Re: [Algorithms] Cache-oblivious layout (was: BSP tree, or what?)

[Marc B. R. <mar...@or...>]

Out of curiosity I did a quick web search looking for recent papers with
some performance numbers, There seem to be a few, here's an example:

"SIMD Packet Techniques for Photon Mapping", 2007
http://www.cs.ucla.edu/~pfal/papers/RTR-07-simd.pdf


> Fairly obviously for very sparse data, there's no coherence to be
> extracted, and neither is significantly better than completely random, 
> in which case you should implement the simplest thing possible 
> (usually linear).

Humm..not sure about this, but I spent a whole 2 minutes thinking about it.