1. Shootout [ATTN: Brent W. Hughes]

Hey, Brent. Are you still lurking out there somewhere? Are you still 
interested in working on the shootout?

-- 
==============================
Too many freaks, not enough circuses.
j.

new topic     » topic index » view message » categorize

2. Re: Shootout [ATTN: Brent W. Hughes]

Yes, Jason, I have been lurking occasionally but I've been involved in some
other things lately that have kept me away from full activity and I may have
missed some postings.  At any rate, let me post some thoughts on the Shootout. 
Maybe you folks are way ahead of me on this, but here goes anyway.

I think we should not go to the official Shootout people until we have at least
15 of the benchmarks implemented.  I suggest we have our own mini-shootout with
the contestants being Euphoria, C/C++, and Python (comparing us to one fast
language and one interpreted language).  Perhaps Rob could post a link to the
results of our mini-shootout on the Euphoria home page.

To this end I would suggest having three types of people involved in our
shootout:
1. A coordinator
2. Implementers
3. Testers

The coordinator would maintain a list of the algorithms together with the names
of those who have implemented them, or checked them out for implementing, or
whether they're still up for grabs.  For example:

Algorithm		Implementer
=============== =============== 
Ackermann		Derek Parnell
Word-Count		Jason Gade
Sieve of Eras.	Robert Craig
Fannkuch		Jason Gade*
Harmonic		----
Nested loops	----
...				...				...
An asterisk means it is checked out but not yet completed.

Implementers would "check out" algorithms from the cooridinator.  When finished
writing the code, the implemeners would hand it off to the coordinator (and post
in on Euforum so everyone can witness the progress of the shootout and maybe
encourage others to join in on the fun).  The coordinator would update his list
and pass the code off to the testers.

Each platform would have a separate tester.  Each tester would have the latest
in Euphoria code and would also have the C and Python code for all of the
algorithms which they can obtain from the official Shootout pages, namely:

The Great Win32 Computer Language Shootout
http://dada.perl.it/shootout/

The Computer Language Shootout Benchmarks
http://shootout.alioth.debian.org/great/index.php?sort=fullc

The testers would return the results to the coordinator who would join together
the results from all of the testers and update our own shootout web site.

After we have at least 15 algorithms finished, we could submit them to the
official Shootout sites.  But even if they won't take them for whatever reason,
we still have our mini-shootout site with all the glory that represents.

Thoughts, anyone?  Volunteers?


Brent

new topic     » goto parent     » topic index » view message » categorize

3. Re: Shootout [ATTN: Brent W. Hughes]

Brent W. Hughes wrote:
> 
> Yes, Jason, I have been lurking occasionally but I've been involved in some
> other things
> lately that have kept me away from full activity and I may have missed some
> postings.
>  At any rate, let me post some thoughts on the Shootout.  Maybe you folks are
>  way ahead
> of me on this, but here goes anyway.
> 
> I think we should not go to the official Shootout people until we have at
> least 15
> of the benchmarks implemented.  

Agreed.

> I suggest we have our own mini-shootout with the contestants
> being Euphoria, C/C++, and Python (comparing us to one fast language and one
> interpreted
> language).  Perhaps Rob could post a link to the results of our mini-shootout
> on the
> Euphoria home page.

Yes, it could be good advertising.

My original thought was to compare the following:
Euphoria 2.4
Euphoria 2.5 interpreted
Euphoria 2.5 translated
eu.ex interpreted
eu.ex translated
C

My thought was to use one of the C programs such as sieve as the timing standard
and rate the other contestants in those units.

When it comes to comparing to other languages, I would like to compare against
C, Python, and Qu. smile

> 
> To this end I would suggest having three types of people involved in our
> shootout:
> 1. A coordinator
> 2. Implementers
> 3. Testers
> 
> The coordinator would maintain a list of the algorithms together with the
> names of
> those who have implemented them, or checked them out for implementing, or
> whether they're
> still up for grabs.  For example:
> 
> Algorithm		Implementer
> =============== =============== 
> Ackermann		Derek Parnell
> Word-Count		Jason Gade
> Sieve of Eras.	Robert Craig
> Fannkuch		Jason Gade*
> Harmonic		----
> Nested loops	----
> ...				...				...
> An asterisk means it is checked out but not yet completed.
> 
> Implementers would "check out" algorithms from the cooridinator.  When
> finished writing
> the code, the implemeners would hand it off to the coordinator (and post in on
> Euforum
> so everyone can witness the progress of the shootout and maybe encourage
> others to
> join in on the fun).  The coordinator would update his list and pass the code
> off to
> the testers.

I don't think it needs to be that formal. Its not like we're writing a standard
library or anything blink

I need to zip up what I have so far and put it in the archive. If someone else
writes a piece they can just send it to me and I'll add it in.

One big challenge that awaits me is how to measure the results? I installed
Linux partially for the purpose of this project, I wonder if I could get the test
framework used by the shootout people.

If not, I would write one in Euphoria.

If a test framework wouldn't work out right, then each benchmark would have to
report its own time at the end of its run.

Another thing, will all programs that run on Linux require the addition of
include dll.e
-- code
free_console()

?

<snip>

> Thoughts, anyone?  Volunteers?
> 
> 
> Brent
> 


=====================================
Too many freaks, not enough circuses.

j.

new topic     » goto parent     » topic index » view message » categorize

4. Re: Shootout [ATTN: Brent W. Hughes]

Anyway, I have the fannkuch done and random benchmarks done.

I did the random benchmark as a prelude to doing the fasta benchmark. But I
think I'll skip fasta for now.

I've been trying to think of how to do the dispatch benchmark.

=====================================
Too many freaks, not enough circuses.

j.

new topic     » goto parent     » topic index » view message » categorize

5. Re: Shootout [ATTN: Brent W. Hughes]

>Jason said:
>I don't think it needs to be that formal. Its not like we're writing a standard
>library
>or anything blink

Yes, I think I got a little carried away there.  Besides, it looks you're the
coordinator, the implementers, and the testers all in one!

>Jason said:
>I wonder if I could get the test framework used by
>the shootout people.

I think I read that their test framework used Perl.  I sure they'd be willing to
share, but maybe you'd just as soon do it in Euphoria, anyway.  Maybe someone
else reading this has some thoughts to share about building the test framework in
Euphoria.

>Jason said:
>I need to zip up what I have so far and put it in the archive. If someone else
>writes
>a piece they can just send it to me and I'll add it in.

Agreed.

Brent

new topic     » goto parent     » topic index » view message » categorize

Search



Quick Links

User menu

Not signed in.

Misc Menu