We examine the performance profile of Convolutional Neural Network training on the current generation of NVIDIA Graphics Processing Units. We introduce two new Fast Fourier Transform convolution implementations: one based on NVIDIA’s cuFFT library, and another based on a Facebook authored FFT implementation, fbfft, that provides significant speedups over cuFFT (over 1.5x) for whole CNNs. […]

December 26, 2014 by hgpu

The worth of finding a general solution for nonsymmetric eigenvalue problems is specified in many areas of engineering and science computations, such as reducing noise to have a quiet ride in automotive industrial engineering or calculating the natural frequency of a bridge in civil engineering. The main objective of this thesis is to design a […]

December 14, 2014 by hgpu

It is well recognized in the computer algebra theory and systems communities that the Fast Fourier Transform (FFT) can be used for multiplying polynomials. Theory predicts that it is fast for "large enough" polynomials. The basic idea is to use fast polynomial multiplication to perform fast integer multiplication. We can achieve really fast FFT multiplication […]

December 7, 2014 by hgpu

We have developed a software-based polarization spectrometer, PolariS, to acquire full-Stokes spectra with a very high spectral resolution of 61 Hz. The primary aim of PolariS is to measure the magnetic fields in dense star-forming cores by detecting the Zeeman splitting of molecular emission lines. The spectrometer consists of a commercially available digital sampler and […]

December 5, 2014 by hgpu

Effective utilization of GPU processing capacity for scientific workloads is often limited by memory throughput and PCIe communication transfer times. This is particularly true for semi-analytic Fourier-domain computations in earthquake modeling (Relax) where operations on large-scale 3D data structures can require moving large volumes of data from storage to the compute in predictable but orthogonal […]

November 9, 2014 by hgpu

The Graphics Processing Unit (GPU) has become an integral part of astronomical instrumentation, enabling high-performance online data reduction and accelerated online signal processing. In this paper, we describe a wide-band reconfigurable spectrometer built using an off-the-shelf GPU card. This spectrometer, when configured as a polyphase filter bank (PFB), supports a dual-polarization bandwidth of up to […]

November 5, 2014 by hgpu

Fast Poisson solvers using the Fast Fourier Transform on uniform grids are especially suited for parallel implementation, making them appropriate for portability on graphical processing unit (GPU) devices. The goal of the following work was to implement, test, and evaluate a fast Poisson solver for periodic boundary conditions for use on a variety of GPU […]

November 3, 2014 by hgpu

For embarrassingly parallel algorithms, a Graphics Processing Unit (GPU) outperforms a traditional CPU on price-per-flop and price-per-watt by at least one order of magnitude. This had led to the mapping of signal and image processing algorithms, and consequently their applications, to run entirely on GPUs. This paper presents CUFFTSHIFT, a ready-to-use GPU-accelerated library, that implements […]

October 24, 2014 by hgpu

An emerging trend in processor architecture seems to indicate the doubling of the number of cores per chip every two years with same or decreased clock speed. Of particular interest to this thesis is the class of many-core processors, which are becoming more attractive due to their high performance, low cost, and low power consumption. […]

October 22, 2014 by hgpu

Molecular dynamics simulations allow us to study the behavior of complex biomolecular systems by modeling the pairwise interaction forces between all atoms. Molecular systems are subject to slowly decaying electrostatic potentials, which turn molecular dynamics into an n-body problem. In this paper, we present a parallel and scalable solution to compute long-range molecular forces, based […]

September 15, 2014 by hgpu

To meet the strict requirements of acquisition time under conditions of high dynamic and low CNR, an acquisition method of spread spectrum signals based on GPU acceleration has put forward through the combination of the spread spectrum signal acquisition and GPU parallel computing. Taking the frequency domain parallel acquisition algorithm based on FFT as the […]

September 5, 2014 by hgpu

Intel Xeon Phi coprocessor-based clusters offer high compute and memory performance for parallel workloads and also support direct network access. Many real world applications are significantly impacted by network characteristics and to maximize the performance of such applications on these clusters, it is particularly important to effectively saturate network bandwidth and/or hide communications latency. We […]

August 15, 2014 by hgpu