Super-linear (super-critical) algorithms needed

Super-linear (super-critical) algorithms needed

Post by Gyorgy Pong » Sat, 13 Aug 1994 10:04:52

Dear Prallelists,

I'm looking for super-optimal (super-linear, super-critical)
parallel algorithms/methods, which are (seem to be)
executed in less time on N processors, than 1/N times
their execution time on 1 processor:

  tN < t1/N

I know the effect of the limited cache size,
the parallel backtracking (B. Monien, O. Vornberger),
the effect of lasy-cancellation at the optimistic parallel
discrete event simulation, but I'm searching for more.

Especially interesting are the "real" super-optimal ones,
which execute faster even if the parallel execution is emulated on
one processor.

Thank you:  Gy. Pongor



Hey people,

Super:Super:... is not a valid construct. Is there some way I can skip an
inheritance level?  I would presume you should be able to (this is
particularly useful for skipping one level of Init(), for example) but don't
see how.

Any ideas?



2. brother hl1040 print problem


4. Super Linear Effect References

5. genicom 3810s

6. Super Linear speed up

7. automatically c.c. a copy

8. need to upgrade nice super eisa motherbard to work with win/95

9. Lead Oracle Technical Projects Consultant (need LEAD X) ///// Dallas, TX //// 1 months=super hot for today

10. I need 'super' compression

11. Need help on Super Computer research.