🚀 go-pugleaf

One thing which you can do is hand the resulting LISP to one of
several code analyzers and find the dependency chains and thus compute
each of the chains concurrently. Some analysis which I did 1988
indicates that these chains are typically short and there are a lot
per page (for typical output generated by formatters such as SCRIBE)
thus there is a lot of potential for parallelism. A side benefit is
that you know which characters in which font and what size to cache
for subsequent rendering of a page - which is great if you are going
to reprint this page later; ideally it allows you to keep multiple
representations of a page and emit previous versions very quickly for
a given device.

Chip