I have committed changes to the locator branch which implements the usage of separate precompiled splitted boundary files. How does it work? * 1st step: Extract the boundary data Download an osm dump (e.g. the europe extract from geobfabrik). Extract the boundary data using osmosis: osmosis --rb europe.osm.pbf -tf accept-ways boundary=admistrative -tf accept-relations boundary=administrative --used-node -wx europe-boundaries.osm (Maybe the osmosis parameters must be tuned to get *all* boundary data but I haven't investigated very long. Please send proposals how to improve it) * 2nd step: Start mkgmap with additional --createboundsfile option e.g. java -jar mkgmap.jar --createboundsfile=europe-boundaries.osm <... your other options ...> Before compiling the common tiles mkgmap reads the europe-boundaries.osm file, performs the multipolygon processing on it and saves special precompiled boundary tiles to the subdirectory bounds. The subdirectory can be changed by adding the new option --boundsdirectory=<directory> It is possible to precompile whole europe with less then 3GB main memory. The europe boundary tiles need around 330MB (bzip2 compressed ~120MB). * 3rd step: Use the precompiled boundary files After precompiling the bounds directory contains several "bounds_<lat>_<lon>.bnd" tile files. These are loaded and used automatically by the LocationHook. You need to compile the boundary tiles once only. Important: The LocationHook does no longer use the boundary information from the tiles. You need the precompiled boundary tiles. I have decided to go this way because I do not expect that splitter will be able to save tiles with complete multipolygons within a foreseeable period of time. WanMil