Large scale benchmark for PAO ML (and/or LS DFT in general).
Bunzip2 all files, have BASIS_MOLOPT and GTH_POTENTIALS available (from cp2k/data.
For tuning purposes, the length of the full benchmark can be reduced in the following ways:
- by reducing the number of MD steps (STEPS 20 -> STEPS 5)
- by doing only an energy calculation (RUN_TYPE MD -> RUN_TYPE ENERGY)
- by doing an energy calculation with few SCF steps (MAX_SCF 50 -> MAX_SCF 5)
Reference energies and timings using CP2K svn:17405, Piz Daint, Cray XC30, 1024 nodes, SB + K20X, +- 8Gb mem used per node
(last column timings per step)
> grep SCF out | head -n 10
------------------------------ Linear scaling SCF -----------------------------
SCF 1 -552962.443966503 -552962.443966503 28.188702
SCF 2 -553391.806569889 -429.362603386 32.410283
SCF 3 -553756.740393045 -794.296426542 12.872345
SCF 4 -554517.794343487 -1555.350376984 13.497826
SCF 5 -554728.247042532 -210.452699045 33.469438
SCF 6 -554849.946612308 -332.152268821 13.510741
SCF 7 -554873.984235220 -356.189891734 17.914532
SCF 8 -554915.784987915 -41.800752695 38.287248
SCF 9 -554918.728930513 -44.744695293 14.047717
> cat pao_ml_md-1.ener
# Step Nr. Time[fs] Kin.[a.u.] Temp[K] Pot.[a.u.] Cons Qty[a.u.] UsedTime[s]
0 0.000000 110.495412573 300.000000000 -554937.745690580 -554827.250278007 0.000000000
1 0.500000 106.078997716 288.009235623 -554933.183067304 -554827.104069588 1606.181380667
2 1.000000 102.069290733 277.122701359 -554929.060740767 -554826.991450034 437.939860202
3 1.500000 100.198165911 272.042513562 -554927.153307400 -554826.955141488 449.515135792
4 2.000000 101.022853760 274.281578054 -554928.022891735 -554827.000037975 375.188747253
5 2.500000 103.892195371 282.071969193 -554930.981726713 -554827.089531342 378.129010544
6 3.000000 107.422614470 291.657215360 -554934.608040349 -554827.185425879 363.554193239
7 3.500000 110.115891922 298.969584414 -554937.373128860 -554827.257236939 365.110804793
8 4.000000 110.802924876 300.834909692 -554938.098364813 -554827.295439937 367.848488379
9 4.500000 108.870951325 295.589514868 -554936.164765656 -554827.293814332 366.621533569
10 5.000000 104.372470458 283.375937591 -554931.619703179 -554827.247232721 371.757427932
11 5.500000 98.092255253 266.324871693 -554925.248720303 -554827.156465049 366.436341383
12 6.000000 91.519001482 248.478192945 -554918.555918355 -554827.036916873 438.594749461
13 6.500000 86.552351413 234.993515291 -554913.479003540 -554826.926652127 363.525124027
14 7.000000 84.878892773 230.450000040 -554911.758150807 -554826.879258034 361.602635782
15 7.500000 87.178039466 236.692286411 -554914.109975162 -554826.931935696 362.720366624
16 8.000000 92.619988766 251.467422791 -554919.692436207 -554827.072447441 360.480368018
17 8.500000 99.082357198 269.013042870 -554926.321931528 -554827.239574330 362.877061196
18 9.000000 104.007271910 282.384407159 -554931.375035372 -554827.367763462 364.628673398
19 9.500000 105.405738067 286.181305486 -554932.815285468 -554827.409547401 365.768221021
20 10.000000 102.522367308 278.352824576 -554929.882289568 -554827.359922261 364.275226861
SUBROUTINE CALLS ASD SELF TIME TOTAL TIME
MAXIMUM AVERAGE MAXIMUM AVERAGE MAXIMUM
CP2K 1 1.0 0.182 0.228 8802.156 8802.158
qs_mol_dyn_low 1 2.0 0.002 0.002 8793.361 8796.208
qs_forces 21 4.0 0.300 0.313 8731.917 8731.935
qs_energies 21 5.0 0.001 0.001 8299.259 8316.296
ls_scf 21 6.0 0.000 0.000 8140.304 8157.313
velocity_verlet 20 3.0 0.014 0.024 7689.100 7689.373
ls_scf_main 21 7.0 0.006 6.929 6214.282 6221.060
dbcsr_multiply_internal 6268 10.6 5.425 5.502 5456.668 5499.570
multiply_cannon 6268 11.6 9.185 9.602 4797.967 4876.009
dm_ls_curvy_optimization 234 7.8 0.002 0.002 3756.519 3757.137
optimization_step 234 8.8 0.001 0.046 3056.137 3059.618
multiply_cannon_multrec 401152 12.6 845.768 1008.921 2235.185 2348.598
mp_waitall_1 3329028 13.7 2028.515 2273.110 2028.515 2273.110
compute_direction_newton 78 9.8 0.101 0.977 2206.565 2206.643
ls_scf_dm_to_ks 255 7.9 0.003 0.004 2046.532 2047.273
multiply_cannon_metrocomm3 401152 12.6 0.957 1.092 1280.957 1642.411
commutator_symm 1536 10.8 0.024 0.026 1393.033 1401.600
dbcsr_mm_accdrv_process 7955916 13.9 862.352 933.296 1051.028 1136.256
rebuild_ks_matrix 277 9.6 0.002 0.004 1013.434 1013.460
qs_ks_build_kohn_sham_matrix 277 10.6 0.037 0.051 1013.433 1013.458
pao_post_scf 21 7.0 0.000 0.000 979.076 979.097
multiply_cannon_metrocomm1 401152 12.6 1.188 1.293 666.837 966.589
pao_add_forces 21 8.0 0.022 0.036 939.777 941.085
qs_ks_update_qs_env 256 8.9 0.002 0.002 915.333 915.356
update_p_exp 234 9.8 0.013 0.032 849.491 852.997
qs_rho_update_rho 256 8.9 0.003 0.004 732.701 747.571
calculate_rho_elec 256 9.9 486.068 514.736 732.698 747.568
sum_up_and_integrate 277 11.6 0.488 0.516 732.856 733.703
integrate_v_rspace 277 12.6 518.242 550.497 732.367 733.214
matrix_qs_to_ls 276 8.0 0.142 0.144 701.236 702.223
pao_calc_outer_grad_lnv 21 9.0 0.100 0.114 657.554 660.960
transform_matrix_orth 333 8.8 0.010 0.011 647.406 651.982
purify_mcweeny_orth 255 10.7 0.009 0.493 635.637 639.508
make_m2s 12536 11.6 5.422 5.562 568.779 607.900
make_images 12536 12.6 19.448 20.772 560.314 599.309
make_images_sizes 12536 13.6 0.028 0.032 428.487 491.794
mp_alltoall_i44 12536 14.6 428.459 491.763 428.459 491.763
matrix_ls_to_qs 276 8.9 0.154 0.157 441.958 469.591
mp_alltoall_d11v 10853 12.8 223.253 388.573 223.253 388.573
dbcsr_new_transposed 3211 11.1 9.536 10.585 333.787 347.908
rs_distribute_matrix 554 12.2 26.205 29.397 187.192 341.592
dbcsr_redistribute 2470 12.3 31.904 35.233 319.945 334.210