Preliminary benchmark:
f=: 256&|@>:&.(a.&i.)
g=: 256&|@:>:&.(a.&i.)
h=: 123&(22 b.)&.(a.&i.)
x=: a.{~1e6 [EMAIL PROTECTED] 256
ts=: 6!:2 , 7!:[EMAIL PROTECTED]
ts 'Expression'
Expression J6.02 J6.01 Ratio
f x 0.004869 1.05344e6 0.634145 9.43834e6 130.25 8.96
g x 0.004700 1.05344e6 0.061089 1.25841e7 13.01 11.95
h x 0.005483 1.05344e6 0.030379 9.43834e6 5.54 8.96
----- Original Message -----
From: Roger Hui <[EMAIL PROTECTED]>
Date: Wednesday, October 17, 2007 7:45
Subject: Re: [Jprogramming] extension to m b.
To: Programming forum <[email protected]>
> Special code for u&.(a.&i.) should prove
> fruitful. For example:
>
> a=: a.{~x=: 1e5 [EMAIL PROTECTED] 256
> b=: a.{~y=: 1e5 [EMAIL PROTECTED] 256
>
> ts 'a 256&|@:+&.(a.&i.) b'
> 0.00533559 2.10118e6
> ts 'x+y'
> 0.00140465 524864
>
> I should have seen it when I did the special code
> for p {~ q i. ] ,
> compared to which u&.(a.&i.) has the
> advantage of being usable
> for monads or dyads.
>
>
>
> ----- Original Message -----
> From: Roger Hui <[EMAIL PROTECTED]>
> Date: Wednesday, October 3, 2007 9:45
> Subject: [Jprogramming] extension to m b.
> To: Programming forum <[email protected]>
>
> > I propose to extend the bitwise operations m
> > b. for m e. 16+i.19
> > to work on character (8-bit or 16-bit) arguments. e.g.
> >
> > 'abc' 23 b. '321'
> > srs
> > ] x=:'dazlious','chthonic',:'patriots'
> > dazlious
> > chthonic
> > patriots
> > 22 b./ x
> > whzvonhc
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
