J-Parallelio – automatic parallelization framework for Java virtual machine code
share
Manual translation of the algorithms from sequential version to its parallel counterpart is time consuming and can be done only with the specific knowledge of hardware accelerator architecture, parallel programming or programming environment. The automation of this process makes porting the code much easier and faster. The key aspect in this case is how efficient the generated parallel code will be. The paper describes J-Parallelio, the framework for automatic analysis of the bytecode source codes and its parallelisation on multicore processors. The process consists of a few steps. First step is a process of decompilation of JVM and its translation to internal abstract syntax tree, the dependency extraction and memory analysis is performed. Finally, the mapping process is performed which consists of a set of rules responsible for translating the input virtual machine source code to its parallel version. The main novelty is that it can deal with pure Java virtual machine and can generate parallel code for multicore processors. This makes the system portable and it can work with different languages based on JVM after some small modifications. The efficiency of automatically translated source codes were compared with their manually written counterparts on chosen benchmarks.
READ FULL TEXT