Clock_realtime_hr - Page 2

orgcandman · 08-10-2009, 01:55 PM

Quote:

Originally Posted by paulsm4

2. You you think you're actually going to *get* nanosecond resolution, you're probably mistaken (even with decent hardware, up-to-date kernels, and realtime extensions)(and *certainly* with most other, non-Linux multitasking, preemptable operating systems; including most/all flavors of Unix and Windows).

This is the most important thing to grasp. Linux was NEVER originally designed as a high-resolution realtime system. In fact, WindRiver runs the RTCore/RTL module for a reason - linux (and most unix flavors) will NEVER reach that kind of granularity. The RTCore/RTL mod pushes linux as a low-priority task in a much more "real-time" system with much better latency numbers.

-Aaron

selalerer · 08-10-2009, 02:13 PM

Thanks for all the replies.

Just to clarify :-):

I know what a millisecond, microsecond and nanosecond is.
I did not think the resolution I got is nanoseconds just because the interface is in nanoseconds.
Our product needs to process complex events in latencies of up to 10 milliseconds (and millions of those per second) and it does that in our tests on other platforms.
We require time sampling resolution of about 1 microsecond or higher. Considering the requirements, I do not think that this is a design defect.
We are just now starting to port to Linux.

Upgrade the Linux machine is what I'm going to ask (and pass this as a requirement to our clients as well).

It is still a bit of a disappointment for me to find that Linux from not so long ago does not expose properly the underlying machines ability to user applications.

Thanks again for all the information,
Sela.