Hello Juergen,
As of svn revision 931, I observe the following for greatest
common divisor calculations
¯3J0 ∨ ¯3J0
2
¯1J0 ∨ ¯3J0
2
¯1J0 ∨ ¯1J0
0
2 does not evenly divide ¯3J0 or ¯1J0 and 0 does not divide.
Please find attached APL code that can be used to identify more
bad results if needed.
Regards,
Fred
⎕pw ← 196
⍝ n is the number of real and imaginary part elements tested, needs to be odd
n ← 7
⍝ r is a vector of the real part elements centered around 0
r ← ( - ⌈ n ÷ 2 ) + ⍳ n
⍝ c is a vector of complex numbers generated from r
c ← , r ∘.+ 0J1 × r
⍝ gcd is matrix of gcd's for all possible pairings of the elements c
gcd ← c ∘.∨ c
" "
⍝ Test the elements on each row or column of gcd are modulo 0 with the
⍝ arguments used to generate it.
∨ / ∨ / mod ← gcd | arg ← ( ( n n ) ⋆ 2 ) ⍴ c
⍝ Count the gcd elements that are not modulo 0 with their arguments
cnt ← + / bad ← , 0 ≠ mod
⍝ Generate a table of the arguments and bad gcd values.
( cnt 1 ⍴ bad / , arg ) , ( cnt 1 ⍴ bad / , ⍉ arg ) , cnt 1 ⍴ bad / , gcd
1
¯3 ¯3 2
¯1 ¯3 2
¯3 ¯1 2
¯1 ¯1 0
¯3 0 2
¯1 0 0
¯3 2 2
¯1 2 2
¯1 3 3
¯3J0 ∨ ¯3J0
2
¯1J0 ∨ ¯3J0
2
¯1 ∨ ¯1
1
¯1J0 ∨ ¯1J0
0
