On Tue, May 01, 2001 at 08:50:51PM +0200, Jaromír Dolecek wrote:
> Sure, more data could never hurt :)
> 
> BTW, it would be quite useful if people tried full hbench benchmark,
> which tests the pipe bandwidth for different buffer sizes. I'd
> really like to find out if the 64KB buffer size is optimal on all archs
> or other archs "like" other values (and hence if PIPE_CHUNK could
> be MI, or it should be machine dependant).

Do you know if hbench requires perl for this pipe tests ?
I don't want to install perl on my older machines :)

--
Manuel Bouyer, LIP6, Universite Paris VI.           Manuel.Bouyer@lip6.fr
--