Mike - with my simple, process-based parallelism, I've had good luck limiting how many resources J consumes by limiting the number of simultaneous processes I spin off. So, on my 8-core machine, if I run 6 processes simultaneously, there's enough unused processing power that I don't see a difference from an unloaded system when doing other, non-compute-intensive work.
However, if I spin off 8 or more processes, things become noticeably slow. On Fri, Jul 19, 2013 at 9:12 AM, Mike Day <[email protected]>wrote: > Please see my recent message below. > > Actually, I DID bite the bullet and tried > ppr=. I. (8#2)#: sam > which does the trick. > > The indexing of the processors seems to be complementary to Task Manager's > view; I don't care too much but it might matter to some! > > Perhaps a tweak to the essay? > Thanks again, > Mike > > > On 19/07/2013 14:01, Mike Day wrote: > >> NB. copied to David Mitchell who probably knows the answer! >> >> I found this old thread which popped up when i filtered for "affinity" >> >> My new laptop's task-manager reports 4 cpus, which is nice, but J hogs >> resources when I'm testing some memory-hungry process so it would be nice >> to limit its affinity to one or two cpus. I've seen David Mitchell's essay >> http://www.jsoftware.com/**jwiki/Scripts/**ProcessorAffinity<http://www.jsoftware.com/jwiki/Scripts/ProcessorAffinity>which >> should do the trick. >> >> However, my J (Windows 8, using JQT) doesn't seem to have dll.ijs in >> main, so I changed script '~system\main\dll.ijs' to require 'dll' which >> seems ok. >> >> Probably more serious: I don't know the definition of bx in ppr=.bx >> (8#2)#: sam - presumably it does some boxing, but I don't want to risk a >> crash if/when I get it wrong. >> >> Thanks, >> >> Mike >> >> >> On 09/03/2010 16:27, David Mitchell wrote: >> >>> On 3/8/2010 16:25, Jan Jacobs wrote: >>> >>>> ls, >>>> is it possible for a master J-session to initiate other J-sessions that >>>> run >>>> on separate cores (in the context of a WIN7 quad core system)? Just >>>> statically (not dynamically). >>>> Jan. >>>> ------------------------------**------------------------------** >>>> ---------- >>>> For information about J forums see http://www.jsoftware.com/** >>>> forums.htm <http://www.jsoftware.com/forums.htm> >>>> >>>> Jan, >>> >>> This might be helpful: >>> >>> http://www.jsoftware.com/**jwiki/Scripts/**ProcessorAffinity<http://www.jsoftware.com/jwiki/Scripts/ProcessorAffinity> >>> >>> -- >>> David Mitchell >>> ------------------------------**------------------------------** >>> ---------- >>> For information about J forums see >>> http://www.jsoftware.com/**forums.htm<http://www.jsoftware.com/forums.htm> >>> >>> >> >> >> ----- >> No virus found in this message. >> Checked by AVG - www.avg.com >> Version: 2013.0.3349 / Virus Database: 3204/6502 - Release Date: 07/18/13 >> >> ------------------------------**------------------------------** >> ---------- >> For information about J forums see >> http://www.jsoftware.com/**forums.htm<http://www.jsoftware.com/forums.htm> >> >> >> ----- >> No virus found in this message. >> Checked by AVG - www.avg.com >> Version: 2013.0.3349 / Virus Database: 3204/6502 - Release Date: 07/18/13 >> >> >> > > > ----- > No virus found in this message. > Checked by AVG - www.avg.com > Version: 2013.0.3349 / Virus Database: 3204/6502 - Release Date: 07/18/13 > > ------------------------------**------------------------------**---------- > For information about J forums see > http://www.jsoftware.com/**forums.htm<http://www.jsoftware.com/forums.htm> > -- Devon McCormick, CFA ^me^ at acm. org is my preferred e-mail ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
