These utilities display the same test values used by the round package, but they are rounded with the C standard library, Java and Go's Float type from the math/big package as a comparison.
This table describes equivalency between various rounding modes.
| Rounding Mode | IEEE 754 | C | Java | Go math/big |
|---|---|---|---|---|
| Round to nearest neighbour, round ties to even | roundTiesToEven | FE_TONEAREST | RoundingMode.HALF_EVEN | big.ToNearestEven |
| Round to nearest neighbour, round ties toward zero | RoundingMode.HALF_DOWN | |||
| Round to nearest neighbour, round ties away from zero | roundTiesToAway | RoundingMode.HALF_UP | big.ToNearestAway | |
| Round toward positive infinity | roundTowardPositive | FE_UPWARD | RoundingMode.CEILING | big.ToPositiveInf |
| Round toward negative infinity | roundTowardNegative | FE_DOWNWARD | RoundingMode.FLOOR | big.ToNegativeInf |
| Round toward zero | roundTowardZero | FE_TOWARDZERO | RoundingMode.DOWN | big.ToZero |
| Round away from zero | RoundingMode.UP | big.AwayFromZero | ||
| Do not round, assert exact result | RoundingMode.UNNECESSARY |
Assumption: A C compiler, such as GCC or Clang, is installed.
To compile and run the C example on Linux or OS X, run:
cc -o rounding rounding.c -lm -frounding-math && ./rounding
Clang may emit a harmless warning or two during compilation.
There are other numeric packages for C, such as the GNU MPFR library, which can perform more accurate rounding than the functions in the standard library.
Assumption: A recent Java Development Kit is installed.
To compile and run the Java example on Linux or OS X, run:
javac Rounding.java && java Rounding
To run the Go example, run:
go run bigfloat.go
Note that this Go example uses the only built-in way to perform floating-point rounding: the SetPrec function in the math/big package.
The trouble with SetPrec is that it takes a number of bits of precision as a parameter, not a number of decimal places. This is not a user friendly method of rounding numbers.