bsc/event/subscription.go

300 lines
8.1 KiB
Go
Raw Permalink Normal View History

2017-04-14 10:29:00 +02:00
// Copyright 2016 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 event
import (
"context"
"sync"
"time"
[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
"github.com/ethereum/go-ethereum/common/gopool"
"github.com/ethereum/go-ethereum/common/mclock"
)
// Subscription represents a stream of events. The carrier of the events is typically a
// channel, but isn't part of the interface.
//
// Subscriptions can fail while established. Failures are reported through an error
// channel. It receives a value if there is an issue with the subscription (e.g. the
// network connection delivering the events has been closed). Only one value will ever be
// sent.
//
// The error channel is closed when the subscription ends successfully (i.e. when the
// source of events is closed). It is also closed when Unsubscribe is called.
//
// The Unsubscribe method cancels the sending of events. You must call Unsubscribe in all
// cases to ensure that resources related to the subscription are released. It can be
// called any number of times.
type Subscription interface {
Err() <-chan error // returns the error channel
Unsubscribe() // cancels sending of events, closing the error channel
}
// NewSubscription runs a producer function as a subscription in a new goroutine. The
// channel given to the producer is closed when Unsubscribe is called. If fn returns an
// error, it is sent on the subscription's error channel.
func NewSubscription(producer func(<-chan struct{}) error) Subscription {
s := &funcSub{unsub: make(chan struct{}), err: make(chan error, 1)}
[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
gopool.Submit(func() {
defer close(s.err)
err := producer(s.unsub)
s.mu.Lock()
defer s.mu.Unlock()
if !s.unsubscribed {
if err != nil {
s.err <- err
}
s.unsubscribed = true
}
[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
})
return s
}
type funcSub struct {
unsub chan struct{}
err chan error
mu sync.Mutex
unsubscribed bool
}
func (s *funcSub) Unsubscribe() {
s.mu.Lock()
if s.unsubscribed {
s.mu.Unlock()
return
}
s.unsubscribed = true
close(s.unsub)
s.mu.Unlock()
// Wait for producer shutdown.
<-s.err
}
func (s *funcSub) Err() <-chan error {
return s.err
}
// Resubscribe calls fn repeatedly to keep a subscription established. When the
// subscription is established, Resubscribe waits for it to fail and calls fn again. This
// process repeats until Unsubscribe is called or the active subscription ends
// successfully.
//
// Resubscribe applies backoff between calls to fn. The time between calls is adapted
// based on the error rate, but will never exceed backoffMax.
func Resubscribe(backoffMax time.Duration, fn ResubscribeFunc) Subscription {
return ResubscribeErr(backoffMax, func(ctx context.Context, _ error) (Subscription, error) {
return fn(ctx)
})
}
// A ResubscribeFunc attempts to establish a subscription.
type ResubscribeFunc func(context.Context) (Subscription, error)
// ResubscribeErr calls fn repeatedly to keep a subscription established. When the
// subscription is established, ResubscribeErr waits for it to fail and calls fn again. This
// process repeats until Unsubscribe is called or the active subscription ends
// successfully.
//
// The difference between Resubscribe and ResubscribeErr is that with ResubscribeErr,
// the error of the failing subscription is available to the callback for logging
// purposes.
//
// ResubscribeErr applies backoff between calls to fn. The time between calls is adapted
// based on the error rate, but will never exceed backoffMax.
func ResubscribeErr(backoffMax time.Duration, fn ResubscribeErrFunc) Subscription {
s := &resubscribeSub{
waitTime: backoffMax / 10,
backoffMax: backoffMax,
fn: fn,
err: make(chan error),
unsub: make(chan struct{}),
}
go s.loop()
return s
}
// A ResubscribeErrFunc attempts to establish a subscription.
// For every call but the first, the second argument to this function is
// the error that occurred with the previous subscription.
type ResubscribeErrFunc func(context.Context, error) (Subscription, error)
type resubscribeSub struct {
fn ResubscribeErrFunc
err chan error
unsub chan struct{}
unsubOnce sync.Once
lastTry mclock.AbsTime
lastSubErr error
waitTime, backoffMax time.Duration
}
func (s *resubscribeSub) Unsubscribe() {
s.unsubOnce.Do(func() {
s.unsub <- struct{}{}
<-s.err
})
}
func (s *resubscribeSub) Err() <-chan error {
return s.err
}
func (s *resubscribeSub) loop() {
defer close(s.err)
var done bool
for !done {
sub := s.subscribe()
if sub == nil {
break
}
done = s.waitForError(sub)
sub.Unsubscribe()
}
}
func (s *resubscribeSub) subscribe() Subscription {
subscribed := make(chan error)
var sub Subscription
for {
s.lastTry = mclock.Now()
ctx, cancel := context.WithCancel(context.Background())
[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
gopool.Submit(func() {
rsub, err := s.fn(ctx, s.lastSubErr)
sub = rsub
subscribed <- err
[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
})
select {
case err := <-subscribed:
cancel()
if err == nil {
if sub == nil {
panic("event: ResubscribeFunc returned nil subscription and no error")
}
return sub
}
// Subscribing failed, wait before launching the next try.
if s.backoffWait() {
return nil // unsubscribed during wait
}
case <-s.unsub:
cancel()
<-subscribed // avoid leaking the s.fn goroutine.
return nil
}
}
}
func (s *resubscribeSub) waitForError(sub Subscription) bool {
defer sub.Unsubscribe()
select {
case err := <-sub.Err():
s.lastSubErr = err
return err == nil
case <-s.unsub:
return true
}
}
func (s *resubscribeSub) backoffWait() bool {
if time.Duration(mclock.Now()-s.lastTry) > s.backoffMax {
s.waitTime = s.backoffMax / 10
} else {
s.waitTime *= 2
if s.waitTime > s.backoffMax {
s.waitTime = s.backoffMax
}
}
t := time.NewTimer(s.waitTime)
defer t.Stop()
select {
case <-t.C:
return false
case <-s.unsub:
return true
}
}
// SubscriptionScope provides a facility to unsubscribe multiple subscriptions at once.
//
// For code that handle more than one subscription, a scope can be used to conveniently
// unsubscribe all of them with a single call. The example demonstrates a typical use in a
// larger program.
//
// The zero value is ready to use.
type SubscriptionScope struct {
mu sync.Mutex
subs map[*scopeSub]struct{}
closed bool
}
type scopeSub struct {
sc *SubscriptionScope
s Subscription
}
// Track starts tracking a subscription. If the scope is closed, Track returns nil. The
// returned subscription is a wrapper. Unsubscribing the wrapper removes it from the
// scope.
func (sc *SubscriptionScope) Track(s Subscription) Subscription {
sc.mu.Lock()
defer sc.mu.Unlock()
if sc.closed {
return nil
}
if sc.subs == nil {
sc.subs = make(map[*scopeSub]struct{})
}
ss := &scopeSub{sc, s}
sc.subs[ss] = struct{}{}
return ss
}
// Close calls Unsubscribe on all tracked subscriptions and prevents further additions to
// the tracked set. Calls to Track after Close return nil.
func (sc *SubscriptionScope) Close() {
sc.mu.Lock()
defer sc.mu.Unlock()
if sc.closed {
return
}
sc.closed = true
for s := range sc.subs {
s.s.Unsubscribe()
}
sc.subs = nil
}
// Count returns the number of tracked subscriptions.
// It is meant to be used for debugging.
func (sc *SubscriptionScope) Count() int {
sc.mu.Lock()
defer sc.mu.Unlock()
return len(sc.subs)
}
func (s *scopeSub) Unsubscribe() {
s.s.Unsubscribe()
s.sc.mu.Lock()
defer s.sc.mu.Unlock()
delete(s.sc.subs, s)
}
func (s *scopeSub) Err() <-chan error {
return s.s.Err()
}