PaPILO for PDLP and CP-SAT
232 views
Skip to first unread message
Stuart Rogers
Feb 6, 2023, 5:58:57 PM2/6/23
to or-tools-discuss
Are there plans to integrate PaPILO as a presolver for PDLP and CP-SAT? My understanding is that PDLP and CP-SAT use GLOP as a presolver, which may not be ideal for very large problems.
Laurent Perron
Feb 7, 2023, 12:08:39 AM2/7/23
to or-tools-discuss
No. I do not see why.
Glop presolve is a very small part of CP-SAT presolve.
Le lun. 6 févr. 2023, 23:59, Stuart Rogers <stum...@gmail.com> a écrit :
Are there plans to integrate PaPILO as a presolver for PDLP and CP-SAT? My understanding is that PDLP and CP-SAT use GLOP as a presolver, which may not be ideal for very large problems.
--
You received this message because you are subscribed to the Google Groups "or-tools-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to or-tools-discu...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/or-tools-discuss/6a6c3434-2791-45a8-9bcd-f8a656ebfa7en%40googlegroups.com.
Stuart Rogers
Feb 7, 2023, 12:18:35 AM2/7/23
to or-tools-discuss
Are there plans to integrate PaPILO as an optional presolver for PDLP?
Laurent Perron
Feb 7, 2023, 12:33:13 AM2/7/23
to or-tools-discuss
Currently no.
Do you have evidence it is doing a better job ?
There are not that many things to do when presolving as a lp, compared to as a mip.
To view this discussion on the web visit https://groups.google.com/d/msgid/or-tools-discuss/dec883e9-01ae-4464-822a-ef4759857678n%40googlegroups.com.
Stuart Rogers
Feb 7, 2023, 12:49:59 AM2/7/23
to or-tools-discuss
For very large problems, the GLOP presolver doesn't work, whereas PaPILO does. Also, PaPILO will presolve a MIP. Then PDLP can solve the LP relaxation of the presolved MIP.
Laurent Perron
Feb 7, 2023, 1:09:56 AM2/7/23
to or-tools-discuss
Please define 'does not work's for glop.
To view this discussion on the web visit https://groups.google.com/d/msgid/or-tools-discuss/6ec390d8-5482-4239-a1ec-d025b8f44a4dn%40googlegroups.com.
Stuart Rogers
Feb 7, 2023, 3:35:20 AM2/7/23
to or-tools...@googlegroups.com
With the GLOP presolver turned off, PDLP solved a very large LP in 3.5 hours. With the GLOP presolver turned on, PDLP was killed after about 15 minutes, probably during presolve. PDLP was able to solve the PaPILO-presolved LP in less time than 3.5 hours (this was several months ago so I don't have the timing).
You received this message because you are subscribed to a topic in the Google Groups "or-tools-discuss" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/or-tools-discuss/TTqSgFu2eXs/unsubscribe.
To unsubscribe from this group and all its topics, send an email to or-tools-discu...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/or-tools-discuss/CABcmEeZS4WUp-0aMFnUB1Tz%3DeWJUjhWA%3D0DqtLXyp%3DLXx-QFcg%40mail.gmail.com.
Laurent Perron
Feb 7, 2023, 3:37:39 AM2/7/23
to or-tools-discuss
Can you send us the lp?
To view this discussion on the web visit https://groups.google.com/d/msgid/or-tools-discuss/CAO6iSeCa1wVaYgADHXENdo_ZaNtkhDjHJk3afz0opwx3GuvTyQ%40mail.gmail.com.
David Applegate
Feb 7, 2023, 12:05:33 PM2/7/23
to or-tools-discuss
For PDLP, we do have some interest in supporting PaPILO, especially for very large problems. However, it would add quite a few dependencies to or-tools (Boost, TBB, Catch2, Fmt, Lusol, Pdqsort, Ska, at least, I believe). An alternative might be letting the user integrate with PaPILO (i.e., use PaPILO and send the presolved instance to PDLP). As is, that wouldn't support checking termination criteria on the original problem, but we've also been considering adding a termination check callback to better support custom termination criteria.
Stuart Rogers
Feb 7, 2023, 12:08:22 PM2/7/23
to or-tools-discuss
Below is the error when the GLOP presolver is turned on, using PDLP in OR-Tools v9.4. The input MPS file is 9.6 GB.
$ solve --solver=pdlp --logtostderr --linear_solver_enable_verbose_output --input=input.mps --params="termination_criteria{eps_optimal_relative:1e-3}, presolve_options{use_glop:true}" --num_threads=24
I0207 08:56:38.110951 186917 solve.cc:150] Read input proto as an MPModelProto.
File : '/data/stuart/gen_MPS_files/p10_fix2_MILP_binary_x.mps'
I0207 08:56:38.117798 186917 solve.cc:181] Set number of threads to 24.
Solver : PDLP_LINEAR_PROGRAMMING
Dimension : 47261300 x 47251695
I0207 08:59:01.050074 186917 primal_dual_hybrid_gradient.cc:1858] Problem stats before presolving and rescaling:
I0207 08:59:01.053852 186917 primal_dual_hybrid_gradient.cc:1298] There are 47251695 variables, 47261300 constraints, and 141744517 constraint matrix nonzeros.
I0207 08:59:01.053862 186917 primal_dual_hybrid_gradient.cc:1304] Absolute values of nonzero constraint matrix elements: largest=1.000000, smallest=1.000000, avg=1.000000
I0207 08:59:01.053876 186917 primal_dual_hybrid_gradient.cc:1309] Constraint matrix, infinity norm: max(row & col)=1.000000, min_col=1.000000, min_row=1.000000
I0207 08:59:01.053882 186917 primal_dual_hybrid_gradient.cc:1314] Constraint bounds statistics (max absolute value per row): largest=10.000000, smallest=1.000000, avg=0.000315, l2_norm=122.425488
I0207 08:59:01.053888 186917 primal_dual_hybrid_gradient.cc:1331] Absolute values of objective vector elements: largest=12800.000000, smallest=20.000000, avg=3561.474902, l2_norm=30055161.602451
I0207 08:59:01.053895 186917 primal_dual_hybrid_gradient.cc:1338] Gaps between variable upper and lower bounds: #finite=47251695 of 47251695, largest=1.000000, smallest=1.000000, avg=1.000000
Killed
I0207 08:56:38.110951 186917 solve.cc:150] Read input proto as an MPModelProto.
File : '/data/stuart/gen_MPS_files/p10_fix2_MILP_binary_x.mps'
I0207 08:56:38.117798 186917 solve.cc:181] Set number of threads to 24.
Solver : PDLP_LINEAR_PROGRAMMING
Dimension : 47261300 x 47251695
I0207 08:59:01.050074 186917 primal_dual_hybrid_gradient.cc:1858] Problem stats before presolving and rescaling:
I0207 08:59:01.053852 186917 primal_dual_hybrid_gradient.cc:1298] There are 47251695 variables, 47261300 constraints, and 141744517 constraint matrix nonzeros.
I0207 08:59:01.053862 186917 primal_dual_hybrid_gradient.cc:1304] Absolute values of nonzero constraint matrix elements: largest=1.000000, smallest=1.000000, avg=1.000000
I0207 08:59:01.053876 186917 primal_dual_hybrid_gradient.cc:1309] Constraint matrix, infinity norm: max(row & col)=1.000000, min_col=1.000000, min_row=1.000000
I0207 08:59:01.053882 186917 primal_dual_hybrid_gradient.cc:1314] Constraint bounds statistics (max absolute value per row): largest=10.000000, smallest=1.000000, avg=0.000315, l2_norm=122.425488
I0207 08:59:01.053888 186917 primal_dual_hybrid_gradient.cc:1331] Absolute values of objective vector elements: largest=12800.000000, smallest=20.000000, avg=3561.474902, l2_norm=30055161.602451
I0207 08:59:01.053895 186917 primal_dual_hybrid_gradient.cc:1338] Gaps between variable upper and lower bounds: #finite=47251695 of 47251695, largest=1.000000, smallest=1.000000, avg=1.000000
Killed
Stuart Rogers
Feb 7, 2023, 12:41:37 PM2/7/23
to or-tools-discuss
When the GLOP presolver is turned off, PDLP is solving the problem.
Solver : PDLP_LINEAR_PROGRAMMING
Dimension : 47261300 x 47251695
$ solve --solver=pdlp --logtostderr --linear_solver_enable_verbose_output --input=input.mps --params="termination_criteria{eps_optimal_relative:1e-3}, presolve_options{use_glop:false}" --num_threads=24
I0207 09:13:37.567702 186946 solve.cc:150] Read input proto as an MPModelProto.
File : 'input.mps'
I0207 09:13:37.578476 186946 solve.cc:181] Set number of threads to 24.
I0207 09:13:37.567702 186946 solve.cc:150] Read input proto as an MPModelProto.
File : 'input.mps'
I0207 09:13:37.578476 186946 solve.cc:181] Set number of threads to 24.
Solver : PDLP_LINEAR_PROGRAMMING
Dimension : 47261300 x 47251695
I0207 09:16:01.121251 186946 primal_dual_hybrid_gradient.cc:1858] Problem stats before rescaling:
I0207 09:16:01.125003 186946 primal_dual_hybrid_gradient.cc:1298] There are 47251695 variables, 47261300 constraints, and 141744517 constraint matrix nonzeros.
I0207 09:16:01.125011 186946 primal_dual_hybrid_gradient.cc:1304] Absolute values of nonzero constraint matrix elements: largest=1.000000, smallest=1.000000, avg=1.000000
I0207 09:16:01.125023 186946 primal_dual_hybrid_gradient.cc:1309] Constraint matrix, infinity norm: max(row & col)=1.000000, min_col=1.000000, min_row=1.000000
I0207 09:16:01.125028 186946 primal_dual_hybrid_gradient.cc:1314] Constraint bounds statistics (max absolute value per row): largest=10.000000, smallest=1.000000, avg=0.000315, l2_norm=122.425488
I0207 09:16:01.125032 186946 primal_dual_hybrid_gradient.cc:1331] Absolute values of objective vector elements: largest=12800.000000, smallest=20.000000, avg=3561.474902, l2_norm=30055161.602451
I0207 09:16:01.125036 186946 primal_dual_hybrid_gradient.cc:1338] Gaps between variable upper and lower bounds: #finite=47251695 of 47251695, largest=1.000000, smallest=1.000000, avg=1.000000
I0207 09:16:02.928667 186946 primal_dual_hybrid_gradient.cc:1927] Problem stats after rescaling:
I0207 09:16:02.928694 186946 primal_dual_hybrid_gradient.cc:1298] There are 47251695 variables, 47261300 constraints, and 141744517 constraint matrix nonzeros.
I0207 09:16:02.928700 186946 primal_dual_hybrid_gradient.cc:1304] Absolute values of nonzero constraint matrix elements: largest=0.840896, smallest=0.010618, avg=0.300228
I0207 09:16:02.928707 186946 primal_dual_hybrid_gradient.cc:1309] Constraint matrix, infinity norm: max(row & col)=0.840896, min_col=0.076125, min_row=0.025097
I0207 09:16:02.928711 186946 primal_dual_hybrid_gradient.cc:1314] Constraint bounds statistics (max absolute value per row): largest=1.172895, smallest=0.117290, avg=0.000045, l2_norm=18.228729
I0207 09:16:02.928715 186946 primal_dual_hybrid_gradient.cc:1331] Absolute values of objective vector elements: largest=10763.474115, smallest=16.817928, avg=2994.831478, l2_norm=25273277.651371
I0207 09:16:02.928718 186946 primal_dual_hybrid_gradient.cc:1338] Gaps between variable upper and lower bounds: #finite=47251695 of 47251695, largest=11.046289, smallest=1.189207, avg=1.190133
iter# kkt_pass time | rel_prim_res rel_dual_res rel_gap | prim_resid dual_resid obj_gap | prim_obj dual_obj | prim_var_l2 dual_var_l2
0 0.0 1.9 | 0.998790 0.00000 0.00000 | 122.278 0.00000 0.00000 | 0.00000 0.00000 | 1.41421 0.00000
64 70.0 55.8 | 0.0280624 0.618417 -0.253533 | 3.43558 1.85866e+07 -2.02184e+07 | 2.97642e+07 4.99826e+07 | 6.22323 1.26147e+07
128 139.0 92.8 | 0.0160167 0.0536887 -0.0107858 | 1.96086 1.61362e+06 -654584. | 3.00175e+07 3.06721e+07 | 7.23091 1.21464e+07
192 208.0 129.8 | 0.00668027 0.0184950 -0.0133601 | 0.817842 555872. -829742. | 3.06381e+07 3.14679e+07 | 8.01165 1.11868e+07
256 275.0 162.4 | 0.00400867 0.0129691 -0.00605125 | 0.490767 389790. -375009. | 3.07986e+07 3.11736e+07 | 8.25760 1.10829e+07
320 342.0 198.7 | 0.00263937 0.0101943 0.00568137 | 0.323129 306393. 351557. | 3.11153e+07 3.07637e+07 | 9.11567 1.10669e+07
384 409.0 231.4 | 0.00336836 0.00784146 0.00508904 | 0.412376 235676. 315286. | 3.11346e+07 3.08193e+07 | 9.38245 1.10504e+07
448 477.0 264.5 | 0.00384927 0.00711738 0.00657943 | 0.471253 213914. 406885. | 3.11244e+07 3.07175e+07 | 9.65453 1.10236e+07
512 544.0 297.4 | 0.00416169 0.00653913 0.00755084 | 0.509501 196535. 466019. | 3.10918e+07 3.06258e+07 | 9.89131 1.09921e+07
576 610.0 333.6 | 0.00507244 0.00568525 0.00967830 | 0.621001 170871. 590991. | 3.08272e+07 3.02363e+07 | 11.2322 1.08502e+07
640 676.0 366.3 | 0.00493565 0.00507531 0.00942726 | 0.604254 152539. 573869. | 3.07236e+07 3.01497e+07 | 11.5011 1.08227e+07
704 742.0 398.8 | 0.00475720 0.00468064 0.00900656 | 0.582407 140677. 546691. | 3.06229e+07 3.00762e+07 | 11.7567 1.07966e+07
768 808.0 431.3 | 0.00455310 0.00445373 0.00820788 | 0.557419 133858. 496964. | 3.05220e+07 3.00251e+07 | 11.9973 1.07714e+07
832 874.0 463.8 | 0.00432830 0.00430867 0.00728883 | 0.529898 129498. 440274. | 3.04221e+07 2.99818e+07 | 12.2291 1.07479e+07
896 939.0 497.2 | 0.00290021 0.00375786 0.00196999 | 0.355062 112943. 117263. | 2.98211e+07 2.97038e+07 | 13.7708 1.06364e+07
iter# kkt_pass time | rel_prim_res rel_dual_res rel_gap | prim_resid dual_resid obj_gap | prim_obj dual_obj | prim_var_l2 dual_var_l2
960 1004.0 529.3 | 0.00262924 0.00353591 0.000869503 | 0.321888 106272. 51639.9 | 2.97209e+07 2.96692e+07 | 14.0495 1.06248e+07
1024 1068.0 562.2 | 0.00192380 0.00296895 -0.00158507 | 0.235524 89232.2 -93436.9 | 2.94274e+07 2.95208e+07 | 15.0478 1.06000e+07
1088 1133.0 594.4 | 0.00173793 0.00277025 -0.00202687 | 0.212769 83260.2 -119184. | 2.93415e+07 2.94607e+07 | 15.4405 1.05952e+07
1152 1198.0 626.6 | 0.00159988 0.00265910 -0.00231884 | 0.195867 79919.8 -136055. | 2.92690e+07 2.94051e+07 | 15.8226 1.05910e+07
1216 1264.0 659.3 | 0.00150470 0.00248462 -0.00262411 | 0.184216 74675.8 -153703. | 2.92098e+07 2.93635e+07 | 16.1956 1.05871e+07
1280 1329.0 692.7 | 0.00143892 0.00211410 -0.00474291 | 0.176162 63539.5 -275809. | 2.89380e+07 2.92138e+07 | 18.2309 1.05725e+07
1344 1394.0 725.0 | 0.00145242 0.00198114 -0.00476626 | 0.177814 59543.4 -276837. | 2.89029e+07 2.91797e+07 | 18.8052 1.05704e+07
1408 1460.0 757.7 | 0.00141101 0.00188438 -0.00476242 | 0.172745 56635.3 -276316. | 2.88719e+07 2.91482e+07 | 19.3431 1.05680e+07
1472 1526.0 790.6 | 0.00142003 0.00182156 -0.00485047 | 0.173849 54747.4 -281183. | 2.88445e+07 2.91257e+07 | 19.8766 1.05653e+07
1536 1590.0 822.6 | 0.00146611 0.00176715 -0.00484904 | 0.179490 53111.9 -280867. | 2.88207e+07 2.91016e+07 | 20.3720 1.05626e+07
1600 1656.0 855.4 | 0.00152361 0.00171531 -0.00474580 | 0.186530 51553.9 -274636. | 2.87973e+07 2.90719e+07 | 20.8793 1.05597e+07
1664 1721.0 888.0 | 0.00157459 0.00167162 -0.00471112 | 0.192772 50240.9 -272414. | 2.87756e+07 2.90480e+07 | 21.3559 1.05569e+07
1728 1787.0 920.8 | 0.00162125 0.00163333 -0.00479021 | 0.198484 49090.0 -276798. | 2.87537e+07 2.90305e+07 | 21.8406 1.05541e+07
1792 1852.0 953.3 | 0.00166127 0.00160652 -0.00485793 | 0.203383 48284.1 -280516. | 2.87317e+07 2.90122e+07 | 22.3126 1.05513e+07
1856 1918.0 987.2 | 0.00179957 0.00163835 -0.00576537 | 0.220315 49240.9 -332571. | 2.86759e+07 2.90084e+07 | 23.4478 1.05502e+07
iter# kkt_pass time | rel_prim_res rel_dual_res rel_gap | prim_resid dual_resid obj_gap | prim_obj dual_obj | prim_var_l2 dual_var_l2
1920 1983.0 1019.8 | 0.00201262 0.00150792 -0.00541294 | 0.246399 45320.9 -311831. | 2.86482e+07 2.89601e+07 | 24.4289 1.05489e+07
1984 2048.0 1052.8 | 0.00217206 0.00142667 -0.00494481 | 0.265917 42878.7 -284495. | 2.86248e+07 2.89093e+07 | 25.3764 1.05474e+07
2048 2113.0 1085.6 | 0.00229857 0.00136409 -0.00423993 | 0.281406 40998.1 -243566. | 2.86010e+07 2.88446e+07 | 26.2823 1.05458e+07
2112 2177.0 1117.9 | 0.00241215 0.00130823 -0.00344102 | 0.295311 39319.0 -197366. | 2.85797e+07 2.87771e+07 | 27.1466 1.05442e+07
2176 2241.0 1150.3 | 0.00252043 0.00124513 -0.00161565 | 0.308567 37422.7 -92430.4 | 2.85585e+07 2.86510e+07 | 27.9985 1.05425e+07
2240 2305.0 1182.8 | 0.00262269 0.00118832 -0.000464597 | 0.321086 35715.2 -26530.1 | 2.85385e+07 2.85650e+07 | 28.8337 1.05407e+07
2304 2371.0 1216.0 | 0.00271645 0.00114950 0.000266402 | 0.332565 34548.4 15190.7 | 2.85184e+07 2.85032e+07 | 29.6446 1.05389e+07
2368 2436.0 1248.7 | 0.00281238 0.00110998 0.000938904 | 0.344310 33360.7 53461.8 | 2.84970e+07 2.84436e+07 | 30.5092 1.05369e+07
2432 2502.0 1283.1 | 0.00313395 0.00104309 0.00445495 | 0.383678 31350.1 252344. | 2.84479e+07 2.81955e+07 | 32.5592 1.05361e+07
2496 2568.0 1316.3 | 0.00374818 0.000890457 0.00800549 | 0.458876 26762.8 451508. | 2.84256e+07 2.79741e+07 | 34.4600 1.05352e+07
2560 2634.0 1349.4 | 0.00416604 0.000829613 0.00921991 | 0.510034 24934.2 518918. | 2.84006e+07 2.78817e+07 | 36.2645 1.05341e+07
2624 2699.0 1381.9 | 0.00462079 0.000781113 0.00999460 | 0.565707 23476.5 561673. | 2.83797e+07 2.78180e+07 | 37.8992 1.05329e+07
2688 2765.0 1415.3 | 0.00502777 0.000744529 0.0105881 | 0.615532 22376.9 594292. | 2.83612e+07 2.77669e+07 | 39.4438 1.05316e+07
I0207 09:16:01.125003 186946 primal_dual_hybrid_gradient.cc:1298] There are 47251695 variables, 47261300 constraints, and 141744517 constraint matrix nonzeros.
I0207 09:16:01.125011 186946 primal_dual_hybrid_gradient.cc:1304] Absolute values of nonzero constraint matrix elements: largest=1.000000, smallest=1.000000, avg=1.000000
I0207 09:16:01.125023 186946 primal_dual_hybrid_gradient.cc:1309] Constraint matrix, infinity norm: max(row & col)=1.000000, min_col=1.000000, min_row=1.000000
I0207 09:16:01.125028 186946 primal_dual_hybrid_gradient.cc:1314] Constraint bounds statistics (max absolute value per row): largest=10.000000, smallest=1.000000, avg=0.000315, l2_norm=122.425488
I0207 09:16:01.125032 186946 primal_dual_hybrid_gradient.cc:1331] Absolute values of objective vector elements: largest=12800.000000, smallest=20.000000, avg=3561.474902, l2_norm=30055161.602451
I0207 09:16:01.125036 186946 primal_dual_hybrid_gradient.cc:1338] Gaps between variable upper and lower bounds: #finite=47251695 of 47251695, largest=1.000000, smallest=1.000000, avg=1.000000
I0207 09:16:02.928667 186946 primal_dual_hybrid_gradient.cc:1927] Problem stats after rescaling:
I0207 09:16:02.928694 186946 primal_dual_hybrid_gradient.cc:1298] There are 47251695 variables, 47261300 constraints, and 141744517 constraint matrix nonzeros.
I0207 09:16:02.928700 186946 primal_dual_hybrid_gradient.cc:1304] Absolute values of nonzero constraint matrix elements: largest=0.840896, smallest=0.010618, avg=0.300228
I0207 09:16:02.928707 186946 primal_dual_hybrid_gradient.cc:1309] Constraint matrix, infinity norm: max(row & col)=0.840896, min_col=0.076125, min_row=0.025097
I0207 09:16:02.928711 186946 primal_dual_hybrid_gradient.cc:1314] Constraint bounds statistics (max absolute value per row): largest=1.172895, smallest=0.117290, avg=0.000045, l2_norm=18.228729
I0207 09:16:02.928715 186946 primal_dual_hybrid_gradient.cc:1331] Absolute values of objective vector elements: largest=10763.474115, smallest=16.817928, avg=2994.831478, l2_norm=25273277.651371
I0207 09:16:02.928718 186946 primal_dual_hybrid_gradient.cc:1338] Gaps between variable upper and lower bounds: #finite=47251695 of 47251695, largest=11.046289, smallest=1.189207, avg=1.190133
iter# kkt_pass time | rel_prim_res rel_dual_res rel_gap | prim_resid dual_resid obj_gap | prim_obj dual_obj | prim_var_l2 dual_var_l2
0 0.0 1.9 | 0.998790 0.00000 0.00000 | 122.278 0.00000 0.00000 | 0.00000 0.00000 | 1.41421 0.00000
64 70.0 55.8 | 0.0280624 0.618417 -0.253533 | 3.43558 1.85866e+07 -2.02184e+07 | 2.97642e+07 4.99826e+07 | 6.22323 1.26147e+07
128 139.0 92.8 | 0.0160167 0.0536887 -0.0107858 | 1.96086 1.61362e+06 -654584. | 3.00175e+07 3.06721e+07 | 7.23091 1.21464e+07
192 208.0 129.8 | 0.00668027 0.0184950 -0.0133601 | 0.817842 555872. -829742. | 3.06381e+07 3.14679e+07 | 8.01165 1.11868e+07
256 275.0 162.4 | 0.00400867 0.0129691 -0.00605125 | 0.490767 389790. -375009. | 3.07986e+07 3.11736e+07 | 8.25760 1.10829e+07
320 342.0 198.7 | 0.00263937 0.0101943 0.00568137 | 0.323129 306393. 351557. | 3.11153e+07 3.07637e+07 | 9.11567 1.10669e+07
384 409.0 231.4 | 0.00336836 0.00784146 0.00508904 | 0.412376 235676. 315286. | 3.11346e+07 3.08193e+07 | 9.38245 1.10504e+07
448 477.0 264.5 | 0.00384927 0.00711738 0.00657943 | 0.471253 213914. 406885. | 3.11244e+07 3.07175e+07 | 9.65453 1.10236e+07
512 544.0 297.4 | 0.00416169 0.00653913 0.00755084 | 0.509501 196535. 466019. | 3.10918e+07 3.06258e+07 | 9.89131 1.09921e+07
576 610.0 333.6 | 0.00507244 0.00568525 0.00967830 | 0.621001 170871. 590991. | 3.08272e+07 3.02363e+07 | 11.2322 1.08502e+07
640 676.0 366.3 | 0.00493565 0.00507531 0.00942726 | 0.604254 152539. 573869. | 3.07236e+07 3.01497e+07 | 11.5011 1.08227e+07
704 742.0 398.8 | 0.00475720 0.00468064 0.00900656 | 0.582407 140677. 546691. | 3.06229e+07 3.00762e+07 | 11.7567 1.07966e+07
768 808.0 431.3 | 0.00455310 0.00445373 0.00820788 | 0.557419 133858. 496964. | 3.05220e+07 3.00251e+07 | 11.9973 1.07714e+07
832 874.0 463.8 | 0.00432830 0.00430867 0.00728883 | 0.529898 129498. 440274. | 3.04221e+07 2.99818e+07 | 12.2291 1.07479e+07
896 939.0 497.2 | 0.00290021 0.00375786 0.00196999 | 0.355062 112943. 117263. | 2.98211e+07 2.97038e+07 | 13.7708 1.06364e+07
iter# kkt_pass time | rel_prim_res rel_dual_res rel_gap | prim_resid dual_resid obj_gap | prim_obj dual_obj | prim_var_l2 dual_var_l2
960 1004.0 529.3 | 0.00262924 0.00353591 0.000869503 | 0.321888 106272. 51639.9 | 2.97209e+07 2.96692e+07 | 14.0495 1.06248e+07
1024 1068.0 562.2 | 0.00192380 0.00296895 -0.00158507 | 0.235524 89232.2 -93436.9 | 2.94274e+07 2.95208e+07 | 15.0478 1.06000e+07
1088 1133.0 594.4 | 0.00173793 0.00277025 -0.00202687 | 0.212769 83260.2 -119184. | 2.93415e+07 2.94607e+07 | 15.4405 1.05952e+07
1152 1198.0 626.6 | 0.00159988 0.00265910 -0.00231884 | 0.195867 79919.8 -136055. | 2.92690e+07 2.94051e+07 | 15.8226 1.05910e+07
1216 1264.0 659.3 | 0.00150470 0.00248462 -0.00262411 | 0.184216 74675.8 -153703. | 2.92098e+07 2.93635e+07 | 16.1956 1.05871e+07
1280 1329.0 692.7 | 0.00143892 0.00211410 -0.00474291 | 0.176162 63539.5 -275809. | 2.89380e+07 2.92138e+07 | 18.2309 1.05725e+07
1344 1394.0 725.0 | 0.00145242 0.00198114 -0.00476626 | 0.177814 59543.4 -276837. | 2.89029e+07 2.91797e+07 | 18.8052 1.05704e+07
1408 1460.0 757.7 | 0.00141101 0.00188438 -0.00476242 | 0.172745 56635.3 -276316. | 2.88719e+07 2.91482e+07 | 19.3431 1.05680e+07
1472 1526.0 790.6 | 0.00142003 0.00182156 -0.00485047 | 0.173849 54747.4 -281183. | 2.88445e+07 2.91257e+07 | 19.8766 1.05653e+07
1536 1590.0 822.6 | 0.00146611 0.00176715 -0.00484904 | 0.179490 53111.9 -280867. | 2.88207e+07 2.91016e+07 | 20.3720 1.05626e+07
1600 1656.0 855.4 | 0.00152361 0.00171531 -0.00474580 | 0.186530 51553.9 -274636. | 2.87973e+07 2.90719e+07 | 20.8793 1.05597e+07
1664 1721.0 888.0 | 0.00157459 0.00167162 -0.00471112 | 0.192772 50240.9 -272414. | 2.87756e+07 2.90480e+07 | 21.3559 1.05569e+07
1728 1787.0 920.8 | 0.00162125 0.00163333 -0.00479021 | 0.198484 49090.0 -276798. | 2.87537e+07 2.90305e+07 | 21.8406 1.05541e+07
1792 1852.0 953.3 | 0.00166127 0.00160652 -0.00485793 | 0.203383 48284.1 -280516. | 2.87317e+07 2.90122e+07 | 22.3126 1.05513e+07
1856 1918.0 987.2 | 0.00179957 0.00163835 -0.00576537 | 0.220315 49240.9 -332571. | 2.86759e+07 2.90084e+07 | 23.4478 1.05502e+07
iter# kkt_pass time | rel_prim_res rel_dual_res rel_gap | prim_resid dual_resid obj_gap | prim_obj dual_obj | prim_var_l2 dual_var_l2
1920 1983.0 1019.8 | 0.00201262 0.00150792 -0.00541294 | 0.246399 45320.9 -311831. | 2.86482e+07 2.89601e+07 | 24.4289 1.05489e+07
1984 2048.0 1052.8 | 0.00217206 0.00142667 -0.00494481 | 0.265917 42878.7 -284495. | 2.86248e+07 2.89093e+07 | 25.3764 1.05474e+07
2048 2113.0 1085.6 | 0.00229857 0.00136409 -0.00423993 | 0.281406 40998.1 -243566. | 2.86010e+07 2.88446e+07 | 26.2823 1.05458e+07
2112 2177.0 1117.9 | 0.00241215 0.00130823 -0.00344102 | 0.295311 39319.0 -197366. | 2.85797e+07 2.87771e+07 | 27.1466 1.05442e+07
2176 2241.0 1150.3 | 0.00252043 0.00124513 -0.00161565 | 0.308567 37422.7 -92430.4 | 2.85585e+07 2.86510e+07 | 27.9985 1.05425e+07
2240 2305.0 1182.8 | 0.00262269 0.00118832 -0.000464597 | 0.321086 35715.2 -26530.1 | 2.85385e+07 2.85650e+07 | 28.8337 1.05407e+07
2304 2371.0 1216.0 | 0.00271645 0.00114950 0.000266402 | 0.332565 34548.4 15190.7 | 2.85184e+07 2.85032e+07 | 29.6446 1.05389e+07
2368 2436.0 1248.7 | 0.00281238 0.00110998 0.000938904 | 0.344310 33360.7 53461.8 | 2.84970e+07 2.84436e+07 | 30.5092 1.05369e+07
2432 2502.0 1283.1 | 0.00313395 0.00104309 0.00445495 | 0.383678 31350.1 252344. | 2.84479e+07 2.81955e+07 | 32.5592 1.05361e+07
2496 2568.0 1316.3 | 0.00374818 0.000890457 0.00800549 | 0.458876 26762.8 451508. | 2.84256e+07 2.79741e+07 | 34.4600 1.05352e+07
2560 2634.0 1349.4 | 0.00416604 0.000829613 0.00921991 | 0.510034 24934.2 518918. | 2.84006e+07 2.78817e+07 | 36.2645 1.05341e+07
2624 2699.0 1381.9 | 0.00462079 0.000781113 0.00999460 | 0.565707 23476.5 561673. | 2.83797e+07 2.78180e+07 | 37.8992 1.05329e+07
2688 2765.0 1415.3 | 0.00502777 0.000744529 0.0105881 | 0.615532 22376.9 594292. | 2.83612e+07 2.77669e+07 | 39.4438 1.05316e+07
Stuart Rogers
Feb 7, 2023, 1:55:58 PM2/7/23
to or-tools-discuss
When the GLOP presolver is turned off, PDLP solved the problem.
$ solve --solver=pdlp --logtostderr --linear_solver_enable_verbose_output --input=input.mps --params="termination_criteria{eps_optimal_relative:1e-3}, presolve_options{use_glop:false}" --num_threads=24
......
Status : MPSOLVER_OPTIMALObjective : 2.705765382474077e+07
BestBound : inf
Iterations : 9920
Time : 5266
Stuart Rogers
Feb 7, 2023, 2:45:07 PM2/7/23
to or-tools-discuss
For that particular problem with the relative error set to 1e-3, PDLP solves the
Original problem with GLOP on: Killed during GLOP presolve
Original problem with GLOP off: in 5266 [s] in 9920 iterations
PaPILO-presolved problem with GLOP off: in 307.1 [s] in 2240 iterations
PaPILO-presolved problem with GLOP on: in 266.4 [s] in 1940 iterations
Stuart Rogers
Feb 8, 2023, 1:40:21 AM2/8/23
to or-tools-discuss
PaPILO is bundled with SCIP, which is provided with OR-Tools. Could OR-Tools, e.g., PDLP and CP-SAT, use the version of PaPILO bundled with SCIP? SCIP's license was updated to Apache 2.0 in November, starting with v8.0.3. https://scipopt.org/
watchdogs132
Feb 8, 2023, 4:14:09 AM2/8/23
to or-tools-discuss
Laurent Perron
Feb 8, 2023, 5:03:40 AM2/8/23
to or-tools-discuss
We are not using papilo in ortools/scip.
We cannot easily deliver an release artifact that contains papilo because of the licence.
To view this discussion on the web visit https://groups.google.com/d/msgid/or-tools-discuss/aa763613-ff71-4816-9900-35aa23c11facn%40googlegroups.com.
Stuart Rogers
Apr 21, 2025, 1:05:38 PMApr 21
to or-tools-discuss
It looks like a future version (3.0.0) of PaPILO will change the license from LGPL-3.0 to Apache License 2.0. The links below are from the develop branch of PaPILO:
https://github.com/scipopt/papilo/tree/develop
Laurent Perron
Apr 21, 2025, 1:21:58 PMApr 21
to or-tools...@googlegroups.com
To view this discussion visit https://groups.google.com/d/msgid/or-tools-discuss/e6ff89e3-7fd1-4c6b-aa0d-b851759420f1n%40googlegroups.com.
Stuart Rogers
Dec 3, 2025, 8:40:50 PM (3 days ago) Dec 3
to or-tools-discuss
PaPILO v3.0.0 has been released, which changed the license to Apache License 2.0. https://github.com/scipopt/papilo/releases/tag/v3.0.0
Laurent Perron
Dec 5, 2025, 7:57:02 AM (2 days ago) Dec 5
to or-tools...@googlegroups.com
Yes, still a nightmare to integrate.
Bazel has no support, I have not yet managed to make it work.
CMake is unfinished.
To view this discussion visit https://groups.google.com/d/msgid/or-tools-discuss/46b46ef5-45e4-4e64-a3db-5cd1d9e3afa6n%40googlegroups.com.
--Laurent
Reply all
Reply to author
Forward
0 new messages