From: Kbwms AT aol DOT com
Message-ID: <7ea92299.36167b51@aol.com>
Date: Sat, 3 Oct 1998 15:30:25 EDT
To: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
Cc: djgpp-workers AT delorie DOT com
Mime-Version: 1.0
Subject: Re: Troubles Compiling the Test Programs
Content-type: text/plain; charset=US-ASCII
Content-transfer-encoding: 7bit
X-Mailer: AOL 3.0 16-bit for Windows sub 38

Dear Eli Zaretskii,

On 09-28-98 at 05:42:15 EST you wrote:
>
> > Sure.  How many other functions do we have to change?
>
> Seems like e_fmod.c, e_remainder.c (and their float brethren), and
> k_standard.c are the only ones that are the candidates.  Can you
> devise test cases that will see if the problems with -0 happen there
> as well?

Here is a demo of remainder()/fmod() "(and their float brethren)":

Testing remainder() vs. fmod():
   X     Y    remainder(X,Y)   fmod(X,Y)
 -0.25  1.75  -2.500000e-01  -2.500000e-01
 -1.25  1.75   5.000000e-01  -1.250000e+00
 -2.25  1.75  -5.000000e-01  -5.000000e-01
 -3.25  1.75   2.500000e-01  -1.500000e+00
 -4.25  1.75  -7.500000e-01  -7.500000e-01
 -5.25  1.75   0.000000e+00   0.000000e+00 (8000000000000000)
(8000000000000000)
 -6.25  1.75   7.500000e-01  -1.000000e+00
 -7.25  1.75  -2.500000e-01  -2.500000e-01
 -8.25  1.75   5.000000e-01  -1.250000e+00
 -9.25  1.75  -5.000000e-01  -5.000000e-01
-10.25  1.75   2.500000e-01  -1.500000e+00
-11.25  1.75  -7.500000e-01  -7.500000e-01
-12.25  1.75   0.000000e+00   0.000000e+00 (8000000000000000)
(8000000000000000)

Testing remainderf() vs. fmodf():
   X     Y    remainderf(X,Y)  fmodf(X,Y)
 -0.25  1.75  -2.500000e-01  -2.500000e-01
 -1.25  1.75   5.000000e-01  -1.250000e+00
 -2.25  1.75  -5.000000e-01  -5.000000e-01
 -3.25  1.75   2.500000e-01  -1.500000e+00
 -4.25  1.75  -7.500000e-01  -7.500000e-01
 -5.25  1.75   0.000000e+00   0.000000e+00  (80000000)  (80000000)
 -6.25  1.75   7.500000e-01  -1.000000e+00
 -7.25  1.75  -2.500000e-01  -2.500000e-01
 -8.25  1.75   5.000000e-01  -1.250000e+00
 -9.25  1.75  -5.000000e-01  -5.000000e-01
-10.25  1.75   2.500000e-01  -1.500000e+00
-11.25  1.75  -7.500000e-01  -7.500000e-01
-12.25  1.75   0.000000e+00   0.000000e+00  (80000000)  (80000000)

The remainder function seems to be working.

A demo program is appended.


K.B. Williams
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
#include <fdlibm.h>
#include <stdio.h>
#include <math.h>

int
main()
{
    double  X, Y;

    Y = 1.75;

    printf("Testing remainder() vs. fmod():\n");
    printf("   X     Y    remainder(X,Y)   fmod(X,Y)\n");

    for (X = 0.25; X <= 12.25; ++X)
    {
	double  Modulo, Remndr;
	ieee_double_shape_type AnyDbl;

	Remndr = remainder(-X, Y);
	Modulo = fmod(-X, Y);

	printf("%6.2f%6.2f%15e%15e", -X, Y, Remndr, Modulo);
	if (Remndr == 0)
	{
	    AnyDbl.value = Remndr;
	    printf(" (%08x%08x)",
		AnyDbl.parts.msw, AnyDbl.parts.lsw);
	    AnyDbl.value = Modulo;
	    printf(" (%08x%08x)",
		AnyDbl.parts.msw, AnyDbl.parts.lsw);
	}
	printf("\n");
    }
    printf("\nTesting remainderf() vs. fmodf():\n");
    printf("   X     Y    remainderf(X,Y)  fmodf(X,Y)\n");

    for (X = 0.25; X <= 12.25; ++X)
    {
	float  Modulo, Remndr;
	ieee_float_shape_type AnyFlt;

	Remndr = remainderf(-X, Y);
	Modulo = fmodf(-X, Y);

	printf("%6.2f%6.2f%15e%15e", -X, Y, Remndr, Modulo);
	if (Remndr == 0)
	{
	    AnyFlt.value = Remndr;
	    printf("  (%08x)", AnyFlt.word);
	    AnyFlt.value = Modulo;
	    printf("  (%08x)", AnyFlt.word);
	}
	printf("\n");
    }
    return 0;
}