Models: all 125 MIPs from the GAMS LINLib
Solvers:
- CPLEX 11.1.1 (as in GAMS 22.8.1)
- CBC 2.2 (stable branch, rev. 1052)
using Clp/stable/1.8 rev. 1265, CoinUtils/stable/2.3 rev. 1070, Cgl/stable/0.53 rev. 680, Osi/stable/0.99 rev. 1288
with extra compiler flags: -DDISALLOW_PRINTING -DCLP_FAST_CODE -DCOIN_FAST_CODE -DCOIN_USE_RESTRICT -DNEW_STYLE_SOLVER=0
- SCIP 1.10
using Clp/stable/1.8 rev. 1285, CoinUtils/stable/2.3 rev. 1093
with extra compiler flags: -DDISALLOW_PRINTING -DCLP_FAST_CODE -DCOIN_FAST_CODE -DCOIN_USE_RESTRICT -DNEW_STYLE_SOLVER=0
with ATLAS 3.8.2 as Blas substitute and Lapack 3.1.1
- SCIP 1.00
using Clp/stable/1.8 rev. 1265, CoinUtils/stable/2.3 rev. 1070
with extra compiler flags: -DDISALLOW_PRINTING -DCLP_FAST_CODE -DCOIN_FAST_CODE -DCOIN_USE_RESTRICT -DNEW_STYLE_SOLVER=0
- GLPK 4.30
Timelimit: 3600 seconds
Gap tolerance: 0
Platform:
- Intel Core 2 Duo CPU T7500, 2.2 GHz, 2 GB RAM
- Linux 2.6.22.18, SuSE Linux 10.3, 32 Bit
- gcc 4.2.1
Results
Notes:
- SCIP 1.1: For the instances air04 and misc06 the solution status has been changed from "solved to optimality" to "integer feasible" and the objective estimate has been set to "NA"
because other solvers found a better solution than the reported one.
- GLPK 4.30: For the instances neos04 and neos05 the solution status has been changed from "solved to optimality" to "integer feasible"
because other solvers found a better solution than the reported one.
- The GAMS/GLPK does not report an objective estimate. Thus, GLPK does not occur the performance profiles which include the gap tolerance into their measure.
- For the logfiles of CPLEX, CBC, SCIP 1.0, and GLPK, see a previous benchmark