LEAP is all about sharing the latest developments in high-performance parallel computing technology on low-power multi-core CPU, GPU and FPGA processors. LEAP kicked-off with a Conference in May 2013, archived here, and continues as an industry news feed & blog, to be followed with the next event scheduled for the Autumn of 2014.
Andreas over at Parallella recently provides his Kickstarter supporters with an update on his blog of the recent challenges faced in getting the first Epiphany boards out, how they have overcome them and the plan going forward, he comments; “Not sure where to start… The last couple of months have been a disaster and I am disappointed to report that we are still weeks away from being able to ship the final Parallella products. What went wrong? Well, pretty much everything… Continue reading
Gaston Hillar over at Dr.Dobbs has published the first of a series of articles to help PyOpenCL developers create OpenCL host applications. PyOpenCL is an open-source package (MIT license) that enables developers to easily access the OpenCL API from Python. The latest stable version of PyOpenCL provides features that make it one of the handiest OpenCL wrappers for Python enabling the creation of OpenCL kernels without leaving the Python environment. In this first article of a two-part series on PyOpenCL, Gaston explains how Continue reading
Altera Corporation has announced that its SDK for OpenCL is conformant to the OpenCL 1.0 standard and is now included on the Khronos Group list of OpenCL conformant products. Altera’s FPGA-optimized OpenCL solution allows software developers to harness the massively parallel architecture of their FPGA’s for system acceleration. Altera will be demonstrating its OpenCL solutions at Continue reading
A recent article in the Journal of Parallel Computing looks at the challenges faced in creating a robust CUDA-to-OpenCL translator, presents an instantiation of a CUDA-to-OpenCL (CU2CL) source-to-source translator, and evaluates its efficacy on real CUDA codes. The researchers conclude that although it is not straightforward and (currently) subject to some important limitations, robust automatic source translation from CUDA to OpenCL is largely achievable. Further they Continue reading
Futuremark have released an update to their PCMark 8 Professional Edition benchmark (8 v1.1.111) that introduces new options for testing OpenCL as well as enabling Windows 8.1 compatibility and adding detailed hardware monitoring graphs. The Home and Creative benchmarks now offer a choice of run modes. Conventional mode does not use OpenCL acceleration, reflecting how the majority of software works today. Accelerated mode uses OpenCL acceleration, anticipating how software may work in the future.
PCMark 8 will automatically find and use the optimal OpenCL device in your system. For troubleshooting or comparison testing, Continue reading
Rahul Garg’s recent article over at the AnandTech site gives an insight into Altera’s OpenCL SDK for harnessing their FPGAs. In summary he concludes that “Altera’s products effectively promise customized hardware for your OpenCL Kernels and the claim is that FPGAs will be more efficient than CPUs or GPUs at many tasks. Applications that are not necessarily floating-point heavy, for example applications relying on custom integer datatypes, heavy bit-manipulation Continue reading
Join Intel for a webinar series that is designed to help you get the most performance out of your OpenCL applications on 4th generation Intel Core processors with Intel Graphics. The first webinar “Taking Advantage of Intel Graphics with OpenCL” will start with an introduction to the new Intel Iris Graphics family and how to maximize performance by using both the CPU and the Intel GPU. The other two webinars should help you optimize your application using the robust OpenCL tools portfolio and then analyze it using Intel VTune Amplifier XE.
The webinars are on 6th Nov, 20th Nov and 4th Dec. Head over to the Intel site for more details and to register.
Dave Smith over at DoubleEncore Inc. notes that the current integration of the new Renderscript Support Library into the build tools is a bit of a work in progress, with no official support yet for it inside of Android Studio or Gradle. The integration with Eclipse also requires a little bit of additional work. The only workflow that functions “out of the box” at the moment is Continue reading
As Adapteva prepare to launch the latest Parallella board, they felt it was only fair to the new hardware for it to receive an identity comparable to that of other famous and well known products like the Linux penguin, the Nike Swoosh or the Open Source Hardware gear-logo. “At Adapteva, parallel processing is the core of our existence and so it was natural to use the purest parallel processor in existence – the Brain – as the Parallella icon of choice” Behold the new logo and read the full story.
In Part 2 of his blog on “neural network design” for the Adapteva Parallella architecture, Nick Oppen looks at Training, or the hard bit as he puts it. Over the two blogs Nick breaks down the process of getting the best performance out of a parallel architecture for processing neural networks, and specifically Feed Forward – Back Propagation networks. This, he claims is not a rigorous academic work, but an experiment and he welcomes feedback. So if you find a better way drop him a line. Read the both parts of the blog at: http://nicksparallellaideas.blogspot.co.uk/ Continue reading