Date:	Wed, 17 Nov 1999 13:16:25 +0200 (IST)
From:	Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
X-Sender:	eliz AT is
To:	Surahyo <surahyo AT pgrad DOT unimelb DOT edu DOT au>
cc:	djgpp AT delorie DOT com
Subject:	Re: math.h or math co-processor?
In-Reply-To:	<4.1.19991117155637.00a4b240@mail.student.unimelb.edu.au>
Message-ID:	<Pine.SUN.3.91.991117131033.17269C-100000@is>
MIME-Version:	1.0
Reply-To:	djgpp AT delorie DOT com
X-Mailing-List:	djgpp AT delorie DOT com
X-Unsubscribes-To:	listserv AT delorie DOT com

On Wed, 17 Nov 1999, Surahyo wrote:

> My system is  386SX-40, no math co-processor, 4 MB SIMM RAM, 
> and 16 MB Flash Disk (the controller for vehicle navigation system).
> 
> I used GCC ver 2.6.3 (gcc263bn.zip) since the latest one doesn't work
> on my system. 
> I've put "set GO32=emu c:\bin\emu387" in my autoexec.bat for 387 emulation.
> 
> I compiled the next source code on my system using the next command:
>     gcc model.c -o model -lm
> and no errors.
> However, I got a wrong result when I run it.

This is a known problem: several trig functions, such as atan, acos and 
asin, give inaccurate results near 90 degrees.  These problems will be 
corrected in DJGPP v2.03.

If you don't have any other solution than to use DJGPP v1.x, try to 
rewrite your code so that it doesn't invoke acos to compute angles near 
90 degrees.  Then the results will be more accurate.

Alternatively, write your own acos function using one of the available 
approximation methods.

> Anyway, I am still hoping to use the latest GCC on my system.
> However, every time I run it, my system reboot automatically.
> I've followed the advice in section 3.9 of the FAQ for the memory-starved 
> systems.

Did you change CWSDPMI parameter "Minimum application memory desired 
before 640K paging" parameter to 512K or larger (using CWSPARAM)?  If 
not, please try it.

Also, what does "mem /c" print on your 386SX?

- Raw text -