msf 0.0.1

multiprecision int/float library
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31

use super::*;
use equator::assert;

#[inline]
pub fn temp_big_float_scratch(precision_bits: u64) -> StackReq {
    let nlimbs = (precision_bits as usize).div_ceil(Limb::BITS as usize);
    StackReq::new::<u64>(2 + nlimbs)
}

#[inline]
#[track_caller]
pub fn temp_big_float_uninit(precision_bits: u64, stack: &mut PodStack) -> (&mut BigFloat, &mut PodStack) {
    assert!(core::mem::size_of::<Limb>() == 8);
    assert!(all(precision_bits > 1, precision_bits < u64::MAX - Limb::BITS as u64,));
    let nlimbs = (precision_bits as usize).div_ceil(Limb::BITS as usize);
    let (buf, stack) = stack.make_raw::<u64>(2 + nlimbs);
    let buf = buf.as_mut_ptr();
    let buf = core::ptr::slice_from_raw_parts_mut(buf, nlimbs) as *mut BigFloat;
    let buf = unsafe { &mut *buf };
    buf.__precision_bits = Some(unsafe { NonZeroU64::new_unchecked(precision_bits) });
    (buf, stack)
}

#[inline]
#[track_caller]
pub fn temp_big_float_zero(precision_bits: u64, stack: &mut PodStack) -> (&mut BigFloat, &mut PodStack) {
    let (buf, stack) = temp_big_float_uninit(precision_bits, stack);
    buf.sign_biased_exponent = 0;
    buf.mantissa_mut().fill(consts::LIMB_ZERO);
    (buf, stack)
}