Raul Miller schreef:
> gcd=: *./@x:
> gcd=: ([:x:>./)&.(_&q:) NB. fast version
>
> F=:gcd@:>:@i.%((]+:@x:@[EMAIL PROTECTED]>:*.-:@[<])(-.]-.]&.(p:inv))@(1 
> _1+2^2<[EMAIL PROTECTED]))
>
>
Impressive speed improvement!

Sorry to bother you again, but this still keeps me busy about
the overall correctness (not mentioning this all is based on a
conjecture).

Showing the results of the development of the verb in the
lower range of n:

   allbut2f"0 [4+i.15
2 10 12 84 60 180 180 1980 1980 25740 25740 25740 _ 360360 360360

   allbut2f"0 [4+i.14
2 10 12 84 60 180 180 1980 1980 25740 25740 25740 _ 360360

   allbut2g"0 [4+i.14
2 10 6 42 60 180 180 1980 1980 25740 _ _ _ 360360

   F"0 [4+i.14
2 6 6 42 60 180 180 1980 1980 25740 _ _ _ 360360


- The following verb covers (tries to) all special cases
- Improved performance by stripping ext. precision
  where possible.
- Verbose output: n,d,d+1,min(X)

ab217i=: 3 : 0
 j=: x:<.&.(2&^.)y
 r=. 1+i.y
 s=. 0$0
 p=: j-1
 v1=: lcm r-.j,p
 if. (0<j|v1)*.0<p|v1 do. s =. v1 end.
 if. j<y do.
   q=: j+1
   v2=: lcm r-.j,q
   if. (0<j|v2)*.0<q|v2 do. s =. v1<.v2 end.
   if. v1<:v2 do. s=. p,j,s else. s=.j,q,s end.
 else.
   s=. p,j,s
 end.
 y,s
)

   ,.ab217i"0 [4+i.14
 4  3  4      2
 5  4  5      6
 6  4  5      6
 7  4  5     42
 8  7  8     60
 9  7  8    180
10  7  8    180
11  7  8   1980
12  7  8   1980
13  7  8  25740
14  8  9  60060
15  8  9  60060
16 15 16      0
17 16 17 360360


checking all divisor pairs for n=14
   (,>:)2+I. 0<*/"1(]|[:*./(1x+i.14)-."1])"1(,>:)"0[2x+i.12
8 9

And this is the special case of divisors I mentioned before:
  
   2^m+1 = prime^2   (2^3+1 = 3^2)

A conclusion could be: in case of divisors 3,4 ; 4,5
(n=5,6) and 7,8 ; 8,9 (n=9..13) you have two X's, in all
other cases, if there's a X for a certain n, you have just
one X. ???

**Things to investigate:
Are there solutions that give an X for divisors other than
2^m and 2^m+/-1, with 2^m+/-1 is prime and n>14?

IOW are there more possibilities like 2^m+/-1 = prime^p
and do they give a solution for X.




=@@i



----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to