Re: optimization/5017: i386 floating point optimization

Re: optimization/5017: i386 floating point optimization

[rodrigc]

Synopsis: i386 floating point optimization

State-Changed-From-To: open->closed
State-Changed-By: rodrigc
State-Changed-When: Wed Dec  5 20:14:20 2001
State-Changed-Why:
    Compile your code and add the -Wall flag.  This will
    give you additional warnings that will point you to the
    error in your code.
    
    If you add #include <math.h> at the top of your
    file, the problem goes away.
    
    Since you did not include the function prototypes
    for those functions, the compiler and linker
    make some guesses (which are incorrect) as to the
    signatures of the math functions which you are
    using.

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&pr=5017&database=gcc

Re: optimization/5017: i386 floating point optimization

[rodrigc]

The following reply was made to PR optimization/5017; it has been noted by GNATS.

From: rodrigc@gcc.gnu.org
To: gcc-bugs@gcc.gnu.org, gcc-gnats@gcc.gnu.org, gcc-prs@gcc.gnu.org,
  ghh@mpq.mpg.de, nobody@gcc.gnu.org
Cc:  
Subject: Re: optimization/5017: i386 floating point optimization
Date: 6 Dec 2001 04:14:21 -0000

 Synopsis: i386 floating point optimization
 
 State-Changed-From-To: open->closed
 State-Changed-By: rodrigc
 State-Changed-When: Wed Dec  5 20:14:20 2001
 State-Changed-Why:
     Compile your code and add the -Wall flag.  This will
     give you additional warnings that will point you to the
     error in your code.
     
     If you add #include <math.h> at the top of your
     file, the problem goes away.
     
     Since you did not include the function prototypes
     for those functions, the compiler and linker
     make some guesses (which are incorrect) as to the
     signatures of the math functions which you are
     using.
 
 http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&pr=5017&database=gcc

optimization/5017: i386 floating point optimization

[ghh]

>Number:         5017
>Category:       optimization
>Synopsis:       i386 floating point optimization
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    unassigned
>State:          open
>Class:          wrong-code
>Submitter-Id:   net
>Arrival-Date:   Wed Dec 05 04:46:00 PST 2001
>Closed-Date:
>Last-Modified:
>Originator:     Gerhard Heinzel
>Release:        3.0.2
>Organization:
Albert Einstein Institut Hannover, Germany
>Environment:
System: Linux lp27 2.4.4-4GB #1 Wed May 16 00:37:55 GMT 2001 i686 unknown
Architecture: i686

	
host: i686-pc-linux-gnu
build: i686-pc-linux-gnu
target: i686-pc-linux-gnu
configured with: ../configure 
>Description:
The included small test program fails when compiled with
gcc -O3 -o mybes mybes.c -lm
It works with -O0.
indication of failure is the following output.
The first line starting with "5.00" is garbage.
There should not be any "nan"

4.00  18      4.0944813915730677499e-11          11.301921952136330496
4.00  19      4.3099804121821765789e-12          11.301921952136330496
4.00  20      4.3099804121821765788e-13          11.301921952136330496
4.00  21      4.1047432496973110274e-14          11.301921952136330496
4.00  22      3.7315847724521009341e-15          11.301921952136330496
4 11.301921952136330773 6.62406158447265625
5.00   1                            nan                              1
5.00   2                            nan                            nan
5.00   3                            nan                            nan
5.00   4                            nan                            nan
5.00   5                            nan                            nan

>How-To-Repeat:

#include <stdio.h>
double
mybesi0 (double x)
{
  int k = 0;
  long double sum = 0;
  long double add = 1;
  long double x2 = x / 2.;

  if (x == 0)
    return 1;
  if (x < 0)
    x = -x;
  if (x > 50)
    {
      fprintf (stderr, "argument to mybesi0 too large:%g\n", x);
      exit (1);
    }
  add = 1;
  for (;;)
    {
      sum += add * add;
      add /= (++k);
      add *= x2;
      printf("%4.2f %3d %30.20Lg %30.20Lg\n",x,k,add,sum);
      if (add / sum < 1e-15)
	break;
    }
  return sum;
}

float
bessi0 (float x)
{
  float ax, ans;
  double y;

  if ((ax = fabs (x)) < 3.75)
    {
      y = x / 3.75;
      y *= y;
      ans = 1.0 + y * (3.5156229 + y * (3.0899424 + y * (1.2067492
		  + y * (0.2659732 + y * (0.360768e-1 + y * 0.45813e-2)))));
    }
  else
    {
      y = 3.75 / ax;
      ans = (exp (ax) / sqrt (ax)) * (0.39894228 + y * (0.1328592e-1
		   + y * (0.225319e-2 + y * (-0.157565e-2 + y * (0.916281e-2
	       + y * (-0.2057706e-1 + y * (0.2635537e-1 + y * (-0.1647633e-1
						  + y * 0.392377e-2))))))));
    }
  return ans;
}

void
main (void)
{
  double x;
  for (x = 0; x <= 50; x += 1)
   printf ("%.20g %.20g %.20g\n", x, mybesi0 (x), bessi0(x)); 
}


>Fix:
>Release-Note:
>Audit-Trail:
>Unformatted: