Actually ts was all confusing, because ts actually
makes two calls, and the first is space, so the
time part already runs on cached data.
Yes, you are right, p j, the self version makes
"less" use of cache.
fib2=: (+&$: <:)@<:`1:@.(<&2) defm
6!:2 'fib2"0 i.5 6'
6.06564
fib3=: (+&fib3 <:)@<:`1:@.(<&2) defm
6!:2 'fib3"0 i.5 6'
0.00466844
--- Oleg Kobchenko <[EMAIL PROTECTED]> wrote:
> For some reason the subsequent definitions
> with name as opposed to $: just take
> more memory on the first call than previous definition.
> Otherwise, they use a fresh cache just after the
> definition (not reusing previous calls of the same name).
>
> fib2=: (+&fib2 <:)@<:`1:@.(<&2) defm NB. first time
> 5 ts'fib2 25'
> 5.43169e_5 16256
> 5 ts'fib2 25'
> 4.86037e_5 4160
>
> fib2=: (+&fib2 <:)@<:`1:@.(<&2) defm NB. second time
> 5 ts'fib2 25'
> 5.16009e_5 23040
> 5 ts'fib2 25'
> 5.32539e_5 4160
>
> fib3=: (+&fib3 <:)@<:`1:@.(<&2) defm NB. new name time
> 5 ts'fib3 25'
> 5.19172e_5 54016
> 5 ts'fib3 25'
> 5.24279e_5 4160
>
> fib3=: (+&$: <:)@<:`1:@.(<&2) defm NB. Self recurse
> 5 ts'fib3 25'
> 5.20134e_5 16256
> 5 ts'fib3 25'
> 4.97499e_5 4160
>
>
>
> --- p j <[EMAIL PROTECTED]> wrote:
>
> > Another note, if you use the verb name instead of $:
> > in the definition, then the function uses the cached
> > data when it recurses.
> >
> > fib4=: (+&fib4 <:)@<:`1:@.(<&2) defm
> >
> > --- Oleg Kobchenko <[EMAIL PROTECTED]> wrote:
> >
> > > NB.
> > >
> > =========================================================
> > > coclass 'pmem'
> > >
> > > create=: 3 : 0
> > > KEY=: ''
> > > VAL=: ''
> > > )
> > > destroy=: codestroy
> > > find=: 3 : 0
> > > if. (#KEY)>i=.KEY i.boxopen y do.i{::VAL else. ''
> > > end.
> > > )
> > > set=: 4 : 0
> > > KEY=: KEY,boxopen x
> > > VAL=: VAL,boxopen y
> > > )
> > > cocurrent 'base'
> > >
> > > mv=: 2 : 0
> > > if. 0=#r=. find__n y do.y set__n r=.u y end. r
> > > )
> > >
> > > defm=: 1 : 0
> > > u mv (''conew'pmem')
> > > )
> > > NB.
> > >
> > =========================================================
> > >
> > >
> > > fib1=: (+&$: <:)@<:`1:@.(<&2)
> > > fib2=: (+&$: <:)@<:`1:@.(<&2) defm
> > >
> > > fib2"0 i.20
> > > 1 1 2 3 5 8 13 21 34 55 89 144 233 377 610 987 1597
> > > 2584 4181 6765
> > >
> > > (fib2 (-: , ]) fib1) 25
> > > 1 121393
> > >
> > >
> > > 5 ts'fib1 25'
> > > 0.343449 14720
> > >
> > > 5 ts'fib2 25'
> > > 5.1219e_5 4160
> > >
> > > 0.343449 % 5.1219e_5
> > > 6705.5
> > >
> > >
> > >
> > >
> > > --- p j <[EMAIL PROTECTED]> wrote:
> > >
> > > > define an adverb A such that:
> > > >
> > > > f A = 'f'
> > > >
> > > > that is for any verb f, f A returns the string
> > > that
> > > > names f.
> > > >
> > > > If this is currently impossible, is there a way to
> > > use
> > > > the 5!: functions such that f A generates a unique
> > > > string for any value of f.
> > > >
> > > > A =: 1 : '5!:5 < ''u.''' NB. generates a string
> > > for
> > > > the functionality of u. instead of the name of u.
> > > >
> > > > If none of this can be done, has anyone come up
> > > with a
> > > > method to memoize results of a verb?
> > > >
> > >
> > http://aspn.activestate.com/ASPN/Cookbook/Python/Recipe/325905
> > > >
__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
