On Dec 16, 2005, at 4:58, Joshua Isom wrote:
I just finished three more shoot outs. Two are rather simple, a
floating point version of ack,
Great, thanks. A comment wrt takfp:
ubstitution. Anyway, the floating point takfp is slow, 364 seconds
for me, which makes it really really slow. The python's the fasted of
perl, tcl, python, ruby, and php. Parrot's not too well.
You din't write how you are testing. Anyway:
* always use svn head
* make -s clean && perl Configure.pl --optimize && make # [1]
* find the appropriate runcore option that is fastest for the benchmark
(-j, -C)
* try to run the compiled program too (should help for very fast
benches)
./parrot -o foo.pbc foo.pir
time ./parrot foo.pbc N
time ./parrot -C takfp.pir 50.1s
time python takfp.py 1m7.3s
Athlon X2 @2000, 512K cache.
[1] There are some asserts in the code, which can slow down an
unoptimized parrot by several magnitudes of times. Usually optimized is
~40 % faster.
leo