Page 3 of 3
Posted: Mon Jan 09, 2006 15:03
by entryway
There is no sense to make it optional on modern NVidia cards because frame rate does not depend on speed of rendering. Or turn on it by default for ATI cards.
Posted: Mon Jan 09, 2006 16:08
by entryway
The easy way to get rid of the dots is to add (or subtract) 0.001f to the vertex coordinates.
Code: Select all
deltax1 = (wall->glseg->x2 - wall->glseg->x1)*0.001f;
deltaz1 = (wall->glseg->z2 - wall->glseg->z1)*0.001f;
glVertex3f(wall->glseg->x1 - deltax1,wall->ytop,wall->glseg->z1 - deltaz1);
glVertex3f(wall->glseg->x1 - deltax1,wall->ybottom,wall->glseg->z1 - deltaz1);
glVertex3f(wall->glseg->x2,wall->ytop,wall->glseg->z2);
glVertex3f(wall->glseg->x2,wall->ybottom,wall->glseg->z2);
I have tried it in the prboom and it works very fine!
Posted: Mon Jan 09, 2006 16:55
by Graf Zahl
I tried that, too, and didn't like the side effects. I'll fix this for real.
Posted: Mon Jan 09, 2006 17:17
by entryway
Graf Zahl wrote:I tried that, too, and didn't like the side effects. I'll fix this for real.
Wie kann ich diese Nebeneffekte in GLBoom sehen? Ich denke, dass 0.001 zu wenig ist, damit es den sichtbaren Effekt in realer Geometrie gibt. Davon werden nur innere Verrechnungen in dem Videoadapter beeinflusst.
Posted: Mon Jan 09, 2006 17:33
by Graf Zahl
Is German easier to read for you?
Anyway, to keep this topic readable for anyone else:
When I tried this some time ago it seemed to work at first but the problem is that the polygons are no longer seamless. The end result on my old Geforce 2 was that I ended up with slight overlaps at the boundaries that were visible on close inspection. And in the far distance the added margin was too small to have any effect so I still got white spots. I never had the chance to properly investigate because at that time I upgraded to a GF3 which was far less prone to these errors and when I got my GF 5900XT they were gone completely.
Posted: Tue Jan 10, 2006 11:04
by entryway
Graf Zahl wrote:Is German easier to read for you?
Ich habe gehofft, dass du mir auf meiner Muttersprache antworten würdest, wenn ich es auf deiner mache
I'll wait for the new version.
Posted: Fri Jan 13, 2006 17:43
by entryway
It works correctly in version 0.9.23
Posted: Fri Jan 13, 2006 17:56
by Graf Zahl
How often do you check my download folder? I have uploaded it just half an hour ago and haven't even had time to update the links!
Posted: Fri Jan 13, 2006 19:15
by entryway
Still buggy with acs_demos.zip\map06
Does not differ from the previous version
Posted: Fri Jan 13, 2006 20:24
by Graf Zahl
Please don't use mirrors to test this. Mirrors suffer from a slight precision problem. From certain angles there can be a small gap between both sides of the mirror that can create these white pixels as well. This is extremely hard to fix because there actually is nothing to render in these gaps.
Posted: Fri Jan 13, 2006 21:43
by lemonzest
gotta say it works perfect here
i did read somewhere about ati cards running at fixed 24bit precision, and nvidia cards running 32bit but then scaling it to 16bit, no idea how much truth is in that tho.
Posted: Fri Jan 13, 2006 21:51
by Graf Zahl
The mirror problem has been mostly solved as well. But I doubt that I will manage to get rid of that completely. Especially with multiple recursions the accumulated errors can become too large to compensate.
Posted: Fri Jan 13, 2006 22:54
by Deathlike2
The precision you are talking about is referring to the shader precision. I doubt that has anything to do with the dots being rendered on ATI cards. I don't think anything is using shaders currently in GZDoom.
Posted: Fri Jan 13, 2006 23:35
by Graf Zahl
I am using fragment programs to get the invulnerability colormaps to work with camera textures but that's it. But somewhere in the geometry pipeline ATI must be considerably less precise than NVidia. I don't get any of the related artifacts on my Geforce.
Posted: Tue Jul 25, 2006 21:03
by entryway
You will laugh :), but white dots still are present in GZDoom ... on slopes
