performance improvement in many aspects by unclezoro · Pull Request #257 · bnb-chain/bsc

unclezoro · 2021-05-26T03:55:19Z

Description

This PR tries to improve the performance of BSC.

Rationale

Part of Block Processing can be parallel, and the layer of the snapshot is too deeper for the current implementation. This PR focus on:

Do BlockBody verification concurrently;
Do the calculation of intermediate root concurrently;
Commit the MPTs concurrently;
Preload accounts before processing blocks;
Make the snapshot layers configurable.
Reuse some objects to reduce GC.
Cherry-pick the RLP improvement from go-ethereum.
Other minor improvements.

Example

The testing result on an 8 core 2.5G Hz Linux VM:

The full sync speed on the main-net is 40% faster.
The execution of an BEP20 transfer is 50% faster.

Changes

New command flags: --triesInMemory

Preflight checks

build passed (make build)
tests passed (make test)
manual transaction test passed

Already reviewed by

...

Related issues

... reference related issue #'s here ...

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.

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)

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)

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

yutianwu · 2021-07-14T03:55:17Z

core/blockchain.go

 			}
 		}
 	}
+	wg.Wait()


why wait here instead of line 1487

I want to state.Commit and blockBatch.Write() run concurrently

core/state/database.go

yutianwu · 2021-07-14T04:00:33Z

core/state/snapshot/difflayer.go

 		if data, ok := storage[storageHash]; ok {
 			snapshotDirtyStorageHitMeter.Mark(1)
-			snapshotDirtyStorageHitDepthHist.Update(int64(depth))
+			//snapshotDirtyStorageHitDepthHist.Update(int64(depth))


should we delete the comments here

It is commented out because of the performance issue, if we need the metrics in some scenery, we get it back without deleting it.

yutianwu · 2021-07-14T10:18:59Z

core/vm/evm.go

-		chainRules:   chainConfig.Rules(blockCtx.BlockNumber),
-		interpreters: make([]Interpreter, 0, 1),
-	}
+	evm := EvmPool.Get().(*EVM)


where do we put the evm back to pool

yutianwu · 2021-07-14T10:36:40Z

eth/downloader/api.go

 	"sync"

+	"github.com/ethereum/go-ethereum/common/gopool"
+


yutianwu · 2021-07-14T10:40:50Z

eth/downloader/peer.go

 	// Issue the header retrieval request (absolute upwards without gaps)
-	go p.peer.RequestHeadersByNumber(from, count, 0, false)

+	go p.peer.RequestHeadersByNumber(from, count, 0, false)


why not use gopool here

It is used when clients doing fast sync and snap sync. Not related to performance

yutianwu · 2021-07-14T10:42:51Z

p2p/rlpx/rlpx.go

 	"github.com/ethereum/go-ethereum/rlp"
-	"github.com/golang/snappy"
-	"golang.org/x/crypto/sha3"
+	"github.com/oxtoacart/bpool"


yutianwu · 2021-07-15T03:44:08Z

core/blockchain.go

+		}
+		if len(accountsSlice) >= preLoadLimit {
+			objsChan := make(chan []*state.StateObject, runtime.NumCPU())
+			for i := 0; i < runtime.NumCPU(); i++ {


should we make sure that len(accountsSlice) > runtime.NumCPU()

Add one more condition: len(accountsSlice) > runtime.NumCPU()

core/blockchain.go

adithyaxx · 2021-07-28T06:26:05Z

Can we get this PR merged?

yvettep321

_thank you _

unclezoro force-pushed the improve_8 branch 13 times, most recently from 62904a0 to 5bdb7b4 Compare May 31, 2021 07:35

unclezoro force-pushed the improve_8 branch 17 times, most recently from 0a1cc10 to e750b5d Compare June 7, 2021 13:19

fjl added 4 commits July 12, 2021 09:41

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

bbd2148

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

unclezoro force-pushed the improve_8 branch from 2fd6ab4 to bbd2148 Compare July 12, 2021 01:41

fix testcase

5427bde

yutianwu reviewed Jul 14, 2021

View reviewed changes

core/state/database.go Show resolved Hide resolved

yutianwu reviewed Jul 14, 2021

View reviewed changes

debug no lazy

90863ac

unclezoro force-pushed the improve_8 branch from 2be9ae2 to 90863ac Compare July 14, 2021 06:13

yutianwu reviewed Jul 14, 2021

View reviewed changes

yutianwu reviewed Jul 15, 2021

View reviewed changes

core/blockchain.go Outdated Show resolved Hide resolved

zfesd approved these changes Jul 19, 2021

View reviewed changes

fix can not repair

94e1505

unclezoro force-pushed the improve_8 branch from 38c7d26 to 94e1505 Compare July 20, 2021 12:25

address comments

8975587

ifstatic approved these changes Jul 26, 2021

View reviewed changes

yutianwu approved these changes Jul 29, 2021

View reviewed changes

j75689 approved these changes Jul 29, 2021

View reviewed changes

unclezoro mentioned this pull request Jul 29, 2021

[R4R]prepare release 1.1.1-beta #334

Merged

3 tasks

j75689 mentioned this pull request Jul 30, 2021

BSC synchronization issues #338

Closed

yvettep321 reviewed Aug 31, 2021

View reviewed changes

brilliant-lx mentioned this pull request Oct 27, 2023

code: remove accountTrieCache and storageTrieCache #1949

Merged

sunny2022da mentioned this pull request Apr 10, 2024

feat: introduce hash cache to improve performance bnb-chain/op-geth#81

Merged

Conversation

unclezoro commented May 26, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Rationale

Example

Changes

Preflight checks

Already reviewed by

Related issues

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

adithyaxx commented Jul 28, 2021

Uh oh!

yvettep321 left a comment

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

8 participants

unclezoro commented May 26, 2021 •

edited

Loading