mathematical optimization - Limits of CPLEX -
we have problem involving 100k variables in objective function , 1500 conditions. cplex support such numbers? if so, limits , performance benchmarks of cplex?
regards,
yash
yes.
quoting reply cplex developers in this thread, dated in 2010:
cplex can handle 30 million variables. more interesting question be: machine have enough memory that. cplex needs store lower, upper bounds , objective function coefficients double precision values each variable. results in storage requirement of @ least 30000000 * (8 + 8 + 8) bytes 680 mb. , have variable names, constraints etc.
and then
this little subtle answer number may depend on features of cplex use (implicitly). however, cplex supports 262,500,000 integral variables (that 2,100,000,000 claimed roland divided sizeof(double)) features -- provided there enough memory that. space required creating many variables 6 gb (one double precision number lower bounds, upper bound , objective function coefficient).
performance benchmarks
here official page of hans mittelman, benchmarks on performance of lp solvers, , here mixed-integer linear programming benchmarks (problem description here).
you notice there few problems in these sets larger problem.
however, in mips (problems integer variables) mileage may vary performance-wise. there exist problems 200 integer variables in cplex cannot below 10% gap in hour of computations. problem's size not predictor solver's performance.
Comments
Post a Comment