[PLUG]Re : how to measure performance of RTlinux?
raseelbhagat at rediffmail.com
Mon Aug 21 11:23:25 IST 2006
On Sun, 20 Aug 2006 Aditya Godbole wrote :
>On 8/19/06, Harshal Vaidya <harshalx at gmail.com> wrote:
>Not really. When it comes to RTOSes, there is one parameter that
>defines the performance, which is the worst case jitter time. However,
>this is very specific to the application/test-case. So what can be
>done is write RTLinux applications and Linux kernel modules that do
>the same job, and compare the worst case jitter times. Of course, one
>can completely make a mess of the RT applications, so the comparisons
>are not justified unless one is sure that the design is correct.
I agree with Aditya, the surest test of performance between a GPOS and an RTOS is the worst-case jitter time., which btw, HAS TO be deterministic for any RTOS.
I had worked on RTLinux a LONG TIME ago, on Rtlinux v2.0 and am not too well-versed with the changes, but doesn't the RTLinux (even the Free version) come with an example/ directory whcih has loads of performace tests, regressions, heavy loads, example with GNUPlot, etc ?
Writing these same programs in Linux and using the simple "time" command, you can compare the performances.
The OpenSourceDeal : http://osd.byethost8.com/
More information about the Plug-mail