** suivant:** Representation by hardware floats
** monter:** Floating point representation.
** précédent:** Floating point representation.
** Table des matières**
** Index**

The `Digits` variable is used to control how real numbers
are represented and also how they are displayed.
When the specified
number of digits is less or equal to 14 (for example ` Digits:=14`), then hardware floating point
numbers are used and they are displayed using the specified
number of digits.
When `Digits` is larger than 14, Xcas uses the MPFR
library, the representation is similar to hardware floats
(cf. infra) but the number of bits of
the mantissa is not fixed and the range of exponents is much larger.
More precisely, the number of bits of the mantissa of created MPFR float
is `ceil(Digits*log(10)/log(2))`.
Note that if you change the value of `Digits`, this will affect
the creation of new real numbers compiled from commandlines
or programs or by instructions like `approx`, but it will
not affect existing real numbers. Hence hardware floats may coexist
with MPFR floats, and even in MPFR floats, some may have 100 bits
of mantissa and some may have 150 bits of mantissa. If operations
mix different kinds of floats, the most precise kind of floats
are coerced to the less precise kind of floats.

** suivant:** Representation by hardware floats
** monter:** Floating point representation.
** précédent:** Floating point representation.
** Table des matières**
** Index**
giac documentation written by Renée De Graeve and Bernard Parisse