Vector Fabrics home page Vector Fabrics Blog RSSView Vector Fabrics' Twitter account View Vector Fabrics' Facebook profile View Vector Farbics' LinkedIn profile Google+

Sharing insights in multicore programming

Multicore programming talks

October was a busy month. Vector Fabrics employees spoke at several conferences and we published an article in Bits & Chips on the complexities of parallelization. We thought it might be a good idea to recap and publish the slides and a link to the article here.

Parallelization is not child’s play

Article: Bits&Chips magazine

The article was published in the leading Dutch embedded magazine Bits & Chips. The paper Parallelization is not child’s play itself is in Dutch, but there is an automatic Google Translation link below. The paper discusses why software parallelization is hard. Amongst others it briefly touches on vectorization, problems people encounter during parallelization like dependencies and hidden state and parallelization helpers like OpenCL, OpenMP, C++11 and Intel TBB.

The original Dutch article by Klaas van Gend.

For your convenience, an automatic translation by Google.

Multithreading in C++11

Talk: T-DOSE 2012, Eindhoven, The Netherlands

On October 28th, Klaas presented a talk about the multithreading capabilities that were added in the new ISO standard of C++: C++11. This presentation is also a part of our training course.

A PDF of slides is available.

Thanks to Eindhoven City TV, it was also recorded and available as a YouTube video.

Application Parallelization for multi-core Android devices

Talk: Embedded Linux Conference Europe 2012, Barcelona, Spain

As a co-organizer of the Embedded Linux Conference Europe, Klaas was present at LinuxCon/ELCE in Barcelona from November 5th to November 7th.

He also presented a talk: Application Parallelization for multi-core Android devices. For your enjoyment, a PDF of the slides is (like all slides from the various ELC conferences) available on the elinux wiki.

Note that this session was recorded, but the video has not been published yet.

Optimizing C/C++ code for multicore systems

Talk: NLUUG nj2012 Talen & Tools, Ede, The Netherlands

The day after ELC-Europe, Klaas did a talk in Ede at the NLUUG conference on programming languages and software tools.

He talked about the various mechanisms available to optimize algorithms, including parallelization, vectorization and various tools and libraries. Klaas presented a taxonomy of several libraries, language extensions and tools available on the market, including OpenCL, FFTW, C++11, Intel TBB and others (see image). The presentation ended with an example project: the parallelization of an MP3-like algorithm.

Again, we have made a PDF of the slides available.

Posted in category: Market & Skills on Wednesday, November 28, 2012 - 13:53


Add comment

Notify me of follow-up comments?

Please enter the word you see in the image below: