| 3 | | In Hydra^VM^ we focus on extracting parallelization from sequential code (e.g. legacy systems) through a set of techniques including: code profiling and visualization, execution analysis, data dependency graphs and software transactional memory. Hydra^VM^ is a virtual machine prototype built on [http://jikesrvm.org Jikes RVM]. Currently, we write our implementation as a modification to baseline compiler, however, our contributions are applicable to other architectures and JIT compilers. At the user level, no changes or recompilation are needed to the underlying source code, all our work are done at the standard bytecode level. [[BR]] |
| | 3 | The Hydra^VM^ project focuses on extracting parallelism from sequential code (e.g., legacy systems) through a set of techniques including code profiling and visualization, data dependency analysis, execution analysis, and software transactional memory, all done at the bytecode-level. Thus, no changes or recompilation are needed to application source codes. Hydra^VM^ is a virtual machine prototype that incorporates these techniques, built by extending [http://jikesrvm.org Jikes RVM]. Though the project's current implementation is a modification to [http://jikesrvm.org Jikes RVM] baseline compiler, the techniques are broadly applicable to other architectures and JIT compilers. [[BR]] |
