[tbnl-devel] TBNL survey / re-org

quasi quasilists at gmail.com
Tue Sep 12 08:54:00 UTC 2006


On 11 Sep 2006, Edi Weitz spake thusly:

>> Unfortunately I could not manage to get SBCL to perform anywhere
>> near what CMUCL could.  My shortcomings, mostly, I guess.
>
> You might see the difference between "green" threads (CMUCL) and OS
> threads (SBCL).  My guess is that thread creation and/or thread
> switching is more expensive on SBCL.  Maybe your tests amplify this
> difference?  Have you tried with other Lisps which use native
> threads?

I am in the process of convincing my CTO to purchase Lispworks.  But
it takes time.  In Indian Rupees, the software cost is a little
high. I have not used any other lisp which uses native threads.  I
tried to profile SBCL, but the statistical profiler used to die.  I
tried asking on #lisp, but I am not an expert lisp programmer and
those guys were a bit of "go read up", which I should.  I think I will
give a shot again when I have more specific questions and understand a
bit more.

We are using lisp software to connect to Abacus air reservation
systems and processing of the data and for all the search logic.  The
good news is our inhouse software is more than 10x faster than the
Java solution purchased from a large software house here and 100x more
stable.

:)

thanks.

-- 
quasi

Utopia Unlimited!



More information about the Tbnl-devel mailing list