💾 Archived View for mirrors.apple2.org.za › archive › ground.icaen.uiowa.edu › Collections › 1WSW › … captured on 2024-12-18 at 01:01:28.
-=-=-=-=-=-=-
<HTML>
<HEAD>
<TITLE>FPE - NC reference page</TITLE>
</HEAD>
<BODY>
Dirk Froehling may be able to advise those interested in ordering and/or
obtaining<BR>
information on the newly available (custom order) Number Cruncher (NC) card
from<BR>
Andres Schmidt. Please note Dirk Froehlings email address follows. <BR>
<BR>
Email: dirk@gaga.maschinenbau.uni-dortmund.de<BR>
<HR WIDTH="50%" SIZE="9"><BR>
See references below - for current Usenet comments on the; FP, FPU and NC<BR>
cards.<BR>
<BR>
(1) "Andreas Schmidt's Number Cruncher are both real coprocs that get
their jobs<BR>
done. The FPE is suffering from a major problem, namely the coproc is crashing<BR>
internally and has to be reset in software. This happens in a non deterministic<BR>
way, and software written for that engineering junk must be adapted to that.
:-)<BR>
The Number Cruncher is compatible with the FPE but is actually what the
FPE was<BR>
supposed to be - a math coproc that works. It perfoms very well.<BR>
<BR>
Applied Engineering once made a coproc. PULL IT FAST! this little card is
only<BR>
sucking power. It's of so limited use that it would make a tremendous doorstop.<BR>
Actually, I have done some testing with it and found it would not accelerate
math<BR>
on a IIgs (*compared with SANE).<BR>
<BR>
Most software will profit from the FPE or Number Cruncher via the software<BR>
patches. A few software titles including one weird math application communicate<BR>
with the coproc directly which can result in a 10% gain..."<BR>
<BR>
<BR>
(2) "Depends on what software you installed. If this is the original
FPE and you<BR>
have the SANE patch installed (more on that later), then all programs that
do<BR>
floating-point through SANE will be accelerated. The original FPE has a
few bugs<BR>
in hardware, though, and the SANE patch is also buggy. If you have the NC<BR>
(NumberCruncher) card, then the hardware bugs are gone, the SANE patch is
the<BR>
same as for the FPE, though. If it's some other card, I don't think it will<BR>
accelerate much.<BR>
<BR>
Regarding bugs: I don't know _which_ bugs the FPE had exactly, I just know
there<BR>
were some (one was suceptibility to heat). The problem of a buggy SANE patch
will<BR>
go away shortly, one guy whom I won't name yet (he can do that himself)
has<BR>
rewritten the patch to be free of all known bugs.<BR>
<BR>
There are also a few programs that access the FPCP (Floating Point Co-Processor)<BR>
directly, must notably GSymbolix (Eureka!, now), which is an _excellent_
maths<BR>
package."<BR>
<BR>
<BR>
(3) "It depends exactly which card it is. If it is a "Number
Cruncher" or an<BR>
Innovative Systems "FPE" card (both of which use the Motorola
68881<BR>
FPU), then any program which uses the IIgs SANE toolset (Standard Apple<BR>
Numerics Environment) will be able to automatically make use of the<BR>
card, PROVIDED you have the necessary patch software installed.<BR>
<BR>
This produces a reasonable speed improvement in floating point<BR>
calculations. For a considerable increase in speed, some programs are<BR>
written to use the card directly (avoiding the overhead of calling<BR>
SANE). ORCA/C and ORCA/Pascal include support for this (a special<BR>
version of the program must be compiled, and it won't work if the card<BR>
is not present).<BR>
<BR>
In theory, the card could also be used by programs which use the 8-bit<BR>
implementation of SANE (e.g. the AppleWorks Spreadsheet module). This<BR>
would require specific patches for each such application.<BR>
<BR>
Applesoft BASIC doesn't use SANE, and will not use the card (unless a<BR>
program is specially written to use it, which would require machine<BR>
language support, and probably wouldn't be worth the effort, since you'd<BR>
have to keep converting floating point numbers between SANE and<BR>
Applesoft formats).<BR>
<BR>
If the card is not either of these types, then I don't know what the<BR>
situation is regarding software that supports it."<BR>
<BR>
--------------------------------------------------------------------------------<BR>
<BR>
References; See Usenet Newgroup articles listed below. <BR>
<BR>
(1)<BR>
Subject: Re: Math co-processor? What a heck?!<BR>
From: GUDATH@EZINFO.VMSMAIL.ETHZ.CH (Henrik 'Ratte' Gudat)<BR>
Date: 1997/01/25<BR>
Message-Id: <5cdem0$9qd@elna.ethz.ch><BR>
Distribution: world<BR>
X-News-Reader: VMS NEWS 1.24<BR>
References: <5ca1kn$6i6@nntp.hut.fi><BR>
Organization: Synergetix<BR>
Newsgroups: comp.sys.apple2<BR>
<BR>
(2)<BR>
Subject: Re: Math co-processor? What a heck?!<BR>
From: sbehrens@arrowweb.com (Sönke Behrens)<BR>
Date: 1997/01/25<BR>
Message-Id: <32e9d4b7.48002307@PersonalNews.Germany.EU.net><BR>
References: <5ca1kn$6i6@nntp.hut.fi><BR>
Content-Type: text/plain; charset=us-Ascii<BR>
Organization: Customer of EUnet Germany; Info: info@Germany.EU.net<BR>
Mime-Version: 1.0<BR>
Newsgroups: comp.sys.apple2<BR>
<BR>
(3)<BR>
Subject: Re: Math co-processor? What a heck?!<BR>
From: dempson@actrix.gen.nz (David Empson)<BR>
Date: 1997/01/25<BR>
Message-Id: <19970125115725210036@dempson.actrix.gen.nz><BR>
Distribution: inet<BR>
References: <5ca1kn$6i6@nntp.hut.fi><BR>
Organization: Empsoft<BR>
Newsgroups comp.sys.apple2<BR>
<HR SIZE="14"><BR>
<BR>
<BR>
</BODY>
</HTML>