SST v6.0.0 Released
SST v6.0.0 Release is Now Available and can be downloaded here!
The SST project teams are pleased to announce that the v6.0.0 of the simulation framework and components are now available for download here.
SST v6.0.0 is one of the most significant updates to the simulation framework since the project began and includes thousands of changes added to the core and elements to improve performance, scalability and simulation accuracy. The most significant of these are overhauled serialization support to reduce large-scale parallel simulation times and the introduction of C++11-based threading to allow for on-node parallel simulations. Further, the core and elements have been separated to allow for future releases to be more flexible and more easily upgradable. The independent SST-Core is also intended to allow for greater community collaboration around a high-performance parallel discrete event core.
The changes in v6.0.0 of the Structure Simulation Toolkit (SST) are divided into three areas: general, PDES-core and integrated elements.
General:
- SST v6.0.0 includes numerous fixes to issues in the core which address end-user feedback and significantly improve performance for some simulations
- SST Core and Elements are now separate downloads and build independently, with per-installation configuration abilities.
- SST project now hosted on GitHub
Core:
- SST Core is an independent application built separately from the elements
- SST Core uses new custom serialization for improved performance (no longer based on Boost serialzation)
- SST Core adds support for C++11 threading allowing parallel simulations to use MPI and on-node threading
- SST Core random number generations are enhanced to improve performance
- Fixes to SST Core statistics handling to improve performance
- Additional support for querying SST configuration in the Python interface
Elements:
- SST Elements is separately built from the SST-Core
- Thornhill element added which will allow for multi-fidelity on-node simulations to be created
- SST-Macro now uses SST-Core for parallel discrete event simulation
- Merlin now supports adaptive routing
- Enhanced DragonFly configuration options for Merlin
- Additional motifs added to Ember to broaden communication pattern modeling in more dynamic cases
- Memory allocation support added to FireFly to improve memory bandwidth modeling
- MemHierarchy includes supports for improved cache modeling including cache-banking
- Memory Controllers now support SubComponent options for request re-ordering
- Ariel supports statistical gathering for application instructions, memory allocations and calls to free
- Fixes in Cassini to reduce memory requirements for prefetching tables
SST Research and Product Teams
Sandia National Laboratories, USA
