Yes, you can, and you can also do lots of other tricks. It was decided in 1989 that it was not worth it. That was the right decision then, and it's the right decision now, even more so because space is less of a premium.
Remember: we are talking about packing dimensions (elements of the shape) from words into bytes (or even bits?). ----- Original Message ----- From: Oleg Kobchenko <[email protected]> Date: Friday, October 23, 2009 13:03 Subject: Re: [Jgeneral] Advantages of using J in 64 bit OS To: General forum <[email protected]> > Can't you pack bytes into words? > Possibly without the high bytes to avoid negative dimension. > > > > > From: Roger Hui <[email protected]> > > > > In the header of an array, there is one word for the rank > > and one word (not byte) per dimension, plus a small, fixed > > number of words for miscellaneous items. So theoretically > > the maximum rank array in J64 is 8 %~ (_1+2^63) - f+1. > > > > > > > > ----- Original Message ----- > > From: Dan Bron > > Date: Friday, October 23, 2009 11:39 > > Subject: Re: [Jgeneral] Advantages of using J in 64 bit OS > > To: General forum > > > > > Roger wrote: > > > > > http://www.jsoftware.com/jwiki/Essays/Covert%20Representation> > > > > Neat! > > > > > > Theoretically speaking, what is the largest rank an array > could > > > have in > > > 64-bit J, if I had "sufficient memory"? > (Assuming > > > all the dimensions are > > > 0 or 1.) > > > > > > Is the structure of a 64-bit J header still "one machine > word" > > > for the rank > > > of the array, and then one byte per dimension? If so, > then > > > is the maximum > > > possible rank equal to the maximum value of a signed > integer, > > > i.e _1+2^63 > > > ? > > > > > > I'm just asking because with my rank 2^28 example, I can > "hide" > > > ~30MB of > > > data, and I'm wondering how much more I could > (theoretically) hide. ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
