[erlang-questions] How small could an Erlang emulator be?
Thu Mar 15 00:38:40 CET 2007
Den 2007-03-15 00:10:59 skrev ok <>:
> On 14 Mar 2007, at 10:55 pm, Matthias Lang wrote:
> > At the risk of guessing wrong, maybe you're thinking of Robert
> > Tjärnström and his ECOMP thingamibob.
> That's it. In 2001 it was "Almost Ready". All a web search can find
> then is a trickle of "whatever happened" messages.
> It is sad that it apparently never came to fruition. If I want a
> free interesting
> processor design to look at, I've got to look at SPARC or OpenRISC
> 1200 instead.
> What's even sadder is how little information was ever published.
> I've never even
> seen an explanation of why work on ECOMP stopped (or even whether it
> did stop).
It was put on ice, partly because Tjärnström was assigned other
duties, and partly because we had difficulty finding a way to
fit it into our products. After all, going from an FPGA prototype
to an ASIC is a pretty expensive project. Ironically, the
architectural changes needed in our products at the time would
only have turned out well if ECOMP turned out as good as the
prototypes suggested, and we'd have to do a large amount of
work on faith.
It would be possible to revive ECOMP and/or release it as Open Source,
but Ericsson would have to be presented a very good reason to do so.
I made a proposal a while ago to MARLOW (www.lowpower.org),
suggesting that a low-power Erlang chip might be a good candidate
for solving their apparent dilemma - needing an order-of-magnitude
improvement in power efficiency in mobile devices, while at the
same time managing the increasing complexity in software. In their
roadmap, they discuss the problem that known techniques for
reducing the power consumption in hardware are difficult to handle
in software. Erlang would address the problem of software complexity,
while embracing parallelism in hardware, and ECOMP should give an
order of magnitude better power efficiency than running Erlang on
a general purpose CPU.
I thought the proposal was well received, but a firm commitment from
Ericsson was asked for. It is possible that that could happen as
a result of sufficient interest on the outside. The part of Ericsson
that I work with has no interest in mobile devices, so I was basically
just pitching the idea over the wall. (:
More information about the erlang-questions