largest, smallest, and closest to 1.0...

[Dr. Ing. Carlos López]

Thank you Pedro and Linas for your quick replies. 
I have read through the page pointed by Pedro, and I would like to ask a
few clarifications.
How could I know the "limb" used? It is mandatory that it is related to hardware?
(i.e. likely to be either 32 or 64)
Are there any formula available, or could it be derived easily describing
a "tolerance"? Maybe some loosely related with the machine epsilon... or
the distance between 1.0 and the next floating point number.
I am working with GMP+MPFR under Matlab for Windows; in such environment
the abovementioned constants are used to set default tolerances for iterations,
convergence criteria, etc. so I need to provide functions to do so.
Any help is appreciated
Regards
Carlos