mirror of
https://codeberg.org/andyscott/ziglings.git
synced 2024-11-08 19:20:47 -05:00
94 lines
3.5 KiB
Zig
94 lines
3.5 KiB
Zig
//
|
|
// Zig has support for IEEE-754 floating-point numbers in these
|
|
// specific sizes: f16, f32, f64, f80, and f128. Floating point
|
|
// literals may be written in the same ways as integers but also
|
|
// in scientific notation:
|
|
//
|
|
// const a1: f32 = 1200; // 1,200
|
|
// const a2: f32 = 1.2e+3; // 1,200
|
|
// const b1: f32 = -500_000.0; // -500,000
|
|
// const b2: f32 = -5.0e+5; // -500,000
|
|
//
|
|
// Hex floats can't use the letter 'e' because that's a hex
|
|
// digit, so we use a 'p' instead:
|
|
//
|
|
// const hex: f16 = 0x2A.F7p+3; // Wow, that's arcane!
|
|
//
|
|
// Be sure to use a float type that is large enough to store your
|
|
// value (both in terms of significant digits and scale).
|
|
// Rounding may or may not be okay, but numbers which are too
|
|
// large or too small become inf or -inf (positive or negative
|
|
// infinity)!
|
|
//
|
|
// const pi: f16 = 3.1415926535; // rounds to 3.140625
|
|
// const av: f16 = 6.02214076e+23; // Avogadro's inf(inity)!
|
|
//
|
|
// When performing math operations with numeric literals, ensure
|
|
// the types match. Zig does not perform unsafe type coercions
|
|
// behind your back:
|
|
//
|
|
// var foo: f16 = 5; // NO ERROR
|
|
//
|
|
// var foo: u16 = 5; // A literal of a different type
|
|
// var bar: f16 = foo; // ERROR
|
|
//
|
|
// Please fix the two float problems with this program and
|
|
// display the result as a whole number.
|
|
|
|
const print = @import("std").debug.print;
|
|
|
|
pub fn main() void {
|
|
// The approximate weight of the Space Shuttle upon liftoff
|
|
// (including boosters and fuel tank) was 4,480,000 lb.
|
|
//
|
|
// We'll convert this weight from pound to kilograms at a
|
|
// conversion of 0.453592kg to the pound.
|
|
const shuttle_weight: f16 = 0.453592 * 4480e6;
|
|
|
|
// By default, float values are formatted in scientific
|
|
// notation. Try experimenting with '{d}' and '{d:.3}' to see
|
|
// how decimal formatting works.
|
|
print("Shuttle liftoff weight: {d:.0}kg\n", .{shuttle_weight});
|
|
}
|
|
|
|
// Floating further:
|
|
//
|
|
// As an example, Zig's f16 is a IEEE 754 "half-precision" binary
|
|
// floating-point format ("binary16"), which is stored in memory
|
|
// like so:
|
|
//
|
|
// 0 1 0 0 0 0 1 0 0 1 0 0 1 0 0 0
|
|
// | |-------| |-----------------|
|
|
// | exponent significand
|
|
// |
|
|
// sign
|
|
//
|
|
// This example is the decimal number 3.140625, which happens to
|
|
// be the closest representation of Pi we can make with an f16
|
|
// due to the way IEEE-754 floating points store digits:
|
|
//
|
|
// * Sign bit 0 makes the number positive.
|
|
// * Exponent bits 10000 are a scale of 16.
|
|
// * Significand bits 1001001000 are the decimal value 584.
|
|
//
|
|
// IEEE-754 saves space by modifying these values: the value
|
|
// 01111 is always subtracted from the exponent bits (in our
|
|
// case, 10000 - 01111 = 1, so our exponent is 2^1) and our
|
|
// significand digits become the decimal value _after_ an
|
|
// implicit 1 (so 1.1001001000 or 1.5703125 in decimal)! This
|
|
// gives us:
|
|
//
|
|
// 2^1 * 1.5703125 = 3.140625
|
|
//
|
|
// Feel free to forget these implementation details immediately.
|
|
// The important thing to know is that floating point numbers are
|
|
// great at storing big and small values (f64 lets you work with
|
|
// numbers on the scale of the number of atoms in the universe),
|
|
// but digits may be rounded, leading to results which are less
|
|
// precise than integers.
|
|
//
|
|
// Fun fact: sometimes you'll see the significand labeled as a
|
|
// "mantissa" but Donald E. Knuth says not to do that.
|
|
//
|
|
// C compatibility fact: There is also a Zig floating point type
|
|
// specifically for working with C ABIs called c_longdouble.
|