RM = Raul Miller, DB = me, RH = Roger Hui

DB>           %floating_zero
DB>        _

RM>  But % is not fixed.

I only included  %  to demonstrate that I'd successfully created a negative 
zero.  The point of my message was the line above this.

RH>  You can not win. 

Au contraire.  It is you who are doomed.

RH>  You have to use primitives. 

You have to use memory.

RH>  If all the primitives take care to 
RH>  eradicate minus zero, 

That was exactly the point; there are an infinite number of ways for me to 
manipulate memory to create a negative zero.  I could even do it from outside 
of J.

Our debate hinges on the definition of "have", used in the sentence:

DB>  J will still have a negative zero 

I define "have" to mean "the bit pattern for an IEEE floating point (FP) 
negative zero exists in an array of FP in J".  

You appear to define it as "any Z that can be distinguised from  0  using  J  
primitives,  such that y = y+Z  where  -. y e. 0,Z  ".

So our discussion is moot.

RH>  I estimate that it would take me less than an hour 
RH>  to build a version of J that can eradicate
RH>  minus_zero, if I really wanted to.

I estimate it would take me less than an hour to find a relevant bug in it, if 
I really wanted to.

-Dan
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to