CPU/Hardware optimized for Erlang

[James Hague]

Thomas Lindgren wrote:

>Well, ECOMP had 30x speedup compared to JAM, not BEAM.
>On similar programs at that time, HIPE had a 10-20x
>speedup over JAM (when compiling JAM bytecodes to
>Sparc asm), if memory serves. So, assuming no further
>compiler improvements, ECOMP would have to remain
>within a factor of 3 or so of the desktop systems, to
>be competitive in speed.

Good points.  I had forgotten about the ECOMP numbers being relative to JAM.

>As someone whose name I can't remember liked to ask in
>situations like this, "what has changed?" Why is it
>going to be different this time?

Three things:

First, CPU designers are running up against physical issues in terms
of simply bumping up clock speed, most notably power dissipation.  For
a long while it was a given that each generation of a processor would
be significantly faster than the last, but this has tapered off
dramatically.  High-end processors keep getting faster, yes, but now
power consumption and heat are growing at prohibitive rates.  So,
optimistically, maybe one alternative is to get some instructions
doing a bit more, rather than having to break things down into a
stream of simpler instructions.  I'm not talking about VAX-style CISC,
however.

Along the same lines, notebook computers have taken off, so lower
power consumption is a big issue.  Sure, you're not going to be
running telecom apps and servers from a notebook, but it's an
interesting trend.

Third, FPGA technology has gotten advanced and commonplace.  There are
people on the web building working CPUs--often with very specific
purposes in mind--as a hobby.  This doesn't solve fundamental issues,
but it sure makes prototyping and refining a design easier.

James