04:11 <rtohid[m]> <matrixb01x> "what problems JIT can have in..." <- Many of HPX facilities are template functions, so it is not possible to create bindings for them ahead of time. Here is a simple example demonstrating the issue: https://github.com/pybind/pybind11/issues/199

04:11 <rtohid[m]> This is where JIT compilation, as suggested by @gonidelis, comes into play. pyhpx must include a code generation framework to generate the bindings on the fly. So, it's the solution not the problem...

04:11 <rtohid[m]> I'm interested to know the thoughts of others and see if there are other solutions

04:44 <satacker[m]> https://cppyy.readthedocs.io/en/latest/

07:38 <matrixb0tx> can someone tell me why there are 2 differnet folders libs/full and libs/core and why libs/src empty

07:45 <gonidelis[m]> full is for distributed hpx and core is for single core hpx

07:46 <gonidelis[m]> what would you exepct to have in libs/src?

07:49 <matrixb0tx> I dont know . but why there empty folder if it has no code in ti

08:01 <gonidelis[m]> i don't see it being empty

08:01 <gonidelis[m]> i think it's there for conformance reasons though

09:06 <matrixb0tx> rtohid[m] what about this Use PYBIND11_MAKE_OPAQUE(T) to disable the template-based conversion

09:06 <matrixb0tx> machinery of types, thus rendering them opaque.

09:07 <matrixb0tx> making the object opaque, I can ensure that it is only accessed through C++ functions that you explicitly expose to Python.

09:07 <matrixb0tx> any thoughts onthis

09:10 <satacker[m]> Instead of this, you can instantiate templates and allow only specific types like numpy arrays, python lists, tuples etc (Iterables supported)

09:11 <satacker[m]> That is to create a predefined interface

09:46 <matrixb0tx> I would have to write custom binding for for every .hpp  containing templates (all of them) and not templates can be . I am thinking a way to automate it this was the easiest way to do it by making is opaque . and cppyy . there so many differnt things you can do deciding which is right out is hard