HOME  |  NEWS  |  BLOGS  |  MESSAGES  |  FEATURES  |  VIDEOS  |  WEBINARS  |  INDUSTRIES  |  FOCUS ON FUNDAMENTALS
  |  REGISTER  |  LOGIN  |  HELP
Blogs
Guest Blogs

Get Ready for Heterogeneous Computing

NO RATINGS
Page 1 / 2 Next >
View Comments: Newest First|Oldest First|Threaded View
Peter Thorne
User Rank
Blogger
A new angle
Peter Thorne   5/7/2014 1:14:56 PM
NO RATINGS
AnandY, Cabe, yes I agree.  I recently led a roundtable at COFES (the Congress on the Future of Engineering Software) on the subject of software development for embedded systems.  One new insight came from from a participant involved in some non-consumer applications of heterogeneous compute systems. It was that these environments seemed to be pushing software developers back towards low-level coding (assembler/machine coding).  This was not a deliberate strategy, it was just that individuals trying to solve synchronization problems seemed to want to work at this level.  I see this as a natural engineering reaction - something along the lines of "it's a difficult problem, let me get my hands on everything that might be relevant", and, in this case, the lower level software tools require you to look at (and handle) more of the workings of the electronics.  However, I believe that eventually, higher  level tools will help solve these problems.  As teams gain experience, more issues (including synchronization between co-operating processors) will be defined or solved at an architecture level.  If the architecture defines the solution, then a software engineer will be able to use high level language and models, and implement according to the architecture.  If the software implements the architecture, then the lower-level synchronization 'must' be OK (if the architecture is right).  Of course, architecture level solutions tend to be quite general, so someone wanting better performance in a special case may find themselves handling a trade-off of development effort and re-usability against product price performance.  Sorry for long post, there's plenty to do to deliver heterogeneous compute potential!

Cabe Atwell
User Rank
Blogger
Re: Re : Get Ready for Heterogeneous Computing
Cabe Atwell   4/29/2014 11:29:33 PM
NO RATINGS
I agree with you AnandY and we're certainly sure to see heterogeneous computing becoming the standard for mobile devices, especially when APUs are packed with more CPU and GPU cores.

AnandY
User Rank
Gold
Re: Re : Get Ready for Heterogeneous Computing
AnandY   4/29/2014 1:45:05 PM
NO RATINGS
It makes sense to combine GPU and DSP processors alongside CPU processors within a Smartphone or a tablet, but especially the latter, since both mobile devices are graphic intensive. Giving them a dedicated graphics processor alongside an independent CPU gives them more overall processing power as the graphics are offloaded from the CPU making it more efficient. This idea is not so far removed from the main principles of parallel computing, if you really break it down.

AnandY
User Rank
Gold
Re : Get Ready for Heterogeneous Computing
AnandY   4/29/2014 1:42:24 PM
NO RATINGS
Creating any software for use in heterogeneous computing will, at least at this stage, require that the software leans more on CPU processing and less on the GPU. Of course, as mentioned in the article above, the specific extent to which the software will rely on either processor will depend on the specific tasks that will need to be completed using the software. But, given that most of the software that the market needs right now are not very specialized, adopting a CPU-centric solution to heterogeneous computing will give the software much more power and versatility when it comes to the execution of complex tasks.

Partner Zone
More Blogs from Guest Blogs
FPGAs use programmable fabric to create custom logic, but this flexibility comes at a cost -- usually around 10 times more silicon real estate and 10 times the power dissipation. Can we really claim any FPGA is low power?
“How can European standards affect me, especially since I only use machines built in the US?” This is a common question, and one way to answer this is to look at how machine safety is enforced, where the information comes from, and how well you can prove you followed the regulations.
In order to keep in line with safety protocols, industrial networks need to be filtered in a semantic way so that only information related to diagnostics is flowing back to the vendor and that any communications that could be used for remote machine operations are suppressed.
While people may talk about the procurement process, the procurement discipline actually encompasses a number of different processes. They include spend analysis, supplier relationship management, and contract management, just to name a few.
As the Industrial Internet of Things and machine-to-machine communications movements gain speed, some companies are asking themselves, “Wait. How much information do we really want to flow in and out of our premises? Aren’t we supposed to be doubling down on cyber security?”
Design News Webinar Series
5/21/2015 11:00 a.m. California / 2:00 p.m. New York
5/7/2015 11:00 a.m. California / 2:00 p.m. New York
6/3/2015 8:00 a.m. California / 11:00 a.m. New York
6/11/2015 11:00 a.m. California / 2:00 p.m. New York
Quick Poll
The Continuing Education Center offers engineers an entirely new way to get the education they need to formulate next-generation solutions.
Jun 8 - 12, Everything You Ever Wanted to Know about Filters
SEMESTERS: 1  |  2  |  3  |  4  |  5  |  6 |  7


Focus on Fundamentals consists of 45-minute on-line classes that cover a host of technologies. You learn without leaving the comfort of your desk. All classes are taught by subject-matter experts and all are archived. So if you can't attend live, attend at your convenience.
Next Course June 2nd-4th:
Sponsored by Proto Labs
Learn More   |   Login   |   Archived Classes
Twitter Feed
Design News Twitter Feed
Like Us on Facebook

Sponsored Content

Technology Marketplace

Copyright © 2015 UBM Canon, A UBM company, All rights reserved. Privacy Policy | Terms of Service