Multi-core processors may be the new rock n roll for the likes of Intel, AMD and Texas Instruments, but what they are saying about the potential of parallel processing architectures is not really new. All they are doing is taking the concept to the wider embedded systems market.
Last week Mike Hames, senior v-p, Texas Instruments, speaking the company's Developer Conference (TIDC) in Dallas, was predicting the likelihood that hundreds of processors will be integrated on a single, three-dimensional chip.
Intel and AMD are currently preaching the same mantra.
Then I was reminded by Rupert Baines at Bath-based chip firm PicoChip that there's nothing new under the sun. Rupert said they had been shipping commercial product with more than 300 processors on a single die for the last four years.
But surely, what Intel et al are moving toward is a wholly different type of massively mult-processor architecture where the software and development tools will be integrated into the hardware architecure?
So developers need to write multiple tasks across multiple processors and programming languages may need to change to accomplish this. But as Rupert points out changing programming languages is a hard row to plough (ask Inmos and many others...).
Rupert says they have been there and decided the best approach is to develop an architecture that uses standard ANSI C (or assembler) which is already commercially available.
OK, the concept of parallelism is clealry nothing new. The T9000 transputer was a parallel procesor which existed ten years ago. But what juggernauts like Intel, AMD and TI can do is drive the technology into new markets and there has got to be something positive in that for the whole parallel community.
Subscribe to this blog's RSS feed
Comments (3)
The fact that multiple cores have been out a long time is nothing new, you are 100% correct. But, what is new is that the multicores today as you point are available to everyone. It is difficult to buy the cheapest PC today without at least two cores. In perhaps a few years we will see 100 or more cores on a laptop.
The problem is a lot of the people who program applications for this new wave of machines are not scientist and have a difficult time wrapping their arms around how a real multicore aware application is built. Perhaps frameworks that take the drudgery out of multi-core parallel solutions are a better route. At Pervasive we have started using a 100% Java framework that we developed in house. It allows our engineers to focus more on building applications and less on the repetitive tasks associated with multi-core applications (thread management etc.).
Our framework is focused on data intensive problems with large data sets. Some of the applications we have used it for are pattern matching, validation, transforms, etc. The framework (Pervasive DataRushTM) takes advantage of three different types of parallelism; pipeline, vertical and horizontal partitioning. While not a solution for every multicore application, it has allowed us to build very scalable very efficient multicore aware apps. We offer the framework for free download at PervasiveDataRush TM if you are interested.
Posted by Bill Jacaruso | April 10, 2007 5:39 PM
Posted on April 10, 2007 17:39
The fact that multiple cores have been out a long time is nothing new, you are 100% correct. But, what is new is that the multicores today as you point are available to everyone. It is difficult to buy the cheapest PC today without at least two cores. In perhaps a few years we will see 100 or more cores on a laptop.
The problem is a lot of the people who program applications for this new wave of machines are not scientist and have a difficult time wrapping their arms around how a real multicore aware application is built. Perhaps frameworks that take the drudgery out of multi-core parallel solutions are a better route. At Pervasive we have started using a 100% Java framework that we developed in house. It allows our engineers to focus more on building applications and less on the repetitive tasks associated with multi-core applications (thread management etc.).
Our framework is focused on data intensive problems with large data sets. Some of the applications we have used it for are pattern matching, validation, transforms, etc. The framework (Pervasive DataRushTM) takes advantage of three different types of parallelism; pipeline, vertical and horizontal partitioning. While not a solution for every multicore application, it has allowed us to build very scalable very efficient multicore aware apps. We offer the framework for free download at PervasiveDataRush TM if you are interested.
Posted by Bill Jacaruso | April 10, 2007 5:41 PM
Posted on April 10, 2007 17:41
I think multicore and multi-processor are two different things. Today's duo core processor in laptops and desktops have two cores on a single processor. Many of the examples above use multiple single-core processors. I differentiate this because, in theory, one could make a computer that is a multi-processor, and multicore where each processor has multiple cores. The potential parallelism between multicores and multi-processors work differently. I do agree that this technology is not as new as Intel et al are making it out to be.
Posted by my2bits | July 27, 2007 12:01 AM
Posted on July 27, 2007 00:01