linearRgbToXyzQ16()

void fl::linearRgbToXyzQ16	(	const SourceMatrixQ16 &	matrix,
		u16	r,
		u16	g,
		u16	b,
		i32(&)	out_xyz[3] )

One pixel: u16 linear RGB -> s16.16 XYZ.

Inputs are the u16 linear values decodeTransferU16 produces and are read as s16.16 raw, so full scale is 65535/65536 rather than exactly 1. That bias is uniform across all three channels and both the matrix and the white point, so it scales the result by 0.9999847 – a luminance error of 0.0015%, against a 0.5% budget – and it buys an exact shift in place of a division by 65535 in the per-pixel path.

Definition at line 114 of file source_xyz.cpp.hpp.

                                                       {
    out_xyz[0] = dotRowQ16(matrix.m[0], r, g, b);
    out_xyz[1] = dotRowQ16(matrix.m[1], r, g, b);
    out_xyz[2] = dotRowQ16(matrix.m[2], r, g, b);
}

References FL_NO_EXCEPT, g, and fl::SourceMatrixQ16::m.

Referenced by processPixelLinearQ16(), and processPixelWideLinearQ16().

Here is the caller graph for this function: