Schválené projekty 2016

Rozdělení přidělené dotace z MŠMT na specifický vysokoškolský výzkum po fakultách se zohledněním celoškolských pracovišť na rok 2016

Celková přidělená částka z MŠMT na specifický vysokoškolský výzkum na VŠB-TUO - 55 896 914 Kč

Z toho 2.5% - 1 397 423 Kč - úhrada způsobilých nákladů spojených s organizací SGS

fakulta přidělená částka v Kč
FBI  1 270 231
EKF  4 459 400
FAST  2 765 016
FS  9 344 371
FEI 13 781 413
HGF  5 130 549
FMMI  7 000 000
VC 10 748 511
CELKEM 54 499 491

Název projektuPERMON toolbox development II
ŘešitelHorák David doc. Ing., Ph.D.
Školitel projektu
Období řešení projektu01.01.2016 - 31.12.2016
Předmět výzkumuProblems described by partial differential equations (PDEs) have to be
discretized to be solved on computers, e.g. with the Finite Element Method
(FEM). Large sparse symmetric linear systems of equations typically arise.
But variational inequalities lead to more general quadratic programming
problems (QPs). QPs also arise in other disciplines like regression, data fitting, support vector machines, control systems and others.
Domain decomposition methods (DDM) come into play in case of large scale
problems not solvable on usual computers. DDM split an original problem into
smaller subdomain problems that are independent, allowing natural
parallelization and efficient use of supercomputers. FETI methods form a
successful subclass of DDM. They allow highly accurate computations scaling
up to tens of thousands of processors by blending iterative and direct solvers.
Our PERMON toolbox aims to turn the theoretical results in the
aforementioned disciplines into practice with respect to the HPC environment.
During this process, new problems appear. To overcome them, we often have
to use sophisticated technical solutions but often also have to come back to
theory. PERMON consists of our own codes on top of renowned open source
libraries. We focus mostly on engineering applications, but also altruistic ones gradually appear.

1. Team members
Ing. Václav Hapla - founding PERMON developer, programming in C/C++, Matlab, OpenMP, MPI, knowledge of Mercurial, optimization and testing of QP algorithms and FETI methods, author of the concept of QP transforms.
Ing. Radim Sojka - involved in hybrid parallelization of FETI using MPI + OpenMP, PERMON optimization and testing, work for READEX project, skills: programming in C/C++, Matlab, OpenMP, MPI.
Bc. Marek Pecha - specialization to image segmentation, recognition, filtering
Jakub Kružík - programming skills: C/C++, Python

The most valuable publications:
- T. Kozubek, V. Vondrak, M. Mensik , D. Horak, Z. Dostal, V. Hapla, P.
Kabelikova, M. Cermak: Total FETI domain decomposition method and its
massively parallel implementation. Advances in Engineering Software 60–61,
pp. 14–22, 2013.DOI: 10.1016/j.advengsoft.2013.04.001
- V. Hapla, D. Horak, M. Merta: Use of direct solvers in TFETI massively
parallel implementation. Proceedings of PARA 2012, Lecture Notes in
Computer Science 7782, pp. 192–205, 2013. DOI: 10.1007/978-3-642-36803-
- M. Cermak, V. Hapla, D. Horak, M. Merta, A. Markopoulos: Total-FETI domain decomposition method for solution of elasto-plastic problems, Advances in Engineering Software 84, pp 48-54, 2015. DOI: 10.1016/j.advengsoft.2014.12.011 (published)
- M. Merta, A. Vašatová, V. Hapla, D. Horák: Parallel implementation of Total-FETI DDM with application to medical image registration, in proceedings of DD 21, LNCSE 98, pp 917-925, 2014. DOI: 10.1007/978-3-319-05789-7_89 (published)
- V. Hapla, M. Cermak, A. Markopoulos, D. Horak: FLLOP: A Massively Parallel Solver Combining FETI Domain Decomposition Method and Quadratic Programming, in proceedings of HPCC 2014, IEEE Xplore, 2014. DOI: 10.1109/HPCC.2014.56 (published)
- V. Hapla, D. Horak, L. Pospisil, M. Cermak, A. Vasatova: Solving contact problems in mechanics and multibody dynamics with PERMON software toolbox, proceedings of HPCSE 2015, LNCS (accepted)
- Numerical libraries solving large-scale problems developed at IT4Innovations Research Programme Supercomputing for Industry, proceedings of 1st Czech-China Scientific Conference 2015 at VSB-TUO, Perspectives in Science (accepted)
- V. Hapla, D. Horák, A. Markopoulos, M. Čermák, L. Pospíšil, A. Vašátová, R. Sojka, PERMON, a new software toolbox for massively parallel computation of real world problems, proceeding of conference PARENG 2015, (published 2015)
- V. Vondrak, D. Horak, T. Kozubek, Z. Dostal: Highly scalable domain decomposition methods in engineering applications, proceeding of conference PARENG 2015, (published 2015)
- SC15 poster “PERMON” (accepted)
- PASC poster with CSCS: Towards the HPC-inference of causality networks from multiscale economical data (Zurich, Switzerland, published)

Awards in 2014:
- V. Hapla – Joseph Fourier Prize 2014 – First Prize
- V. Hapla – rector's award for the results in doctoral studies 2013/14
- V. Hapla – Department of Applied Mathematics Best Doctoral Student

Awards in 2015:
- V. Hapla – Department of Applied Mathematics Best Doctoral Student
- R. Sojka - 3rd position in Babuska prize in MSc. category
Členové řešitelského týmuIng. Václav Hapla, Ph.D.
doc. Ing. David Horák, Ph.D.
Bc. Lukáš Kresta
Ing. Jakub Kružík
Ing. Marek Pecha
Ing. Ivo Peterek
Ing. Radim Sojka
Specifikace výstupů projektu (cíl projektu)-----------------------------------------------
2. The aim of the project and expected outputs
- Software development: PERMON library
* Writing the documentation having a form of the tutorials and show-cases.
* PermonQP and PermonFLLOP realeses.
* Optimization of hybrid parallelization.
* Optimization of iterative methods for solution of linear
systems and quadratic programming problems with communication hiding and
avoiding techniques especially the application of PIPECG for the coarse problem solution and employing accelerators.
* Interface with Elmer as the preprocessing tool to generate real-world problems.
* Energy consumption optimization of FETI solvers.
* Massively parallel benchmarking and scalability testing of the solvers on
various applications and architectures (Archer, Marenostrum, FERMI, Salomon).
* Starting new topic dealing with knowledge processing.

- Publications and conferences of PERMON team:
* SIAM Conference on Parallel Processing for Scientific Computing in Paris (J. Kruzik)
* Contact Mechanics International Symposium (CMIS 2016, Warsaw) (D. Horak)
* SNA in Prague (R. Sojka, M. Pecha, V. Hapla, J. Kruzik)
* ESCO 2016 in Pilsen (R. Sojka, M. Pecha, V. Hapla, J. Kruzik)
* PANM 18 (R. Sojka, M. Pecha, V. Hapla, J. Kruzik)

3. Impact of the project
QP problems arise in many disciplines like contact mechanics, least-squares
regression, data fitting, data mining, support vector machines, control systems and many others. But it is hard to find maintained freely available QP codes.
This motivates us to carry on PERMON development. Its applications already
include engineering (linear elasticity, contact problems with friction, elastoplasticity, shape optimization) as well as altruistic ones (medical imaging, icesheet melting modelling, and climate changes modelling). Additionally, we employ domain decomposition methods, which allow efficient and robust utilization of parallel computers up to tens of thousands processor cores. By these means we are able to shorten the solution time or even to solve problems whose huge dimension make them unsolvable on conventional personal computers.

4. Goal attainment
Team has reached high level of organizing. Team members start every day by
stand-up meeting at 9:30 having a joint coffee, and meet regularly once per
week on PermonMeeting. The first mathematical ideas and inventions are
implemented and verified in MATLAB, promising features are then
implemented in C/C++ enabling massively parallel implementation. As the
team members participate on other projects (PRACE, EXA2CT, READEX, GACR, etc.), they would like to adopt the best known approaches into PERMON. For project management we use PermonTrac. Trac is an enhanced wiki and issue tracking system for software development projects. The project’s progress is reflected here in milestones with deadlines. Each milestone is fulfilled with particular tickets – tasks, bugfixes and enhancements. We assume this system will help us to progress continuously and simultaneously. See the project objectives above.

Rozpočet projektu - uznané náklady

1. Osobní náklady
Z toho
1.1. Mzdy (včetně pohyblivých složek)0,-0,-
1.2. Odvody pojistného na veřejné zdravotně pojištění a pojistného na sociální zabezpečení a příspěvku na státní politiku zaměstnanosti0,-0,-
2. Stipendia92000,-92000,-
3. Materiálové náklady4800,-0,-
4. Drobný hmotný a nehmotný majetek0,-0,-
5. Služby5000,-0,-
6. Cestovní náhrady80000,-89800,-
7. Doplňkové (režijní) náklady max. do výše 10% poskytnuté podpory20200,-20200,-
8. Konference pořádané VŠB-TUO k prezentaci výsledků studentského grantu (max. do výše 10% poskytnuté podpory)0,-0,-
9. Pořízení investic0,-0,-
Plánované náklady202000,-
Uznané náklady202000,-
Celkem běžné finanční prostředky202000,-202000,-