i think some thing like this
History lesson
in the 1960's and 1970's Iann Barron and chums produced the first true parallel processing environment
The still revolutionary Transputer and Occam
in the late 1970's and 1980's Transmeta produced
The Code morphing compiler [and recently reincarnated in Russia ??]
in recent times Andreas Olofsson tried to reinvent the transputer and produced
The Epiphany co processor and associates the accompanying Parrallela language [sadly its next incarnation .. the one they should have gone for seems to have stalled .. being a cycnic im not surprised by that ]
What will Happen ....
The latter idea is sort of where it will go in my opinion ....
My prediction is that the FGPA system will enable individuals to make specialised cluster "chips" .. eventually this will gather enough momentum that these groups will be able to form some thing like a computer-by-process and that enough of these processes"at a reasonable price" [its prohibitive at the moment] will be the basis of the next generation computer in the same way that Linux formed
an operating system first from a kernel and then the mish-mash of contributions by lots of enthusiasts mostly from the scientific and academic communities [due to their requirements as ordinary people only want 'to do a bit of keeping in touch with each other' and maybe a few basic accounts so they just need posh phones ] ...
The problem with computing has always been the Hardware software interface and this is partly relieved with FGPA [see below ]
unfortunately you will have noticed that ....
politics and funding killed off the Transputer
IBM bought Red Hat Linux [the only decent Linux !!!] for 36 billion recently i think this will cast a long shadow over Linux
you will notice that Transmeta's offering was balatantly plaguarised by Intel and AMD in order to kill it off [and i would too if i was in their position]
that the incarnation of Epiphany [1024 cores 64 bit with a Zylinx chip attached ] isnt happening as far as i can see
Nvidia own many of the Linux based GUI API's
so you can fully expect that developments likely to cause issues with what is prodominantly American Big Business may well be bought out or swallowed up .. in fact if you come up with something innovative then dont expect to get much out of it [fact !!] and probably dont expect it to see much daylight as large companies are a bit like "collectors" so it might sit in a vault for a long time
you will also be aware
much of the parallelisation and optimisation of code in the x86 machine is in hardware not software as with the Transmeta offering
[ see various articles on the Itanium for a pleasant contrast in idealogy The Transmeta offering wasnt brilliant either but it was a bloody good first attempt considering the difficulties involved and how far evolved the competion was]
incidentally some of this has moved into ML64 {along with what i reckon is a lot of paranoia} which you will notice makes it rather bulky compared to its 32 bit parent
regards mike b
