GHC's internals

the equations inside a Rec binding are mutually recursive (and hence there are as few of them as possible)? Olaf -- OLAF CHITIL, Lehrstuhl fuer Informatik II, RWTH Aachen, 52056 Aachen, Germany Tel: (+49/0)241/80-21212; Fax: (+49/0)241/-217 URL: http://www-i2

Notes on Adding an Optimisation Pass to GHC

pages http://www-i2.informatik.rwth-aachen.de/~chitil/PUBLICATIONS/extendGHC.html (not to be confused with Will Partain's document of similar title) Cheers, Olaf -- OLAF CHITIL, Lehrstuhl fuer Informatik II, RWTH Aachen, 52056 Aachen, Germany Tel: (+49/0)241/80-21212; Fax: (+49/0)241

Strange time behaviour without optimisations

time0.02s ( 0.07s elapsed) MUT time2.79s ( 2.99s elapsed) GCtime2.13s ( 2.31s elapsed) Total time4.94s ( 5.37s elapsed) Olaf -- OLAF CHITIL, Lehrstuhl fuer Informatik II, RWTH Aachen, 52056 Aachen, Germany Tel: (+49/0)241/80-21212; Fax: (+49/0)241

problems with nofib benchmarks

of `anna' is completely different from the given `big.sum.out' (with unmodified ghc-2.08). I obtain: 1279649 5. For `make runtests' the program `symalg' requires more heap on my sparc station. It works with `-H70'. Greetings, Olaf -- OLAF CHITIL, Lehrstuhl fuer Informatik II, RWTH Aachen

Re: Bus Error from ghc-4.01 binary on Solaris 2.6

naries. We are running SunOS 5.5 Generic sun4m and use gcc version 2.8.1. -- OLAF CHITIL, Lehrstuhl fuer Informatik II, RWTH Aachen, 52056 Aachen, Germany Tel: (+49/0)241/80-21212; Fax: (+49/0)241/-217 URL: http://www-i2.informatik.rwth-aachen.de/~chitil/

Re: ghc 4.01 solaris binary dist: bus error on Hello World!

gcc 2.7.2. If you have that under the name gcc-2.7.2 you can ask ghc to call this version instead of gcc by writing: ghc -pgmcgcc-2.7.2 -o Test Test.hs -- OLAF CHITIL, Lehrstuhl fuer Informatik II, RWTH Aachen, 52056 Aachen, Germany Tel: (+49/0)241/80-21212; Fax: (+49/0)24

Hugs error messages

or message "Program error: erroneous term" only unqualified names are used. This can be very confusing when the error occurs inside an imported module, the top level module does not import some names of the erroneous term, and possibly the names are even defined differently in the top lev

Re: Predicativity

a problem with using an explicitly typed calculus in a compiler. GHC should probably use hash-consing. Any type view tool based on final types (instead of principal typings like mine) faces the same problem. Olaf -- OLAF CHITIL, Dept. of Computer Science, The University of York, York YO10 5DD