Re: Minimum integer values

[Scott Robert Ladd <coyote at coyotegulch dot com>]

Tobias Schlüter wrote:
> Scott Robert Ladd wrote:
> |
> | Looking at arith.c, I find this comment:
> |
> |       mpz_init (int_info->min_int);
> |       mpz_neg (int_info->min_int, int_info->huge);
> |       /* No -1 here, because the representation is symmetric.  */
> |
> |
> | Yet I can't find anything in the Fortran 95 standard that supports the
> | contention that potitive and negative values are symmetric (as suggested
> | by the comment.
> |
> | So is this a bug or not? It certainly is different from other Fortran 95
> | compilers...
>
> See section 13.7.1. This is a fairly strict interpretation of the
> standard if I recall all the details correctly.

From a platform-independence standpoint, 13.7.1 makes sense; however, 
it does pose some definite problems in regard to interoperability 
between languages.

In my code, I use an "all bits set" mask in a bitwise calculation; under 
the strict interpretation of 13.7.1, this is not possible with Fortran 
on a two's complement system, since 0xFFFFFFFF lies outside the defined 
boundaries of a 32-bit Fortran integer.

In addition, other languages (and the processor itself) accept 
-2,147,483,648 is a valid value for a 32-bit signed integer on a two's 
complement platform. This incompatibility could lead to problems with 
code compiled by other Fortran compilers, or with values supplied 
directly or indirectly from code compiled from other programming languages.

Given that other compilers accept this silently, I think there should at 
least be a switch to control strict adherence to 13.7.1 (or section 13.4 
in the Oct 2003 final committee draft of the new standard).

--
Scott Robert Ladd
Coyote Gulch Productions (http://www.coyotegulch.com)
Software Invention for High-Performance Computing