BSC contributes to the beta version of the TANGO Toolbox with programming models

03 May 2018

The TANGO project has released the beta version of the TANGO Toolbox, which aims to make heterogeneous hardware architectures easier to program

BSC contributed by integrating COMPSs and OmpSs in the TANGO Toolbox programming model

The TANGO Toolbox is the main outcome of TANGO project. It is a set of modular, open and interoperable tools to design, model, develop and execute software in distributed heterogeneous hardware architectures. The TANGO Toolbox also allows the optimization of various dimensions of software design and operations, such as energy efficiency, performance and dependability on target architectures.

Barcelona Supercomputing Center (BSC) has contributed to the development of the beta version of the TANGO Toolbox developing the TANGO Programming Model which is an integration of its in-house programming models COMPSs and OmpSs. “Distributed platforms are becoming more and more heterogeneous. Computing clusters and Cloud providers are offering computing nodes with different accelerators like GPUs and FPGA, With the TANGO Programming Model, developers can easily implement parallel applications for these kind of platforms, where COMPSs transparently manages the execution in the different heterogeneous computing nodes of the platform and OmpSs manages the execution inside each node transparently spawning the computation in the available computing devices (CPU, GPUS or FPGAs)” says BSC researcher Jorge Ejarque from the Workflows and Distributed Computing group in the Computer Sciences department, who is involved in the TANGO project.

The components of the TANGO Toolbox can be used stand-alone or combined with other ones to support complex scenarios or use cases, facilitating and enhancing their performance. In future, the number of components will grow as the number of supported use cases increases. Currently, three different use cases have been identified where TANGO plays a part (efficient apps programming in HPC, programming embedded apps efficiently, and the design and programming of heterogeneous platforms with configurable hardware).

In this beta version of the project, the focus has been on performing on time and saving energy. This could be seen at several levels. At the IDE layer, the tools tell you whether it is better or not to offload tasks to an FPGA and allow you to visualize which architectures gives you the better performance for a given task. At the runtime level, the application is monitored in terms of energy and performance, if possible by the underlying infrastructure, and it self-adapts to optimize the application in this aspects. This beta version also supports new heterogeneous hardware like the Intel Xeon Phi. Finally, this version also focused on improvements in productivity on the developer’s side, offering robust feedback on tools trained in this version and the first version (Alpha version), to achieve better results and correct possible mistakes during the training period for development.

TANGO is an EU project, coordinated by Atos, launched in January 2016 with the purpose of paving the way for the new world that emerges from the new exploiting possibilities that the use of new powerful computing hardware offer by customized heterogeneous hardware. The project aims to pursue maximum success through the benefits of upgrades to fulfill the objectives of the entire consortium, to finally present a great solution to be implemented outside of the context of the project. TANGO shows the #TANGOtoolbox as the way for developers to get closer to the next generation of applications, enabling them to work better and faster in mobile, IoT (Internet of things), Cyber Physical Systems, Wearables, Big Data and HPC (High Performance Computing).
TANGO is the founding member of
Heterogeneity Alliance, which aims at uniting the efforts of organizations interested in the development of future technologies and tools to advance, and take full advantage, of computing and applications using heterogeneous hardware.

For further information visit