high performance computing on graphics processing units: hgpu.org

Modern applications including graphics, multimedia, web search, and data analytics not only can benefit from acceleration, but also exhibit significant degrees of tolerance to imprecise computation. This amenability to approximation provides an opportunity to trade quality of the results for higher performance and better resource utilization. Exploiting this opportunity is particularly important for FPGA accelerators […]

There are numerous way to express parallelism which can make it challenging for developers to verify these programs. Many tools only target a single dialect but the Concurrency Intermediate Verification Language (CIVL) targets MPI, Pthreads, and CUDA. CIVL provides a general concurrency model that can represent pro- grams in a variety of concurrency dialects. CIVL […]

High performance graph analytics are critical for a long list of application domains. In recent years, the rapid advancement of many-core processors, in particular graphical processing units (GPUs), has sparked a broad interest in developing high performance parallel graph programs on these architectures. However, the SIMT architecture used in GPUs places particular constraints on both […]

The paper describes an approach to the parallel natural join execution on computing clusters with GPU and MIC Coprocessors. This approach is based on a decomposition of natural join relational operator using the column indices and domain-interval fragmentation. This decomposition admits parallel executing the resource-intensive relational operators without data transfers. All column index fragments are […]

We show that an end-to-end deep learning approach can be used to recognize either English or Mandarin Chinese speech–two vastly different languages. Because it replaces entire pipelines of hand-engineered components with neural networks, end-to-end learning allows us to handle a diverse variety of speech including noisy environments, accents and different languages. Key to our approach […]

Using the histogram procedure, this work studies performance determining factors in computing in parallel on SIMD and SIMT devices. Modern graphics pro-cessing units (GPUs) support SIMT, multiple threads running the same instruction, whereas central processing units (CPUs) use SIMD, in which one instruction op-erates on multiple operands. As part of this work, a cross-technology framework […]

Classic path finding algorithms are not adequate in real world path planning, where environment information is incomplete or dynamic and Markov Decision Processes have been used as an alternative. The problem with the MDP formalism is that its state space grows exponentially with the number of domain variables, and its inference methods grow with the […]

We propose a tool-flow methodology that can be applied to analyze and track the performance of OpenCL applications on heterogeneous platforms. Using a case study on a datacenter representative workload, we evaluate our tool flow on three distinct heterogeneous platforms and demonstrate how it can be employed more widely to provide insight and track attainable […]

MXNet is a multi-language machine learning (ML) library to ease the development of ML algorithms, especially for deep neural networks. Embedded in the host language, it blends declarative symbolic expression with imperative tensor computation. It offers auto differentiation to derive gradients. MXNet is computation and memory efficient and runs on various heterogeneous systems, ranging from […]

We present an idea of efficient query answering over an RDF dataset employing a consumer-grade graphic card for an efficient computation. We consider tree-shaped SPARQL queries and static datasets, to facilitate data mining over RDF graphs in warehouse-like setups. Reasons to see the poster: a) presentation of the approach with examples; b) possibility of discussion […]

Deep learning methods are proven to be state-of-theart in addressing many challenges in machine learning domains. However, it comes at the cost of high computational requirements and energy consumption. The emergence of Processing In Memory (PIM) with diestacking technology presents an opportunity to speed up deep learning computation and reduce energy consumption by providing low-cost […]

A graphics processing unit (GPU) parallelization approach was implemented to improve the efficiency of nonlinear dynamic analysis. The GPU parallelization approach speeded up the computation of implicit time integration and reduced total calculation time. In addition, a parallel equations solver is introduced to solve the equation system. Numerical examples of reinforced concrete (RC) frames were […]