Multiprocessed architectures have been making each time more frequent on the current computational context, by the need for greater computating power and processing restrictions in only one central unit. Consequently, the lack of software development more consistent, safe, and mostly complex, is real. The exploitation of parallelism is consensually the most suitable resource to be used, however, until the moment it isn’t a complete groundwork on the treatment of non-deterministic software. We still have to observe, assess and determine more efficient methods for such cases. Transactional memory is a new mechanism focused on the correct handling of shared memory, reducing the occurrence of some typical problems of concurrent and parallel computing, such as deadlocks, difficulties on synchronization, race conditions and inconsistent states, yet facilitating the programming and concurrency control. This project introduces tLua, a software transactional memory system developed and made for the Lua programming language.
