List view
This version of SMLP's API is meant take care of the following goals: - allow user to specify an algorithm solving their specific in terms of building blocks, such as - basic: SMT solutions - sequential: (single-objective) optimization results - allow for not-only machine level parallelism in the basic building blocks The main means to achieve these are: - separate workers from server: - workers are separate processes on different machines running SMT solvers - the server is a single process sending/receiving messages to/from workers in form of network messages - the user code starts the server and schedules tasks to workers by means of the server's API (the v2 API) - asynchronous functions using Python's single-threaded async/await feature (requires Python >= v3.7) to let workers run while the server does something else / "awaits" some results from workers This constitutes a complete redesign of the solver. The reasoning is laid out in https://github.com/fbrausse/smlp/blob/fb/v2/redesign
No due date•0/3 issues closed