The HECToR Service is now closed and has been superceded by ARCHER.

CASINO

This Distributed Computational Science and Engineering (dCSE) project was to improve the performance of the Quantum Monte Carlo (QMC) code CASINO. Three main areas were improved with : i) better memory utilisation using shared data between tasks belonging to the same processor, ii) acceleration of the computation speed by implementing second level parallelism iii) improved input reading speed for large sets of initial data.

  • Performance gains from this dCSE work show a speedup of up to 100% for certain large scale models. Mixed mode parallelism will now speed up the computation by a factor of 1.6-1.8. A user running CASINO with 1000 cores on HECToR will now save at least 3000 AUs per run.

The individual achievements of the project are summarised below:

  • The System V shared memory solution allows sharing of the orbitals data and therefore each core of a processor can run a computing task even for very large scale models. In quantitative terms the speed of models using more than 2GB of orbital data was increased by 100%.
  • The input optimisations have eliminated unnecessary waiting time at the start of a computation which wasted between 3000-6000 AU per 1000 cores per run.
  • The computation using mixed mode second level parallelism speeds up the computation by a factor of 1.6-1.8 on the quad core processor for models with more than 1000 electrons. The mixed mode parallelism offers the possibility to extend the range of CASINO towards models with 10,000 electrons.
  • The shared memory and optimized input were incorporated in release 2.4 of CASINO(July 2009). The CASINO manual was updated with detailed information about the new features and the Makefile was adapted in order to provide users with easy access to the new features. The mixed mode version is schedule for the 2010 release and it is also available upon request from development branches.

Please see PDF or HTML for a report which summarises this project.