Help needed with Math of setting gl_TextureMatrix
Author
I would like to setup 4 different texture matricies to various texture coordinates. e.g.
1st Matrix I would like to have 0,0; .5,0; .5,.5; 0,.5; as my texture coordinates so I can access the lower left of my texture unit...
Thanks for any help BTW I am using GLSL...
As you've posted here I'll reply here instead of via email [smile]
OK, as I mentioned in the email, the theory is pretty much the same as when you make a remap matrix for shadow mapping.
In the case of shadow mapping you want to divide by 2 (scale by .5) and then add 0.5 (translate by .5).
For for the first texture coords to access the bottom left of the texture, you supply a matrix which simply scales by 0.5.
To get to the other corners the same priniple applies, scale by 0.5 and then translate it on the correct axis so that its in the right range for that corner.
You should be able to work the numbers out for yourself as its not really that hard [smile]
OK, as I mentioned in the email, the theory is pretty much the same as when you make a remap matrix for shadow mapping.
In the case of shadow mapping you want to divide by 2 (scale by .5) and then add 0.5 (translate by .5).
For for the first texture coords to access the bottom left of the texture, you supply a matrix which simply scales by 0.5.
To get to the other corners the same priniple applies, scale by 0.5 and then translate it on the correct axis so that its in the right range for that corner.
You should be able to work the numbers out for yourself as its not really that hard [smile]
Author
Quote:Original post by phantom
As you've posted here I'll reply here instead of via email [smile]
OK, as I mentioned in the email, the theory is pretty much the same as when you make a remap matrix for shadow mapping.
In the case of shadow mapping you want to divide by 2 (scale by .5) and then add 0.5 (translate by .5).
For for the first texture coords to access the bottom left of the texture, you supply a matrix which simply scales by 0.5.
To get to the other corners the same priniple applies, scale by 0.5 and then translate it on the correct axis so that its in the right range for that corner.
You should be able to work the numbers out for yourself as its not really that hard [smile]
What I was thinking on doing is having 4 different Texture matricies so I can just upload the matrix as I want it on the CPU side and just call one of the four matricies I need in GLSL... Or are you saying just modify the texture matrix in the vertex program everytime?
Well,it really depends on how flexible you want it to be. Are you going to need to change the values on the fly all that much? or will you always want to pull the same 4 values from the same for corners of the texture? If so, you might as well just put the maths inline in the shader and be done with it.
Ofcourse, if you want to change the range and/or part of the texture you want to pull the image data from a matrix via a uniform would be the best solution.
// assuming coord is a vec2 with your texture coords invec4 color1 = texture2D(tex,coord*vec2(0.5,0.5)); // bottom leftvec4 color2 = texture2D(tex,coord*vec2(0.5,0.5)+vec2(0.5,0.0)) // top left// etcOfcourse, if you want to change the range and/or part of the texture you want to pull the image data from a matrix via a uniform would be the best solution.
Author
Quote:Original post by phantom
Well,it really depends on how flexible you want it to be. Are you going to need to change the values on the fly all that much? or will you always want to pull the same 4 values from the same for corners of the texture? If so, you might as well just put the maths inline in the shader and be done with it.// assuming coord is a vec2 with your texture coords invec4 color1 = texture2D(tex,coord*vec2(0.5,0.5)); // bottom leftvec4 color2 = texture2D(tex,coord*vec2(0.5,0.5)+vec2(0.5,0.0)) // top left// etc
Ofcourse, if you want to change the range and/or part of the texture you want to pull the image data from a matrix via a uniform would be the best solution.
Ok here is what I got and it works. I am looking to access 4 corners right now, but might move to 16.
mat4 upperRightTexMatrix = mat4(vec4(.5, .0, .0, .0), vec4(.0, .5, .0, .0), vec4(.0, .0, .5, .0), vec4(.5, .5, .5, .5)); mat4 upperLeftTexMatrix = mat4(vec4(.5, .0, .0, .0), vec4(.0, .5, .0, .0), vec4(.0, .0, .5, .0), vec4(.0, .5, .0, .0)); mat4 lowerLeftTexMatrix = mat4(vec4(.5, .0, .0, .0), vec4(.0, .5, .0, .0), vec4(.0, .0, .5, .0), vec4(.0, .0, .0, .0)); mat4 lowerRightTexMatrix = mat4(vec4(.5, .0, .0, .0), vec4(.0, .5, .0, .0), vec4(.0, .0, .5, .0), vec4(.5, .0, .0, .5)); I don't plan on changing the matricies, but wouldn't it seem faster to compute this on the CPU side and store them in gl_TextureMatrix[1] though gl_TextureMatrix[4]? that way all I have to do in the shader is multiply the matrix by the texture coordinates? :) I am looking from a stand point of which way is faster. I can see if I was to change them alot, putting a if/else in the vertex program and uploading some type to determine the output, but as of now I don't need that. Now here is the kicker how do I tile it?? I tried to multiply it by 16 or insert amount here and that blows up the code...
If you dont change the matrix then the faster method is just to inline the maths in the code (like I did) and leave it at that.
Although over all thats the method which requires more work to change, but as you said you dont plan on changing it then I dont see the point in going the matrix route as you are doing more maths than you really need todo.
Although over all thats the method which requires more work to change, but as you said you dont plan on changing it then I dont see the point in going the matrix route as you are doing more maths than you really need todo.
Author
Hey Phantom, I re-read your post and it says if I want the range out of 0-1 I might need to use matricies? I am hoping to not have to do that now that I have my code setup to do the 0-1 range... I have tried the math and can't seem to figure out how to come up with any good solutions. I am trying to tile the texture 16x. Thanks
errm... i've no idea where I said that, mainly because I'm sure I didnt :)
when you say 'tile the texture 16 times' do you mean 16 time across the whole object?
Atm I cant think of a way to allow you todo what you want...
when you say 'tile the texture 16 times' do you mean 16 time across the whole object?
Atm I cant think of a way to allow you todo what you want...
Author
Quote:Original post by phantom
errm... i've no idea where I said that, mainly because I'm sure I didnt :)
when you say 'tile the texture 16 times' do you mean 16 time across the whole object?
Atm I cant think of a way to allow you todo what you want...
Yeah, I would like to tile the texture I grab out of the larger one using the texture coordinates and tile that one 16x...
I can't either but it seems so simple that it should be doable. I guess I can't understand why I am not allowed to or limited to not being able to do this???
the problem is that you are trying to map an arbatry range into a smaller repeating range.
So, some how 0->1 needs to be worked out for each block of floats between each interger (0->1, 1->2, 2->3 etc all need to be initally mapped into 0->1).. there probably is a way, but atm I'll be buggered if I can see it
So, some how 0->1 needs to be worked out for each block of floats between each interger (0->1, 1->2, 2->3 etc all need to be initally mapped into 0->1).. there probably is a way, but atm I'll be buggered if I can see it
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement
