After being prodded by several people (and I thank you all ;-) I did a quick and dirty compile of the latest plib/SimGear/FlightGear. This contains an updated version of my plib optimization that results in a similar boost in frame rate, on Mac OS X in particular. I haven't done any testing to prove it, but fgfs appears to be working OK and doing what it is supposed to do.
With your plib optimisations fgfs is much faster on Mac OS X ! On a imac (G3 500 MHz, 384 Mb RAM) with an ATI Rage pro (16 Mb VRAM):
with plib optim.: without: at KFSO, daylight: 13 fps 6 fps at KFSO, night: 9 fps not tested over the SF area: 8 fps 2-3 fps (during the day)
As far as compiling, .... The second one is required because the default stack limit in Mac OS X is 256K. The "unlimit" command is the other option, but I don't feel anyone should be forced to use it just to get fgfs to run.
I don't think that it is a stack limit problem: matlib.cx compiles very well on my imac with Xcode, but in order to have a working executable, one must turn optimization off with gcc 3.3 (ie -O0). This known bug of gcc 3.3 has been reported with a function very similar to gen_vasi_light_map().
I also applied the old ssgVtxTable optimization to plib, which keeps OpenGL from going down a very slow path and substantially boosts frame rates. This patch is different from the original immediate-mode hack but shows similar results and still uses drawarrays. At some point (now a long time ago) I concluded that the glPushClientAttrib()/glPopClientAttrib() calls and their consequences were the real killer so I've eliminated them in this patch.
Thanks a lot for this hint! It is really, really faster!
Olivier A.
_______________________________________________ Flightgear-devel mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-devel
