Areas of Expertise

Successful HPC solutions by definition are the product of a number of technologies combined together to create an integrated, highly functioning single platform. Acceleware has deep experience that focuses on several key areas:

  • Accelerating the ‘Hard’ Parallel Problems
    • Some applications are relatively easy to run in parallel. These embarrassingly parallel problems do not require particular expertise and are able to run simply by scheduling the jobs out to different processors. However problems that do not lend themselves to simplistic parallelization are faced with a crisis as the industry moves to a multi-core world. These ‘Hard” parallel problems are where Acceleware’s unique experience in designing successful solutions for algorithms and/or applications provide tremendous value to your project.
  • Accelerating the ‘Whole’ Problem
    • Acceleration examples of 100x or 200x speed-ups are commonly publicized.  However these examples generally only focus on a narrow part of the entire problem and do not take into account issues like compute intensive initialization, pre/post-processing or I/O bottlenecks. Acceleware provides acceleration value to the entire solution stack because what really matters is how long the entire problem takes to solve, not just one part of it!
  • Programming Languages and 3rd Party Tool Experience
    • CUDA/OpenCL – Acceleware launched the industries first commercial CUDA solution and we have subsequently ported and built several CUDA products that run in production environments with Fortune 500 companies.
    • OpenGL – Acceleware delivered the first general purpose commercial GPU acceleration product on OpenGL and continues to lead the way.
    • C++, Fortran
    • MPI, OpenMP, Boost (pThreads), Intel MKL
    • General scientific and HPC programming environments
  • Hardware
    • Acceleware has been assisting its global customers adopt GPU computing technology since its infancy. Our System Engineers specialize in integrating HPC hardware spanning a wide range of servers, GPUs, networking equipment, storage and clusters from many vendors.
  • Scalability
    • Our software libraries run on 16 node (equals 64 GPUs with 15,360 processing cores) clusters of GPUs with InfiniBand. If your problem is large, look to Acceleware as the solution in creating scalable solutions to meet any need.