Splotch is a rendering algorithm for exploration and visual discovery in particlebased datasets coming from astronomical observations or numerical simulations. The strengths of the approach are production of high quality imagery and support for very large-scale datasets through an effective mix of the OpenMP and MPI parallel programming paradigms. This article reports our experiences in re-designing Splotch for exploiting emerging HPC architectures nowadays increasingly populated with GPUs. A performance model is introduced to guide our re-factoring of Splotch. A number of parallelization issues are discussed, in particular relating to race conditions and workload balancing, towards achieving optimal performances. Our implementation was accomplished by using the CUDA programming paradigm. Our strategy is founded on novel schemes achieving optimised data organisation and classification of particles. We deploy a reference cosmological simulation to present performance results on acceleration gains and scalability. We finally outline our vision for future work developments including possibilities for further optimisations and exploitation of hybrid systems and emerging accelerators.

Virtual Observatories (VOs) are online hubs of scientific knowledge. They encompass a collection of platforms dedicated to the storage and dissemination of astronomical data, from simple data archives to e-research platforms offering advanced tools for data exploration and analysis. Whilst the more mature platforms within VOs primarily serve the observational community, there are also services fulfilling a similar role for theoretical data. Scientific visualization can be an effective tool for analysis and exploration of datasets made accessible through web platforms for theoretical data, which often contain spatial dimensions and properties inherently suitable for visualization via e.g. mock imaging in 2d or volume rendering in 3d. We analyze the current state of 3d visualization for big theoretical astronomical datasets through scientific web portals and virtual observatory services. We discuss some of the challenges for interactive 3d visualization and how it can augment the workflow of users in a virtual observatory context. Finally we showcase a lightweight client-server visualization tool for particle-based datasets allowing quantitative visualization via data filtering, highlighting two example use cases within the Theoretical Astrophysical Observatory.

With the increasing size and complexity of data produced by large-scale numerical simulations, it is of primary importance for scientists to be able to exploit all available hardware in heterogenous high-performance computing environments for increased throughput and efficiency. We focus on the porting and optimization of Splotch, a scalable visualization algorithm, to utilize the Xeon Phi, Intel's coprocessor based upon the new many integrated core architecture. We discuss steps taken to offload data to the coprocessor and algorithmic modifications to aid faster processing on the many-core architecture and make use of the uniquely wide vector capabilities of the device, with accompanying performance results using multiple Xeon Phi. Finally we compare performance against results achieved with the Graphics Processing Unit (GPU) based implementation of Splotch.

No abstract

One of the most computationally demanding aspects of the hydrodynamical modelling of Astrophysical phenomena is the transport of energy by radiation or relativistic particles. Physical processes involving energy transport are ubiquitous and of capital importance in many scenarios ranging from planet formation to cosmic structure evolution, including explosive events like core collapse supernova or gamma-ray bursts. Moreover, the ability to model and hence understand these processes has often been limited by the approximations and incompleteness in the treatment of radiation and relativistic particles. The diaphane project has focused in developing a portable and scalable library that handles the transport of radiation and particles (in particular neutrinos) independently of the underlying hydrodynamic code. In this work, we present the computational framework and the functionalities of the first version of the diaphane library, which has been successfully ported to three different smoothed-particle hydrodynamic codes, gadget2, gasoline and sphynx. We also present validation of different modules solving the equations of radiation and neutrino transport using different numerical schemes.