MathGroup Archive 2007

[Date Index] [Thread Index] [Author Index]

Search the Archive

Re: Mathematica to .NET compiler

  • To: mathgroup at smc.vnet.net
  • Subject: [mg79223] Re: Mathematica to .NET compiler
  • From: David Bailey <dave at Remove_Thisdbailey.co.uk>
  • Date: Sat, 21 Jul 2007 04:31:24 -0400 (EDT)
  • References: <f7pon7$prc$1@smc.vnet.net>

Jon Harrop wrote:
> Would anyone be interested in a compiler that translates Mathematica code
> into code that runs on Microsoft's .NET?
> 
> This would allow general purpose programs written in Mathematica to execute
> much more efficiently.
> 
Translation to Java would make more sense, because it would work 
cross-platform, but Java or .NET, there are serious problems as I see it.

You would only be able to compile about as much as the Compile command 
within Mathematica can achieve. True, that command only compiles to a 
form of P-code, but it has the advantage that it can invoke Mathematica 
for the bits it cannot handle without needing to switch process.

I would imagine that this would put such code at a disadvantage WRT the 
output of Compile - at least in many cases.

There is, of course, MathCode, which translates a subset of Mathematica 
to C++:

http://www.mathcore.com/products/mathcode/mathcodec++.php

The subset in question is described on a link, which was unfortunately 
dead when I pressed it. I would imagine their subset would be much the 
same as yours would need to be.

Because I have a background in writing compilers, I once gave some 
serious thought to the compile-to-Java approach.

It would be interesting to know other people's experiences with MathCode.

David Bailey
http://www.dbaileyconsultancy.co.uk


  • Prev by Date: Re: strange bug?
  • Next by Date: Re: N-dimensional NIntegrate
  • Previous by thread: Re: Mathematica to .NET compiler
  • Next by thread: Re: Mathematica to .NET compiler