Re: Fw: maybey an error in mpf_cmp

[Paul Zimmermann]

> Indeed.  But in your case, the numbers were slightly different.
> The mpf_cmp function will find numbers equal that are equal even
> if they are not canonically represented,

Looking more closely at the example from Toine Bommelijn, there is clearly
a bug, either in mpf_cmp or in mpf_out_str, since two numbers with the same
binary representation (according to mpf_out_str) are said different by mpf_cmp:

lucrezia(zimmerma): ./a.out 2
x=0.101000111101011100001010001111010111000010100011110101110000101001e-3
y=0.101000111101011100001010001111010111000010100011110101110000101001e-3
cmp (x, y) = -1

Paul

#include <stdio.h>
#include "gmp.h"

int
main(int argc, char *argv[])
{
  mpf_t x, y;
  int iprec = atoi(argv[1]);

  mpf_init2 (x, iprec);
  mpf_init2 (y, iprec);

  mpf_set_ui (x, 1);
  mpf_div_ui (x, x, 12);
  mpf_mul_ui (x, x, 100);
  mpf_floor (x, x);
  mpf_div_ui (x, x, 100);

  mpf_set_str (y, "0.08", 10);

  printf ("x=");
  mpf_out_str (stdout, 2, x->_mp_size * mp_bits_per_limb, x);
  printf ("\n");

  printf ("y=");
  mpf_out_str (stdout, 2, y->_mp_size * mp_bits_per_limb, y);
  printf ("\n");

  printf ("cmp (x, y) = %d\n", mpf_cmp (x, y));

  mpf_clear (x);
  mpf_clear (y);

  return 0;
}