Hi Aai, That is very peculiar. There is nothing in the patches that would speed up either of those two factorials.
Can you tell us more information. What was the original and patched OS and cpu architecture? What was the compiler and flags used for the patched version? If you compile yourself with and without those patches the speed on the corresponding factorial implementations should be the same. On Tue, Jun 4, 2013 at 6:31 AM, Aai <[email protected]> wrote: > After having applied the patches for J701 from > http://www.2bestsystems.com/**foundation/j/<http://www.2bestsystems.com/foundation/j/> > > and playing with a divide and conquer version of factorial > > > facf=: fac43 @(1x&,)`1:@.(2>]) > > fac43 =: (($:@,&{. * $:@,~&{:) (,>:)@<.@-:@+/)`{.@.({.={:) > > > it noticed the following timings for facf and !: > > > original J7 engine: > > timer '! 30000x' > > 19.9206 > > timer 'facf 30000x' > > 11.8613 > > > > patched J7 engine: > > timer '! 30000x' > > 7.25327 > > timer 'facf 30000x' > > 3.71744 > > > FYI > > > > -- > Met vriendelijke groet, > @@i = Arie Groeneveld > > ------------------------------**------------------------------**---------- > For information about J forums see > http://www.jsoftware.com/**forums.htm<http://www.jsoftware.com/forums.htm> > ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
