[Patch v1] improve splitter
Hi all, I've finally found a way to test a lot more possible solutions in splitter in a reasonable time. As a result, splitting a single country like Germany is now done with a full search, means, all posible splits are tested and the best is chosen. For large files (e.g. Europe), the program can't search all possible solutions, but it still finds better ones than r389. This takes a bit more time than before, but is likely to find really good solutions. If needed, I can implement new options to allow stopping the search if e.g. a solution is found that has at least 85% of max-nodes in each tile. A few numbers: Split Germany with max-nodes=1200000 r 389: Solution is nice. Can't find a better solution: 164 tile(s) and a rating of 2756. The smallest node count is 903153, the worst aspect ratio is near 3.99 patched: Solution is nice. Can't find a better solution: 150 tile(s). The smallest node count is 1081130 (90 %), the worst aspect ratio is near 3.83 Split Germany with max-nodes=1000000 r389: Solution is nice. Can't find a better solution: 192 tile(s) and a rating of 3014. The smallest node count is 742452, the worst aspect ratio is near 3.9 patched: Solution is nice. Can't find a better solution: 181 tile(s). The smallest node count is 894244 (89 %), the worst aspect ratio is near 3.93 Please try with your workload and let me know the results. If you are not happy with the new result, please send the densities_out.txt and the splitter.log. A compiled binary is available here: http://files.mkgmap.org.uk/download/208/splitter.jar Gerd
Am Donnerstag, 22. Mai 2014, 20:03:48 schrieb Gerd Petermann: Hi Gerd
Please try with your workload and let me know the results. If you are not happy with the new result, please send the densities_out.txt and the splitter.log.
I've made some tests with the patched version, you find my logs here: http://files.mkgmap.org.uk/download/209/splitter.zip For me, it's seems to be a great improvement Bernd
Hi Bernd, thanks for the feedback and the data. I am working on some more improvements for large inputs or those that contain large empty areas. Gerd
From: weigelt.bernd@web.de To: mkgmap-dev@lists.mkgmap.org.uk Date: Fri, 23 May 2014 12:19:13 +0200 Subject: Re: [mkgmap-dev] [Patch v1] improve splitter
Am Donnerstag, 22. Mai 2014, 20:03:48 schrieb Gerd Petermann:
Hi Gerd
Please try with your workload and let me know the results. If you are not happy with the new result, please send the densities_out.txt and the splitter.log.
I've made some tests with the patched version, you find my logs here:
http://files.mkgmap.org.uk/download/209/splitter.zip
For me, it's seems to be a great improvement
Bernd
_______________________________________________ mkgmap-dev mailing list mkgmap-dev@lists.mkgmap.org.uk http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
Hi Gerd, i splat ( ;-) ) Germany.o5m with --max-nodes=2000000 per tile. With version 389 i get 96 tiles, with the improved version i get only 91 tiles with extraordinary equality in size, also the final gmapsupp.img is smaller. I think, it's a good improvement! Best regards, Uli -- View this message in context: http://gis.19327.n5.nabble.com/Patch-v1-improve-splitter-tp5807051p5807163.h... Sent from the Mkgmap Development mailing list archive at Nabble.com.
Hi Uli, yes, in most cases the patched version is much better, and I have found some nice improvements so I have an even better patch, BUT it is not very good when splitting a whole planet. I still try to understand why the old version is able to find a rather usable split with only a few tests, while the patched version is running much longer to end up with a bad result. Gerd
Date: Sat, 24 May 2014 04:55:58 -0700 From: ulibaer@gmail.com To: mkgmap-dev@lists.mkgmap.org.uk Subject: Re: [mkgmap-dev] [Patch v1] improve splitter
Hi Gerd,
i splat ( ;-) ) Germany.o5m with --max-nodes=2000000 per tile. With version 389 i get 96 tiles, with the improved version i get only 91 tiles with extraordinary equality in size, also the final gmapsupp.img is smaller. I think, it's a good improvement!
Best regards, Uli
-- View this message in context: http://gis.19327.n5.nabble.com/Patch-v1-improve-splitter-tp5807051p5807163.h... Sent from the Mkgmap Development mailing list archive at Nabble.com. _______________________________________________ mkgmap-dev mailing list mkgmap-dev@lists.mkgmap.org.uk http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
Hi Gerd, splitter version 392 is even better, it splat Germany.o5m to only 90 tiles! :) Best regards, Uli -- View this message in context: http://gis.19327.n5.nabble.com/Patch-v1-improve-splitter-tp5807051p5807253.h... Sent from the Mkgmap Development mailing list archive at Nabble.com.
Hi Bernd,
splitter version 392 is even better, it splat Germany.o5m to only 90 tiles! :)
just a note: The algo doesn't minimize the number of tiles, it maximizes "the number of nodes in the least populated tile". With the 1st strategy it could happen that you have 87 tiles with a 99% fill ratio and 2 with < 50%. The algo also stops if the worst fill ratio is 93% or higher. Gerd
participants (3)
-
Bernd Weigelt -
Gerd Petermann -
UliBaer