[Haskell-cafe] benchmarking pure code

[Paul Brauner]

Hello,

actually I don't know if I can. I totally wouldn't mind but this is
mainly my co-author work and I don't know if he would (I suppose not but
since he is sleeping right now I can't check). However let's assume it's
a deBruijn representation for instance, I can tell you the scheme I
used:

data Term = Lam Term | App Term Term | Var Int

instance NFData where
  rnf (Lam t)     = rnf t
  rnf (App t1 t2) = rnf t1 `seq` rnf t2
  rnf (Var x)     = rnf x

the actual datatype doesn't have fancy stuff like higher-order
types for constructors, it's really similar. The only difference
is that it is a GADT, but this souldn't change anything right?

Did I make some mistake in instancing NFData ?

Regards,
Paul

On Wed, Mar 31, 2010 at 09:32:29AM +0200, Bas van Dijk wrote:
> On Wed, Mar 31, 2010 at 9:17 AM, Paul Brauner <paul.brauner at loria.fr> wrote:
> > Does anyone have an idea why calling rnf before the bench
> > doesn't seem to "cache" the result as calling show does?
> > (my instances of NFData follow the scheme described in strictbench
> > documentation).
> 
> Is it possible you could show us your term type and your NFData instance?
> 
> regards,
> 
> Bas