64 lines
		
	
	
		
			2.2 KiB
		
	
	
	
		
			V
		
	
	
			
		
		
	
	
			64 lines
		
	
	
		
			2.2 KiB
		
	
	
	
		
			V
		
	
	
// Copyright (c) 2019-2022 Alexander Medvednikov. All rights reserved.
 | 
						||
// Use of this source code is governed by an MIT license
 | 
						||
// that can be found in the LICENSE file.
 | 
						||
module math
 | 
						||
 | 
						||
const (
 | 
						||
	uvnan                   = u64(0x7FF8000000000001)
 | 
						||
	uvinf                   = u64(0x7FF0000000000000)
 | 
						||
	uvneginf                = u64(0xFFF0000000000000)
 | 
						||
	uvone                   = u64(0x3FF0000000000000)
 | 
						||
	mask                    = 0x7FF
 | 
						||
	shift                   = 64 - 11 - 1
 | 
						||
	bias                    = 1023
 | 
						||
	normalize_smallest_mask = u64(u64(1) << 52)
 | 
						||
	sign_mask               = u64(0x8000000000000000) // (u64(1) << 63)
 | 
						||
	frac_mask               = u64((u64(1) << u64(shift)) - u64(1))
 | 
						||
)
 | 
						||
 | 
						||
// inf returns positive infinity if sign >= 0, negative infinity if sign < 0.
 | 
						||
pub fn inf(sign int) f64 {
 | 
						||
	v := if sign >= 0 { math.uvinf } else { math.uvneginf }
 | 
						||
	return f64_from_bits(v)
 | 
						||
}
 | 
						||
 | 
						||
// nan returns an IEEE 754 ``not-a-number'' value.
 | 
						||
pub fn nan() f64 {
 | 
						||
	return f64_from_bits(math.uvnan)
 | 
						||
}
 | 
						||
 | 
						||
// is_nan reports whether f is an IEEE 754 ``not-a-number'' value.
 | 
						||
pub fn is_nan(f f64) bool {
 | 
						||
	// IEEE 754 says that only NaNs satisfy f != f.
 | 
						||
	// To avoid the floating-point hardware, could use:
 | 
						||
	// x := f64_bits(f);
 | 
						||
	// return u32(x>>shift)&mask == mask && x != uvinf && x != uvneginf
 | 
						||
	return f != f
 | 
						||
}
 | 
						||
 | 
						||
// is_inf reports whether f is an infinity, according to sign.
 | 
						||
// If sign > 0, is_inf reports whether f is positive infinity.
 | 
						||
// If sign < 0, is_inf reports whether f is negative infinity.
 | 
						||
// If sign == 0, is_inf reports whether f is either infinity.
 | 
						||
pub fn is_inf(f f64, sign int) bool {
 | 
						||
	// Test for infinity by comparing against maximum float.
 | 
						||
	// To avoid the floating-point hardware, could use:
 | 
						||
	// x := f64_bits(f);
 | 
						||
	// return sign >= 0 && x == uvinf || sign <= 0 && x == uvneginf;
 | 
						||
	return (sign >= 0 && f > max_f64) || (sign <= 0 && f < -max_f64)
 | 
						||
}
 | 
						||
 | 
						||
pub fn is_finite(f f64) bool {
 | 
						||
	return !is_nan(f) && !is_inf(f, 0)
 | 
						||
}
 | 
						||
 | 
						||
// normalize returns a normal number y and exponent exp
 | 
						||
// satisfying x == y × 2**exp. It assumes x is finite and non-zero.
 | 
						||
pub fn normalize(x f64) (f64, int) {
 | 
						||
	smallest_normal := 2.2250738585072014e-308 // 2**-1022
 | 
						||
	if abs(x) < smallest_normal {
 | 
						||
		return x * math.normalize_smallest_mask, -52
 | 
						||
	}
 | 
						||
	return x, 0
 | 
						||
}
 |