Intel® Software Network

Hi there,

My name is Zander Sprague and I am the Americas Geography Manager for Intel Software College.   This my first foray into the forums and I hope not my last.

Although I have a good understanding of Multicore and the technology that Intel makes, I would not say that I am very technical, so I will not be talking about technical aspects of this subject.

What I really wanted to share was my perspective of what Multicore means to me and possibly you. Going forward all the processors that are made will be at least 2 cores if not more. This is exciting to me, there is hope that my computer will be able to do a virus scan and e-mail and gaming all at once. In the past, when my virus checker started I went to get a coffee, because it slammed my system.

I have been a technical instructor for many years. I have had to learn a lot of new programs and then turn around and teach them. Some were very complex and the technology behind it so new that I was helping to define how to teach it.  I see the same thing for University professors today. You are faced with the task of integrating Multicore and Parallel programming into your curriculum. You maybe asking "How do I do that"?

I know how you are feeling; I have been in similar situations. Luckily, there is the Intel Academic Community is here to help you. There is a cornucopia of resources for you, and I want to be your partner in bringing Multicore, Manycores, and parallel programming to your students.

Why not MASTER MULTICORE WITH INTEL?

 

Well said, Zander.

Anybody out there have any experiences with many core programming and want to discuss or compare with others? This would be an ideal thread for that! Don't be shy: post away!

Cheers,

jdg

 

Zander,

Before joining Intel, I worked as a research and teaching assistant. I’ve also given several ISC classes, so I know both worlds. The ISC MC classes focus on what I call “handcrafting”. Threading APIs, debugging tools, libraries capabilities, and OpenMP syntax are not very scientific subjects that are suited for a University class. Honestly speaking, I cannot imagine most of the professors I know to teach the “MC for Academia” class. They try to teach concepts that hopefully survive the next decade. However, a lot of the ISC content is the usage of tools, knowledge that might be outdated with the next release. In fact, the whole thing feels like a long sales pitch for the Intel tools.

Another problem is that professors like to define their own content. It is part of their job to develop a class and their way of forming their students. I therefore suggest not positioning the class as self-contained, but creating building blocks that professors can add to their existing classes. A class about algorithms can be extended to cover parallel algorithms, a class about programming languages can also discuss how parallelism can be expressed, a compiler class can discuss auto-parallelization, and so on. However, you should still be prepared that they would like to modify the content to adopt it to their way of presenting. You might even embrace the bazaar model in order to address this.

Kind regards

Thomas

Hi Thomas,

I'm the person constructing the agenda for version 2 of these workshops, and I agree with you on all points. For a durable academic curriculum, we need to focus on fundamentals, with concurrency infused into the whole range of CS topics.

We'll be posting the outline for comment Real Soon Now (probably next Wed), but the sense of it: one part Concurrency Models (the Berkeley 13 motifs, Parallel Design Patterns), and one part Implementation (language types and extensions (imperative, OO, functional etc) and their suitability/mapping to concurrent algorithms and/or hardware). In this approach, the collection of threading material to which you refer remains useful -- as supplemental content, for one particular implementation.

cheers,
Michael