New OmpSs release 16.06

Mercurium compiler: 2.0.0
Nanos++ RT Library: 0.10
Download this version here

The programming models team is glad to announce you the release of the new stable version of OmpSs, which is based on the latest Mercurium and Nanos++ 0.10.

Apart from several bug-fixes in both tools, the main features introduced in this new version are:
  • New cluster support
  • Execute OmpSs programs transparently on top of a distributed memory system (CUDA & OpenCL devices are also supported).
  • Network communication is implemented using GASNet, which provides support for modern High Performance networking technologies.
  • Several optimization mechanisms allow to maximize the performance of applications running on a cluster: a data affinity scheduling policy distributes and minimizes the network activity, task presend allows the OmpSs run-time to overlap communication with computation.
  • Support for non-contiguous data
    • Tasks can reference non-contiguous, multi-dimensional data, which eases the implementation of some applications.
  • Thread manager
    • The Thread Manager module dynamically controls the amount of working threads needed for a specific amount of workload (info)
  • Task reductions
    • Extend the task construct adding support to the reduction clause (info)
    • Enhance the support of user-defined reductions (info)
  • Enhanced support for Dynamic Load Balancing

 

If you have any doubt or question, feel free to contact us at: pm-tools [at] bsc.es