onedrive: improve speed of quickxorhash

This commits ports a fast C-implementation from https://github.com/namazso/QuickXorHash

It uses new crypto/subtle code from go1.20 to avoid the use of unsafe.

Typical speedups are about 25x  when using go1.20

    goos: linux
    goarch: amd64
    cpu: Intel(R) Celeron(R) N5105 @ 2.00GHz
    QuickXorHash-Before  2.49ms   422MB/s ±11%   100.00%
    QuickXorHash-Subtle  87.9µs 11932MB/s ± 5% +2730.83% + 42.17%

Co-Author: @namazso
This commit is contained in:
LXY 2023-01-26 19:50:12 +08:00 committed by GitHub
parent 41cc4530f3
commit dc5d5de35c
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
4 changed files with 93 additions and 116 deletions

View File

@ -7,51 +7,40 @@
// See: https://docs.microsoft.com/en-us/onedrive/developer/code-snippets/quickxorhash
package quickxorhash
// This code was ported from the code snippet linked from
// https://docs.microsoft.com/en-us/onedrive/developer/code-snippets/quickxorhash
// Which has the copyright
// This code was ported from a fast C-implementation from
// https://github.com/namazso/QuickXorHash
// which has licenced as BSD Zero Clause License
//
// BSD Zero Clause License
//
// Copyright (c) 2022 namazso <admin@namazso.eu>
//
// Permission to use, copy, modify, and/or distribute this software for any
// purpose with or without fee is hereby granted.
//
// THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES WITH
// REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY
// AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY SPECIAL, DIRECT,
// INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM
// LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR
// OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR
// PERFORMANCE OF THIS SOFTWARE.
// ------------------------------------------------------------------------------
// Copyright (c) 2016 Microsoft Corporation
//
// 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.
// ------------------------------------------------------------------------------
import (
"hash"
)
import "hash"
const (
// BlockSize is the preferred size for hashing
BlockSize = 64
// Size of the output checksum
Size = 20
bitsInLastCell = 32
shift = 11
widthInBits = 8 * Size
dataSize = (widthInBits-1)/64 + 1
Size = 20
shift = 11
widthInBits = 8 * Size
dataSize = shift * widthInBits
)
type quickXorHash struct {
data [dataSize]uint64
lengthSoFar uint64
shiftSoFar int
data [dataSize]byte
size uint64
}
// New returns a new hash.Hash computing the quickXorHash checksum.
@ -70,94 +59,37 @@ func New() hash.Hash {
//
// Implementations must not retain p.
func (q *quickXorHash) Write(p []byte) (n int, err error) {
currentshift := q.shiftSoFar
// The bitvector where we'll start xoring
vectorArrayIndex := currentshift / 64
// The position within the bit vector at which we begin xoring
vectorOffset := currentshift % 64
iterations := len(p)
if iterations > widthInBits {
iterations = widthInBits
var i int
// fill last remain
lastRemain := int(q.size) % dataSize
if lastRemain != 0 {
i += xorBytes(q.data[lastRemain:], p)
}
for i := 0; i < iterations; i++ {
isLastCell := vectorArrayIndex == len(q.data)-1
var bitsInVectorCell int
if isLastCell {
bitsInVectorCell = bitsInLastCell
} else {
bitsInVectorCell = 64
}
// There's at least 2 bitvectors before we reach the end of the array
if vectorOffset <= bitsInVectorCell-8 {
for j := i; j < len(p); j += widthInBits {
q.data[vectorArrayIndex] ^= uint64(p[j]) << uint(vectorOffset)
}
} else {
index1 := vectorArrayIndex
var index2 int
if isLastCell {
index2 = 0
} else {
index2 = vectorArrayIndex + 1
}
low := byte(bitsInVectorCell - vectorOffset)
xoredByte := byte(0)
for j := i; j < len(p); j += widthInBits {
xoredByte ^= p[j]
}
q.data[index1] ^= uint64(xoredByte) << uint(vectorOffset)
q.data[index2] ^= uint64(xoredByte) >> low
}
vectorOffset += shift
for vectorOffset >= bitsInVectorCell {
if isLastCell {
vectorArrayIndex = 0
} else {
vectorArrayIndex = vectorArrayIndex + 1
}
vectorOffset -= bitsInVectorCell
if i != len(p) {
for len(p)-i >= dataSize {
i += xorBytes(q.data[:], p[i:])
}
xorBytes(q.data[:], p[i:])
}
// Update the starting position in a circular shift pattern
q.shiftSoFar = (q.shiftSoFar + shift*(len(p)%widthInBits)) % widthInBits
q.lengthSoFar += uint64(len(p))
q.size += uint64(len(p))
return len(p), nil
}
// Calculate the current checksum
func (q *quickXorHash) checkSum() (h [Size]byte) {
// Output the data as little endian bytes
ph := 0
for i := 0; i < len(q.data)-1; i++ {
d := q.data[i]
_ = h[ph+7] // bounds check
h[ph+0] = byte(d >> (8 * 0))
h[ph+1] = byte(d >> (8 * 1))
h[ph+2] = byte(d >> (8 * 2))
h[ph+3] = byte(d >> (8 * 3))
h[ph+4] = byte(d >> (8 * 4))
h[ph+5] = byte(d >> (8 * 5))
h[ph+6] = byte(d >> (8 * 6))
h[ph+7] = byte(d >> (8 * 7))
ph += 8
func (q *quickXorHash) checkSum() (h [Size + 1]byte) {
for i := 0; i < dataSize; i++ {
shift := (i * 11) % 160
shiftBytes := shift / 8
shiftBits := shift % 8
shifted := int(q.data[i]) << shiftBits
h[shiftBytes] ^= byte(shifted)
h[shiftBytes+1] ^= byte(shifted >> 8)
}
// remaining 32 bits
d := q.data[len(q.data)-1]
h[Size-4] = byte(d >> (8 * 0))
h[Size-3] = byte(d >> (8 * 1))
h[Size-2] = byte(d >> (8 * 2))
h[Size-1] = byte(d >> (8 * 3))
h[0] ^= h[20]
// XOR the file length with the least significant bits in little endian format
d = q.lengthSoFar
d := q.size
h[Size-8] ^= byte(d >> (8 * 0))
h[Size-7] ^= byte(d >> (8 * 1))
h[Size-6] ^= byte(d >> (8 * 2))
@ -174,7 +106,7 @@ func (q *quickXorHash) checkSum() (h [Size]byte) {
// It does not change the underlying hash state.
func (q *quickXorHash) Sum(b []byte) []byte {
hash := q.checkSum()
return append(b, hash[:]...)
return append(b, hash[:Size]...)
}
// Reset resets the Hash to its initial state.
@ -196,8 +128,10 @@ func (q *quickXorHash) BlockSize() int {
}
// Sum returns the quickXorHash checksum of the data.
func Sum(data []byte) [Size]byte {
func Sum(data []byte) (h [Size]byte) {
var d quickXorHash
_, _ = d.Write(data)
return d.checkSum()
s := d.checkSum()
copy(h[:], s[:])
return h
}

View File

@ -4,6 +4,7 @@ import (
"encoding/base64"
"fmt"
"hash"
"math/rand"
"testing"
"github.com/stretchr/testify/assert"
@ -166,3 +167,16 @@ func TestReset(t *testing.T) {
// check interface
var _ hash.Hash = (*quickXorHash)(nil)
func BenchmarkQuickXorHash(b *testing.B) {
b.SetBytes(1 << 20)
buf := make([]byte, 1<<20)
rand.Read(buf)
h := New()
b.ResetTimer()
for i := 0; i < b.N; i++ {
h.Reset()
h.Write(buf)
h.Sum(nil)
}
}

View File

@ -0,0 +1,20 @@
//go:build !go1.20
package quickxorhash
func xorBytes(dst, src []byte) int {
n := len(dst)
if len(src) < n {
n = len(src)
}
if n == 0 {
return 0
}
dst = dst[:n]
//src = src[:n]
src = src[:len(dst)] // remove bounds check in loop
for i := range dst {
dst[i] ^= src[i]
}
return n
}

View File

@ -0,0 +1,9 @@
//go:build go1.20
package quickxorhash
import "crypto/subtle"
func xorBytes(dst, src []byte) int {
return subtle.XORBytes(dst, src, dst)
}