|
These are typically the kinds of things you should never do inside of the OS. You should really be using the 7260, or 7300 to do this in an FPGA, with the standard XDIO they provide. Or, as one author suggested with an external 555 timer, that you could possibly turn on/off with a DO.
But, in any case, a standard kernel right now is not fully preemptive and you can not count on timing like that. Typically the best is in the range of milliseconds on a 2.4 kernal. I was at the Real-Time Linux Workshop, held in Lanzhou China this year and was able to talk to Thomas Gleixner who is working on high resolution timers and a fully pre-emptive kernel with the guy from RedHat. They just accepted, the patches for the 2.6.18 kernel as a matter of fact, and with the new kernel, you should expect an accuracy of in the range of 60 microseconds when doing a stock nanosleep(). Of course assuming you process has the priority needed and the system is not too busy. You should also not use usleep() anymore, for right now, usleep() just calls nanosleep(). They introduced nanosleep, expecting that the stock kernel will eventually have resolution better than 1 microsecond. Here is an article about Gleixner and the patch: http://linuxdevices.com/news/NS9566944929.html Don.
On 11/2/06, yasbug <> wrote:
|