search

jk-jeon / dragonbox

Reference implementation of Dragonbox in C++
Apache License 2.0
588 stars 37 forks source link

Completely expel the implicit assumption "`float` = binary32 and `double` = binary64" #59

Closed jk-jeon closed 5 months ago

jk-jeon commented 5 months ago

Now the main library does not rely on this assumption, but many testing and meta codes still do. Dependence on the traits also need to be minimized, and ideally as many as possible templates should be parameterized by the format & carrier pair.

jk-jeon commented 5 months ago

This is impossible because other libraries (used for test/benchmark) are relying on that assumption. All the other parts of this project is now free of the assumption, so closing. In the future it might be a good idea to add some remarks on this into the source codes that are relying on it, but currently it's not planned.