FC4 kernel performance

[Malita, Florin]

On Tue, 2005-06-21 at 11:54 -0400, Colin Walters wrote:
> Ah, but is Unixbench a relevant benchmark for a desktop? 

Probably not as relevant as for servers, but I'm sure the kernel does
make a difference. I would assume I/O performance affects application
startup time a whole lot more than process creation.

On a server though, that 62% (taken with a grain of salt until I get the
2.6.12-rc6 results) is really worrying me.

> I'm skeptical that (apart from preemption) having e.g. a "desktop" 
> kernel would be useful.

How about the kernel debug options - shouldn't that be a special-purpose
kernel?