A first attempt at using SDPA-GMP in Julia. Call SDPA_GMP.Optimizer() to use this wrapper via MathOptInterface.
This package is not yet registered in METADATA.jl. To install, type ] in the julia command prompt, then execute
(v1.1) pkg> add https://github.com/ericphanson/SDPA_GMP.jl.git
After installing the package, make sure that the sdpa_gmp binary is in system PATH. Alternatively, modify the sdpa_gmp_path constant in SDPA_GMP.jl to a path that points to the sdpa_gmp binary. Further information about compiling SDPA-GMP binary can be found here.
Note that before compilation it is often necessary to modify the SDPA-GMP source code so that the output values have sufficient precision. To do this, modify the P_FORMAT string at line 23 in sdpa_struct.h so that the output has a precision no less than the default precision (200 bits) or precision specified by the parameter file.
SDPA-GMP uses the parameter file named param.sdpa in the same folder as the binary if present. For details please refer to the SDPA users manual.
SDPA-GMP.jl uses Float64 for problem data and solution by default. To use, for example, BigFloat instead, simply call SDPA_GMP.Optimizer{BigFloat}(). Note that with MathOptInterface all the problem data must be of the same type, i.e. BigFloat in this case.
SDPA-GMP will emit cholesky miss condition :: not positive definite error if the problem data contain linearly dependent constraints. By default, a presolver will try to detect such constraints by Gaussian elimination. The redundant constraints are omitted from the problem formulation and the corresponding decision variables are set to 0 in the final result. The time taken to perform the presolve step depends on the size of the problem. To disable, call SDPA_GMP.Optimizer(presolve = false) instead.