Some senseless babbling from Jonathan De Boyne Pollard to Murray Lesser
on 11-22-99  09:45 about The 80486 "requirement"...

 ML> According to the books, it takes more (minimum) RAM to run Warp 4
 ML> than it did to run Warp 3 (but 20 MB should be more than sufficient)
 ML> and at least a 80486 chip (which you have), [...]

 JDBP> An 80486 (or equivalent, such as a AMD or Cyrix 5x86) is not a
 JDBP> *requirement*, however.  It would be for OS/2 SMP, but it isn't for the
 JDBP> uniprocessor kernel.  Apparently, from the discussions that ensued when
 JDBP> this first became known several years ago, the 80486 "requirement" is
 JDBP> nothing more than the fact that IBM switched on the "optimise for the
 JDBP> 486" flag to the compiler.  Previously it had been optimising for the
 JDBP> 80386.  When one generates code optimised for the 486, compilers
 JDBP> schedule instructions slightly differently (to better use the
 JDBP> pipelining that is in the 80486), but they don't actually use any
 JDBP> instructions that aren't also available on the 80386.  (There aren't
 JDBP> any, really, that are of any practical use outside of certain very
 JDBP> specialised areas like SMP.) 

The real reason for the 486 "minimum" requirement was to disinherit the
responsibility of supporting 386-based machines, from what I've heard.  Not
only does it alleviate some pressure on their support department, but it
removes an entire generation of machines from their testing department,
since they simply don't do any testing on 386 machines (officially,
anyway).

Mike Ruskai
thannymeister@yahoo.com


... Americans call it fast food because it speeds them to the grave.

___ Blue Wave/QWK v2.20
--- Platinum Xpress/Win/Wildcat5! v3.0pr2
114/477
143/1