Some opengl functions (like glColor) require an integer to float conversion for some parameter.
Unsigned integers (GLuint, GLushort, GLubyte) are mapped to [0.0, 1.0] and signed integers (GLint, GLshort, GLbyte) to [-1.0,1.0].
Most of the time the values are encoded in IEEE 754 FP32 in the FIFO. But sometimes the values are compressed to 16 or 8 bits.
16 bits conversions
We have A and B, 2 integers converted to float in [-1.0, 1.0].
If A1 and B1 are signed short and A1=A*USHRT_MAX/2 ; B1=B*USHRT_MAX/2, we have in the FIFO
B1 << 16 | A1
example :
glNormal3i(0 ,INT_MAX,INT_MIN);
2b5 0x00000000 0x00082c40 {size: 0x2 channel: 0x1 cmd: NV15_NORMAL_SHORT_BYTE}
2b6 0x00000000 0x7fff0000 (FP16: 0.000000 | 0.999985)
2b7 0x00000000 0x00008000 (FP16: -1.000015 | 0.000000)
8 bits conversions
We have A, B, C, D 4 unsigned integers converted to float in [0, 1.0].
If A1, B1, C1, D1 are unsigned char and A1=A/UCHAR_MAX; B1=B/UCHAR_MAX, C1=C/UCHAR_MAX, D1=D/UCHAR_MAX we have in the FIFO
D1 << 24 | C1 << 16 | B1 << 8 | A1
example :
glColor4ub(0,UCHAR_MAX, 0, UCHAR_MAX);
2b7 0x00000000 0x00042c6c {size: 0x1 channel: 0x1 cmd: NV15_COLOR}
2b8 0x00000000 0xff00ff00 (FP8: 0.000000 | 1.000000 | 0.000000 | 1.000000)
Note
The conversion can have some errors. For example in 16 bits, MAX_INT = 0x7fff but SCHAR_MAX = 0x7f7f.