Generate n cosine-weighted unit direction vectors over the
upper hemisphere (Z >= 0). Uses the Hammersley low-discrepancy
sequence — better stratification than uniform random for small
n, so a 64-sample hemisphere produces a smoother sky-view
factor estimate than 64 random samples would.
Generate
ncosine-weighted unit direction vectors over the upper hemisphere (Z >= 0). Uses the Hammersley low-discrepancy sequence — better stratification than uniform random for smalln, so a 64-sample hemisphere produces a smoother sky-view factor estimate than 64 random samples would.Returned as a flat
Float32Arrayof lengthn * 3.