Kofler, K.; Grasso, I.; Cosenza, B.; Fahringer, T.: An automatic input-sensitive approach for heterogeneous task partitioning. ABSTRACTUnleashing the full potential of heterogeneous systems, consisting of multi-core CPUs and GPUs, is a challenging task due to the difference in processing capabilities, memory availability, and communication latencies of different computational resources.In this paper we propose a novel approach that automatically optimizes task partitioning for different (input) problem sizes and different heterogeneous multi-core architectures. We use the Insieme source-to-source compiler to translate a single-device OpenCL program into a multi-device OpenCL program. The Insieme Runtime System then performs dynamic task partitioning based on an offline-generated prediction model. In order to derive the prediction model, we use a machine learning approach based on Artificial Neural Networks (ANN) that incorporates static program features as well as dynamic, input sensitive features. Principal component analysis have been used to further improve the task partitioning. Our approach has been evaluated over a suite of 23 programs and respectively achieves a performance improvement of 22% and 25% compared to an execution of the benchmarks on a single CPU and a single GPU which is equal to 87.5% of the optimal performance.

No abstract

Large-scale compute clusters of heterogeneous nodes equipped with multi-core CPUs and GPUs are getting increasingly popular in the scientific community. However, such systems require a combination of different programming paradigms making application development very challenging.In this article we introduce libWater, a library-based extension of the OpenCL programming model that simplifies the development of heterogeneous distributed applications. libWater consists of a simple interface, which is a transparent abstraction of the underlying distributed architecture, offering advanced features such as inter-context and inter-node device synchronization. It provides a runtime system which tracks dependency information enforced by event synchronization to dynamically build a DAG of commands, on which we automatically apply two optimizations: collective communication pattern detection and device-host-device copy removal.We assess libWater’s performance in three compute clusters available from the Vienna Scientific Cluster, the Barcelona Supercomputing Center and the University of Innsbruck, demonstrating improved performance and scaling with different test applications and configurations.

N-body simulations represent an important class of numerical simulations in order to study a wide range of physical phenomena for which researchers demand fast and accurate implementations. Due to the computational complexity, simple brute-force methods to solve the longdistance interaction between bodies can only be used for smallscale simulations. Smarter approaches utilize neighbor lists, tree methods or other hierarchical data structures to reduce the complexity of the force calculations. However, such data structures have complex building algorithms which hamper their parallelization for GPUs.In this paper, we introduce a novel method to effectively parallelize N-body simulations for GPU architectures. Our method is based on an efficient, three-phase, parallel Kd-tree building algorithm and a novel volume-mass heuristic to reduce the simulation time and increase accuracy.Experiments demonstrate that our approach is the fastest monopole implementation with an accuracy that is comparable with state of the art implementations (GADGET-2). In particular, we are able to reach a simulation speed of up to 3 Mparticles/s on a single GPU for the force calculation, while still having a relative force error below 0.4% for 99% of the particles. We also show competitive performance with existing GPU implementations, while our competitor shows worse accuracy behavior as well as a higher energy error during time integration.

Abstract. Extreme Data is an incarnation of Big Data concept distinguished by the massive amounts of data that must be queried, communicated and analyzed in near real-time by using a very large number of memory or storage elements and exascale computing systems. Immediate examples are the scientific data produced at a rate of hundreds of gigabits-per-second that must be stored, filtered and analyzed, the millions of images per day that must be analyzed in parallel, the one billion of social data posts queried in real-time on an in-memory components database. Traditional disks or commercial storage nowadays cannot handle the extreme scale of such application data.Following the need of improvement of current concepts and technologies, we focus in this paper on the needs of data intensive applications running on systems composed of up to millions of computing elements (exascale systems). We propose in this paper a methodology to advance the state-of-the-art. The starting point is the definition of new programming paradigms, APIs, runtime tools and methodologies for expressing data-intensive tasks on exascale systems. This will pave the way for the exploitation of massive parallelism over a simplified model of the system architecture, thus promoting high performance and efficiency, offering powerful operations and mechanisms for processing extreme data sources at high speed and/or real time.