vals_val_get(cp_fm_gemm_test_section, "GRID_2D", i_rep_section=i_rep, i_vals=grid_2d)
1273 CALL section_vals_val_get(cp_fm_gemm_test_section, "FORCE_BLOCKSIZE", i_rep_section=i_rep, l_val=force_blocksize)
1274 transa = "N"
1275 transb = "N"
1276 IF (transa_p) transa = "T"
1277 IF (transb_p) transb = "T"
1278
1279 IF (iw > 0) THEN
1280 WRITE (iw, '(T2,A)') "----------- TESTING PARALLEL MATRIX MULTIPLY -------------"
1281 WRITE (iw, '(T2,A)', advance="NO") "C = "
1282 IF (transa_p) THEN
1283 WRITE (iw, '(A)', advance="NO") "TRANSPOSE(A) x"
1284 ELSE
1285 WRITE (iw, '(A)', advance="NO") "A x "
1286 END IF
1287 IF (transb_p) THEN
1288 WRITE (iw, '(A)') "TRANSPOSE(B) "
1289 ELSE
1290 WRITE (iw, '(A)') "B "
1291 END IF
1292 WRITE (iw, '(T2,A,T50,I5,A,I5)') 'requested block size', nrow_block, ' by ', ncol_block
1293 WRITE (iw, '(T2,A,T50,I5)') 'number of repetitions of cp_fm_gemm ', n_loop
1294 WRITE (iw, '(T2,A,T50,L5)') 'Row Major', row_major
1295 WRITE (iw, '(T2,A,T50,2I7)') 'GRID_2D ', grid_2d
1296 WRITE (iw, '(T2,A,T50,L5)') 'Force blocksize ', force_blocksize
1297 ! check the return value of pilaenv, too small values limit the performance (assuming pdgemm is the vanilla variant)
1298 np = cp_fm_pilaenv(0, 'D')
1299 IF (np > 0) THEN
1300 WRITE (iw, '(T2,A,T50,I5)') 'PILAENV blocksize', np
1301 END IF
1302 END IF
1303
1304 NULLIFY (blacs_env)
1305 CALL cp_blacs_env_create(blacs_env=blacs_env, &
1306 para_env=para_env, &
1307 row_major=row_major, &
1308 grid_2d=grid_2d)
1309
1310 NULLIFY (fmstruct_a)
1311 IF (transa_p) THEN
1312 nrow_global = m; ncol_global = k
1313 ELSE
1314 nrow_global = k; ncol_global = m
1315 END IF
1316 CALL cp_fm_struct_create(fmstruct=fmstruct_a, para_env=para_env, context=blacs_env, &
1317 nrow_global=nrow_global, ncol_global=ncol_global, &
1318 nrow_block=nrow_block, ncol_block=ncol_block, force_block=force_blocksize)
1319 CALL cp_fm_struct_get(fmstruct_a, nrow_block=nrow_block_actual, ncol_block=ncol_block_actual)
1320 IF (iw > 0) WRITE (iw, '(T2,A,I9,A,I9,A,I5,A,I5)') 'matrix A ', nrow_global, " by ", ncol_global, &
1321 ' using blocks of ', nrow_block_actual, ' by ', ncol_block_actual
1322
1323 IF (transb_p) THEN
1324 nrow_global = n; ncol_global = m
1325 ELSE
1326 nrow_global = m; ncol_global = n
1327 END IF
1328 NULLIFY (fmstruct_b)
1329 CALL cp_fm_struct_create(fmstruct=fmstruct_b, para_env=para_env, context=blacs_env, &
1330 nrow_global=nrow_global, ncol_global=ncol_global, &
1331 nrow_block=nrow_block, ncol_block=ncol_block, force_block=force_blocksize)
1332 CALL cp_fm_struct_get(fmstruct_b, nrow_block=nrow_block_actual, ncol_block=ncol_block_actual)
1333 IF (iw > 0) WRITE (iw, '(T2,A,I9,A,I9,A,I5,A,I5)') 'matrix B ', nrow_global, " by ", ncol_global, &
1334 ' using blocks of ', nrow_block_actual, ' by ', ncol_block_actual
1335
1336 NULLIFY (fmstruct_c)
1337 nrow_global = k
1338 ncol_global = n
1339 CALL cp_fm_struct_create(fmstruct=fmstruct_c, para_env=para_env, context=blacs_env, &
1340 nrow_global=nrow_global, ncol_global=ncol_global, &
1341 nrow_block=nrow_block, ncol_block=ncol_block, force_block=force_blocksize)
1342 CALL cp_fm_struct_get(fmstruct_c, nrow_block=nrow_block_actual, ncol_block=ncol_block_actual)
1343 IF (iw > 0) WRITE (iw, '(T2,A,I9,A,I9,A,I5,A,I5)') 'matrix C ', nrow_global, " by ", ncol_global, &
1344 ' using blocks of ', nrow_block_actual, ' by ', ncol_block_actual
1345
1346 CALL cp_fm_create(matrix=matrix_a, matrix_struct=fmstruct_a, name="MATRIX A")
1347 CALL cp_fm_create(matrix=matrix_b, matrix_struct=fmstruct_b, name="MATRIX B")
1348 CALL cp_fm_create(matrix=matrix_c, matrix_struct=fmstruct_c, name="MATRIX C")
1349
1350 CALL random_number(matrix_a%local_data)
1351 CALL random_number(matrix_b%local_data)
1352 CALL random_number(matrix_c%local_data)
1353
1354 IF (iw > 0) CALL m_flush(iw)
1355
1356 t1 = m_walltime()
1357 DO i_loop = 1, n_loop
1358 t3 = m_walltime()
1359 CALL parallel_gemm(transa, transb, k, n, m, 1.0_dp, matrix_a, matrix_b, 0.0_dp, matrix_c)
1360 t4 = m_walltime()
1361 IF (iw > 0) THEN
1362 WRITE (iw, '(T2,A,T50,F12.6)') "cp_fm_gemm timing: ", (t4 - t3)
1363 CALL m_flush(iw)
1364 END IF
1365 END DO
1366 t2 = m_walltime()
1367
1368 IF (iw > 0) THEN
1369 WRITE (iw, '(T2,A,T50,F12.6)') "average cp_fm_gemm timing: ", (t2 - t1)/n_loop
1370 IF (t2 > t1) THEN
1371 WRITE (iw, '(T2,A,T50,F12.6)') "cp_fm_gemm Gflops per MPI task: ", &
1372 2*real(m, kind=dp)*real(n, kind=dp)*real(k, kind=dp)*n_loop/max(0.001_dp, t2 - t1)/1.0e9_dp/para_env%num_pe
1373 END IF
1374 END IF
1375
1376 CALL cp_fm_release(matrix=matrix_a)
1377 CALL cp_fm_release(matrix=matrix_b)
1378 CALL cp_fm_release(matrix=matrix_c)
1379 CALL cp_fm_struct_release(fmstruct=fmstruct_a)
1380 CALL cp_fm_struct_release(fmstruct=fmstruct_b)
1381 CALL cp_fm_struct_release(fmstruct=fmstruct_c)
1382 CALL cp_blacs_env_release(blacs_env=blacs_env)
1383
1384 END DO
1385
1386 END SUBROUTINE cp_fm_gemm_test
1387
1388! **************************************************************************************************
1389!> \brief Tests the DBCSR interface.
1390!> \param para_env ...
1391!> \param iw ...
1392!> \param input_section ...
1393! **************************************************************************************************
1394 SUBROUTINE cp_dbcsr_tests(para_env, iw, input_section)
1395
1396 TYPE(mp_para_env_type), POINTER :: para_env
1397 INTEGER :: iw
1398 TYPE(section_vals_type), POINTER :: input_section
1399
1400 CHARACTER, DIMENSION(3) :: types
1401 INTEGER :: data_type, i_rep, k, m, n, n_loop, &
1402 n_rep, test_type
1403 INTEGER, DIMENSION(:), POINTER :: bs_k, bs_m, bs_n, nproc
1404 LOGICAL :: always_checksum, retain_sparsity, &
1405 transa_p, transb_p
1406 REAL(kind=dp) :: alpha, beta, filter_eps, s_a, s_b, s_c
1407
1408! ---------------------------------------------------------------------------
1409
1410 NULLIFY (bs_m, bs_n, bs_k)
1411 CALL section_vals_get(input_section, n_repetition=n_rep)
1412 CALL dbcsr_reset_randmat_seed()
1413 DO i_rep = 1, n_rep
1414 ! how often should we do the multiply
1415 CALL section_vals_val_get(input_section, "N_loop", i_rep_section=i_rep, i_val=n_loop)
1416
1417 ! matrices def.
1418 CALL section_vals_val_get(input_section, "TEST_TYPE", i_rep_section=i_rep, i_val=test_type)
1419 CALL section_vals_val_get(input_section, "DATA_TYPE", i_rep_section=i_rep, i_val=data_type)
1420 CALL section_vals_val_get(input_section, "K", i_rep_section=i_rep, i_val=k)
1421 CALL section_vals_val_get(input_section, "N", i_rep_section=i_rep, i_val=n)
1422 CALL section_vals_val_get(input_section, "M", i_rep_section=i_rep, i_val=m)
1423 CALL section_vals_val_get(input_section, "transa", i_rep_section=i_rep, l_val=transa_p)
1424 CALL section_vals_val_get(input_section, "transb", i_rep_section=i_rep, l_val=transb_p)
1425 CALL section_vals_val_get(input_section, "bs_m", i_rep_section=i_rep, &
1426 i_vals=bs_m)
1427 CALL section_vals_val_get(input_section, "bs_n", i_rep_section=i_rep, &
1428 i_vals=bs_n)
1429 CALL section_vals_val_get(input_section, "bs_k", i_rep_section=i_rep, &
1430 i_vals=bs_k)
1431 CALL section_vals_val_get(input_section, "keepsparse", i_rep_section=i_rep, l_val=retain_sparsity)
1432 CALL section_vals_val_get(input_section, "asparsity", i_rep_section=i_rep, r_val=s_a)
1433 CALL section_vals_val_get(input_section, "bsparsity", i_rep_section=i_rep, r_val=s_b)
1434 CALL section_vals_val_get(input_section, "csparsity", i_rep_section=i_rep, r_val=s_c)
1435 CALL section_vals_val_get(input_section, "alpha", i_rep_section=i_rep, r_val=alpha)
1436 CALL section_vals_val_get(input_section, "beta", i_rep_section=i_rep, r_val=beta)
1437 CALL section_vals_val_get(input_section, "nproc", i_rep_section=i_rep, &
1438 i_vals=nproc)
1439 CALL section_vals_val_get(input_section, "atype", i_rep_section=i_rep, &
1440 c_val=types(1))
1441 CALL section_vals_val_get(input_section, "btype", i_rep_section=i_rep, &
1442 c_val=types(2))
1443 CALL section_vals_val_get(input_section, "ctype", i_rep_section=i_rep, &
1444 c_val=types(3))
1445 CALL section_vals_val_get(input_section, "filter_eps", &
1446 i_rep_section=i_rep, r_val=filter_eps)
1447 CALL section_vals_val_get(input_section, "ALWAYS_CHECKSUM", i_rep_section=i_rep, l_val=always_checksum)
1448
1449 CALL dbcsr_run_tests(para_env%get_handle(), iw, nproc, &
1450 [m, n, k], &
1451 [transa_p, transb_p], &
1452 bs_m, bs_n, bs_k, &
1453 [s_a, s_b, s_c], &
1454 alpha, beta, &
1455 data_type=data_type, &
1456 test_type=test_type, &
1457 n_loops=n_loop, eps=filter_eps, retain_sparsity=retain_sparsity, &
1458 always_checksum=always_checksum)
1459 END DO
1460 END SUBROUTINE cp_dbcsr_tests
1461
1462! **************************************************************************************************
1463!> \brief Tests the DBM library.
1464!> \param para_env ...
1465!> \param iw ...
1466!> \param input_section ...
1467! **************************************************************************************************
1468 SUBROUTINE run_dbm_tests(para_env, iw, input_section)
1469
1470 TYPE(mp_para_env_type), POINTER :: para_env
1471 INTEGER :: iw
1472 TYPE(section_vals_type), POINTER :: input_section
1473
1474 INTEGER :: i_rep, k, m, n, n_loop, n_rep
1475 INTEGER, DIMENSION(:), POINTER :: bs_k, bs_m, bs_n
1476 LOGICAL :: always_checksum, retain_sparsity, &
1477 transa_p, transb_p
1478 REAL(kind=dp) :: alpha, beta, filter_eps, s_a, s_b, s_c
1479
1480! ---------------------------------------------------------------------------
1481
1482 NULLIFY (bs_m, bs_n, bs_k)
1483 CALL section_vals_get(input_section, n_repetition=n_rep)
1484 CALL dbcsr_reset_randmat_seed()
1485 DO i_rep = 1, n_rep
1486 CALL section_vals_val_get(input_section, "N_loop", i_rep_section=i_rep, i_val=n_loop)
1487 CALL section_vals_val_get(input_section, "K", i_rep_section=i_rep, i_val=k)
1488 CALL section_vals_val_get(input_section, "N", i_rep_section=i_rep, i_val=n)
1489 CALL section_vals_val_get(input_section, "M", i_rep_section=i_rep, i_val=m)
1490 CALL section_vals_val_get(input_section, "transa", i_rep_section=i_rep, l_val=transa_p)
1491 CALL section_vals_val_get(input_section, "transb", i_rep_section=i_rep, l_val=transb_p)
1492 CALL section_vals_val_get(input_section, "bs_m", i_rep_section=i_rep, i_vals=bs_m)
1493 CALL section_vals_val_get(input_section, "bs_n", i_rep_section=i_rep, i_vals=bs_n)
1494 CALL section_vals_val_get(input_section, "bs_k", i_rep_section=i_rep, i_vals=bs_k)
1495 CALL section_vals_val_get(input_section, "keepsparse", i_rep_section=i_rep, l_val=retain_sparsity)
1496 CALL section_vals_val_get(input_section, "asparsity", i_rep_section=i_rep, r_val=s_a)
1497 CALL section_vals_val_get(input_section, "bsparsity", i_rep_section=i_rep, r_val=s_b)
1498 CALL section_vals_val_get(input_section, "csparsity", i_rep_section=i_rep, r_val=s_c)
1499 CALL section_vals_val_get(input_section, "alpha", i_rep_section=i_rep, r_val=alpha)
1500 CALL section_vals_val_get(input_section, "beta", i_rep_section=i_rep, r_val=beta)
1501 CALL section_vals_val_get(input_section, "filter_eps", i_rep_section=i_rep, r_val=filter_eps)
1502 CALL section_vals_val_get(input_section, "ALWAYS_CHECKSUM", i_rep_section=i_rep, l_val=always_checksum)
1503
1504 CALL dbm_run_tests(mp_group=para_env, &
1505 io_unit=iw, &
1506 matrix_sizes=[m, n, k], &
1507 trs=[transa_p, transb_p], &
1508 bs_m=bs_m, &
1509 bs_n=bs_n, &
1510 bs_k=bs_k, &
1511 sparsities=[s_a, s_b, s_c], &
1512 alpha=alpha, &
1513 beta=beta, &
1514 n_loops=n_loop, &
1515 eps=filter_eps, &
1516 retain_sparsity=retain_sparsity, &
1517 always_checksum=always_checksum)
1518 END DO
1519 END SUBROUTINE run_dbm_tests
1520
1521END MODULE library_tests
dgemm
static void dgemm(const char transa, const char transb, const int m, const int n, const int k, const double alpha, const double *a, const int lda, const double *b, const int ldb, const double beta, double *c, const int ldc)
Convenient wrapper to hide Fortran nature of dgemm_, swapping a and b.
Definition grid_cpu_task_list.c:225
cp_fm_types::cp_fm_release
Definition cp_fm_types.F:88
cp_fm_types::cp_fm_to_fm
Definition cp_fm_types.F:83
fft_tools::fft3d
Definition fft_tools.F:162
parallel_gemm_api::parallel_gemm
Definition parallel_gemm_api.F:39
pw_grids::pw_grid_create
Definition pw_grids.F:81
pw_methods::pw_transfer
Definition pw_methods.F:304
pw_methods::pw_zero
Definition pw_methods.F:83
ai_coulomb_test
Test of Electron Repulsion Routines (ERI).
Definition ai_coulomb_test.F:14
ai_coulomb_test::threshold
real(kind=dp), parameter threshold
Definition ai_coulomb_test.F:29
ai_coulomb_test::eri_test
subroutine, public eri_test(iw)
...
Definition ai_coulomb_test.F:43
cell_methods
Handles all functions related to the CELL.
Definition cell_methods.F:15
cell_methods::init_cell
subroutine, public init_cell(cell, hmat, periodic)
Initialise/readjust a simulation cell after hmat has been changed.
Definition cell_methods.F:185
cell_methods::cell_create
subroutine, public cell_create(cell, hmat, periodic, tag)
allocates and initializes a cell
Definition cell_methods.F:95
cell_types
Handles all functions related to the CELL.
Definition cell_types.F:15
cell_types::cell_release
subroutine, public cell_release(cell)
releases the given cell (see doc/ReferenceCounting.html)
Definition cell_types.F:704
cg_test
Test of the Gaunt coefficients.
Definition cg_test.F:21
cg_test::gaunt_coefficient_test
subroutine, public gaunt_coefficient_test()
Compare the Gaunt coefficients against a Lebedev quadrature of the triple product of spherical harmon...
Definition cg_test.F:47
cp_blacs_env
methods related to the blacs parallel environment
Definition cp_blacs_env.F:15
cp_blacs_env::cp_blacs_env_release
subroutine, public cp_blacs_env_release(blacs_env)
releases the given blacs_env
Definition cp_blacs_env.F:274
cp_blacs_env::cp_blacs_env_create
subroutine, public cp_blacs_env_create(blacs_env, para_env, blacs_grid_layout, blacs_repeatable, row_major, grid_2d)
allocates and initializes a type that represent a blacs context
Definition cp_blacs_env.F:123
cp_dbcsr_api
Definition cp_dbcsr_api.F:8
cp_eri_mme_interface
Interface to Minimax-Ewald method for periodic ERI's to be used in CP2K.
Definition cp_eri_mme_interface.F:12
cp_eri_mme_interface::cp_eri_mme_perf_acc_test
subroutine, public cp_eri_mme_perf_acc_test(para_env, iw, eri_mme_test_section)
...
Definition cp_eri_mme_interface.F:754
cp_files
Utility routines to open and close files.
Definition cp_files.F:22
cp_files::open_file
subroutine, public open_file(file_name, unit_number, file_status, file_form, file_action, file_position, file_pad, file_access, debug)
Opens a file and returns its logical unit number.
Definition cp_files.F:196
cp_files::close_file
subroutine, public close_file(unit_number, file_status)
Closes an open file given by its logical unit number.
Definition cp_files.F:63
cp_fm_basic_linalg
Basic linear algebra operations for full matrices.
Definition cp_fm_basic_linalg.F:14
cp_fm_basic_linalg::cp_fm_gemm
subroutine, public cp_fm_gemm(transa, transb, m, n, k, alpha, matrix_a, matrix_b, beta, matrix_c, a_first_col, a_first_row, b_first_col, b_first_row, c_first_col, c_first_row)
computes matrix_c = beta * matrix_c + alpha * ( matrix_a ** transa ) * ( matrix_b ** transb )
Definition cp_fm_basic_linalg.F:476
cp_fm_diag
used for collecting some of the diagonalization schemes available for cp_fm_type. cp_fm_power also mo...
Definition cp_fm_diag.F:17
cp_fm_diag::cp_fm_syevd
subroutine, public cp_fm_syevd(matrix, eigenvectors, eigenvalues, info)
Computes all eigenvalues and vectors of a real symmetric matrix significantly faster than syevx,...
Definition cp_fm_diag.F:590
cp_fm_diag::cp_fm_syevx
subroutine, public cp_fm_syevx(matrix, eigenvectors, eigenvalues, neig, work_syevx)
compute eigenvalues and optionally eigenvectors of a real symmetric matrix using scalapack....
Definition cp_fm_diag.F:814
cp_fm_struct
represent the structure of a full matrix
Definition cp_fm_struct.F:14
cp_fm_struct::cp_fm_struct_create
subroutine, public cp_fm_struct_create(fmstruct, para_env, context, nrow_global, ncol_global, nrow_block, ncol_block, descriptor, first_p_pos, local_leading_dimension, template_fmstruct, square_blocks, force_block)
allocates and initializes a full matrix structure
Definition cp_fm_struct.F:132
cp_fm_struct::cp_fm_struct_get
subroutine, public cp_fm_struct_get(fmstruct, para_env, context, descriptor, ncol_block, nrow_block, nrow_global, ncol_global, first_p_pos, row_indices, col_indices, nrow_local, ncol_local, nrow_locals, ncol_locals, local_leading_dimension)
returns the values of various attributes of the matrix structure
Definition cp_fm_struct.F:444
cp_fm_struct::cp_fm_struct_release
subroutine, public cp_fm_struct_release(fmstruct)
releases a full matrix structure
Definition cp_fm_struct.F:355
cp_fm_types
represent a full matrix distributed on many processors
Definition cp_fm_types.F:15
cp_fm_types::cp_fm_set_submatrix
subroutine, public cp_fm_set_submatrix(fm, new_values, start_row, start_col, n_rows, n_cols, alpha, beta, transpose)
sets a submatrix of a full matrix fm(start_row:start_row+n_rows,start_col:start_col+n_cols) = alpha*o...
Definition cp_fm_types.F:756
cp_fm_types::cp_fm_set_all
subroutine, public cp_fm_set_all(matrix, alpha, beta)
set all elements of a matrix to the same value, and optionally the diagonal to a different one
Definition cp_fm_types.F:534
cp_fm_types::cp_fm_create
subroutine, public cp_fm_create(matrix, matrix_struct, name, nrow, ncol, set_zero)
creates a new full matrix with the given structure
Definition cp_fm_types.F:165
cp_fm_types::cp_fm_pilaenv
integer function, public cp_fm_pilaenv(ictxt, prec)
...
Definition cp_fm_types.F:2447
cp_log_handling
various routines to log and control the output. The idea is that decisions about where to log should ...
Definition cp_log_handling.F:41
cp_log_handling::cp_get_default_logger
type(cp_logger_type) function, pointer, public cp_get_default_logger()
returns the default logger
Definition cp_log_handling.F:234
cp_output_handling
routines to handle the output, The idea is to remove the decision of wheter to output and what to out...
Definition cp_output_handling.F:25
cp_output_handling::cp_print_key_unit_nr
integer function, public cp_print_key_unit_nr(logger, basis_section, print_key_path, extension, middle_name, local, log_filename, ignore_should_output, file_form, file_position, file_action, file_status, do_backup, on_file, is_new_file, mpi_io, fout)
...
Definition cp_output_handling.F:853
cp_output_handling::cp_print_key_finished_output
subroutine, public cp_print_key_finished_output(unit_nr, logger, basis_section, print_key_path, local, ignore_should_output, on_file, mpi_io)
should be called after you finish working with a unit obtained with cp_print_key_unit_nr,...
Definition cp_output_handling.F:1073
cp_realspace_grid_init
Definition cp_realspace_grid_init.F:15
cp_realspace_grid_init::init_input_type
subroutine, public init_input_type(input_settings, nsmax, rs_grid_section, ilevel, higher_grid_layout)
parses an input section to assign the proper values to the input type
Definition cp_realspace_grid_init.F:48
dbm_tests
Definition dbm_tests.F:8
dbm_tests::dbm_run_tests
subroutine, public dbm_run_tests(mp_group, io_unit, matrix_sizes, trs, bs_m, bs_n, bs_k, sparsities, alpha, beta, n_loops, eps, retain_sparsity, always_checksum)
Tests the DBM library.
Definition dbm_tests.F:54
fft_tools
Definition fft_tools.F:29
fft_tools::fft_radix_operations
subroutine, public fft_radix_operations(radix_in, radix_out, operation)
Determine the allowed lengths of FFT's '''.
Definition fft_tools.F:233
fft_tools::bwfft
integer, parameter, public bwfft
Definition fft_tools.F:146
fft_tools::finalize_fft
subroutine, public finalize_fft(para_env, wisdom_file, export_wisdom)
does whatever is needed to finalize the current fft setup
Definition fft_tools.F:209
fft_tools::fft_radix_closest
integer, parameter, public fft_radix_closest
Definition fft_tools.F:147
fft_tools::init_fft
subroutine, public init_fft(fftlib, alltoall, pool_limit, wisdom_file, plan_style)
...
Definition fft_tools.F:181
fft_tools::fwfft
integer, parameter, public fwfft
Definition fft_tools.F:146
global_types
Define type storing the global information of a run. Keep the amount of stored data small....
Definition global_types.F:21
input_constants
collects all constants needed in input so that they can be used without circular dependencies
Definition input_constants.F:18
input_constants::do_diag_syevd
integer, parameter, public do_diag_syevd
Definition input_constants.F:78
input_constants::do_mat_read
integer, parameter, public do_mat_read
Definition input_constants.F:81
input_constants::do_pwgrid_ns_fullspace
integer, parameter, public do_pwgrid_ns_fullspace
Definition input_constants.F:254
input_constants::do_diag_syevx
integer, parameter, public do_diag_syevx
Definition input_constants.F:78
input_constants::do_pwgrid_spherical
integer, parameter, public do_pwgrid_spherical
Definition input_constants.F:254
input_constants::do_mat_random
integer, parameter, public do_mat_random
Definition input_constants.F:81
input_constants::do_pwgrid_ns_halfspace
integer, parameter, public do_pwgrid_ns_halfspace
Definition input_constants.F:254
input_section_types
objects that represent the structure of input sections and the data contained in an input section
Definition input_section_types.F:15
input_section_types::section_vals_get_subs_vals
recursive type(section_vals_type) function, pointer, public section_vals_get_subs_vals(section_vals, subsection_name, i_rep_section, can_return_null)
returns the values of the requested subsection
Definition input_section_types.F:735
input_section_types::section_vals_get
subroutine, public section_vals_get(section_vals, ref_count, n_repetition, n_subs_vals_rep, section, explicit)
returns various attributes about the section_vals
Definition input_section_types.F:708
input_section_types::section_vals_val_get
subroutine, public section_vals_val_get(section_vals, keyword_name, i_rep_section, i_rep_val, n_rep_val, val, l_val, i_val, r_val, c_val, l_vals, i_vals, r_vals, c_vals, explicit)
returns the requested value
Definition input_section_types.F:1052
kinds
Defines the basic variable types.
Definition kinds.F:23
kinds::dp
integer, parameter, public dp
Definition kinds.F:34
library_tests
Performance tests for basic tasks like matrix multiplies, copy, fft.
Definition library_tests.F:18
library_tests::lib_test
subroutine, public lib_test(root_section, para_env, globenv)
Master routine for tests.
Definition library_tests.F:123
machine
Machine interface based on Fortran 2003 and POSIX.
Definition machine.F:17
machine::m_flush
subroutine, public m_flush(lunit)
flushes units if the &GLOBAL flag is set accordingly
Definition machine.F:126
machine::m_walltime
real(kind=dp) function, public m_walltime()
returns time from a real-time clock, protected against rolling early/easily
Definition machine.F:143
mathconstants
Definition of mathematical constants and functions.
Definition mathconstants.F:16
mathconstants::gaussi
complex(kind=dp), parameter, public gaussi
Definition mathconstants.F:143
message_passing
Interface to the message passing library MPI.
Definition message_passing.F:27
minimax_exp
Routines to calculate the minimax coefficients in order to approximate 1/x as a sum over exponential ...
Definition minimax_exp.F:29
minimax_exp::validate_exp_minimax
subroutine, public validate_exp_minimax(n_r, iw)
Unit test checking that numerical error of minimax approximations generated using any k15 or k53 coef...
Definition minimax_exp.F:246
mp_perf_test
Interface to the message passing library MPI.
Definition mp_perf_test.F:23
mp_perf_test::mpi_perf_test
subroutine, public mpi_perf_test(comm, npow, output_unit)
Tests the MPI library.
Definition mp_perf_test.F:57
parallel_gemm_api
basic linear algebra operations for full matrixes
Definition parallel_gemm_api.F:14
parallel_rng_types
Parallel (pseudo)random number generator (RNG) for multiple streams and substreams of random numbers.
Definition parallel_rng_types.F:51
parallel_rng_types::advance
subroutine advance(self, e, c)
Advance the state by n steps, i.e. jump n steps forward, if n > 0, or backward if n < 0.
Definition parallel_rng_types.F:228
parallel_rng_types::uniform
integer, parameter, public uniform
Definition parallel_rng_types.F:73
pw_grid_types
Definition pw_grid_types.F:13
pw_grid_types::halfspace
integer, parameter, public halfspace
Definition pw_grid_types.F:26
pw_grid_types::fullspace
integer, parameter, public fullspace
Definition pw_grid_types.F:26
pw_grids
This module defines the grid data type and some basic operations on it.
Definition pw_grids.F:36
pw_grids::pw_grid_release
subroutine, public pw_grid_release(pw_grid)
releases the given pw grid
Definition pw_grids.F:2204
pw_methods
Definition pw_methods.F:25
pw_types
Definition pw_types.F:24
realspace_grid_types
Definition realspace_grid_types.F:18
realspace_grid_types::rs_grid_print
subroutine, public rs_grid_print(rs, iounit)
Print information on grids to output.
Definition realspace_grid_types.F:579
realspace_grid_types::rs_grid_create
subroutine, public rs_grid_create(rs, desc)
...
Definition realspace_grid_types.F:478
realspace_grid_types::rs_grid_create_descriptor
subroutine, public rs_grid_create_descriptor(desc, pw_grid, input_settings, border_points)
Determine the setup of real space grids - this is divided up into the creation of a descriptor and th...
Definition realspace_grid_types.F:202
realspace_grid_types::transfer_pw2rs
subroutine, public transfer_pw2rs(rs, pw)
...
Definition realspace_grid_types.F:695
realspace_grid_types::rs_grid_release_descriptor
subroutine, public rs_grid_release_descriptor(rs_desc)
releases the given rs grid descriptor (see doc/ReferenceCounting.html)
Definition realspace_grid_types.F:2040
realspace_grid_types::transfer_rs2pw
subroutine, public transfer_rs2pw(rs, pw)
...
Definition realspace_grid_types.F:652
realspace_grid_types::rs_grid_release
subroutine, public rs_grid_release(rs_grid)
releases the given rs grid (see doc/ReferenceCounting.html)
Definition realspace_grid_types.F:2020
realspace_grid_types::rs_grid_zero
subroutine, public rs_grid_zero(rs)
Initialize grid to zero.
Definition realspace_grid_types.F:1913
shg_integrals_test
Calculates 2-center integrals for different r12 operators comparing the Solid harmonic Gaussian integ...
Definition shg_integrals_test.F:13
shg_integrals_test::shg_integrals_perf_acc_test
subroutine, public shg_integrals_perf_acc_test(iw, shg_integrals_test_section)
Unit test for performance and accuracy of the SHG integrals.
Definition shg_integrals_test.F:205
time_frequency_grids
Definition and construction of time/frequency grids for correlation methods.
Definition time_frequency_grids.F:13
time_frequency_grids::test_least_square_ft
subroutine, public test_least_square_ft(nr, iw)
test the singular value decomposition for the computation of integration weights for the Fourier tran...
Definition time_frequency_grids.F:450
cell_types::cell_type
Type defining parameters related to the simulation cell.
Definition cell_types.F:60
cp_blacs_env::cp_blacs_env_type
represent a blacs multidimensional parallel environment (for the mpi corrispective see cp_paratypes/m...
Definition cp_blacs_env.F:53
cp_fm_struct::cp_fm_struct_type
keeps the information about the structure of a full matrix
Definition cp_fm_struct.F:82
cp_fm_types::cp_fm_type
represent a full matrix
Definition cp_fm_types.F:114
cp_log_handling::cp_logger_type
type of a logger, at the moment it contains just a print level starting at which level it should be l...
Definition cp_log_handling.F:140
global_types::global_environment_type
contains the initially parsed file and the initial parallel environment
Definition global_types.F:66
input_section_types::section_vals_type
stores the values of a section
Definition input_section_types.F:127
message_passing::mp_para_env_type
stores all the informations relevant to an mpi environment
Definition message_passing.F:768
parallel_rng_types::rng_stream_type
Definition parallel_rng_types.F:157
pw_grid_types::pw_grid_type
Definition pw_grid_types.F:53
pw_types::pw_c1d_gs_type
Definition pw_types.F:127
pw_types::pw_c3d_rs_type
Definition pw_types.F:88
pw_types::pw_r3d_rs_type
Definition pw_types.F:62
realspace_grid_types::realspace_grid_desc_type
Definition realspace_grid_types.F:90
realspace_grid_types::realspace_grid_input_type
Definition realspace_grid_types.F:80
realspace_grid_types::realspace_grid_type
Definition realspace_grid_types.F:136