Re: R200 DOT3 EXT problems (was Re: [Dri-devel] mesa 4.1 branch /NO go on 2.5.48)

2002-11-21 Thread Adam K Kirchhoff

 On Wed, 20 Nov 2002, Ian Romanick wrote:
 
  On Tue, Nov 19, 2002 at 11:06:16PM -0800, Linus Torvalds wrote:
   
   On Wed, 20 Nov 2002, Dieter Nützel wrote:
Am Mittwoch, 20. November 2002 00:55 schrieb Ian Molton:
 
  Linus updated 2.5.48 with Brian's latest DRM r200 stuff so I should do
  some testing.

No go so far.

Modules are somewhat broken in 2.5.48.
   
   One approach is to not use modules, just compile the thing in. Works for
   me (damn, I'd like to see how the commercial tuxracer looks with bump 
   mapping. But apparently DRI doesn't support the right extension or 
   something ;)
  
  The problem is that it uses EXT_texture_env_dot3 (which the driver does
  advertise), but the driver doesn't actually implement only implements
  EXT_texture_env_dot3 or ARB_texture_env_dot3 correctly.  I noted this in an
  earlier thread about glaxium, but I haven't posted a fix yet.
  
  I have attached a patch to the DRI CVS trunk that should fix the problem.  I
  don't have access to an 8500, so I haven't actually tested it.  I would
  really appreciate it if people with 8500 (or 9000) cards could try this
  patch and tell me if it works.  If I get a few positive results and no
  negatives, I'll commit the patch.

I've applied the patch to two machines, one running Linux and one running 
FreeBSD.  I compiled on both machines and installed.  Unfortunately, I 
won't be able to test the Linux box till I get home, but I'm going to 
assume that the results are going to be the same :-)

http://memory.visualtech.com/glaxium.png

Unfortunately, it doesn't look like the patch worked :-(

Adam




---
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
___
Dri-devel mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/dri-devel



Re: R200 DOT3 EXT problems (was Re: [Dri-devel] mesa 4.1 branch /NO go on 2.5.48)

2002-11-21 Thread Brian Paul
Adam K Kirchhoff wrote:

On Wed, 20 Nov 2002, Ian Romanick wrote:



On Tue, Nov 19, 2002 at 11:06:16PM -0800, Linus Torvalds wrote:


On Wed, 20 Nov 2002, Dieter N?tzel wrote:


Am Mittwoch, 20. November 2002 00:55 schrieb Ian Molton:


Linus updated 2.5.48 with Brian's latest DRM r200 stuff so I should do
some testing.



No go so far.

Modules are somewhat broken in 2.5.48.


One approach is to not use modules, just compile the thing in. Works for
me (damn, I'd like to see how the commercial tuxracer looks with bump 
mapping. But apparently DRI doesn't support the right extension or 
something ;)

The problem is that it uses EXT_texture_env_dot3 (which the driver does
advertise), but the driver doesn't actually implement only implements
EXT_texture_env_dot3 or ARB_texture_env_dot3 correctly.  I noted this in an
earlier thread about glaxium, but I haven't posted a fix yet.

I have attached a patch to the DRI CVS trunk that should fix the problem.  I
don't have access to an 8500, so I haven't actually tested it.  I would
really appreciate it if people with 8500 (or 9000) cards could try this
patch and tell me if it works.  If I get a few positive results and no
negatives, I'll commit the patch.




I've applied the patch to two machines, one running Linux and one running 
FreeBSD.  I compiled on both machines and installed.  Unfortunately, I 
won't be able to test the Linux box till I get home, but I'm going to 
assume that the results are going to be the same :-)

http://memory.visualtech.com/glaxium.png

Unfortunately, it doesn't look like the patch worked :-(

I'm not sure what it's supposed to look like.  Can you try software
rendering, just to get a screen-shot.?

-Brian





---
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
___
Dri-devel mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/dri-devel



Re: R200 DOT3 EXT problems (was Re: [Dri-devel] mesa 4.1 branch / NO go on 2.5.48)

2002-11-21 Thread Ian Romanick
On Thu, Nov 21, 2002 at 08:28:01AM -0500, Adam K Kirchhoff wrote:
 
  On Wed, 20 Nov 2002, Ian Romanick wrote:
  
   On Tue, Nov 19, 2002 at 11:06:16PM -0800, Linus Torvalds wrote:

On Wed, 20 Nov 2002, Dieter Nützel wrote:
 Am Mittwoch, 20. November 2002 00:55 schrieb Ian Molton:
  
   Linus updated 2.5.48 with Brian's latest DRM r200 stuff so I should do
   some testing.
 
 No go so far.
 
 Modules are somewhat broken in 2.5.48.

One approach is to not use modules, just compile the thing in. Works for
me (damn, I'd like to see how the commercial tuxracer looks with bump 
mapping. But apparently DRI doesn't support the right extension or 
something ;)
   
   The problem is that it uses EXT_texture_env_dot3 (which the driver does
   advertise), but the driver doesn't actually implement only implements
   EXT_texture_env_dot3 or ARB_texture_env_dot3 correctly.  I noted this in an
   earlier thread about glaxium, but I haven't posted a fix yet.
   
   I have attached a patch to the DRI CVS trunk that should fix the problem.  I
   don't have access to an 8500, so I haven't actually tested it.  I would
   really appreciate it if people with 8500 (or 9000) cards could try this
   patch and tell me if it works.  If I get a few positive results and no
   negatives, I'll commit the patch.
 
 I've applied the patch to two machines, one running Linux and one running 
 FreeBSD.  I compiled on both machines and installed.  Unfortunately, I 
 won't be able to test the Linux box till I get home, but I'm going to 
 assume that the results are going to be the same :-)
 
 http://memory.visualtech.com/glaxium.png
 
 Unfortunately, it doesn't look like the patch worked :-(

My bad!  Change the  at the end of line 1082 in r200_texstate.c to ||.
The if-statement that compares against GL_DOT3_{RGB,RGBA}_EXT should match
the one that compares against GL_DOT3_{RGB,RGBA}.  Sorry.

-- 
Smile!  http://antwrp.gsfc.nasa.gov/apod/ap990315.html


---
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
___
Dri-devel mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/dri-devel



Re: R200 DOT3 EXT problems (was Re: [Dri-devel] mesa 4.1 branch /NO go on 2.5.48)

2002-11-21 Thread Adam K Kirchhoff

On Thu, 21 Nov 2002, Ian Romanick wrote:

 On Thu, Nov 21, 2002 at 08:28:01AM -0500, Adam K Kirchhoff wrote:
  
   On Wed, 20 Nov 2002, Ian Romanick wrote:
The problem is that it uses EXT_texture_env_dot3 (which the driver does
advertise), but the driver doesn't actually implement only implements
EXT_texture_env_dot3 or ARB_texture_env_dot3 correctly.  I noted this in an
earlier thread about glaxium, but I haven't posted a fix yet.

I have attached a patch to the DRI CVS trunk that should fix the problem.  I
don't have access to an 8500, so I haven't actually tested it.  I would
really appreciate it if people with 8500 (or 9000) cards could try this
patch and tell me if it works.  If I get a few positive results and no
negatives, I'll commit the patch.
  
  I've applied the patch to two machines, one running Linux and one running 
  FreeBSD.  I compiled on both machines and installed.  Unfortunately, I 
  won't be able to test the Linux box till I get home, but I'm going to 
  assume that the results are going to be the same :-)
  
  http://memory.visualtech.com/glaxium.png
  
  Unfortunately, it doesn't look like the patch worked :-(
 
 My bad!  Change the  at the end of line 1082 in r200_texstate.c to ||.
 The if-statement that compares against GL_DOT3_{RGB,RGBA}_EXT should match
 the one that compares against GL_DOT3_{RGB,RGBA}.  Sorry.

Unfortunately, that only seemed to make things worse.  If you check out 
the same URL I posted above, you'll see what I mean.

Adam




---
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
___
Dri-devel mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/dri-devel



Re: R200 DOT3 EXT problems (was Re: [Dri-devel] mesa 4.1 branch / NO go on 2.5.48)

2002-11-21 Thread Ian Romanick
On Thu, Nov 21, 2002 at 11:57:32AM -0500, Adam K Kirchhoff wrote:
 
 On Thu, 21 Nov 2002, Ian Romanick wrote:
 
  On Thu, Nov 21, 2002 at 08:28:01AM -0500, Adam K Kirchhoff wrote:
   
On Wed, 20 Nov 2002, Ian Romanick wrote:
 The problem is that it uses EXT_texture_env_dot3 (which the driver does
 advertise), but the driver doesn't actually implement only implements
 EXT_texture_env_dot3 or ARB_texture_env_dot3 correctly.  I noted this in an
 earlier thread about glaxium, but I haven't posted a fix yet.
 
 I have attached a patch to the DRI CVS trunk that should fix the problem.  I
 don't have access to an 8500, so I haven't actually tested it.  I would
 really appreciate it if people with 8500 (or 9000) cards could try this
 patch and tell me if it works.  If I get a few positive results and no
 negatives, I'll commit the patch.
   
   I've applied the patch to two machines, one running Linux and one running 
   FreeBSD.  I compiled on both machines and installed.  Unfortunately, I 
   won't be able to test the Linux box till I get home, but I'm going to 
   assume that the results are going to be the same :-)
   
   http://memory.visualtech.com/glaxium.png
   
   Unfortunately, it doesn't look like the patch worked :-(
  
  My bad!  Change the  at the end of line 1082 in r200_texstate.c to ||.
  The if-statement that compares against GL_DOT3_{RGB,RGBA}_EXT should match
  the one that compares against GL_DOT3_{RGB,RGBA}.  Sorry.
 
 Unfortunately, that only seemed to make things worse.  If you check out 
 the same URL I posted above, you'll see what I mean.

Hmm...could you try two quick tests for me?  Both would be with an unpatched
driver.

1. Try running with LIBGL_ALWAYS_INDIRECT=1.  I expect that will produce the
same results I see on a Radeon M6.

2. Replace GL_DOT3_RGB_EXT on line 592 of scene.cpp (in glaxium 0.5) with
GL_DOT3_RGB and rebuild glaxium.  If that doesn't work, then dot3
bumpmapping is totally hosed in the R200 driver.  I expect that this will
also produce correct results.

-- 
Smile!  http://antwrp.gsfc.nasa.gov/apod/ap990315.html


---
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
___
Dri-devel mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/dri-devel



Re: R200 DOT3 EXT problems (was Re: [Dri-devel] mesa 4.1 branch / NO go on 2.5.48)

2002-11-21 Thread Ian Romanick
On Thu, Nov 21, 2002 at 01:27:03PM -0500, Adam K Kirchhoff wrote:

 Hmmm... Well, line 592 of scene.cpp (in glaxium 0.5, just downloaded from
 the website) seems to be:
 
   glEnable(GL_TEXTURE_2D);
 
 Line 586, though, makes reference to GL_DOT3_RGB_EXT.  I've made the 
 change there.
 
 Now, with the unpatched driver and the patched game, the results look
 exactly look they did with the patched driver and the unpatched game!

I believe that we may have discovered one of the subtle differences between
the R100 family of chips and the R200 family.  On the R100, when DOT3
bumpmapping is selected the driver has to manually select the implicit 4X
multiplier required by the ARB  EXT specs.  It seems that when the R200
this is not needed.

The unpatched driver programs the chip to do DOT3 and a 4X scale.  It seems
that the result is a 4X + 4X scale, or a 16X scale.  The result is that all
fragments end up saturated to white.  If this theory is correct, then this
patch should fix it.

I'm crossing my fingers...

-- 
Smile!  http://antwrp.gsfc.nasa.gov/apod/ap990315.html



r200_DOT3_EXT.patch.gz
Description: application/gunzip


Re: R200 DOT3 EXT problems (was Re: [Dri-devel] mesa 4.1 branch /NO go on 2.5.48)

2002-11-21 Thread Adam K Kirchhoff

On Thu, 21 Nov 2002, Ian Romanick wrote:

 On Thu, Nov 21, 2002 at 01:27:03PM -0500, Adam K Kirchhoff wrote:

  Hmmm... Well, line 592 of scene.cpp (in glaxium 0.5, just downloaded from
  the website) seems to be:
 
  glEnable(GL_TEXTURE_2D);
 
  Line 586, though, makes reference to GL_DOT3_RGB_EXT.  I've made the
  change there.
 
  Now, with the unpatched driver and the patched game, the results look
  exactly look they did with the patched driver and the unpatched game!

 I believe that we may have discovered one of the subtle differences between
 the R100 family of chips and the R200 family.  On the R100, when DOT3
 bumpmapping is selected the driver has to manually select the implicit 4X
 multiplier required by the ARB  EXT specs.  It seems that when the R200
 this is not needed.

 The unpatched driver programs the chip to do DOT3 and a 4X scale.  It seems
 that the result is a 4X + 4X scale, or a 16X scale.  The result is that all
 fragments end up saturated to white.  If this theory is correct, then this
 patch should fix it.

 I'm crossing my fingers...

Well, it looks better than your last patch, but still not right, and
worse than it does with the unpatched driver.

http://memory.visualtech.com/glaxium.png

Adam




---
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
___
Dri-devel mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/dri-devel



R200 DOT3 EXT problems (was Re: [Dri-devel] mesa 4.1 branch / NO go on 2.5.48)

2002-11-20 Thread Ian Romanick
On Tue, Nov 19, 2002 at 11:06:16PM -0800, Linus Torvalds wrote:
 
 On Wed, 20 Nov 2002, Dieter Nützel wrote:
  Am Mittwoch, 20. November 2002 00:55 schrieb Ian Molton:
   
Linus updated 2.5.48 with Brian's latest DRM r200 stuff so I should do
some testing.
  
  No go so far.
  
  Modules are somewhat broken in 2.5.48.
 
 One approach is to not use modules, just compile the thing in. Works for
 me (damn, I'd like to see how the commercial tuxracer looks with bump 
 mapping. But apparently DRI doesn't support the right extension or 
 something ;)

The problem is that it uses EXT_texture_env_dot3 (which the driver does
advertise), but the driver doesn't actually implement only implements
EXT_texture_env_dot3 or ARB_texture_env_dot3 correctly.  I noted this in an
earlier thread about glaxium, but I haven't posted a fix yet.

I have attached a patch to the DRI CVS trunk that should fix the problem.  I
don't have access to an 8500, so I haven't actually tested it.  I would
really appreciate it if people with 8500 (or 9000) cards could try this
patch and tell me if it works.  If I get a few positive results and no
negatives, I'll commit the patch.

-- 
Smile!  http://antwrp.gsfc.nasa.gov/apod/ap990315.html



r200_DOT3_EXT.patch.gz
Description: application/gunzip