We were able to get three identical machines where gentoo, Mandrake and debian were installed for a basic i386/i586/optimised by gentoo (-march=petium3 -pipe -O3) comparison. Nothing fancy, roughly standard installs, of similar configuration (where debian was a mixture of various versions) so that if you run the same application, on the same data using the same command, you expect to see an indication which system is faster. cpu's were celeries.
My problem is that gentoo seems to be trailing in performance (sometimes by as much as 1/3 behind debian!). Opening the same large spreadsheet on each machine is slower on gentoo, gimp scripts are slower etc. Tests are not super sophisticated, but real world in that the gimp script is used by someone in their day to day work, the spreadsheet is one of mine etc. Mandrake and debian (monolithic, no modules) used the 2.4.21 kernel, against 2.4.20 gentoo-sources with preempt etc (it seems gentoo is behind here?) xfree versions did differ with gentoo being latest 4.3, but that seems an unlikely culprit. The tests I witnessed did hit the swap a lot, due to having only 256M ram, but that was not does not account for all of the difference - should be similar. The only real difference seems to be the modular gentoo-sources 2.4.20 kernel - could that be the problem - but I would not expect applications to be overly affected by the kernel? Items such as dma, disk performance, swap etc are similar. Any suggestions where the cause could be? BillK -- [EMAIL PROTECTED] mailing list
