On 26/01/2019 20:38, Robert Osfield wrote:
I have a few more fixes to the 3.6 branch, so have gone ahead and tagged 3.6.4-rc3:
I've extensively tested FlightGear against 3.6.4-rc3 (several hundred hours of flight time) and so far it's been rock solid; the change (from our side) to using the ref methods appears to have fixed the problem, hard to be 100% certain but I'm 99% certain.
Overall I find that the certain aspects of the run time performance of 3.6.4 are improved compared to 3.4.2 and to possibly also to 3.7dev.
As a background I recently modified FG to have an automatic (on disk) DDS texture cache; I'm using osg_nvtt to convert the images into DDS using the filehash of the original to get a filename. This makes a massive difference to FG (as most textures are png, jpg, or sgi format) - as the generation of mipmaps give us seriously bad frame pauses (up to 5 seconds on my system). The DDS cache fixes all of this. On 3.6 the initial pan around from inside the cockpit is smoother than both 3.4 and 3.7. This is an observation rather than any sort of request for help as it's not really a problem, just nice when things get faster.
On FG we are using OSGText so I had to do a bit of interim hackery to get it to compile. Is there any sort of documentation that explains the changes to OSGText. We also have problems with the version that uses shaders because that interferes with our shader framework.
In my F-15 all of the CRT devices are using FG's canvas; which basically allows rendering of an SVG (or other vector) onto a texture. In 3.6 (and 3.7) I'm seeing problems with this that weren't present in 3.4. https://i.imgur.com/2s9ztJH.png https://i.imgur.com/rOW4wYe.png
Thanks for the improvements; I've started an initiative to move FG to 3.6.4 hopefully by our next release cycle.
_______________________________________________ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
