Andrew, thanks for joining the list. I hope we will be able to work out something interesting.
A couple of notes. On Mon, Nov 16, 2009 at 4:25 PM, Andrew Whitworth <[email protected]> wrote: >> On http://www.parrot.org/languages I found a project called Matrixy >> which is an implementation of Octave. I don't know how complete it is >> and how easy or difficult it will be to create PDL code from its parse >> tree but I hope I could tease the author into subscribing to this list so >> he will be able to give us an idea where things are standing. > > I'm the primary author of Matrixy right now, so maybe I can help with > some information. Matrixy is really two projects: > > 1) http://github.com/Whiteknight/matrixy, the Matrixy parser and > runtime (aims to be a more-or-less faithful compiler for the M > language used by Matlab/Octave) Once I have a little more time I'll check what is missing from the Parrot based syntax highlighting in Padre. Once that's done any language that has a parser using the Parrot Compiler Tools will automatically get a syntax highlighting and context sensitive help in Padre. > 2) http://github.com/Whiteknight/parrot-linear-algebra, which intends > to implement a number of "high performance" matrix data types for use > generally by Parrot. This includes bindings to CBLAS and CLAPACK > libraries for linear algebra (very preliminary). > > It is unlikely that Matrixy will ever be as performant or as capable > as Matlab or Octave are. The primary goals of the project were to > bring quality mathematics and linear algebra support to Parrot. Since > M is so focused on linear algebra and has syntax and routines > specially for building and manipulating matrices, it is a very natural > choice for writing those kinds of routines in. Ideally I would like to > see Matrixy used to write mathematics libraries which in turn can be > used by other HLLs on Parrot. > I talked with Gabor yesterday on IRC, and it seems another good use > for Matrixy is to serve as a parser for M, which can possibly > repurposed as a code translator to PDL. That would be fine too. With > the completed parser, the only remaining task would be a PDL code > generator. I don't know enough about PDL, so if anybody was interested > in working on that I would be happy to hand out commit bits to the > Matrixy repo. Before doing that we might want to evaluate if the execution over Parrot is really slower and if it is, how bad is that. If we can run it on Parrot we can then embed Parrot in any Perl 5 code which would mean the code written in the M language and the code written in PDL will run in the same process. I hope that make some sense. Gabor _______________________________________________ Perldl mailing list [email protected] http://mailman.jach.hawaii.edu/mailman/listinfo/perldl
