[CP2K:5796] CP2K running slow
Teodoro Laino
teodor... at gmail.com
Thu Oct 30 13:28:07 UTC 2014
I cannot see the point..
-) First you are not running a minimization but simply an MD (at least this is what I see from your input and timings)
-) You are running on 8 mpi task.
-) Your system contains 150 QM atoms
-) Your QM cell is 20 Ang cubic
-) The reported timing is for 10 MD steps.
For 10 MD steps you get 9476 seconds overall. A good part of it (approximately > 3000 secs) are for the first step (difficult to be sure without a more extended output).
This means that at regime you can do 1 MD step in 550 seconds.
What is the problem? Does it look too slow?
Why don’t you increase # MPI task? Your system definitely is not made for a workstation. You should run that on more than just 8 tasks.
Criticism:
-) a cutoff of 250 with BLYP… very much on the very low side..
-) This setup:
SCF_GUESS ATOMIC
MAX_SCF 300
EPS_SCF 1.0E-06
&OUTER_SCF
EPS_SCF 1.0E-06
MAX_SCF 20
&END OUTER_SCF
The 300 in the MAX_SCF is difficult to understand. The idea is that you do a certain number of steps (few tens) and then you reset with a new PC.
Try to decrease 300 to something more meaningful, like 30. You’ll also definitely notice an improvement in terms of time to solution.
Other things, like the setup of the PERIODIC QMMM section will affect performance, but they will pop up in the timing only when you arrive at the scaling limit. 8 is very far from being the scaling limit for this job.
Final comment:
-) if you are a beginner I would try with something more simple.. You combined, QS, QMMM and the periodic version of it.. Even for experts all this s**t requires some careful tuning, before going in production.
Regards,
Teo
> On 30 Oct 2014, at 06:30, Tanmoy Paul <tanmo... at gmail.com> wrote:
>
> Dear CP2K users,
> I checked the output files and to be frank I have understood a very little about it . Last few lines of the output file says...
>
> -------------------------------------------------------------------------------
> ---- MULTIGRID INFO ----
> -------------------------------------------------------------------------------
> count for grid 1: 317951 cutoff [a.u.] 125.00
> count for grid 2: 302978 cutoff [a.u.] 31.25
> count for grid 3: 76551 cutoff [a.u.] 7.81
> count for grid 4: 684 cutoff [a.u.] 1.95
> total gridlevel count : 698164
>
> -------------------------------------------------------------------------------
> - -
> - MESSAGE PASSING PERFORMANCE -
> - -
> -------------------------------------------------------------------------------
>
> ROUTINE CALLS TOT TIME [s] AVE VOLUME [Bytes] PERFORMANCE [MB/s]
> MP_Group 6 0.000
> MP_Bcast 1834 0.213 5713507. 49195.17
> MP_Allreduce 10132 130.663 142451. 11.05
> MP_Sync 101392 0.269
> MP_Alltoall 9590 652.514 46235404. 679.52
> MP_SendRecv 15776 87.207 294912. 53.35
> MP_ISendRecv 23730 0.566 4071645. 170706.96
> MP_Wait 24093 402.879
> MP_ISend 275 0.038 728596. 5272.74
> MP_IRecv 264 0.001 746667. 197120.00
> MP_Recv 76412 0.793 3234. 311.65
> -------------------------------------------------------------------------------
>
>
> -------------------------------------------------------------------------------
> - -
> - T I M I N G -
> - -
> -------------------------------------------------------------------------------
> SUBROUTINE CALLS ASD SELF TIME TOTAL TIME
> AVERAGE MAXIMUM AVERAGE MAXIMUM
> CP2K 1 1.0 1.346 1.804 9476.129 9476.130
> qs_mol_dyn_low 1 2.0 0.095 0.314 9465.241 9465.854
> qs_forces 11 3.9 0.027 0.040 8706.050 8706.055
> qs_energies_scf 11 4.9 0.001 0.002 8569.698 8569.705
> scf_env_do_scf 11 5.9 0.004 0.004 8453.922 8453.929
> scf_env_do_scf_inner_loop 639 6.5 0.218 0.273 8289.159 8289.172
> velocity_verlet 10 3.0 0.894 3.202 5147.091 5147.116
> qs_ks_build_kohn_sham_matrix 650 8.5 12.649 13.425 4123.288 4123.607
> qs_rho_update_rho 650 7.5 0.026 0.031 4077.287 4077.571
> calculate_rho_elec 1300 8.5 592.558 649.730 4077.261 4077.552
> qs_ks_update_qs_env 651 7.5 0.014 0.027 3995.187 3995.463
> fft_wrap_pw1pw2 9590 11.0 0.464 0.476 3630.990 3654.319
> fft_wrap_pw1pw2_130 3630 11.8 366.371 376.203 3397.033 3418.254
> density_rs2pw 1300 9.5 0.135 0.159 3331.163 3383.539
> fft3d_ps 9590 13.0 1685.880 1712.933 2797.864 2836.568
> qs_vxc_create 650 9.5 0.047 0.054 1589.974 1590.199
> sum_up_and_integrate 336 9.4 46.390 55.122 1355.401 1368.971
> integrate_v_rspace 672 10.4 301.274 329.909 1309.006 1315.670
> xc_rho_set_and_dset_create 650 11.5 202.175 205.516 1171.640 1217.797
> potential_pw2rs 672 11.4 0.450 0.575 1007.402 1032.221
> xc_vxc_pw_create 336 10.4 118.323 123.799 1030.731 1030.957
> rs_pw_transfer 7954 11.1 0.263 0.272 886.089 941.457
> qs_ks_ddapc 650 9.5 0.014 0.016 682.186 683.086
> rs_pw_transfer_RS2PW_130 1311 11.5 529.773 582.666 529.773 582.666
> x_to_yz 3696 14.5 561.963 566.914 561.963 566.914
> yz_to_x 5894 13.7 549.928 566.657 549.928 566.657
> xc_exc_calc 314 10.5 30.516 71.193 559.196 559.200
> xc_functional_eval 1300 12.5 0.114 0.123 501.124 543.504
> pw_nn_compose_r 7920 12.1 443.318 502.878 443.318 502.878
> cp_dbcsr_multiply_d 18824 11.3 0.056 0.060 440.552 455.131
> dbcsr_multiply_anytype 18824 13.3 6.863 8.968 438.921 453.471
> qmmm_forces 11 3.9 5.057 7.224 418.833 418.835
> qmmm_forces_with_gaussian 11 4.9 8.177 8.259 402.521 404.942
> qs_scf_loop_do_ot 639 7.5 0.006 0.010 397.926 398.436
> cp_ddapc_apply_CD 650 10.5 65.894 68.079 357.733 357.740
> ot_scf_mini 639 8.5 0.129 0.142 357.246 357.589
> qmmm_force_with_gaussian_low 11 5.9 0.000 0.000 345.037 345.171
> lyp_lsd_eval 650 13.5 314.342 343.915 314.342 343.915
> qmmm_forces_with_gaussian_LG 11 7.9 340.844 340.922 340.844 340.922
> qmmm_forces_gaussian_low_R 11 6.9 0.000 0.000 340.844 340.922
> dbcsr_mm_cannon_multiply 18824 14.3 10.537 23.151 264.680 278.647
> qmmm_el_coupling 11 3.9 0.005 0.007 272.055 272.925
> rs_pw_transfer_PW2RS_130 683 13.3 263.671 264.213 263.671 264.213
> qmmm_elec_with_gaussian 11 4.9 10.833 11.870 259.977 262.493
> pw_scatter_p 3696 13.5 250.884 255.081 250.884 255.081
> pw_axpy 11933 9.6 219.576 219.933 219.576 219.933
> fft_wrap_pw1pw2_40 1972 12.2 15.362 15.579 211.444 216.042
> cp_dbcsr_mult_NS_NR 2831 11.3 0.015 0.017 202.511 214.845
> pw_zero 12765 10.0 211.188 214.613 211.188 214.613
> qmmm_elec_with_gaussian_low 11 5.9 0.000 0.000 203.561 204.576
> pw_copy 5587 11.3 197.645 200.647 197.645 200.647
> qmmm_elec_with_gaussian_LG 11 7.9 199.115 200.356 199.115 200.356
> qmmm_elec_gaussian_low_R 11 6.9 0.000 0.000 199.115 200.356
> pw_poisson_solve 661 9.4 115.353 117.174 200.327 200.345
> pw_gather_p 5894 12.7 197.867 200.180 197.867 200.180
> xb88_lsd_eval 650 13.5 186.668 199.483 186.668 199.483
> ot_mini 639 9.5 0.017 0.022 198.207 198.562
> -------------------------------------------------------------------------------
>
> **** **** ****** ** PROGRAM ENDED AT 2014-10-30 00:27:35.638
> ***** ** *** *** ** PROGRAM RAN ON user
> ** **** ****** PROGRAM RAN BY user
> ***** ** ** ** ** PROGRAM PROCESS ID 16608
> **** ** ******* ** PROGRAM STOPPED IN /home/user/cp2k_run/cp2k_input/for_co
> 2_HCA
>
> Any idea ? And one more added question to it........... Where can I find the file formats of the cp2k output files?
> Regards
> Tanmoy
>
>
> On Wed, Oct 29, 2014 at 3:37 PM, Tanmoy Paul <tanmo... at gmail.com <mailto:tanmo... at gmail.com>> wrote:
> Thanks Samuel and Juerg for your inputs .
>
> On Wed, Oct 29, 2014 at 3:32 PM, <hut... at chem.uzh.ch <mailto:hut... at chem.uzh.ch>> wrote:
> Hi
>
> yes, check the timing information at the end of a short run.
> This will give some information on possible problems.
>
> You should also change the EWALD method from PME to SPME.
>
> 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 <mailto:hut... at chem.uzh.ch>
> Winterthurerstrasse 190
> CH-8057 Zürich, Switzerland
> ---------------------------------------------------------------
>
> -----cp... at googlegroups.com <mailto:cp... at googlegroups.com> wrote: -----
> To: cp... at googlegroups.com <mailto:cp... at googlegroups.com>
> From: Samuel Andermatt
> Sent by: cp... at googlegroups.com <mailto:cp... at googlegroups.com>
> Date: 10/29/2014 10:45AM
> Subject: [CP2K:5788] Re: CP2K running slow
>
> Sadly I do not know what exactly the reason for the performance issue is. An additional method to figure out the problem would be to make a short run and at the end of the output you will see the timings. It will give some information about the performance bottlenecks.
>
> --
> 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 <mailto:cp2k%2Bun... at googlegroups.com>.
> To post to this group, send email to cp... at googlegroups.com <mailto:cp... at googlegroups.com>.
> Visit this group at http://groups.google.com/group/cp2k <http://groups.google.com/group/cp2k>.
> For more options, visit https://groups.google.com/d/optout <https://groups.google.com/d/optout>.
>
>
> --
> 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 <mailto:cp2k%2Bun... at googlegroups.com>.
> To post to this group, send email to cp... at googlegroups.com <mailto:cp... at googlegroups.com>.
> Visit this group at http://groups.google.com/group/cp2k <http://groups.google.com/group/cp2k>.
> For more options, visit https://groups.google.com/d/optout <https://groups.google.com/d/optout>.
>
>
>
> --
> 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 <mailto:cp2k+uns... at googlegroups.com>.
> To post to this group, send email to cp... at googlegroups.com <mailto:cp... at googlegroups.com>.
> Visit this group at http://groups.google.com/group/cp2k <http://groups.google.com/group/cp2k>.
> For more options, visit https://groups.google.com/d/optout <https://groups.google.com/d/optout>.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20141030/1ca76a1b/attachment.htm>
More information about the CP2K-user
mailing list