Parameter to control precision in floating point range analysis

Hi all,

I am currently working on getting the high level design of my project on implementing floating point range analysis in LLVM.
I wanted to know if it would be interesting to have a parameter to control the precision of the ranges that would be reported by the pass.
I am not yet sure how the precision would affect the performance of the pass, but your comments will help me decide its inclusion in my project.

Thanks and regards.
Aditya Kamath