Louis Pasteur's theory of germs is ridiculous fiction - Pierre Pachet
|
|
|
Microsoft and Intel finally team up on multi-core programming
WALL STREET JOURNAL reader Mike Magee says Microsoft and Intel are expected to announce a joint research project to engage in programming for multicore processors.
The research and development of the multicore programming project will purportedly be carried out by boffins from the University of California, Berkeley, where the Vole and Intel are funding a laboratory for research into parallel computing.
The lab would let researchers imitate 1000 core systems and could cost a cool $2 million a year for the next five years, the Journal learned today and the rest of us will learn later on.
The materialisation of processors with two or more cores has led to many calls for more research into bettering a device's processing ability using less power. The research hopes to lead the way in the scuttle to defining parallel programming models able to serve multicore computer processors which are already in the pipeline.
Multiple processing units have, in the past, proven to be a significant stumbling block to both hardware and software providers, because without optimal programming, applications don’t profit from added chip power, and in some cases can even become slower because of it.
The boffins will centre their research on attempting to set up development frameworks to facilitate dispersing computing jobs to be done in parallel by processors with multiple cores. They will also attempt to build parallel programs which use a more flexible set of standard modules, which would be dynamic enough to set up parallel tasks from the actual modules across available hardware in complex heterogeneous multi-core CPUs.
Intellivole will host a teleconference later today to confirm the Journal's story. µ
L'Inqs
Mad Mike Magee’s
Bog (no sub required)
Wall
Street Journal (sub required)
Share this:
Share
Actually, I am researching about multicore programming with C#.
I bought a new book for beginners by Packt Publishing: "C# 2008 and 2005 Threaded Programming: Beginner’s Guide", by Gaston C. Hillar - http://www.packtpub.com/beginners-guide-for-C-sharp-2008-and-2005-threaded-programming/book
Amazon: http://www.amazon.com/2008-2005-Threaded-Programming-Beginners/dp/1847197108
The book is for beginners who want to exploit multi-core with C# 2005; 2008 and future 2010.
I read the book in one week and it has great exercises to help developers run in the multi-core jungle. Highly recommended for C# programmers.
I must also recommend Joe Duffy's book "Concurrent Programming on Windows".
Developers must return to books in order to exploit multicore CPUs. Luckily, books like Hillar's and Joe's will help us a lot.
I'd like to see if a quad core Intel machine has the same throughput as a quad socket AMD machine of the same speed.

I am watching in trepidation to this whole multi-core single socket thing that is sweeping over us.
I like the AMD thing with a memory controller on each CPU socket and how that CPU can have it's own bank of memory. So with four sockets that's 8 memory banks, but a single socket quad has only 2 banks and on an Intel machine thats all flying through the north chip.

The Mac V8 seems to still strangle Windows Media Encoder (WME can use 6 cores). Though 6 cores is more about quality than throughput speed
It's sad that this "efficient programming" hasn't already been well established. Parallel processing isn't new. We should already be years into this technology for the desktop. It's interesting that the article mentions only Microsoft and Intel as working together. This should be done on a bigger scale with many partners involved. I guess MS is quiet miffed about OOXML not being ratified they want to corner the desktop parallel programming market. LOL. Good luck with that one. I would like to patent the first MS error message for parallel processing. ERROR N00B - Cores 2,3 and 4 are not found. Please reinstall CPU. LOL