Struct f16b
#[repr(transparent)]
pub struct f16b(pub(in ::num::bfloat) u16);
A 16-bit brain floating-point value.
This type stores values using the bfloat16 encoding. It deliberately
exposes only raw-bit construction, comparison, formatting, and lossless
widening to f32.
The 16-bit brain floating-point intends to preserve the dynamic range of
a 32-bit floating-point value while using half the storage. It does
this by using 8 bits for the exponent, the same as f32, but only
using 7 bits for the mantissa. See Wikipedia on bfloat16 for
more information.
Fields
0: u16
Implementations
impl f16b
const fn from_bits(bits: u16) -> SelfRaw transmutation from
u16.This is currently identical to
transmute::<u16, f16b>(v)on all platforms. It turns out this is incredibly portable, for two reasons:- Floats and Ints have the same endianness on all supported platforms.
- IEEE 754 very precisely specifies the bit layout of floats.
However there is one caveat: prior to the 2008 version of IEEE 754, how to interpret the NaN signaling bit wasn't actually specified. Most platforms (notably x86 and ARM) picked the interpretation that was ultimately standardized in 2008, but some didn't (notably MIPS). As a result, all signaling NaNs on MIPS are quiet NaNs on x86, and vice-versa.
Rather than trying to preserve signaling-ness cross-platform, this implementation favors preserving the exact bits. This means that any payloads encoded in NaNs will be preserved even if the result of this method is sent over the network from an x86 machine to a MIPS one.
If the results of this method are only manipulated by the same architecture that produced them, then there is no portability concern.
If the input isn't NaN, then there is no portability concern.
If you don't care about signalingness (very likely), then there is no portability concern.
Note that this function is distinct from
ascasting, which attempts to preserve the numeric value, and not the bitwise value.#![feature(f16b)] # #[cfg(target_has_reliable_f16b)] { use core::num::f16b; let v = f16b::from_bits(0x4148); assert_eq!(f32::from(v), 12.5); # }const fn to_bits(self) -> u16Raw transmutation to
u16.This is currently identical to
transmute::<f16b, u16>(self)on all platforms.See
from_bitsfor some discussion of the portability of this operation (there are almost no issues).Note that this function is distinct from
ascasting, which attempts to preserve the numeric value, and not the bitwise value.#![feature(f16b)] # #[cfg(target_has_reliable_f16b)] { use core::num::f16b; assert_eq!(f16b::from_bits(0x4148).to_bits(), 0x4148); # }
Trait Implementations
impl Clone for f16b
fn clone(&self) -> Self
impl Copy for f16b
impl Debug for f16b
fn fmt(&self, formatter: &mut Formatter<'_>) -> Result
impl Default for f16b
fn default() -> Self
impl LowerExp for f16b
fn fmt(&self, formatter: &mut Formatter<'_>) -> Result
impl PartialEq for f16b
fn eq(&self, other: &Self) -> bool
impl PartialOrd for f16b
fn partial_cmp(&self, other: &Self) -> Option<Ordering>
impl UpperExp for f16b
fn fmt(&self, formatter: &mut Formatter<'_>) -> Result
Auto Trait Implementations
impl Freeze for f16b
impl RefUnwindSafe for f16b
impl Send for f16b
impl Sync for f16b
impl Unpin for f16b
impl UnsafeUnpin for f16b
impl UnwindSafe for f16b
Blanket Implementations
impl<T> Any for f16b
where
T: 'static + ?Sized,
fn type_id(&self) -> TypeId
impl<T> Borrow<T> for f16b
where
T: ?Sized,
fn borrow(&self) -> &T
impl<T> BorrowMut<T> for f16b
where
T: ?Sized,
fn borrow_mut(&mut self) -> &mut T
impl<T> CloneToUninit for f16b
where
T: Clone,
unsafe fn clone_to_uninit(&self, dest: *mut u8)
impl<T> From<T> for f16b
fn from(t: T) -> TReturns the argument unchanged.
impl<T> Printable for f16b
where
T: Copy + Debug,
impl<T> SizeHint for f16b
where
T: ?Sized,
fn lower_bound(&self) -> usizefn upper_bound(&self) -> Option<usize>
impl<T> SizedTypeProperties for f16b
impl<T, U> Into<U> for f16b
where
U: From<T>,
fn into(self) -> UCalls
U::from(self).That is, this conversion is whatever the implementation of
[From]<T> for Uchooses to do.
impl<T, U> TryFrom<U> for f16b
where
U: Into<T>,
type Error = never;fn try_from(value: U) -> Result<T, never>
impl<T, U> TryInto<U> for f16b
where
U: TryFrom<T>,
type Error = <U as TryFrom<T>>::Error;fn try_into(self) -> Result<U, <U as TryFrom<T>>::Error>