Re: [glas] introducing the storage concept, a first interface specification
From: Karl Meerbergen (Karl.Meerbergen_at_[hidden])
Date: 2005-01-25 12:43:09
On Tuesday 25 January 2005 15:44, Patrick Kowalzick wrote:
> > The motivation for discussing storage concepts are the following:
> > 1) To allow algorithms to optimize for different ways of storage, so
> > algorithms are not orthogonal to storage
> Ok. In fact my questions belong here:
> a.) About which optimizations do we talk?
Blocking strategies for example. Files on disk can be accessed randomly, but
it usually is more efficient to access data in a forward_iterator fashion.
Out of core algorithms can reduce the disk access to a minimum. I suppose
some technologies from cache blocking can come in handy here as well.
If data are distributed over a network, you may also want to code algorithms
in a different way.