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

Previous topic

Glossary

Next topic

Bibliography

This Page