Double-precision Pd uses 64-bit floating point as its native number format for floats and signals. Examples of what double precision enables are illustrated on www.katjaas.nl/doubleprecision/doubleprecision.html. Test- and demo-patches are available for download.

Double-precision Pd core + extra's (a double-precision equivalent of Pd-vanilla) can be built with the code from https://github.com/pd-projects/pd-double. It's tested to work fine under OSX and Linux, and should work under Windows as well.

The pd-double sources allow for single and double precision builds of Pd. Based on it, nightly builds are produced which feature a single precision Pd core for i386 builds and double precision for x86_64 and amd64 builds, in order to test them both. Double precision builds try to include Pd-extended libs compiled in double precision as well. Be not surprised if externals are still single precision, or even absent from the build. A lot of code in external libs is not yet double-precision-ready. If 'alpha' were preceded by another character in the alphabet, we'd use it to designate pd-double-extended in this stage. For your safety, turn amp level looooow when testing.

http://autobuild.puredata.info/auto-build/latest/