Recently David Goblirsch asked the following on the Haskell
mailing list:
>What is the status of the Haskell 1.3 and pH efforts announced in this
>group several months ago? Are there separate mailing lists for these
>efforts?
pH development has been continuing both at MIT (in my group) and
at Cambridge Research Lab, DEC. We are collaborating with Lennart Augustsson
and Nikhil at DEC is collaborating with Simon Peyton Jones. I will briefly
describe the pH effort at MIT.
There are two Id compiler substrates at MIT. One is written in Lisp
and the other in Id. The Lisp version has been the main work horse for years
while the Id-in-Id compiler is still in development. The target of both
these compilers were various dataflow machines but now considerable effort
is being spent to target them to Unix workstations and commercial parallel
machines. The pH compiler will be based on the Lisp substrate.
There are several components in this developmental work:
1. A translator from Id to pH. This work is almost complete
for the functional subset of Id. We have been running functional programs
translated from Id to pH on various Haskell compilers. We plan to
release some these Haskell programs in the near future.
2. Extensions to the Haskell front end to accept the full pH syntax, and
Haskell type-checking to accept imperative types in pH. Lennart is doing most
of this work.
3. New desugaring modules for comprehensions to take full
advantage of I-structures and parallelism.
4. Connecting the Haskell/pH front end to the middle part of the
Id compiler written in Lisp.
No significant changes are expected in the middle part or the
code generator of the Id compiler to tackle pH.
Work is continuing on 2, 3 and 4. So far we have code to handle the
functional part of pH almost ready, and are able to compile some
small example programs, although a lot more engineering efforts
are expected. If all goes well we may have pH running by the summer of 1994.
So far only a few technical problems have been encountered,
all these have to do with arrays and the Haskell prelude.
Please feel free to contact me for further information.
Arvind
