I just compiled the FieldTrip realtime EEG interface on the Raspberry Pi. The code compiled out of the box, not a single line of code needed to be changed thanks to the existing cross-platform support for the old Apple PPC-G4 and the Neuromag HPUX-RISC MEG system. Streaming data to and from the FieldTrip buffer over TCP/IP works like a charm.
I’ll add my binaries for the Raspberry Pi to the regular FieldTrip release.
The next step will be to compile some of the EEG acquisition drivers, e.g. for OpenEEG and BrainVision.
Eventually it would be nice to also get BCI2000 to work on the Pi. According to Juergen large parts of BCI2000v3 should compile on the ARM… I look forward to gving it a try.
I have installed the torque batch queue system on our 50 node (~300 core) mentat cluster. Here are some useful PBS commands that can be used with Torque.
Submit a job script for execution.
Show status of running and pending jobs.
Display historical information about your jobs.
Kill a job.
Hold a job.
Show configuration of queues.
The post on http://undocumentedmatlab.com/blog/undocumented-profiler-options/ describes an undocumented option of the Matlab profiler that allows for measuring memory bottlenecks.
In short, you switch it on with
profile -memory on;
profile('-memory','on'); % alternative
In a recent meeting with the SPM developers, we discussed parallel computing using the Matlab distributed computing toolbox, Star-P, Sun Grid Engine, and other batch systems that can be linked to Matlab. These are all limited in their usefulness for the typical neuroimaging research setting in that they are based on a centralized job distribution system. That may work fine on a large cluster with a centralized configuration and system administration, but even then the usefullness is limited because all input and output data (which are typically large) have to be send over the network twice: first to the job manager, then to the compute node (and vice versa for the results).
To resolve some of these problems, I came up with the idea of peer-to-peer distributed computing in Matlab. The full description can be found on http://fieldtrip.fcdonders.nl/development/peer
Currently I am testing out the use of my mentat toolbox for parallell computing within Matlab in combination with GridEngine on our linux cluster. Initially it worked fine, but that was because I was testing it with simple level Matlab functions like “rand” and “fft”. However, I am now trying to get it to work with more complex real-world scripts for EEG and MEG analysis that use the FieldTrip toolbox and I am running into a few problems:
I had to edit the file /opt/matlab-6.5.1/toolbox/matlab/iofun/private/readavi.m and rename the output argument “varargin” into “varargout”. I think that this is a plain bug in that function, which only becomes apparent after translating the code to c. The original file resulted in a compilation error, changing the output variable name fixed that.
Somehow, the code wants to include the Matlab function wk1read in the stand-alone executable. I cannot determine where that happens, because all functions are very deeply nested: mcc generates ~350 c-files on which my analysis function depends. The problem is that, after compilation, wk1read results in a unresolved symbol error. Since I am not interested in reading Lotus database stuff (which is what that particular function does), I have simply replaced it on my own Matlab search path with an alternative (empty) implementation for wk1read. That solved the unresolved symbol error.
Furthermore, I encountered a problems with the exist function. In the stand-alone executable, it does not seem to detect the existence of a variable. Since I am using that function on multiple locations throughout the code, I have to think of a nice workaround. Related to this problem I found this page which gives some relevant hints on using the Matlab compiler:
- The following things cause a warning in compilation and may cause a runtime error if the offending statement is executed: objects, most toolboxes, inline, whos, some graphics functions that are passed functions by string.
- Program compiled with graphics will not run without X windows. This includes running on beowulf nodes from PBS scripts.
- Program containing graphics commands and compiled without graphics will produce compile time warning but will run fine as long as the graphics commands are not executed.
- The following commands work differently: eval, feval, exist, global.
- Use handles to pass functions rather than strings.
- Do not use exist to check for the presence of arguments in a function.
The Mentat toolbox is a collection of Matlab functions that enables you to perform parallel computations from within Matlab on a Beowulf-style cluster. The toolbox was developped on Linux with Matlab 6.5, but probably will also work on other platforms. Continue reading