For legacy sequential code, Parallel Xtractor will analyze a region and under user direction, refactor the source code, and parallelize it for multi-core environments. OpenCL, OpenMP and OpenACC offload are user selectable code parallelization options. For new software targeted towards parallel systems, engineers may choose to write in a parallel language. There are also many good reasons for a developer to continue writing sequential code in C or C++. Perhaps code is intended for a sequential system today but will be run on both sequential and multi-core systems in the future. Perhaps the target is multi-core but the target platform is not yet ready. Perhaps there is simply a desire to program in a known language, in a known or existing environment.
For these engineers, Parallel Xtractor will monitor and review the sequential software development process, providing guidance during code creation. Code developed in this manner allows it to be created and debugged in a known environment but also ensures this software is free of dependencies and ready for parallelism.
Why use Parallel Xtractor
- Parallelize legacy sequential code
- Prepare sequential code for parallelization using executable refactoring proposals
- Apply pre-defined parallel transformations to a code region
- Benefit from platform-aware optimizations
Features of Parallel Xtractor
- Leverage multiple proven parallelization approaches using best-in-class algorithms
- Coarse or fine grain capabilities
- Source-to-source code translation
- Synergistic with existing development flows and tools platforms