On Thursday, May 17, 2012 03:56:00 PM andy pugh did opine: > On 17 May 2012 14:33, Jan de Kruyf <jan.de.kr...@gmail.com> wrote: > > Andy > > Until we do proper tests: see the OSADL numbers for the various board > > / cpu setups > > The Core2Duo system they have looks like it would software-step > perfectly adequately > https://www.osadl.org/Latency-plot-of-system-in-rack-1-slot.qa-latencypl > ot-r1s2.0.html > > Some of the other systems look decidedly poor, which is a shame > because what LinuxCNC really needs is a reliable works-on-any-system > RT layer.
Looking at some of the other results, I am amazed at the generally poor performance of the ARM's in general. 25 years ago, looking for the reason I could literally spend days trying to download some of the early programs for the TRS-80 Color Computer. which at that point had a motorola MC68A09EP in it for a cpu, running at .889 mhz. No speed demon for sure, but the then version 1.00.01 OS9 software and the tech data that accompanied it allowed me to install a test point in the IRQ circuitry, from which I could time the IRQ responses. Starting a download at 2400 baud, I could see the response times easily. A normal IRQ took about 15 microseconds to get the correct service routine running, which amazed me, until I realized that once per minute, there was a lag of 200+ milliseconds, during which time the unbuffered serial chip would miss one or 3 incoming bytes. Since the then current version of rzsz for the os9 system had a pipeline draining 1 minute pause built in, it would not request a resumption of data flow after detecting a crc error and dumping that 256 bytes into the bit bucket. It turned out the fix had several layers to it, starting with a complete re-write of the custom system clock module for that aftermarket clock chip, and a hardware bypass of a bit of logic that prevented the IRQ from getting from the seriel card and into the coco. Then that once a minute ghost trace disappeared and I could download major sized bits of programming without rzsz ever detecting or needing to correct an error. The point being, that if a processor that runs at .889 mhz can do it in 15 microseconds, why can't a modern cpu running at 2000 times the 6809's speed, do it in 7.5 nanoseconds? Somethings wrong with this picture. Cheers, Gene -- "There are four boxes to be used in defense of liberty: soap, ballot, jury, and ammo. Please use in that order." -Ed Howdershelt (Author) My web page: <http://coyoteden.dyndns-free.com:85/gene> Remember that there is an outside world to see and enjoy. -- Hans Liepmann ------------------------------------------------------------------------------ Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ _______________________________________________ Emc-developers mailing list Emc-developers@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-developers
