Re: [Flightgear-devel] Trees, alpha, and the sky

2002-05-05 Thread Wolfram Kuss

Andy Ross wrote:

Wolfram Kuss wrote:
 Andy Ross wrote:
  You could also experiment with turning off backface culling instead
  of rendering two quads for each direction.  In principle, it should
  be faster.  In practice, it's probably a good way to detect driver
  bugs. :)

 The speed difference will be very, very small.

Are you sure?  It reduces the vertex count by a full factor of two.

IMHO you can use indexes (sp?) into the same vertices, so the vertex
count would stay the same. The pixel count would off course also be
the same.

Andy

Bye bye,
Wolfram.

___
Flightgear-devel mailing list
[EMAIL PROTECTED]
http://mail.flightgear.org/mailman/listinfo/flightgear-devel



Re: [Flightgear-devel] Trees, alpha, and the sky

2002-05-05 Thread Martin Spott

   http://www.megginson.com/flightsim/tree.png
[...]
 2. While other scenery shows through the alpha parts of the texture,
the sky often doesn't, as you can see in the above screenshot (note
how you can see the grass and river near the bottom of the trunk, but
not the sky near the crown).

This is very similar to what I saw with the propeller disk,

Martin.
-- 
 Unix _IS_ user friendly - it's just selective about who its friends are !
--

___
Flightgear-devel mailing list
[EMAIL PROTECTED]
http://mail.flightgear.org/mailman/listinfo/flightgear-devel



Re: [Flightgear-devel] Trees, alpha, and the sky

2002-05-02 Thread Wolfram Kuss

Andy wrote:

Also, that's an awfully small texture.  While it's good to not waste
space, this thing is tiny -- one eighth the size of a single panel
instrument's face.  IMHO, it would look much better at 128x256 or so,
with no measurable loss of performance.

I agree. Some modders I know use one 512 x 512 texture file per
instrument (for the face including the bezel (sp?) and the needle).
This works very well, with GeForce 2 and higher you get no measurable
slow down compared to smaller textures.

You could also experiment with turning off backface culling instead of
rendering two quads for each direction.  In principle, it should be
faster.  In practice, it's probably a good way to detect driver
bugs. :)

The speed difference will be very, very small.

Andy

Bye bye,
Wolfram.


___
Flightgear-devel mailing list
[EMAIL PROTECTED]
http://mail.flightgear.org/mailman/listinfo/flightgear-devel



Re: [Flightgear-devel] Trees, alpha, and the sky

2002-05-02 Thread Wolfram Kuss

Jim wrote:

 I've been thinking about that: how about not at the top, but halfway?
 Essentially, you'd have the XY, YZ, and XZ planes, one unit wide, all
 intersecting at the origin. Is that clear?

Yes. You want to make the horizontal polygon at the largest extension
of the leaf canopy. This means the chance that in your view the
horizontal poly extends over the other two polygons is lowest.

BTW, I recommend setting the origin to the place where the stem goes
into the earth.

 
 Jon

That might work...was even thinking about at the bottom so you'd only really
see it from above.

No - if you do not look directly from above, it will look very bad.
Say the leaf canopy is a sphere (two polys with circular mask
textures) and the horizontal poly is a circular disc as well. When
looking with a pitch of say 20 degrees down onto it, you will see the
sphere above and a ellipse below the shere.


Best,

Jim

Bye bye,
Wolfram.

___
Flightgear-devel mailing list
[EMAIL PROTECTED]
http://mail.flightgear.org/mailman/listinfo/flightgear-devel



Re: [Flightgear-devel] Trees, alpha, and the sky

2002-05-02 Thread Andy Ross

Wolfram Kuss wrote:
 Andy Ross wrote:
  You could also experiment with turning off backface culling instead
  of rendering two quads for each direction.  In principle, it should
  be faster.  In practice, it's probably a good way to detect driver
  bugs. :)

 The speed difference will be very, very small.

Are you sure?  It reduces the vertex count by a full factor of two.
For a single tree, that's a savings of 12 vertices; immesurable noise
for a single tree, but for a forest of thousands?  I could easily see
us running into GPU throughput issues with this.

It's probably not something to worry about for a good while, though.
As David points out, it can't be done without hacking plib's
internals.

Andy

-- 
Andrew J. RossNextBus Information Systems
Senior Software Engineer  Emeryville, CA
[EMAIL PROTECTED]  http://www.nextbus.com
Men go crazy in conflagrations.  They only get better one by one.
 - Sting (misquoted)


___
Flightgear-devel mailing list
[EMAIL PROTECTED]
http://mail.flightgear.org/mailman/listinfo/flightgear-devel



RE: [Flightgear-devel] Trees, alpha, and the sky

2002-05-02 Thread Norman Vine

Andy Ross

Wolfram Kuss wrote:
 Andy Ross wrote:
  You could also experiment with turning off backface culling instead
  of rendering two quads for each direction.  In principle, it should
  be faster.  In practice, it's probably a good way to detect driver
  bugs. :)

 The speed difference will be very, very small.

Are you sure?  It reduces the vertex count by a full factor of two.
For a single tree, that's a savings of 12 vertices; immesurable noise
for a single tree, but for a forest of thousands?  

Well if we are worried about vertex counts then
by using impostors for all execpt the very closest trees 
you would only have 4 vertices per for the vast majority :-))

Norman

___
Flightgear-devel mailing list
[EMAIL PROTECTED]
http://mail.flightgear.org/mailman/listinfo/flightgear-devel



Re: [Flightgear-devel] Trees, alpha, and the sky

2002-05-02 Thread Marcio Shimoda

 Or draw the trees as step four?
 

But, how do I know where I can draw a tree?

Marcio Shimoda


___
Flightgear-devel mailing list
[EMAIL PROTECTED]
http://mail.flightgear.org/mailman/listinfo/flightgear-devel



Re: [Flightgear-devel] Trees, alpha, and the sky

2002-05-01 Thread Charles Puffer

Would small dense green clouds work?

Jon Berndt wrote:

Jon Berndt writes:

  can you use a sphere?

That's a lot of triangles for each tree.



Oh, blast it! That's right. OGL has no real sphere.

Jon




___
Flightgear-devel mailing list
[EMAIL PROTECTED]
http://mail.flightgear.org/mailman/listinfo/flightgear-devel



Re: [Flightgear-devel] Trees, alpha, and the sky

2002-05-01 Thread David Megginson

Marcio Shimoda writes:

  What function draws the sky and the tile?

The sky is drawn in SimGear -- see simgear/sky/dome.cxx.  It's not
integrated with the SSG graph, and that may or may not be part of the
problem.


All the best,


David

-- 
David Megginson, [EMAIL PROTECTED], http://www.megginson.com/

___
Flightgear-devel mailing list
[EMAIL PROTECTED]
http://mail.flightgear.org/mailman/listinfo/flightgear-devel



Re: [Flightgear-devel] Trees, alpha, and the sky

2002-05-01 Thread Curtis L. Olson

Marcio Shimoda writes:
 What function draws the sky and the tile?

The code to draw the sky is in simgear.

Curt.
-- 
Curtis Olson   IVLab / HumanFIRST Program   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] Trees, alpha, and the sky

2002-05-01 Thread Curtis L. Olson

David Megginson writes:
 Marcio Shimoda writes:
 
   What function draws the sky and the tile?
 
 The sky is drawn in SimGear -- see simgear/sky/dome.cxx.  It's not
 integrated with the SSG graph, and that may or may not be part of the
 problem.

The scene drawing is split into several phases:

Step 1: the sky dome, the sun, moon, stars, planets.  This forms the
back drop.  Everything else is drawn in front of these sky items.

Step 2: draw the terrain.  This needs to go in front of sky back drop
so it is drawn next.

Step 3: draw the clouds.  The clouds have alpha so they are drawn
last.

However, if the tree has been already drawn, then the sky won't show
throw because the depth buffering of the tree polygons masks it out.

Everything with alpha really needs to be thrown in a big bin and all
drawn back to front in one step.  However, with things like the clouds
that are huge relative everything else, how do you calculate the
distance of the cloud layer?  Probably to the nearest point is what we
want, but I doubt that is what plib uses ... it probably goes off the
'reference' point or center of the bounding sphere of each partially
transparent leaf node.

This is a problem ... this might be a good one to ask on the plib list
to see if they can provide suggestions?

Curt.
-- 
Curtis Olson   IVLab / HumanFIRST Program   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] Trees, alpha, and the sky

2002-05-01 Thread Jim Wilson

David Megginson [EMAIL PROTECTED] said:

 Curtis L. Olson writes:
 
   Step 1: the sky dome, the sun, moon, stars, planets.  This forms the
   back drop.  Everything else is drawn in front of these sky items.
   
   Step 2: draw the terrain.  This needs to go in front of sky back drop
   so it is drawn next.
   
   Step 3: draw the clouds.  The clouds have alpha so they are drawn
   last.
 
 I assume that steps 2 and 3 are in this order so that you'll be able
 to see a mountain (for example) through gaps in the clouds.  If you
 reversed the two, we'd lose that but alpha on trees (etc.) would start
 working.  Ouch.

Or draw the trees as step four?

Best,

Jim

___
Flightgear-devel mailing list
[EMAIL PROTECTED]
http://mail.flightgear.org/mailman/listinfo/flightgear-devel



Re: [Flightgear-devel] Trees, alpha, and the sky

2002-05-01 Thread Curtis L. Olson

Jim Wilson writes:
 David Megginson [EMAIL PROTECTED] said:
 
  Curtis L. Olson writes:
  
Step 1: the sky dome, the sun, moon, stars, planets.  This forms the
back drop.  Everything else is drawn in front of these sky items.

Step 2: draw the terrain.  This needs to go in front of sky back drop
so it is drawn next.

Step 3: draw the clouds.  The clouds have alpha so they are drawn
last.
  
  I assume that steps 2 and 3 are in this order so that you'll be able
  to see a mountain (for example) through gaps in the clouds.  If you
  reversed the two, we'd lose that but alpha on trees (etc.) would start
  working.  Ouch.
 
 Or draw the trees as step four?

If we do that, then you won't see any of the trees through the breaks
in the clouds if you are above a cloud layer.

Curt.
-- 
Curtis Olson   IVLab / HumanFIRST Program   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] Trees, alpha, and the sky

2002-05-01 Thread Jim Wilson

Jim Wilson [EMAIL PROTECTED] said:


 Or draw the trees as step four?
 
Nope that won't work...uggh.

Best,

Jim

___
Flightgear-devel mailing list
[EMAIL PROTECTED]
http://mail.flightgear.org/mailman/listinfo/flightgear-devel



[Flightgear-devel] Trees, alpha, and the sky

2002-04-30 Thread David Megginson

I've been experimenting with a simple tree model in FlightGear.
Here's the result:

  http://www.megginson.com/flightsim/tree.png

The tree consists of a 32x64 RGBA texture mapped onto three sets of
back-to-back rectangles at 120-degree angles to each other.  From any
given viewing angle (other than directly overhead) it looks
more-or-less 3D, but there are two problems:

1. I need to clean up the edges of the texture to avoid artifacts.

2. While other scenery shows through the alpha parts of the texture,
   the sky often doesn't, as you can see in the above screenshot (note
   how you can see the grass and river near the bottom of the trunk, but
   not the sky near the crown).

I suspect that the sky problem has something to do with drawing order,
clip planes, or a combination of the above.  We may need to draw the
sky before the scenery so that it will show through transparent parts.

Any hints?  This model is light-weight enough that we could build
fairly large groves, at least near airports.


All the best,


David

-- 
David Megginson, [EMAIL PROTECTED], http://www.megginson.com/

___
Flightgear-devel mailing list
[EMAIL PROTECTED]
http://mail.flightgear.org/mailman/listinfo/flightgear-devel



Re: [Flightgear-devel] Trees, alpha, and the sky

2002-04-30 Thread Jim Wilson

David Megginson [EMAIL PROTECTED] said:

 
 What about overhead?  Should I plop one more rectangle on the top?
 You don't usually look straight down from a plane, but it will happen
 during aerobatics.
 

Hmmm...it'd be hard to make that not look weird from some angle.

Best,

Jim


___
Flightgear-devel mailing list
[EMAIL PROTECTED]
http://mail.flightgear.org/mailman/listinfo/flightgear-devel



RE: [Flightgear-devel] Trees, alpha, and the sky

2002-04-30 Thread Jon Berndt

 Jon Berndt writes:
 
   can you use a sphere?
 
 That's a lot of triangles for each tree.


Oh, blast it! That's right. OGL has no real sphere.

Jon



smime.p7s
Description: application/pkcs7-signature


RE: [Flightgear-devel] Trees, alpha, and the sky

2002-04-30 Thread David Megginson

Jim Wilson writes:

  That might work...was even thinking about at the bottom so you'd
  only really see it from above.

That might actually be the best idea, but it will make placement very
sensitive (I usually sink models slightly into the ground to avoid
gaps).


All the best,


David

-- 
David Megginson, [EMAIL PROTECTED], http://www.megginson.com/

___
Flightgear-devel mailing list
[EMAIL PROTECTED]
http://mail.flightgear.org/mailman/listinfo/flightgear-devel