Closed bas-rustenburg closed 7 years ago
Looping in @pgrinaway, he has some cython experience.
One thing that might let you avoid memory allocation manually but still get full C is to return the output via a parameter reference (then you can allocate the array elsewhere and the hot loop can be just C).
I can take a stab at this tomorrow morning if you want.
This didn't end up being very helpful, and I think we've found other ways of solving the problem computationally.
This is not much faster at the moment. Opening this PR to share my progress.
All the numpy arrays make use of the python interpreter. I might be able to gain some speedup if I replace them by C arrays.
See here for some details on why it is slow. Yellow means calling it is calling the python interpreter. You can click to see the C code underneath. https://rawgit.com/bas-rustenburg/bayesian-itc/cython/bitc/heats.html