Re: [Mesa3d-users] GL_POINT_DISTANCE_ATTENUATION

[Brian P. <br...@vm...>]

OK, the problem is we're not doing sub-pixel adjustment of texcoords 
in the sprite rasterization code.  Look in sprite_point() in 
s_points.c if interested.  This can be fixed but it will take some 
work.  I'll see if I can fix it when I get some spare time.

-Brian


Owen Kaluza wrote:
> Hi Brian,
> 
> Sure, here is the best illustration of the issue I could produce: 1000 
> points, aligned as you suggested.
> If I bring the point size up from 1.0 the issue isn't as obvious 
> although still noticeable.
> The attenuation seems to drop the point size then gradually increase, 
> you can see the points towards the back disappear then reappear.
> Attached modified program and two screen shots, one using osmesa and the 
> other with glut+video card gl drivers.
> 
> Thanks,
> Owen.
> 
> Brian Paul wrote:
>> Owen Kaluza wrote:
>>> Hello,
>>> I'm having trouble with point distance attenuation using OSMesa.
>>> I'm rendering a lot of depth sorted, alpha blended, textured points and
>>> dark bands are appearing that are not there when I render with the
>>> system GL.
>>>
>>> I found the problem only occurs with point distance attenuation 
>>> turned on.
>>> If you look at the attached image you can see there are clearly defined
>>> bands, possibly the point size calculation is incorrect at certain
>>> distances resulting in size jumps.
>>>
>>> I've attached a sample program that reproduces the problem. Also tried
>>> the latest Mesalib code (Mesa-7.7-devel-20091105) and is still 
>>> occurring.
>> Could you prune down your test program a bit?  Perhaps you could draw 
>> a series of points between the min/max Z positions and see how they look.
>>
>> -Brian
>>
>>
>> ------------------------------------------------------------------------
>>
>>
>> ------------------------------------------------------------------------
>>