Sverker Holmgren
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Sverker Holmgren has a total of 16 co-authors. Here are the top co-authors by number of shared publications:
- Anders Goude
Division of Electricity Department of Engineering Sciences Uppsala University SE-751 21 Uppsala Sweden
1 shared publication - Behrang Mahjani
Division of Scientific Computing Department of Information Technology Uppsala University Uppsala Sweden
1 shared publication - Carl Nettelblad
Division of Scientific Computing, Department of Information Technology, Science for Life Laboratory, Uppsala University , Uppsala SE-752 37
1 shared publication - Dan Wallin
Uppsala University
1 shared publication Henrik Brandén
Division of Scientific Computing, Department of Information Technology, Uppsala University, Uppsala, Sweden
1 shared publication- Jarmo Rantakokko
Uppsala University
1 shared publication
Latest Publications
- Open Access
Dynamic autotuning of adaptive fast multipole methods on hybrid multicore CPU & GPU systems
We discuss an implementation of adaptive fast multipole methods targeting hybrid multicore CPU- and GPU-systems. From previous experiences with the computational profile of our version of the fast multipole algorithm, suitable parts are off-loaded to the GPU, while the remaining parts are threaded and executed concurrently by the CPU. The parameters defining the algorithm affects the performance and by measuring this effect we are able to dynamically balance the algorithm towards optimal performance. Our setup uses the dynamic nature of the computations and is therefore of general character.
Open AccessDynamic autotuning of adaptive fast multipole methods on hybrid multicore CPU & GPU systems
We discuss an implementation of adaptive fast multipole methods targeting hybrid multicore CPU- and GPU-systems. From previous experiences with the computational profile of our version of the fast multipole algorithm, suitable parts are off-loaded to the GPU, while the remaining parts are threaded and executed concurrently by the CPU. The parameters defining the algorithm affects the performance and by measuring this effect we are able to dynamically balance the algorithm towards optimal performance. Our setup uses the dynamic nature of the computations and is therefore of general character.
- Open Access
Fast and Accurate Detection of Multiple QTL
We present a new computational scheme that enables efficient and reliable Quantitative Trait Loci (QTL) scans for experimental populations. Using a standard brute-force exhaustive search effectively prohibits accurate QTL scans involving more than two loci to be performed in practice, at least if permutation testing is used to determine significance. Some more elaborate global optimization approaches, e.g. DIRECT, have earlier been adopted to QTL search problems. Dramatic speedups have been reported for high-dimensional scans. However, since a heuristic termination criterion must be used in these types of algorithms the accuracy of the optimization process cannot be guaranteed. Indeed, earlier results show that a small bias in the significance thresholds is sometimes introduced. Our new optimization scheme, PruneDIRECT, is based on an analysis leading to a computable (Lipschitz) bound on the slope of a transformed objective function. The bound is derived for both infinite size and finite size populations. Introducing a Lipschitz bound in DIRECT leads to an algorithm related to classical Lipschitz optimization. Regions in the search space can be permanently excluded (pruned) during the optimization process. Heuristic termination criteria can thus be avoided. Hence, PruneDIRECT has a well-defined error bound and can in practice be guaranteed to be equivalent to a corresponding exhaustive search. We present simulation results that show that for simultaneous mapping of three QTL using permutation testing, PruneDIRECT is typically more than 50 times faster than exhaustive search. The speedup is higher for stronger QTL. This could be used to quickly detect strong candidate eQTL networks.
Open AccessFast and Accurate Detection of Multiple QTL
We present a new computational scheme that enables efficient and reliable Quantitative Trait Loci (QTL) scans for experimental populations. Using a standard brute-force exhaustive search effectively prohibits accurate QTL scans involving more than two loci to be performed in practice, at least if permutation testing is used to determine significance. Some more elaborate global optimization approaches, e.g. DIRECT, have earlier been adopted to QTL search problems. Dramatic speedups have been reported for high-dimensional scans. However, since a heuristic termination criterion must be used in these types of algorithms the accuracy of the optimization process cannot be guaranteed. Indeed, earlier results show that a small bias in the significance thresholds is sometimes introduced. Our new optimization scheme, PruneDIRECT, is based on an analysis leading to a computable (Lipschitz) bound on the slope of a transformed objective function. The bound is derived for both infinite size and finite size populations. Introducing a Lipschitz bound in DIRECT leads to an algorithm related to classical Lipschitz optimization. Regions in the search space can be permanently excluded (pruned) during the optimization process. Heuristic termination criteria can thus be avoided. Hence, PruneDIRECT has a well-defined error bound and can in practice be guaranteed to be equivalent to a corresponding exhaustive search. We present simulation results that show that for simultaneous mapping of three QTL using permutation testing, PruneDIRECT is typically more than 50 times faster than exhaustive search. The speedup is higher for stronger QTL. This could be used to quickly detect strong candidate eQTL networks.
- 30 July 2010
- Published by Springer Nature
- p. 417-424
An Implementation Framework for Solving High-Dimensional PDEs on Massively Parallel Computers
- 30 July 2010
- Published by Springer Nature
- p. 417-424
An Implementation Framework for Solving High-Dimensional PDEs on Massively Parallel Computers
- 28 February 2010
- Published by Elsevier in Computational Biology and Chemistry
- Vol. 34 (1) , 34-41
Assessing a multiple QTL search using the variance component model
- 28 February 2010
- Published by Elsevier in Computational Biology and Chemistry
- Vol. 34 (1) , 34-41
Assessing a multiple QTL search using the variance component model
- 1 July 2006
- Published by Springer Nature in Mediterranean Journal of Mathematics
- Vol. 3 (2) , 227-249
Can One Hear the Composition of a Drum?
- 1 July 2006
- Published by Springer Nature in Mediterranean Journal of Mathematics
- Vol. 3 (2) , 227-249
Can One Hear the Composition of a Drum?
- 2 December 2005
- Published by Springer Nature in Journal of Scientific Computing
- Vol. 30 (1) , 35-60
Discrete Fundamental Solution Preconditioning for Hyperbolic Systems of PDE
- 2 December 2005
- Published by Springer Nature in Journal of Scientific Computing
- Vol. 30 (1) , 35-60
Discrete Fundamental Solution Preconditioning for Hyperbolic Systems of PDE
- 1 April 2002
- Published by Taylor & Francis in Parallel Algorithms and Applications
- Vol. 17 (4) , 285-299
Performance of PDE solvers on a self-optimizing NUMA architecture
The performance of shared-memory (OpenMP) implementa- tions of three dierent PDE solver kernels representing finite dierence methods, finite volume methods, and spectral methods has been investi- gated. The experiments have been performed on a self-optimizing NUMA system, the Sun Orange prototype, using dierent data placement and thread scheduling strategies. The results show that correct data place- ment is very important for the performance for all solvers. However, the Orange system has a unique capababilty of automatically changing the data distribution at run time through both migration and replication of data. For reasonable large PDE problems, we find that the time to do this is negligible compared to the total solve time. Also, the performance after the migration and replication process has reached steady-state is the same as what is achieved if data is optimally placed at the begin- ning of the execution using hand tuning. This shows that, for the ap- plication studied, the self-optimizing features are successful, and shared memory code without explicit data distribution directives yields good performance.
- 1 April 2002
- Published by Taylor & Francis in Parallel Algorithms and Applications
- Vol. 17 (4) , 285-299
Performance of PDE solvers on a self-optimizing NUMA architecture
The performance of shared-memory (OpenMP) implementa- tions of three dierent PDE solver kernels representing finite dierence methods, finite volume methods, and spectral methods has been investi- gated. The experiments have been performed on a self-optimizing NUMA system, the Sun Orange prototype, using dierent data placement and thread scheduling strategies. The results show that correct data place- ment is very important for the performance for all solvers. However, the Orange system has a unique capababilty of automatically changing the data distribution at run time through both migration and replication of data. For reasonable large PDE problems, we find that the time to do this is negligible compared to the total solve time. Also, the performance after the migration and replication process has reached steady-state is the same as what is achieved if data is optimally placed at the begin- ning of the execution using hand tuning. This shows that, for the ap- plication studied, the self-optimizing features are successful, and shared memory code without explicit data distribution directives yields good performance.