(j3.2006) BEQ, BNE?

[Van Snyder]

On Thu, 2013-01-31 at 17:18 -0600, Bill Long wrote:
> Please site examples, along with which Fortran 2008 compiler supports 
> (or might even remotely contemplate supporting) those processors.

The focus on who is presently manufacturing processors with multiple
representations of zero, and who is making Fortran processors for those
platforms, is misplaced.  As Stan observed, this "my dick is bigger than
yours" contest is tiresome.

The standard clearly contemplates multiple representations of zero.  The
only way the standard presently provides to detect equality or
inequality of bit strings that would represent zero if considered as an
integer are clumsy circumlocutions that disguise the true intent,
thereby increasing everybody's cost.

I don't understand the difficulty to complete the picture of intrinsic
functions providing relational tests for bit strings.

When we replaced the BIT type with integers-as-bits we left a hole.
Markham resolution 6 clearly advocates "removal of ... discrepancies
between existing features."  I consider it to be a discrepancy that one
can construct clearly different bit strings for which there is no simple
way later to determine whether they are different.

An alternative to completing the picture of bit string relational
testing is to enable the SIGN function to detect the sign of integer
zero, a proposition that has repeatedly been defeated.  It would,
furthermore, simply lead (in this case) to yet another clumsy
circumlocution:

  if ( i == j .and. sign(i,1) == sign(j,1) ) ...