Is it correct to use p_view.z as depth?

[Jianjin Xu]

In forward.cu line 250, you calculate the depth as:

	depths[idx] = p_view.z;

And p_view is obtained in line 190

in_frustum(idx, orig_points, viewmatrix, projmatrix, prefiltered, p_view)

where p_view = transformPoint4x3(p_orig, viewmatrix);

This means that p_view is the point transformed such that the camera is in canonical position. but not by the perspective matrix, projmatrix.
I think this will create a misalignment in the rendered data, because XYZ are transformed by perspective matrix.

Do I misunderstand your code?