Massive paralell junk (was: Classic Computer Auctions List)

From: Hans Franke <Hans.Franke_at_mch20.sbs.de>
Date: Mon Feb 1 09:19:07 1999

> > a) what if there is no peer type network, but rather a structured
> > hirarchy of domains ?

> You then have a tree structure, which at first improves speed, but you end
> up with bottle necks the higher you go in the tree (again, depending upon
> problem type and communications, you get the bottle neck of you allow
> arbitrary communication between nodes).

Of course, since we have only realy slow machines (less than
300 kop(6502)/s*unit) and slow communication (between ~10 kBit
for the working slaves and 100 kByte for the Main controler),
the problem has to be ideal - so low (or beter none) communication
between the slaves, and low dependancies between work pieces.

For the bottleneck we plan to use different networks with different
speeds in ech hirarchy - and for the problem, things like cracking
or rendering comes to mind .... (cracking especialy is _very_ good
scalable). Also the overhead might decrease in the higer levels _if_
we can manage to concate transfers (lower arbitation).

> > Come in and join the definition team.
> How does one join then?

write to majordomo_at_nut.ner, message subscribe vcf-c64

> -spc (And I never did really learn 6502 Assembly code)

We'll give you _ONE_ last chance :)

Now come!

Gruss
H.

--
Traue keinem Menschen der 5 Tage blutet und immer noch nicht tod ist.
Received on Mon Feb 01 1999 - 09:19:07 GMT

This archive was generated by hypermail 2.3.0 : Fri Oct 10 2014 - 23:31:58 BST