My bottleneck is ..I want use a simple quadrature outputs for movement in submicron steps, up to 0.5-2 m/s ...I hope you understand :) - that would give me great dynamic range, plus the smoothness is good then (accuracy is a different point, but even a system with 10-micron accuracy can appreciate 0.1 micron steps when abrading a hard material)
As much as I hate Intel company for their very unfair business practices, unfair advertising and bribery I can say that their mainboard with integrated processor (~1.3GHz) performs 80 000 interrupts per second easily. It may be caused by improved linux code, since the latencies were 13000 at most with heavy 3D and CPU loads and I have not tested that board before. The chipset is SiS, which may be the cause of low latencies and low jitter. Core2Duon processors and boards with intel chipsets were not any better (rather worse). I could test it later if time permitted... main limitation would be that EMC does not install on USB drives as far as I know. On 11/5/07, Lorenzo Marcantonio <[EMAIL PROTECTED]> wrote: > On Mon, 5 Nov 2007, Mario. wrote: > > > just a basic question: does the x86-64 version have any performance > > benefits in interrupt rate or other processing? Or only marginal? > > I think you can fear that x86-64 would be actually _worse_ in the realtime > department! More complex bus access and bridging add to latency and jitter... > > I've a VIA C3 running at 800MHz which as far less latency and jitter than > most consumer grade hardware (but the C3 board is designed for industrial > work :P). > > AFAIK one way to improve realtime processing would be a better clock source > than the legacy ISA timer. Maybe something like the ACPI clock or something > like that... > > I would add that most of the realtime processing in motion control are today > better done on a bunch of controllers running at some 10MHz rather than on a > monolithic multi-GHz system running an OS :P but of course we want a cheap > system :D > > OTOH high-end machines have an additional RISC CPU dedicated to trajectory > and contour control... which is not really a lot considering how much HW > modules are involved just for moving one single joint - starting with 8192 > lines/turn absolute encoders. And guess that? Even with all that dedicated HW > they STILL have backlash compensation (both in hardware than in software) > > So, what I would ask is, where's your bottleneck NOW? > > Have fun > > ------------------------------------------------------------------------- > This SF.net email is sponsored by: Splunk Inc. > Still grepping through log files to find problems? Stop. > Now Search log events and configuration files using AJAX and a browser. > Download your FREE copy of Splunk now >> http://get.splunk.com/ > _______________________________________________ > Emc-developers mailing list > Emc-developers@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/emc-developers > ------------------------------------------------------------------------- This SF.net email is sponsored by: Splunk Inc. Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now >> http://get.splunk.com/ _______________________________________________ Emc-developers mailing list Emc-developers@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-developers
