Hi Sigbjorn,
thanks for your advice. Your fixes solved the problems.
I just came across a new one. `make runtests' in directory `nofib' just gives
==nofib== nofib: time to link nofib follows...
Enter your Haskell program, end with ^D (on a line of its own):
Maybe I just accidently messed up the recursive decent in `mk/target.mk' when
adding your fixes? However, `make all' makes the measurements correctly.
Some further thoughts on nofib:
I want to compare programs that are compiled with two completely different
compiler options. Hence I use way_a and way_b. It would be nice if I could avoid
having to compile all programs also the normal way (and having to keep the
code).
I'd like to measure heap residency. The information given by the standard run is
not useful, because there are too few samples. I use
SRC_RUNTEST_OPTS += +RTS -j10k -RTS
to cause more major garbage collections to gain better residency information.
However, different programs require different settings of -jnumber to obtain a
result in reasonable time.
It would be nice to have something like `make testresidency' to gain all
residency information automatically.
Finally it would be a good idea to extend `target.mk' for the use of the
spixtools. It seems to be quite impossible to obtain precise run-time
measurements without it. (You could add a pointer to http://sw.sun.com/shade/
to the profiling section of the user guide).
I intend to tackle the last two points when I have a better understanding of how
the whole make system works.
Olaf
--
OLAF CHITIL, Lehrstuhl fuer Informatik II, RWTH Aachen, 52056 Aachen, Germany
Tel: (+49/0)241/80-21212; Fax: (+49/0)241/8888-217
URL: http://www-i2.informatik.rwth-aachen.de/~chitil/
