Future Work
Work in progress concerning this project includes the following:
- To increase the number of algorithms parallelized using yaCF, with particular attention to commercial applications
- To study and implement additional compiler optimizations that will enhance the performance of the target code
- To extend the llc syntax to capture additional information from the programmer for better adaption of the translation to the target architecture
- To study the generation of hybrid CUDA+OpenMP code
Some future work may be done within the CUDA backend.
- To improve locality through a better use of the memory hierarchy
- To use the texture memory to store read-only data
- An intelligent load balance between host and device