On Oct 26, 2006, at 10:27 AM, Elad Efrat wrote:

> Jason Thorpe wrote:
>
>> What is the performance impact of "in there, but not enabled"?
>
> Okay, I ran lmbench multiple times on my GENERIC+DIAGNOSTIC+LOCKDEBUG
> kernel. The "no-veriexec" results are when Veriexec options are not  
> even
> compiled in the kernel. The "veriexec-nop" results are when Veriexec  
> is
> compiled in, but not used.

How about results for just GENERIC?  LOCKDEBUG, especially, is going  
to skew the results.

>
> no-veriexec-1:Process fork+exit: 1300.6000 microseconds
> no-veriexec-1:Process fork+execve: 2558.3333 microseconds
> no-veriexec-1:Process fork+/bin/sh -c: 5557.0000 microseconds
>
> veriexec-nop-1:Process fork+exit: 1281.6000 microseconds
> veriexec-nop-1:Process fork+execve: 2588.3333 microseconds
> veriexec-nop-1:Process fork+/bin/sh -c: 5583.0000 microseconds
>
> no-veriexec-2:Process fork+exit: 1296.6000 microseconds
> no-veriexec-2:Process fork+execve: 2566.3333 microseconds
> no-veriexec-2:Process fork+/bin/sh -c: 5563.0000 microseconds
>
> veriexec-nop-2:Process fork+exit: 1283.6000 microseconds
> veriexec-nop-2:Process fork+execve: 2601.3333 microseconds
> veriexec-nop-2:Process fork+/bin/sh -c: 5610.0000 microseconds
>
> no-veriexec-3:Process fork+exit: 1298.6000 microseconds
> no-veriexec-3:Process fork+execve: 2566.3333 microseconds
> no-veriexec-3:Process fork+/bin/sh -c: 5567.0000 microseconds
>
> veriexec-nop-3:Process fork+exit: 1300.4000 microseconds
> veriexec-nop-3:Process fork+execve: 2584.6667 microseconds
> veriexec-nop-3:Process fork+/bin/sh -c: 5609.0000 microseconds
>
> -e.
>
> -- 
> Elad Efrat

-- thorpej