Ada Lovelace Symposium
Abstract
For full details please visit:
http://blogs.bodleian.ox.ac.uk/adalovelace/files/2015/10/Ada-Lovelace-S…
A GPU Implementation of the Filtered Lanczos Procedure
Abstract
This talk describes a graphics processing unit (GPU) implementation of the Filtered Lanczos Procedure for the solution of large, sparse, symmetric eigenvalue problems. The Filtered Lanczos Procedure uses a carefully chosen polynomial spectral transformation to accelerate the convergence of the Lanczos method when computing eigenvalues within a desired interval. This method has proven particularly effective when matrix-vector products can be performed efficiently in parallel. We illustrate, via example, that the Filtered Lanczos Procedure implemented on a GPU can greatly accelerate eigenvalue computations for certain classes of symmetric matrices common in electronic structure calculations and graph theory. Comparisons against previously published CPU results suggest a typical speedup of at least a factor of $10$.
Characterising the integers in the rationals
Abstract
Starting from Hilbert's 10th problem, I will explain how to characterise the set of integers by non-solubility of a set of polynomial equations and discuss related challenges. The methods needed are almost entirely elementary; ingredients from algebraic number theory will be explained as we go along. No knowledge of first-order logic is necessary.
A fast hierarchical direct solver for singular integral equations defined on disjoint boundaries and application to fractal screens
Abstract
Our starting point for specialized linear algebra is an alternative algorithm based on a recursive block LU factorization recently developed by Aminfar, Ambikasaran, and Darve. This algorithm specifically exploits the hierarchically off-diagonal low-rank structure arising from coercive singular integral operators of elliptic partial differential operators. The hierarchical solver involves a pre-computation phase independent of the right-hand side. Once this pre-computation factorizes the operator, the solution of many right-hand sides takes a fraction of the original time. Our fast direct solver allows for the exploration of reduced-basis problems, where the boundary density for any incident plane wave can be represented by a periodic Fourier series whose coefficients are in turn expanded in weighted Chebyshev or ultraspherical bases.
Exploitation of the parareal algorithm in divertor physics simulations
Abstract
Parallelizing the time domain in numerical simulations is non-intuitive, but has been proven to be possible using various algorithms like parareal, PFASST and RIDC. Temporal parallelizations adds an entire new dimension to parallelize and significantly enhances use of super computing resources. Exploiting this technique serves as a big step towards exascale computation.
Starting with relatively simple problems, the parareal algorithm (Lions et al, A ''parareal'' in time discretization of PDE's, 2001) has been successfully applied to various complex simulations in the last few years (Samaddar et al, Parallelization in time of numerical simulations of fully-developed plasma turbulence using the parareal algorithm, 2010). The algorithm involves a predictor-corrector technique.
Numerical studies of the edge of magnetically confined, fusion plasma are an extremely challenging task. The complexity of the physics in this regime is particularly increased due to the presence of neutrals as well as the interaction of the plasma with the wall. These simulations are extremely computationally intensive but are key to rapidly achieving thermonuclear breakeven on ITER-like machines.
The SOLPS code package (Schneider et al, Plasma Edge Physics with B2‐Eirene, 2006) is widely used in the fusion community and has been used to design the ITER divertor. A reduction of the wallclock time for this code has been a long standing goal and recent studies have shown that a computational speed-up greater than 10 is possible for SOLPS (Samaddar et al, Greater than 10x Acceleration of fusion plasma edge simulations using the Parareal algorithm, 2014), which is highly significant for a code with this level of complexity.
In this project, the aim is to explore a variety of cases of relevance to ITER and thus involving more complex physics to study the feasibility of the algorithm. Since the success of the parareal algorithm heavily relies on choosing the optimum coarse solver as a predictor, the project will involve studying various options for this purpose. The tasks will also include performing scaling studies to optimize the use of computing resources yielding maximum possible computational gain.
The monoidal structure on strict polynomial functors and adjoints of the Schur functor
Abstract
Firstly, we will discuss how the category of strict polynomial functors can be endowed with a monoidal structure, including concrete calculations. It is well-known that the above category is equivalent to the category of modules over the Schur algebra. The so-called Schur functor in turn relates the category of modules over the Schur algebra to the category of representations of the symmetric group which posseses a monoidal structure given by the Kronecker product. We show that the Schur functor is monoidal with respect to these structures.
Finally, we consider the right and left adjoints of the Schur functor. We explain how these can be expressed in terms of one another using Kuhn duality and the central role the monoidal structure on strict polynomial functors plays in this context.