cuda - GPGPU performance in high-level languages -

cuda - GPGPU performance in high-level languages -

for science fair project have write computationally-intensive algorithm suited parallelization. have read opencl , cuda , seems used c/c++. while not hard me pick bit of c write simple main, wondering how big performance nail if used java or python bindings gpu computation? specifically, more interested in performance nail using cuda because that's framework i'm planning on using.

in general, every time add together abstraction layer you're loosing performance but, in case of cuda not true because, whether using python or java you'll end writing cuda kernels on c/fortran, performance in gpu side same using c/fortran (check pycuda examples here)

the bad news java , python never accomplish performance of compiled languages such c on tasks, see this reply more detailed give-and-take topic. here give-and-take c versus java, on so.

there many questions , discussions performance comparing between interpreted , compiled languages, encourage read of them.

cuda opencl gpgpu pycuda jocl