Solving for directional irradiance

There are two options for solving directional irradiance instead of just irradiance alone. For a single directional output texture containing the luminance-based dominant direction of the indirect lighting, add the following to the code sample in the Solving for irradiance section:

irradianceTask.m_DirectionalOutput       = directionalOutput;

irradianceTask.m_DirectionalOutputStride = radCore->m_MetaData.m_OutputWidth;
// lazily assuming stride = width, not always true.

Alternatively, to solve directional irradiance separately for the red, green and blue components of the indirect lighting, add the following to the code sample in the Solving for irradiance section (note how m_DirectionalOutput is re-used for red directional output data):

irradianceTask.m_DirectionalOutput       = directionalOutputR;
irradianceTask.m_DirectionalOutputG      = directionalOutputG;
irradianceTask.m_DirectionalOutputB      = directionalOutputB;

irradianceTask.m_DirectionalOutputStride = radCore->m_MetaData.m_OutputWidth;
// lazily assuming stride = width, not always true.

In both cases, make the same function call (SolveIrradianceTask) as for irradiance alone. The texture format of the directional output data is always 8 bits per channel per pixel. The dominant direction { x, y, z } is stored in the { r, g, b } channels respectively, and the rescaling factor is stored in the alpha channel. The byte order of the output can be controlled with the m_OutputFormatByteOrder parameter of the task.

The length of the directional vector corresponds to how "directional" the incoming light is. If the light is concentrated entirely from a single direction, then the length of the directional vector equals 1. If the length is less than 1, light is coming from multiple directions. Note that if the light is distributed uniformly across the hemisphere, the length of the directional vector is 0.5 and that in some lighting distributions the length can indeed be 0.