[Icfp04-discuss] Re: simulator

[Alan Schmitt]

* Eijiro Sumii (sumii at saul.cis.upenn.edu) wrote:
> From: Alan Schmitt <alan.schmitt at polytechnique.org>
> > We tried ant compilation here (one of the too many things we did instead 
> > of doing ant strategy ;-) ), and we found that it was not that 
> > impressive (something around a 10% gain for a 1.4 s run on a 1.7 GHz 
> > machine).
> 
> How did you compile?  E.g., did you use function pointers for calling
> each step function specialized to a particular instruction?

(I did not write the ant compiler, so I just looked at the code to see 
how it works.) I realize that we did not really write a compiler, as 
what it does is simply load the ant brain in an array as a C file that 
is then compiled (so this just saves us parsing time, basically). We 
were already using function pointers for calling each function (then we 
switched to macros and it went 25% faster).

> 
> Another merit of compiling is, you can implement all the interpretive
> (more complex) part in ML, while benefiting from the speed of C for
> the other (very plain) part.

Definitely.

> 
> Yes, you shouldn't have spent much time on the simulator, anyway...:)

I know :( But it was fun ;-)

Alan Schmitt

-- 
The hacker: someone who figured things out and made something cool happen.
.O.
..O
OOO
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://lists.seas.upenn.edu/pipermail/icfp04-discuss/attachments/20040609/2463a72d/attachment.bin