[Tile-serving] Parallelizing more of osm2pgsql

[Kai Krueger]

On 07/18/2013 03:39 AM, Peter Körner wrote:
> Hi
> 
> Am 17.07.2013 00:49, schrieb Paul Norman:
>> Something for discussion: Do we want to optimize non-slim mode or drop it?
>> Right now it's slower and has a very high RAM requirement, so it really
>> doesn't make sense to keep it as-is.
> 
> That's only true when talking about planet-imports. For small-area
> imports it can look very different here, couldn't it?

One of the main reasons non-slim is slower is that the "going over
pending way" stage is not parallelized in non-slim mode, and that part
does parallelize pretty well (at least up to 8 cores)

However, if you are importing on weaker machines like a small laptop,
you might not have as many cores and I think (although I'd need to
double check that), that with --number-processes=1 non-slim should still
be quite a bit faster.

On the other hand, is it worth keeping that code, just so that the
liechtenstein import runs in 1 rather than 2 seconds... ;-) ?

However, as long as it isn't broken, I don't think we should remove
functionality. I am just not planning on spending much time to improve
non-slim mode.

Kai

> 
> Peter
> 
> 
> 
> _______________________________________________
> Tile-serving mailing list
> Tile-serving at openstreetmap.org
> http://lists.openstreetmap.org/listinfo/tile-serving
>