Please excuse my previous post, I found the answer soon afterward. I was
stupidly not selecting the projection matrix before my gluPerspective call.
Works now. <sheepish grin>

Jason Maskell
Visiware, France.

"As soon as you make an assumption, it's wrong" - Assumed to be a Murphy's
Law.

Disclaimer: I don't speak for my company. You'll know when I do because it
will be in French.


--
Author:
   INET: backov@...

Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

I'm Looking for a good site for more info on MCD/ICD drivers.....help!

-Suryan
InXight Software Inc.

--
Author:
   INET: stalin@...

Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

Here's how I did a pick... this seems to work when the window is resized ok...
I'm using glOrtho rather than gluPerspective, but maybe this will help...

   wglMakeCurrent(m_hDC,m_hRC));
   glSelectBuffer(1024,selectBuff);

   // Get current viewport and save projection matrix
   glGetIntegerv(GL_VIEWPORT,viewport);
   glMatrixMode(GL_PROJECTION);
   glPushMatrix();

   // Set the render mode for selection and establish a
   // new small clipping volume around the mouse location
   glRenderMode(GL_SELECT);
   glLoadIdentity();
   gluPickMatrix(point.x,viewport[3] - point.y,2,2,viewport); // OpenGL measures
from bottom
   glOrtho (  m_LeftClip*m_wAspect, m_RightClip*m_wAspect,
     m_BottomClip*m_hAspect, m_TopClip*m_hAspect,
     m_NearClip,m_FarClip);

   // Draw the current mesh, with names
   RenderWithNames();

then I pop back the original matrix...

Robert

At 08:51 AM 11/3/98 -0800, you wrote:
>Well, I'm having a bit of a strange problem. (i740 ICD).
>
>The problem is, my OpenGL window is resizable. How I resize it is I simply
>call glViewport and gluPerspective again. It seems to work fine, however,
>when I'm using my selection code, it starts screwing up after the window is
>resized. It looks like it's still stuck in the old RC settings. Is there
>some other call I should be making? What's the "right" way to resize an OGL
>window?
>
>FYI, the selection code in question uses gluPickMatrix and the selection
>buffer, with pushnames, etc, etc. Like I said, it works perfectly until I
>resize the window.
>
>I'll keep working on it, but any wisdom would be appreciated.
>
>
>
>--
>Author:
>  INET: backov@...
>
>Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
>San Diego, California        -- Public Internet access / Mailing Lists
>--------------------------------------------------------------------
>To REMOVE yourself from this mailing list, send an E-Mail message
>to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
>the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
>(or the name of mailing list you want to be removed from).  You may
>also send the HELP command for other information (like subscribing).
>
--
Author: Robert Taylor
   INET: r.taylor@...

Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

Okay,
I got some money to spend and I'm looking for a good OpenGL card.
(development and gaming, not 3d design like Cad and stuff).  I've heard
some good things about TNT, but Riva will soon release a better version
of TNT.  Also, ATI is coming out with its new line of boards, called
Rage I believe (32MBs on board) and an "OpenGL ICD". Does ICD stand for
Isomthing Client Driver ?  This provides the most stable and fastest
OpenGL support, right?
Any response will be greatly appreciated, Paul.

--
Author: Paul R Suszko
   INET: susz7857@...

Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

Well, I'm having a bit of a strange problem. (i740 ICD).

The problem is, my OpenGL window is resizable. How I resize it is I simply
call glViewport and gluPerspective again. It seems to work fine, however,
when I'm using my selection code, it starts screwing up after the window is
resized. It looks like it's still stuck in the old RC settings. Is there
some other call I should be making? What's the "right" way to resize an OGL
window?

FYI, the selection code in question uses gluPickMatrix and the selection
buffer, with pushnames, etc, etc. Like I said, it works perfectly until I
resize the window.

I'll keep working on it, but any wisdom would be appreciated.



--
Author:
   INET: backov@...

Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

Hi Alex,

The FireGL Pro uses an ICD - it's called 3DOGLDRV.DLL.

Note that once you have an OpenGL context, you will have access to the
extensions that are supported by the underlying OpenGL implementation.
  The support for extensions will differ.  For example if you select a
generic pixel format, you will get whatever software extensions the
Microsoft implementation exports.  If you select an accelerated ICD,
you will get what that implementation supports.  Extensions by their
very nature are not a part of core OpenGL, so are not required by all
OpenGL implementations.

Are you looking for the extension for glTexSubImage or the 1.1 version
?  Note that in OpenGL 1.1 this IS core functionality and should exist
on all 1.1 implementations.

Phil.

-----Original Message-----
From: Alex [SMTP:alex_ket@...]
Sent: 01 November 1998 11:07
To: Multiple recipients of list OPENGL-GAMEDEV-L
Subject: [Q] Cannot access 3DOGLDRV

Hi, GL'ers!

I have severe problem with FireGL 1000 Pro ICD.

I have installed opengl2.exe containing some GL extentions I need.
Then
I've installed latest ISD version, nt41kp6f.exe.
I link to OpenGL.lib and Glu.lib. My NT4 registry at
HKEY_LOCAL_MACHINE\SOFTWARE\Silicon Graphics\OpenGL contains now 3
keys:
1.0, 1.1 and 1.1.1.

Former state (one key only: 1.0) with OverrideDispatch==0/1/2 worked
almost adequately: in any case, my DLL's were both OpenGL/Glu and
OpenGL32/Glu32 libraries together with 3DOGLDRV.DLL - Diamond's ICD.
Thus, I had HW acceleration without some important extentions.

Today there are desirable extentions (say, glTexSubimage2D) but at
the
same time there is no HW accel! Windows loads SGI's OpenGL/Glu and no
ICD.

What to do?

Alex
--
Author: Alex
   INET: alex_ket@...

Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
San Diego, California        -- Public Internet access / Mailing
Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

--
Author: Phil Huxley
   INET: Phil.Huxley@...

Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

This probably doesn't interest a whole lot of you, but it's still nice to
see that OpenGL is available for even more platforms

> -----Original Message-----
> This message comes to you from MacDev-1(tm).
> See below for more info on this list (including sub/unsub details).
> __________________________
>
> OPENGL FOR THE MACINTOSH 1.5 (SGI API 1.1) IS NOW AVAILABLE
>
> OpenGL for the Macintosh by Conix 1.5
> 25% performance increase
> 400% performance increase for some texture map modes
>
>
> San Luis Obispo, CA (November 2, 1998) Conix Enterprises, Inc. today
> announced the availability of OpenGL for the Mac 1.5. This is the OpenGL
> the Macintosh graphic's community has been waiting for. With incredible
> performance gains and improved support for hardware acceleration, Conix
> provides the ultimate solution for 3D graphics on the Macintosh.
> ATI Fully Supports OpenGL
>
> "Conix is an excellent partner in bringing accelerated OpenGL to the
> Macintosh platform. ATI has been proud of our working relationship with
> Conix and feel that their modular implementation of OpenGL is elegant in
> design.... The growing number of 3D Rendering Applications and Games being
> developed using the Conix solution is a testament to Conix's robustness
> and
> easy of use."
>  Henry Nienhuis
>  ATI Technologies Inc.
>  Manager, Software Engineering
>
>
> New Features:
>
>       * Improved AGL API: support for the MP library and improved
> off-screen
>         and full-screen support
>       * An overall average of 25% performance increase
>       * Four hundred percent performance increase for some texture map
> modes
>       * Color index mode support
>       * Automatic recovery from insufficient hardware memory failures
>       * Better multi-monitor support
>       * Finer control over hardware renderer selection
>       * Improved documentation
>
> To celebrate the occasion, Conix is offering $100 off the regular SDK
> price
> of $389. This special offer is good through November 20.
>
> Conix provides affordable distribution licensing for application vendors,
> and internal distribution agreements for our commercial customers who are
> developing apps for internal distribution.
>
> See for yourself! Try OpenGL for 10 days with no cost or obligation.
> http://www.conix3d.com/download_request.html
> Conix Enterprises, Inc.
> janet@...
> http://www.conix3d.com
>
> PO Box 4113
> San Luis Obispo, CA 93403
>
> 805.546.1040 voice
> 805.546.1017 fax
>
> __________________________
>
--
Author: Rebecca  McFarland
   INET: RMcFarland@...

Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

On Mon, 2 Nov 1998, brent christopher spillner wrote:

> Right, but going from 320x240 to 800x600 is maybe a 6X slowdown.  He
> seems to be having bigger problems than that.  (I forget what his
> original resolution was, but I think 320x240 is about right).

Well, to this group, he said 200x200 and to comp.graphics.api.opengl
he said 100x100 - and *that* is a factor 48 which is more than enough
to account for blindingly fast performance in a tiny window going
down to 1Hz fullscreen.

Steve Baker                (817)619-2657 (Vox/Vox-Mail)
Raytheon Systems Inc.      (817)619-2466 (Fax)
Work: SBaker@...      http://www.hti.com
Home: SJBaker1@... http://web2.airmail.net/sjbaker1

--
Author: Steve Baker
   INET: sbaker@...

Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

This could be considered advocacy, so I'll keep it short. :-)

Those interested in an example of Microsoft tactics in forcing
a competitor to abandon a market segment or merge technologies
may find the following link enlightening.  This is the transcript
of Apple VP of Software Engineering Avie Tevanian's testimony before
the DOJ from earlier this month.

http://www.usdoj.gov/atr/cases/f2000/2010.htm

Cheers,

Jason Asbahr
jason@...

--
Author:
   INET: jason@...

Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

Hello,

I am trying to implement dynamic lightmaps, and I thought the best way to
download a new texture is with glTexSubImage2D to "replace all or part of a
texture". But how come in 3dfx drivers these functions are not exported? Yet
in any 3dfx quake mini driver they are? These seem like a very important
functions. Is there a better way to do this other than gltexSubImage?

Whats more, on my starfighterAGP the driver seems to allocate more memory
the first time I call this for each texture, and I'm not changing the
texture size. This has the effect of SERIOUSLY degrading the performance in
my app, yet when I run quake the driver doesn't suffer this problem?

Thx for any help.

Hamilton


--
Author:
   INET: hamilton@...

Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

That's been discussed at length on the list.. I suggest that you check the
archives or download the GLUT source and check out the examples..

Mainly, chaning the resolution is an OS problem, not an OpenGL problem.

I'm not sure that I understand why the windows98 OpenGL would be in Beta, I
haven't heard anything about that.

Wilbur


      --------------------------------------------
         Putting A Human Face On Technology ;-)
      --------------------------------------------
--
Author: Wilbur Streett
   INET: WStreett@...

Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

>>The card has an ATI 3D Rage Pro chipset.  It's an 8MB AGP card.  It
>doesn't have OpenGL support in the drivers, but at the same time...it's
>just a rectangle.  I'm running at 800x600x15.  Nothing major.
>
>At that resolution you're not able to double buffer..  that takes up the
>entire memory of the card.

With 8MB he should be able to accellerate at 800x600x15bit, or is ATI 3D Rage
Pro a bit special in this area, because I can accellerate up to 1024x768x24bit,
on my 8MB FireGL1000 PRO.

   Isaack Rasmussen
     - Mayan | Geniue

   "You said that you would love me, until you died... but you are still alive"
            - Alanis Morissette


--
Author:
   INET: geniue@...

Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

>>The card has an ATI 3D Rage Pro chipset.  It's an 8MB AGP card.  It
>>doesn't have OpenGL support in the drivers, but at the same time...it's
>>just a rectangle.  I'm running at 800x600x15.  Nothing major.
>
>At that resolution you're not able to double buffer..  that takes up the
>entire memory of the card..  My own testing has shown the software only

???  800x600 = 480,000, so even at 32bpp two pages would fit in less
than 4MB, leaving all kinds of room for a Z-buffer, textures, etc.

I wonder if the problem is that he doesn't resize his viewport when
the window size is changed, and his OpenGL implementation is doing a
poor job of stretching the image out (possibly trying to dither it or
something).

>OpenGL to be very fill limited, as in the amount of drawing in the frame
>makes a big difference..  texturing is very slow as well in the software
>only OpenGL.

Right, but going from 320x240 to 800x600 is maybe a 6X slowdown.  He
seems to be having bigger problems than that.  (I forget what his
original resolution was, but I think 320x240 is about right).


--
Author: brent christopher spillner
   INET: spillner@...

Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

Here is how I think about it:

Envision the "current" view volume as a 3D solid.
Construct another solid, the "portal volume", as the space visible through
and beyond the window or door.
The new clipping volume is the intersection of the current view volume and
the portal voulme.

Here is how to actually do it:
Clip the portal to the current view volume.
If the portal was not clipped away completely, the new view volume is
bounded by the following planes*:

the plane of the portal
a plane through each edge of the portal and the eye point.

Note that you can't clip to these weird planes using a "clipping coordinate
system", you must have a fully general clipper.

-Gil

*I lied, if the portal is not between the near and far clip planes, there
are some special cases to handle, not difficult, but needs to be considered.

-----Original Message-----
To: Multiple recipients of list OPENGL-GAMEDEV-L
<OPENGL-GAMEDEV-L@...>
Date: Monday, November 02, 1998 3:58 PM


>Hey Tom,
>
>I got (at least I'm pretty confident) the main viewport view frustum
>clipping working correctly.  It was a logic bug.  However, I am doing what
>I said of clipping corners 1 and 4 against the right edge and corners 2 and
>3 against the
>left edge.
>
>I need some help though on how I would create a new view volume via the
>coordinates of a door or window, and then compare them to the objects in
>the adjoining room for clipping.
>
>Thanks
>
>At 09:40 AM 11/2/98 -0800, you wrote:
>>Yay! Clipping!
>>
>>I was talking about this stuff a couple weeks ago and was referred to the
>>following EXCELLENT text (thanks BTW!)
>>
>>Procedural Elements for Computer Graphics
>>David F. Rogers
>>ISBN: 0-07-053548-5
>>
>>Here's the skinny of it though ... In most cases you don't need to do the
>>clipping yourself. If you're clipping to the view volume then let OpenGL
do
>>it for you. If you're clipping a region that could be culled with
z-buffer,
>>then let the z-buffer handle it. If you're clipping to some smaller region
>>of the screen (a portal while doing a reflection or something like that)
and
>>you need to do your own clipping so you don't need to use stencil or
>>z-buffer tricks then you should pick up the book and check out Chapter 3 -
>>Clipping. Sutherland Hodgman clipping is probably gonna work best (still
>>reading in spare time).
>>
>>It sounds like you're doing view frustum clipping so don't even bother
with
>>it, after you've culled out the non-intersecting triangles just let OpenGL
>>handle the clipping for you.
>>
>>Tom
>>
>>-----Original Message-----
>>To: Multiple recipients of list OPENGL-GAMEDEV-L
>><OPENGL-GAMEDEV-L@...>
>>Date: Saturday, October 31, 1998 12:52 PM
>>
>>
>>>I'm already doing back face culling, and full polygon clipping of polys
>>>that are completely outside the view frustum. Now while I'm trying to
>>>improve things, I was wondering:
>>>
>>>1. Is it worth performing partial poly clipping of poly's which are
>>>partially seen and partially off the edge of the view?  Since OpenGL
>>>handles it well anyway, and performs correct clipping, and since a poly
>>>needs to be rendered anyway, does it matter if the poly is larger then
>>>needed and clipped by OpenGL, or would it help performance noticeably if
I
>>>perform further clipping before sending it to OpenGL?
>>>
>>>2. Normally, partial poly clipping is performed using screen coords (in
>>>2D), not world or camera coords.  Therefore, is it even possible to do
>>>partial poly clipping before sending to OpenGL, since I'm sending OpenGL
>>>camera coords?  If so, can anyone explain how it would be performed.  All
I
>>>can think of would be:
>>>
>>> 1. to convert the camera coords to screen coords
>>> 2. then perform 2d clipping
>>> 3. then convert the new coords back to camera coords and
>>> 4. send to OpenGL for rendering
>>>
>>>This seems very inefficient and overkill.  Second, would be to perform
>>>steps 1 and 2, then render the object to OpenGL as a 2D object.  However,
>>>then I'd suspect I'd run into problems with texture mapping and proper
>>>lighting.
>>>
>>>Thanks,
>>>
>>>Larry
>>>
>>>--
>>>Author: Larry Grant
>>>  INET: lgrant@...
>>>
>>>Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
>>>San Diego, California        -- Public Internet access / Mailing Lists
>>>--------------------------------------------------------------------
>>>To REMOVE yourself from this mailing list, send an E-Mail message
>>>to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
>>>the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
>>>(or the name of mailing list you want to be removed from).  You may
>>>also send the HELP command for other information (like subscribing).
>>>
>>>
>>
>>--
>>Author:
>>  INET: tomh@...
>>
>>Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
>>San Diego, California        -- Public Internet access / Mailing Lists
>>--------------------------------------------------------------------
>>To REMOVE yourself from this mailing list, send an E-Mail message
>>to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
>>the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
>>(or the name of mailing list you want to be removed from).  You may
>>also send the HELP command for other information (like subscribing).
>
>--
>Author: Laurence Grant
>  INET: lgrant@...
>
>Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
>San Diego, California        -- Public Internet access / Mailing Lists
>--------------------------------------------------------------------
>To REMOVE yourself from this mailing list, send an E-Mail message
>to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
>the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
>(or the name of mailing list you want to be removed from).  You may
>also send the HELP command for other information (like subscribing).

--
Author:
   INET: ggribb@...

Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

Hey Tom,

I got (at least I'm pretty confident) the main viewport view frustum
clipping working correctly.  It was a logic bug.  However, I am doing what
I said of clipping corners 1 and 4 against the right edge and corners 2 and
3 against the
left edge.

I need some help though on how I would create a new view volume via the
coordinates of a door or window, and then compare them to the objects in
the adjoining room for clipping.

Thanks

At 09:40 AM 11/2/98 -0800, you wrote:
>Yay! Clipping!
>
>I was talking about this stuff a couple weeks ago and was referred to the
>following EXCELLENT text (thanks BTW!)
>
>Procedural Elements for Computer Graphics
>David F. Rogers
>ISBN: 0-07-053548-5
>
>Here's the skinny of it though ... In most cases you don't need to do the
>clipping yourself. If you're clipping to the view volume then let OpenGL do
>it for you. If you're clipping a region that could be culled with z-buffer,
>then let the z-buffer handle it. If you're clipping to some smaller region
>of the screen (a portal while doing a reflection or something like that) and
>you need to do your own clipping so you don't need to use stencil or
>z-buffer tricks then you should pick up the book and check out Chapter 3 -
>Clipping. Sutherland Hodgman clipping is probably gonna work best (still
>reading in spare time).
>
>It sounds like you're doing view frustum clipping so don't even bother with
>it, after you've culled out the non-intersecting triangles just let OpenGL
>handle the clipping for you.
>
>Tom
>
>-----Original Message-----
>To: Multiple recipients of list OPENGL-GAMEDEV-L
><OPENGL-GAMEDEV-L@...>
>Date: Saturday, October 31, 1998 12:52 PM
>
>
>>I'm already doing back face culling, and full polygon clipping of polys
>>that are completely outside the view frustum. Now while I'm trying to
>>improve things, I was wondering:
>>
>>1. Is it worth performing partial poly clipping of poly's which are
>>partially seen and partially off the edge of the view?  Since OpenGL
>>handles it well anyway, and performs correct clipping, and since a poly
>>needs to be rendered anyway, does it matter if the poly is larger then
>>needed and clipped by OpenGL, or would it help performance noticeably if I
>>perform further clipping before sending it to OpenGL?
>>
>>2. Normally, partial poly clipping is performed using screen coords (in
>>2D), not world or camera coords.  Therefore, is it even possible to do
>>partial poly clipping before sending to OpenGL, since I'm sending OpenGL
>>camera coords?  If so, can anyone explain how it would be performed.  All I
>>can think of would be:
>>
>> 1. to convert the camera coords to screen coords
>> 2. then perform 2d clipping
>> 3. then convert the new coords back to camera coords and
>> 4. send to OpenGL for rendering
>>
>>This seems very inefficient and overkill.  Second, would be to perform
>>steps 1 and 2, then render the object to OpenGL as a 2D object.  However,
>>then I'd suspect I'd run into problems with texture mapping and proper
>>lighting.
>>
>>Thanks,
>>
>>Larry
>>
>>--
>>Author: Larry Grant
>>  INET: lgrant@...
>>
>>Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
>>San Diego, California        -- Public Internet access / Mailing Lists
>>--------------------------------------------------------------------
>>To REMOVE yourself from this mailing list, send an E-Mail message
>>to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
>>the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
>>(or the name of mailing list you want to be removed from).  You may
>>also send the HELP command for other information (like subscribing).
>>
>>
>
>--
>Author:
>  INET: tomh@...
>
>Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
>San Diego, California        -- Public Internet access / Mailing Lists
>--------------------------------------------------------------------
>To REMOVE yourself from this mailing list, send an E-Mail message
>to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
>the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
>(or the name of mailing list you want to be removed from).  You may
>also send the HELP command for other information (like subscribing).

--
Author: Laurence Grant
   INET: lgrant@...

Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

| > There are some specialized processing techniques at texture edges
| > that are useful -- clamping to a constant border color, for example.
|
| Yes - but putting a one-texel border *into* the texture and then using
| GL_CLAMP solves most of those problems.

True for some markets.  People using video or digitized film as texture,
on the other hand, can't always afford to lose information on the edges
and don't have enough time on each frame to rescale the image to make room
for a border.  The clamp-to-border-color solution is easy to implement
and covers a lot of applications with less hassle for software developers.

| ...still - hindsight is an overly precise weapon - and I can't imagine
| features ever vanishing from future OpenGL specs.

On the other hand, if sufficiently many vendors choose not to implement
something, then it may become effectively unusable.  (This applies to
any API, obviously.)  It can be a tough call.  Texture borders are a
good example.

Allen
--
Author:
   INET: akin@...

Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

Rainbow 6 is a Direct3D app, and the ATI 3D Rage Pro chipset does support
accelerated Direct3D rendering.  Hence, why it runs fast.  Also, just for
information's sake you can download an openGL beta driver
for your card that works for most things (especially with what you are doing :)
.   It is on their site somewhere under game information.

Thanks,
Ryan


Ron Barbosa wrote:

> > Depending on the amount of memory that you have on your graphics card, you
> > may end up dropping back to a software only version of OpenGL because the
> > graphics card can't support the resolution and acceleration at the same
time.
> >
> > What card are you using at what resolution and pixel depth?  You may be
> > able to reduce the memory requirements enough by > changing either setting.
>
> The card has an ATI 3D Rage Pro chipset.  It's an 8MB AGP card.  It doesn't
have OpenGL support in the drivers, but at the same time...it's just a
rectangle.  I'm running at 800x600x15.  Nothing major.
>
> I'm missing something, because the card (without OpenGL drivers) plays Rainbow
6 (an OpenGL app) in 640x480 without a hitch.  Very fluid...looks great.
>
> The guys at Red Storm obviously know something that I don't. =)
>
> Thanks in advance for any additional info.
> --RB
>
> -----
> See the original message at
http://www.egroups.com/list/opengl-gamedev-l/?start=8348
> --
> Free e-mail group hosting at http://www.eGroups.com/
> --
> Author:
>   INET: rbarbosa@...
>
> Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
> San Diego, California        -- Public Internet access / Mailing Lists
> --------------------------------------------------------------------
> To REMOVE yourself from this mailing list, send an E-Mail message
> to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
> the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
> (or the name of mailing list you want to be removed from).  You may
> also send the HELP command for other information (like subscribing).



--
Author:
   INET: stradlir@...

Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

OK, I'm convinced that I should use an FFT or at least a better linear
filter for my mipmaps. However, I, like I'm sure most of the people on this
list, have to work on a game, and don't have time to work this out. So, does
anybody have code that is plug-compatible with gluBuild2DMipMap()? Actually,
that's not what I want. I want a program that will build a file that I can
then load with a replacement for gluBuild2DMipMap(), since presumably it
will be too slow to do this at game load time.

> -----Original Message-----
> From: root@... [mailto:root@...]On Behalf Of Gil Gribb
> Sent: Monday, November 02, 1998 1:54 PM
> To: Multiple recipients of list OPENGL-GAMEDEV-L
> Subject: Re: Are Texture Borders widely supported?
>
>
> But be advised that any FFT based filter method can be done with a linear
> filter. The linear filter corresponding to a FFT filter is obtained as the
> inverse fourier transform of the smoothing kernel. Unfortuantely
> the linear
> filter is the same size as the image, leaving us with a
> width*width*height*height filtering algorithm.
>
> However the filters used for mipping can be well approximated by a small
> linear filter, since the filter coefficients typically decay exponentially
> from the center out. These small filters are easier to code and
> faster than
> the fft method.
> -Gil
>
>
>
>
>
> >Hello,
> >
> >Yes, a box filter works.  As with anything, though, there is always
> >something better.
> >It was pointed out to me that one can use a Fourier transform on the
> >image.
> >Perform a low pass filter (eliminate the high frequencies), and inverse
> >transform to
> >get the lower MIP.  I hadn't bothered working out the specifics as I
> >have to create
> >MIPMAPs at runtime and a box filter is fast.
> >
> >Regards,
> >
> >David S. Lee Swenson
> >
> > -----Original Message-----
> > From: Mayers, P J
> > Sent: Monday, November 02, 1998 8:54 AM
> > To: Multiple recipients of list OPENGL-GAMEDEV-L
> > Subject: RE: Are Texture Borders widely supported?
> >
> > I believe it uses a box filter. Someone commented that this (or
> >whatever
> > filter it does use) was Wrong(tm) for mipmap generation. I don't
> >remember
> > who, but I seem to remember they were one of those people you
> >normally
> > listen to. Then again, someone was just saying the other week
> >that a box
> > filter was ideal. Go figure.
> >
> > Anyone care to comment?
> >
> > Cheers,
> > Phil
> >
> >
> >--
> >Author: David Lee
> >  INET: davidl@...
> >
> >Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
> >San Diego, California        -- Public Internet access / Mailing Lists
> >--------------------------------------------------------------------
> >To REMOVE yourself from this mailing list, send an E-Mail message
> >to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
> >the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
> >(or the name of mailing list you want to be removed from).  You may
> >also send the HELP command for other information (like subscribing).
>
> --
> Author:
>   INET: ggribb@...
>
> Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
> San Diego, California        -- Public Internet access / Mailing Lists
> --------------------------------------------------------------------
> To REMOVE yourself from this mailing list, send an E-Mail message
> to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
> the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
> (or the name of mailing list you want to be removed from).  You may
> also send the HELP command for other information (like subscribing).
>

--
Author:
   INET: benton@...

Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

At 12:24 PM 11/2/98 -0800, you wrote:
>The card has an ATI 3D Rage Pro chipset.  It's an 8MB AGP card.  It
doesn't have OpenGL support in the drivers, but at the same time...it's
just a rectangle.  I'm running at 800x600x15.  Nothing major.

At that resolution you're not able to double buffer..  that takes up the
entire memory of the card..  My own testing has shown the software only
OpenGL to be very fill limited, as in the amount of drawing in the frame
makes a big difference..  texturing is very slow as well in the software
only OpenGL.

>I'm missing something, because the card (without OpenGL drivers) plays
Rainbow 6 (an OpenGL app) in 640x480 without a hitch.  Very fluid...looks
great.

Could be double buffered, and optimized for use in software only.. There's
also winswaphint and some other tricks.. depends on the implementation.

Wilbur

      --------------------------------------------
         Putting A Human Face On Technology ;-)
      --------------------------------------------
--
Author: Wilbur Streett
   INET: WStreett@...

Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

On Mon, Tony Cox wrote:

>NT, then NT5 beta 2 has full DX6 support which is a much better way to go.

Unless you want hardware, which still isn't supported (or at least,
no-one has thought to do any drivers...) - so probably best sticking
to OpenGL.. (hey look, we're back on topic... kind of :)


Jase.

--
Author: Jason G Doig
   INET: jase@...

Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

> If you already have hardware 3D then perhaps you are somehow not getting it
turned
> on - post your question again but with more hardware details and information
about
> which display resolutions, bit depths, etc you have chosen. With many
implementations,
> choosing a resolution/colour-depth that the hardware can't support results in
it
> falling back on software rendering.

Right now, I'm using a card based on the ATI 3d Rage Pro Turbo chipset.  It's an
8MB AGP card.  I run it in 800x600x15.  The OpenGL driver for Win98 is available
in beta but it's very buggy.  I understand what you're saying about the pixel
increase.  So maybe I should ask my question this way.  Is there a way for
OpenGL to switch me out to full screen and change my resolution to something
more reasonable?  I'm sure this is possible because OpenGL apps like Rainbow Six
and GLQuake are running quite well on my machine WITHOUT OpenGL drivers (believe
it or not).  I guess what I'm looking for is some kind of analog to exclusive
mode in DirectX.

Thanks again for any info!
--RB

-----
See the original message at
http://www.egroups.com/list/opengl-gamedev-l/?start=8356
--
Free e-mail group hosting at http://www.eGroups.com/
--
Author:
   INET: rbarbosa@...

Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

> Depending on the amount of memory that you have on your graphics card, you
> may end up dropping back to a software only version of OpenGL because the
> graphics card can't support the resolution and acceleration at the same time.
>
> What card are you using at what resolution and pixel depth?  You may be
> able to reduce the memory requirements enough by > changing either setting.

The card has an ATI 3D Rage Pro chipset.  It's an 8MB AGP card.  It doesn't have
OpenGL support in the drivers, but at the same time...it's just a rectangle. 
I'm running at 800x600x15.  Nothing major.

I'm missing something, because the card (without OpenGL drivers) plays Rainbow 6
(an OpenGL app) in 640x480 without a hitch.  Very fluid...looks great.

The guys at Red Storm obviously know something that I don't. =)

Thanks in advance for any additional info.
--RB


-----
See the original message at
http://www.egroups.com/list/opengl-gamedev-l/?start=8348
--
Free e-mail group hosting at http://www.eGroups.com/
--
Author:
   INET: rbarbosa@...

Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

On Mon, 2 Nov 1998 akin@... wrote:

> Steve Baker wrote:
> | I think texture borders were a severe mistake and should never have been in
> | the OpenGL spec in the first place.
>
> I think most of us who were involved in OpenGL in the early days would
> agree.  The original texture border mechanism was an attempt to solve
> a couple of different problems (using textures as decals, and tiling
> textures to get around texture memory size limitations) but turned out
> to be both difficult to implement in hardware and not as effective as
> other solutions.
>
> There are some specialized processing techniques at texture edges
> that are useful -- clamping to a constant border color, for example.

Yes - but putting a one-texel border *into* the texture and then using
GL_CLAMP solves most of those problems.

...still - hindsight is an overly precise weapon - and I can't imagine
features ever vanishing from future OpenGL specs.

Steve Baker                (817)619-2657 (Vox/Vox-Mail)
Raytheon Systems Inc.      (817)619-2466 (Fax)
Work: SBaker@...      http://www.hti.com
Home: SJBaker1@... http://web2.airmail.net/sjbaker1

--
Author: Steve Baker
   INET: sbaker@...

Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

But be advised that any FFT based filter method can be done with a linear
filter. The linear filter corresponding to a FFT filter is obtained as the
inverse fourier transform of the smoothing kernel. Unfortuantely the linear
filter is the same size as the image, leaving us with a
width*width*height*height filtering algorithm.

However the filters used for mipping can be well approximated by a small
linear filter, since the filter coefficients typically decay exponentially
from the center out. These small filters are easier to code and faster than
the fft method.
-Gil





>Hello,
>
>Yes, a box filter works.  As with anything, though, there is always
>something better.
>It was pointed out to me that one can use a Fourier transform on the
>image.
>Perform a low pass filter (eliminate the high frequencies), and inverse
>transform to
>get the lower MIP.  I hadn't bothered working out the specifics as I
>have to create
>MIPMAPs at runtime and a box filter is fast.
>
>Regards,
>
>David S. Lee Swenson
>
> -----Original Message-----
> From: Mayers, P J
> Sent: Monday, November 02, 1998 8:54 AM
> To: Multiple recipients of list OPENGL-GAMEDEV-L
> Subject: RE: Are Texture Borders widely supported?
>
> I believe it uses a box filter. Someone commented that this (or
>whatever
> filter it does use) was Wrong(tm) for mipmap generation. I don't
>remember
> who, but I seem to remember they were one of those people you
>normally
> listen to. Then again, someone was just saying the other week
>that a box
> filter was ideal. Go figure.
>
> Anyone care to comment?
>
> Cheers,
> Phil
>
>
>--
>Author: David Lee
>  INET: davidl@...
>
>Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
>San Diego, California        -- Public Internet access / Mailing Lists
>--------------------------------------------------------------------
>To REMOVE yourself from this mailing list, send an E-Mail message
>to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
>the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
>(or the name of mailing list you want to be removed from).  You may
>also send the HELP command for other information (like subscribing).

--
Author:
   INET: ggribb@...

Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

Steve Baker wrote:
| I think texture borders were a severe mistake and should never have been in
| the OpenGL spec in the first place.

I think most of us who were involved in OpenGL in the early days would
agree.  The original texture border mechanism was an attempt to solve
a couple of different problems (using textures as decals, and tiling
textures to get around texture memory size limitations) but turned out
to be both difficult to implement in hardware and not as effective as
other solutions.

There are some specialized processing techniques at texture edges
that are useful -- clamping to a constant border color, for example.
I believe most of these have been incorporated in extensions implemented
by HP and SGI, and several of them found their way into OpenGL 1.2.

Allen
--
Author:
   INET: akin@...

Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

Dominic Curran asked:
| Whats wrong with gluBuild2DMipmaps ?  Anyone ?

To ensure consistency across all OpenGL implementations, gluBuild2DMipmaps
is required to use box filtering.  It's possible to do better (in some
situations, a lot better).

Also, since gluBuild2DMipmaps has to deal with a large number of image
sizes and formats, it doesn't have the fastest possible code for every
case.  (There's a good opportunity here for vendors to use dynamic code
generation, but I don't know if any do.)

Allen
--
Author:
   INET: akin@...

Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

On Mon, 2 Nov 1998, Mayers, P J wrote:

> I believe it uses a box filter. Someone commented that this (or whatever
> filter it does use) was Wrong(tm) for mipmap generation. I don't remember
> who, but I seem to remember they were one of those people you normally
> listen to. Then again, someone was just saying the other week that a box
> filter was ideal. Go figure.
>
> Anyone care to comment?

OK, one more time.

* glBuild2DMipmaps uses a box filter. That's a fancy way to say
   that it adds up four texels and divides by four.

* The argument that this is A Good Thing (tm) is that the MIPmap
   GL_LINEAR filters are also (essentially) box filters. So somehow,
   it's a good thing if they match.

* I argue that since the whole idea of MIPmapping is to dispose of
   the high frequencies in the map that would make it alias, the most
   mathematically "pure" way to do that is to convert the map into the
   frequency domain, delete the high frequencies and convert it back
   into the spatial domain. This technique has sound theoretical backing.

* Erm - maybe some fancy wavelet thingy would be *more* mathematically
   sound for GL_CLAMP maps since FFT's really want to assume we have
   GL_REPEAT.  I have a good kludge for that so FFT's still work - but
   theoretically, it's still a kludge. I'm not a wavelet expert - so I
   stick with FFT's.

* Who cares about the math? FFT'ed maps look noticable better - end
   of theoretical discussion.

* Not a LOT better - so if you are in a hurry (or mathematically
   challenged), go ahead and use glBuild2DMipmaps.

* FFT's are S-L-O-W (quite ironic given their name) so not a
   realtime technique.

* FFT'ed maps look kinda nasty when you look at them in a
   paint program. They have weird 'ringing' effects around the
   edges of high contrast edges - but when properly rendered
   with GL_LINEAR_MIPMAP_LINEAR, they look good.

* I need to experiment with GL_LINEAR_MIPMAP_NEAREST (which is
   all that some PC cards can produce) - FFT's *may* not be
   appropriate under those circumstances (but there is no
   reason to assume that Box Filtering is any better).

* There are probably issues to be dealt with when you start
   thinking about crap 3D cards that don't do proper gamma
   correction - but things are already so screwed up under
   those circumstances that all bets are off anyway.

Steve Baker                (817)619-2657 (Vox/Vox-Mail)
Raytheon Systems Inc.      (817)619-2466 (Fax)
Work: SBaker@...      http://www.hti.com
Home: SJBaker1@... http://web2.airmail.net/sjbaker1

--
Author: Steve Baker
   INET: sbaker@...

Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

Hello,

Yes, a box filter works.  As with anything, though, there is always
something better.
It was pointed out to me that one can use a Fourier transform on the
image.
Perform a low pass filter (eliminate the high frequencies), and inverse
transform to
get the lower MIP.  I hadn't bothered working out the specifics as I
have to create
MIPMAPs at runtime and a box filter is fast.

Regards,

David S. Lee Swenson

	 -----Original Message-----
	 From: Mayers, P J
	 Sent: Monday, November 02, 1998 8:54 AM
	 To: Multiple recipients of list OPENGL-GAMEDEV-L
	 Subject: RE: Are Texture Borders widely supported?

	 I believe it uses a box filter. Someone commented that this (or
whatever
	 filter it does use) was Wrong(tm) for mipmap generation. I don't
remember
	 who, but I seem to remember they were one of those people you
normally
	 listen to. Then again, someone was just saying the other week
that a box
	 filter was ideal. Go figure.

	 Anyone care to comment?

	 Cheers,
	 Phil


--
Author: David Lee
   INET: davidl@...

Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).

On Mon, 2 Nov 1998, Dominic Curran wrote:

> Steve Baker wrote:

> Why do you say 'ugh' here Steve ?

I like the sound of the word 'ugh' - altough 'ick' is also becoming a
personal favorite   :-)

> Whats wrong with gluBuild2DMipmaps ?  Anyone ?

Havn't we had that thread already?

Basically, glBuild2Dmipmaps uses a box filter - which is a pretty terrible
choice. It was probably chosen for *speed* and because it matches what
the old IrisGL did automagically - which was obviously of some importance
to SGI at the time. You almost always need to do something different for
alpha-modulation textures.

I use an off-line FFT (Fast Fourier Transform) approach - which I'm not
going into *again* on this list - check the recent archives. Our FFT
MIPmapper generates noticably sharper images than glBuild2DMipmaps.

Steve Baker                (817)619-2657 (Vox/Vox-Mail)
Raytheon Systems Inc.      (817)619-2466 (Fax)
Work: SBaker@...      http://www.hti.com
Home: SJBaker1@... http://web2.airmail.net/sjbaker1

--
Author: Steve Baker
   INET: sbaker@...

Fat City Network Services    -- (619) 538-5051  FAX: (619) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru@... (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB OPENGL-GAMEDEV-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).