i386 and InterlockedIncrement inline asm

[Nicholas Wourms]

cgf@redhat.com wrote:
> On Sat, Jan 25, 2003 at 08:44:26AM +1100, Danny Smith wrote:
> 
>>Although I haven't find any bug reports about running cygwin exes on
>>i386 (nor do I know of anyone with an i386 box to test), it could be a
>>problem.  Has this potential problem been acknowledged anywhere?
> 
> 
> Yes.  It's been brought up.  I've been waiting to see if there are any
> complaints and so far I'm not aware of any.
> 

I know that Win95a *barely* runs at all on a 386/dx 33, so I 
can seriously say that it is highly improbable that anyone 
would run cygwin on that platform.  Even if it does run, it 
would be so slow that you couldn't use it in any productive 
manner.  Good grief, imagine how long it would take to run a 
configure script!  Since cygwin doesn't run on WinCE, the 
NT/Alpha and NT/PPC ports being pretty much dead, and the 
fact that it won't run on 3.X pretty much underscores my 
assertion.  The grey area might be in some of the early 
non-intel processors/processor upgrades.

Since we're on the subject, Chuck and I were discussing the 
floating point routines in newlib a few months back.  FWICT, 
we are currently using Software FPU emulation.  The reason 
we were discussing this was because we had noted how poorly 
sh performs when it does FP division (used by libtool when 
calculating the max number of arguments that can be passed 
to the command line).  NT seems to have a much better handle 
on this than does Win9X/ME, but it is still slower then true 
HW FPU.  If we do decide to start cutting out earlier 
platforms, how about enabling Hardware FPU?  Surely this 
would provide a performance boost?

Cheers,
Nicholas