bsc/core/state/snapshot/conversion.go
zjubfd 2ce00adb55
[R4R] performance improvement in many aspects (#257)
* focus on performance improvement in many aspects.

1. Do BlockBody verification concurrently;
2. Do calculation of intermediate root concurrently;
3. Preload accounts before processing blocks;
4. Make the snapshot layers configurable.
5. Reuse some object to reduce GC.

add

* rlp: improve decoder stream implementation (#22858)

This commit makes various cleanup changes to rlp.Stream.

* rlp: shrink Stream struct

This removes a lot of unused padding space in Stream by reordering the
fields. The size of Stream changes from 120 bytes to 88 bytes. Stream
instances are internally cached and reused using sync.Pool, so this does
not improve performance.

* rlp: simplify list stack

The list stack kept track of the size of the current list context as
well as the current offset into it. The size had to be stored in the
stack in order to subtract it from the remaining bytes of any enclosing
list in ListEnd. It seems that this can be implemented in a simpler
way: just subtract the size from the enclosing list context in List instead.

* rlp: use atomic.Value for type cache (#22902)

All encoding/decoding operations read the type cache to find the
writer/decoder function responsible for a type. When analyzing CPU
profiles of geth during sync, I found that the use of sync.RWMutex in
cache lookups appears in the profiles. It seems we are running into
CPU cache contention problems when package rlp is heavily used
on all CPU cores during sync.

This change makes it use atomic.Value + a writer lock instead of
sync.RWMutex. In the common case where the typeinfo entry is present in
the cache, we simply fetch the map and lookup the type.

* rlp: optimize byte array handling (#22924)

This change improves the performance of encoding/decoding [N]byte.

    name                     old time/op    new time/op    delta
    DecodeByteArrayStruct-8     336ns ± 0%     246ns ± 0%  -26.98%  (p=0.000 n=9+10)
    EncodeByteArrayStruct-8     225ns ± 1%     148ns ± 1%  -34.12%  (p=0.000 n=10+10)

    name                     old alloc/op   new alloc/op   delta
    DecodeByteArrayStruct-8      120B ± 0%       48B ± 0%  -60.00%  (p=0.000 n=10+10)
    EncodeByteArrayStruct-8     0.00B          0.00B          ~     (all equal)

* rlp: optimize big.Int decoding for size <= 32 bytes (#22927)

This change grows the static integer buffer in Stream to 32 bytes,
making it possible to decode 256bit integers without allocating a
temporary buffer.

In the recent commit 088da24, Stream struct size decreased from 120
bytes down to 88 bytes. This commit grows the struct to 112 bytes again,
but the size change will not degrade performance because Stream
instances are internally cached in sync.Pool.

    name             old time/op    new time/op    delta
    DecodeBigInts-8    12.2µs ± 0%     8.6µs ± 4%  -29.58%  (p=0.000 n=9+10)

    name             old speed      new speed      delta
    DecodeBigInts-8   230MB/s ± 0%   326MB/s ± 4%  +42.04%  (p=0.000 n=9+10)

* eth/protocols/eth, les: avoid Raw() when decoding HashOrNumber (#22841)

Getting the raw value is not necessary to decode this type, and
decoding it directly from the stream is faster.

* fix testcase

* debug no lazy

* fix can not repair

* address comments

Co-authored-by: Felix Lange <fjl@twurst.com>
2021-07-29 17:16:53 +08:00

379 lines
12 KiB
Go

// Copyright 2020 The go-ethereum Authors
// This file is part of the go-ethereum library.
//
// The go-ethereum library is free software: you can redistribute it and/or modify
// it under the terms of the GNU Lesser General Public License as published by
// the Free Software Foundation, either version 3 of the License, or
// (at your option) any later version.
//
// The go-ethereum library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public License
// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
package snapshot
import (
"bytes"
"encoding/binary"
"errors"
"fmt"
"math"
"runtime"
"sync"
"time"
"github.com/ethereum/go-ethereum/common/gopool"
"github.com/ethereum/go-ethereum/common"
"github.com/ethereum/go-ethereum/core/rawdb"
"github.com/ethereum/go-ethereum/ethdb"
"github.com/ethereum/go-ethereum/log"
"github.com/ethereum/go-ethereum/rlp"
"github.com/ethereum/go-ethereum/trie"
)
// trieKV represents a trie key-value pair
type trieKV struct {
key common.Hash
value []byte
}
type (
// trieGeneratorFn is the interface of trie generation which can
// be implemented by different trie algorithm.
trieGeneratorFn func(db ethdb.KeyValueWriter, in chan (trieKV), out chan (common.Hash))
// leafCallbackFn is the callback invoked at the leaves of the trie,
// returns the subtrie root with the specified subtrie identifier.
leafCallbackFn func(db ethdb.KeyValueWriter, accountHash, codeHash common.Hash, stat *generateStats) (common.Hash, error)
)
// GenerateAccountTrieRoot takes an account iterator and reproduces the root hash.
func GenerateAccountTrieRoot(it AccountIterator) (common.Hash, error) {
return generateTrieRoot(nil, it, common.Hash{}, stackTrieGenerate, nil, newGenerateStats(), true)
}
// GenerateStorageTrieRoot takes a storage iterator and reproduces the root hash.
func GenerateStorageTrieRoot(account common.Hash, it StorageIterator) (common.Hash, error) {
return generateTrieRoot(nil, it, account, stackTrieGenerate, nil, newGenerateStats(), true)
}
// GenerateTrie takes the whole snapshot tree as the input, traverses all the
// accounts as well as the corresponding storages and regenerate the whole state
// (account trie + all storage tries).
func GenerateTrie(snaptree *Tree, root common.Hash, src ethdb.Database, dst ethdb.KeyValueWriter) error {
// Traverse all state by snapshot, re-generate the whole state trie
acctIt, err := snaptree.AccountIterator(root, common.Hash{})
if err != nil {
return err // The required snapshot might not exist.
}
defer acctIt.Release()
got, err := generateTrieRoot(dst, acctIt, common.Hash{}, stackTrieGenerate, func(dst ethdb.KeyValueWriter, accountHash, codeHash common.Hash, stat *generateStats) (common.Hash, error) {
// Migrate the code first, commit the contract code into the tmp db.
if codeHash != emptyCode {
code := rawdb.ReadCode(src, codeHash)
if len(code) == 0 {
return common.Hash{}, errors.New("failed to read contract code")
}
rawdb.WriteCode(dst, codeHash, code)
}
// Then migrate all storage trie nodes into the tmp db.
storageIt, err := snaptree.StorageIterator(root, accountHash, common.Hash{})
if err != nil {
return common.Hash{}, err
}
defer storageIt.Release()
hash, err := generateTrieRoot(dst, storageIt, accountHash, stackTrieGenerate, nil, stat, false)
if err != nil {
return common.Hash{}, err
}
return hash, nil
}, newGenerateStats(), true)
if err != nil {
return err
}
if got != root {
return fmt.Errorf("state root hash mismatch: got %x, want %x", got, root)
}
return nil
}
// generateStats is a collection of statistics gathered by the trie generator
// for logging purposes.
type generateStats struct {
head common.Hash
start time.Time
accounts uint64 // Number of accounts done (including those being crawled)
slots uint64 // Number of storage slots done (including those being crawled)
slotsStart map[common.Hash]time.Time // Start time for account slot crawling
slotsHead map[common.Hash]common.Hash // Slot head for accounts being crawled
lock sync.RWMutex
}
// newGenerateStats creates a new generator stats.
func newGenerateStats() *generateStats {
return &generateStats{
slotsStart: make(map[common.Hash]time.Time),
slotsHead: make(map[common.Hash]common.Hash),
start: time.Now(),
}
}
// progressAccounts updates the generator stats for the account range.
func (stat *generateStats) progressAccounts(account common.Hash, done uint64) {
stat.lock.Lock()
defer stat.lock.Unlock()
stat.accounts += done
stat.head = account
}
// finishAccounts updates the gemerator stats for the finished account range.
func (stat *generateStats) finishAccounts(done uint64) {
stat.lock.Lock()
defer stat.lock.Unlock()
stat.accounts += done
}
// progressContract updates the generator stats for a specific in-progress contract.
func (stat *generateStats) progressContract(account common.Hash, slot common.Hash, done uint64) {
stat.lock.Lock()
defer stat.lock.Unlock()
stat.slots += done
stat.slotsHead[account] = slot
if _, ok := stat.slotsStart[account]; !ok {
stat.slotsStart[account] = time.Now()
}
}
// finishContract updates the generator stats for a specific just-finished contract.
func (stat *generateStats) finishContract(account common.Hash, done uint64) {
stat.lock.Lock()
defer stat.lock.Unlock()
stat.slots += done
delete(stat.slotsHead, account)
delete(stat.slotsStart, account)
}
// report prints the cumulative progress statistic smartly.
func (stat *generateStats) report() {
stat.lock.RLock()
defer stat.lock.RUnlock()
ctx := []interface{}{
"accounts", stat.accounts,
"slots", stat.slots,
"elapsed", common.PrettyDuration(time.Since(stat.start)),
}
if stat.accounts > 0 {
// If there's progress on the account trie, estimate the time to finish crawling it
if done := binary.BigEndian.Uint64(stat.head[:8]) / stat.accounts; done > 0 {
var (
left = (math.MaxUint64 - binary.BigEndian.Uint64(stat.head[:8])) / stat.accounts
speed = done/uint64(time.Since(stat.start)/time.Millisecond+1) + 1 // +1s to avoid division by zero
eta = time.Duration(left/speed) * time.Millisecond
)
// If there are large contract crawls in progress, estimate their finish time
for acc, head := range stat.slotsHead {
start := stat.slotsStart[acc]
if done := binary.BigEndian.Uint64(head[:8]); done > 0 {
var (
left = math.MaxUint64 - binary.BigEndian.Uint64(head[:8])
speed = done/uint64(time.Since(start)/time.Millisecond+1) + 1 // +1s to avoid division by zero
)
// Override the ETA if larger than the largest until now
if slotETA := time.Duration(left/speed) * time.Millisecond; eta < slotETA {
eta = slotETA
}
}
}
ctx = append(ctx, []interface{}{
"eta", common.PrettyDuration(eta),
}...)
}
}
log.Info("Iterating state snapshot", ctx...)
}
// reportDone prints the last log when the whole generation is finished.
func (stat *generateStats) reportDone() {
stat.lock.RLock()
defer stat.lock.RUnlock()
var ctx []interface{}
ctx = append(ctx, []interface{}{"accounts", stat.accounts}...)
if stat.slots != 0 {
ctx = append(ctx, []interface{}{"slots", stat.slots}...)
}
ctx = append(ctx, []interface{}{"elapsed", common.PrettyDuration(time.Since(stat.start))}...)
log.Info("Iterated snapshot", ctx...)
}
// runReport periodically prints the progress information.
func runReport(stats *generateStats, stop chan bool) {
timer := time.NewTimer(0)
defer timer.Stop()
for {
select {
case <-timer.C:
stats.report()
timer.Reset(time.Second * 8)
case success := <-stop:
if success {
stats.reportDone()
}
return
}
}
}
// generateTrieRoot generates the trie hash based on the snapshot iterator.
// It can be used for generating account trie, storage trie or even the
// whole state which connects the accounts and the corresponding storages.
func generateTrieRoot(db ethdb.KeyValueWriter, it Iterator, account common.Hash, generatorFn trieGeneratorFn, leafCallback leafCallbackFn, stats *generateStats, report bool) (common.Hash, error) {
var (
in = make(chan trieKV) // chan to pass leaves
out = make(chan common.Hash, 1) // chan to collect result
stoplog = make(chan bool, 1) // 1-size buffer, works when logging is not enabled
wg sync.WaitGroup
)
// Spin up a go-routine for trie hash re-generation
wg.Add(1)
go func() {
defer wg.Done()
generatorFn(db, in, out)
}()
// Spin up a go-routine for progress logging
if report && stats != nil {
wg.Add(1)
go func() {
defer wg.Done()
runReport(stats, stoplog)
}()
}
// Create a semaphore to assign tasks and collect results through. We'll pre-
// fill it with nils, thus using the same channel for both limiting concurrent
// processing and gathering results.
threads := runtime.NumCPU()
results := make(chan error, threads)
for i := 0; i < threads; i++ {
results <- nil // fill the semaphore
}
// stop is a helper function to shutdown the background threads
// and return the re-generated trie hash.
stop := func(fail error) (common.Hash, error) {
close(in)
result := <-out
for i := 0; i < threads; i++ {
if err := <-results; err != nil && fail == nil {
fail = err
}
}
stoplog <- fail == nil
wg.Wait()
return result, fail
}
var (
logged = time.Now()
processed = uint64(0)
leaf trieKV
)
// Start to feed leaves
for it.Next() {
if account == (common.Hash{}) {
var (
err error
fullData []byte
)
if leafCallback == nil {
fullData, err = FullAccountRLP(it.(AccountIterator).Account())
if err != nil {
return stop(err)
}
} else {
// Wait until the semaphore allows us to continue, aborting if
// a sub-task failed
if err := <-results; err != nil {
results <- nil // stop will drain the results, add a noop back for this error we just consumed
return stop(err)
}
// Fetch the next account and process it concurrently
account, err := FullAccount(it.(AccountIterator).Account())
if err != nil {
return stop(err)
}
hash := it.Hash()
gopool.Submit(func() {
subroot, err := leafCallback(db, hash, common.BytesToHash(account.CodeHash), stats)
if err != nil {
results <- err
return
}
if !bytes.Equal(account.Root, subroot.Bytes()) {
results <- fmt.Errorf("invalid subroot(path %x), want %x, have %x", hash, account.Root, subroot)
return
}
results <- nil
})
fullData, err = rlp.EncodeToBytes(account)
if err != nil {
return stop(err)
}
}
leaf = trieKV{it.Hash(), fullData}
} else {
leaf = trieKV{it.Hash(), common.CopyBytes(it.(StorageIterator).Slot())}
}
in <- leaf
// Accumulate the generation statistic if it's required.
processed++
if time.Since(logged) > 3*time.Second && stats != nil {
if account == (common.Hash{}) {
stats.progressAccounts(it.Hash(), processed)
} else {
stats.progressContract(account, it.Hash(), processed)
}
logged, processed = time.Now(), 0
}
}
// Commit the last part statistic.
if processed > 0 && stats != nil {
if account == (common.Hash{}) {
stats.finishAccounts(processed)
} else {
stats.finishContract(account, processed)
}
}
return stop(nil)
}
func stackTrieGenerate(db ethdb.KeyValueWriter, in chan trieKV, out chan common.Hash) {
t := trie.NewStackTrie(db)
for leaf := range in {
t.TryUpdate(leaf.key[:], leaf.value)
}
var root common.Hash
if db == nil {
root = t.Hash()
} else {
root, _ = t.Commit()
}
out <- root
}