Re: Interpreters and (math) speed

stefan.wils@zorro.ruca.ua.ac.be said:
> I'm kind of new to this area, but I am in the process of designing a
> 'graphics language'
> [...]
> what do people, who are thinking about
> compiler/interpreter technology all the time, think of using an
> interpreter for such a task ? Could it be made fast enough, especially
> for calculations


One solution to building implementations of domain-specific languages
is indeed to write a interpreter on top of an abstract machine
(basically, a library), i.e., to offer a programmable interface on top
of a computation model.


If the overhead of the interpretation layer happens to be a
performance bottleneck, you might consider removing it automatically
using a partial evaluator, i.e., specializing your interpreter with
respect to given 'graphics programs'.


Some links:


  - How to architecture and implement a DSL
http://www.irisa.fr/compose/dsl/ssr97_abstract.html
  - Examples
http://www.irisa.fr/compose/dsl/dsl97_abstract.html
http://www.irisa.fr/compose/dsl/planp_abstract.html
  - DSL related projects
http://www.irisa.fr/compose/dsl/


  - Short introduction to partial evaluation
http://www.irisa.fr/compose/tempo/#specialization
  - Tempo Specializer, a partial evaluator for C
http://www.irisa.fr/compose/tempo/
  - C-Mix, a partial evaluator for C
http://www.diku.dk/research-groups/topps/activities/cmix.html


Renaud Marlet
--

Return to the comp.compilers page.
Search the comp.compilers archives again.