Matthew Dillon dillon at apollo.backplane.com
Wed Jul 23 23:53:17 PDT 2003

:Matt (or others who can answer of course :)),
:I remember talks when 5.x was being made about how the 386 and 486
:(IIRC) didn't have certain essential assembly commands for ease of
:atomic locking.
:How will this affect your ideas?  Will we only go with Pentium and above
:(and likewise CPUs on other platforms) or will it allow to use 386/486
:as well?
:Jeroen Ruigrok van der Werven <asmodai(at)wxs.nl> / asmodai

    I think the argument was more along the lines of whether to support
    'MP' on a 486, because you only need those locking primitives
    (i.e. the cmpxchgl instruction) in MP systems.  The 486 supports
    cmpxchgl, but 486 systems do not have MP tables or APICs, and they
    are god aweful slow which makes it just plain silly to spend any
    effort on MP/486.

    Plain old vanilla UP 486 will continue to be supported, primarily
    because the support doesn't interfere with development.  

    I am not particularly interested in continuing suppor for the 386,
    not when competitor's cpus (like the VIA C3) are at a minimum
    pentium compatible.  That said I see no particular reason to rip
    386 support out for UP (though I don't even know if it still works!
    My %fs globaldata access code might have broken it!), because it
    does not generally interfere with development.

					Matthew Dillon 
					<dillon at xxxxxxxxxxxxx>

More information about the Kernel mailing list