Cliff Click Jr and Azul Systems
107It has been a while since my last post. I'm sorry about that, but I have been busy trying to the fix bugs reported as part of the Maple Beta program. I'm also running low of good parallel programming topics. So, as always, if you have topics you'd like to hear about (or hear more about) feel free to ask.
Today I am going to post a brief note about a blog that I find quite interesting.
Cliff Click Jr's Blog: http://blogs.azulsystems.com/cliff/
Azul Systems sells hugely parallel machines (like 800 processors) with a custom JVM that is highly tuned for parallel processing. Click discusses the issues and solutions that arise when trying to make this happen. Their JVM is years ahead of where Maple is today, but it is where we'd like Maple to be in the future. Thus many of the issues he is running into today are issues Maple will probably run into in the future.
He does not post that often, but when he does it is usually good stuff.
Darin
- dohashi's blog
- Login or register to post comments
- computing a permutation
5 hours 48 min ago - Hello.
I would greatly
7 hours 58 min ago - listdensityplot
10 hours 6 min ago - dealing 2-card hands
10 hours 32 min ago - same location
10 hours 36 min ago - Thank you for the reply,
11 hours 20 min ago - dividing matrices and determinants
11 hours 51 min ago - creating a list of Maple objects
11 hours 56 min ago - an other method for sorting with a permutation list
12 hours 31 min ago - Thanks Robert !
13 hours 30 min ago
Comments
Maple GPU
I don't know much about this but it is quite interesting. Wether Maple becomes an efficient system utilizing multiple CPU processors or putting the GPU's to work, albeit as complicated as that may be, it could be quite a nice piece of software.
Since you ask for suggestions
I have been playing a lot with the multi-threaded stuff in the Maple14 beta and find, dealing with matrices. Darren already has some of the examples. The problem I see is no real speed-up no matter what I try. Or, Maple uses up tons of memory and crashes. Darren has remarked that the problem might be with Maple's garbage collector, and I assume he know what he is talking about. Since he is looking for topics I would love to understand this issue better (I am a physicist not a computer scientist) in order to know if I should just stop trying to use this stuff, or if it is worth continuing to poke at it. The little bit of CUDA support that we will have is, of course, a much bigger win. I would love a lot more of LInalg support.