From: apm233m@vaxc.cc.monash.edu.au Subject: Re: AMD Cyrix chips etc. Date: 1 Aug 1992 14:16:56 +1000
In article <joef.712572287@extro.ucc.su.OZ.AU>, joef@extro.ucc.su.OZ.AU () writes:
>
> I've seen a few posts [and mailed a few back] about problems with the AMD
> 386 chips and their FPU exception's. Since I saw 3 messages abouth them when
> I just had a read through I think I might as well post it to the net for
> world-wide flaming by people who own 386/40's.
>
> ....
> .... When you
> are asking the AMD chip to do a FPU instruction you can get delays of up
> to _5_ milliseconds before the chip returns an error so that the Software
> FPU emulation can kick in. This [unfortunatly] means that AMD chips go at
> about the speed of a 8mhz XT .....
> ........
> ---
> Disclaimer: My machine is a 386/33 - with an AMD chip w/o FPU. ....
I don't understand. Are you saying that this is a problem just with 40MHz
parts or just with 33MHz parts... or both?
My machine has a 33MHZ AMD 386 cpu and an OPTI chipset. I have written an
80387 emulator which works with djgpp and I have done quite a bit of work in
trying to get reasonable speed. Whenever I have checked the details of timings,
I find that I can account for most of the cycles taken to emulate any function.
The user-program-level performance of my emulator is: less than 100 microsec
for basic arithmetic, less than one millisec for any other function.
Maybe the timing problem you have found is due to the support chipset? Or
perhaps the BIOS has not been told that an 80387 is not installed?
My only disappointment with the AMD 386 (so far) was finding that it has the
popad bug :-(
-- =============== Bill Metzenthen Mathematics Department Monash University Australia