WG Leader: Prof. Wolfgang Karl
There has been an increasing number of papers illustrating the potential of transactions in parallel programs, introducing TM implementation techniques, and proposing hardware mechanisms to accelerate common case behavior.
Nevertheless, TMs are still not widely used by application developers as the TM technology has not yet been integrated with any of the widely used software development frameworks. It is unlikely that programmers will abandon the many advantages of mature frameworks in order to try this new technology. As a result, there are very few large applications that use transactions and there is no clear assessment on whether TM actually improves programmability.
In addition, there is little convergence or sharing of technology between TM researchers. Each group uses a different programming model, code generation, and TM system approach. This leads to incompatible and immature modules that are unlikely to attract significant attention. The activities carried out withing this WG will be aimed at removing this major technology roadblock by focusing on the issues related to the integration of TM supports within mainstream languages and development frameworks. This encompasses the development of a set of extensions to compilers (e.g. the GNU gcc compiler for C/C++ languages), virtual machines for languages relying on managed runtime systems (e.g. the Jikes RVM for the Java language), as well as companion tools for assisting the development process such as debuggers and profilers.
This WG will also endeavor to boost the international standardization process of a basic set of APIs to be provided at the language level to support execution of TM-based applications.
Working Groups >