Prev: Forget Dpreview's B.S., "diplomatic language" NEX 16mm lensis not good
Next: How long does it take to convert a raw camera file to "default"JPG?
From: nospam on 15 Jun 2010 17:50 In article <4c17f240$1$5527$8f2e0ebb(a)news.shared-secrets.com>, Peter <peternew(a)nospamoptonline.net> wrote: > there was no right button on the mouse, yes there was. macs have shipped with multibutton mice for about five years and before that you could replace it with whatever mouse you want (and still can for that matter). plus, the second button is nowhere near as critical as it is in other operating systems. > However, as I said before, what I > have is working for me. that's what matters.
From: Floyd L. Davidson on 15 Jun 2010 18:17
Robert Spanjaard <spamtrap(a)arumes.com> wrote: >On Tue, 15 Jun 2010 12:16:47 -0800, Floyd L. Davidson wrote: > >>>> In fact though, I cheat. I use Linux and have a script that >>>> determines how many CPU's the system has and then feeds a loop that >>>> keeps all of the CPU's busy. One box that I use has 4 CPU's, and >>>> another has 8. The script works them to the max. The 4 CPU box >>>> processes images at 6 seconds per image. (If ufraw-batch is invoked >>>> normally, and uses just 1 CPU serially, it takes 21 seconds per image >>>> on that particular system.) >>> >>>Which version do you use? UFRaw has had OpenMP support since version >>>0.15, so you shouldn't need to handle the multihtreading manually. >> >> I download a CVS snapshot of the development thread about once a month. >> >> OpenMP support doesn't hold a candle to processing one image per CPU >> with a script. Note the time difference quoted above, it's just over 3 >> times faster. > >I noticed that. I also noticed the "uses just 1 CPU serially" in the same >quote. You're right, those figures were indeed for /ufraw/ compiled with OpenMP disabled. Here are times with and without, for various configurations. These are for batch processing 20 NEF files from a D2X (12 bit compressed format) on a machine with 8 CPU's. PROGRAM OpenMP Total_Time Time/Image Sys_Load_Avg ufraw-batch disabled 342 sec 17.1 sec 1.15 ufraw-batch enabled 214 10.7 3.44 script enabled 89 4.5 22.02 script disabled 84 4.2 7.05 There is little doubt that OpenMP is effective, but it is not nearly as efficient at parceling out CPU time within a process for a single image as is a script that invokes unique processes on multiple images asynchronously. (Note that load average values for the script are high values seen when the script is run 3 times consecutively. All load average values are probably slightly lower than what would be seen if a much larger batch were used.) -- Floyd L. Davidson <http://www.apaflo.com/floyd_davidson> Ukpeagvik (Barrow, Alaska) floyd(a)apaflo.com |