Introduction to OpenCL2CUDA

We all know that software is replacing many people functions. Said that, many very complicated data processing are now made by computers, that are getting better and better ways to do those tasks. There are many libraries and frameworks that helps programmers and engineers writing code to process some big amount of data. Two of these well known libraries are OpenCL, developed for heterogeneous computing (gpu, processors, fpga), and CUDA, a NVIDIA library created so people can write code to run on their GPUs. These libraries have some similar routines, cause there are many steps you have to do on both of them. Thinking about it, I started writing a OpenCL to CUDA converter.

Implementation

The implementation of this code still really simple, since all I am doing is searching for some OpenCL functions and replacing it for its equivalent on CUDA. Besides, if its not a direct translation, the converter suggests some possible fixes for you code. To find the suggestions on your code search for the #tranlation# word. We are using Python 3. To run this code all you have to do is:

chmod +x createCUDAkernel.py (just the first time)
./createCUDAkernel.py --opencl_name="name of the opencl file" --main_name="name of the C/C++ file"

How to contribute

Now, I am searching for CUDA and OpenCL codes that do the same thing, so I can go on this project. Besides, you can fork the project on Github.

Thanks a lot.