(git:9111030)
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: clebsch_gordon_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 mp2_grids, ONLY: test_least_square_ft
80 USE mp_perf_test, ONLY: mpi_perf_test
81 USE parallel_gemm_api, ONLY: parallel_gemm
82 USE parallel_rng_types, ONLY: uniform,&
83 rng_stream_type
84 USE pw_grid_types, ONLY: fullspace,&
85 halfspace,&
86 pw_grid_type
87 USE pw_grids, ONLY: pw_grid_create,&
88 pw_grid_release
89 USE pw_methods, ONLY: pw_transfer,&
90 pw_zero
91 USE pw_types, ONLY: pw_c1d_gs_type,&
92 pw_c3d_rs_type,&
93 pw_r3d_rs_type
94 USE realspace_grid_types, ONLY: &
95 realspace_grid_desc_type, realspace_grid_input_type, realspace_grid_type, rs_grid_create, &
96 rs_grid_create_descriptor, rs_grid_print, rs_grid_release, rs_grid_release_descriptor, &
97 rs_grid_zero, transfer_pw2rs, transfer_rs2pw
98 USE shg_integrals_test, ONLY: shg_integrals_perf_acc_test
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 !
158 IF (runtest(4) /= 0) CALL eri_test(iw)
159 !
160 IF (runtest(6) /= 0) CALL clebsch_gordon_test()
161 !
162 ! runtest 7 has been deleted and can be recycled
163 !
164 IF (runtest(8) /= 0) CALL mpi_perf_test(para_env, runtest(8), iw)
165 !
166 IF (runtest(10) /= 0) CALL validate_exp_minimax(runtest(10), iw)
167 !
168 IF (runtest(11) /= 0) CALL test_least_square_ft(runtest(11), iw)
169 !
170
171 rs_pw_transfer_section => section_vals_get_subs_vals(root_section, "TEST%RS_PW_TRANSFER")
172 CALL section_vals_get(rs_pw_transfer_section, explicit=explicit)
173 IF (explicit) THEN
174 CALL rs_pw_transfer_test(para_env, iw, globenv, rs_pw_transfer_section)
175 END IF
176
177 pw_transfer_section => section_vals_get_subs_vals(root_section, "TEST%PW_TRANSFER")
178 CALL section_vals_get(pw_transfer_section, explicit=explicit)
179 IF (explicit) THEN
180 CALL pw_fft_test(para_env, iw, globenv, pw_transfer_section)
181 END IF
182
183 cp_fm_gemm_test_section => section_vals_get_subs_vals(root_section, "TEST%CP_FM_GEMM")
184 CALL section_vals_get(cp_fm_gemm_test_section, explicit=explicit)
185 IF (explicit) THEN
186 CALL cp_fm_gemm_test(para_env, iw, cp_fm_gemm_test_section)
187 END IF
188
189 eigensolver_section => section_vals_get_subs_vals(root_section, "TEST%EIGENSOLVER")
190 CALL section_vals_get(eigensolver_section, explicit=explicit)
191 IF (explicit) THEN
192 CALL eigensolver_test(para_env, iw, eigensolver_section)
193 END IF
194
195 eri_mme_test_section => section_vals_get_subs_vals(root_section, "TEST%ERI_MME_TEST")
196 CALL section_vals_get(eri_mme_test_section, explicit=explicit)
197 IF (explicit) THEN
198 CALL cp_eri_mme_perf_acc_test(para_env, iw, eri_mme_test_section)
199 END IF
200
201 shg_integrals_test_section => section_vals_get_subs_vals(root_section, "TEST%SHG_INTEGRALS_TEST")
202 CALL section_vals_get(shg_integrals_test_section, explicit=explicit)
203 IF (explicit) THEN
204 CALL shg_integrals_perf_acc_test(iw, shg_integrals_test_section)
205 END IF
206
207 ! DBCSR tests
208 cp_dbcsr_test_section => section_vals_get_subs_vals(root_section, "TEST%CP_DBCSR")
209 CALL section_vals_get(cp_dbcsr_test_section, explicit=explicit)
210 IF (explicit) THEN
211 CALL cp_dbcsr_tests(para_env, iw, cp_dbcsr_test_section)
212 END IF
213
214 ! DBM tests
215 dbm_test_section => section_vals_get_subs_vals(root_section, "TEST%DBM")
216 CALL section_vals_get(dbm_test_section, explicit=explicit)
217 IF (explicit) THEN
218 CALL run_dbm_tests(para_env, iw, dbm_test_section)
219 END IF
220
221 CALL cp_print_key_finished_output(iw, logger, root_section, "TEST%PROGRAM_RUN_INFO")
222
223 CALL timestop(handle)
224
225 END SUBROUTINE lib_test
226
227! **************************************************************************************************
228!> \brief Reads input section &TEST ... &END
229!> \param root_section ...
230!> \param para_env ...
231!> \author JGH 30-NOV-2000
232!> \note
233!> I---------------------------------------------------------------------------I
234!> I SECTION: &TEST ... &END I
235!> I I
236!> I MEMORY max_memory I
237!> I COPY n I
238!> I MATMUL n I
239!> I FFT n I
240!> I ERI n I
241!> I PW_FFT n I
242!> I Clebsch-Gordon n I
243!> I RS_GRIDS n I
244!> I MPI n I
245!> I RNG n -> Parallel random number generator I
246!> I---------------------------------------------------------------------------I
247! **************************************************************************************************
248 SUBROUTINE test_input(root_section, para_env)
249 TYPE(section_vals_type), POINTER :: root_section
250 TYPE(mp_para_env_type), POINTER :: para_env
251
252 TYPE(section_vals_type), POINTER :: test_section
253
254!
255!..defaults
256! using this style is not recommended, introduce sections instead (see e.g. cp_fm_gemm)
257
258 runtest = 0
259 test_section => section_vals_get_subs_vals(root_section, "TEST")
260 CALL section_vals_val_get(test_section, "MEMORY", r_val=max_memory)
261 CALL section_vals_val_get(test_section, 'COPY', i_val=runtest(1))
262 CALL section_vals_val_get(test_section, 'MATMUL', i_val=runtest(2))
263 CALL section_vals_val_get(test_section, 'DGEMM', i_val=runtest(5))
264 CALL section_vals_val_get(test_section, 'FFT', i_val=runtest(3))
265 CALL section_vals_val_get(test_section, 'ERI', i_val=runtest(4))
266 CALL section_vals_val_get(test_section, 'CLEBSCH_GORDON', i_val=runtest(6))
267 CALL section_vals_val_get(test_section, 'MPI', i_val=runtest(8))
268 CALL section_vals_val_get(test_section, 'MINIMAX', i_val=runtest(10))
269 CALL section_vals_val_get(test_section, 'LEAST_SQ_FT', i_val=runtest(11))
270
271 CALL para_env%sync()
272 END SUBROUTINE test_input
273
274! **************************************************************************************************
275!> \brief Tests the performance to copy two vectors.
276!> \param para_env ...
277!> \param iw ...
278!> \par History
279!> none
280!> \author JGH 6-NOV-2000
281!> \note
282!> The results of these tests allow to determine the size of the cache
283!> of the CPU. This can be used to optimize the performance of the
284!> FFTSG library.
285! **************************************************************************************************
286 SUBROUTINE copy_test(para_env, iw)
287 TYPE(mp_para_env_type), POINTER :: para_env
288 INTEGER :: iw
289
290 INTEGER :: i, j, len, ntim, siz
291 REAL(kind=dp) :: perf, t, tend, tstart
292 REAL(kind=dp), ALLOCATABLE, DIMENSION(:) :: ca, cb
293
294! test for copy --> Cache size
295
296 siz = abs(runtest(1))
297 IF (para_env%is_source()) WRITE (iw, '(//,A,/)') " Test of copy ( F95 ) "
298 DO i = 6, 24
299 len = 2**i
300 IF (8.0_dp*real(len, kind=dp) > max_memory*0.5_dp) EXIT
301 ALLOCATE (ca(len))
302 ALLOCATE (cb(len))
303
304 CALL random_number(ca)
305 ntim = nint(1.e7_dp/real(len, kind=dp))
306 ntim = max(ntim, 1)
307 ntim = min(ntim, siz*10000)
308
309 tstart = m_walltime()
310 DO j = 1, ntim
311 cb(:) = ca(:)
312 ca(1) = real(j, kind=dp)
313 END DO
314 tend = m_walltime()
315 t = tend - tstart + threshold
316 IF (t > 0.0_dp) THEN
317 perf = real(ntim, kind=dp)*real(len, kind=dp)*1.e-6_dp/t
318 ELSE
319 perf = 0.0_dp
320 END IF
321
322 IF (para_env%is_source()) THEN
323 WRITE (iw, '(A,i2,i10,A,T59,F14.4,A)') " Copy test: Size = 2^", i, &
324 len/1024, " Kwords", perf, " Mcopy/s"
325 END IF
326
327 DEALLOCATE (ca)
328 DEALLOCATE (cb)
329 END DO
330 CALL para_env%sync()
331 END SUBROUTINE copy_test
332
333! **************************************************************************************************
334!> \brief Tests the performance of different kinds of matrix matrix multiply
335!> kernels for the BLAS and F95 intrinsic matmul.
336!> \param para_env ...
337!> \param test_matmul ...
338!> \param test_dgemm ...
339!> \param iw ...
340!> \par History
341!> none
342!> \author JGH 6-NOV-2000
343! **************************************************************************************************
344 SUBROUTINE matmul_test(para_env, test_matmul, test_dgemm, iw)
345 TYPE(mp_para_env_type), POINTER :: para_env
346 LOGICAL :: test_matmul, test_dgemm
347 INTEGER :: iw
348
349 INTEGER :: i, j, len, ntim, siz
350 REAL(kind=dp) :: perf, t, tend, tstart, xdum
351 REAL(kind=dp), ALLOCATABLE, DIMENSION(:, :) :: ma, mb, mc
352
353! test for matrix multpies
354
355 IF (test_matmul) THEN
356 siz = abs(runtest(2))
357 IF (para_env%is_source()) WRITE (iw, '(//,A,/)') " Test of matmul ( F95 ) "
358 DO i = 5, siz, 2
359 len = 2**i + 1
360 IF (8.0_dp*real(len*len, kind=dp) > max_memory*0.3_dp) EXIT
361 ALLOCATE (ma(len, len))
362 ALLOCATE (mb(len, len))
363 ALLOCATE (mc(len, len))
364 mc = 0.0_dp
365
366 CALL random_number(xdum)
367 ma = xdum
368 CALL random_number(xdum)
369 mb = xdum
370 ntim = nint(1.e8_dp/(2.0_dp*real(len, kind=dp)**3))
371 ntim = max(ntim, 1)
372 ntim = min(ntim, siz*200)
373 tstart = m_walltime()
374 DO j = 1, ntim
375 mc(:, :) = matmul(ma, mb)
376 ma(1, 1) = real(j, kind=dp)
377 END DO
378 tend = m_walltime()
379 t = tend - tstart + threshold
380 perf = real(ntim, kind=dp)*2.0_dp*real(len, kind=dp)**3*1.e-6_dp/t
381 IF (para_env%is_source()) THEN
382 WRITE (iw, '(A,i6,T59,F14.4,A)') &
383 " Matrix multiply test: c = a * b Size = ", len, perf, " Mflop/s"
384 END IF
385 tstart = m_walltime()
386 DO j = 1, ntim
387 mc(:, :) = mc + matmul(ma, mb)
388 ma(1, 1) = real(j, kind=dp)
389 END DO
390 tend = m_walltime()
391 t = tend - tstart + threshold
392 IF (t > 0.0_dp) THEN
393 perf = real(ntim, kind=dp)*2.0_dp*real(len, kind=dp)**3*1.e-6_dp/t
394 ELSE
395 perf = 0.0_dp
396 END IF
397
398 IF (para_env%is_source()) THEN
399 WRITE (iw, '(A,i6,T59,F14.4,A)') &
400 " Matrix multiply test: a = a * b Size = ", len, perf, " Mflop/s"
401 END IF
402
403 tstart = m_walltime()
404 DO j = 1, ntim
405 mc(:, :) = mc + matmul(ma, transpose(mb))
406 ma(1, 1) = real(j, kind=dp)
407 END DO
408 tend = m_walltime()
409 t = tend - tstart + threshold
410 IF (t > 0.0_dp) THEN
411 perf = real(ntim, kind=dp)*2.0_dp*real(len, kind=dp)**3*1.e-6_dp/t
412 ELSE
413 perf = 0.0_dp
414 END IF
415
416 IF (para_env%is_source()) THEN
417 WRITE (iw, '(A,i6,T59,F14.4,A)') &
418 " Matrix multiply test: c = a * b(T) Size = ", len, perf, " Mflop/s"
419 END IF
420
421 tstart = m_walltime()
422 DO j = 1, ntim
423 mc(:, :) = mc + matmul(transpose(ma), mb)
424 ma(1, 1) = real(j, kind=dp)
425 END DO
426 tend = m_walltime()
427 t = tend - tstart + threshold
428 IF (t > 0.0_dp) THEN
429 perf = real(ntim, kind=dp)*2.0_dp*real(len, kind=dp)**3*1.e-6_dp/t
430 ELSE
431 perf = 0.0_dp
432 END IF
433
434 IF (para_env%is_source()) THEN
435 WRITE (iw, '(A,i6,T59,F14.4,A)') &
436 " Matrix multiply test: c = a(T) * b Size = ", len, perf, " Mflop/s"
437 END IF
438
439 DEALLOCATE (ma)
440 DEALLOCATE (mb)
441 DEALLOCATE (mc)
442 END DO
443 END IF
444
445 ! test for matrix multpies
446 IF (test_dgemm) THEN
447 siz = abs(runtest(5))
448 IF (para_env%is_source()) WRITE (iw, '(//,A,/)') " Test of matmul ( BLAS ) "
449 DO i = 5, siz, 2
450 len = 2**i + 1
451 IF (8.0_dp*real(len*len, kind=dp) > max_memory*0.3_dp) EXIT
452 ALLOCATE (ma(len, len))
453 ALLOCATE (mb(len, len))
454 ALLOCATE (mc(len, len))
455 mc = 0.0_dp
456
457 CALL random_number(xdum)
458 ma = xdum
459 CALL random_number(xdum)
460 mb = xdum
461 ntim = nint(1.e8_dp/(2.0_dp*real(len, kind=dp)**3))
462 ntim = max(ntim, 1)
463 ntim = min(ntim, 1000)
464
465 tstart = m_walltime()
466 DO j = 1, ntim
467 CALL dgemm("N", "N", len, len, len, 1.0_dp, ma, len, mb, len, 1.0_dp, mc, len)
468 END DO
469 tend = m_walltime()
470 t = tend - tstart + threshold
471 IF (t > 0.0_dp) THEN
472 perf = real(ntim, kind=dp)*2.0_dp*real(len, kind=dp)**3*1.e-6_dp/t
473 ELSE
474 perf = 0.0_dp
475 END IF
476
477 IF (para_env%is_source()) THEN
478 WRITE (iw, '(A,i6,T59,F14.4,A)') &
479 " Matrix multiply test: c = a * b Size = ", len, perf, " Mflop/s"
480 END IF
481
482 tstart = m_walltime()
483 DO j = 1, ntim
484 CALL dgemm("N", "N", len, len, len, 1.0_dp, ma, len, mb, len, 1.0_dp, mc, len)
485 END DO
486 tend = m_walltime()
487 t = tend - tstart + threshold
488 IF (t > 0.0_dp) THEN
489 perf = real(ntim, kind=dp)*2.0_dp*real(len, kind=dp)**3*1.e-6_dp/t
490 ELSE
491 perf = 0.0_dp
492 END IF
493
494 IF (para_env%is_source()) THEN
495 WRITE (iw, '(A,i6,T59,F14.4,A)') &
496 " Matrix multiply test: a = a * b Size = ", len, perf, " Mflop/s"
497 END IF
498
499 tstart = m_walltime()
500 DO j = 1, ntim
501 CALL dgemm("N", "T", len, len, len, 1.0_dp, ma, len, mb, len, 1.0_dp, mc, len)
502 END DO
503 tend = m_walltime()
504 t = tend - tstart + threshold
505 IF (t > 0.0_dp) THEN
506 perf = real(ntim, kind=dp)*2.0_dp*real(len, kind=dp)**3*1.e-6_dp/t
507 ELSE
508 perf = 0.0_dp
509 END IF
510
511 IF (para_env%is_source()) THEN
512 WRITE (iw, '(A,i6,T59,F14.4,A)') &
513 " Matrix multiply test: c = a * b(T) Size = ", len, perf, " Mflop/s"
514 END IF
515
516 tstart = m_walltime()
517 DO j = 1, ntim
518 CALL dgemm("T", "N", len, len, len, 1.0_dp, ma, len, mb, len, 1.0_dp, mc, len)
519 END DO
520 tend = m_walltime()
521 t = tend - tstart + threshold
522 IF (t > 0.0_dp) THEN
523 perf = real(ntim, kind=dp)*2.0_dp*real(len, kind=dp)**3*1.e-6_dp/t
524 ELSE
525 perf = 0.0_dp
526 END IF
527
528 IF (para_env%is_source()) THEN
529 WRITE (iw, '(A,i6,T59,F14.4,A)') &
530 " Matrix multiply test: c = a(T) * b Size = ", len, perf, " Mflop/s"
531 END IF
532
533 DEALLOCATE (ma)
534 DEALLOCATE (mb)
535 DEALLOCATE (mc)
536 END DO
537 END IF
538
539 CALL para_env%sync()
540
541 END SUBROUTINE matmul_test
542
543! **************************************************************************************************
544!> \brief Tests the performance of all available FFT libraries for 3D FFTs
545!> \param para_env ...
546!> \param iw ...
547!> \param fftw_plan_type ...
548!> \param wisdom_file where FFTW3 should look to save/load wisdom
549!> \par History
550!> none
551!> \author JGH 6-NOV-2000
552! **************************************************************************************************
553 SUBROUTINE fft_test(para_env, iw, fftw_plan_type, wisdom_file)
554
555 TYPE(mp_para_env_type), POINTER :: para_env
556 INTEGER :: iw, fftw_plan_type
557 CHARACTER(LEN=*), INTENT(IN) :: wisdom_file
558
559 INTEGER, PARAMETER :: ndate(3) = [12, 48, 96]
560
561 INTEGER :: iall, it, j, len, n(3), ntim, radix_in, &
562 radix_out, siz, stat
563 COMPLEX(KIND=dp), DIMENSION(4, 4, 4) :: zz
564 COMPLEX(KIND=dp), ALLOCATABLE, DIMENSION(:, :, :) :: ca, cb, cc
565 CHARACTER(LEN=7) :: method
566 REAL(kind=dp) :: flops, perf, scale, t, tdiff, tend, &
567 tstart
568 REAL(kind=dp), ALLOCATABLE, DIMENSION(:, :, :) :: ra
569
570! test for 3d FFT
571
572 IF (para_env%is_source()) WRITE (iw, '(//,A,/)') " Test of 3D-FFT "
573 siz = abs(runtest(3))
574
575 DO iall = 1, 100
576 SELECT CASE (iall)
577 CASE DEFAULT
578 EXIT
579 CASE (1)
580 CALL init_fft("FFTSG", alltoall=.false., fftsg_sizes=.true., wisdom_file=wisdom_file, &
581 pool_limit=10, plan_style=fftw_plan_type)
582 method = "FFTSG "
583 CASE (2)
584 cycle
585 CASE (3)
586 CALL init_fft("FFTW3", alltoall=.false., fftsg_sizes=.true., wisdom_file=wisdom_file, &
587 pool_limit=10, plan_style=fftw_plan_type)
588 method = "FFTW3 "
589 END SELECT
590 n = 4
591 zz = 0.0_dp
592 CALL fft3d(fwfft, n, zz, status=stat)
593 IF (stat == 0) THEN
594 DO it = 1, 3
595 radix_in = ndate(it)
596 CALL fft_radix_operations(radix_in, radix_out, fft_radix_closest)
597 len = radix_out
598 n = len
599 IF (16.0_dp*real(len*len*len, kind=dp) > max_memory*0.5_dp) EXIT
600 ALLOCATE (ra(len, len, len))
601 ALLOCATE (ca(len, len, len))
602 CALL random_number(ra)
603 ca(:, :, :) = ra
604 CALL random_number(ra)
605 ca(:, :, :) = ca + gaussi*ra
606 flops = real(len**3, kind=dp)*15.0_dp*log(real(len, kind=dp))
607 ntim = nint(siz*1.e7_dp/flops)
608 ntim = max(ntim, 1)
609 ntim = min(ntim, 200)
610 scale = 1.0_dp/real(len**3, kind=dp)
611 tstart = m_walltime()
612 DO j = 1, ntim
613 CALL fft3d(fwfft, n, ca)
614 CALL fft3d(bwfft, n, ca)
615 END DO
616 tend = m_walltime()
617 t = tend - tstart + threshold
618 IF (t > 0.0_dp) THEN
619 perf = real(ntim, kind=dp)*2.0_dp*flops*1.e-6_dp/t
620 ELSE
621 perf = 0.0_dp
622 END IF
623
624 IF (para_env%is_source()) THEN
625 WRITE (iw, '(T2,A,A,i6,T59,F14.4,A)') &
626 adjustr(method), " test (in-place) Size = ", len, perf, " Mflop/s"
627 END IF
628 DEALLOCATE (ca)
629 DEALLOCATE (ra)
630 END DO
631 IF (para_env%is_source()) WRITE (iw, *)
632 ! test if input data is preserved
633 len = 24
634 n = len
635 ALLOCATE (ra(len, len, len))
636 ALLOCATE (ca(len, len, len))
637 ALLOCATE (cb(len, len, len))
638 ALLOCATE (cc(len, len, len))
639 CALL random_number(ra)
640 ca(:, :, :) = ra
641 CALL random_number(ra)
642 ca(:, :, :) = ca + gaussi*ra
643 cc(:, :, :) = ca
644 CALL fft3d(fwfft, n, ca, cb)
645 tdiff = maxval(abs(ca - cc))
646 IF (tdiff > 1.0e-12_dp) THEN
647 IF (para_env%is_source()) THEN
648 WRITE (iw, '(T2,A,A,A)') adjustr(method), " FWFFT ", &
649 " Input array is changed in out-of-place FFT !"
650 END IF
651 ELSE
652 IF (para_env%is_source()) THEN
653 WRITE (iw, '(T2,A,A,A)') adjustr(method), " FWFFT ", &
654 " Input array is not changed in out-of-place FFT !"
655 END IF
656 END IF
657 ca(:, :, :) = cc
658 CALL fft3d(bwfft, n, ca, cb)
659 tdiff = maxval(abs(ca - cc))
660 IF (tdiff > 1.0e-12_dp) THEN
661 IF (para_env%is_source()) THEN
662 WRITE (iw, '(T2,A,A,A)') adjustr(method), " BWFFT ", &
663 " Input array is changed in out-of-place FFT !"
664 END IF
665 ELSE
666 IF (para_env%is_source()) THEN
667 WRITE (iw, '(T2,A,A,A)') adjustr(method), " BWFFT ", &
668 " Input array is not changed in out-of-place FFT !"
669 END IF
670 END IF
671 IF (para_env%is_source()) WRITE (iw, *)
672
673 DEALLOCATE (ra)
674 DEALLOCATE (ca)
675 DEALLOCATE (cb)
676 DEALLOCATE (cc)
677 END IF
678 CALL finalize_fft(para_env, wisdom_file=wisdom_file)
679 END DO
680
681 END SUBROUTINE fft_test
682
683! **************************************************************************************************
684!> \brief test rs_pw_transfer performance
685!> \param para_env ...
686!> \param iw ...
687!> \param globenv ...
688!> \param rs_pw_transfer_section ...
689!> \author Joost VandeVondele
690!> 9.2008 Randomise rs grid [Iain Bethune]
691!> (c) The Numerical Algorithms Group (NAG) Ltd, 2008 on behalf of the HECToR project
692! **************************************************************************************************
693 SUBROUTINE rs_pw_transfer_test(para_env, iw, globenv, rs_pw_transfer_section)
694
695 TYPE(mp_para_env_type), POINTER :: para_env
696 INTEGER :: iw
697 TYPE(global_environment_type), POINTER :: globenv
698 TYPE(section_vals_type), POINTER :: rs_pw_transfer_section
699
700 CHARACTER(LEN=*), PARAMETER :: routinen = 'rs_pw_transfer_test'
701
702 INTEGER :: halo_size, handle, i_loop, n_loop, ns_max
703 INTEGER, DIMENSION(3) :: no, np
704 INTEGER, DIMENSION(:), POINTER :: i_vals
705 LOGICAL :: do_rs2pw
706 REAL(kind=dp) :: tend, tstart
707 TYPE(cell_type), POINTER :: box
708 TYPE(pw_grid_type), POINTER :: grid
709 TYPE(pw_r3d_rs_type) :: ca
710 TYPE(realspace_grid_desc_type), POINTER :: rs_desc
711 TYPE(realspace_grid_input_type) :: input_settings
712 TYPE(realspace_grid_type) :: rs_grid
713 TYPE(section_vals_type), POINTER :: rs_grid_section
714
715 CALL timeset(routinen, handle)
716
717 !..set fft lib
718 CALL init_fft(globenv%default_fft_library, alltoall=.false., fftsg_sizes=.true., &
719 pool_limit=globenv%fft_pool_scratch_limit, &
720 wisdom_file=globenv%fftw_wisdom_file_name, &
721 plan_style=globenv%fftw_plan_type)
722
723 ! .. set cell (should otherwise be irrelevant)
724 NULLIFY (box)
725 CALL cell_create(box)
726 box%hmat = reshape([20.0_dp, 0.0_dp, 0.0_dp, 0.0_dp, 20.0_dp, 0.0_dp, &
727 0.0_dp, 0.0_dp, 20.0_dp], [3, 3])
728 CALL init_cell(box)
729
730 ! .. grid type and pw_grid
731 NULLIFY (grid)
732 CALL section_vals_val_get(rs_pw_transfer_section, "GRID", i_vals=i_vals)
733 np = i_vals
734 CALL pw_grid_create(grid, para_env, box%hmat, grid_span=fullspace, npts=np, fft_usage=.true., iounit=iw)
735 no = grid%npts
736
737 CALL ca%create(grid)
738 CALL pw_zero(ca)
739
740 ! .. rs input setting type
741 CALL section_vals_val_get(rs_pw_transfer_section, "HALO_SIZE", i_val=halo_size)
742 rs_grid_section => section_vals_get_subs_vals(rs_pw_transfer_section, "RS_GRID")
743 ns_max = 2*halo_size + 1
744 CALL init_input_type(input_settings, ns_max, rs_grid_section, 1, [-1, -1, -1])
745
746 ! .. rs type
747 NULLIFY (rs_desc)
748 CALL rs_grid_create_descriptor(rs_desc, pw_grid=grid, input_settings=input_settings)
749 CALL rs_grid_create(rs_grid, rs_desc)
750 CALL rs_grid_print(rs_grid, iw)
751 CALL rs_grid_zero(rs_grid)
752
753 ! Put random values on the grid, so summation check will pick up errors
754 CALL random_number(rs_grid%r)
755
756 CALL section_vals_val_get(rs_pw_transfer_section, "N_loop", i_val=n_loop)
757 CALL section_vals_val_get(rs_pw_transfer_section, "RS2PW", l_val=do_rs2pw)
758
759 ! go for the real loops, sync to get max timings
760 IF (para_env%is_source()) THEN
761 WRITE (iw, '(T2,A)') ""
762 WRITE (iw, '(T2,A)') "Timing rs_pw_transfer routine"
763 WRITE (iw, '(T2,A)') ""
764 WRITE (iw, '(T2,A)') "iteration time[s]"
765 END IF
766 DO i_loop = 1, n_loop
767 CALL para_env%sync()
768 tstart = m_walltime()
769 IF (do_rs2pw) THEN
770 CALL transfer_rs2pw(rs_grid, ca)
771 ELSE
772 CALL transfer_pw2rs(rs_grid, ca)
773 END IF
774 CALL para_env%sync()
775 tend = m_walltime()
776 IF (para_env%is_source()) THEN
777 WRITE (iw, '(T2,I9,1X,F12.6)') i_loop, tend - tstart
778 END IF
779 END DO
780
781 !cleanup
782 CALL rs_grid_release(rs_grid)
783 CALL rs_grid_release_descriptor(rs_desc)
784 CALL ca%release()
785 CALL pw_grid_release(grid)
786 CALL cell_release(box)
787 CALL finalize_fft(para_env, wisdom_file=globenv%fftw_wisdom_file_name)
788
789 CALL timestop(handle)
790
791 END SUBROUTINE rs_pw_transfer_test
792
793! **************************************************************************************************
794!> \brief Tests the performance of PW calls to FFT routines
795!> \param para_env ...
796!> \param iw ...
797!> \param globenv ...
798!> \param pw_transfer_section ...
799!> \par History
800!> JGH 6-Feb-2001 : Test and performance code
801!> Made input sensitive [Joost VandeVondele]
802!> \author JGH 1-JAN-2001
803! **************************************************************************************************
804 SUBROUTINE pw_fft_test(para_env, iw, globenv, pw_transfer_section)
805
806 TYPE(mp_para_env_type), POINTER :: para_env
807 INTEGER :: iw
808 TYPE(global_environment_type), POINTER :: globenv
809 TYPE(section_vals_type), POINTER :: pw_transfer_section
810
811 REAL(kind=dp), PARAMETER :: toler = 1.e-11_dp
812
813 INTEGER :: blocked_id, grid_span, i_layout, i_rep, &
814 ig, ip, itmp, n_loop, n_rep, nn, p, q
815 INTEGER, ALLOCATABLE, DIMENSION(:, :) :: layouts
816 INTEGER, DIMENSION(2) :: distribution_layout
817 INTEGER, DIMENSION(3) :: no, np
818 INTEGER, DIMENSION(:), POINTER :: i_vals
819 LOGICAL :: debug, is_fullspace, odd, &
820 pw_grid_layout_all, spherical
821 REAL(kind=dp) :: em, et, flops, gsq, perf, t, t_max, &
822 t_min, tend, tstart
823 REAL(kind=dp), ALLOCATABLE, DIMENSION(:) :: t_end, t_start
824 TYPE(cell_type), POINTER :: box
825 TYPE(pw_c1d_gs_type) :: ca, cc
826 TYPE(pw_c3d_rs_type) :: cb
827 TYPE(pw_grid_type), POINTER :: grid
828
829!..set fft lib
830
831 CALL init_fft(globenv%default_fft_library, alltoall=.false., fftsg_sizes=.true., &
832 pool_limit=globenv%fft_pool_scratch_limit, &
833 wisdom_file=globenv%fftw_wisdom_file_name, &
834 plan_style=globenv%fftw_plan_type)
835
836 !..the unit cell (should not really matter, the number of grid points do)
837 NULLIFY (box, grid)
838 CALL cell_create(box)
839 box%hmat = reshape([10.0_dp, 0.0_dp, 0.0_dp, 0.0_dp, 8.0_dp, 0.0_dp, &
840 0.0_dp, 0.0_dp, 7.0_dp], [3, 3])
841 CALL init_cell(box)
842
843 CALL section_vals_get(pw_transfer_section, n_repetition=n_rep)
844 DO i_rep = 1, n_rep
845
846 ! how often should we do the transfer
847 CALL section_vals_val_get(pw_transfer_section, "N_loop", i_rep_section=i_rep, i_val=n_loop)
848 ALLOCATE (t_start(n_loop))
849 ALLOCATE (t_end(n_loop))
850
851 ! setup of the grids
852 CALL section_vals_val_get(pw_transfer_section, "GRID", i_rep_section=i_rep, i_vals=i_vals)
853 np = i_vals
854
855 CALL section_vals_val_get(pw_transfer_section, "PW_GRID_BLOCKED", i_rep_section=i_rep, i_val=blocked_id)
856 CALL section_vals_val_get(pw_transfer_section, "DEBUG", i_rep_section=i_rep, l_val=debug)
857
858 CALL section_vals_val_get(pw_transfer_section, "PW_GRID_LAYOUT_ALL", i_rep_section=i_rep, &
859 l_val=pw_grid_layout_all)
860
861 ! prepare to loop over all or a specific layout
862 IF (pw_grid_layout_all) THEN
863 ! count layouts that fit
864 itmp = 0
865 ! start from 2, (/1,para_env%num_pe/) is not supported
866 DO p = 2, para_env%num_pe
867 q = para_env%num_pe/p
868 IF (p*q == para_env%num_pe) THEN
869 itmp = itmp + 1
870 END IF
871 END DO
872 ! build list
873 ALLOCATE (layouts(2, itmp))
874 itmp = 0
875 DO p = 2, para_env%num_pe
876 q = para_env%num_pe/p
877 IF (p*q == para_env%num_pe) THEN
878 itmp = itmp + 1
879 layouts(:, itmp) = [p, q]
880 END IF
881 END DO
882 ELSE
883 CALL section_vals_val_get(pw_transfer_section, "PW_GRID_LAYOUT", i_rep_section=i_rep, i_vals=i_vals)
884 ALLOCATE (layouts(2, 1))
885 layouts(:, 1) = i_vals
886 END IF
887
888 DO i_layout = 1, SIZE(layouts, 2)
889
890 distribution_layout = layouts(:, i_layout)
891
892 CALL section_vals_val_get(pw_transfer_section, "PW_GRID", i_rep_section=i_rep, i_val=itmp)
893
894 ! from cp_control_utils
895 SELECT CASE (itmp)
896 CASE (do_pwgrid_spherical)
897 spherical = .true.
898 is_fullspace = .false.
899 CASE (do_pwgrid_ns_fullspace)
900 spherical = .false.
901 is_fullspace = .true.
902 CASE (do_pwgrid_ns_halfspace)
903 spherical = .false.
904 is_fullspace = .false.
905 END SELECT
906
907 ! from pw_env_methods
908 IF (spherical) THEN
909 grid_span = halfspace
910 spherical = .true.
911 odd = .true.
912 ELSE IF (is_fullspace) THEN
913 grid_span = fullspace
914 spherical = .false.
915 odd = .false.
916 ELSE
917 grid_span = halfspace
918 spherical = .false.
919 odd = .true.
920 END IF
921
922 ! actual setup
923 CALL pw_grid_create(grid, para_env, box%hmat, grid_span=grid_span, odd=odd, spherical=spherical, &
924 blocked=blocked_id, npts=np, fft_usage=.true., &
925 rs_dims=distribution_layout, iounit=iw)
926
927 IF (iw > 0) CALL m_flush(iw)
928
929 ! note that the number of grid points might be different from what the user requested (fft-able needed)
930 no = grid%npts
931
932 CALL ca%create(grid)
933 CALL cb%create(grid)
934 CALL cc%create(grid)
935
936 ! initialize data
937 CALL pw_zero(ca)
938 CALL pw_zero(cb)
939 CALL pw_zero(cc)
940 nn = SIZE(ca%array)
941 DO ig = 1, nn
942 gsq = grid%gsq(ig)
943 ca%array(ig) = exp(-gsq)
944 END DO
945
946 flops = product(no)*30.0_dp*log(real(maxval(no), kind=dp))
947 tstart = m_walltime()
948 DO ip = 1, n_loop
949 CALL para_env%sync()
950 t_start(ip) = m_walltime()
951 CALL pw_transfer(ca, cb, debug)
952 CALL pw_transfer(cb, cc, debug)
953 CALL para_env%sync()
954 t_end(ip) = m_walltime()
955 END DO
956 tend = m_walltime()
957 t = tend - tstart + threshold
958 IF (t > 0.0_dp) THEN
959 perf = real(n_loop, kind=dp)*2.0_dp*flops*1.e-6_dp/t
960 ELSE
961 perf = 0.0_dp
962 END IF
963
964 em = maxval(abs(ca%array(:) - cc%array(:)))
965 CALL para_env%max(em)
966 et = sum(abs(ca%array(:) - cc%array(:)))
967 CALL para_env%sum(et)
968 t_min = minval(t_end - t_start)
969 t_max = maxval(t_end - t_start)
970
971 IF (para_env%is_source()) THEN
972 WRITE (iw, *)
973 WRITE (iw, '(A,T67,E14.6)') " Parallel FFT Tests: Maximal Error ", em
974 WRITE (iw, '(A,T67,E14.6)') " Parallel FFT Tests: Total Error ", et
975 WRITE (iw, '(A,T67,F14.0)') &
976 " Parallel FFT Tests: Performance [Mflops] ", perf
977 WRITE (iw, '(A,T67,F14.6)') " Best time : ", t_min
978 WRITE (iw, '(A,T67,F14.6)') " Worst time: ", t_max
979 IF (iw > 0) CALL m_flush(iw)
980 END IF
981
982 ! need debugging ???
983 IF (em > toler .OR. et > toler) THEN
984 cpwarn("The FFT results are not accurate ... starting debug pw_transfer")
985 CALL pw_transfer(ca, cb, .true.)
986 CALL pw_transfer(cb, cc, .true.)
987 END IF
988
989 ! done with these grids
990 CALL ca%release()
991 CALL cb%release()
992 CALL cc%release()
993 CALL pw_grid_release(grid)
994
995 END DO
996
997 ! local arrays
998 DEALLOCATE (layouts)
999 DEALLOCATE (t_start)
1000 DEALLOCATE (t_end)
1001
1002 END DO
1003
1004 ! cleanup
1005 CALL cell_release(box)
1006 CALL finalize_fft(para_env, wisdom_file=globenv%fftw_wisdom_file_name)
1007
1008 END SUBROUTINE pw_fft_test
1009
1010! **************************************************************************************************
1011!> \brief Tests the eigensolver library routines
1012!> \param para_env ...
1013!> \param iw ...
1014!> \param eigensolver_section ...
1015!> \par History
1016!> JGH 6-Feb-2001 : Test and performance code
1017!> \author JGH 1-JAN-2001
1018! **************************************************************************************************
1019 SUBROUTINE eigensolver_test(para_env, iw, eigensolver_section)
1020
1021 TYPE(mp_para_env_type), POINTER :: para_env
1022 INTEGER :: iw
1023 TYPE(section_vals_type), POINTER :: eigensolver_section
1024
1025 INTEGER :: diag_method, i, i_loop, i_rep, &
1026 init_method, j, n, n_loop, n_rep, &
1027 neig, unit_number
1028 REAL(kind=dp) :: t1, t2
1029 REAL(kind=dp), ALLOCATABLE, DIMENSION(:) :: eigenvalues
1030 REAL(kind=dp), ALLOCATABLE, DIMENSION(:, :) :: buffer
1031 TYPE(cp_blacs_env_type), POINTER :: blacs_env
1032 TYPE(cp_fm_struct_type), POINTER :: fmstruct
1033 TYPE(cp_fm_type) :: eigenvectors, matrix, work
1034 TYPE(rng_stream_type), ALLOCATABLE :: rng_stream
1035
1036 IF (iw > 0) THEN
1037 WRITE (unit=iw, fmt="(/,/,T2,A,/)") "EIGENSOLVER TEST"
1038 END IF
1039
1040 ! create blacs env corresponding to para_env
1041 NULLIFY (blacs_env)
1042 CALL cp_blacs_env_create(blacs_env=blacs_env, &
1043 para_env=para_env)
1044
1045 ! loop over all tests
1046 CALL section_vals_get(eigensolver_section, n_repetition=n_rep)
1047 DO i_rep = 1, n_rep
1048
1049 ! parse section
1050 CALL section_vals_val_get(eigensolver_section, "N", i_rep_section=i_rep, i_val=n)
1051 CALL section_vals_val_get(eigensolver_section, "EIGENVALUES", i_rep_section=i_rep, i_val=neig)
1052 CALL section_vals_val_get(eigensolver_section, "DIAG_METHOD", i_rep_section=i_rep, i_val=diag_method)
1053 CALL section_vals_val_get(eigensolver_section, "INIT_METHOD", i_rep_section=i_rep, i_val=init_method)
1054 CALL section_vals_val_get(eigensolver_section, "N_loop", i_rep_section=i_rep, i_val=n_loop)
1055
1056 ! proper number of eigs
1057 IF (neig < 0) neig = n
1058 neig = min(neig, n)
1059
1060 ! report
1061 IF (iw > 0) THEN
1062 WRITE (iw, *) "Matrix size", n
1063 WRITE (iw, *) "Number of eigenvalues", neig
1064 WRITE (iw, *) "Timing loops", n_loop
1065 SELECT CASE (diag_method)
1066 CASE (do_diag_syevd)
1067 WRITE (iw, *) "Diag using syevd"
1068 CASE (do_diag_syevx)
1069 WRITE (iw, *) "Diag using syevx"
1070 CASE DEFAULT
1071 ! stop
1072 END SELECT
1073
1074 SELECT CASE (init_method)
1075 CASE (do_mat_random)
1076 WRITE (iw, *) "using random matrix"
1077 CASE (do_mat_read)
1078 WRITE (iw, *) "reading from file"
1079 CASE DEFAULT
1080 ! stop
1081 END SELECT
1082 END IF
1083
1084 ! create matrix struct type
1085 NULLIFY (fmstruct)
1086 CALL cp_fm_struct_create(fmstruct=fmstruct, &
1087 para_env=para_env, &
1088 context=blacs_env, &
1089 nrow_global=n, &
1090 ncol_global=n)
1091
1092 ! create all needed matrices, and buffers for the eigenvalues
1093 CALL cp_fm_create(matrix=matrix, &
1094 matrix_struct=fmstruct, &
1095 name="MATRIX")
1096 CALL cp_fm_set_all(matrix, 0.0_dp)
1097
1098 CALL cp_fm_create(matrix=eigenvectors, &
1099 matrix_struct=fmstruct, &
1100 name="EIGENVECTORS")
1101 CALL cp_fm_set_all(eigenvectors, 0.0_dp)
1102
1103 CALL cp_fm_create(matrix=work, &
1104 matrix_struct=fmstruct, &
1105 name="WORK")
1106 CALL cp_fm_set_all(matrix, 0.0_dp)
1107
1108 ALLOCATE (eigenvalues(n))
1109 eigenvalues = 0.0_dp
1110 ALLOCATE (buffer(1, n))
1111
1112 ! generate initial matrix, either by reading a file, or using random numbers
1113 IF (para_env%is_source()) THEN
1114 SELECT CASE (init_method)
1115 CASE (do_mat_random)
1116 rng_stream = rng_stream_type( &
1117 name="rng_stream", &
1118 distribution_type=uniform, &
1119 extended_precision=.true.)
1120 CASE (do_mat_read)
1121 CALL open_file(file_name="MATRIX", &
1122 file_action="READ", &
1123 file_form="FORMATTED", &
1124 file_status="OLD", &
1125 unit_number=unit_number)
1126 END SELECT
1127 END IF
1128
1129 DO i = 1, n
1130 IF (para_env%is_source()) THEN
1131 SELECT CASE (init_method)
1132 CASE (do_mat_random)
1133 DO j = i, n
1134 buffer(1, j) = rng_stream%next() - 0.5_dp
1135 END DO
1136 !MK activate/modify for a diagonal dominant symmetric matrix:
1137 !MK buffer(1,i) = 10.0_dp*buffer(1,i)
1138 CASE (do_mat_read)
1139 READ (unit=unit_number, fmt=*) buffer(1, 1:n)
1140 END SELECT
1141 END IF
1142 CALL para_env%bcast(buffer)
1143 SELECT CASE (init_method)
1144 CASE (do_mat_random)
1145 CALL cp_fm_set_submatrix(fm=matrix, &
1146 new_values=buffer, &
1147 start_row=i, &
1148 start_col=i, &
1149 n_rows=1, &
1150 n_cols=n - i + 1, &
1151 alpha=1.0_dp, &
1152 beta=0.0_dp, &
1153 transpose=.false.)
1154 CALL cp_fm_set_submatrix(fm=matrix, &
1155 new_values=buffer, &
1156 start_row=i, &
1157 start_col=i, &
1158 n_rows=n - i + 1, &
1159 n_cols=1, &
1160 alpha=1.0_dp, &
1161 beta=0.0_dp, &
1162 transpose=.true.)
1163 CASE (do_mat_read)
1164 CALL cp_fm_set_submatrix(fm=matrix, &
1165 new_values=buffer, &
1166 start_row=i, &
1167 start_col=1, &
1168 n_rows=1, &
1169 n_cols=n, &
1170 alpha=1.0_dp, &
1171 beta=0.0_dp, &
1172 transpose=.false.)
1173 END SELECT
1174 END DO
1175
1176 DEALLOCATE (buffer)
1177
1178 IF (para_env%is_source()) THEN
1179 SELECT CASE (init_method)
1180 CASE (do_mat_read)
1181 CALL close_file(unit_number=unit_number)
1182 END SELECT
1183 END IF
1184
1185 DO i_loop = 1, n_loop
1186 eigenvalues = 0.0_dp
1187 CALL cp_fm_set_all(eigenvectors, 0.0_dp)
1188 CALL cp_fm_to_fm(source=matrix, &
1189 destination=work)
1190
1191 ! DONE, now testing
1192 t1 = m_walltime()
1193 SELECT CASE (diag_method)
1194 CASE (do_diag_syevd)
1195 CALL cp_fm_syevd(matrix=work, &
1196 eigenvectors=eigenvectors, &
1197 eigenvalues=eigenvalues)
1198 CASE (do_diag_syevx)
1199 CALL cp_fm_syevx(matrix=work, &
1200 eigenvectors=eigenvectors, &
1201 eigenvalues=eigenvalues, &
1202 neig=neig, &
1203 work_syevx=1.0_dp)
1204 END SELECT
1205 t2 = m_walltime()
1206 IF (iw > 0) WRITE (iw, *) "Timing for loop ", i_loop, " : ", t2 - t1
1207 END DO
1208
1209 IF (iw > 0) THEN
1210 WRITE (iw, *) "Eigenvalues: "
1211 WRITE (unit=iw, fmt="(T3,5F14.6)") eigenvalues(1:neig)
1212 WRITE (unit=iw, fmt="(T3,A4,F16.6)") "Sum:", sum(eigenvalues(1:neig))
1213 WRITE (iw, *) ""
1214 END IF
1215
1216 ! Clean up
1217 DEALLOCATE (eigenvalues)
1218 CALL cp_fm_release(matrix=work)
1219 CALL cp_fm_release(matrix=eigenvectors)
1220 CALL cp_fm_release(matrix=matrix)
1221 CALL cp_fm_struct_release(fmstruct=fmstruct)
1222
1223 END DO
1224
1225 CALL cp_blacs_env_release(blacs_env=blacs_env)
1226
1227 END SUBROUTINE eigensolver_test
1228
1229! **************************************************************************************************
1230!> \brief Tests the parallel matrix multiply
1231!> \param para_env ...
1232!> \param iw ...
1233!> \param cp_fm_gemm_test_section ...
1234! **************************************************************************************************
1235 SUBROUTINE cp_fm_gemm_test(para_env, iw, cp_fm_gemm_test_section)
1236
1237 TYPE(mp_para_env_type), POINTER :: para_env
1238 INTEGER :: iw
1239 TYPE(section_vals_type), POINTER :: cp_fm_gemm_test_section
1240
1241 CHARACTER(LEN=1) :: transa, transb
1242 INTEGER :: i_loop, i_rep, k, m, n, n_loop, n_rep, ncol_block, ncol_block_actual, &
1243 ncol_global, np, nrow_block, nrow_block_actual, nrow_global
1244 INTEGER, DIMENSION(:), POINTER :: grid_2d
1245 LOGICAL :: force_blocksize, row_major, transa_p, &
1246 transb_p
1247 REAL(kind=dp) :: t1, t2, t3, t4
1248 TYPE(cp_blacs_env_type), POINTER :: blacs_env
1249 TYPE(cp_fm_struct_type), POINTER :: fmstruct_a, fmstruct_b, fmstruct_c
1250 TYPE(cp_fm_type) :: matrix_a, matrix_b, matrix_c
1251
1252 CALL section_vals_get(cp_fm_gemm_test_section, n_repetition=n_rep)
1253 DO i_rep = 1, n_rep
1254
1255 ! how often should we do the multiply
1256 CALL section_vals_val_get(cp_fm_gemm_test_section, "N_loop", i_rep_section=i_rep, i_val=n_loop)
1257
1258 ! matrices def.
1259 CALL section_vals_val_get(cp_fm_gemm_test_section, "K", i_rep_section=i_rep, i_val=k)
1260 CALL section_vals_val_get(cp_fm_gemm_test_section, "N", i_rep_section=i_rep, i_val=n)
1261 CALL section_vals_val_get(cp_fm_gemm_test_section, "M", i_rep_section=i_rep, i_val=m)
1262 CALL section_vals_val_get(cp_fm_gemm_test_section, "transa", i_rep_section=i_rep, l_val=transa_p)
1263 CALL section_vals_val_get(cp_fm_gemm_test_section, "transb", i_rep_section=i_rep, l_val=transb_p)
1264 CALL section_vals_val_get(cp_fm_gemm_test_section, "nrow_block", i_rep_section=i_rep, i_val=nrow_block)
1265 CALL section_vals_val_get(cp_fm_gemm_test_section, "ncol_block", i_rep_section=i_rep, i_val=ncol_block)
1266 CALL section_vals_val_get(cp_fm_gemm_test_section, "ROW_MAJOR", i_rep_section=i_rep, l_val=row_major)
1267 CALL section_vals_val_get(cp_fm_gemm_test_section, "GRID_2D", i_rep_section=i_rep, i_vals=grid_2d)
1268 CALL section_vals_val_get(cp_fm_gemm_test_section, "FORCE_BLOCKSIZE", i_rep_section=i_rep, l_val=force_blocksize)
1269 transa = "N"
1270 transb = "N"
1271 IF (transa_p) transa = "T"
1272 IF (transb_p) transb = "T"
1273
1274 IF (iw > 0) THEN
1275 WRITE (iw, '(T2,A)') "----------- TESTING PARALLEL MATRIX MULTIPLY -------------"
1276 WRITE (iw, '(T2,A)', advance="NO") "C = "
1277 IF (transa_p) THEN
1278 WRITE (iw, '(A)', advance="NO") "TRANSPOSE(A) x"
1279 ELSE
1280 WRITE (iw, '(A)', advance="NO") "A x "
1281 END IF
1282 IF (transb_p) THEN
1283 WRITE (iw, '(A)') "TRANSPOSE(B) "
1284 ELSE
1285 WRITE (iw, '(A)') "B "
1286 END IF
1287 WRITE (iw, '(T2,A,T50,I5,A,I5)') 'requested block size', nrow_block, ' by ', ncol_block
1288 WRITE (iw, '(T2,A,T50,I5)') 'number of repetitions of cp_fm_gemm ', n_loop
1289 WRITE (iw, '(T2,A,T50,L5)') 'Row Major', row_major
1290 WRITE (iw, '(T2,A,T50,2I7)') 'GRID_2D ', grid_2d
1291 WRITE (iw, '(T2,A,T50,L5)') 'Force blocksize ', force_blocksize
1292 ! check the return value of pilaenv, too small values limit the performance (assuming pdgemm is the vanilla variant)
1293 np = cp_fm_pilaenv(0, 'D')
1294 IF (np > 0) THEN
1295 WRITE (iw, '(T2,A,T50,I5)') 'PILAENV blocksize', np
1296 END IF
1297 END IF
1298
1299 NULLIFY (blacs_env)
1300 CALL cp_blacs_env_create(blacs_env=blacs_env, &
1301 para_env=para_env, &
1302 row_major=row_major, &
1303 grid_2d=grid_2d)
1304
1305 NULLIFY (fmstruct_a)
1306 IF (transa_p) THEN
1307 nrow_global = m; ncol_global = k
1308 ELSE
1309 nrow_global = k; ncol_global = m
1310 END IF
1311 CALL cp_fm_struct_create(fmstruct=fmstruct_a, para_env=para_env, context=blacs_env, &
1312 nrow_global=nrow_global, ncol_global=ncol_global, &
1313 nrow_block=nrow_block, ncol_block=ncol_block, force_block=force_blocksize)
1314 CALL cp_fm_struct_get(fmstruct_a, nrow_block=nrow_block_actual, ncol_block=ncol_block_actual)
1315 IF (iw > 0) WRITE (iw, '(T2,A,I9,A,I9,A,I5,A,I5)') 'matrix A ', nrow_global, " by ", ncol_global, &
1316 ' using blocks of ', nrow_block_actual, ' by ', ncol_block_actual
1317
1318 IF (transb_p) THEN
1319 nrow_global = n; ncol_global = m
1320 ELSE
1321 nrow_global = m; ncol_global = n
1322 END IF
1323 NULLIFY (fmstruct_b)
1324 CALL cp_fm_struct_create(fmstruct=fmstruct_b, para_env=para_env, context=blacs_env, &
1325 nrow_global=nrow_global, ncol_global=ncol_global, &
1326 nrow_block=nrow_block, ncol_block=ncol_block, force_block=force_blocksize)
1327 CALL cp_fm_struct_get(fmstruct_b, nrow_block=nrow_block_actual, ncol_block=ncol_block_actual)
1328 IF (iw > 0) WRITE (iw, '(T2,A,I9,A,I9,A,I5,A,I5)') 'matrix B ', nrow_global, " by ", ncol_global, &
1329 ' using blocks of ', nrow_block_actual, ' by ', ncol_block_actual
1330
1331 NULLIFY (fmstruct_c)
1332 nrow_global = k
1333 ncol_global = n
1334 CALL cp_fm_struct_create(fmstruct=fmstruct_c, para_env=para_env, context=blacs_env, &
1335 nrow_global=nrow_global, ncol_global=ncol_global, &
1336 nrow_block=nrow_block, ncol_block=ncol_block, force_block=force_blocksize)
1337 CALL cp_fm_struct_get(fmstruct_c, nrow_block=nrow_block_actual, ncol_block=ncol_block_actual)
1338 IF (iw > 0) WRITE (iw, '(T2,A,I9,A,I9,A,I5,A,I5)') 'matrix C ', nrow_global, " by ", ncol_global, &
1339 ' using blocks of ', nrow_block_actual, ' by ', ncol_block_actual
1340
1341 CALL cp_fm_create(matrix=matrix_a, matrix_struct=fmstruct_a, name="MATRIX A")
1342 CALL cp_fm_create(matrix=matrix_b, matrix_struct=fmstruct_b, name="MATRIX B")
1343 CALL cp_fm_create(matrix=matrix_c, matrix_struct=fmstruct_c, name="MATRIX C")
1344
1345 CALL random_number(matrix_a%local_data)
1346 CALL random_number(matrix_b%local_data)
1347 CALL random_number(matrix_c%local_data)
1348
1349 IF (iw > 0) CALL m_flush(iw)
1350
1351 t1 = m_walltime()
1352 DO i_loop = 1, n_loop
1353 t3 = m_walltime()
1354 CALL parallel_gemm(transa, transb, k, n, m, 1.0_dp, matrix_a, matrix_b, 0.0_dp, matrix_c)
1355 t4 = m_walltime()
1356 IF (iw > 0) THEN
1357 WRITE (iw, '(T2,A,T50,F12.6)') "cp_fm_gemm timing: ", (t4 - t3)
1358 CALL m_flush(iw)
1359 END IF
1360 END DO
1361 t2 = m_walltime()
1362
1363 IF (iw > 0) THEN
1364 WRITE (iw, '(T2,A,T50,F12.6)') "average cp_fm_gemm timing: ", (t2 - t1)/n_loop
1365 IF (t2 > t1) THEN
1366 WRITE (iw, '(T2,A,T50,F12.6)') "cp_fm_gemm Gflops per MPI task: ", &
1367 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
1368 END IF
1369 END IF
1370
1371 CALL cp_fm_release(matrix=matrix_a)
1372 CALL cp_fm_release(matrix=matrix_b)
1373 CALL cp_fm_release(matrix=matrix_c)
1374 CALL cp_fm_struct_release(fmstruct=fmstruct_a)
1375 CALL cp_fm_struct_release(fmstruct=fmstruct_b)
1376 CALL cp_fm_struct_release(fmstruct=fmstruct_c)
1377 CALL cp_blacs_env_release(blacs_env=blacs_env)
1378
1379 END DO
1380
1381 END SUBROUTINE cp_fm_gemm_test
1382
1383! **************************************************************************************************
1384!> \brief Tests the DBCSR interface.
1385!> \param para_env ...
1386!> \param iw ...
1387!> \param input_section ...
1388! **************************************************************************************************
1389 SUBROUTINE cp_dbcsr_tests(para_env, iw, input_section)
1390
1391 TYPE(mp_para_env_type), POINTER :: para_env
1392 INTEGER :: iw
1393 TYPE(section_vals_type), POINTER :: input_section
1394
1395 CHARACTER, DIMENSION(3) :: types
1396 INTEGER :: data_type, i_rep, k, m, n, n_loop, &
1397 n_rep, test_type
1398 INTEGER, DIMENSION(:), POINTER :: bs_k, bs_m, bs_n, nproc
1399 LOGICAL :: always_checksum, retain_sparsity, &
1400 transa_p, transb_p
1401 REAL(kind=dp) :: alpha, beta, filter_eps, s_a, s_b, s_c
1402
1403! ---------------------------------------------------------------------------
1404
1405 NULLIFY (bs_m, bs_n, bs_k)
1406 CALL section_vals_get(input_section, n_repetition=n_rep)
1407 CALL dbcsr_reset_randmat_seed()
1408 DO i_rep = 1, n_rep
1409 ! how often should we do the multiply
1410 CALL section_vals_val_get(input_section, "N_loop", i_rep_section=i_rep, i_val=n_loop)
1411
1412 ! matrices def.
1413 CALL section_vals_val_get(input_section, "TEST_TYPE", i_rep_section=i_rep, i_val=test_type)
1414 CALL section_vals_val_get(input_section, "DATA_TYPE", i_rep_section=i_rep, i_val=data_type)
1415 CALL section_vals_val_get(input_section, "K", i_rep_section=i_rep, i_val=k)
1416 CALL section_vals_val_get(input_section, "N", i_rep_section=i_rep, i_val=n)
1417 CALL section_vals_val_get(input_section, "M", i_rep_section=i_rep, i_val=m)
1418 CALL section_vals_val_get(input_section, "transa", i_rep_section=i_rep, l_val=transa_p)
1419 CALL section_vals_val_get(input_section, "transb", i_rep_section=i_rep, l_val=transb_p)
1420 CALL section_vals_val_get(input_section, "bs_m", i_rep_section=i_rep, &
1421 i_vals=bs_m)
1422 CALL section_vals_val_get(input_section, "bs_n", i_rep_section=i_rep, &
1423 i_vals=bs_n)
1424 CALL section_vals_val_get(input_section, "bs_k", i_rep_section=i_rep, &
1425 i_vals=bs_k)
1426 CALL section_vals_val_get(input_section, "keepsparse", i_rep_section=i_rep, l_val=retain_sparsity)
1427 CALL section_vals_val_get(input_section, "asparsity", i_rep_section=i_rep, r_val=s_a)
1428 CALL section_vals_val_get(input_section, "bsparsity", i_rep_section=i_rep, r_val=s_b)
1429 CALL section_vals_val_get(input_section, "csparsity", i_rep_section=i_rep, r_val=s_c)
1430 CALL section_vals_val_get(input_section, "alpha", i_rep_section=i_rep, r_val=alpha)
1431 CALL section_vals_val_get(input_section, "beta", i_rep_section=i_rep, r_val=beta)
1432 CALL section_vals_val_get(input_section, "nproc", i_rep_section=i_rep, &
1433 i_vals=nproc)
1434 CALL section_vals_val_get(input_section, "atype", i_rep_section=i_rep, &
1435 c_val=types(1))
1436 CALL section_vals_val_get(input_section, "btype", i_rep_section=i_rep, &
1437 c_val=types(2))
1438 CALL section_vals_val_get(input_section, "ctype", i_rep_section=i_rep, &
1439 c_val=types(3))
1440 CALL section_vals_val_get(input_section, "filter_eps", &
1441 i_rep_section=i_rep, r_val=filter_eps)
1442 CALL section_vals_val_get(input_section, "ALWAYS_CHECKSUM", i_rep_section=i_rep, l_val=always_checksum)
1443
1444 CALL dbcsr_run_tests(para_env%get_handle(), iw, nproc, &
1445 [m, n, k], &
1446 [transa_p, transb_p], &
1447 bs_m, bs_n, bs_k, &
1448 [s_a, s_b, s_c], &
1449 alpha, beta, &
1450 data_type=data_type, &
1451 test_type=test_type, &
1452 n_loops=n_loop, eps=filter_eps, retain_sparsity=retain_sparsity, &
1453 always_checksum=always_checksum)
1454 END DO
1455 END SUBROUTINE cp_dbcsr_tests
1456
1457! **************************************************************************************************
1458!> \brief Tests the DBM library.
1459!> \param para_env ...
1460!> \param iw ...
1461!> \param input_section ...
1462! **************************************************************************************************
1463 SUBROUTINE run_dbm_tests(para_env, iw, input_section)
1464
1465 TYPE(mp_para_env_type), POINTER :: para_env
1466 INTEGER :: iw
1467 TYPE(section_vals_type), POINTER :: input_section
1468
1469 INTEGER :: i_rep, k, m, n, n_loop, n_rep
1470 INTEGER, DIMENSION(:), POINTER :: bs_k, bs_m, bs_n
1471 LOGICAL :: always_checksum, retain_sparsity, &
1472 transa_p, transb_p
1473 REAL(kind=dp) :: alpha, beta, filter_eps, s_a, s_b, s_c
1474
1475! ---------------------------------------------------------------------------
1476
1477 NULLIFY (bs_m, bs_n, bs_k)
1478 CALL section_vals_get(input_section, n_repetition=n_rep)
1479 CALL dbcsr_reset_randmat_seed()
1480 DO i_rep = 1, n_rep
1481 CALL section_vals_val_get(input_section, "N_loop", i_rep_section=i_rep, i_val=n_loop)
1482 CALL section_vals_val_get(input_section, "K", i_rep_section=i_rep, i_val=k)
1483 CALL section_vals_val_get(input_section, "N", i_rep_section=i_rep, i_val=n)
1484 CALL section_vals_val_get(input_section, "M", i_rep_section=i_rep, i_val=m)
1485 CALL section_vals_val_get(input_section, "transa", i_rep_section=i_rep, l_val=transa_p)
1486 CALL section_vals_val_get(input_section, "transb", i_rep_section=i_rep, l_val=transb_p)
1487 CALL section_vals_val_get(input_section, "bs_m", i_rep_section=i_rep, i_vals=bs_m)
1488 CALL section_vals_val_get(input_section, "bs_n", i_rep_section=i_rep, i_vals=bs_n)
1489 CALL section_vals_val_get(input_section, "bs_k", i_rep_section=i_rep, i_vals=bs_k)
1490 CALL section_vals_val_get(input_section, "keepsparse", i_rep_section=i_rep, l_val=retain_sparsity)
1491 CALL section_vals_val_get(input_section, "asparsity", i_rep_section=i_rep, r_val=s_a)
1492 CALL section_vals_val_get(input_section, "bsparsity", i_rep_section=i_rep, r_val=s_b)
1493 CALL section_vals_val_get(input_section, "csparsity", i_rep_section=i_rep, r_val=s_c)
1494 CALL section_vals_val_get(input_section, "alpha", i_rep_section=i_rep, r_val=alpha)
1495 CALL section_vals_val_get(input_section, "beta", i_rep_section=i_rep, r_val=beta)
1496 CALL section_vals_val_get(input_section, "filter_eps", i_rep_section=i_rep, r_val=filter_eps)
1497 CALL section_vals_val_get(input_section, "ALWAYS_CHECKSUM", i_rep_section=i_rep, l_val=always_checksum)
1498
1499 CALL dbm_run_tests(mp_group=para_env, &
1500 io_unit=iw, &
1501 matrix_sizes=[m, n, k], &
1502 trs=[transa_p, transb_p], &
1503 bs_m=bs_m, &
1504 bs_n=bs_n, &
1505 bs_k=bs_k, &
1506 sparsities=[s_a, s_b, s_c], &
1507 alpha=alpha, &
1508 beta=beta, &
1509 n_loops=n_loop, &
1510 eps=filter_eps, &
1511 retain_sparsity=retain_sparsity, &
1512 always_checksum=always_checksum)
1513 END DO
1514 END SUBROUTINE run_dbm_tests
1515
1516END 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:166
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:184
cell_methods::cell_create
subroutine, public cell_create(cell, hmat, periodic, tag)
allocates and initializes a cell
Definition cell_methods.F:94
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:608
cg_test
Test of Clebsch-Gordon Coefficients.
Definition cg_test.F:14
cg_test::clebsch_gordon_test
subroutine, public clebsch_gordon_test()
...
Definition cg_test.F:42
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. Tracking of preconnections.
Definition cp_files.F:16
cp_files::open_file
subroutine, public open_file(file_name, file_status, file_form, file_action, file_position, file_pad, unit_number, debug, skip_get_unit_number, file_access)
Opens the requested file using a free unit number.
Definition cp_files.F:311
cp_files::close_file
subroutine, public close_file(unit_number, file_status, keep_preconnection)
Close an open file given by its logical unit number. Optionally, keep the file and unit preconnected.
Definition cp_files.F:122
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:475
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:573
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:797
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:750
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:2441
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:239
fft_tools::bwfft
integer, parameter, public bwfft
Definition fft_tools.F:146
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, fftsg_sizes, pool_limit, wisdom_file, plan_style)
...
Definition fft_tools.F:186
fft_tools::fwfft
integer, parameter, public fwfft
Definition fft_tools.F:146
fft_tools::finalize_fft
subroutine, public finalize_fft(para_env, wisdom_file)
does whatever is needed to finalize the current fft setup
Definition fft_tools.F:216
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:75
input_constants::do_mat_read
integer, parameter, public do_mat_read
Definition input_constants.F:78
input_constants::do_pwgrid_ns_fullspace
integer, parameter, public do_pwgrid_ns_fullspace
Definition input_constants.F:243
input_constants::do_diag_syevx
integer, parameter, public do_diag_syevx
Definition input_constants.F:75
input_constants::do_pwgrid_spherical
integer, parameter, public do_pwgrid_spherical
Definition input_constants.F:243
input_constants::do_mat_random
integer, parameter, public do_mat_random
Definition input_constants.F:78
input_constants::do_pwgrid_ns_halfspace
integer, parameter, public do_pwgrid_ns_halfspace
Definition input_constants.F:243
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:124
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:141
mathconstants
Definition of mathematical constants and functions.
Definition mathconstants.F:16
mathconstants::gaussi
complex(kind=dp), parameter, public gaussi
Definition mathconstants.F:142
message_passing
Interface to the message passing library MPI.
Definition message_passing.F:23
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
mp2_grids
Routines to calculate frequency and time grids (integration points and weights) for correlation metho...
Definition mp2_grids.F:14
mp2_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 mp2_grids.F:1149
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:2163
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:2034
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:2014
realspace_grid_types::rs_grid_zero
subroutine, public rs_grid_zero(rs)
Initialize grid to zero.
Definition realspace_grid_types.F:1907
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
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:746
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