Resource Time Comparison

Resource Times: Considers all models.

Date / Time: 10/09/03 17:08:12

Total resource time for all models.

See the solver return definitions for optimal/feasible model and solve statuses. Solutions are not checked for how close they are to eachother. Resource times are considered the same, if they are within 10% of eachother. A solver is considered faster than another, if it is less than 50% faster than the other. A solver is considered much faster than another, if it is more than 50% faster than the other.

If a model was not solved optimal/feasible by both solvers, the resource time is considered the same. If a model was solved optimal/feasible by solver A but not by solver B, then solver A is considered infinitely faster than solver B. Similarly, if one solver has trace data and the other has no data, then the first solver is consideredinfinitely faster.

A solver is considered to have a better objective function value, if the relative objective value difference is greater than 1.00E-05 For objective values below 1e-1 we use absolute differences.

Tracefile 1 :	`mosek_cone.pvr`
Tracefile 2 :	`snopt_cone.pvr`
Solvers used :	`MOSEK`
	`SNOPT`
Modeltype(s)	NLP

	Total	Obj MOSEK better	Obj same	Obj SNOPT better
Solver MOSEK infinitely faster :	`3`	`3`	-	-
Solver MOSEK much faster :	`14`	`4`	`4`	`6`
Solver MOSEK faster :	`1`	-	`1`	-
Solvers perform the same :	`2`	-	`2`	-
Solver SNOPT faster :	`5`	`1`	`4`	-
Solver SNOPT much faster :	`2`	-	`2`	-
Solver SNOPT infinitely faster :	-	-	-	-
Both solvers failed to solve optimally :	`1`	-	`1`	-
Total models: :	`28`	`8`	`14`	`6`

Solver return definition:

Outcome	Model Status	Solver Status
globally optimal	1 or 15	1
locally optimal/feasible	2 or 7 or 16 or 17	1 or 2 or 3 or 4 or 5

Resource Times:

Models for each resource time category. Listed are the solver resource times TIME(.) in seconds, as well as the ratio of resource times RATIO(.)for the two solvers if both solved optimally. Models not solved optimally by a solver have status listed instead of resource time.
Also listed are the objective values OBJ(.) using both solvers. The better solution found is listed in boldface. A solution is considered better, if the relative objective function difference is greater than 1.00E-05.
Solver resource time ratios for a particular model are listed only if both solvers are optimal and if one solver has resource time greater than 5.00E-02.

Solver MOSEK infinitely faster - Obj of MOSEK better:

Modelname	Time (MOSEK)	Time (SNOPT)	Ratio (MOSEK / SNOPT)	Obj (MOSEK)	Obj (SNOPT)
`emfl_socp_250_2`	`127.6800`	`fail`	--	`1.00751732E+02`	0.00000000E+00
`sched_200_100_o`	`25.0200`	`fail`	--	`1.41360366E+05`	2.68125525E+04
`sched_200_100_s`	`26.8900`	`fail`	--	`5.18119732E+01`	8.36818560E+00

Solver MOSEK much faster - Obj of MOSEK better:

Modelname	Time (MOSEK)	Time (SNOPT)	Ratio (MOSEK / SNOPT)	Obj (MOSEK)	Obj (SNOPT)
`nb_nlp`	`6.2000`	`122.1600`	`0.051`	`-5.07030948E-02`	`-4.81927695E-02`
`nb_l2_nlp`	`20.3000`	`123.0600`	`0.165`	`-1.62897291E+00`	`-1.47534758E+00`
`nb_l2_bessel_nl`	`3.7700`	`122.0900`	`0.031`	`-1.02569509E-01`	`1.68124257E-01`
`sched_100_100_s`	`11.5700`	`120.0800`	`0.096`	`2.73308254E+01`	`2.85885509E+01`

Solver MOSEK much faster - Obj same for both solvers:

Modelname	Time (MOSEK)	Time (SNOPT)	Ratio (MOSEK / SNOPT)	Obj (MOSEK)	Obj (SNOPT)
`sched_50_50_ori`	`1.9100`	`17.2100`	`0.111`	`2.66729916E+04`	`2.66729936E+04`
`sched_50_50_sca`	`1.4600`	`13.3900`	`0.109`	`7.85203848E+00`	`7.85203845E+00`
`sched_100_50_sc`	`4.4500`	`90.8200`	`0.049`	`6.71650801E+01`	`6.71649827E+01`
`springs_100_nlp`	`0.1800`	`7.5100`	`0.024`	`-4.15910422E+04`	`-4.15910401E+04`

Solver MOSEK much faster - Obj of SNOPT better:

Modelname	Time (MOSEK)	Time (SNOPT)	Ratio (MOSEK / SNOPT)	Obj (MOSEK)	Obj (SNOPT)
`emfl_socp_200_2`	`3.6500`	`120.1200`	`0.030`	`4.68675434E+01`	`3.45561667E+01`
`emfl_socp_200_1`	`29.4800`	`120.7300`	`0.244`	`8.07643056E+01`	`9.42799449E+00`
`emfl_socp_500_1`	`49.5200`	`120.7700`	`0.410`	`2.06376237E+02`	`7.93508938E+00`
`nb_l1_nlp`	`6.8400`	`123.6900`	`0.055`	`-1.30122707E+01`	`-1.56522407E+02`
`sched_100_50_or`	`4.2700`	`120.0300`	`0.036`	`1.81889918E+05`	`1.12922069E+05`
`sched_100_100_o`	`9.0300`	`120.0500`	`0.075`	`7.17367750E+05`	`3.62534232E+05`

Solver MOSEK faster - Obj same for both solvers:

Modelname	Time (MOSEK)	Time (SNOPT)	Ratio (MOSEK / SNOPT)	Obj (MOSEK)	Obj (SNOPT)
`fir_socp_20_nlp`	`0.0700`	`0.1200`	`0.583`	`1.04648765E+00`	`1.04648765E+00`

Solvers perform the same: Back to top

Modelname	Time (MOSEK)	Time (SNOPT)	Ratio (MOSEK / SNOPT)	Obj (MOSEK)	Obj (SNOPT)
`qp7_nlp`	`0.0400`	`0.0200`	`---`	`8.09315462E-04`	`8.09315050E-04`
`springs_10_nlp`	`0.0300`	`0.0800`	`---`	`-1.85446065E+02`	`-1.85446062E+02`

Solver SNOPT faster - Obj of MOSEK better:

Modelname	Time (MOSEK)	Time (SNOPT)	Ratio (MOSEK / SNOPT)	Obj (MOSEK)	Obj (SNOPT)
`springs_1000_nl`	`41.3500`	`32.4700`	`1.273`	`-4.00326051E+07`	`-3.62174061E+06`

Solver SNOPT faster - Obj same for both solvers:

Modelname	Time (MOSEK)	Time (SNOPT)	Ratio (MOSEK / SNOPT)	Obj (MOSEK)	Obj (SNOPT)
`fir_socp_10_nlp`	`0.0600`	`0.0500`	`1.200`	`1.73823742E+00`	`1.73823302E+00`
`qp7_50_100_nlp`	`0.0800`	`0.0700`	`1.143`	`1.81608587E-03`	`1.81606351E-03`
`qp7_100_100_nlp`	`0.1700`	`0.1300`	`1.308`	`4.29072481E-03`	`4.29072457E-03`
`qp7_100_170_nlp`	`0.2700`	`0.1500`	`1.800`	`3.41240580E-03`	`3.41240484E-03`

Solver SNOPT much faster - Obj same for both solvers:

Modelname	Time (MOSEK)	Time (SNOPT)	Ratio (MOSEK / SNOPT)	Obj (MOSEK)	Obj (SNOPT)
`fir_socp_40_nlp`	`0.5400`	`0.1300`	`4.154`	`1.00008181E+00`	`1.00008181E+00`
`fir_socp_80_nlp`	`0.7700`	`0.3700`	`2.081`	`1.00000002E+00`	`1.00000071E+00`

Both solvers failed to solve optimally - Obj same for both solvers:

Modelname	Time (MOSEK)	Time (SNOPT)	Ratio (MOSEK / SNOPT)	Obj (MOSEK)	Obj (SNOPT)
`springs_10000_n`	`fail`	`fail`	--	`6.81631145E+02`	`-4.89951186E+04`