[mapguide-users] Re: CUDA?

[Hans Milling]

The reason I ask is because I have a "simple" map with 23 layers. All MapInfo
TAB files. If I open all the 23 layers in a single map in MapInfo, zooping
etc is quite fast. Also when reacing zoom level 1:250 it is extreamly fast
(as one would expect). But with the MapGuide server tile generation I
experience the opposite. When at zoom level 1:350000 the tiles are generated
fast (I know less tiles are generated but the average speed of each tile)
but at 1:250 each tiles takes 700 ms in average. I don't understand why the
tile building is so slow.
Next I tried to put all the MapInfo files into an SQL Server 2008 R2 Spacial
database (using the EasyLoader). But the OGR provider still is a bit faster
than the SQL Spacial provider it seems.

My idea was, not to get rid of the tile cache, but remove the need to
pregenerate all the tiles. If tiles could be generated on the fly almost as
fast as a cache could deliver them, but still be cached once generated. That
would be great.

Even though you say the tile generation is all about memory troughput, i
believe that a graphics card can still do a lot better than your average
desktop/server. GPU's and their memory are optimized for paralel processing
and they are a lot faster than your standard CPU and memory.

Hans...

--
View this message in context: http://osgeo-org.1560.n6.nabble.com/CUDA-tp4965248p4965329.html
Sent from the MapGuide Users mailing list archive at Nabble.com.