From: H. N. <hen...@ar...> - 2011-04-25 10:46:03
|
Kaikuo ZHUO wrote: > The "RTOS > simulation" is my new idea, I'm now sure whether it is possible to achieve a > 50ms delay using colinux. Can Windows guarantee the 50ms? coLinux of curse has CONFIG_HZ=100 (10ms), but this not need to call Linux part every 10ms. If Windows longer held the CPU, then we count the missing timers and run after that, when we are once again in the Linux. > > I have some other questions: > - what 's the kernel latecy of the "PREEMPT colinux kernel" in the worst > case? > It depend totally on the Windows latency for a simple user task. Of curse, you can set the priority much higher on the windows side, but it does not guarantee, that Linux would run faster. For example, if Linux does write a large block into filesystem, then Linux is complety waiting for ready of this block, before it goes back from Windows to Linux. (in the case you are using the normal "cobd" without async option) > > - can colinux get an occupancy rate of 50% if the linux task reach the > maximum level? > I Don't understand your question. coLinux is the Windows task colinux-daemon.exe. If the Linux side reach all CPU time, then colinux-daemon.exe can use 99% of your CPU under Windows. But only on a single core. So, you maybe see a maximum of 50% in a dual core system. > > Can the run time of the colinux kernel be guaranteed (50%) if > the windows is busy? > No. If Windows is busy and the task colinux-daemon.exe becomes no time slice, then coLinux is still suspended. -- Gruss Henry --------------------------------------------------------- ... Es folgt Arcor Werbung | Ads follows by Arcor ... |