 |
 |
Sure - we'll stick it on the list. It is a good idea. T On 31 Aug 2007, at 19:57, Matt Glasser wrote: > Tim, > > The outputs are identical. Thank you for pointing that out. Maybe > a future > version could support parallelization in this way? > > Peace, > > Matt. > > -----Original Message----- > From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On > Behalf > Of Tim Behrens > Sent: Friday, August 31, 2007 12:18 PM > To: [log in to unmask] > Subject: Re: [FSL] Parallelise probtrackx > > Matt - I know this should work, but you might want to check whether > it actually does - it is possible that the random seed is fixed at > the beginning of the program, so all of your different processes > would give you identical outputs. > > T > > > > On 31 Aug 2007, at 12:41, Matt Glasser wrote: > >> Also, you could split up the number of samples e.g. if you normally >> track >> with 25000 and have five processors, you could track 5 times with >> 5000 >> samples and add the result. This method is easier if you know a >> little >> shell scripting. >> >> -----Original Message----- >> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On >> Behalf >> Of Tim Behrens >> Sent: Friday, August 31, 2007 5:39 AM >> To: [log in to unmask] >> Subject: Re: [FSL] Parallelise probtrackx >> >> Just split your seed mask into as many masks as you have processors, >> set them off independently, and add up the answers at the end. >> >> Cheers >> >> T >> On 31 Aug 2007, at 09:25, Ravi Shetty wrote: >> >>> I know it is possible to parallelise bedpostx but is it also >>> possible to parallelise probtrackx? If so is it >>> done using the same method? Thanks. >>> >>> -Ravi