From: ········@gmail.com
Subject: About graphics and big amounts of data.
Date: 
Message-ID: <1168617655.240538.181940@s34g2000cwa.googlegroups.com>
I've got a task: filter significant amount of simple data (big picture)
with specific filters. Usually I used C/C++ mixed with assembler, but
eventually I decided to try Lisp. Can anybody recommend me some
literature / already completed programs / something else in this
direction?
And, by the way, is there any sense to try to avoid mixing with other
languages like C (for fast interface to target data, for example)?

Tia.

From: Ken Tilton
Subject: Re: About graphics and big amounts of data.
Date: 
Message-ID: <vyPph.280$3w7.166@newsfe12.lga>
········@gmail.com wrote:
> I've got a task: filter significant amount of simple data (big picture)
> with specific filters. Usually I used C/C++ mixed with assembler, but
> eventually I decided to try Lisp. Can anybody recommend me some
> literature / already completed programs / something else in this
> direction?
> And, by the way, is there any sense to try to avoid mixing with other
> languages like C (for fast interface to target data, for example)?

No. You can fuss with Lisp declarations to improve the performance of 
lisp on tasks like this, but Lisp does not have much (any?) value to add 
to a straightforward task like filtering a kazillion bits all the same 
way. Since you know C and how to work with it outside C (in asm), 
learning FFI is not all that hard and probably the easiest way to go. 
CFFI is the best portable FFI solution.

kt

-- 
The Dalai Lama gets the same crap all the time.
   -- Kenny Tilton on c.l.l when accused of immodesty
From: Lars Rune Nøstdal
Subject: Re: About graphics and big amounts of data.
Date: 
Message-ID: <pan.2007.01.12.19.02.49.842711@gmail.com>
On Fri, 12 Jan 2007 08:00:55 -0800, alkhimov wrote:

> I've got a task: filter significant amount of simple data (big picture)
> with specific filters. Usually I used C/C++ mixed with assembler, but
> eventually I decided to try Lisp. Can anybody recommend me some
> literature / already completed programs / something else in this
> direction?
> And, by the way, is there any sense to try to avoid mixing with other
> languages like C (for fast interface to target data, for example)?
> 
> Tia.

I saw this a while back:
  http://www.lrde.epita.fr/cgi-bin/twiki/view/Publications/200606-IMECS

-- 
Lars Rune Nøstdal
http://nostdal.org/
From: Ken Tilton
Subject: Re: About graphics and big amounts of data.
Date: 
Message-ID: <7sRph.162$rG4.111@newsfe11.lga>
Lars Rune Nøstdal wrote:
> On Fri, 12 Jan 2007 08:00:55 -0800, alkhimov wrote:
> 
> 
>>I've got a task: filter significant amount of simple data (big picture)
>>with specific filters. Usually I used C/C++ mixed with assembler, but
>>eventually I decided to try Lisp. Can anybody recommend me some
>>literature / already completed programs / something else in this
>>direction?
>>And, by the way, is there any sense to try to avoid mixing with other
>>languages like C (for fast interface to target data, for example)?
>>
>>Tia.
> 
> 
> I saw this a while back:
>   http://www.lrde.epita.fr/cgi-bin/twiki/view/Publications/200606-IMECS
> 

Cool. I saved a copy for future ref. Never saw anything good on 
optimizing Lisp -- PAIP had some, IIRC. Well, I guess thx to google 
groups I would just look for the periodic thread where Lisp code gets 
taken from 100x slower to faster... there was one beaut a few years ago.

kt


-- 
The Dalai Lama gets the same crap all the time.
   -- Kenny Tilton on c.l.l when accused of immodesty
From: Harold
Subject: Re: About graphics and big amounts of data.
Date: 
Message-ID: <1168643399.618452.261580@38g2000cwa.googlegroups.com>
Lars Rune Nøstdal wrote:
> I saw this a while back:
>   http://www.lrde.epita.fr/cgi-bin/twiki/view/Publications/200606-IMECS

Very good read. My favorite part is the mention that although C and
Lisp are running at the same speed, the Lisp functions can be
dynamically redefined if needed. Try making the C array lookups go
through a function pointer instead of the inline [] operator and see
who wins ;-)

- Harold
From: ········@gmail.com
Subject: Re: About graphics and big amounts of data.
Date: 
Message-ID: <1168699443.619066.324680@v45g2000cwv.googlegroups.com>
Lars Rune Nøstdal wrote:
> On Fri, 12 Jan 2007 08:00:55 -0800, alkhimov wrote:
>
> > I've got a task: filter significant amount of simple data (big picture)
> > with specific filters. Usually I used C/C++ mixed with assembler, but
> > eventually I decided to try Lisp. Can anybody recommend me some
> > literature / already completed programs / something else in this
> > direction?
> > And, by the way, is there any sense to try to avoid mixing with other
> > languages like C (for fast interface to target data, for example)?
> >
> > Tia.
>
> I saw this a while back:
>   http://www.lrde.epita.fr/cgi-bin/twiki/view/Publications/200606-IMECS
>
> --
> Lars Rune Nøstdal
> http://nostdal.org/

Thank you for this perfect answer: it's precisely what I needed!!!

WRT this task: I think that it is closer to fractals than to just
trivial filtering (like "Gaussian blur"- or "sharpen"-like filters).
Its logic is not trivial. For now I experiment with details, so C/C++
or assembler are inappropriate. The first thing I need on this stage is
flexibility, but not the speed itself. It's really surprising that Lisp
can be powerful enough even on so small tasks as shown in publication
in your answer.
From: Nicolas Neuss
Subject: Re: About graphics and big amounts of data.
Date: 
Message-ID: <87lkk83s25.fsf@ma-patru.mathematik.uni-karlsruhe.de>
········@gmail.com writes:

> I've got a task: filter significant amount of simple data (big picture)
> with specific filters. Usually I used C/C++ mixed with assembler, but
> eventually I decided to try Lisp. Can anybody recommend me some
> literature / already completed programs / something else in this
> direction?
> And, by the way, is there any sense to try to avoid mixing with other
> languages like C (for fast interface to target data, for example)?
> 
> Tia.

I have written something in this direction ("On using Common Lisp in
Scientific Computing") which you can find here:

http://www.mathematik.uni-karlsruhe.de/user/~neuss/publications.html

Helping someone else with a PhD thesis, I have implemented also some
nonlinear filter (nonlinear anisotropic diffusion) some years ago.  ISTR
that C/C++ was faster in this case.  (It should also be noted that for this
sort of task performance often depends on the memory hierarchy of the
computer architecture.)

I guess that the benefits of Lisp are not important enough for this
algorithmically relatively trivial number crunching task, and the
disadvantages (especially incompatibility with other developments) probably
dominate.

Nicolas