Current work in compiler/language design.

In article 91-11-030 Harris L. Gilliam - MIT Project Athena <hackeron@Athena.MIT.EDU> writes:
// What are the current hot spots in compiler/language design ?


I think the previous responses (on explicit and automatic parallelization,
memory hierarchies, etc.) are best categorized as a EE view of writing
optimizing compilers for their latest RISC/MIMD/SIMD architectures (and
they left out object-oriented architectures at that). I thought I'd put
in a few cents worth for the programming language community.


The hot spot in programming languages today is object-oriented, with
(well-known) future directions and alternatives including dynamic typing,
dynamic inheritance (delegation), dynamic classes, method combinations and
better parameter modes (named and default at run-time), persistent
objects, generics (with dynamic class parameters), active objects, good
exception handling, dynamic linking, integrated environments, etc. ad
infinitum. Most of this stuff is well known, it's just that almost no
one's gotten around to providing them, at least not together in a usable
form. I've done a lot of these and have designed a new architecture that
can easily generate efficient compiled code for all of the above, but not
through 'dynamic compilation', which is what you'll inevitably read about.


Dynamic compilation gives the power of dynamic constructs in a compiled
form, but derived through interpreted intermediate code (yech!!!). I'm
simply pointing out that the aforementioned 'hot-spots' can be efficiently
compiled without any mess (don't let any interpreted prototype systems
throw you, is my advice. I plan to have a paper out to show how to do
this soon, since programming language and compiler design is my primary
research interest, too. You can look for it in the next OOPSLA conference
papers through sigplan, assuming its accepted).


Anyway, read up on Sigplan Notices and the quarterly OO-messenger and
OOPSLA proceedings. I was just reading up on type roles in the messenger,
a primitive view not generalized (in the article anyway) up a
multiple-inheritance hierarchy, but still worth reading. Dynamic typing
solves some of the problems addressed in the article too, but they are
trying for static typing as an efficiency issue, alas... I prefer both
static and dynamic facilities as alternatives, and yes, orthogonally in
the same language with no cost to static facilities (*My* opinion: if you
can't do this, don't write yet another programming language!!!).


Bob Hathaway
Object Systems
objsys@netcom.com
--

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