Lead Developers: Jing Guo (2012)
Most modern desktop machines contain very powerful multiprocessor sub-systems on their graphics cards. Nvidia has now developed a freely accessible development environment named CUDA. It enables C programmers to conveniently program Nvidia's line of graphics processors.
The goal of this project is to investigate the suitability of the CUDA system as a back-end target of the high-performance compiler sac2c which compiles programs written in SaC into C programs.
The intended approach is to take a few prototypical C program kernels (less than 100 lines of code each), to translate these into the CUDA framework, and to analyze the potential performance gains in depth. The chosen kernels should represent the sac2c compiler output for the most important language constructs of SaC so that the performance analyzes can be used as an indication of the potential performance benefits.
Current Status: finished.