Logo of Science Foundation Ireland  Logo of the Higher Education Authority, Ireland7 CapacitiesGPGPU Research Projects
Ireland's High-Performance Computing Centre | ICHEC
Home | News | Infrastructure | Outreach | Services | Research | Support | Education & Training | Consultancy | About Us | Login

Consortium Support

ICHEC provides direct consortium support to a number of projects. This support takes the form of a member of the ICHEC scientific staff committing a portion of their time to that individual project over a given period. Below you will see listed the activities that are underway in 2009 along with some details. This type of work complements the normal issue specific Helpdesk support by being broader in nature and more long term.

Typically the process begins by the user and ICHEC staff defining the scope of the work. All users are strongly encouraged to engage in this activiity. It is a great opportunity to work hand in hand with ICHEC staff, pick up tips and tricks and help ICHEC to better understand your objectives.

If you already have an ICHEC project and feel consortium support is something that could be valuable to your project or would like to discuss the possibilities for such support in the context of preparing a project application please feel free to contact us.

Selection of activities underway in 2010 & 2009

Consortium Support Activities
 
Project title: Effects of electric and electromagnetic fields on nanoparticle-protein systems
PI: Dr. Niall English Project Code: ndphy010a ICHEC computational scientists: Mr. Gilles Civario
Duration: March 2009 - March 2011
Aim: By working in close collaboration with Dr English, we aim to provide assistance in code installation, script development, job creation and adaptation, data preparation and analysis.
Outcome: Numerous scientific packages have been installed and tested on all the ICHEC machines, and several PRACE ones. Furthermore, for all of them, some sample submission scripts and job settings have been created and/or adapted from existing ones. The PI is now in a position to submit efficiently his jobs on almost whatever machine he has access to.
 
Project title: Minimum Number of Clues Problem
PI: Prof. Gary McGuire Project Code: ndmat003c ICHEC computational scientists: Mr. Gilles Civario
Duration: July 2009 - July 2010
Aim: Evaluating the feasibility of a class A application to solve the underlying hitting set problem, according to the current computing time limits in ICHEC. Parallelising and optimising the code so that it could run efficiently in as many different machines as possible. Defining a proper data representation and storage format for both the inputs and outputs.
Outcome: The code has successfully been parallelised. Furthermore, in close collaboration with its initial developer, it has been optimised and debugged in a way that wouldn't have been possible otherwise. The code has been successfully used on all ICHEC machines and several PRACE ones. Some more work has to be carried out to prepare it ready for a class A application, or for an international grand challenge application.
 
Project title: Parallel algorithms for protein molecular dynamics
PI: Professor Giovanni Ciccotti Project Code: ucd01 ICHEC computational scientists: Mr. Alin Elena
Duration: December 2009 - July 2010
Aim: Together with Dr Grazia Cottone the aim is to explore and revisit parallel algorithms to speed up protein molecular dynamics. The long term goal is to specialise a general molecular dynamics code eg dl_poly3 and/or 4. A set of utilities is to be created and updated.
Outcome: Progress was made in the investigation of adding different constraints to a protein. dl_protein utilities to prepare the input for dl_poly were updated.
 
Project title: Electronic Properties of Oxidised Silicon Nanowires
PI: Dr. Giorgos Fagas Project Code: tiphy012b ICHEC computational scientists: Mr. Alin Elena
Duration: December 2009 - July 2010
Aim: Parallelisation of TIMES, code written by Dr Fagas. The work is to be carried in 3 distinct phases. Phase 1: Data replication parallelism. Phase 2: New algorithms investigations, optimisation, refactoring and documentation. Phase 3: Data parallelism implementation.
Outcome: Phase 1 was completed. Preparing to start phase 2.
 
 
Project title: Development of a large scale ab initio electron transport simulation program
PI: Dr. Ivan Rungger Project Code: tcphy016b ICHEC computational scientist: Mr. Alin Elena
Duration: January 2010 - January 2011
Aim: Working together with Dr Rungger we aim to develop an efficient way of computing the inverse of a large tridiagonal matrix for applications in electronic structure and transport calculations and integrate it with the group’s existing code. This is a continuation of a previous collaboration with the group. We intend to enable the investigation of systems with 100000 orbitals, increasing by a factor of 3 what can currently be achieved.
Outcome: Work in progress.
 
Project title: Seismic source modelling and propagation in complex 3D Earth models
PI: Dr. Gareth O'Brien Project Code: ndear008b ICHEC computational scientist: Mr. Ivan_Girotto
Duration: June 2009 - June 2010
Aim: The aim is to assist research activity that already ICHEC computational resources. The optimisation of the codes now running on Stokes will be first step. Then, the analysis will also focus on the scalability performance to possibly examine a problem which may be suitable for the Blue Gene architecture. Furthermore we will move data to standard formats, improving portability and reducing the effort necessary to access both standard data visualisation tools and the advanced ICHEC data visualisation hardware.

The second objective will be to integrate all the codes now used from the research group in an unique collection of tools. A central code repository will be initially hosted by ICHEC providing shared access to the research group and collaborators.
Outcome: Initial code optimisation has shown good improvement. Code analysis pointed out that enhancement and reorganisation of the code will be beneficial and complementary to a 3D data distribution which is to be implemented. It is foreseen that this data distribution and associated communication pattern should give greater scalability to satisfy next scientific challenge planned by the group.
 
Project title: Performance Analysis of Low-Density Parity-Check Codes Derived from Finite Inversive Spaces
PI: Dr. Marcus Greferath Project Code: ndeng003a ICHEC computational scientist: Mr. Gilles Civario
Duration: September 2008 - Dec 2009
Aim: The aim of this activity is to port the PI's code to ICHEC systems and to develop and/or re-develop some modules of the existing code.
Outcome: New code has been developed and is being used for production runs.
 
Project title: Topological phases in quantum lattice systems
PI: Dr. Jiri Vala Project Code: nmphy002a ICHEC computational scientist: Mr. Gilles Civario
Duration: July 2006 - June 2009
Aim: The aim of this activity is to port their code on ICHEC HPC systems and help them in the development of new features in this code.
Outcome: Completed.
 
Project title: The Tree of One Hundred Percent
PI: Dr. James McInerney Project Code: nmlif004b ICHEC computational scientists: Dr. Simon Wong
Duration: February 2009 - June 2009
Aim: Concaterpillar is a package used for phylogenetic analyses that may be executed in parallel using an MPI library written in the Python language. It had consistently failed to complete when executed on the local compute cluster in Maynooth, as well as on Stokes, owing to obscure memory issues. The problem appeared to be a conflict between the MPI library and thread handling, or issues with the MPI library itself.
Outcome: Concaterpillar (v1.3) was installed on Stokes along with corrective code modifications. The user in question was able to carry out the originally-intended analysis to completion. The results from the analysis have been incorporated into the user’s PhD thesis and they are likely to be included in a journal publication.
 
Project title: Identifying Taxon Restricted Genes in Higher Plants and Epigenomics of genome dosage in Arabidopsis thaliana
PI: Dr. Charlie Spillane Project Code: uclif007b ICHEC computational scientists: Dr. Simon Wong
Duration: : February 2009 - April 2009
Aim: Interproscan is a software package that combines a range of protein domain/motif searching algorithms and databases that is designed to run on compute clusters; by default it achieves parallelism by multiple job submissions to a batch scheduling system. The default set-up, while possible to implement on Stokes, is very inefficient, to the point where sequential runs may be preferable. Hence modification of the code was carried out in order to improve efficiency.
Outcome: Interproscan (v4.4) was installed on Stokes along with code modifications which allows it to carry out different searches in parallel. While there is room for improvements in the speed-up and it is still limited in its scalability, the installed software module on Stokes have enabled the group to carry out domain searches on large datasets (full genomes with ~28,000 queries) which would have otherwise taken weeks/months instead of days. It is expected that the output from the analyses will be incorporated into a publication. The package on Stokes is also easily available to other users, saving the required effort for what is arguably a time-consuming installation process. At least one other user from a different project has taken advantage of the installed package on Stokes.
 
Project title: Identifying Taxon Restricted Genes in Higher Plants
PI: Dr. Charlie Spillane Project Code: uclif006c ICHEC computational scientists: Dr. Simon Wong
Duration: January 2009 - April 2009
Aim: Modification & configuration of Interproscan 4.4 for use on Stokes. Set-up of sequence databases. Assistance with running packages such as mpiBLAST.
Outcome: Completed, work carried forward to uclif007b.
 
Project title: Annotation & analysis of the Panagrolaimus superbus genome.
PI: Prof. Ann Burnell Project Code: nmlif007c ICHEC computational scientists: Dr. Simon Wong
Duration: September 2009 - March 2009
Aim: Gene repertoire comparisons among species (especially those with more than two species) are not straight-forward due to difficulties in determining homologous/orthologous relationships. Following discussions with the user, a scheme based on graphs and BLAST reciprocal best hits was devised to generate comparative gene repertoire results among three nematode species.
Outcome: The new method had resolved some incongruent numbers previously encountered by the user; it avoids unnecessary complexity. The results from the analysis was to be used for a conference presentation.
 
Project title: Coupled Ocean-atmosphere Climate Modelling
PI: Dr. Michael Hartnett Project Code: ngear003b ICHEC computational scientists: Mr. Alastair McKinstry
Duration: April 2009 - April 2011
Aim: Coupled ocean-atmosphere model with improved geochemical processes
Outcome: REMOTE20 model has been OpemMP-optimised, resulting in a 5-fold speedup on Stokes nodes. The ECOMSED chemistry code has been translated to Fortran90 enabling vector-speedup; It is being parallelized with MPI now.
 
Project title: Present and future statistics of extreme European storms
PI: Dr Rodrigo Caballero Project Code: ndear009b ICHEC computational scientists: Mr. Alastair McKinstry
Duration: March 2009 - March 2010
Aim: Ensemble prediction of extreme storms over the next century
Outcome: Work in progress on optimisation and optimal nested scheduling of runs.
 
Project title: Optimisation of WRF community Weather model using GPGPU units
PI: Prof. John Sweeney ICHEC computational scientists: Ms. Nicola McDonnell, Mr. Alastair McKinstry
Duration: September 2009 - September 2010
Aim: Experimental porting and scaling of WRF weather model used by NUI Maynooth to Nvidia GPGPU hardware
Outcome: Initial porting work in completed and benchmarking carried out.
 
Project title: Theory of ferromagnetism in diluted and defective oxides
PI: Prof. Stefano Sanvito Project Code: tcphy004a ICHEC computational scientists: Mr. Alin Elena & Dr. Michael Browne
Duration: January 2009 - July 2009
Aim: Working together with Dr Ivan Rungger we aim to develop an efficient way of computing the inverse of a large tridiagonal matrix for applications in electronic structure and transport calculations and integrate it with the group’s existing code.
Outcome: A serial and parallel OpenMP enabled MKL method using MKL has been implemented. The code is now in production use. Systems up to 10 times larger than before can now be investigated.
 
Project title: Spin and charge polarons and their impact on magnetism and charge transport in oxides
PI: Dr. Charles Patterson Project Code: tcphy008b ICHEC computational scientists: Mr. Alin Elena
Duration: March 2009 - March 2010
Aim: The aim is to refactor and optimize exciton code written by Dr Charles Patterson.
Outcome: This work has been completed.
 
Project title: A finite element analysis of the microscale forces that drive cell-derived tissue formation: a tissue engineering solution
PI: Dr. Noel Harrison Project Code: ngeng012b ICHEC computational scientists: Dr. Michael Browne
Duration: November 2008 - March 2009
Aim: The aim is to assist in optimising memory usage and environment configuration to allow increasingly large jobs to run.
Outcome: Completed. The work enabled the team to complete an important series of simulations which has lead to the following publication: A prediction of cell differentiation and proliferation within a collagen-glycosaminoglycan scaffold subjected to mechanical strain and perfusive fluid flow. By: AJF Stops, KB Heraty, M Browne, FJ O’Brien, PE McHugh. Journal of Biomechanics, November 2009
 
Project title: First Principles Simulations of Shape Memory Alloys Based on NiTi
PI: Dr. Michal Nolan Project Code: tiche010b ICHEC computational scientists: Dr. Michael Browne
Duration: October 2009 - December 2009
Aim: The aim is to use compiler based bounds checking and tools such as Valgrind to uncover a number of bugs in the lastest release of the VASP package. Where possible workaround or fixes will be produced and feedback to the maintainers.
Outcome: A number of overrun bugs have been found that effect parrallel runs, some of which can be masked by differing compiler output.

In addition ICHEC has supported the following users who have been successful in gaining access to European systems via the PRACE prototype programme. Due to the nature of the systems involved the work involves porting codes to new systems and scaling to take advantage of those systems.

PRACE Prototype Access Project Support
 
PI: Dr. Niall English, University College Dublin
PI: Dr. Turlough Downes, Dublin City University
PI: Dr. Damien Thompson, Tyndall National Institute
PI: Prof. James Greer, Tyndall National Institute
PI: Prof. Gary McGuire, University College Dublin