Archives Discussions

[lerimini]

Our application uses glTexSubImage3d function to transfer 3d data from cpu to gpu. Starting from Catalyst 12.6  data transfer doesn't work correctly with the following dataset dimensions: 192x172x161 (it worked fine on earlier versions but not on the last one -12.8). Opengl returns no error but data are completely wrong. The problem isn't present with other daset dimensions.

Do you have open issues about that? Thank you for your help.

[lerimini]

Add some information and a code snippet to reproduce te problem.

ATI E2400 with driver version 8.960; OS: Windows XP Embedded.

Code:

	//Clear all pixels
	glClear(GL_COLOR_BUFFER_BIT);

	//good (128x128x128 volume dataset works fine)
	//int Width = 128;
	//int Height = 128;
	//int Depth = 128;

	////bad
	int Width = 192;
	int Height = 172;
	int Depth = 161;

	GLuint my3dTexture = 0;
	unsigned char* my3DBuffer = NULL;

	//Generate 3d texture and allocate video memory and texture properties
	glGenTextures(1, &my3dTexture);
	glBindTexture(GL_TEXTURE_3D, my3dTexture);

	glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
	glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
	glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_WRAP_R, GL_CLAMP_TO_EDGE);
	glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
	glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
	glPixelStorei(GL_PACK_ALIGNMENT, 1);

	//Allocate memory client size
	my3DBuffer = new unsigned char[Width * Height * Depth];

	//Grayscale gradient along z-direction
	for(int z=0; z<Depth; z++)
	{
		float fValue = (float)z / (float)Depth;
		unsigned char value = (unsigned char) (fValue * 255.0f);
		for(int x=0; x<Width; x++)
			for(int y=0; y<Height; y++)
			{
				my3DBuffer[(z*Width*Height + y*Width + x)] = value;
			}
	}

//Transferring to gpu

	glTexImage3D(GL_TEXTURE_3D, 0, GL_R8, Width, Height, Depth, 0, GL_RED, GL_UNSIGNED_BYTE, NULL);
	glTexSubImage3D(GL_TEXTURE_3D, 0, 0, 0, 0, Width, Height, Depth, GL_RED, GL_UNSIGNED_BYTE, my3DBuffer);

	//Draw a quad using texure mapping (diagonal)
	// (0.25, 0.25, 0.0) and (0.75, 0.75, 0.0)
	glEnable(GL_TEXTURE_3D);
	glColor3f(1.0,1.0,1.0);

	glBegin(GL_QUADS);
		glTexCoord3f(0.0,0.0,0.0);
		glVertex3f(0.25, 0.25, 0.0);
		glTexCoord3f(1.0,0.0,0.0);
		glVertex3f(0.75, 0.25, 0.0);
		glTexCoord3f(1.0,1.0,1.0);
		glVertex3f(0.75, 0.75, 0.0);
		glTexCoord3f(0.0,1.0,1.0);
		glVertex3f(0.25, 0.75, 0.0);
	glEnd();
	glDisable(GL_TEXTURE_3D);
	glBindTexture(GL_TEXTURE_3D, 0);

	// Don't wait start processing buffered OpenGL routines)
	glFlush();

	if(my3DBuffer!=NULL)
	{
		delete[] my3DBuffer;
		my3DBuffer = NULL;
	}

	if(my3dTexture!=0)
	{
		glDeleteTextures(1, &my3dTexture);
		my3dTexture = 0;
	}

Output:

Changing format to GL_LUMINANCE works fine, but the nicest thing is that it works fine also if i add the instruction

glPixelTransferf(GL_RED_BIAS, 0.0000001f);

just before executing glTexImage3D.

Also the combination GL_RED + GL_FLOAT or GL_RED + GL_BYTE works fine. It seems that is not working only the one that I need.