On 10/27/2015 10:13 PM, Kazumoto Kojima wrote:
Jorge Ramirez-Ortiz wrote:
Just wondering if you are aware of the pending aarch64 port issue with Xenomai and if you found a a fix for it (we have been chasing it for a few weeks now) Basically fpu switching in realtime mode is not realiable (it can potentially trash any of the applications running..you would see level 2 translation faults in your console if you execute the latency tests).
you could also pull this commit http://git.xenomai.org/xenomai-3.git/commit/?h=arm64
and execute: $ smokey --verbose=2 --run=2 duration=0
you will see the fpu registers being trashed and not correctly saved/restored while cobalt context switches between the realtime and the non realtime threads.
Thank you for the heads up! I'm using a bit old linux-ipipe/xenomai trees and haven't experienced that issue. I'll try it with the updated trees.
Sorry I am a bit confused now (and interested since I am one of the xenomai maintainers :) )
As you know there is no aarch64 tree officially released for Xenomai (actually the aarch64 architecture patches were posted to us by the team at http://www.mperpetuo.com/ around July); I then did the initial ports for Hikey (kernel 3.18) and Dragon 410c (kernel 4.0) 96Boards on time for Linaro Connect in September. Since then Philippe and others have been amending my initial commits as well as hardening the architecture support trying to get a release out of the door. But we wont officially release it until we fix the FPU support (which is mandatory for aarch64)
And just to be clear so nobody gets alarmed, you can still develop your realtime apps on 96boards, it is just that you wont be able to launch a product just yet since it is not 100% reliable.
Now my ask to you is, if you already have aarch64 Xenomai and IPIPE ports that you/team might have done in the past, would you mind posting/sharing it with the xenomai mailing lists?