| 30 Oct 2025 | ==================== |
| AMPL-NLP Benchmark | |
| ==================== | |
| (mittelmann@asu.edu) |
Logfiles are at: plato.asu.edu/ftp/ampl-nlp_logs/
The (local) codes
IPOPT-3.14.5 github.com/coin-or/Ipopt KNITRO-15.0 www.artelys.com/knitro/ SNOPT-7.7 www.scicomp.ucsd.edu/~peg/ CONOPT-4.38.1 www.conopt.com/ WORHP-1.16 worhp.de/ FMINCON-2024a www.mathworks.com COPT-8.0.0 www.shanshu.ai/copt/ UNO-2.0.3 github.com/cvanaret/Unoare used to solve some medium size NLP instances coded in AMPL and run as NL files.
plato.asu.edu/ftp/ampl-nlp-source/The codes were run in default mode and with a time limit of 2hrs on an AMD Ryzen 9 5900X (12 cores, 128GB).
========================================================================== scaled shifted geom mean 11.6 1.94 103 38.4 11.8 37.0 1 276 solved 46 46 30 39 45 36 47 22 47 problems nv nc IPOPT$ KNIT SNOPT CONPT WORHP MATLB COPT UNO ========================================================================== arki0003 2237 2500 1 1 1 1 2 i 1 2 arki0009 6220 5924 6 1 119 8 t 36 3 i bearing_400 160000 0 5 3 f 73 7 11 9 t camshape_6400 6400 12800 3 2 8 10 5 12 1 40 clnlbeam 59999 40000 536 1 t 3 1 162 3 t cont5_1_l 90600 90300 13 11 t t 69 f 8 844 cont5_2_1_l 90600 90300 35 16 t 5902 65 27 3 341 cont5_2_2_l 90600 90300 42 27 t 6875 429 t 19 880 cont5_2_3_l 90600 90300 47 12 t t 512 f 20 t cont5_2_4_l 90600 90300 12 10 716 45 154 77 8 1039 corkscrw 44997 35000 28 1 16 272 33 i 3 3195 dirichlet_120 53881 241 106 80 t 1209 315 i 41 t dtoc1nd 6705 4470 3 1 27 12 8 14 1 111 dtoc2 64950 38970 2076 3 1493 74 5 2766 4 t elec_400 1200 400 36 21 26 4 10 32 13 i ex1_160 50562 25281 1 1 103 129 4 15 2 t ex1_320 203522 101761 7 4 2668 320 31 35 10 t ex4_2_160 51198 25917 2 1 79 109 11 39 1 t ex4_2_320 204798 103037 7 11 1589 2634 121 250 6 t ex8_2_2 7510 1943 1 1 12 2 5 7 1 1 ex8_2_3 15636 3155 1 1 36 3 4 9 1 3 gasoil_3200 32001 31998 69 8 33 4 2 77 1 t henon_120 32401 241 131 75 t 1261 74 i 15 i lane_emden_120 57721 241 116 36 t 861 131 408 16 2017 marine_1600 38415 38392 3 1 f 54 98 14 1 t NARX_CFy 43973 46744 938 i t f 117 1070 17 t nql180 162001 130080 2050 7 f t 27 f 8 t optmass 60006 50005 7 3 24 11 1 20 1 2131 pinene_3200 64000 63995 7 1 1233 11 108 f 1 t qcqp500-3c 500 120 769 1 337 64 539 42 2 382 qcqp500-3nc 500 120 143 1 24 96 216 37 1 26 qcqp750-2c 750 138 281 1 1693 184 176 174 5 1441 qcqp750-2nc 750 138 543 1 67 133 20 113 5 69 qcqp1000-1nc 1000 154 20 3 149 15 831 82 6 24 qcqp1000-2c 1000 5107 31 1 7 141 12 51 3 18 qcqp1000-2nc 1000 5107 68 1 78 119 189 236 2 605 qcqp1500-1c 1500 10508 122 2 f 1156 46 791 5 143 qcqp1500-1nc 1500 10508 t 140 338 981 156 i 10 f qssp180 261365 139141 15 9 t t 156 241 10 t robot_1600 14399 9601 1 1 8 5 2 10 1 574 robot_a 1001 52013 688 1 7 121 18 24 3 f robot_b 1001 52013 717 9 322 144 1200 25 4 i robot_c 1001 52013 683 1 f 147 13 24 3 i rocket_12800 51201 38400 6 2 t 581 24 12 2 t steering_12800 32000 25601 14 1 678 78 52 36 2 t svanberg 50000 50000 31 3 t f 43 23 5 t WM_CFy 8520 9826 1243 1612 2922 f 5526 i 13 t =========================================================================== nv/nc: no. of var./cons.; t/m/i: tim/mem/itlim exceeded $: Higher performance of IPOPT may be achieved by using different linear algebra routines (both architecture and problem dependent) ===========================================================================