From:	abx AT abx DOT art DOT pl (Wlodzimierz ABX Skiba)
Newsgroups:	comp.os.msdos.djgpp
Subject:	Re: FPU
Date:	12 Mar 2003 23:49:57 -0800
Organization:	http://groups.google.com/
Lines:	27
Message-ID:	<e8982227.0303122349.702cfac3@posting.google.com>
References:	<19001N297 AT web2news DOT com>
NNTP-Posting-Host:	195.205.156.51
X-Trace:	posting.google.com 1047541797 20928 127.0.0.1 (13 Mar 2003 07:49:57 GMT)
X-Complaints-To:	groups-abuse AT google DOT com
NNTP-Posting-Date:	13 Mar 2003 07:49:57 GMT
To:	djgpp AT delorie DOT com
DJ-Gateway:	from newsgroup comp.os.msdos.djgpp
Reply-To:	djgpp AT delorie DOT com

"Joel_S" <jbs30000 DOT news DOT invalid AT web2news DOT net> wrote in message news:<19001N297 AT web2news DOT com>...
> Since DJGPP uses AT&T assembly syntax, then that probably means that FPU
> commands would be slightly different than Intel syntax.  Is there a web
> site that goes over using FPU commands in DJGPP?

I also interested in references to FPU handling on the same platform
by different compilers. I recently tested C++ application (MegaPOV)
with set of compilers: OpenWatcom 1.0, DMC 8.32, DJGPP gcc 3.2,
Borland 5.51, Borland 6.0, MinGW gcc 3.2. All were compiled and
started on the same platform - NT4 on Pentium 2. In particular I was
concerned on simple loop with heavily used doubles to control its
execution. I noticed that each compiler caused different number of
iterations (only B5.51==B6.0 which can be expected).

If I write something like

  #define M_PI   3.1415926535897932384626
  double my_PI = M_PI;

should I expect the same internall bit representation of 'my_PI' on
the same platform with mentioned compilers or not (considering all
compilers use the same number of bytes for double) ?

Since it is not DJGPP specific issue (but related) is there any other
forum where this issue can be continued ?

ABX

- Raw text -