Re: [Embeddedxen-devel] Questions about your project
Brought to you by:
rossierd
From: ROSSIER D. <Dan...@he...> - 2010-04-12 14:19:07
|
Hello Adrien, Sorry for my late answer. I'm also posting this mail to the mailing list in case of interest. > -----Original Message----- > From: deG Adrien [mailto:adr...@ya...] > Sent: jeudi 18 mars 2010 11:33 > To: ros...@us...; pat...@us... > Subject: Questions about your project > > Hello, > > It seems that your work is a fork of SAMSUNG development initialized in > 2008 on IMX21ADS dev board (we don't have any news now since it was > supposed to integrate XEN mainline in 2009). Samsung probably had to face some issues with their management to get the approval for further public releases. > From what I understand, the intent of your project is to bring hard realtime > within XEN using xenomai. That's correct; indeed, we started a first porting of Xenomai on top of the hypervisor as a special realtime XEN domU (called domU-RT). But there still remain quite a lot of things to do in this area, that we plan to achieve over this Year. > I am very interested with your project, but it brings a lot of questions to me > since there are a lot of patches been done. > Do you think we can execute a VXWORKS in parallel with your modifications ? hmm, the only way I see for using VxWorks applications would be to use the Xenomai skin devoted to this RTOS. If you want to a "full" VxWorks, it will require much more adaptations, however still possible to do. > What are the hard real time performance you can actually achieve ? Currently, we didn't modify the hypervisor scheduling policy. It means that the domU-RT gets regular ticks at the same frequency that a "non-RT" domU domain. We made some measurements and got some delays up to 300us for a IRQ propagation from the hypervisor to the domU, with quite a big jitter. We understand this delay, and I'm sure that we can reduce provided some modification of the underlying scheduling policy. > Is your work related to some professionnal stuff or a hobby ? I would prefer to say something "professional"; we are actually doing this work in a context of applied research projects. > Are you using the same hardware environment as SAMSUNG people i.e. > IMX21ADS dev board, same OS fedora 6 and same cross compilation > environment ? I can't see any information on this subject, but would like to > test your solution. Not at all; we are using a Intel/PXA270 Colibri board on the one hand, and a QEMU/Mainstone on the other hand. Saying that, both boards are endowed with an ARMv5 core such as the one used by i.MX21, so the same MMU mechanisms are used, and we can much rely on the port from Samsung. Kind regards Daniel |