Date: Wed, 9 Dec 92 11:00:59 CST
From: csaba AT vuse DOT vanderbilt DOT edu (Csaba A. Biegl)
To: djgpp AT sun DOT soe DOT clarkson DOT edu
Subject: Re: on -O and -O2, again

>> 
>> (Stuff deleted)
>> 
>> >int argc;
>> >char **argv;
>> 
>> >{  matrix morg;
>> 
>> >   static int lact, lm, ln;
>> >   ^
>> >   add this here 
>> >	
>> >   for (lact=0; lact <= maxdim; ++lact)
>> >     { printf("lact=%5d ",lact); fflush(stdout);
>> 
>> >       for (lm=0; lm <= 2*lact; ++lm)
>> >         for (ln=0; ln <= 2*lact; ++ln)
>> >           morg[lm][ln] = 1.0;
>> >     }
>> >}
>> >/*-----------------------------------------------------------------------*/
>> 
>> >Add "static" in front of your declaration "int lact, lm, ln" solves
>> >the problem.
>> 
>> Does it solve it, or just make it go away?  (I.e., is there a
>> theoretical basis for this fix?  I don't think so...)
>> 
>> This problem seems somewhat similar to the one I posted some time ago,
>> regarding local variables in nested functions.  That one was a problem
>> without -O or -O2, but went away when either was present.  That one was
>> generally conceded to be real and to be native to 386 based versions of GCC.
>> (I.e., not unique to DJGPP.  Has anyone tested this later version of the
>> orig poster's program on GCC under a 386 Unix?)

I compiled the original (integer variables not 'static') program without
optimization, with -O and with -O2. I ran the executables on five
different systems:

	386/33		8MB RAM
	486/66DX2	8MB RAM
	386/20		4MB (32bit) + 2.5MB (16bit) RAM
	486/33		16MB RAM
	386SX/25	2MB RAM

I used DJGPP 1.09. Out of the five systems four executed all three 
versions of the program without error. The fifth system, the 
486/33 with 16Megs, died when running the -O2 compiled program
with Exception 14. It executed the other two programs without 
error.

Conclusion: Hardware error on some motherboards???


Csaba Biegl
csaba AT vuse DOT vanderbilt DOT edu