I have observed a puzzling timing result which may be CPU dependent.
Please try the following and tell me what ratio you get, and the kind of
machine you ran it on. Thanks.
On my machine, an XPS L521X Intel Core i7-3632QM @ 2.2 GHz,
x=: 1e6 $ ' '
y=: a. {~ ? 1e6 $ 256
%/ 100 (6!:2)&> '/:~x' ; '/:~y'
2.44133
Basically, sorting a constant vector of 1-byte ints takes 2.4 times as long
as sorting a random vector of 1-byte ints.
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm