About system call latency ... anyone tried the Linux Trace Toolkit?
This will give you exactly how much time is spent in any system
call and what the kernel does during that system call. Total system
overhead is below 1.5% for non-fs intensive systems. This effectively
closes the gprof loop-hoole and reports micro-second accurate system
call / trap / irq information and much more ... Take a look for yourself
LTT's home page:
http://www.info.polymtl.ca/~karym/trace
> calls are not sampled, but I am not sure I like it. High-latency
> system calls can have a big effect on an application's performance.
> If your application is slow because it makes too many slow system
> calls, there does not seem to be any easy way to find that out; the
> results of gprof can even be misleading.
===================================================
Karim Yaghmour
[EMAIL PROTECTED]
Operating System Consultant
(Linux kernel, real-time and distributed systems)
===================================================
