I wrote:
> 13 : 'y i. >./ y'
> ] i. >./
>
> rand =: ?~ 3e6
>
> (-: |.)@:(ways`:0) rand NB. Identical results
> 1
>
> '5.2d' 8!:2 (%"1 <./) (6!:2, 7!:2@:])&> ways ,L:0 ' rand' NB.
> Non-identical performance
> 5.06 1.07
> 5.09 1.07
> 1.00 1.00
Sorry, I forgot to include the definition of "ways" (the heart of the example):
w0 =: 13 : 'y i. >./ y'
w1 =: ] i. >./
w2 =: i. >./
ways =: w0`w1`w2
That is, the tacit output of 13 : is significantly slower than the native
tacit code (which is often true of any code generator or
automatic translator). In fact, it is just as slow as the original explicit
code:
w_1 =: 3 : 'y i. >./ y'
w0 =: 13 : 'y i. >./ y'
w1 =: ] i. >./
w2 =: i. >./
ways =: w_1`w0`w1`w2
rand =: ?~ 3e6
(-: |.)@:(ways`:0) rand NB. Identical results
1
'5.2d' 8!:2 (%"1 <./) (6!:2, 7!:2@:])&> ways ,L:0 ' rand' NB.
Non-identical performance
4.47 2.86
4.60 1.07
4.45 1.07
1.00 1.00
Though it does have a space advantage (which surprises me).
-Dan
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
