On Aug 5, 2007, at 8:52 PM, Simon Burge wrote:

> Andrew Doran wrote:
>
>> There's a global flag that's tested (ktrace_on) before any function  
>> calls
>> are made. In the common case (tracing off) it should be marginally  
>> cheaper.
>
> Have you measured this with say lat_syscall from lmbench?

for platforms that have syscall_intern and a plain/fancy syscall entry  
point, how would lat_syscall show any difference?

>
>
> Cheers,
> Simon.

-- thorpej