3 ms·
If you think you want to use a Hilbert Curve for cache locality (on a square grid), you might want to check out the Z-Order Curve instead - https://en.wikipedia
by jmuhlich 7y ago
If you think you want to use a Hilbert Curve for cache locality (on a square grid), you might want to check out the Z-Order Curve instead - https://en.wikipedia.org/wiki/Z-order_curve https://en.wikipedia.org/wiki/Z-order_curve . It has less ideal locality properties but can be computed more efficiently.
- lichtenberger 7y agoCan we access the publication for free? Someone changed the link, originally it was: https://scibib.dbvis.de/uploadedFiles/MotionRugsPreprint.pdf https://scibib.dbvis.de/uploadedFiles/MotionRugsPreprint.pdf "For the Z-Order Curve we found that it is prone toproduce “Phantom-Splits”, visual artifacts creating the impression of a physical split of the entities when there is none. The Hilbert Curve insteaddisplays the spatial dynamic better than the Point-QuadTree and shows fewer artifacts than the Z-Order."
- jmuhlich 7y agoI have pretty comprehensive journal access through my job but I still couldn’t get this one. It makes sense that the Hilbert curve gave more accurate results, and in that application it probably makes sense to spend more compute time for a better visualization.