Henrik Grubbstr?m wrote:
>I've now implemented a codec for use when decoding the master object.
>I've now fixed this issue as well and fixed a few compiler bugs
>along the way (reversing the compiler isn't exactly easy)...
>Loading of the dumped master now seems to work for me:
Preliminary tests confirm this.
As for timings... *** drum roll ***
# time pike -e ''
real 0m0.141s
user 0m0.110s
sys 0m0.010s
Compared to the timings I got in the beginning, this shows that pike now
needs only 20% of the time that it originally took to startup.
Not quite as snappy as Perl yet:
# time perl -e ''
real 0m0.012s
user 0m0.010s
sys 0m0.000s
But, these timings (for Pike) I can live with. I would guess that the
largest difference might be that perl doesn't interpret anything at startup,
it just starts the C-compiled binary; whereas Pike is running master.pike
already. Anyone who knows more about Perl innards who can confirm this?
--
Sincerely,
Stephen R. van den Berg.
Currently unemployed programmer:
"What do you mean, I can't initialize things in an assert()?"
|