ITEA2/COLOC-FOI
Reference number | |
Coordinator | TOTALFÖRSVARETS FORSKNINGSINSTITUT, FOI - Avdelningen för informations- och aerosystem, Stockholm |
Funding from Vinnova | SEK 1 677 600 |
Project duration | May 2014 - September 2017 |
Status | Completed |
Venture | Eureka cluster co-funding |
Call | EUREKA medfinansiering kluster 2014-2015 |
Important results from the project
The FOI objectives in COLOC were, 1. extract the massively parallelised domain-decomposing direct linear equation solver from the FOI in-house monolitic Finite Element (FE) code STRIPE used for stress analysis. 2. to modularise the solver (called FOISOL) , i.e. make it stand-alone and devide it into its elementary steps. If feasible extend it for complex equation systems 3. to equip it and provide it with A Programable Interface (API) available for COLOC partners at the project cluster platform NOVA at Bull/ATOS. The stated objectives and aims are fulfilled.
Expected long term effects
In COLOC FOISOL was developed and packaged into an object library equipped with an API which was provided in source code form to the partners giving them possibility to adapt it to their FEA applications. FOISOL has an element-by-element API similar to the MUMPS solver from INRIA, used as reference in COLOC. We showed the advantages with the domain-decomposition approach used in FOISOL in large scale HPC computations. We were able to demonstrate that we can solve huge equation systems with FOISOL with small resources. Thanks to COLOC FOISOL can now be used in other external projects.
Approach and implementation
The FOI activities were conducted within the WP3 which concerned development to basic tools. FOISOL attracted a substantial attention. FOI performed its commitment and work, reached its goals and made its deliverables timely in WP3. Application by partners were scheduled in WP4. These were conducted during the last year of COLOC after that FOI had delivered the FOISOL tool for others to test. In retrospect we see that the by regulations limited financial support alloted for FOI did not cover enough assistance, pushing for others to test FOISOL in the desired extent. A lesson learned.