I don't see anything in the dictionary that states how the processing of
elements of arrays is ordered. It looks to me like the dictionary allows for
parallel processing already. That is good. If we, when programming in J,
avoid making any assumptions based on the way J works today, then in the
future, if J does implement parallel processing, you and whoever, the
implementers of the J interpreter, have more freedom on how it is done.
On Dec 18, 2007 11:16 AM, Roger Hui <[EMAIL PROTECTED]> wrote:
> What is needed in theory is a description of how a given implementation
> of J differs from the dictionary. The only implementation that currently
> exists is for a serial machine. At some point in the future there may
> be an implementation for a parallel machine. Presumably the
> two descriptions will differ in places.
>
> Somewhere in the description there will be a section on evaluation order.
> What the dictionary permits for the monad u"r is
>
> rankmonad=: 2 : 0
> z=. a: $~ m=. */$ t=. <"n y
> for_i. ?~m do. z=. z i}~ u&.> i{,t end.
> >($t)$z
> )
>
> What the current serial implementation does is:
>
> rankmonad=: 2 : 0
> z=. a: $~ m=. */$ t=. <"n y
> for_i. i.m do. z=. z i}~ u&.> i{,t end.
> >($t)$z
> )
>
>
>
> ----- Original Message -----
> From: Don Guinn <[EMAIL PROTECTED]>
> Date: Tuesday, December 18, 2007 5:08
> Subject: Re: [Jprogramming] Aiming at a multi-core future
> To: Programming forum <[email protected]>
>
> > So is there or is there not a problem with f and h running in
> > either order
> > or in parallel?
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
>
>
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
