Description:

• If used with an integer parameter n, round(x,n,mode) rounds x to a floating-point number with precision n, according to rounding-mode mode.
• If used with a format parameter format, round(x,format,mode) rounds x to a floating-point number in the floating-point format format, according to rounding-mode mode.
• Subnormal numbers are handled for the case when format is one of halfprecision, single, double, doubleextended, doubledouble, quad or tripledouble. Otherwise, when format is an integer, round does not take any exponent range into consideration, i.e. typically uses the full exponent range of the underlying MPFR library.
• It is worth mentionning that the result of round does not depend on the current global precision of Sollya, unless a warning is displayed. As a matter of fact, round rounds the given constant or constant expression x applying all rules of IEEE 754 correct rounding, unless a warning is displayed. The result of round is hence the floating-point value of the given precision n or format format that is nearest to x (resp. just below or just above, depending on mode), computed as if infinite precision were used for evaluating the constant x, unless a warning is displayed.