It is also evident from the left panel that most of the time is
spent on simulating stars in the field. In principle, one can split the
full star catalog into smaller files and work on them in parallel. It
may even be possible to use a GPU card or a Xeon-phi co-processor
card to accelerate these calculations by running through several
stars in parallel. However, we have not implemented it, as it is
already running in a rather optimal configuration. As we move to
smaller pixel sizes, time taken to simulate diffuse UV and zodiacal
light increases sharply, since a reduction in pixel size by a factor of
2 results in the increase of the total number of pixels by a factor
of 4 within the field. This does not happen for stars however, since
they are point sources (one pixel per source) and the total number
of pixels occupied by stars is the same regardless of pixel size.