GPU accelerated particle visualization with Splotch

Abstract

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.

An integrated visualization environment for the virtual observatory: Current status and future directions

“…An optimization would be to split and pipeline the rendering into multiple channels. The Splotch rendering algorithm for exploration and visual discovery included into the VisIVO packages has been already redesigned for exploiting HPC architectures by using the CUDA programming paradigm (Rivi et al, 2014). This approach is currently fully supported by VTK, so it should be easy to incorporate it in the proposed framework.…”

“…The VisIVO Filters are a collection of data processing modules to assist the extraction of hidden properties of the dataset. Finally, the VisIVO Viewer is the visualization core component based on the Visualization ToolKit(VTK) and Splotch (Dolag et al, 2008), a customized volume ray casting algorithm recently redesigned in view of the CUDA paradigm 10 to fully exploit modern HPC infrastructures (Rivi et al, 2014). VisIVO Viewer can also be used to produce images from a given set of camera positions that can be assembled externally to generate movies.…”

Interactive 3D visualization for theoretical virtual observatories

“…Splotch is an open source scientific visualization tool designed for high performance computing environments. Written in C++, with minimal external dependencies, Splotch is implemented with a variety of parallel models including hybrid OpenMP and MPI for x86 64 environments such as Xeon and Xeon Phi CPUs, as well as CUDA (NVIDIA Corporation 2007) for GPU based systems (Rivi et al 2014) (Jin et al 2010. The lack of dependencies and inclusion of multiple parallel models means Splotch can be built to exploit almost any type of system, from a standalone desktop system with or without a GPU to large heterogeneous HPC architectures.…”

Splotch

“…Finally when all chunks of data have been processed and accumulated, the resultant device image is copied back to the host for output. The original OpenMP rendering process described in Rivi et al’s (2014) study has not been conceptually modified, rather the implementation has been optimized for MIC.…”

“…Splotch (Dolag et al, 2008) is an algorithm for visualizing large particle-based data sets, providing high quality imagery while exploiting a broad variety of HPC systems such as multi-core processors, multi-node supercomputing systems (Jin et al, 2010), and also GPUs (Rivi et al, 2014). The variety of implementations is due to the fact that many HPC systems of today are exploiting not only standard CPUs but accelerators to achieve maximum computational power with low energy usage.…”