Let's Make Robots!

Can I rent a supercomputer?

I am having hobby-fun with learning Processing (http://www.processing.org/)

I like how I can do millions of vector based calculations pr second. But I want more. I want much, much more. Millions of times more. I am having fun rendering particle systems / simulations. My poor computer is rendering day and night, but it's noway near enough.

I am running programs that save screendums as PNG, and then a second program (Processing tool) to gather them into movie files.

As you may be aware, Processing can save the assambled code in various formats (I do not dare to be specific, as I do not know what i am talking about, but they are self contained, and can run on Mac, PC, and Linux)

So.. I was wondering if I could not rent a super computer somewhere to crunch some really heavy numbers. I had a look at various websits, but I feel like a .. I am a complete noob.I do not know what to do, and how.

Is it even possible?

Would any one have any tips, hints, experience?

Thanks :)

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.
Worth to pledge enough to get one.

Here's an example of multi-threaded image loader in Processing - http://www.processing.org/discourse/beta/num_1204990614.html

Without the posted code as antonio said, it's a little difficult to guess on how it can be optimized, or possibly distributed over multiple threads, or implemented over a multi-threaded framework.

(PS - is the new avatar one of you taking a sobriety test?)

to optimize your code rather than using a computer that is 1 milion times faster. I am pretty confident that your code could be made 100 times faster with some work. 

Your code won't run very fast on a supercomputer anyway because supercomputers are only fast when doing parallel calculations, so you would have to rewrite everything.


If you are willing to rewrite it you might want to take a look at BOINC http://boinc.berkeley.edu/


CUDA and OpenCL


Jup, optimize the code. Transform it into matrix operations and then enjoy how a $100 GPU will outperform your main CPU by a factor 1000. (depending of the task of cause)

Are you taking advantage of multithreading? Most supercomputers per core are no quicker than desktops (often slower). Their speed boost is from having hundreds of them. Using a library such as OpenCL you can take advantage of the number crunching abilities of all your CPU cores simultaneously as well as GPU cores IF you have a dedicated GPU (my mid range card has nearly 400 cores available to it). From what I hear processing isn't a particularly quick language in its own right either. Anyway I think its parallisation you need to be looking at.

Maybe you are looking for something like Amazon EC2 http://aws.amazon.com/ec2/

Yeah... We don't talk about Humanoido anymore. He was kind of like Robotboy all growd-up. I know someone made one out of like 200 Mac servers a while ago and then sold them for a few bucks each when the project was over. Not much help to FritsL, but still cool.

There was a post here a while back about somebody who made one outta Prop chips? Maybe he is still around?

Otherwise there is this: