I meet a problem with the speedup cvxpy solve method(basically use SCS solver),the task contains lots of constraints and a big matrix. I have tried the following methods:
- use
mklblas&lapacklibrary to replace the original library. (got a little improvement) - use GPU and
scsto do matrix calculation. (made a worse performance) - update to the latest python and the corresponding library. (got a little improvement)
those solutions didn't improve the performance enough(even the GPU indirect calculation get a worse result), Did anybody has some advice about this project.