Simon, you wrote
> I made some improvements to the Integer representation in the current
> sources of GHC (which you can get from CVS), based on some patches by Marc
> Van Dongen.
What is mean CVS ?
"Current sources" is not an official release, right?
For i would rather wait for the official release.
> Integer performance is now close to Int for small integers
> (minBound :: Int <= x <= maxBound :: Int), and about the same as before for
> large Integers.
This is a very helpful thing.
Here follows certain benchmark for Int vs Integer performance
- to run with Z = Integer, then with Z = Int.
If it shows the time ratio less then 2, then i would gladly ignore
Int.
------------------
Sergey Mechveliani
[EMAIL PROTECTED]
------------------------------------------------------------------
type Z = Integer -- Int
main = -- compute extended gcd for many x,y and sum the results
let
(d,n,m) = (40,5000,10000) :: (Z,Z,Z)
ns = [n..(n+d)]
ms = [m..(m+d)]
pairs = [(x,y)| x<-ns, y<-ms]
tripls = map (\ (x,y)->(x,y,gcdE x y)) pairs
s = sum (map (\ (_,_,(d,u,v))->d+u-v) tripls)
-- boo = all test tripls -- tests gcdE
in
putStr (shows s "\n")
-- gcdE x y -> (d,u,v): d = gcd(x,y) = u*x + v*y
gcdE :: Integral a => a -> a -> (a,a,a)
gcdE 0 y = (y,0,1)
gcdE x y = g (1,0,x) (0,1,y)
where
g (u1,u2,u3) (v1,v2,v3) =
if v3==0 then (u3,u1,u2)
else
case quotRem u3 v3
of
(q,r) -> g (v1,v2,v3) (u1-q*v1, u2-q*v2, r)
test (x,y,(d,u,v)) = d==(u*x+v*y) && d==(gcd x y)
-------------------------------------------------------------------
