[Haskell] Beyond ASCII only editors for Haskell

Bill Wood william.wood3 at comcast.net
Mon May 23 18:00:43 EDT 2005


  . . .
> It is true that we do not see CASE-tools being used very much, but we do
> see tools go beyond ASCII and still be very popular. What I am thinking
> of is the visuals environment, like those produced by MS.
> 

You might be interested in looking at work done at Oxford and Chalmers
U. on graphical notations for describing hardware architecture, in
particular networks of functional units.  The interesting thing was that
the notation mapped essentially 1-1 onto expressions in an
application-specific language built on top of Haskell.  One variant,
called Lava, was adopted for use by Xylinx (?sp), an FPGA house.  At
least part of the reason it worked was that the functional units were
combinational, which is nicely compatible with a (stateless) pure
functional language.

In a similar vein, you might look at a heterogeneous simulation system
called "Ptolemy"/"Ptolemy II" developed by Prof. Edward Lee at Berkeley.
The system supports, among others, simulation models based on
Finite-State Automata, Kahn Process Networks and Token Flow Networks (a
generalization of dataflow).  The user interface is graphical, using
drag&drop to populate the functions and dragging to interconnect.  I was
experimenting with it as a tool for prototyping computations to be
performed on massively-parallel networks of functional units, similar to
FPGAs.  One really interesting fact to come out of the experiments was
that the EEs understood immediately what was going on (it looked sort of
like circuit and logic diagrams).  The difficulty was with the software
people, who insisted on viewing the functional units as little stateful
processes, so they started suggesting mechanisms for synchronization and
communication -- all that complexity.  I found myself siding with the
EES -- the dataflow model was simple, and design for performance could
be based on indicators such as length (latency) versus width
(parallelism) of the network.

 -- Bill Wood
    bill.wood at acm.org




More information about the Haskell mailing list