| |
| |
| |
|
Page: 1 2
Comments:
<0> androvosky ?
<0> do you gotta pay ?
<1> havent done much with shaders in a long time, but how do you look up specific texels in a pixel shader? since the lookup functions take normalized coords, do i have to load 1/width and 1/height in constants and use those or is there some easier way?
<0> last I seen.... nvidia is still closed source
<2> Some people where superman pajamas. Superman wears chuck norris pajamas.
<3> anyone here familiar with the primitive_restart extension nvidia provides?
<4> I just got to know about it a few days ago :)
<5> Never heard of it.
<4> Not widely supported I'm afraid.
<4> The better option is to make primitive restarts uneccesary, by resorting/recomputing the triangle list.
<5> what *is* a primitive restart?
<4> Stevie-O: What points which defines a triangle, based on the primitive type you are rendering.
<4> Imagine drawing a bunch of triangle strips.
<5> (as in /\/\/\/\ ?)
<4> Now imagine you have two arrays of triangle strips and want to concaternate those two together.
<4> That won't work.
<5> where /\/\ is three triangles
<5> two up, one down
<4> Since the points depends on the previous one.
<4> In order to make that work, you have to "restart" the algorithm.
<4> The extension lets you do that inside the drawing call itself.
<5> Right. You'd have to have two respecified vertices, right?
<4> Yes. Simply concaternate two triangle strip arrays would mess things up. Badly.
<4> Best case scenario would be one invalid triangle.
<5> My most sophisticated creation so far has three quads.
<4> I came acros this issue when I made a MD2 loader recently.
<2> try changing it to three tristrips, quads are bad mojo
<4> Where the model was stored as a bunch of tristrips and triangle fans.
<4> And I mean ALOT of them.
<5> Quads are bad mojo?
<2> heh, you're special, madsy :} *duck*
<4> ``Erik: Heh :-)
<4> In what way?
<2> on modern consumer grade hw, a quad is handled like this... take 3 points, get the plane, verify that the fourth point is on the plane, then convert it to either a tristrip or trifan, THEN send it to the card
<5> (I *have* noticed that e.g. RTCW renders *everything* as triangles, even flat rectangular surfaces. God bless that console and r_* commands)
<4> That I use a model format which was rendered obsolete four years ago? :-D
<2> cards only handle triangles, dude
<5> I reiterate, I'm new at this :)
<2> no, madsy, that you got confused on how to deal with tristrips and trifans :) it's not a "problem", it's how they work
<2> stevie: that's why i'm telling you :)
<5> hehe, thanks
<4> ``Erik: Well, sure. It was a problem in my particular context though.
<5> Would you be so kind as to tell me how to mix texturing into the triangle striping?
<4> Quake 2 rendered all those strips with several batches. I just wanted two.
<4> One for fans, and one for strips.
<5> since if I have a triangle strip I only specify each vertex once
<4> I eventually just sorted everything as triangles and never looked back.
<2> stevie: using what, va's, display lists, vbo's, or immediate mode?
<5> I'm just working with immediate mode for now.
<2> glTextureCoord2d(); glVertex3d(); glTextureCoord2d(); glVertex3d(); ...
<5> right now I'm just trying to figure out why one of my surfaces needs to be twice as wide as I think it should be, in order for it to look "right".
<2> opengl is a state machine
<5> hah! I startede out by just changing GL_QUADS to GL_TRIANGLE_STRIP. The resulting effect is... quite interesting.
<2> you need to do three glBegin()/glEnd() pairs to do three rectangles
<5> wait, I do?
<2> ja
<5> does it matter that the three quads I have are connected in a chain?
<2> oh, fi they're a chain, then you're all good
<2> do you have the red book?
<5> the OpenGL Programming Guide? I went through that last night.
<2> in chapter two or three, there's a figure showing the draw order of the primitives
<2> that might be useful to you at this point :)
<5> The object I'm rendering is basically in a 'U' shape. I'm plastering a scanned-in image of a DVD case insert onto it
<5> hrm, I remember the picture you're talking about.
<5> ah, figure 2-7
<5> I have to reverse the inner two vertex definitions so it goes zig-zag like an 'N' instead of going around the perimeter of the quad
<5> No, that didn't fix it. wtf
<5> aha!
<5> I chose the wrong vertices to swap
<5> hey Erik, you still here? Does this look right for the vertex placing order? http://www.klozoff.ms11.net/gltest/lines.png
<5> hrm, no, that doesn't poly right
<2> looks right to me, um
<2> are you backface culling?
<5> I am not.
<5> I was at first, because I couldn't get it to occlude properly. Then I learned about GL_DEPTH_TEST :)
<2> you don't have lighting turned on, do you?
<5> nope.
<5> That gets into surface normals and all that crap
<5> okay, this is strange. My first two triangles (the first quad) seem to draw correctly. But then things start to fall apart.
<5> http://www.klozoff.ms11.net/gltest/textured-back.png <- this is the back, looking right
<5> http://www.klozoff.ms11.net/gltest/textured-front.png <- this is the front. it is quite apparent that something's wrong
<2> that almost looks like backface culling to me :/
<2> mebbe this is something that if you put it in paste.lisp.org or rafb.net/paste for people to look at, someone can spot omething
<5> nope. But if I turn *on* backface culling, I learn that the rendered triangle on the spine is facing into the case instead of out
<5> http://rafb.net/paste/results/gLujtL57.html
<6> i spy nehe code
<5> anything labeled NFI means I had NFI what was going on and worked straight from whatever tutorial or guide I was reading at the time.
<5> So it's quite likely something in there came from nehe :)
<5> ah, yes, the fog code.
<5> ZeZu: anything stand out as obviously broken?
<6> dunno, trying to debug my own problems, the simplest vertex shader wont work in my renderer, think its not liking orthographic projection but i dont why it would matter as long as the matrices are handled correctly
<5> over my head already :D
<5> oh, I see the problem
<6> me too, you can to use the runtime lib to p*** it the matrix
<6> have to *
<5> you mean... the runtime lib is not psychic?
<5> My problem is that when I converted from separate quads to a triangle strip, instead of deleting the redundant vertices, I deleted some important ones.
<6> your about as bright as a penny in the sewer, obviously i'm speaking about the vertex program not being psychic
<5> Oh. It would appear, then, that I am also not psychic.
<5> ;)
<7> have any of you herd of a 3d-operating system before? does one exist? I heard someone took the source to doom and turned a bot into a kill process command.
<3> 3d-operating system? that makes no sense
<7> it doesn't though does it?
<7> it's like trying to teach an ant how to fix a pc.
<6> JohnnyL, have you been taking any lsd or other interesting drugs?
<7> ZeZu, why should I start?
<6> should i parse that as, "why should I start?" or "why, should I start?" ?
<7> ZeZu, Syntax Error, goto 10.
<6> i dont contain line numbers
<5> I've heard of that program.
<7> What Stevie the Kill Process/Doom program?
<5> yes, that one. And would it be too much trouble to use a bit of punctuation? Maybe a comma here and there?
<8> is this channel geared towards opengl programming?
<5> That's one theory.
<5> Another theory is that this is about how well Gary Lee golfed at the last U.S. Open
<7> i read about that back in like 1999!!!! (and partied as such as well!) :)
<9> _Olympus - yes
<7> what ever happened to lightflowtech.com? anyone know?
<10> yo, who's around?
<11> gentry
<6> ok, how do i p*** multiple sets of colors ?
<9> ZeZu - p*** them as tex coords
<9> (maybe?)
<6> hmm
<6> there is a glMultiTexCoord or something that is used for that '?
<6> i have color, offset color, fog color, and two sets of tex coords i need to input to shader ;)
<9> fog color is per-vertex?
<6> some of it
<6> i have two diff. fog types
<6> table based and something else i'd have to look
<9> hm
<9> i would probably p*** stuff with glTexCoordPointer
<6> well i do use vertex arrays
<12> Just p*** vertex attributes, it doesn't have to be bound to standard attributes..
<6> but can i just p*** to to 8 to GL_COLOR ?
<6> its gotta figure out which set i want somehow ;)
<9> so you p*** in your texcoords normally
<9> you p*** in color normally
<9> then you p*** offsetcolor and fogcolor in two more texcoord stages
<9> in the shader you know your layout
<6> i dont get the stages part, but then again i've never done any multitexturing, so do i just p*** a larger # to glTexCoordPointer to acheive that or what ?
<9> glActiveClientTexture( GL_TEXTURE0 ); glTexCoordPointer( ... real_texcoords_0 );
<9> glActiveClientTexture( GL_TEXTURE1 ); glTexCoordPointer( ... real_texcoords_1 );
<9> glActiveClientTexture( GL_TEXTURE2 ); glTexCoordPointer( ... offset_colors );
<9> glActiveClientTexture( GL_TEXTURE3 ); glTexCoordPointer( ... fog_colors );
Return to
#opengl
or
Go to some related
logs:
tcfd11
qmail-showctl plesk
'fuck redchairsoftware'
#nhl
#linux-noob
#politics
the language of the pharoes
#windows
#politics
#computers
|
|
