Re: Re[2]: Bookworm Adventures Postmortem

[Brent Fulgham <bfulg@...>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1


On Mar 18, 2007, at 3:51 AM, Gé Weijers wrote:

> On Mar 18, 2007, at 4:39 PM, Lothar Scholz wrote:
>
> > I must say i think the same same. I like the programming model but it
> > is very slow and i really don't know why the design of lua wants to
> > use slow hashtable access anywhere. My performance tests show that  
> > for
> > function calls there is a lot (i really mean a lot) room for
> > optimization.
>
> Warning for those reading this message: "There are lies, damn lies,  
> and benchmarks"
>
> [...]

> I also implemented the same function in a number of other  
> languages. On my PowerPC Mac laptop I obtained the following results:
>
> 	C: 97301516   (optimization flag: -O3)
> 	C: 34003218   (no optimization)
> 	Lua: 3119095
> 	Python: 1077014
> 	Perl: 453289
>
> Relative to Lua's performance:
>
> 	C: 31.20   (optimized)
> 	C: 10.90   (unoptimized)
> 	Lua: 1.00
> 	Python: 0.35
> 	Perl: 0.15
>
> Lua is quite a bit faster than the common interpreted languages I  
> tried. The optimized C performance number is unrealistically high  
> because the compiler performs a kind of inline substitution to  
> lower the number of recursive calls made. If your turn that off the  
> speed advantage relative to Lua drops to 10.9.
>
> It seems to me that Lua is doing very well relative to the  
> established scripting languages, at least in the function call  
> department.

For another take on this topic, consider http:// 
shootout.alioth.debian.org/.  In my tests, Lua is faster than Python  
and Perl in almost every category.  The places where it is slower, is  
probably due to poor Lua test implementations, an not Lua itself.

- -Brent
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (Darwin)

iD8DBQFF/eQ+zGDdrzfvUpURAgE7AJ41hhsU18m+XwwxlimvB/8ElzZcugCeMdk4
9BqTogY5nfj7VOYjBXM58cU=
=Lq/g
-----END PGP SIGNATURE-----