> Cool stuff, but you might want to compare to this example in PyPy:
>
> http://morepypy.blogspot.com/2011/04/tutorial-part-2-adding-jit.html
>
> If you get numbers, please share them here!
Ah, I see! I missed the link to the tutorial that explained out to
generate the JITed versions of example1.py through example5.py. Ok,
so I'll use
$ python ./pypy/pypy/translator/goal/translate.py example5.py
to produce a custom BF evaluator through PyPy, and use that for my
future tests. I checked out a clone of pypy this afternoon.
Here are the numbers I see from the PyPy tutorial's evaluator for prime.b.
$ echo 100 | time ./example5-c prime.b
Primes up to: 2 3 5 7 11 13 17 19 23 29 31 37 41 43 47 53 59 61 67
71 73 79 83 89 97
1.16user 0.00system 0:01.21elapsed 96%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (0major+777minor)pagefaults 0swaps
My local bf language for Racket performs almost identically on this workload.
fermi ~/work/bf/examples $ raco make prime.rkt && (echo 100 | time
racket prime.rkt)
Primes up to: 2 3 5 7 11 13 17 19 23 29 31 37 41 43 47 53 59 61 67
71 73 79 83 89 97
1.15user 0.06system 0:01.35elapsed 89%CPU (0avgtext+0avgdata 0maxresident)k
120inputs+0outputs (0major+10365minor)pagefaults 0swaps
--------------------------------------------------------------------------------------------------------------
Under the mandelbrot benchmark, PyPy produces the following numbers:
$ time ./example5-c mandel.b
[after it produces the fractal...]
real 1m9.891s
user 1m9.128s
sys 0m0.612s
A run with my bf language produces:
$ raco make mandel.rkt && time racket mandel.rkt
[after it produces the fractal...]
real 0m48.837s
user 0m47.435s
sys 0m0.384s
As soon as I finish the rest of my tutorial, I'll update the reported
numbers accordingly.
_________________________________________________
For list-related administrative tasks:
http://lists.racket-lang.org/listinfo/users
