I am trying to solve SDP using cvx. I tried first using single processor i7 cpu. Now I used the 4 cpu xeon processor which should be faster. But it seems to take the same time? Is there some way to make it faster?
Which part of CVX is slow for you? The specification process, or the solution process? If it’s the solution process, then you should try the other solvers—including MOSEK and Gurobi, if you have a CVX professional license. But I am not certain if any of the solvers can truly handle a multiprocessor system very well.
It must be the solution process. But I was hoping at least the computation should be faster on the Xeon CPU. I can try other solvers add well. But looks like cvx might not be the best choice to solve an SDP for say 1000 cross 1000 matrix case.
Well, if CVX is not suitable it is only because the solvers it uses are not. There may be other solvers out there that take advantage of multiprocessor setups better; if you find one, let us know and we can consider adding support for it to CVX.
I tried to use Gurobi but it says ; gurobi does not support semidefinite cones larger than 2x2.
You must use another solver for this problem.
Ah yes, sorry, MOSEK is it as far as CVX goes.