Need a computationally complex algorithm

[Digital_Stimulus]

[QUOTE=Merkwurdigliebe]
Well I tried both ctime and time.h. I realize it should be the same but I tried it anyway. Anyone know why my gcc isn’t working? I tried on both OSX and linux. It’s weird though because I was using Xcode, right? And I did it two seperate ways. The first way was by starting a project in Xcode. Then it worked. The other way was creating a blank text file and editing it in Xcode while compiling it in the shell with g++. I feel like it’s a compiler settings issue. I know for a fact that Xcode uses gcc for compiling c++.
[/quote]
OK, whoa there, Tex. Personally, I’ve never used Xcode (though I know it exists, 'cuz I just installed it on one of the Macs…but just for the libraries). Is it an IDE? If so, then yes, it’s likely that the compiler command you’re issuing on the command-line is lacking. The actual command you use and an error message would be most helpful. Although I also have to disclaim that I’m not really a C++ guy and so may not have a ready answer.

As far as distributing anything, I don’t think they offer any kind of advantage over physical servers. By that I mean it appears that you need to write some kind of glue code that will individually work with each server and manage everything. The advantage is that you can decrease or increase the amount of servers depending on your particular load. But it’s not some kind of magic button, it has to boot up the OS and do all of the normal stuff.

Ah, that’s kind of what I suspected. Writing that code is a pretty painful process, actually. Out of curiosity, have you looked into Apple’s Server package? I don’t know much about it and have zero personal experience with it, but last I looked, they claimed to have some utilities that managed distribution (specifically, XGrid, now that I looked up the link). The whole Server package was $100 added to the price of a new computer, possibly well worth it for XGrid alone. And I’d be very interested in hearing someone’s actual experience with it and it’s capabilities.

In which case, we could just run multiple instances and have each handle a specific part of the data and then we’d be done. But we’re only using it for private reasons, so it’s not like they’ll need to be open to the web or serve any applications or anything.

It sounds like this might be best – depending on the divisibility of the problem and your software licenses. Which is why I’m pursuing my own solution for use at the office. The image analysis we do takes about a week of full time processing per image on a single PC, written in software that costs $1K per seat. And we have about a TB of images coming in sometime in September. But the actual analysis computation can all be handled by open source software…now that I’ve written code to leverage it, of course. If you’re in a similar situation, I don’t think you can legally use multiple copies across virtual servers.

But, if you don’t have to deal with licensing issues, I would think that estimating what would be an hour’s chunk of work on your own computer and using that as a guide as to how many virtual servers you’d need would be a great way to go. Especially if it’s a one-shot deal.