C++ advocates exceptions as the preferred way to handle unexpected behaviour of an implementation in the code. This does not integrate well with the error handling of MPI, which more or less always results in program termination in case of MPI failures. In particular, a local C++ exception can currently lead to a deadlock due to unfinished communication requests on remote hosts. At the same time, future MPI implementations are expected to include an API to continue computations even after a hard fault (node loss), i.e. the worst possible unexpected behaviour. In this paper we present an approach that adds extended exception propagation support to C++ MPI programs. Our technique allows to propagate local exceptions to remote hosts to avoid deadlocks, and to map MPI failures on remote hosts to local exceptions. A use case of particular interest are asynchronous 'local failure local recovery' resilience approaches. Our prototype implementation uses MPI-3.0 features only. In addition we present a dedicated implementation, which integrates seamlessly with MPI-ULFM, i.e. the most prominent proposal for extending MPI towards fault tolerance. Our implementation is available at https://gitlab.dune-project.org/christi/test-mpi-exceptions.

We have previously suggested mixed precision iterative solvers specifically tailored to the iterative solution of sparse linear equation systems as they typically arise in the finite element discretization of partial differential equations. These schemes have been evaluated for a number of hardware platforms, in particular, single-precision GPUs as accelerators to the general purpose CPU. This paper reevaluates the situation with new mixed precision solvers that run entirely on the GPU: We demonstrate that mixed precision schemes constitute a significant performance gain over native double precision. Moreover, we present a new implementation of cyclic reduction for the parallel solution of tridiagonal systems and employ this scheme as a line relaxation smoother in our GPU-based multigrid solver. With an alternating direction implicit variant of this advanced smoother, we can extend the applicability of the GPU multigrid solvers to very ill-conditioned systems arising from the discretization on anisotropic meshes, that previously had to be solved on the CPU. The resulting mixed-precision schemes are always faster than double precision alone, and outperform tuned CPU solvers consistently by almost an order of magnitude.

Abstract FEAST (Finite Element Analysis and Solutions Tools) is a Finite Element‐based solver toolkit for the simulation of PDE problems on parallel HPC systems, which implements the concept of ‘hardware‐oriented numerics’, a holistic approach aiming at optimal performance for modern numerics. In this paper, we describe this concept and the modular design that enables applications built on top of FEAST to execute efficiently, without any code modifications, on commodity‐based clusters, the NEC SX 8 and GPU‐accelerated clusters. We demonstrate good performance and weak and strong scalability for the prototypical Poisson problem and more challenging applications from solid mechanics and fluid dynamics. Copyright © 2010 John Wiley & Sons, Ltd.

FPGAs are becoming more and more attractive for high precision scientific computations. One of the main problems in efficient resource utilization is the quadratically growing resource usage of multipliers depending on the operand size. Many research efforts have been devoted to the optimization of individual arithmetic and linear algebra operations. In this paper the authors take a higher level approach and seek to reduce the intermediate computational precision on the algorithmic level by optimizing the accuracy towards the final result of an algorithm. In our case this is the accurate solution of partial differential equations (PDEs). Using the Poisson problem as a typical PDE example the authors show that most intermediate operations can be computed with floats or even smaller formats and only very few operations (e.g. 1%) must be performed in double precision to obtain the same accuracy as a full double precision solver. Thus the FPGA can be configured with many parallel float rather than few resource hungry double operations. To achieve this, the authors adapt the general concept of mixed precision iterative refinement methods to FPGAs and develop a fully pipelined version of the conjugate gradient solver. The authors combine this solver with different iterative refinement schemes and precision combinations to obtain resource efficient mappings of the pipelined algorithm core onto the FPGA