[Ecls-list] ECL shootout

[Marko Kocić]

Seems like there are a couple of people interested in ECL performance.

I wonder if maybe we can help Juan by providing some patches if we can make
some perf improvenments by changing / adding functionality to some isolated
parts of code that don't require intimate knowledge of the full
compliler system,
if it is possible.

Regards,
Marko

2009/6/30 Gabriel Dos Reis <gdr at integrable-solutions.net>:
> On Tue, Jun 30, 2009 at 6:01 AM, Marko Kocić<marko.kocic at gmail.com> wrote:
>> And here are updated results with unithreaded ECL. Seems like threads
>> add big overhead to benchmarks.
>
> I would love to be able to translate those to speed ups when building
> OpenAxiom  http://www.open-axiom.org/ with ECL.
> The build-time with ECL (2h34min) has consistenly trailed SBCL (33min),
> and GCL (35min) in that order.
>
> Also, recently, I got a very curious report about some testcase, which
> I ultimately traced to be dependent on the Lisp implementation.
> Detailed numbers are at
>
> http://sourceforge.net/tracker/?func=detail&aid=2809350&group_id=203172&atid=984524
>
> One thing to note is that OpenAxiom makes uses of lot of indirect
> calls, and calls
> to sequence generic functions.  I know, Juanjo has improved indirect calls
> recently -- still they seem at least two times slower than GCL.
>
> -- Gaby
>