[mkgmap-dev] Regression causing Heap space exhaustion?

Sat Aug 22 10:48:28 BST 2009

Hi

On 08/22/2009 02:00 AM, Morten Kjeldgaard wrote:
> I've tried at 512M, 1024M, all the way up to 4096M, with the same
> result. The only difference is at 4G it takes much longer before
> mkgmap gives up. I don't think there's much point in going further
> than 4G, the problem must lie elsewhere.

Oh, in that case something is wrong.  There are limits on how big an 
output map can be and so the input osm file must be split into tiles.
Any tile that is small enough that it doesn't overflow the created map
can be compiled with a heap size of 512M or less(*)

The current map extract of Denmark is certainly too big and should be
split.  So that is normal.

The surprising thing is that r630 could deal with and also produce a
working map.  I'm not going to speculate on that just yet because there 
is definitely some problem with the Danish extract as I am unable to
process it even after splitting it properly.

Information on splitting here:
   http://www.mkgmap.org.uk/page/tile-splitter

I will follow up when I have investigated the large memory usage of
the Danish tiles.  Anyone else have problems?

(*) If there was a lot of stuff mapped that is not going to be in the
output file, then clearly the input file could be bigger and therefore
require more memory while not overflowing the map that is created.
I believe that 512M is true for areas within Europe - certainly less 
than 1024M.

Cheers,

..Steve