[lkml]   [1998]   [Jun]   [1]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: Compatible with i386/UP but optimised for i686/SMP [was Re: test_and_set_bit() not atomic forever?]

> > >The point was that lots of people are running SMP kernels on UP machines,
> > >where it's gratuitous overhead.
> >
> > Then they shouldn't do that.
> Unless they're a distributor who wants to run one kernel per
> architecture, + modules.

You cannot avoid significant slowdown with SMP kernel on UP machines
due to extra locking needed.

> Or a sysadmin, in a situation where managing different kernels for that
> extra bit of efficiency isn't worth the administrative overhead.

When you need to have a single kernel for lots of similar machines,
they are usually UP, so compiling a SMP kernel doesn't make sense.

> These points apply also to the i386/i486/i586/i686 optimisation issue.
> Is it worth making a kernel nearly optimised for i686 but compatible
> with i386? (Perhaps using fixups in the same way to blank out calls to
> do the "verify put_user" type stuff, vs. blanking out flush_tlb and so
> forth?)

It would make sense to distribute three kernels:

- i386

- i586

- i586 SMP

Have a nice fortnight
Martin `MJ' Mares <>
Faculty of Math and Physics, Charles University, Prague, Czech Rep., Earth
"The computer is mightier than the pen, the sword, and usually, the programmer."

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to

 \ /
  Last update: 2005-03-22 13:42    [W:0.057 / U:3.108 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site