glReadPixels type

[David Doria]

I would expect the following to give me same result

float pixelf[3];
glReadPixels(x, y, 1, 1, GL_RGB, GL_FLOAT, pixelf);

GLuint pixelui[3];
glReadPixels(x, y, 1, 1, GL_RGB, GL_UNSIGNED_INT, pixelui);

unsigned char pixelub[3];
glReadPixels(x, y, 1, 1, GL_RGB, GL_UNSIGNED_BYTE, pixelub);

cout << "r: " << pixelf[0] << " g: " << pixelf[1] << " b: " << pixelf[2] << endl;

cout << "r: " << pixelui[0] << " g: " << pixelui[1] << " b: " << pixelui[2] << endl;

cout << "r: " << pixelub[0] << " g: " << pixelub[1] << " b: " << pixelub[2] << endl;

however, my result is this:

r: 0 g: 1 b: 0
r: 0 g: 4294967295 b: 0
r: g: � b:

the pixel is green, so the float result is correct (0,1,0). Why are the other results so crazy!?

Thanks,
Dave

[trinitrotoluene]

They are not crazy. Floating point number are normalized in the range [0 1]. For unsigned int [0 2^32-1] => 2^32-1 = 4294967295. Finnaly, for unsigned byte [0 255], I suspect that the glyph � is equivalent as 2^8-1 = 255.

Edit: After testing, the character &#65533 = 255.

[ZbuffeR]

Colors are typically normalized: 0-1 for floats and doubles.
0 to max integer value for integer formats.
Read the OpenGL specification for details.

So results are perfectly logical :
1.0f is the largest normalized value for floats
4294967295 is the highest number for 32 bit unsigned integer.
For unsigned bytes, you a seeing characters according to ASCII char tables instead of numbers ;)
Try the good old :
printf("r: %u g: %u b: %u\n", pixelub[0], pixelub[1], pixelub[2]);

you should see 255.

[David Doria]

ah, got it. I thought cout would have converted the number to something readable. Also, is a char the same as a byte? ie I declared an unsigned char array and told opengl to give me back unsigned bytes - is that ok?

Thanks for the help!

Dave

[trinitrotoluene]

In C/C++ for a x86 cpu, the size for a char is 8 bits => a byte. At the start of the gl.h header I have this:


typedef unsigned int GLenum;
typedef unsigned char GLboolean;
typedef unsigned int GLbitfield;
typedef signed char GLbyte;
typedef short GLshort;
typedef int GLint;
typedef int GLsizei;
typedef unsigned char GLubyte;
typedef unsigned short GLushort;
typedef unsigned int GLuint;
typedef float GLfloat;
typedef float GLclampf;
typedef double GLdouble;
typedef double GLclampd;
typedef void GLvoid;


When you pass or get data to OpenGL functions, its better to use the predefined typedef (GLubyte, GLint, etc). Maybe a different processor architecture will have a different typedef set for those.So this can ease to port application.


Also, is a char the same as a byte? ie I declared an unsigned char array and told opengl to give me back unsigned bytes - is that ok?

So to answer your question, for a x86 cpu is yes, for another architecture maybe not.

[_NK47]

float pixelf[3];
glReadPixels(x, y, 1, 1, GL_RGB, GL_FLOAT, pixelf);

GLuint pixelui[3];
glReadPixels(x, y, 1, 1, GL_RGB, GL_UNSIGNED_INT, pixelui);

unsigned char pixelub[3];
glReadPixels(x, y, 1, 1, GL_RGB, GL_UNSIGNED_BYTE, pixelub);

Thats why you should use GL types 24/7. OpenGL typedefed them not for fun.
float - GLfloat
GLuint - ok
unsigned char - GLubyte

[David Doria]

got it - thanks everyone!