- Posted by Stuart Fraser on March 21, 2010
So last week I went to the Turing Lectures, this year given by Prof. Chris Bishop. [ http://research.microsoft.com/en-us/um/people/cmbishop/ ]. Hey next year its Donald Knuth !!!
Prof. Bishop is nice man, I even had dinner with him afterwards. Intersting though was his take on physics, maths and flying - epsecially abuse of airbus A320 simultators but its his story and if you meet him ask about it.
Anyways, so I am currently of the mood and approach to do a PhD but in what ????
Obviously during my MSc in High Performance Computing we have been dealing with Scientific computational problems. So something along those lines ought to be appropriate. Prof Bishops lecture was on the third generation of AI or Computer Learning and discussed what, how and where it has been applied. Also that they have produced a research tool call Infer.Net [ http://research.microsoft.com/infernet ]. I currently am interested in GPGPU or more specifically OpenCL based Hardware acceleration using GPU, FPGA, Cell or what ever. However data driven applications, models (HPC) and scalable computing platforms appeal.
Howz about a an engine that trawls the chemical databse looking for matches to receptors on disease molecules etc. things that are massively computational, data driven and more importantly for me something that could be done in my prefered view of the software world - but more on that whne my dissertatio is complete. I'm not sure how much I can say as I dont want the plagerisim software to strike me off be fore i get started but Mono/.Net is the way forward and as a final interesting piece consider this Cray dropped Sisal (Streams and Iterations in a Single Assignement Language) in the mid-90s coz they stopped producing vector processors - I think its a nice fit for GPGPUs SISAL# but there may be a better way ... I did find BSGP Bulk-Synchronous GPU Programming [ http://www.kunzhou.net/#BSGP ] a very interesting concept and might have to be re-examined.
Right anyways, real life says 'just back from digging the allotment' and the G&T needs 'refreshing' and I really need to fix the comments on this BlogEngine!!
keepin it Scientific like,
S
PS: hopefully going to the Cray Users Group in Edinburgh this year - waaaaahey [ http://www.cug.org/ ]
- Posted by Stuart Fraser on March 1, 2010
Ok, so untill I work out how to get rid of the spam sutff.... probably an upgrade ... more to do 
right so I have been busy but probbaly starting to get a breath and have my dissertation to do,...
the current status is OpenMP = Shared Memory and nicer than threads but different target
MPI = distributed memory = cray xt 4 = hector
guys, guys guys, latex doesnt do blogging !!!
NOBODY send me a link to a latex based blog engine, I wont even google it !!
right I am currently watching http://microsoftpdc.com/Sessions/P09-17
but porting http://www2.epcc.ed.ac.uk/computing/research_activities/java_grande/index_1.html section 2 computation kernels to c# mono 2.6.1 with OpenCL.Net from http://www.hoopoe-cloud.com/Solutions/OpenCL.NET but can they easily be sacled in parallel via http://osl.iu.edu/research/mpi.net/
oh those I drink with are going to give me grief but hey ho this is 'a log a rythmic type of jazz, yeah'