We help our customers get faster, more responsive and/or more precise software. But what does that mean? What is it what we do here in Amsterdam?
Unfortunately we cannot share what we have done for most of our customers the past 6 years, as we are their secret weapon to make them more competitive. Below is an overview of blog posts discussing our demos, open source software and even some work for customers – most of the stories are not published yet, but then you now what’s coming!
- Porting Gromacs from CUDA to OpenCL. Till we ported the simulation software end of 2014, it has been CUDA-only. Porting took several man-months to manually port all code. You can now download the source, build it and run it on AMD/Intel hardware – see here for more info. All is open source, so you can see our code.
- Porting Manchester’s UNIFAC to OpenCL@XeonPhi. Even though XeonPhi Knights Corner is not a very performant accelerator, we managed to get a 160x speedup from single threaded code. Most of the speedup is due to clever code-optimisations.
- Android video filter demo [to be published]. Real-time Android-app, where the webcam stream has several OpenGL filters applied to make the it look like an old movie.
- Speeding up Excel [to be published]. A heavy financial algorithm is offloaded to a GPU, resulting in a big speedup.
- Flooding simulation. Software that simulates flooding of land, which we ported to multi-GPU on OpenCL and got a 35x speedup over MPI.
- Cartoonizer [to be published]. The webcam or video stream is made as if it was a cartoon using several image filters.
- Boost.compute port from OpenCL to HSA [to be published]. Using cloc, we ported Boost.compute to AMD HSA.
- Real-time network packet processing using OpenCL-on-FPGAs [to be published].
We work together with hardware partners to show hardware-optimised demos, so many of the demos are visual. Actually most work is done on speeding up simulations, but such interesting work has no visual output.