Thorsten (with an H) wrote > > > I haven't fully tested all the options, but in general the frame rate > > cost seems very heavy. > > > Regarding frame-rates: yes, the Local-Weather implementation kills > > non-high-end systems (mine, too). > > I'm getting slightly frustrated here. I've spent months to improve > performance, and I feel by now that I at least deserve a fair judgement > for my effords, i.e. an apples-to-apples comparison rather than a highly > skewed comparison. > > The default for the standard 3d clouds is about 16 km visibility of the > terrain and clouds are visible out to 20 km, clouds do not drift in the > wind. > > The default comparable local weather call, high pressure, has about 35 km > visibility (that's a factor 4 more terrain to render), clouds are visible > out to 30 km (that's a factor 2 more area to cover) and clouds do drift in > the wind. > > If you want to compare apples to apples, then you have to level the field. > I've just tested 'fair weather' conditions seen from the same location > (above KLSV), no cloud movement, same altitude of cloud layers and same > visibility of terrain (35 km) and clouds (20 km). Under these conditions I > get 80 fps with the standard 3d clouds and 75 fps with local weather. > > Yes, it's a bit slower under equal conditions. But hardly dramatically so. > > The reason why it appears slower should now be obvious - it attempts to do > more. If you want to render multiple overcast layers and clouds which > drift in the wind and flow in a semi-realistic way obstacles in the > terrain, then yes, that costs framerate. But if you compare the default > settings, then this is like flying the Cessna over Munich and the Piper > Seneca over the Paris scenery, and conclude that the Seneca is really > slow. > > > It has not been commited to the repository and I would not recommend > > doing so. We should stick to implementing algorithms and systems in the > > C++ core and use Nasal for prototyping or aircraft supporting logic. The > > terrainsampler is a good example, after being prototyped in Nasal, it's > > now a C++ coded subsystem configurable through the property-tree. > > I beg to disagree here. The terrain presampler is a highly specialized > tool, which is best for what it is supposed to do. A fast vector-method to > retrieve terrain elevations is a good multi-purpose tool quite independent > of terrain presampling. > > * weather could use it to model the deflection of the wind around major > obstacles > > * any ground radar mode could use it to get a picture of the terrain > > * low-flying AI planes could use it to navigate around obstacles for which > some more global picture of the terrain is needed > > * ... > > All this could be done faster having a fast vector method to get > elevations in Nasal, and unless you want to hard-code everything, Nasal > would probably be the platform of choice. What you want to include in GIT > and what not is of course not my business, but I'd certainly have use for > the vector elevation sampling beyond the terrain presampling. >
I for one would really like a faster, less framerate costly method of getting elevations. For my purposes, it would need to be hard-coded. Can't be soon enough. As for frame rates - I couldn't actually compare like for like because global weather was bugged here - I understand it is now fixed - so I will try again. Boy, would I like to see any framerate over 40 with 3D clouds - and that's with a Core 2 Quad, 4G of RAM and a nVidia 285. Hardly low end. What do I need to get 80? But keep on - it was I who first put this in CVS - there IS great potential here. Oh - and btw - I did persistent contrails as we discussed a while back. Vivian ------------------------------------------------------------------------------ Start uncovering the many advantages of virtual appliances and start using them to simplify application deployment and accelerate your shift to cloud computing http://p.sf.net/sfu/novell-sfdev2dev _______________________________________________ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
