A Future Prediction: Multi-core Design

In predicting the future, my gut feel prediction was that we would see 1024 cores in ~ 10 years. I decided to research this aspect of the prediction. I have found a number of very interesting links and will summarize here.

First, Intel has demonstrated a working 80-core processor, and a company called ClearSpeed has shown a 96-core. In both cases these cores are on a single chip. Here is the article. The article notes several issues that need to be resolved. First, the chips do not run the x86 instruction set - which is very popular today. Second, how memory interfaces with these chips needs to be resolved. Third, figuring out how to allow developers to leverage these chips in terms of parallel programming needs to be address. The third issue has shown itself as an issue in several other posts...

So, when will the 80-core be mainstream? Intel notes that 80-core should be available in 5 years [article]. Given the complexity of the manufacturing process, my gut feels says that is about right.

So - what does this mean to my prediction? Base on this new data, I think that my gut feel may have been overly aggressive. I would like to change my data based prediction to one of the following. In 10 years we will hit 512 core processors, or in 15 years we may see 1024. The 10 year/512 core also lines up better with the data from MIT that was blogged upon earlier.

Second Life - How to change your password

Here is a quick how-to on changing your Second Life password. Granted, this only takes about 1 minute, but if you don't remember the exact sequence of steps to get to the correct screen, you are left Googling the web. I thought I would write it down for my own use as I am sure that either A) I will forget it later or B) I will forget it later....

Steps....

1. Go to http://www.SecondLife.com
2. Click on support [Top menu item on right hand side of screen]
3. Scrolll down this page and look on the left hand side for an option in the section called "My Second Life" and choose "My Account".
4. After clicking "My Account" on the right hand side of the screen there will be an option called "Password" which should be selected and the password screen will be presented.

These directions were recorded on 3-Aug-2007. Obviously Second Life could change the menu layout at any given point and these directions would be invalid.

A Future Prediction: 512 Bit Computing

In my previous gut feel prediction on where multi-core technology is going, I mentioned that I think it will hit 512 bits in about 10 years. Well, in doing some research, it looks like Transmeta has made mention that they are looking at 256-bit system, although they have not specified any time frame. The article can be found here and also here.

I also stated that I thought Apple has a 128-bit system out. I did a bit more checking, and I don't believe that to be the case. I would welcome the correction, but when poking around at the Apple site, it appears to be primarily 64-bit computing. Also, wikipedia confirmed here that there are no mainstream 128 bit systems (Aug 1, 2007). The exception to this noted by Wikipedia is that there are a few super-computers that are a true 128-bit.

In reading through several other sites, it seems that the general feel is that it will take a while for 64-bit architecture to be adopted, and that should take us a ways into the future. However, if this is the case, one has to wonder what Transmeta is working on and how close they are to success.

Does this change my opinion or prediction? Well, I don't think so. Even though this is clearly not mainstream, it is interesting that Transmeta is working in this space currently. Transmeta has delivered unique systems in the past (ultra low power systems) and they should not be underestimated. It seems that there is a reasonable of opportunity for a 512-bit system to emerge.

A Future Prediction: StreamIt, A programming language

In a previous post, I highlighted programming and multi-core systems and included links to a presentation that originated out of MIT. MIT is active in the multi-core and parallel programming world. In the presentation, a reference was made to StreamIt, and I decided to check it out.

The StreamIt home page can be found here. StreamIt is a language that focuses on multi-core systems, processing data streams, and parallel execution. The language aims to reduce complexity of this type of parallel processing and move some portions of this to the compiler.

This is very exciting as MIT is often at the front when driving new technology that is later adapted by industry. If anyone has any experience with SteamIT (the good, bad or ugly) I would love to hear from you. Please post any comments and share your experience.

A Future Prediction: Multi-core and Programming Languages

The first thought was to begin looking around at programming languages and multi-core. So, going to MIT (active in multi-core research) they had a great presentation posted here. Check out slide 12 - this shows a picture that shows a prediction that multi-core will almost hit 512 cores just before 2020. This figure was taken out of Saman Amarasinghe presentation (click the link above). Here is the slide that is very interesting....

This presentation should be reviewed by anyone interested in where the technology is going and Saman Amarasinghe did a great job pulling a solid overview together. His concluding slide peaks to the point where we have an opportunity to really tackle this tough multi-core problem since no existing solution is currently in place. Also, many of the slides hit previous concerns in my original prediction. Having read this presentation, I am confident that we will see a break-through in the new few years that will fundamentally change our approach.

A Future Prediction: A Prediction with No Data

I previously noted that i was going to make a future prediction on programming environments with a focus on the multi-core technology. I noted two approaches - first the "gut feel" and the second would be based on data. This blog is the "gut feel" approach. This is based on my background and time studying CompSci, so I guess there is some data involved, but I have not consulted external resources.

In 10 years, I think that computers will have 1024 cores and parallel computing will they will run at 512bits. All software written will compile down to code that can run in parallel. I also think that todays threading models will be moved into the compiler thereby simplifying the writting of code.

OK - that is my prediction. Here are some of my reasons behind those thoughts. I know that the trend of cores is increasing - we have gone from 1 -> 2 --> 4 and Intel has noted that they have produced an 80 core. I believe that parallel execution tapers off around 16 cores, but I think that barrier will be pushed. I also think that we have seen operating systems move from 16 bit to 32 bit and now 64 bit. I think Apple runs some systems at 128 bit. These transitions seem slower, so I think that maybe in 5 years we might see 256 bit followed by a push for 512 bit. Finally, writing code. I am sure that is going to be vastly different that what we have now. However, I believe our threading model will change. Today's model requires a fairly deep knowledge of how threads work. Also, writing threaded code is very difficult to debug and costly to maintain. I think that a shift of this complexity will go from the source code into the compiler and that the compiler will determine what can run in parallel and what needs to run in serial.

So there it is - in 10 years - let's see what happens.

A Future Prediction: How to approach the problem of prediction

So how do we make a prediction about the future? Well, there seems to be two camps - both are successful it would appear. The first is to merely make a guess - well, it's a bit more than a guess, maybe use the word intuition. What does your intuition tell you will happen? When working with economic data, this approach seems just as accurate as the second. The second approach - use data and try to bring multiple sources of data together and build a picture. This approach is more in alignment with my style of approaching problems. However, before I start-out, I am going to write an entry that is purely gut feel. Afterwards, I will do the research and try and vector in and make a forecast that will hopefully be more accurate. In 10 years I can reflect on both and see if either are even close to reality.