Re: [Flightgear-devel] Clouds artefacts
Frederic Bouvier wrote: I manage to implement this algorithm tonight. The results are here : 1. /sim/rendering/multi-pass-clouds=false : http://perso.wanadoo.fr/frbouvi/flightsim/fg-multip-off-1.jpg (204kb) 2. /sim/rendering/multi-pass-clouds=true : http://perso.wanadoo.fr/frbouvi/flightsim/fg-multip-on-1.jpg (198kb) 3 /sim/rendering/multi-pass-clouds=true ( billboard tree detail ) : http://perso.wanadoo.fr/frbouvi/flightsim/fg-multip-on-2.jpg (106kb) The previous try was not addressing hole made by prop disc or exhaust plume in layers beneath. Only artefact created by upper layers. This time, with more passes, the result is quite correct in all cases : Without multi passes, depth buffer blocks cloud rendering through the prop disk/quad : http://perso.wanadoo.fr/frbouvi/flightsim/fg-multip-off-2.jpg (131kb) with multi pass, over a broken layer : http://perso.wanadoo.fr/frbouvi/flightsim/fg-multip-on-3.jpg over an overcast layer : http://perso.wanadoo.fr/frbouvi/flightsim/fg-multip-on-4.jpg It seems there is a significant framerate penalty ( 14 - 11 fps ) : cloud layers are drawn two or three times instead of one. -Fred ___ Flightgear-devel mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-devel
Re: [Flightgear-devel] Clouds artefacts
Frederic Bouvier said: The previous try was not addressing hole made by prop disc or exhaust plume in layers beneath. Only artefact created by upper layers. This time, with more passes, the result is quite correct in all cases : Without multi passes, depth buffer blocks cloud rendering through the prop disk/quad : http://perso.wanadoo.fr/frbouvi/flightsim/fg-multip-off-2.jpg (131kb) with multi pass, over a broken layer : http://perso.wanadoo.fr/frbouvi/flightsim/fg-multip-on-3.jpg over an overcast layer : http://perso.wanadoo.fr/frbouvi/flightsim/fg-multip-on-4.jpg It seems there is a significant framerate penalty ( 14 - 11 fps ) : cloud layers are drawn two or three times instead of one. FWIW I don't think prop disks and plumes (as they are being done now) are worth the extra cost. Is there any problem with ignoring this particular issue for now, or can the triple pass be a separate command line option? Best, Jim ___ Flightgear-devel mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-devel
Re: [Flightgear-devel] Clouds artefacts
Frederic Bouvier wrote: Melchior FRANZ wrote: * Jim Wilson -- Sunday 28 March 2004 18:07: FWIW I don't think prop disks and plumes (as they are being done now) are worth the extra cost. Is there any problem with ignoring this particular issue for now, or can the triple pass be a separate command line option? FWIW I think that this bug has to be fixed, but also that it should be a render option. http://perso.wanadoo.fr/frbouvi/flightsim/fg-multip-off-2.jpg I am leaning to a 4 level quality choice : 0 - actual state : clouds are making holes in mountains when seen from below, translucent surfaces are making hole in clouds underneath. http://perso.wanadoo.fr/frbouvi/flightsim/fg-multip-lvl0.jpg 1 - clouds are blended with mountains http://perso.wanadoo.fr/frbouvi/flightsim/fg-multip-lvl1.jpg 2 - translucent aircraft surfaces are blended with clouds but with problem with underlying 3d static objects ( clouds between the propeller disc and 3d objects are not seen ) 3 - total correctness http://perso.wanadoo.fr/frbouvi/flightsim/fg-multip-lvl2-1.jpg the propeller disc doesn't appear well on overcast so the same over broken to show that it is still there : http://perso.wanadoo.fr/frbouvi/flightsim/fg-multip-lvl2-2.jpg Perhaps 2 and 3 can be collapsed with no extra cost. Yes, they can, it is just a matter of drawing the aircraft after the other models In this area of the world, there is no impact on framerate with my hardware setup. At KSFO though, the difference is real. -Fred ___ Flightgear-devel mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-devel
RE: [Flightgear-devel] Clouds artefacts
Frederic Bouvier wrote Curtis L. Olson wrote: David Megginson wrote: Whether it goes in the release or not is up to Curt, but it looks great, and I notice from the screenshots that your framerate stayed the same. On IRC Fred mentioned that they didn't look correct from above. Let's see if he can track down the problem before the official release ... I'm trying to roll out the next pre release tonight. I discovered since then that I started FG with --disable-clouds (!) With --enable-clouds, it is ok. This property ( /environment/clouds/status ) seems to be only tested on clouds drawn below the viewer. So to make it short, it seems to work ok. What this change does not address yet is the fact that we can see the ground over overcast layer through the exhaust beam of the hunter. I can try to look at this tonight and cure the --disable-clouds to really disable clouds. Cheers, -Fred The exhaust plume of the Hunter is a crude simulation - using a nearly transparent material. Perhaps there's a better way of simulating the exhaust? Regards Vivian Meazza ___ Flightgear-devel mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-devel
Re: [Flightgear-devel] Clouds artefacts
Vivian Meazza wrote: So to make it short, it seems to work ok. What this change does not address yet is the fact that we can see the ground over overcast layer through the exhaust beam of the hunter. I can try to look at this tonight and cure the --disable-clouds to really disable clouds. The exhaust plume of the Hunter is a crude simulation - using a nearly transparent material. Perhaps there's a better way of simulating the exhaust? No, I don't think of a better way to draw something transparent with transparent material ;-) The only problem here is that it seems that the aircraft seems to be drawn before the lower clouds ( although the source contradict this, but I must be misled - perhaps the exterior model is in the terrain scene graph ). For better result, transparent objects must be drawn back to front. -Fred ___ Flightgear-devel mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-devel
RE: [Flightgear-devel] Clouds artefacts
Frederic BOUVIER wrote Vivian Meazza wrote: So to make it short, it seems to work ok. What this change does not address yet is the fact that we can see the ground over overcast layer through the exhaust beam of the hunter. I can try to look at this tonight and cure the --disable-clouds to really disable clouds. The exhaust plume of the Hunter is a crude simulation - using a nearly transparent material. Perhaps there's a better way of simulating the exhaust? No, I don't think of a better way to draw something transparent with transparent material ;-) The only problem here is that it seems that the aircraft seems to be drawn before the lower clouds ( although the source contradict this, but I must be misled - perhaps the exterior model is in the terrain scene graph ). For better result, transparent objects must be drawn back to front. -Fred The canopy doesn't do this - it has a semi-transparent texture as well. I'll experiment some more. Regards Vivian ___ Flightgear-devel mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-devel
Re: [Flightgear-devel] Clouds artefacts
Vivian Meazza wrote: Frederic BOUVIER wrote Vivian Meazza wrote: So to make it short, it seems to work ok. What this change does not address yet is the fact that we can see the ground over overcast layer through the exhaust beam of the hunter. I can try to look at this tonight and cure the --disable-clouds to really disable clouds. The exhaust plume of the Hunter is a crude simulation - using a nearly transparent material. Perhaps there's a better way of simulating the exhaust? No, I don't think of a better way to draw something transparent with transparent material ;-) The only problem here is that it seems that the aircraft seems to be drawn before the lower clouds ( although the source contradict this, but I must be misled - perhaps the exterior model is in the terrain scene graph ). For better result, transparent objects must be drawn back to front. -Fred The canopy doesn't do this - it has a semi-transparent texture as well. I'll experiment some more. Beware: the interior is not drawn at the same place that the exterior. To convince yourself, you can see that exterior model is affected by fog ( in cloud layer for instance ) and the interior not. -Fred ___ Flightgear-devel mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-devel
RE: [Flightgear-devel] Clouds artefacts
Frederic BOUVIER writes: Vivian Meazza wrote: Frederic BOUVIER wrote Vivian Meazza wrote: So to make it short, it seems to work ok. What this change does not address yet is the fact that we can see the ground over overcast layer through the exhaust beam of the hunter. I can try to look at this tonight and cure the --disable-clouds to really disable clouds. The exhaust plume of the Hunter is a crude simulation - using a nearly transparent material. Perhaps there's a better way of simulating the exhaust? No, I don't think of a better way to draw something transparent with transparent material ;-) The only problem here is that it seems that the aircraft seems to be drawn before the lower clouds ( although the source contradict this, but I must be misled - perhaps the exterior model is in the terrain scene graph ). For better result, transparent objects must be drawn back to front. The canopy doesn't do this - it has a semi-transparent texture as well. I'll experiment some more. Beware: the interior is not drawn at the same place that the exterior. To convince yourself, you can see that exterior model is affected by fog ( in cloud layer for instance ) and the interior not. Exactly, as was mentioned earlier in this thread To render transparent Objects with OpenGL 'correctly', they need to be rendered in sorted order from back to front. The only way I know of doing this in a 'generic' way requires detecting those objects with tranparency somehow and adding them to their own scenegraph which is then only drawn after all non-transparent objects. The easiest way I can see to do this in FGFS would be to require all models with transparent features to have two models 1) the opaque parts 2) the transparent parts and then render the transparent parts after the clouds are drawn, this assumes that all the clouds are further away then any of (2) above, which might not be the case with eshaust trails esp. with 3D clouds. This is 'tricky' stuff :-) Cheers Norman ___ Flightgear-devel mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-devel
RE: [Flightgear-devel] Clouds artefacts
Frederic BOUVIER wrote Vivian Meazza wrote: Frederic BOUVIER wrote Vivian Meazza wrote: So to make it short, it seems to work ok. What this change does not address yet is the fact that we can see the ground over overcast layer through the exhaust beam of the hunter. I can try to look at this tonight and cure the --disable-clouds to really disable clouds. The exhaust plume of the Hunter is a crude simulation - using a nearly transparent material. Perhaps there's a better way of simulating the exhaust? No, I don't think of a better way to draw something transparent with transparent material ;-) The only problem here is that it seems that the aircraft seems to be drawn before the lower clouds ( although the source contradict this, but I must be misled - perhaps the exterior model is in the terrain scene graph ). For better result, transparent objects must be drawn back to front. -Fred The canopy doesn't do this - it has a semi-transparent texture as well. I'll experiment some more. Beware: the interior is not drawn at the same place that the exterior. To convince yourself, you can see that exterior model is affected by fog ( in cloud layer for instance ) and the interior not. -Fred It's hard to spot, but indeed the cockpit canopy does render clouds transparent when viewed from outside not sure about fog - can't see through it :-)). Transparent textures seem to make no difference. I was going to say if it was just an exhaust plume problem, I could delete them until I could make them better, but since the problem is wider we should seek a solution. Vivian ___ Flightgear-devel mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-devel
Re: [Flightgear-devel] Clouds artefacts
Frederic BOUVIER wrote: I wrote: We can also begin to think about a multi pass method that would : 1. draw the clouds above the viewer without depth update, 2. draw the scene, 3 .redraw the clouds above with depth test Thinking more about it, for the record if someone wants to toy with this : 0. clear the stencil buffer 1. draw the clouds above the viewer without depth update, 2. draw the scene with stencil buffer write enabled ( terrain and objects overwrite all clouds, even those that are between the viewer and the terrain ) 3 .redraw the clouds above with depth test and stencil test to only update terrain that can be obscured by clouds - this should clear the transparency 'drawn twice' problem With an impact on framerate due to double writing and problem like the one Melchior is experiencing with overlapping semi transparent objects. Haven't thought about it much of that. Obviously, this should be optional to accomodate less capable hardware I manage to implement this algorithm tonight. The results are here : 1. /sim/rendering/multi-pass-clouds=false : http://perso.wanadoo.fr/frbouvi/flightsim/fg-multip-off-1.jpg (204kb) 2. /sim/rendering/multi-pass-clouds=true : http://perso.wanadoo.fr/frbouvi/flightsim/fg-multip-on-1.jpg (198kb) 3 /sim/rendering/multi-pass-clouds=true ( billboard tree detail ) : http://perso.wanadoo.fr/frbouvi/flightsim/fg-multip-on-2.jpg (106kb) If you think it's worth being in the release ( being optional thanks to property ), just speak and I will prepare a patch tomorrow evening ( sky.[ch]xx was touched in SimGear, main.cxx in flightgear, preferences.xml in fgfsbase ) Otherwise, I'll hold off until the release is out. Cheers, -Fred ___ Flightgear-devel mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-devel
Re: [Flightgear-devel] Clouds artefacts
Frederic Bouvier said: Frederic BOUVIER wrote: I wrote: We can also begin to think about a multi pass method that would : 1. draw the clouds above the viewer without depth update, 2. draw the scene, 3 .redraw the clouds above with depth test Thinking more about it, for the record if someone wants to toy with this : 0. clear the stencil buffer 1. draw the clouds above the viewer without depth update, 2. draw the scene with stencil buffer write enabled ( terrain and objects overwrite all clouds, even those that are between the viewer and the terrain ) 3 .redraw the clouds above with depth test and stencil test to only update terrain that can be obscured by clouds - this should clear the transparency 'drawn twice' problem With an impact on framerate due to double writing and problem like the one Melchior is experiencing with overlapping semi transparent objects. Haven't thought about it much of that. Obviously, this should be optional to accomodate less capable hardware I manage to implement this algorithm tonight. The results are here : 1. /sim/rendering/multi-pass-clouds=false : http://perso.wanadoo.fr/frbouvi/flightsim/fg-multip-off-1.jpg (204kb) 2. /sim/rendering/multi-pass-clouds=true : http://perso.wanadoo.fr/frbouvi/flightsim/fg-multip-on-1.jpg (198kb) 3 /sim/rendering/multi-pass-clouds=true ( billboard tree detail ) : http://perso.wanadoo.fr/frbouvi/flightsim/fg-multip-on-2.jpg (106kb) If you think it's worth being in the release ( being optional thanks to property ), just speak and I will prepare a patch tomorrow evening ( sky.[ch]xx was touched in SimGear, main.cxx in flightgear, preferences.xml in fgfsbase ) Otherwise, I'll hold off until the release is out. Cheers, -Fred That looks really nice. I hope we can include it. Best, Jim ___ Flightgear-devel mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-devel
Re: [Flightgear-devel] Clouds artefacts
Frederic Bouvier wrote: If you think it's worth being in the release ( being optional thanks to property ), just speak and I will prepare a patch tomorrow evening ( sky.[ch]xx was touched in SimGear, main.cxx in flightgear, preferences.xml in fgfsbase ) Otherwise, I'll hold off until the release is out. Whether it goes in the release or not is up to Curt, but it looks great, and I notice from the screenshots that your framerate stayed the same. All the best, David ___ Flightgear-devel mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-devel
Re: [Flightgear-devel] Clouds artefacts
David Megginson wrote: Whether it goes in the release or not is up to Curt, but it looks great, and I notice from the screenshots that your framerate stayed the same. On IRC Fred mentioned that they didn't look correct from above. Let's see if he can track down the problem before the official release ... I'm trying to roll out the next pre release tonight. Curt. -- Curtis Olson Intelligent Vehicles Lab FlightGear Project Twin Cities[EMAIL PROTECTED] [EMAIL PROTECTED] Minnesota http://www.menet.umn.edu/~curt http://www.flightgear.org ___ Flightgear-devel mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-devel
Re: [Flightgear-devel] Clouds artefacts
David Megginson wrote : Frederic Bouvier wrote: If you think it's worth being in the release ( being optional thanks to property ), just speak and I will prepare a patch tomorrow evening ( sky.[ch]xx was touched in SimGear, main.cxx in flightgear, preferences.xml in fgfsbase ) Otherwise, I'll hold off until the release is out. Whether it goes in the release or not is up to Curt, but it looks great, and I notice from the screenshots that your framerate stayed the same. I have a GeForce FX5900. I think it is where the newer cards can give us the additional benefit we are not seeing in triangle crunching. -Fred ___ Flightgear-devel mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-devel
Re: [Flightgear-devel] Clouds artefacts
Curtis L. Olson wrote: David Megginson wrote: Whether it goes in the release or not is up to Curt, but it looks great, and I notice from the screenshots that your framerate stayed the same. On IRC Fred mentioned that they didn't look correct from above. Let's see if he can track down the problem before the official release ... I'm trying to roll out the next pre release tonight. I discovered since then that I started FG with --disable-clouds (!) With --enable-clouds, it is ok. This property ( /environment/clouds/status ) seems to be only tested on clouds drawn below the viewer. So to make it short, it seems to work ok. What this change does not address yet is the fact that we can see the ground over overcast layer through the exhaust beam of the hunter. I can try to look at this tonight and cure the --disable-clouds to really disable clouds. Cheers, -Fred ___ Flightgear-devel mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-devel
