16 Aug
2013
16 Aug
'13
10:01 p.m.
On 8/16/13 4:12 PM, "BE Arnold" <bearnold at outlook.com> wrote:
On Aug 16, 2013, at 7:01 PM, "Dan Gahlinger"
<dgahling at hotmail.com> wrote:
Nah.
The transputer was one response to what I like to call the 'topology wars'
of distributed parallel computing. At one time, everyone was trying to
argue that a star vs. a ring vs. a hypertoroid vs. a hypercube topology
was the best way to do parallel computing. The transputer was very cool
in that it incorporated the communication element into both the hardware
and the instruction set of the part.
But not too long after, networking advanced to the degree that it was
possible to create massively parallel computers that were effectively
all-way connected. One might say that no one won the topology wars,
because fast networking offered a Gordian solution.
I wonder what might have happened if Inmos hadn't expected everyone to
learn Occam in order to use their part? Not that Occam wasn't pretty
cool, but I still remember the immense effort into 'parallelizing'
existing languages. Funny enough, FORTRAN showed the most benefit, which
most folks attributed to the belief that it was so awfully slow to begin
with. -- Ian
I remember at the end or near the end of the
commodore-64 era,around
the time that Transactor Magazine died,
there were discussions about "trans-puters" and this lasted a while,
but not too long.
Even shorter (perhaps a month or 3) was "trans-networking"
Now, googling for any of these things comes up with nothing.
Does anyone have any recollection of this stuff, or has it really died
out that much?
Dan.
My snarky answer is "Jack Tramiel happened to transputers."
Actually that's a good question. I read an article not too long ago
suggesting that the complicated design of a transputer was just pushed
aside by the sheer brute force of 'x86 processors.