1.2.5 pre 4 Comments and suggestions

[Paolo]

Yesterday I've downloaded and installed the version 1.2.5 pre 4, but I have experienced some problems with Saturn.
I have an Hercules 3D Prophet Graphics adapter with the ATI Radeon 8500 GPU, an AMD Athlon XP 1900+ CPU and Windows XP.
On my PC With the 1.2.4 version Saturn is visualized correctly with the planet shadow on the rings and a framerate of about 90.

With the 1.2.5 pre 4 the framerate in the solar system remains around 100-120. But when I go to saturn it slows down to 2-3 , and the ring shadow on the planet have discontinuities. Also if I disable the rings shadows the framerate remains very low (3-4). It increases to about 20-30 if I rotate the visualization from an equatorial point of view so the saturn rings are viewed from side.

Comet ion tails are displayed gracefully. But I think that the ion tail should disappear when the comet is very far from the sun because of the solar wind intensity decreases. It should be not very difficult to implement this feature I think, I've noticed that the ion tail is rotated correctly toward to the Sun so the ion tail intensity calculations should be done when the object is reoriented.

When I display orbits in close range visualizations the line does not fit with the object position and the lines tilt fastidiously. This problem is present in 1.2.4 too. I think that is caused by the approximations dued of the limited number of points of the shape of the orbit. Perhaps should help if the orbit points of the selected object should be recalculated each frame forcing the first to the center of the object. It should be not very difficult to implement and not very expensive in computing time terms.

Obviously with my PC I still don't see Pixel shaders and Vertex shaders effects. I've seen some OpenGL demos from the ATI Radeon that uses these effects. I'll investigate over this fact and if I'll discover something all'give you the references.

I hope that these suggestions could help.
Congratulations to Chris.
Celestia is great! I love this program.

[chris]

Yesterday I've downloaded and installed the version 1.2.5 pre 4, but I have experienced some problems with Saturn.
I have an Hercules 3D Prophet Graphics adapter with the ATI Radeon 8500 GPU, an AMD Athlon XP 1900+ CPU and Windows XP.
On my PC With the 1.2.4 version Saturn is visualized correctly with the planet shadow on the rings and a framerate of about 90.

With the 1.2.5 pre 4 the framerate in the solar system remains around 100-120. But when I go to saturn it slows down to 2-3 , and the ring shadow on the planet have discontinuities. Also if I disable the rings shadows the framerate remains very low (3-4). It increases to about 20-30 if I rotate the visualization from an equatorial point of view so the saturn rings are viewed from side.

I think that I figured out what's going on here . . . Ring shadows were supposed to be disabled for Radeons (actually for anything other than GeForce3 and GeForce4 Ti . . . but see below for more on this). I accidentally left them enabled, causing wacky things to happen. It will be fixed soon--it's a big problem, so I'll try and make another prerelease tonight to correct it.

Comet ion tails are displayed gracefully. But I think that the ion tail should disappear when the comet is very far from the sun because of the solar wind intensity decreases. It should be not very difficult to implement this feature I think, I've noticed that the ion tail is rotated correctly toward to the Sun so the ion tail intensity calculations should be done when the object is reoriented.

Comet tails are still a work in progress, and I definitely will decrease the visibility of the ion tail when the comet is far from the Sun.

When I display orbits in close range visualizations the line does not fit with the object position and the lines tilt fastidiously. This problem is present in 1.2.4 too. I think that is caused by the approximations dued of the limited number of points of the shape of the orbit. Perhaps should help if the orbit points of the selected object should be recalculated each frame forcing the first to the center of the object. It should be not very difficult to implement and not very expensive in computing time terms.

Recalculating the orbit path each frame actually can be very expensive. The orbit of the Earth is modeled with over 400 periodic terms. If I sample the orbit 100 times to render the path, that's 40,000 cosine/multiply/add passes. And there's actually, there's even more work than that. This is why orbits paths aren't recomputed dynamically. One compromise would be to only recalculate the orbit path of the currently selected object.

Obviously with my PC I still don't see Pixel shaders and Vertex shaders effects. I've seen some OpenGL demos from the ATI Radeon that uses these effects. I'll investigate over this fact and if I'll discover something all'give you the references.

It's a known issue . . . Basically, until very recently there was no standard way to do pixel and vertex shaders in OpenGL. Since I own nVIDIA hardware (and now work for nVIDIA ), I used nVIDIA's pixel and vertex shader extensions. A couple months ago, vertex shaders were incorporated into OpenGL 1.4 and in the next Celestia version I'll switch to using the standardized shaders, making Celestia's vertex shader effects available for ATI users too. A few days ago, the OpenGL ARB approved a standard pixel shader extension. Once this appears in ATI's and nVIDIA's OpenGL drivers, I'll start using it. However, the new pixel shaders will only work on NV30 and Radeon 9700 graphics cards.

--Chris

[MKruer]

So can you give us any word on the NV50? :lol: Na, that’s cool, thanks for the info on the cards. I don’t know how things work over an nVIDIA but try to talk them into using Celestia as a benchmark for the video cards.

-Matt-