mirror of
https://github.com/superseriousbusiness/gotosocial.git
synced 2024-12-23 10:42:12 +00:00
332 lines
9.9 KiB
Go
332 lines
9.9 KiB
Go
// Copyright 2020 The Libc Authors. All rights reserved.
|
|
// Use of this source code is governed by a BSD-style
|
|
// license that can be found in the LICENSE file.
|
|
|
|
// Some code is copied and adjusted from
|
|
// https://github.com/lukechampine/uint128, the original LICENSE file
|
|
// reproduced below in full as of 2021-01-19:
|
|
|
|
/*
|
|
The MIT License (MIT)
|
|
|
|
Copyright (c) 2019 Luke Champine
|
|
|
|
Permission is hereby granted, free of charge, to any person obtaining a copy
|
|
of this software and associated documentation files (the "Software"), to deal
|
|
in the Software without restriction, including without limitation the rights
|
|
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
|
copies of the Software, and to permit persons to whom the Software is
|
|
furnished to do so, subject to the following conditions:
|
|
|
|
The above copyright notice and this permission notice shall be included in
|
|
all copies or substantial portions of the Software.
|
|
|
|
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
|
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
|
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
|
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
|
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
|
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
|
|
THE SOFTWARE.
|
|
*/
|
|
|
|
package libc // import "modernc.org/libc"
|
|
|
|
import (
|
|
mbits "math/bits"
|
|
|
|
"modernc.org/mathutil"
|
|
)
|
|
|
|
type Int128 mathutil.Int128
|
|
|
|
var (
|
|
int128Minus1 = Int128{-1, -1}
|
|
int128Plus1 = Int128{Lo: 1}
|
|
)
|
|
|
|
func Int128FromFloat32(n float32) Int128 { return Int128(mathutil.NewInt128FromFloat32(n)) }
|
|
func Int128FromFloat64(n float64) Int128 { return Int128(mathutil.NewInt128FromFloat64(n)) }
|
|
func Int128FromInt16(n int16) Int128 { return Int128(mathutil.NewInt128FromInt64(int64(n))) }
|
|
func Int128FromInt32(n int32) Int128 { return Int128(mathutil.NewInt128FromInt64(int64(n))) }
|
|
func Int128FromInt64(n int64) Int128 { return Int128(mathutil.NewInt128FromInt64(n)) }
|
|
func Int128FromInt8(n int8) Int128 { return Int128(mathutil.NewInt128FromInt64(int64(n))) }
|
|
func Int128FromUint16(n uint16) Int128 { return Int128(mathutil.NewInt128FromInt64(int64(n))) }
|
|
func Int128FromUint32(n uint32) Int128 { return Int128(mathutil.NewInt128FromInt64(int64(n))) }
|
|
func Int128FromUint64(n uint64) Int128 { return Int128(mathutil.NewInt128FromUint64(n)) }
|
|
func Int128FromUint8(n uint8) Int128 { return Int128(mathutil.NewInt128FromInt64(int64(n))) }
|
|
func Int128FromUint128(n Uint128) Int128 { return Int128{Lo: int64(n.Lo), Hi: int64(n.Hi)} }
|
|
|
|
func (n *Int128) LValueDec() { *n = n.Add(int128Minus1) }
|
|
func (n *Int128) LValueInc() { *n = n.Add(int128Plus1) }
|
|
func (n *Int128) LValueShl(c int32) { *n = n.Shl(c) }
|
|
func (n *Int128) LValueShr(c int32) { *n = n.Shr(c) }
|
|
func (n Int128) And(v Int128) Int128 { return Int128{n.Lo & v.Lo, n.Hi & v.Hi} }
|
|
func (n Int128) Cmp(y Int128) int { return mathutil.Int128(n).Cmp(mathutil.Int128(y)) }
|
|
func (n Int128) Int16() int16 { return int16(n.Lo) }
|
|
func (n Int128) Int32() int32 { return int32(n.Lo) }
|
|
func (n Int128) Int64() int64 { return n.Lo }
|
|
func (n Int128) Int8() int8 { return int8(n.Lo) }
|
|
func (n Int128) Or(v Int128) Int128 { return Int128{n.Lo | v.Lo, n.Hi | v.Hi} }
|
|
func (n Int128) Uint128() (r Uint128) { return Uint128{uint64(n.Lo), uint64(n.Hi)} }
|
|
func (n Int128) Uint16() uint16 { return uint16(n.Lo) }
|
|
func (n Int128) Uint32() uint32 { return uint32(n.Lo) }
|
|
func (n Int128) Uint64() uint64 { return uint64(n.Lo) }
|
|
func (n Int128) Uint8() uint8 { return uint8(n.Lo) }
|
|
func (n Int128) Xor(v Int128) Int128 { return Int128{n.Lo ^ v.Lo, n.Hi ^ v.Hi} }
|
|
|
|
func (n Int128) Neg() Int128 {
|
|
n.Lo ^= -1
|
|
n.Hi ^= -1
|
|
return n.Add(int128Plus1)
|
|
}
|
|
|
|
func (n Int128) Float32() float32 {
|
|
switch n.Hi {
|
|
case 0:
|
|
return float32(uint64(n.Lo))
|
|
case -1:
|
|
return -float32(uint64(n.Lo))
|
|
}
|
|
|
|
if n.Hi < 0 {
|
|
n = n.Neg()
|
|
return -float32(n.Hi)*(1<<64) + float32(uint64(n.Lo))
|
|
}
|
|
|
|
return -float32(n.Hi)*(1<<64) + float32(uint64(n.Lo))
|
|
}
|
|
|
|
func (n Int128) Float64() float64 {
|
|
switch n.Hi {
|
|
case 0:
|
|
return float64(uint64(n.Lo))
|
|
case -1:
|
|
return -float64(uint64(n.Lo))
|
|
}
|
|
|
|
if n.Hi < 0 {
|
|
n = n.Neg()
|
|
return -float64(n.Hi)*(1<<64) + float64(uint64(n.Lo))
|
|
}
|
|
|
|
return float64(n.Hi)*(1<<64) + float64(uint64(n.Lo))
|
|
}
|
|
|
|
func (n Int128) Add(m Int128) (r Int128) {
|
|
r.Lo = n.Lo + m.Lo
|
|
r.Hi = n.Hi + m.Hi
|
|
if uint64(r.Lo) < uint64(n.Lo) {
|
|
r.Hi++
|
|
}
|
|
return r
|
|
}
|
|
|
|
func (n Int128) Mul(m Int128) Int128 {
|
|
hi, lo := mbits.Mul64(uint64(n.Lo), uint64(m.Lo))
|
|
_, p1 := mbits.Mul64(uint64(n.Hi), uint64(m.Lo))
|
|
_, p3 := mbits.Mul64(uint64(n.Lo), uint64(m.Hi))
|
|
hi, _ = mbits.Add64(hi, p1, 0)
|
|
hi, _ = mbits.Add64(hi, p3, 0)
|
|
return Int128{int64(lo), int64(hi)}
|
|
}
|
|
|
|
func (n Int128) Shl(c int32) (r Int128) {
|
|
if c > 64 {
|
|
r.Lo = 0
|
|
r.Hi = n.Lo << (c - 64)
|
|
} else {
|
|
r.Lo = n.Lo << c
|
|
r.Hi = n.Hi<<c | n.Lo>>(64-c)
|
|
}
|
|
return r
|
|
}
|
|
|
|
func (n Int128) Shr(c int32) (r Int128) {
|
|
if c > 64 {
|
|
r.Lo = n.Hi >> (c - 64)
|
|
switch {
|
|
case n.Hi < 0:
|
|
r.Hi = -1
|
|
default:
|
|
r.Hi = 0
|
|
}
|
|
} else {
|
|
r.Lo = n.Lo>>c | n.Hi<<(64-c)
|
|
r.Hi = n.Hi >> c
|
|
}
|
|
return r
|
|
}
|
|
|
|
type Uint128 mathutil.Uint128
|
|
|
|
func Uint128FromFloat32(n float32) Uint128 { return Uint128(mathutil.NewUint128FromFloat32(n)) }
|
|
func Uint128FromFloat64(n float64) Uint128 { return Uint128(mathutil.NewUint128FromFloat64(n)) }
|
|
func Uint128FromInt128(n Int128) Uint128 { return Uint128{Lo: uint64(n.Lo), Hi: uint64(n.Hi)} }
|
|
func Uint128FromInt16(n int16) Uint128 { return Uint128FromInt64(int64(n)) }
|
|
func Uint128FromInt32(n int32) Uint128 { return Uint128FromInt64(int64(n)) }
|
|
func Uint128FromInt8(n int8) Uint128 { return Uint128FromInt64(int64(n)) }
|
|
func Uint128FromUint16(n uint16) Uint128 { return Uint128{Lo: uint64(n)} }
|
|
func Uint128FromUint32(n uint32) Uint128 { return Uint128{Lo: uint64(n)} }
|
|
func Uint128FromUint64(n uint64) Uint128 { return Uint128{Lo: n} }
|
|
func Uint128FromUint8(n uint8) Uint128 { return Uint128{Lo: uint64(n)} }
|
|
|
|
func Uint128FromInt64(n int64) (r Uint128) {
|
|
r.Lo = uint64(n)
|
|
if n < 0 {
|
|
r.Hi = ^uint64(0)
|
|
}
|
|
return r
|
|
}
|
|
|
|
func (n *Uint128) LValueShl(c int32) { *n = n.Shl(c) }
|
|
func (n *Uint128) LValueShr(c int32) { *n = n.Shr(c) }
|
|
func (n Uint128) And(m Uint128) Uint128 { return Uint128{n.Lo & m.Lo, n.Hi & m.Hi} }
|
|
func (n Uint128) Int128() Int128 { return Int128{int64(n.Lo), int64(n.Hi)} }
|
|
func (n Uint128) Int16() int16 { return int16(n.Lo) }
|
|
func (n Uint128) Int32() int32 { return int32(n.Lo) }
|
|
func (n Uint128) Int64() int64 { return int64(n.Lo) }
|
|
func (n Uint128) Int8() int8 { return int8(n.Lo) }
|
|
func (n Uint128) Or(m Uint128) Uint128 { return Uint128{n.Lo | m.Lo, n.Hi | m.Hi} }
|
|
func (n Uint128) Uint16() uint16 { return uint16(n.Lo) }
|
|
func (n Uint128) Uint32() uint32 { return uint32(n.Lo) }
|
|
func (n Uint128) Uint64() uint64 { return n.Lo }
|
|
func (n Uint128) Uint8() uint8 { return uint8(n.Lo) }
|
|
func (n Uint128) Xor(m Uint128) Uint128 { return Uint128{n.Lo ^ m.Lo, n.Hi ^ m.Hi} }
|
|
|
|
func (n Uint128) Add(m Uint128) (r Uint128) {
|
|
var carry uint64
|
|
r.Lo, carry = mbits.Add64(n.Lo, m.Lo, 0)
|
|
r.Hi, _ = mbits.Add64(n.Hi, m.Hi, carry)
|
|
return r
|
|
}
|
|
|
|
func (n Uint128) Mul(m Uint128) Uint128 {
|
|
hi, lo := mbits.Mul64(n.Lo, m.Lo)
|
|
_, p1 := mbits.Mul64(n.Hi, m.Lo)
|
|
_, p3 := mbits.Mul64(n.Lo, m.Hi)
|
|
hi, _ = mbits.Add64(hi, p1, 0)
|
|
hi, _ = mbits.Add64(hi, p3, 0)
|
|
return Uint128{lo, hi}
|
|
}
|
|
|
|
func (n Uint128) Shr(c int32) (r Uint128) {
|
|
if c > 64 {
|
|
r.Lo = n.Hi >> (c - 64)
|
|
r.Hi = 0
|
|
} else {
|
|
r.Lo = n.Lo>>c | n.Hi<<(64-c)
|
|
r.Hi = n.Hi >> c
|
|
}
|
|
return r
|
|
}
|
|
|
|
func (n Uint128) mulOvf(m Uint128) (_ Uint128, ovf bool) {
|
|
hi, lo := mbits.Mul64(n.Lo, m.Lo)
|
|
p0, p1 := mbits.Mul64(n.Hi, m.Lo)
|
|
p2, p3 := mbits.Mul64(n.Lo, m.Hi)
|
|
hi, c0 := mbits.Add64(hi, p1, 0)
|
|
hi, c1 := mbits.Add64(hi, p3, 0)
|
|
ovf = p0 != 0 || p2 != 0 || c0 != 0 || c1 != 0
|
|
return Uint128{lo, hi}, ovf
|
|
}
|
|
|
|
func (n Uint128) quoRem(m Uint128) (q, r Uint128) {
|
|
if m.Hi == 0 {
|
|
var r64 uint64
|
|
q, r64 = n.quoRem64(m.Lo)
|
|
r = Uint128FromUint64(r64)
|
|
} else {
|
|
// generate a "trial quotient," guaranteed to be within 1 of the actual
|
|
// quotient, then adjust.
|
|
nz := mbits.LeadingZeros64(m.Hi)
|
|
v1 := m.Shl(int32(nz))
|
|
u1 := n.Shr(1)
|
|
tq, _ := mbits.Div64(u1.Hi, u1.Lo, v1.Hi)
|
|
tq >>= 63 - nz
|
|
if tq != 0 {
|
|
tq--
|
|
}
|
|
q = Uint128FromUint64(tq)
|
|
// calculate remainder using trial quotient, then adjust if remainder is
|
|
// greater than divisor
|
|
r = n.Sub(m.mul64(tq))
|
|
if r.Cmp(m) >= 0 {
|
|
q = q.add64(1)
|
|
r = r.Sub(m)
|
|
}
|
|
}
|
|
return
|
|
}
|
|
|
|
func (n Uint128) quoRem64(m uint64) (q Uint128, r uint64) {
|
|
if n.Hi < m {
|
|
q.Lo, r = mbits.Div64(n.Hi, n.Lo, m)
|
|
} else {
|
|
q.Hi, r = mbits.Div64(0, n.Hi, m)
|
|
q.Lo, r = mbits.Div64(r, n.Lo, m)
|
|
}
|
|
return
|
|
}
|
|
|
|
func (n Uint128) Div(m Uint128) (r Uint128) {
|
|
r, _ = n.quoRem(m)
|
|
return r
|
|
}
|
|
|
|
func (n Uint128) Shl(c int32) (r Uint128) {
|
|
if c > 64 {
|
|
r.Lo = 0
|
|
r.Hi = n.Lo << (c - 64)
|
|
} else {
|
|
r.Lo = n.Lo << c
|
|
r.Hi = n.Hi<<c | n.Lo>>(64-c)
|
|
}
|
|
return
|
|
}
|
|
|
|
func (n Uint128) Sub(m Uint128) Uint128 {
|
|
lo, borrow := mbits.Sub64(n.Lo, m.Lo, 0)
|
|
hi, _ := mbits.Sub64(n.Hi, m.Hi, borrow)
|
|
return Uint128{lo, hi}
|
|
}
|
|
|
|
func (n Uint128) mul64(m uint64) Uint128 {
|
|
hi, lo := mbits.Mul64(n.Lo, m)
|
|
_, p1 := mbits.Mul64(n.Hi, m)
|
|
hi, _ = mbits.Add64(hi, p1, 0)
|
|
return Uint128{lo, hi}
|
|
}
|
|
|
|
func (n Uint128) Cmp(m Uint128) int {
|
|
if n == m {
|
|
return 0
|
|
} else if n.Hi < m.Hi || (n.Hi == m.Hi && n.Lo < m.Lo) {
|
|
return -1
|
|
} else {
|
|
return 1
|
|
}
|
|
}
|
|
|
|
func (n Uint128) add64(m uint64) Uint128 {
|
|
lo, carry := mbits.Add64(n.Lo, m, 0)
|
|
hi, _ := mbits.Add64(n.Hi, 0, carry)
|
|
return Uint128{lo, hi}
|
|
}
|
|
|
|
func (n Uint128) Float32() float32 {
|
|
if n.Hi == 0 {
|
|
return float32(n.Lo)
|
|
}
|
|
|
|
return float32(n.Hi)*(1<<64) + float32(n.Lo)
|
|
}
|
|
|
|
func (n Uint128) Float64() float64 {
|
|
if n.Hi == 0 {
|
|
return float64(n.Lo)
|
|
}
|
|
|
|
return float64(n.Hi)*(1<<64) + float64(n.Lo)
|
|
}
|