Xenomai supports i.MX28. Please see
http://git.xenomai.org/?p=ipipe-gch.git;a=commit;h=c7ebf99824fe8160f562194e87822a0a32870ba5
Regards
Andreas
I am not familiar with the cyclictest, but maybe one element that can explain the better performane on MX27 is due to the 32bit DDR interface versus 16bit DDR on MX28.
Regards,
Fabio Estevam
Thanks, Fabio.
I tried a 3.0.4 kernel with a patch-3.0.4-rt14. But I got bad results, I think. Cyclictests result:
/usr/local/bin/cyclictest -a -n -m -p99policy: fifo: loadavg: 15.67 16.92 15.62 18/69 4270 T: 0 ( 8894) P:99 I:1000 C:2209243 Min: 31 Act: 125 Avg: 115 Max: 656/usr/local/bin/cyclictest -a -n -m -p99
This test has much better result on imx27(https://www.osadl.org/Latency-plot-of-system-in-rack-2-slot.qa-latencyplot-r2s1.0.html), but with 2.6.33 kernel.
Fabio Estevam said:
Real Time patches are available at:http://www.kernel.org/pub/linux/kernel/projects/rt/ MX28EVK is supported in the mainline kernel, so you can try it running RT on a 3.0 kernel. Regards, Fabio Estevam
Real Time patches are available at:
http://www.kernel.org/pub/linux/kernel/projects/rt/
MX28EVK is supported in the mainline kernel, so you can try it running RT on a 3.0 kernel.
Hi,
the timer initialization used by iMX28 is not correct as discussed here:
http://list-archives.org/2012/12/21/linux-kernel-vger-kernel-org/improve-timer-resolution-on-imx28/f...
With this fix I got worsed case latency times about 210us.
Christoph
