Instead of assuming there won't be errors in the C++ XLA code, pass these errors up through FFI to Idris and handle them there. This is mainly to make debugging and development of spidr much easier. Errors are thrown frequently when working on spidr, be it refactoring or adding features, and it's a real pain debugging them.
Instead of assuming there won't be errors in the C++ XLA code, pass these errors up through FFI to Idris and handle them there. This is mainly to make debugging and development of spidr much easier. Errors are thrown frequently when working on spidr, be it refactoring or adding features, and it's a real pain debugging them.