bsc/rlp/typecache.go

274 lines
6.9 KiB
Go
Raw Normal View History

2015-07-07 03:54:22 +03:00
// Copyright 2014 The go-ethereum Authors
// This file is part of the go-ethereum library.
2015-07-07 03:54:22 +03:00
//
// The go-ethereum library is free software: you can redistribute it and/or modify
2015-07-07 03:54:22 +03:00
// 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,
2015-07-07 03:54:22 +03:00
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2015-07-07 03:54:22 +03:00
// 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/>.
2015-07-07 03:54:22 +03:00
package rlp
import (
2015-12-21 22:05:20 +02:00
"fmt"
"reflect"
2015-12-21 22:05:20 +02:00
"strings"
"sync"
[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 12:16:53 +03:00
"sync/atomic"
)
[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 12:16:53 +03:00
// typeinfo is an entry in the type cache.
type typeinfo struct {
decoder decoder
decoderErr error // error from makeDecoder
writer writer
writerErr error // error from makeWriter
}
rlp: improve nil pointer handling (#20064) * rlp: improve nil pointer handling In both encoder and decoder, the rules for encoding nil pointers were a bit hard to understand, and didn't leave much choice. Since RLP allows two empty values (empty list, empty string), any protocol built on RLP must choose either of these values to represent the null value in a certain context. This change adds choice in the form of two new struct tags, "nilString" and "nilList". These can be used to specify how a nil pointer value is encoded. The "nil" tag still exists, but its implementation is now explicit and defines exactly how nil pointers are handled in a single place. Another important change in this commit is how nil pointers and the Encoder interface interact. The EncodeRLP method was previously called even on nil values, which was supposed to give users a choice of how their value would be handled when nil. It turns out this is a stupid idea. If you create a network protocol containing an object defined in another package, it's better to be able to say that the object should be a list or string when nil in the definition of the protocol message rather than defining the encoding of nil on the object itself. As of this commit, the encoding rules for pointers now take precedence over the Encoder interface rule. I think the "nil" tag will work fine for most cases. For special kinds of objects which are a struct in Go but strings in RLP, code using the object can specify the desired encoding of nil using the "nilString" and "nilList" tags. * rlp: propagate struct field type errors If a struct contained fields of undecodable type, the encoder and decoder would panic instead of returning an error. Fix this by propagating type errors in makeStruct{Writer,Decoder} and add a test.
2019-09-13 12:10:57 +03:00
// tags represents struct tags.
type tags struct {
2015-12-21 22:05:20 +02:00
// rlp:"nil" controls whether empty input results in a nil pointer.
[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 12:16:53 +03:00
// nilKind is the kind of empty value allowed for the field.
rlp: improve nil pointer handling (#20064) * rlp: improve nil pointer handling In both encoder and decoder, the rules for encoding nil pointers were a bit hard to understand, and didn't leave much choice. Since RLP allows two empty values (empty list, empty string), any protocol built on RLP must choose either of these values to represent the null value in a certain context. This change adds choice in the form of two new struct tags, "nilString" and "nilList". These can be used to specify how a nil pointer value is encoded. The "nil" tag still exists, but its implementation is now explicit and defines exactly how nil pointers are handled in a single place. Another important change in this commit is how nil pointers and the Encoder interface interact. The EncodeRLP method was previously called even on nil values, which was supposed to give users a choice of how their value would be handled when nil. It turns out this is a stupid idea. If you create a network protocol containing an object defined in another package, it's better to be able to say that the object should be a list or string when nil in the definition of the protocol message rather than defining the encoding of nil on the object itself. As of this commit, the encoding rules for pointers now take precedence over the Encoder interface rule. I think the "nil" tag will work fine for most cases. For special kinds of objects which are a struct in Go but strings in RLP, code using the object can specify the desired encoding of nil using the "nilString" and "nilList" tags. * rlp: propagate struct field type errors If a struct contained fields of undecodable type, the encoder and decoder would panic instead of returning an error. Fix this by propagating type errors in makeStruct{Writer,Decoder} and add a test.
2019-09-13 12:10:57 +03:00
nilKind Kind
[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 12:16:53 +03:00
nilOK bool
// rlp:"optional" allows for a field to be missing in the input list.
// If this is set, all subsequent fields must also be optional.
optional bool
rlp: improve nil pointer handling (#20064) * rlp: improve nil pointer handling In both encoder and decoder, the rules for encoding nil pointers were a bit hard to understand, and didn't leave much choice. Since RLP allows two empty values (empty list, empty string), any protocol built on RLP must choose either of these values to represent the null value in a certain context. This change adds choice in the form of two new struct tags, "nilString" and "nilList". These can be used to specify how a nil pointer value is encoded. The "nil" tag still exists, but its implementation is now explicit and defines exactly how nil pointers are handled in a single place. Another important change in this commit is how nil pointers and the Encoder interface interact. The EncodeRLP method was previously called even on nil values, which was supposed to give users a choice of how their value would be handled when nil. It turns out this is a stupid idea. If you create a network protocol containing an object defined in another package, it's better to be able to say that the object should be a list or string when nil in the definition of the protocol message rather than defining the encoding of nil on the object itself. As of this commit, the encoding rules for pointers now take precedence over the Encoder interface rule. I think the "nil" tag will work fine for most cases. For special kinds of objects which are a struct in Go but strings in RLP, code using the object can specify the desired encoding of nil using the "nilString" and "nilList" tags. * rlp: propagate struct field type errors If a struct contained fields of undecodable type, the encoder and decoder would panic instead of returning an error. Fix this by propagating type errors in makeStruct{Writer,Decoder} and add a test.
2019-09-13 12:10:57 +03:00
[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 12:16:53 +03:00
// rlp:"tail" controls whether this field swallows additional list elements. It can
// only be set for the last field, which must be of slice type.
2015-12-21 22:05:20 +02:00
tail bool
rlp: improve nil pointer handling (#20064) * rlp: improve nil pointer handling In both encoder and decoder, the rules for encoding nil pointers were a bit hard to understand, and didn't leave much choice. Since RLP allows two empty values (empty list, empty string), any protocol built on RLP must choose either of these values to represent the null value in a certain context. This change adds choice in the form of two new struct tags, "nilString" and "nilList". These can be used to specify how a nil pointer value is encoded. The "nil" tag still exists, but its implementation is now explicit and defines exactly how nil pointers are handled in a single place. Another important change in this commit is how nil pointers and the Encoder interface interact. The EncodeRLP method was previously called even on nil values, which was supposed to give users a choice of how their value would be handled when nil. It turns out this is a stupid idea. If you create a network protocol containing an object defined in another package, it's better to be able to say that the object should be a list or string when nil in the definition of the protocol message rather than defining the encoding of nil on the object itself. As of this commit, the encoding rules for pointers now take precedence over the Encoder interface rule. I think the "nil" tag will work fine for most cases. For special kinds of objects which are a struct in Go but strings in RLP, code using the object can specify the desired encoding of nil using the "nilString" and "nilList" tags. * rlp: propagate struct field type errors If a struct contained fields of undecodable type, the encoder and decoder would panic instead of returning an error. Fix this by propagating type errors in makeStruct{Writer,Decoder} and add a test.
2019-09-13 12:10:57 +03:00
2017-03-07 14:37:53 +03:00
// rlp:"-" ignores fields.
ignored bool
}
rlp: improve nil pointer handling (#20064) * rlp: improve nil pointer handling In both encoder and decoder, the rules for encoding nil pointers were a bit hard to understand, and didn't leave much choice. Since RLP allows two empty values (empty list, empty string), any protocol built on RLP must choose either of these values to represent the null value in a certain context. This change adds choice in the form of two new struct tags, "nilString" and "nilList". These can be used to specify how a nil pointer value is encoded. The "nil" tag still exists, but its implementation is now explicit and defines exactly how nil pointers are handled in a single place. Another important change in this commit is how nil pointers and the Encoder interface interact. The EncodeRLP method was previously called even on nil values, which was supposed to give users a choice of how their value would be handled when nil. It turns out this is a stupid idea. If you create a network protocol containing an object defined in another package, it's better to be able to say that the object should be a list or string when nil in the definition of the protocol message rather than defining the encoding of nil on the object itself. As of this commit, the encoding rules for pointers now take precedence over the Encoder interface rule. I think the "nil" tag will work fine for most cases. For special kinds of objects which are a struct in Go but strings in RLP, code using the object can specify the desired encoding of nil using the "nilString" and "nilList" tags. * rlp: propagate struct field type errors If a struct contained fields of undecodable type, the encoder and decoder would panic instead of returning an error. Fix this by propagating type errors in makeStruct{Writer,Decoder} and add a test.
2019-09-13 12:10:57 +03:00
// typekey is the key of a type in typeCache. It includes the struct tags because
// they might generate a different decoder.
type typekey struct {
reflect.Type
tags
}
type decoder func(*Stream, reflect.Value) error
type writer func(reflect.Value, *encbuf) error
[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 12:16:53 +03:00
var theTC = newTypeCache()
type typeCache struct {
cur atomic.Value
// This lock synchronizes writers.
mu sync.Mutex
next map[typekey]*typeinfo
}
func newTypeCache() *typeCache {
c := new(typeCache)
c.cur.Store(make(map[typekey]*typeinfo))
return c
}
func cachedDecoder(typ reflect.Type) (decoder, error) {
[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 12:16:53 +03:00
info := theTC.info(typ)
return info.decoder, info.decoderErr
}
func cachedWriter(typ reflect.Type) (writer, error) {
[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 12:16:53 +03:00
info := theTC.info(typ)
return info.writer, info.writerErr
}
[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 12:16:53 +03:00
func (c *typeCache) info(typ reflect.Type) *typeinfo {
key := typekey{Type: typ}
if info := c.cur.Load().(map[typekey]*typeinfo)[key]; info != nil {
return info
}
[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 12:16:53 +03:00
// Not in the cache, need to generate info for this type.
return c.generate(typ, tags{})
}
[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 12:16:53 +03:00
func (c *typeCache) generate(typ reflect.Type, tags tags) *typeinfo {
c.mu.Lock()
defer c.mu.Unlock()
cur := c.cur.Load().(map[typekey]*typeinfo)
if info := cur[typekey{typ, tags}]; info != nil {
return info
}
// Copy cur to next.
c.next = make(map[typekey]*typeinfo, len(cur)+1)
for k, v := range cur {
c.next[k] = v
}
// Generate.
info := c.infoWhileGenerating(typ, tags)
// next -> cur
c.cur.Store(c.next)
c.next = nil
return info
}
func (c *typeCache) infoWhileGenerating(typ reflect.Type, tags tags) *typeinfo {
key := typekey{typ, tags}
[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 12:16:53 +03:00
if info := c.next[key]; info != nil {
return info
}
[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 12:16:53 +03:00
// Put a dummy value into the cache before generating.
// If the generator tries to lookup itself, it will get
// the dummy value and won't call itself recursively.
[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 12:16:53 +03:00
info := new(typeinfo)
c.next[key] = info
info.generate(typ, tags)
return info
}
type field struct {
[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 12:16:53 +03:00
index int
info *typeinfo
optional bool
}
[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 12:16:53 +03:00
// structFields resolves the typeinfo of all public fields in a struct type.
func structFields(typ reflect.Type) (fields []field, err error) {
[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 12:16:53 +03:00
var (
lastPublic = lastPublicField(typ)
anyOptional = false
)
for i := 0; i < typ.NumField(); i++ {
if f := typ.Field(i); f.PkgPath == "" { // exported
tags, err := parseStructTag(typ, i, lastPublic)
2015-12-21 22:05:20 +02:00
if err != nil {
return nil, 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 12:16:53 +03:00
// Skip rlp:"-" fields.
2017-03-07 14:37:53 +03:00
if tags.ignored {
continue
}
[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 12:16:53 +03:00
// If any field has the "optional" tag, subsequent fields must also have it.
if tags.optional || tags.tail {
anyOptional = true
} else if anyOptional {
return nil, fmt.Errorf(`rlp: struct field %v.%s needs "optional" tag`, typ, f.Name)
}
info := theTC.infoWhileGenerating(f.Type, tags)
fields = append(fields, field{i, info, tags.optional})
}
}
return fields, nil
}
rlp: improve nil pointer handling (#20064) * rlp: improve nil pointer handling In both encoder and decoder, the rules for encoding nil pointers were a bit hard to understand, and didn't leave much choice. Since RLP allows two empty values (empty list, empty string), any protocol built on RLP must choose either of these values to represent the null value in a certain context. This change adds choice in the form of two new struct tags, "nilString" and "nilList". These can be used to specify how a nil pointer value is encoded. The "nil" tag still exists, but its implementation is now explicit and defines exactly how nil pointers are handled in a single place. Another important change in this commit is how nil pointers and the Encoder interface interact. The EncodeRLP method was previously called even on nil values, which was supposed to give users a choice of how their value would be handled when nil. It turns out this is a stupid idea. If you create a network protocol containing an object defined in another package, it's better to be able to say that the object should be a list or string when nil in the definition of the protocol message rather than defining the encoding of nil on the object itself. As of this commit, the encoding rules for pointers now take precedence over the Encoder interface rule. I think the "nil" tag will work fine for most cases. For special kinds of objects which are a struct in Go but strings in RLP, code using the object can specify the desired encoding of nil using the "nilString" and "nilList" tags. * rlp: propagate struct field type errors If a struct contained fields of undecodable type, the encoder and decoder would panic instead of returning an error. Fix this by propagating type errors in makeStruct{Writer,Decoder} and add a test.
2019-09-13 12:10:57 +03:00
type structFieldError struct {
typ reflect.Type
field int
err error
}
func (e structFieldError) Error() string {
return fmt.Sprintf("%v (struct field %v.%s)", e.err, e.typ, e.typ.Field(e.field).Name)
}
type structTagError struct {
typ reflect.Type
field, tag, err string
}
func (e structTagError) Error() string {
return fmt.Sprintf("rlp: invalid struct tag %q for %v.%s (%s)", e.tag, e.typ, e.field, e.err)
}
func parseStructTag(typ reflect.Type, fi, lastPublic int) (tags, error) {
2015-12-21 22:05:20 +02:00
f := typ.Field(fi)
var ts tags
for _, t := range strings.Split(f.Tag.Get("rlp"), ",") {
switch t = strings.TrimSpace(t); t {
case "":
2017-03-07 14:37:53 +03:00
case "-":
ts.ignored = true
rlp: improve nil pointer handling (#20064) * rlp: improve nil pointer handling In both encoder and decoder, the rules for encoding nil pointers were a bit hard to understand, and didn't leave much choice. Since RLP allows two empty values (empty list, empty string), any protocol built on RLP must choose either of these values to represent the null value in a certain context. This change adds choice in the form of two new struct tags, "nilString" and "nilList". These can be used to specify how a nil pointer value is encoded. The "nil" tag still exists, but its implementation is now explicit and defines exactly how nil pointers are handled in a single place. Another important change in this commit is how nil pointers and the Encoder interface interact. The EncodeRLP method was previously called even on nil values, which was supposed to give users a choice of how their value would be handled when nil. It turns out this is a stupid idea. If you create a network protocol containing an object defined in another package, it's better to be able to say that the object should be a list or string when nil in the definition of the protocol message rather than defining the encoding of nil on the object itself. As of this commit, the encoding rules for pointers now take precedence over the Encoder interface rule. I think the "nil" tag will work fine for most cases. For special kinds of objects which are a struct in Go but strings in RLP, code using the object can specify the desired encoding of nil using the "nilString" and "nilList" tags. * rlp: propagate struct field type errors If a struct contained fields of undecodable type, the encoder and decoder would panic instead of returning an error. Fix this by propagating type errors in makeStruct{Writer,Decoder} and add a test.
2019-09-13 12:10:57 +03:00
case "nil", "nilString", "nilList":
2015-12-21 22:05:20 +02:00
ts.nilOK = true
rlp: improve nil pointer handling (#20064) * rlp: improve nil pointer handling In both encoder and decoder, the rules for encoding nil pointers were a bit hard to understand, and didn't leave much choice. Since RLP allows two empty values (empty list, empty string), any protocol built on RLP must choose either of these values to represent the null value in a certain context. This change adds choice in the form of two new struct tags, "nilString" and "nilList". These can be used to specify how a nil pointer value is encoded. The "nil" tag still exists, but its implementation is now explicit and defines exactly how nil pointers are handled in a single place. Another important change in this commit is how nil pointers and the Encoder interface interact. The EncodeRLP method was previously called even on nil values, which was supposed to give users a choice of how their value would be handled when nil. It turns out this is a stupid idea. If you create a network protocol containing an object defined in another package, it's better to be able to say that the object should be a list or string when nil in the definition of the protocol message rather than defining the encoding of nil on the object itself. As of this commit, the encoding rules for pointers now take precedence over the Encoder interface rule. I think the "nil" tag will work fine for most cases. For special kinds of objects which are a struct in Go but strings in RLP, code using the object can specify the desired encoding of nil using the "nilString" and "nilList" tags. * rlp: propagate struct field type errors If a struct contained fields of undecodable type, the encoder and decoder would panic instead of returning an error. Fix this by propagating type errors in makeStruct{Writer,Decoder} and add a test.
2019-09-13 12:10:57 +03:00
if f.Type.Kind() != reflect.Ptr {
return ts, structTagError{typ, f.Name, t, "field is not a pointer"}
}
switch t {
case "nil":
ts.nilKind = defaultNilKind(f.Type.Elem())
case "nilString":
ts.nilKind = String
case "nilList":
ts.nilKind = List
}
[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 12:16:53 +03:00
case "optional":
ts.optional = true
if ts.tail {
return ts, structTagError{typ, f.Name, t, `also has "tail" tag`}
}
2015-12-21 22:05:20 +02:00
case "tail":
ts.tail = true
if fi != lastPublic {
rlp: improve nil pointer handling (#20064) * rlp: improve nil pointer handling In both encoder and decoder, the rules for encoding nil pointers were a bit hard to understand, and didn't leave much choice. Since RLP allows two empty values (empty list, empty string), any protocol built on RLP must choose either of these values to represent the null value in a certain context. This change adds choice in the form of two new struct tags, "nilString" and "nilList". These can be used to specify how a nil pointer value is encoded. The "nil" tag still exists, but its implementation is now explicit and defines exactly how nil pointers are handled in a single place. Another important change in this commit is how nil pointers and the Encoder interface interact. The EncodeRLP method was previously called even on nil values, which was supposed to give users a choice of how their value would be handled when nil. It turns out this is a stupid idea. If you create a network protocol containing an object defined in another package, it's better to be able to say that the object should be a list or string when nil in the definition of the protocol message rather than defining the encoding of nil on the object itself. As of this commit, the encoding rules for pointers now take precedence over the Encoder interface rule. I think the "nil" tag will work fine for most cases. For special kinds of objects which are a struct in Go but strings in RLP, code using the object can specify the desired encoding of nil using the "nilString" and "nilList" tags. * rlp: propagate struct field type errors If a struct contained fields of undecodable type, the encoder and decoder would panic instead of returning an error. Fix this by propagating type errors in makeStruct{Writer,Decoder} and add a test.
2019-09-13 12:10:57 +03:00
return ts, structTagError{typ, f.Name, t, "must be on last field"}
2015-12-21 22:05:20 +02:00
}
[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 12:16:53 +03:00
if ts.optional {
return ts, structTagError{typ, f.Name, t, `also has "optional" tag`}
}
2015-12-21 22:05:20 +02:00
if f.Type.Kind() != reflect.Slice {
rlp: improve nil pointer handling (#20064) * rlp: improve nil pointer handling In both encoder and decoder, the rules for encoding nil pointers were a bit hard to understand, and didn't leave much choice. Since RLP allows two empty values (empty list, empty string), any protocol built on RLP must choose either of these values to represent the null value in a certain context. This change adds choice in the form of two new struct tags, "nilString" and "nilList". These can be used to specify how a nil pointer value is encoded. The "nil" tag still exists, but its implementation is now explicit and defines exactly how nil pointers are handled in a single place. Another important change in this commit is how nil pointers and the Encoder interface interact. The EncodeRLP method was previously called even on nil values, which was supposed to give users a choice of how their value would be handled when nil. It turns out this is a stupid idea. If you create a network protocol containing an object defined in another package, it's better to be able to say that the object should be a list or string when nil in the definition of the protocol message rather than defining the encoding of nil on the object itself. As of this commit, the encoding rules for pointers now take precedence over the Encoder interface rule. I think the "nil" tag will work fine for most cases. For special kinds of objects which are a struct in Go but strings in RLP, code using the object can specify the desired encoding of nil using the "nilString" and "nilList" tags. * rlp: propagate struct field type errors If a struct contained fields of undecodable type, the encoder and decoder would panic instead of returning an error. Fix this by propagating type errors in makeStruct{Writer,Decoder} and add a test.
2019-09-13 12:10:57 +03:00
return ts, structTagError{typ, f.Name, t, "field type is not slice"}
2015-12-21 22:05:20 +02:00
}
default:
return ts, fmt.Errorf("rlp: unknown struct tag %q on %v.%s", t, typ, f.Name)
}
}
return ts, nil
}
func lastPublicField(typ reflect.Type) int {
last := 0
for i := 0; i < typ.NumField(); i++ {
if typ.Field(i).PkgPath == "" {
last = i
}
}
return last
}
func (i *typeinfo) generate(typ reflect.Type, tags tags) {
i.decoder, i.decoderErr = makeDecoder(typ, tags)
i.writer, i.writerErr = makeWriter(typ, tags)
}
rlp: improve nil pointer handling (#20064) * rlp: improve nil pointer handling In both encoder and decoder, the rules for encoding nil pointers were a bit hard to understand, and didn't leave much choice. Since RLP allows two empty values (empty list, empty string), any protocol built on RLP must choose either of these values to represent the null value in a certain context. This change adds choice in the form of two new struct tags, "nilString" and "nilList". These can be used to specify how a nil pointer value is encoded. The "nil" tag still exists, but its implementation is now explicit and defines exactly how nil pointers are handled in a single place. Another important change in this commit is how nil pointers and the Encoder interface interact. The EncodeRLP method was previously called even on nil values, which was supposed to give users a choice of how their value would be handled when nil. It turns out this is a stupid idea. If you create a network protocol containing an object defined in another package, it's better to be able to say that the object should be a list or string when nil in the definition of the protocol message rather than defining the encoding of nil on the object itself. As of this commit, the encoding rules for pointers now take precedence over the Encoder interface rule. I think the "nil" tag will work fine for most cases. For special kinds of objects which are a struct in Go but strings in RLP, code using the object can specify the desired encoding of nil using the "nilString" and "nilList" tags. * rlp: propagate struct field type errors If a struct contained fields of undecodable type, the encoder and decoder would panic instead of returning an error. Fix this by propagating type errors in makeStruct{Writer,Decoder} and add a test.
2019-09-13 12:10:57 +03:00
// defaultNilKind determines whether a nil pointer to typ encodes/decodes
// as an empty string or empty list.
func defaultNilKind(typ reflect.Type) Kind {
k := typ.Kind()
if isUint(k) || k == reflect.String || k == reflect.Bool || isByteArray(typ) {
return String
}
return List
}
func isUint(k reflect.Kind) bool {
return k >= reflect.Uint && k <= reflect.Uintptr
}
rlp: improve nil pointer handling (#20064) * rlp: improve nil pointer handling In both encoder and decoder, the rules for encoding nil pointers were a bit hard to understand, and didn't leave much choice. Since RLP allows two empty values (empty list, empty string), any protocol built on RLP must choose either of these values to represent the null value in a certain context. This change adds choice in the form of two new struct tags, "nilString" and "nilList". These can be used to specify how a nil pointer value is encoded. The "nil" tag still exists, but its implementation is now explicit and defines exactly how nil pointers are handled in a single place. Another important change in this commit is how nil pointers and the Encoder interface interact. The EncodeRLP method was previously called even on nil values, which was supposed to give users a choice of how their value would be handled when nil. It turns out this is a stupid idea. If you create a network protocol containing an object defined in another package, it's better to be able to say that the object should be a list or string when nil in the definition of the protocol message rather than defining the encoding of nil on the object itself. As of this commit, the encoding rules for pointers now take precedence over the Encoder interface rule. I think the "nil" tag will work fine for most cases. For special kinds of objects which are a struct in Go but strings in RLP, code using the object can specify the desired encoding of nil using the "nilString" and "nilList" tags. * rlp: propagate struct field type errors If a struct contained fields of undecodable type, the encoder and decoder would panic instead of returning an error. Fix this by propagating type errors in makeStruct{Writer,Decoder} and add a test.
2019-09-13 12:10:57 +03:00
func isByte(typ reflect.Type) bool {
return typ.Kind() == reflect.Uint8 && !typ.Implements(encoderInterface)
}
rlp: improve nil pointer handling (#20064) * rlp: improve nil pointer handling In both encoder and decoder, the rules for encoding nil pointers were a bit hard to understand, and didn't leave much choice. Since RLP allows two empty values (empty list, empty string), any protocol built on RLP must choose either of these values to represent the null value in a certain context. This change adds choice in the form of two new struct tags, "nilString" and "nilList". These can be used to specify how a nil pointer value is encoded. The "nil" tag still exists, but its implementation is now explicit and defines exactly how nil pointers are handled in a single place. Another important change in this commit is how nil pointers and the Encoder interface interact. The EncodeRLP method was previously called even on nil values, which was supposed to give users a choice of how their value would be handled when nil. It turns out this is a stupid idea. If you create a network protocol containing an object defined in another package, it's better to be able to say that the object should be a list or string when nil in the definition of the protocol message rather than defining the encoding of nil on the object itself. As of this commit, the encoding rules for pointers now take precedence over the Encoder interface rule. I think the "nil" tag will work fine for most cases. For special kinds of objects which are a struct in Go but strings in RLP, code using the object can specify the desired encoding of nil using the "nilString" and "nilList" tags. * rlp: propagate struct field type errors If a struct contained fields of undecodable type, the encoder and decoder would panic instead of returning an error. Fix this by propagating type errors in makeStruct{Writer,Decoder} and add a test.
2019-09-13 12:10:57 +03:00
func isByteArray(typ reflect.Type) bool {
return (typ.Kind() == reflect.Slice || typ.Kind() == reflect.Array) && isByte(typ.Elem())
}