Re: [TRIVIAL] kstrdup

[Alan Cox]

On Sad, 2003-04-19 at 05:48, Jeff Garzik wrote:
> > Completely disagree.  Write the most straightforward code possible,
> > and then if there proves to be a problem, optimize.  Optimizations
> > where there's no actual performance problem should be left to the
> > compiler.
> 
> Since the kernel does its own string ops, the compiler does not have 
> enough information to deduce that further optimization is possible.
> 
> 
> > Case in point: gcc-3.2 on -O2 on Intel is one instruction longer for
> > your version.
> 
> And?  It's still slower.

You are arguing over a 1 instruction, probably sub 1 clock scheduling
matter on a call which is not used on any fast or common path. If you
shaved 1 clock off the timer handling instead you'd make a lot more
difference..