Xref: utzoo comp.unix.xenix.sco:1118 comp.lang.c:34876
Path: utzoo!attcan!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!uunet!wshb!michaelb
From: michaelb@wshb.csms.com ( WSHB Operations Eng)
Newsgroups: comp.unix.xenix.sco,comp.lang.c
Subject: Problems compiling the ABC posting in c.u.sources
Keywords: floating point exception
Message-ID: <914@wshb.csms.com>
Date: 21 Dec 90 19:51:40 GMT
Followup-To: comp.unix.xenix.sco
Organization: WSHB, Christian Science Mon. Synd., Cypress Creek, SC
Lines: 125


Help,

I'm trying to compile the ABC program which was just posted to c.u.sources
but I can't get past a part of it.

My setup is an AST 20 MHz 386 with 9 meg memory, SCO Xenix 2.3.2, Development
package release 2.3.

When I try to do 'make depend' there is a mkconfig.c program which gets
compiled and run to set up the config.h file. This program core dumps
with this:

sh: 6797 Floating point exception - core dumped

I have run the program and it is dying while trying to calculate
the maximum possible number on my box. Here are the relevant sections of code.


=====================================================
begin code fragement that blows up
=====================================================

/* These routines are intended to defeat any attempt at optimisation */
Dstore(a, b) double a, *b; { *b=a; }
double Dsum(a, b) double a, b; { double r; Dstore(a+b, &r); return (r); }
double Ddiff(a, b) double a, b; { double r; Dstore(a-b, &r); return (r); }
double Dmul(a, b) double a, b; { double r; Dstore(a*b, &r); return (r); }
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
This is the line which blows up.  Line 38.
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
double Ddiv(a, b) double a, b; { double r; Dstore(a/b, &r); return (r); }

double power(x, n) int x, n; {
	double r=1.0;
	for (;n>0; n--) r*=x;
	return r;
}

int floor_log(base, x) int base; double x; { /* return floor(log base(x)) */
	int r=0;
	while (x>=base) { r++; x/=base; }
	return r;
}

int ceil_log(base, x) int base; double x; {
	int r=0;
	while (x>1.0) { r++; x/=base; }
	return r;
}

============================================================================
end of the offending section
============================================================================

The preceeding function crashed with this from sdb

Dmul(1.3482698511467e+308,2.0000000000000), line 38


============================================================================
beginning of code fragment which calls Dmul function
============================================================================

/****** Minimum exponent ***************************************************/
	if (setjmp(lab)==0) /* in case of underflow trap */
		do {
			xminner=y;
			y=Ddiv(y,base);
			a=Dmul(y,1.0);
			if ((Dsum(a,a) == 0.0) || (absval(y) >= xminner)) break;
		} while (1);

	if (setjmp(lab)!=0) { printf("Unexpected over/underflow\n"); exit(1); }

	if (xminner != 0.0 && xminner != xmin) {
		printf("/\* The smallest numbers are not kept normalised *\/\n");
		printf("/\* Smallest unnormalised positive number=%g *\/\n",
			xminner);
	}

/****** Maximum exponent ***************************************************/
	maxexp=2; xmax=1.0; newxmax=base+1.0;
	if (setjmp(lab) == 0) {
		while (xmax<newxmax) {
			xmax=newxmax;
			newxmax=Dmul(newxmax, base);
!!!!!!!!!!              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^          !!!!!!!!!!
          ((This is the line sdb says calls the final straw))
!!!!!!!!!!                                                    !!!!!!!!!!
			if (Ddiv(newxmax, base) != xmax) break; /* ieee infinity */
			maxexp++;
		}
	}
	if (setjmp(lab)!=0) { printf("Unexpected over/underflow\n"); exit(1); }

	printf("/\* Maximum exponent=%d *\/\n", maxexp);

/****** Largest and smallest numbers ************************************/
	xmax=Ddiff(1.0, epsneg);
	if (Dmul(xmax,1.0) != xmax) xmax=Ddiff(1.0, Dmul(base,epsneg));
	for (i=1; i<=maxexp; i++) xmax=Dmul(xmax, base);
	printf("/\* Maximum number=%g *\/\n", xmax);

============================================================================
end of code fragment
============================================================================


Everything seems to run fine through the Minimum exponent section as I get the
'Smallest unnormalised positive number' report. It core dumps immediatly
after that. I don't understand is what is different about the Dmul
function in the Maximun exponent call. It seems to work correctly several 
times before the program gets to that point.

Does anyone have any pointers? If so, please send me an answer which will 
be meaningful to a knowledgable amateur. I don't do 'c' for a living. 
I just get things working.

Michael
-- 
Michael Batchelor--Systems/Operations Engineer #compliments and complaints
WSHB - An International Broadcast Station of   #   letterbox@csms.com
 The Christian Science Monitor Syndicate, Inc. #technical questions and reports
michaelb@wshb.csms.com         +1 803 625 4880 #   letterbox-tech@csms.com