tell you want - I'll write a simple test shader that takes a parameter and change it several times a begin/end block and see what it does on my environment.
I guess as long as the variable was not marked as a uniform then it should be valid to change at any time. What version of Cg are you running ? Just a thought - but there is a new Cg Version 2.1 out now.
[OpenGL + CG] Ignored cgSetParameter
Author
Quote:Original post by BionicBytes
tell you want - I'll write a simple test shader that takes a parameter and change it several times a begin/end block and see what it does on my environment.
I guess as long as the variable was not marked as a uniform then it should be valid to change at any time. What version of Cg are you running ? Just a thought - but there is a new Cg Version 2.1 out now.
That would be great! I'm running 2.1.0016 (the November build, which I think is the latest). Thanks for the effort!
Sorry for the lateness in my reply. I've had a bad cold and couldn't face starring at a PC screen for any length of time. Anyway better now.
So - I've set up a simple shader test case.
The test case is: Can I draw two different rectangles in two different colors by changing the value of one shader parameter (t) twice whilst a single shader is bound.
If the test is successful two rect's are draw in different shades of red.
Here is the main render loop where the rects are drawn.
And the result is TWO different colored RED rectangles on the screen.
Now, if I change the shader code and make the #if 1 to a #if 0, and re-run the test the result is 2 BLACK rectangles.
So... I think in this second case the parameter t is being ignored.
But ... in the debugger the named Cg param is still valid - meaning (from what I know about Cg etc) is that the param was found in the list of enumerated Cg parameters. If the param was invalid the param value would have been 0.
I have no explanation for the t param when its specified as part of the main function.
B.T.W if i go back to the Uniform param t version, if i remove the uniform keyword, the test still works as expected and 2 red rects are displayed.
Does this help?
So - I've set up a simple shader test case.
The test case is: Can I draw two different rectangles in two different colors by changing the value of one shader parameter (t) twice whilst a single shader is bound.
If the test is successful two rect's are draw in different shades of red.
// the vp shaderstruct vp_in{ float3 Position : POSITION;};struct vp2fp{ float4 HPos : POSITION;};vp2fp main(vp_in IN){ vp2fp OUT; float4 pos=float4 (IN.Position.xyz,1.0); OUT.HPos=mul(glstate.matrix.mvp,pos); return OUT;} // now the fragment shaderstruct vp2fp{ float4 HPos : POSITION;};// Two interesting variations of the same thing. Here we have declared// a uniform float param, and in the second case param t is a function param.#if 1uniform float t;void main( vp2fp IN,out float4 color : COLOR){ color = float4 (t,0,0,1);}#elsevoid main( vp2fp IN,in float t,out float4 color : COLOR){ color = float4 (t,0,0,1);}#endifHere is the main render loop where the rects are drawn.
// my lib code to get a shader referenceShader aShader = aMaterial.GetBaseShader();// Get a named Cg parameterSystem.IntPtr param = Cg.cgGetNamedParameter(aShader.GetFragmentProgram(),"t"); aMaterial.Bind(); // my lib code to grab and activate the above shader Cg.cgSetParameter1f(param,0.8f);Gl.glRectf (10,10,100,100);Cg.cgSetParameter1f(param,0.4f);Gl.glRectf (10,110,100,210); aMaterial.Unbind();Gl.glColor3f (1,1,1);And the result is TWO different colored RED rectangles on the screen.
Now, if I change the shader code and make the #if 1 to a #if 0, and re-run the test the result is 2 BLACK rectangles.
So... I think in this second case the parameter t is being ignored.
But ... in the debugger the named Cg param is still valid - meaning (from what I know about Cg etc) is that the param was found in the list of enumerated Cg parameters. If the param was invalid the param value would have been 0.
I have no explanation for the t param when its specified as part of the main function.
B.T.W if i go back to the Uniform param t version, if i remove the uniform keyword, the test still works as expected and 2 red rects are displayed.
Does this help?
One other thing -- I compile my code using:
// vertex
cgc.exe -profile vp40
//frag
cgc.exe -profile fp40
I am running cg 2.0.0015 on a windows XP environment using C# and Tao framework.
I have however, also, run similar type of code in a Windows with c++ environment.
// vertex
cgc.exe -profile vp40
//frag
cgc.exe -profile fp40
I am running cg 2.0.0015 on a windows XP environment using C# and Tao framework.
I have however, also, run similar type of code in a Windows with c++ environment.
Author
Quote:Original post by BionicBytes
Sorry for the lateness in my reply. I've had a bad cold and couldn't face starring at a PC screen for any length of time. Anyway better now.
So - I've set up a simple shader test case.
The test case is: Can I draw two different rectangles in two different colors by changing the value of one shader parameter (t) twice whilst a single shader is bound.
If the test is successful two rect's are draw in different shades of red.
*** Source Snippet Removed ***
Here is the main render loop where the rects are drawn.
*** Source Snippet Removed ***
And the result is TWO different colored RED rectangles on the screen.
Now, if I change the shader code and make the #if 1 to a #if 0, and re-run the test the result is 2 BLACK rectangles.
So... I think in this second case the parameter t is being ignored.
But ... in the debugger the named Cg param is still valid - meaning (from what I know about Cg etc) is that the param was found in the list of enumerated Cg parameters. If the param was invalid the param value would have been 0.
I have no explanation for the t param when its specified as part of the main function.
B.T.W if i go back to the Uniform param t version, if i remove the uniform keyword, the test still works as expected and 2 red rects are displayed.
Does this help?
Thanks for making the effort!
As for your first test (using 1), I get the same behaviour. However, this is because the parameter set calls are made after each rectangle is rendered. That is, I assume the glRectf methods use something similar to glBegin() and glEnd() calls. What I'm interested in is setting the parameters inside this glBegin() - glEnd() block, instead of after each block. So far, I have not been able to get this working; a I explained in my previous posts, only my last parameter set inside the glBegin() - glEnd() block is used.
Secondly, you seem to have the same problems as I do when you change your 1 to 0. I pass my parameters in the same way (using main), and get the same result as you: although the cg parameter is still valid, it appears to be set to 0, as evidenced by the two black rectangles.
So, still two questions remain:
- Is it possible to pass different values to the same cg parameter in a glBegin() - glEnd() block, and if so, how?
- How is it possible to pass values to cg parameters in a shader main function?
okay - i need to be quick because the international rugby is on.
So ... within a BEGIN/END block any call to Cg.cgSetParameter1f(param,0.8f); is invalid. Resulting in two black quads.
Placing the Cg.cgSetParameter1f(param,0.8f); outside of a BEGIN/END block is valid and results in RED rects. (of course both the same color).
So ... rule is must set Cg params outside of BEGIN/END blocks.
And for the second point - never set additional cg params as a parameter to main.
So ... within a BEGIN/END block any call to Cg.cgSetParameter1f(param,0.8f); is invalid. Resulting in two black quads.
Placing the Cg.cgSetParameter1f(param,0.8f); outside of a BEGIN/END block is valid and results in RED rects. (of course both the same color).
So ... rule is must set Cg params outside of BEGIN/END blocks.
And for the second point - never set additional cg params as a parameter to main.
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement
