For the record: purr data has completed a series of changes to make the core work in single or double precision. Changes were largely based on the pd-double project, but done systematically per topic, improved where needed, and following pd's coding style. The work was done by Pranay Gupta and supervised by Jonathan Wilkes (by my own choice I was not involved). The series of merged pull requests provides an overview of precision-bottlenecks, some of which aren't addressed yet in vanilla Pd.
https://git.purrdata.net/jwilkes/purr-data/merge_requests?author_username=pr...
Katja