An Automated Tool for Converting Directive Based C Code Into Parallel CUDA Code
Computer Engineering, Sandip Institute of Technology and Research Centre, Nashik
IRF International Conference, 2014
@article{saifuddin2014automated,
title={An Automated Tool for Converting Directive Based C Code Into Parallel CUDA Code},
author={Saifuddin, Batliwala and Lalit, Kadtan and Mujjammil, Khan and Pratik, Khandagale and Walunj, S. M.},
year={2014}
}
Parallel programming has become simple and reasonable with the preamble of GPGPUs. Now a day’s many programmers transfer their application to GPGPUs with the accessibility of APIs such as NVIDIA’s CUDA. But it is very tricky task to write CUDA program. Most of the industry extensively uses the immense serial C code, and they are unable to take any advantages of this additional computing power available. Some tools, allow programmers to add "hints" to their sequential programs, while another approach has been to build an interactive system between parallelizing tools/compilers and programmers. But none of these are really automatic tools, because programmer is fully involved in the process. We present an automatic parallelization tool with modest involvement of the programmer in the process of parallelization. This tool will provide better graphical user interface with higher flexibility. The process followed in this tool will produce better quality result.
April 19, 2014 by hgpu
