> > That reminds me of the fact that we should probably include cpphs in
> > our GHC from HEAD...
>
> - test performance first
Yes, good point. Compiled with nhc98, the heap profile is virtually
flat, i.e. live heapsize is constant around 8kb, for any size of input
file. (Actually, I reckon it grows linearly in the number of symbols
#defined, but overall their contribution is usually very slight.)
With ghc, the heap profile seems very jagged, but the basic underlying
trend is flat as far as I can tell. The jagged peaks of : applications
reach about 80kb, but everything else is down around 5kb. I suspect
there is some kind of buffering effect in ghc's IO monad that makes
the heap usage of strings uneven, and since the trend is flat, I'm
not too worried.
Time-wise, on my 2.4GHz Linux machine, 1Mb of text input is processed
in the following times:
ghc-5.04.3 1.21s
ghc-5.04.3 -O2 0.74s
ghc-6.2.1 1.25s
ghc-6.2.1 -O2 0.77s
cpp -traditional 0.05s
Regards,
Malcolm
_______________________________________________
Cvs-ghc mailing list
[EMAIL PROTECTED]
http://www.haskell.org/mailman/listinfo/cvs-ghc
