[CP2K:8418] DFT-D3 Timing

hut... at chem.uzh.ch hut... at chem.uzh.ch
Sat Nov 19 15:21:35 CET 2016


Hi

although the implementation of the C9 term is linear scaling,
it has a very large prefactor, especially for the force term.
That is why we implemented the possibility to keep the 
coordination numbers for C9 fixed:

REFERENCE_C9_TERM = .TRUE.

This already reduces the computational time by order(s) of magnitude.

I don't know why you have such a bad load balancing in your calculation
(that is the reason of your bad scaling).
I haven't looked into this for a long time and without more information
on your system it is impossible to guess.

regards

Juerg
--------------------------------------------------------------
Juerg Hutter                         Phone : ++41 44 635 4491
Institut für Chemie C                FAX   : ++41 44 635 6838
Universität Zürich                   E-mail: hut... at chem.uzh.ch
Winterthurerstrasse 190
CH-8057 Zürich, Switzerland
---------------------------------------------------------------

-----cp... at googlegroups.com wrote: -----To: cp2k <cp... at googlegroups.com>
From: Mike Ruggiero 
Sent by: cp... at googlegroups.com
Date: 11/18/2016 09:22PM
Subject: [CP2K:8418] DFT-D3 Timing

Hello everyone,
I seem to have stumbled on something. I noticed some funny scaling (or rather, a lack of scaling) when performing AIMD simulations with DFT-D3. I've tracked the problem to the calculation of the C9 term, when I have that set to be calculated the simulation doesn't scale at all and it takes a long time versus when I have it set to not be calculated, in which case I do get linear scaling. Below is a timing report for the same exact INPUT file, with the only difference being the calculation of the C9 term being set to T and F, respectively. Something interesting is how different they are. Again, these are from the exact same INPUT file, with the exact same number of processors, and the only difference is the CALCULATE_C9_TERM  is set to T in the first case and F in the second. 
timing with C9 calculation:
 ------------------------------------------------------------------------------- -                                                                             - -                                T I M I N G                                  - -                                                                             - ------------------------------------------------------------------------------- SUBROUTINE                       CALLS  ASD         SELF TIME        TOTAL TIME                                MAXIMUM       AVERAGE  MAXIMUM  AVERAGE  MAXIMUM CP2K                                 1  1.0    0.008    0.019  772.815  772.821 qs_mol_dyn_low                       1  2.0    0.002    0.003  772.503  772.517 qs_forces                           11  3.9    0.002    0.002  771.252  771.258 qs_energies                         11  4.9    0.001    0.001  765.300  765.307 velocity_verlet                     10  3.0    0.002    0.002  686.306  686.323 qs_energies_init_hamiltonians       11  5.9    0.000    0.000  686.298  686.303 mp_sum_d                          4362 11.6  619.789  685.783  619.789  685.783 calculate_dispersion_pairpot        11  6.9   65.911  685.372  685.486  685.491 scf_env_do_scf                      11  5.9    0.000    0.001   76.654   76.655 scf_env_do_scf_inner_loop          139  6.7    0.004    0.030   71.107   71.468 rebuild_ks_matrix                  150  8.5    0.000    0.001   52.308   52.341 qs_ks_build_kohn_sham_matrix       150  9.5    0.017    0.021   52.307   52.340 qs_ks_update_qs_env                151  7.7    0.001    0.001   47.661   47.693 pw_transfer                       4061 12.2    0.239    0.267   35.026   35.154 fft_wrap_pw1pw2                   3761 13.2    0.028    0.031   34.447   34.575 fft_wrap_pw1pw2_350               1661 14.3    2.499    2.594   32.640   32.789 qs_vxc_create                      150 10.5    0.003    0.004   30.531   30.553 xc_vxc_pw_create                   150 11.5    1.027    1.065   30.528   30.551 fft3d_ps                          3761 15.2   12.507   12.766   28.236   28.478 qs_rho_update_rho                  150  7.8    0.001    0.001   18.349   18.356 calculate_rho_elec                 150  8.8    3.941    4.247   18.348   18.355 xc_rho_set_and_dset_create         150 12.5    0.188    0.193   16.380   16.402 sum_up_and_integrate               150 10.5    0.079    0.082   15.635   15.650 integrate_v_rspace                 150 11.5    4.076    4.275   15.556   15.572 -------------------------------------------------------------------------------
Without c9 calculation 
 ------------------------------------------------------------------------------- -                                                                             - -                                T I M I N G                                  - -                                                                             - ------------------------------------------------------------------------------- SUBROUTINE                       CALLS  ASD         SELF TIME        TOTAL TIME                                MAXIMUM       AVERAGE  MAXIMUM  AVERAGE  MAXIMUM CP2K                                 1  1.0    0.010    0.023   98.369   98.371 qs_mol_dyn_low                       1  2.0    0.002    0.003   98.009   98.019 qs_forces                           11  3.9    0.002    0.002   97.695   97.697 qs_energies                         11  4.9    0.001    0.001   91.392   91.395 scf_env_do_scf                      11  5.9    0.000    0.001   84.901   84.903 scf_env_do_scf_inner_loop          139  6.7    0.004    0.028   78.845   79.075 velocity_verlet                     10  3.0    0.002    0.002   67.867   67.871 rebuild_ks_matrix                  150  8.5    0.000    0.001   55.485   55.544 qs_ks_build_kohn_sham_matrix       150  9.5    0.017    0.020   55.484   55.543 qs_ks_update_qs_env                151  7.7    0.001    0.001   50.557   50.612 pw_transfer                       3461 12.2    0.206    0.259   37.722   37.838 fft_wrap_pw1pw2                   3161 13.2    0.027    0.033   37.180   37.297 fft_wrap_pw1pw2_350               1661 14.3    2.439    2.559   34.568   34.695 qs_vxc_create                      150 10.5    0.003    0.004   31.774   31.801 xc_vxc_pw_create                   150 11.5    1.015    1.048   31.771   31.798 fft3d_ps                          3161 15.2   12.404   12.688   31.074   31.311 qs_rho_update_rho                  150  7.8    0.001    0.001   20.139   20.150 calculate_rho_elec                 150  8.8    3.909    4.230   20.138   20.149 sum_up_and_integrate               150 10.5    0.077    0.081   17.172   17.191 integrate_v_rspace                 150 11.5    4.070    4.271   17.094   17.111 xc_rho_set_and_dset_create         150 12.5    0.185    0.190   16.948   16.973 rs_pw_transfer                    1822 12.1    0.020    0.024   15.129   15.437 mp_alltoall_z22v                  3161 17.2   15.000   15.294   15.000   15.294 density_rs2pw                      150  9.8    0.007    0.009   14.718   14.977 qs_scf_new_mos                     139  7.7    0.001    0.001   14.120   14.173 qs_scf_loop_do_ot                  139  8.7    0.001    0.001   14.120   14.173 dbcsr_multiply_generic            2770 12.6    0.073    0.082   13.810   13.940 ot_scf_mini                        139  9.7    0.005    0.006   13.225   13.279 potential_pw2rs                    150 12.5    0.010    0.013   12.324   12.362 x_to_yz                           1050 16.5    1.546    1.618    9.447    9.550 multiply_cannon                   2770 13.6    0.293    0.307    8.197    8.790 yz_to_x                            911 16.0    1.073    1.104    8.140    8.355 mp_waitall_1                    427928 15.4    7.704    8.344    7.704    8.344 ot_mini                            139 10.7    0.001    0.001    8.062    8.132 qs_ot_get_derivative               139 11.7    0.002    0.002    6.714    6.768 xc_functional_eval                 150 13.5    0.002    0.004    6.023    6.043 pbe_lda_eval                       150 14.5    6.021    6.041    6.021    6.041 init_scf_loop                       12  6.8    0.000    0.000    5.780    5.780 qs_ks_update_qs_env_forces          11  4.9    0.000    0.000    4.981    4.985 rs_pw_transfer_PW2RS_350           161 14.2    2.271    2.444    4.256    4.435 mp_waitany                       18660 14.0    3.701    4.433    3.701    4.433 make_m2s                          5540 13.6    0.087    0.097    4.247    4.358 rs_pw_transfer_RS2PW_350           161 11.6    2.137    2.399    3.706    3.947 make_images                       5540 14.6    0.322    0.341    3.720    3.811 qs_ot_get_derivative_taylor        110 12.9    0.003    0.004    3.646    3.691 qs_energies_init_hamiltonians       11  5.9    0.000    0.000    3.259    3.259 multiply_cannon_multrec          33240 14.6    2.695    3.049    2.705    3.060 qs_ot_get_p                        151 10.5    0.001    0.001    2.871    2.945 init_scf_run                        11  5.9    0.000    0.001    2.925    2.925 scf_env_initial_rho_setup           11  6.9    0.000    0.000    2.925    2.925 multiply_cannon_metrocomm1       33240 14.6    0.049    0.053    2.285    2.914 multiply_cannon_metrocomm3       33240 14.6    0.040    0.045    0.735    2.682 mp_sum_d                          4362 11.6    2.357    2.664    2.357    2.664 wfi_extrapolate                     11  7.9    0.000    0.001    2.650    2.650 calculate_dispersion_pairpot        11  6.9    0.219    2.076    2.327    2.327 make_images_sizes                 5540 15.6    0.006    0.007    1.394    1.992 mp_alltoall_i44                   5540 16.6    1.388    1.986    1.388    1.986 -------------------------------------------------------------------------------






-- 

You received this message because you are subscribed to the Google Groups "cp2k" group.

To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns... at googlegroups.com.

To post to this group, send email to cp... at googlegroups.com.

Visit this group at https://groups.google.com/group/cp2k.

For more options, visit https://groups.google.com/d/optout.





More information about the CP2K-user mailing list