search

pycket / pycket-bench

Benchmarking pycket against some Schemes
6 stars 9 forks source link

Gambit optimiziation as suggested #5

Open krono opened 10 years ago

krono commented 10 years ago

Quotes from a mail by Marc Feeley:

Some declarations, such as (fixnum) and (flonum), indicate the type of numbers being used in arithmetic functions (was a similar type declaration used with the other systems?).

I used (declare (not interrupts-enabled) (standard-bindings) (fixnum) (not safe) (inlining-limit 0)) and gcc -O2.

If you keep an “8 bits per char” representation, at least compile Gambit with --enable-char-size=1 .

[@samth]

Sixth, we didn't use (not interrupts-enabled) -- in our upcoming revised submission, we will hopefully also support raising an exception in Pycket on Ctrl-C, which is Racket's behavior. If we don't have this, we'll make sure to use that option.

samth commented 10 years ago

We should just fix strings and fix ctrl-c, and then not change the benchmarks. But we should make sure to revisit this for the paper.

krono commented 10 years ago

I would include

samth commented 9 years ago

There's still an issue about ctrl-c. @cfbolz what do you predict for fixing this by the deadline?

cfbolz commented 9 years ago

I think at this point it looks somewhat unlikely that I'll manage ctrl-c. So let's disable it in Gambit and explain in the paper?

samth commented 9 years ago

Ok. We should also investigate this for other Schemes. We can't change Racket's behavior, so we'll just have a slight unfair advantage there.

krono commented 9 years ago

I had to reenable interrupts, regardless of pycket's state. Some benchmarks do not run without that and I want a level playing ground for all benches.