[PD] puredata system architecture

[Mathieu Bouchard]

On Mon, 16 Jul 2007, Frank Barknecht wrote:
> Miller Puckette hat gesagt: // Miller Puckette wrote:
>> There's a sketchy description in:
>> http://crca.ucsd.edu/~msp/Publications/icmc96.ps
>> Some of the details didn't come out as planned there but the client/server
>> setup at least didn't change.
> It's interesting to re-read section 5 "Implementation" of this 1996
> paper in the light of the "renaissance" of multiprocessor systems
> today. ;)

Yeah, it's not really a "renaissance", it's that FTS/ISPW systems were 
designed as elite systems for elite musicians backed by elite budgets. 
When Opcode and Pd and jMax appeared, it was to bring patchers to personal 
computers.

On personal computers, multiprocessing never caught on because the fastest 
thing in the medium-price range was single-processor. This only changed at 
the time of the introduction of the G5 and the Pentium-D processor (which 
I won't abbreviate to "PD"...). At that time, the impending era of 
multiprocessing had been announced for well over 10 years and one can only 
cry wolf for so long...

At about the same time as that, though, personal computers got a heavily 
parallel coprocessor known as a GPU. Multithread support would be quite 
moot if the GPU could be used with the same level of flexibility that the 
CPU can. I don't know enough about GPUs to tell yet, but I guess that this 
is one of the next things I'll have to learn.

  _ _ __ ___ _____ ________ _____________ _____________________ ...
| Mathieu Bouchard - tél:+1.514.383.3801, Montréal QC Canada