2010-08-29 18:58:15 +02:00
// Copyright (c) 2009-2010 Satoshi Nakamoto
2023-08-16 19:27:31 +02:00
// Copyright (c) 2009-2020 The Bitcoin Core developers
2014-10-31 04:34:30 +01:00
// Distributed under the MIT software license, see the accompanying
2012-05-18 16:02:28 +02:00
// file COPYING or http://www.opensource.org/licenses/mit-license.php.
2013-04-13 07:13:08 +02:00
2011-05-15 09:11:04 +02:00
# ifndef BITCOIN_SERIALIZE_H
# define BITCOIN_SERIALIZE_H
2010-08-29 18:58:15 +02:00
2020-03-19 23:46:56 +01:00
# include <compat/endian.h>
2015-03-27 13:19:49 +01:00
2013-04-13 07:13:08 +02:00
# include <algorithm>
2021-09-30 23:00:52 +02:00
# include <atomic>
2022-10-20 11:50:51 +02:00
# include <cstdint>
2020-02-26 19:00:18 +01:00
# include <cstring>
2013-04-13 07:13:08 +02:00
# include <ios>
2014-09-14 12:43:56 +02:00
# include <limits>
2016-11-13 18:52:34 +01:00
# include <list>
2010-08-29 18:58:15 +02:00
# include <map>
2016-11-21 10:51:32 +01:00
# include <memory>
2010-09-06 23:03:04 +02:00
# include <set>
2013-04-13 07:13:08 +02:00
# include <string>
# include <string.h>
Collection of minor performance optimizations (#2855)
* Merge #13176: Improve CRollingBloomFilter performance: replace modulus with FastMod
9aac9f90d5e56752cc6cbfac48063ad29a01143c replace modulus with FastMod (Martin Ankerl)
Pull request description:
Not sure if this is optimization is necessary, but anyway I have some spare time so here it is. This replaces the slow modulo operation with a much faster 64bit multiplication & shift. This works when the hash is uniformly distributed between 0 and 2^32-1. This speeds up the benchmark by a factor of about 1.3:
```
RollingBloom, 5, 1500000, 3.73733, 4.97569e-07, 4.99002e-07, 4.98372e-07 # before
RollingBloom, 5, 1500000, 2.86842, 3.81630e-07, 3.83730e-07, 3.82473e-07 # FastMod
```
Be aware that this changes the internal data of the filter, so this should probably
not be used for CBloomFilter because of interoperability problems.
Tree-SHA512: 04104f3fb09f56c9d14458a6aad919aeb0a5af944e8ee6a31f00e93c753e22004648c1cd65bf36752b6addec528d19fb665c27b955ce1666a85a928e17afa47a
* Use unordered_map in CSporkManager
In one of my profiling sessions with many InstantSend transactions
happening, calls into CSporkManager added up to about 1% of total CPU time.
This is easily avoidable by using unordered maps.
* Use std::unordered_map instead of std::map in limitedmap
* Use unordered_set for CNode::setAskFor
* Add serialization support for unordered maps and sets
* Use unordered_map for mapArgs and mapMultiArgs
* Let limitedmap prune in batches and use unordered_multimap
Due to the batched pruning, there is no need to maintain an ordered map
of values anymore. Only when nPruneAfterSize, there is a need to create
a temporary ordered vector of values to figure out what can be removed.
* Instead of using a multimap for mapAskFor, use a vector which we sort on demand
CNode::AskFor will now push entries into an initially unordered vector
instead of an ordered multimap. Only when we later want to use vecAskFor in
SendMessages, we sort the vector.
The vector will actually be mostly sorted in most cases as insertion order
usually mimics the desired ordering. Only the last few entries might need
some shuffling around. Doing the sort on-demand should be less wasteful
then trying to maintain correct order all the time.
* Fix compilation of tests
* Fix limitedmap tests
* Rename limitedmap to unordered_limitedmap to ensure backports conflict
This ensures that future backports that depends on limitedmap's ordering
conflict so that we are made aware of needed action.
* Fix compilation error on Travis
2019-04-11 14:42:14 +02:00
# include <unordered_map>
# include <unordered_set>
2013-04-13 07:13:08 +02:00
# include <utility>
# include <vector>
2011-05-15 09:11:04 +02:00
2020-03-19 23:46:56 +01:00
# include <prevector.h>
2018-04-08 17:57:22 +02:00
# include <span.h>
2013-04-13 07:13:08 +02:00
2021-05-29 22:24:52 +02:00
/**
* The maximum size of a serialized object in bytes or number of elements
* ( for eg vectors ) when the size is encoded as CompactSize .
*/
static constexpr uint64_t MAX_SIZE = 0x02000000 ;
2010-08-29 18:58:15 +02:00
2017-09-06 20:25:03 +02:00
/** Maximum amount of memory (in bytes) to allocate at once when deserializing vectors. */
static const unsigned int MAX_VECTOR_ALLOCATE = 5000000 ;
2016-11-21 10:51:32 +01:00
/**
* Dummy data type to identify deserializing constructors .
*
* By convention , a constructor of a type T with signature
*
* template < typename Stream > T : : T ( deserialize_type , Stream & s )
*
* is a deserializing constructor , which builds the type by
* deserializing it from s . If T contains const fields , this
* is likely the only way to do so .
*/
struct deserialize_type { } ;
constexpr deserialize_type deserialize { } ;
2014-12-19 11:41:50 +01:00
/*
* Lowest - level serialization and conversion .
*/
template < typename Stream > inline void ser_writedata8 ( Stream & s , uint8_t obj )
{
2024-02-24 08:36:25 +01:00
s . write ( AsBytes ( Span { & obj , 1 } ) ) ;
2014-12-19 11:41:50 +01:00
}
template < typename Stream > inline void ser_writedata16 ( Stream & s , uint16_t obj )
{
obj = htole16 ( obj ) ;
2024-02-24 08:36:25 +01:00
s . write ( AsBytes ( Span { & obj , 1 } ) ) ;
2014-12-19 11:41:50 +01:00
}
2020-12-17 10:02:44 +01:00
template < typename Stream > inline void ser_writedata16be ( Stream & s , uint16_t obj )
{
obj = htobe16 ( obj ) ;
2024-02-24 08:36:25 +01:00
s . write ( AsBytes ( Span { & obj , 1 } ) ) ;
2020-12-17 10:02:44 +01:00
}
2014-12-19 11:41:50 +01:00
template < typename Stream > inline void ser_writedata32 ( Stream & s , uint32_t obj )
{
obj = htole32 ( obj ) ;
2024-02-24 08:36:25 +01:00
s . write ( AsBytes ( Span { & obj , 1 } ) ) ;
2014-12-19 11:41:50 +01:00
}
2016-03-16 19:00:50 +01:00
template < typename Stream > inline void ser_writedata32be ( Stream & s , uint32_t obj )
{
obj = htobe32 ( obj ) ;
2024-02-24 08:36:25 +01:00
s . write ( AsBytes ( Span { & obj , 1 } ) ) ;
2016-03-16 19:00:50 +01:00
}
2014-12-19 11:41:50 +01:00
template < typename Stream > inline void ser_writedata64 ( Stream & s , uint64_t obj )
{
obj = htole64 ( obj ) ;
2024-02-24 08:36:25 +01:00
s . write ( AsBytes ( Span { & obj , 1 } ) ) ;
2014-12-19 11:41:50 +01:00
}
template < typename Stream > inline uint8_t ser_readdata8 ( Stream & s )
{
uint8_t obj ;
2024-02-24 08:36:25 +01:00
s . read ( AsWritableBytes ( Span { & obj , 1 } ) ) ;
2014-12-19 11:41:50 +01:00
return obj ;
}
template < typename Stream > inline uint16_t ser_readdata16 ( Stream & s )
{
uint16_t obj ;
2024-02-24 08:36:25 +01:00
s . read ( AsWritableBytes ( Span { & obj , 1 } ) ) ;
2014-12-19 11:41:50 +01:00
return le16toh ( obj ) ;
}
2020-12-17 10:02:44 +01:00
template < typename Stream > inline uint16_t ser_readdata16be ( Stream & s )
{
uint16_t obj ;
2024-02-24 08:36:25 +01:00
s . read ( AsWritableBytes ( Span { & obj , 1 } ) ) ;
2020-12-17 10:02:44 +01:00
return be16toh ( obj ) ;
}
2014-12-19 11:41:50 +01:00
template < typename Stream > inline uint32_t ser_readdata32 ( Stream & s )
{
uint32_t obj ;
2024-02-24 08:36:25 +01:00
s . read ( AsWritableBytes ( Span { & obj , 1 } ) ) ;
2014-12-19 11:41:50 +01:00
return le32toh ( obj ) ;
}
2016-03-16 19:00:50 +01:00
template < typename Stream > inline uint32_t ser_readdata32be ( Stream & s )
{
uint32_t obj ;
2024-02-24 08:36:25 +01:00
s . read ( AsWritableBytes ( Span { & obj , 1 } ) ) ;
2016-03-16 19:00:50 +01:00
return be32toh ( obj ) ;
}
2014-12-19 11:41:50 +01:00
template < typename Stream > inline uint64_t ser_readdata64 ( Stream & s )
{
uint64_t obj ;
2024-02-24 08:36:25 +01:00
s . read ( AsWritableBytes ( Span { & obj , 1 } ) ) ;
2014-12-19 11:41:50 +01:00
return le64toh ( obj ) ;
}
2010-08-29 18:58:15 +02:00
/////////////////////////////////////////////////////////////////
//
// Templates for serializing to anything that looks like a stream,
2024-02-24 08:36:25 +01:00
// i.e. anything that supports .read(Span<std::byte>) and .write(Span<const std::byte>)
2010-08-29 18:58:15 +02:00
//
2016-11-09 12:32:57 +01:00
class CSizeComputer ;
2010-08-29 18:58:15 +02:00
enum
{
// primary actions
SER_NETWORK = ( 1 < < 0 ) ,
SER_DISK = ( 1 < < 1 ) ,
SER_GETHASH = ( 1 < < 2 ) ,
} ;
2018-04-10 19:56:22 +02:00
//! Convert the reference base type to X, without changing constness or reference type.
template < typename X > X & ReadWriteAsHelper ( X & x ) { return x ; }
template < typename X > const X & ReadWriteAsHelper ( const X & x ) { return x ; }
2020-12-17 03:07:07 +01:00
# define READWRITE(...) (::SerReadWriteMany(s, ser_action, __VA_ARGS__))
2018-04-10 19:56:22 +02:00
# define READWRITEAS(type, obj) (::SerReadWriteMany(s, ser_action, ReadWriteAsHelper<type>(obj)))
2020-05-20 13:30:21 +02:00
# define SER_READ(obj, code) ::SerRead(s, ser_action, obj, [&](Stream& s, typename std::remove_const<Type>::type& obj) { code; })
# define SER_WRITE(obj, code) ::SerWrite(s, ser_action, obj, [&](Stream& s, const Type& obj) { code; })
2010-08-29 18:58:15 +02:00
2020-01-18 16:32:31 +01:00
/**
* Implement the Ser and Unser methods needed for implementing a formatter ( see Using below ) .
*
* Both Ser and Unser are delegated to a single static method SerializationOps , which is polymorphic
* in the serialized / deserialized type ( allowing it to be const when serializing , and non - const when
* deserializing ) .
*
* Example use :
* struct FooFormatter {
* FORMATTER_METHODS ( Class , obj ) { READWRITE ( obj . val1 , VARINT ( obj . val2 ) ) ; }
* }
* would define a class FooFormatter that defines a serialization of Class objects consisting
* of serializing its val1 member using the default serialization , and its val2 member using
* VARINT serialization . That FooFormatter can then be used in statements like
* READWRITE ( Using < FooFormatter > ( obj . bla ) ) .
*/
# define FORMATTER_METHODS(cls, obj) \
template < typename Stream > \
static void Ser ( Stream & s , const cls & obj ) { SerializationOps ( obj , s , CSerActionSerialize ( ) ) ; } \
template < typename Stream > \
static void Unser ( Stream & s , cls & obj ) { SerializationOps ( obj , s , CSerActionUnserialize ( ) ) ; } \
template < typename Stream , typename Type , typename Operation > \
static inline void SerializationOps ( Type & obj , Stream & s , Operation ser_action ) \
2017-07-08 00:48:13 +02:00
/**
* Implement the Serialize and Unserialize methods by delegating to a single templated
* static method that takes the to - be - ( de ) serialized object as a parameter . This approach
* has the advantage that the constness of the object becomes a template parameter , and
* thus allows a single implementation that sees the object as const for serializing
* and non - const for deserializing , without casts .
*/
# define SERIALIZE_METHODS(cls, obj) \
template < typename Stream > \
void Serialize ( Stream & s ) const \
{ \
static_assert ( std : : is_same < const cls & , decltype ( * this ) > : : value , " Serialize type mismatch " ) ; \
2020-01-18 16:32:31 +01:00
Ser ( s , * this ) ; \
2017-07-08 00:48:13 +02:00
} \
template < typename Stream > \
void Unserialize ( Stream & s ) \
{ \
static_assert ( std : : is_same < cls & , decltype ( * this ) > : : value , " Unserialize type mismatch " ) ; \
2020-01-18 16:32:31 +01:00
Unser ( s , * this ) ; \
2017-07-08 00:48:13 +02:00
} \
2020-01-18 16:32:31 +01:00
FORMATTER_METHODS ( cls , obj )
2017-07-08 00:48:13 +02:00
2023-06-21 10:13:08 +02:00
// clang-format off
2018-07-05 13:21:37 +02:00
# ifndef CHAR_EQUALS_INT8
2024-02-24 08:36:25 +01:00
template < typename Stream > void Serialize ( Stream & , char ) = delete ; // char serialization forbidden. Use uint8_t or int8_t
2018-07-05 13:21:37 +02:00
# endif
2023-06-30 11:30:59 +02:00
template < typename Stream > void Serialize ( Stream & s , std : : byte a ) { ser_writedata8 ( s , uint8_t ( a ) ) ; }
2016-11-09 12:32:57 +01:00
template < typename Stream > inline void Serialize ( Stream & s , int8_t a ) { ser_writedata8 ( s , a ) ; }
template < typename Stream > inline void Serialize ( Stream & s , uint8_t a ) { ser_writedata8 ( s , a ) ; }
template < typename Stream > inline void Serialize ( Stream & s , int16_t a ) { ser_writedata16 ( s , a ) ; }
template < typename Stream > inline void Serialize ( Stream & s , uint16_t a ) { ser_writedata16 ( s , a ) ; }
template < typename Stream > inline void Serialize ( Stream & s , int32_t a ) { ser_writedata32 ( s , a ) ; }
template < typename Stream > inline void Serialize ( Stream & s , uint32_t a ) { ser_writedata32 ( s , a ) ; }
template < typename Stream > inline void Serialize ( Stream & s , int64_t a ) { ser_writedata64 ( s , a ) ; }
template < typename Stream > inline void Serialize ( Stream & s , uint64_t a ) { ser_writedata64 ( s , a ) ; }
2024-02-24 08:36:25 +01:00
template < typename Stream , int N > inline void Serialize ( Stream & s , const char ( & a ) [ N ] ) { s . write ( MakeByteSpan ( a ) ) ; }
template < typename Stream , int N > inline void Serialize ( Stream & s , const unsigned char ( & a ) [ N ] ) { s . write ( MakeByteSpan ( a ) ) ; }
2023-06-21 10:13:08 +02:00
template < typename Stream , typename B > void Serialize ( Stream & s , Span < B > span ) { ( void ) /* force byte-type */ UCharCast ( span . data ( ) ) ; s . write ( AsBytes ( span ) ) ; }
2016-11-09 12:32:57 +01:00
2018-07-05 13:21:37 +02:00
# ifndef CHAR_EQUALS_INT8
2024-02-24 08:36:25 +01:00
template < typename Stream > void Unserialize ( Stream & , char ) = delete ; // char serialization forbidden. Use uint8_t or int8_t
2018-07-05 13:21:37 +02:00
# endif
2023-06-30 11:30:59 +02:00
template < typename Stream > void Unserialize ( Stream & s , std : : byte & a ) { a = std : : byte { ser_readdata8 ( s ) } ; }
2016-11-09 12:32:57 +01:00
template < typename Stream > inline void Unserialize ( Stream & s , int8_t & a ) { a = ser_readdata8 ( s ) ; }
template < typename Stream > inline void Unserialize ( Stream & s , uint8_t & a ) { a = ser_readdata8 ( s ) ; }
template < typename Stream > inline void Unserialize ( Stream & s , int16_t & a ) { a = ser_readdata16 ( s ) ; }
template < typename Stream > inline void Unserialize ( Stream & s , uint16_t & a ) { a = ser_readdata16 ( s ) ; }
template < typename Stream > inline void Unserialize ( Stream & s , int32_t & a ) { a = ser_readdata32 ( s ) ; }
template < typename Stream > inline void Unserialize ( Stream & s , uint32_t & a ) { a = ser_readdata32 ( s ) ; }
template < typename Stream > inline void Unserialize ( Stream & s , int64_t & a ) { a = ser_readdata64 ( s ) ; }
template < typename Stream > inline void Unserialize ( Stream & s , uint64_t & a ) { a = ser_readdata64 ( s ) ; }
2024-02-24 08:36:25 +01:00
template < typename Stream , int N > inline void Unserialize ( Stream & s , char ( & a ) [ N ] ) { s . read ( MakeWritableByteSpan ( a ) ) ; }
template < typename Stream , int N > inline void Unserialize ( Stream & s , unsigned char ( & a ) [ N ] ) { s . read ( MakeWritableByteSpan ( a ) ) ; }
2023-06-21 10:13:08 +02:00
template < typename Stream , typename B > void Unserialize ( Stream & s , Span < B > span ) { ( void ) /* force byte-type */ UCharCast ( span . data ( ) ) ; s . read ( AsWritableBytes ( span ) ) ; }
2016-11-09 12:32:57 +01:00
2021-05-31 14:57:32 +02:00
template < typename Stream > inline void Serialize ( Stream & s , bool a ) { uint8_t f = a ; ser_writedata8 ( s , f ) ; }
template < typename Stream > inline void Unserialize ( Stream & s , bool & a ) { uint8_t f = ser_readdata8 ( s ) ; a = f ; }
2023-06-21 10:13:08 +02:00
// clang-format on
2010-08-29 18:58:15 +02:00
2014-10-31 04:34:30 +01:00
/**
* Compact Size
* size < 253 - - 1 byte
* size < = USHRT_MAX - - 3 bytes ( 253 + 2 bytes )
* size < = UINT_MAX - - 5 bytes ( 254 + 4 bytes )
* size > UINT_MAX - - 9 bytes ( 255 + 8 bytes )
*/
2013-04-13 07:13:08 +02:00
inline unsigned int GetSizeOfCompactSize ( uint64_t nSize )
2010-08-29 18:58:15 +02:00
{
2010-09-30 18:23:07 +02:00
if ( nSize < 253 ) return sizeof ( unsigned char ) ;
2022-10-20 11:50:51 +02:00
else if ( nSize < = std : : numeric_limits < uint16_t > : : max ( ) ) return sizeof ( unsigned char ) + sizeof ( uint16_t ) ;
2011-12-19 23:08:25 +01:00
else if ( nSize < = std : : numeric_limits < unsigned int > : : max ( ) ) return sizeof ( unsigned char ) + sizeof ( unsigned int ) ;
2013-04-13 07:13:08 +02:00
else return sizeof ( unsigned char ) + sizeof ( uint64_t ) ;
2010-08-29 18:58:15 +02:00
}
2016-11-09 12:32:57 +01:00
inline void WriteCompactSize ( CSizeComputer & os , uint64_t nSize ) ;
2010-08-29 18:58:15 +02:00
template < typename Stream >
2013-04-13 07:13:08 +02:00
void WriteCompactSize ( Stream & os , uint64_t nSize )
2010-08-29 18:58:15 +02:00
{
2010-09-30 18:23:07 +02:00
if ( nSize < 253 )
2010-08-29 18:58:15 +02:00
{
2014-12-19 11:41:50 +01:00
ser_writedata8 ( os , nSize ) ;
2010-08-29 18:58:15 +02:00
}
2022-10-20 11:50:51 +02:00
else if ( nSize < = std : : numeric_limits < uint16_t > : : max ( ) )
2010-08-29 18:58:15 +02:00
{
2014-12-19 11:41:50 +01:00
ser_writedata8 ( os , 253 ) ;
ser_writedata16 ( os , nSize ) ;
2010-08-29 18:58:15 +02:00
}
2011-12-19 23:08:25 +01:00
else if ( nSize < = std : : numeric_limits < unsigned int > : : max ( ) )
2010-08-29 18:58:15 +02:00
{
2014-12-19 11:41:50 +01:00
ser_writedata8 ( os , 254 ) ;
ser_writedata32 ( os , nSize ) ;
2010-08-29 18:58:15 +02:00
}
else
{
2014-12-19 11:41:50 +01:00
ser_writedata8 ( os , 255 ) ;
ser_writedata64 ( os , nSize ) ;
2010-08-29 18:58:15 +02:00
}
return ;
}
2021-05-29 22:24:52 +02:00
/**
* Decode a CompactSize - encoded variable - length integer .
*
* As these are primarily used to encode the size of vector - like serializations , by default a range
* check is performed . When used as a generic number encoding , range_check should be set to false .
*/
2010-08-29 18:58:15 +02:00
template < typename Stream >
2021-05-29 22:24:52 +02:00
uint64_t ReadCompactSize ( Stream & is , bool range_check = true )
2010-08-29 18:58:15 +02:00
{
2014-12-19 11:41:50 +01:00
uint8_t chSize = ser_readdata8 ( is ) ;
2013-04-13 07:13:08 +02:00
uint64_t nSizeRet = 0 ;
2010-09-30 18:23:07 +02:00
if ( chSize < 253 )
2010-08-29 18:58:15 +02:00
{
nSizeRet = chSize ;
}
2010-09-30 18:23:07 +02:00
else if ( chSize = = 253 )
2010-08-29 18:58:15 +02:00
{
2014-12-19 11:41:50 +01:00
nSizeRet = ser_readdata16 ( is ) ;
Reject non-canonically-encoded sizes
The length of vectors, maps, sets, etc are serialized using
Write/ReadCompactSize -- which, unfortunately, do not use a
unique encoding.
So deserializing and then re-serializing a transaction (for example)
can give you different bits than you started with. That doesn't
cause any problems that we are aware of, but it is exactly the type
of subtle mismatch that can lead to exploits.
With this pull, reading a non-canonical CompactSize throws an
exception, which means nodes will ignore 'tx' or 'block' or
other messages that are not properly encoded.
Please check my logic... but this change is safe with respect to
causing a network split. Old clients that receive
non-canonically-encoded transactions or blocks deserialize
them into CTransaction/CBlock structures in memory, and then
re-serialize them before relaying them to peers.
And please check my logic with respect to causing a blockchain
split: there are no CompactSize fields in the block header, so
the block hash is always canonical. The merkle root in the block
header is computed on a vector<CTransaction>, so
any non-canonical encoding of the transactions in 'tx' or 'block'
messages is erased as they are read into memory by old clients,
and does not affect the block hash. And, as noted above, old
clients re-serialize (with canonical encoding) 'tx' and 'block'
messages before relaying to peers.
2013-08-07 04:21:34 +02:00
if ( nSizeRet < 253 )
throw std : : ios_base : : failure ( " non-canonical ReadCompactSize() " ) ;
2010-08-29 18:58:15 +02:00
}
2010-09-30 18:23:07 +02:00
else if ( chSize = = 254 )
2010-08-29 18:58:15 +02:00
{
2014-12-19 11:41:50 +01:00
nSizeRet = ser_readdata32 ( is ) ;
Reject non-canonically-encoded sizes
The length of vectors, maps, sets, etc are serialized using
Write/ReadCompactSize -- which, unfortunately, do not use a
unique encoding.
So deserializing and then re-serializing a transaction (for example)
can give you different bits than you started with. That doesn't
cause any problems that we are aware of, but it is exactly the type
of subtle mismatch that can lead to exploits.
With this pull, reading a non-canonical CompactSize throws an
exception, which means nodes will ignore 'tx' or 'block' or
other messages that are not properly encoded.
Please check my logic... but this change is safe with respect to
causing a network split. Old clients that receive
non-canonically-encoded transactions or blocks deserialize
them into CTransaction/CBlock structures in memory, and then
re-serialize them before relaying them to peers.
And please check my logic with respect to causing a blockchain
split: there are no CompactSize fields in the block header, so
the block hash is always canonical. The merkle root in the block
header is computed on a vector<CTransaction>, so
any non-canonical encoding of the transactions in 'tx' or 'block'
messages is erased as they are read into memory by old clients,
and does not affect the block hash. And, as noted above, old
clients re-serialize (with canonical encoding) 'tx' and 'block'
messages before relaying to peers.
2013-08-07 04:21:34 +02:00
if ( nSizeRet < 0x10000u )
throw std : : ios_base : : failure ( " non-canonical ReadCompactSize() " ) ;
2010-08-29 18:58:15 +02:00
}
else
{
2014-12-19 11:41:50 +01:00
nSizeRet = ser_readdata64 ( is ) ;
2014-09-29 00:22:44 +02:00
if ( nSizeRet < 0x100000000ULL )
Reject non-canonically-encoded sizes
The length of vectors, maps, sets, etc are serialized using
Write/ReadCompactSize -- which, unfortunately, do not use a
unique encoding.
So deserializing and then re-serializing a transaction (for example)
can give you different bits than you started with. That doesn't
cause any problems that we are aware of, but it is exactly the type
of subtle mismatch that can lead to exploits.
With this pull, reading a non-canonical CompactSize throws an
exception, which means nodes will ignore 'tx' or 'block' or
other messages that are not properly encoded.
Please check my logic... but this change is safe with respect to
causing a network split. Old clients that receive
non-canonically-encoded transactions or blocks deserialize
them into CTransaction/CBlock structures in memory, and then
re-serialize them before relaying them to peers.
And please check my logic with respect to causing a blockchain
split: there are no CompactSize fields in the block header, so
the block hash is always canonical. The merkle root in the block
header is computed on a vector<CTransaction>, so
any non-canonical encoding of the transactions in 'tx' or 'block'
messages is erased as they are read into memory by old clients,
and does not affect the block hash. And, as noted above, old
clients re-serialize (with canonical encoding) 'tx' and 'block'
messages before relaying to peers.
2013-08-07 04:21:34 +02:00
throw std : : ios_base : : failure ( " non-canonical ReadCompactSize() " ) ;
2010-08-29 18:58:15 +02:00
}
2021-05-29 22:24:52 +02:00
if ( range_check & & nSizeRet > MAX_SIZE ) {
2015-01-08 11:44:25 +01:00
throw std : : ios_base : : failure ( " ReadCompactSize(): size too large " ) ;
2021-05-29 22:24:52 +02:00
}
2010-08-29 18:58:15 +02:00
return nSizeRet ;
}
2014-10-31 04:34:30 +01:00
/**
* Variable - length integers : bytes are a MSB base - 128 encoding of the number .
* The high bit in each byte signifies whether another digit follows . To make
* sure the encoding is one - to - one , one is subtracted from all but the last digit .
* Thus , the byte sequence a [ ] with length len , where all but the last byte
* has bit 128 set , encodes the number :
2020-07-29 03:23:12 +02:00
*
2014-10-31 04:34:30 +01:00
* ( a [ len - 1 ] & 0x7F ) + sum ( i = 1. . len - 1 , 128 ^ i * ( ( a [ len - i - 1 ] & 0x7F ) + 1 ) )
2020-07-29 03:23:12 +02:00
*
2014-10-31 04:34:30 +01:00
* Properties :
* * Very small ( 0 - 127 : 1 byte , 128 - 16511 : 2 bytes , 16512 - 2113663 : 3 bytes )
* * Every integer has exactly one encoding
* * Encoding does not depend on size of original integer type
* * No redundancy : every ( infinite ) byte sequence corresponds to a list
* of encoded integers .
2020-07-29 03:23:12 +02:00
*
2014-10-31 04:34:30 +01:00
* 0 : [ 0x00 ] 256 : [ 0x81 0x00 ]
* 1 : [ 0x01 ] 16383 : [ 0xFE 0x7F ]
* 127 : [ 0x7F ] 16384 : [ 0xFF 0x00 ]
2016-08-24 10:41:31 +02:00
* 128 : [ 0x80 0x00 ] 16511 : [ 0xFF 0x7F ]
* 255 : [ 0x80 0x7F ] 65535 : [ 0x82 0xFE 0x7F ]
2014-10-31 04:34:30 +01:00
* 2 ^ 32 : [ 0x8E 0xFE 0xFE 0xFF 0x00 ]
*/
2012-06-15 14:19:11 +02:00
2020-12-17 13:20:31 +01:00
/**
* Mode for encoding VarInts .
*
* Currently there is no support for signed encodings . The default mode will not
* compile with signed values , and the legacy " nonnegative signed " mode will
* accept signed values , but improperly encode and decode them if they are
* negative . In the future , the DEFAULT mode could be extended to support
* negative numbers in a backwards compatible way , and additional modes could be
* added to support different varint formats ( e . g . zigzag encoding ) .
*/
enum class VarIntMode { DEFAULT , NONNEGATIVE_SIGNED } ;
template < VarIntMode Mode , typename I >
struct CheckVarIntMode {
constexpr CheckVarIntMode ( )
{
static_assert ( Mode ! = VarIntMode : : DEFAULT | | std : : is_unsigned < I > : : value , " Unsigned type required with mode DEFAULT. " ) ;
static_assert ( Mode ! = VarIntMode : : NONNEGATIVE_SIGNED | | std : : is_signed < I > : : value , " Signed type required with mode NONNEGATIVE_SIGNED. " ) ;
}
} ;
template < VarIntMode Mode , typename I >
2012-06-15 14:19:11 +02:00
inline unsigned int GetSizeOfVarInt ( I n )
{
2020-12-17 13:20:31 +01:00
CheckVarIntMode < Mode , I > ( ) ;
2012-06-15 14:19:11 +02:00
int nRet = 0 ;
while ( true ) {
nRet + + ;
if ( n < = 0x7F )
break ;
n = ( n > > 7 ) - 1 ;
}
return nRet ;
}
2016-11-09 12:32:57 +01:00
template < typename I >
inline void WriteVarInt ( CSizeComputer & os , I n ) ;
2020-12-17 13:20:31 +01:00
template < typename Stream , VarIntMode Mode , typename I >
2012-06-15 14:19:11 +02:00
void WriteVarInt ( Stream & os , I n )
{
2020-12-17 13:20:31 +01:00
CheckVarIntMode < Mode , I > ( ) ;
2012-06-15 14:19:11 +02:00
unsigned char tmp [ ( sizeof ( n ) * 8 + 6 ) / 7 ] ;
int len = 0 ;
while ( true ) {
tmp [ len ] = ( n & 0x7F ) | ( len ? 0x80 : 0x00 ) ;
if ( n < = 0x7F )
break ;
n = ( n > > 7 ) - 1 ;
len + + ;
}
do {
2014-12-19 11:41:50 +01:00
ser_writedata8 ( os , tmp [ len ] ) ;
2012-06-15 14:19:11 +02:00
} while ( len - - ) ;
}
2010-08-29 18:58:15 +02:00
2020-12-17 13:20:31 +01:00
template < typename Stream , VarIntMode Mode , typename I >
2012-06-15 14:19:11 +02:00
I ReadVarInt ( Stream & is )
{
2020-12-17 13:20:31 +01:00
CheckVarIntMode < Mode , I > ( ) ;
2012-06-15 14:19:11 +02:00
I n = 0 ;
while ( true ) {
2014-12-19 11:41:50 +01:00
unsigned char chData = ser_readdata8 ( is ) ;
2017-04-17 13:46:34 +02:00
if ( n > ( std : : numeric_limits < I > : : max ( ) > > 7 ) ) {
throw std : : ios_base : : failure ( " ReadVarInt(): size too large " ) ;
}
2012-06-15 14:19:11 +02:00
n = ( n < < 7 ) | ( chData & 0x7F ) ;
2017-04-17 13:46:34 +02:00
if ( chData & 0x80 ) {
if ( n = = std : : numeric_limits < I > : : max ( ) ) {
throw std : : ios_base : : failure ( " ReadVarInt(): size too large " ) ;
}
2012-06-15 14:19:11 +02:00
n + + ;
2017-04-17 13:46:34 +02:00
} else {
2012-06-15 14:19:11 +02:00
return n ;
2017-04-17 13:46:34 +02:00
}
2012-06-15 14:19:11 +02:00
}
}
2010-08-29 18:58:15 +02:00
2021-05-27 17:04:17 +02:00
/** TODO: describe FixedBitSet */
inline unsigned int GetSizeOfFixedBitSet ( size_t size )
{
return ( size + 7 ) / 8 ;
}
template < typename Stream >
void WriteFixedBitSet ( Stream & s , const std : : vector < bool > & vec , size_t size )
{
2024-02-24 08:36:25 +01:00
std : : vector < uint8_t > vBytes ( ( size + 7 ) / 8 ) ;
2021-05-27 17:04:17 +02:00
size_t ms = std : : min ( size , vec . size ( ) ) ;
for ( size_t p = 0 ; p < ms ; p + + )
vBytes [ p / 8 ] | = vec [ p ] < < ( p % 8 ) ;
2024-02-24 08:36:25 +01:00
s . write ( AsBytes ( Span { vBytes } ) ) ;
2021-05-27 17:04:17 +02:00
}
template < typename Stream >
void ReadFixedBitSet ( Stream & s , std : : vector < bool > & vec , size_t size )
{
vec . resize ( size ) ;
2024-02-24 08:36:25 +01:00
std : : vector < uint8_t > vBytes ( ( size + 7 ) / 8 ) ;
s . read ( AsWritableBytes ( Span { vBytes } ) ) ;
2021-05-27 17:04:17 +02:00
for ( size_t p = 0 ; p < size ; p + + )
vec [ p ] = ( vBytes [ p / 8 ] & ( 1 < < ( p % 8 ) ) ) ! = 0 ;
if ( vBytes . size ( ) * 8 ! = size ) {
size_t rem = vBytes . size ( ) * 8 - size ;
uint8_t m = ~ ( uint8_t ) ( 0xff > > rem ) ;
if ( vBytes [ vBytes . size ( ) - 1 ] & m ) {
throw std : : ios_base : : failure ( " Out-of-range bits set " ) ;
}
}
}
/**
* Stores a fixed size bitset as a series of VarInts . Each VarInt is an offset from the last entry and the sum of the
* last entry and the offset gives an index into the bitset for a set bit . The series of VarInts ends with a 0.
*/
template < typename Stream >
void WriteFixedVarIntsBitSet ( Stream & s , const std : : vector < bool > & vec , size_t size )
{
int32_t last = - 1 ;
for ( int32_t i = 0 ; i < ( int32_t ) vec . size ( ) ; i + + ) {
if ( vec [ i ] ) {
WriteVarInt < Stream , VarIntMode : : DEFAULT , uint32_t > ( s , ( uint32_t ) ( i - last ) ) ;
last = i ;
}
}
WriteVarInt < Stream , VarIntMode : : DEFAULT , uint32_t > ( s , 0 ) ; // stopper
}
template < typename Stream >
void ReadFixedVarIntsBitSet ( Stream & s , std : : vector < bool > & vec , size_t size )
{
vec . assign ( size , false ) ;
int32_t last = - 1 ;
while ( true ) {
uint32_t offset = ReadVarInt < Stream , VarIntMode : : DEFAULT , uint32_t > ( s ) ;
if ( offset = = 0 ) {
break ;
}
int32_t idx = last + offset ;
2022-02-11 17:15:26 +01:00
if ( idx > = int32_t ( size ) ) {
2021-05-27 17:04:17 +02:00
throw std : : ios_base : : failure ( " out of bounds index " ) ;
}
if ( last ! = - 1 & & idx < = last ) {
throw std : : ios_base : : failure ( " offset overflow " ) ;
}
vec [ idx ] = true ;
last = idx ;
}
}
/**
* Serializes either as a CFixedBitSet or CFixedVarIntsBitSet , depending on which would give a smaller size
*/
typedef std : : pair < std : : vector < bool > , size_t > autobitset_t ;
struct CFixedBitSet
{
const std : : vector < bool > & vec ;
size_t size ;
CFixedBitSet ( const std : : vector < bool > & vecIn , size_t sizeIn ) : vec ( vecIn ) , size ( sizeIn ) { }
template < typename Stream >
void Serialize ( Stream & s ) const { WriteFixedBitSet ( s , vec , size ) ; }
} ;
struct CFixedVarIntsBitSet
{
const std : : vector < bool > & vec ;
size_t size ;
CFixedVarIntsBitSet ( const std : : vector < bool > & vecIn , size_t sizeIn ) : vec ( vecIn ) , size ( sizeIn ) { }
template < typename Stream >
void Serialize ( Stream & s ) const { WriteFixedVarIntsBitSet ( s , vec , vec . size ( ) ) ; }
} ;
2024-02-22 19:01:13 +01:00
/* Forward declaration for WriteAutoBitSet */
template < typename T > size_t GetSerializeSize ( const T & t , int nVersion = 0 ) ;
2021-05-27 17:04:17 +02:00
template < typename Stream >
void WriteAutoBitSet ( Stream & s , const autobitset_t & item )
{
auto & vec = item . first ;
auto & size = item . second ;
assert ( vec . size ( ) = = size ) ;
2018-09-11 09:09:57 +02:00
size_t size1 = : : GetSerializeSize ( CFixedBitSet ( vec , size ) , s . GetVersion ( ) ) ;
size_t size2 = : : GetSerializeSize ( CFixedVarIntsBitSet ( vec , size ) , s . GetVersion ( ) ) ;
2021-05-27 17:04:17 +02:00
assert ( size1 = = GetSizeOfFixedBitSet ( size ) ) ;
if ( size1 < size2 ) {
ser_writedata8 ( s , 0 ) ;
WriteFixedBitSet ( s , vec , vec . size ( ) ) ;
} else {
ser_writedata8 ( s , 1 ) ;
WriteFixedVarIntsBitSet ( s , vec , vec . size ( ) ) ;
}
}
template < typename Stream >
void ReadAutoBitSet ( Stream & s , autobitset_t & item )
{
uint8_t isVarInts = ser_readdata8 ( s ) ;
if ( isVarInts ! = 0 & & isVarInts ! = 1 ) {
throw std : : ios_base : : failure ( " invalid value for isVarInts byte " ) ;
}
auto & vec = item . first ;
auto & size = item . second ;
if ( ! isVarInts ) {
ReadFixedBitSet ( s , vec , size ) ;
} else {
ReadFixedVarIntsBitSet ( s , vec , size ) ;
}
}
2020-01-08 17:56:19 +01:00
/** Simple wrapper class to serialize objects using a formatter; used by Using(). */
template < typename Formatter , typename T >
class Wrapper
{
static_assert ( std : : is_lvalue_reference < T > : : value , " Wrapper needs an lvalue reference type T " ) ;
protected :
T m_object ;
public :
explicit Wrapper ( T obj ) : m_object ( obj ) { }
template < typename Stream > void Serialize ( Stream & s ) const { Formatter ( ) . Ser ( s , m_object ) ; }
template < typename Stream > void Unserialize ( Stream & s ) { Formatter ( ) . Unser ( s , m_object ) ; }
} ;
/** Cause serialization/deserialization of an object to be done using a specified formatter class.
*
* To use this , you need a class Formatter that has public functions Ser ( stream , const object & ) for
* serialization , and Unser ( stream , object & ) for deserialization . Serialization routines ( inside
* READWRITE , or directly with < < and > > operators ) , can then use Using < Formatter > ( object ) .
*
* This works by constructing a Wrapper < Formatter , T > - wrapped version of object , where T is
* const during serialization , and non - const during deserialization , which maintains const
* correctness .
*/
template < typename Formatter , typename T >
static inline Wrapper < Formatter , T & > Using ( T & & t ) { return Wrapper < Formatter , T & > ( t ) ; }
2021-05-27 17:04:17 +02:00
# define DYNBITSET(obj) Using<DynamicBitSetFormatter>(obj)
# define AUTOBITSET(obj) Using<AutoBitSetFormatter>(obj)
2020-02-11 07:51:08 +01:00
# define VARINT_MODE(obj, mode) Using<VarIntFormatter<mode>>(obj)
# define VARINT(obj) Using<VarIntFormatter<VarIntMode::DEFAULT>>(obj)
2021-05-29 22:24:52 +02:00
# define COMPACTSIZE(obj) Using<CompactSizeFormatter<true>>(obj)
2021-05-27 16:53:41 +02:00
# define LIMITED_STRING(obj,n) Using<LimitedStringFormatter<n>>(obj)
2012-03-26 16:48:23 +02:00
2021-05-27 17:04:17 +02:00
/** TODO: describe DynamicBitSet */
struct DynamicBitSetFormatter
2018-09-14 17:53:49 +02:00
{
template < typename Stream >
2021-05-27 17:04:17 +02:00
void Ser ( Stream & s , const std : : vector < bool > & vec ) const
2018-09-14 17:53:49 +02:00
{
WriteCompactSize ( s , vec . size ( ) ) ;
2021-05-27 17:04:17 +02:00
WriteFixedBitSet ( s , vec , vec . size ( ) ) ;
2019-01-16 16:24:49 +01:00
}
template < typename Stream >
2021-05-27 17:04:17 +02:00
void Unser ( Stream & s , std : : vector < bool > & vec )
2019-01-16 16:24:49 +01:00
{
2021-05-27 17:04:17 +02:00
ReadFixedBitSet ( s , vec , ReadCompactSize ( s ) ) ;
2019-01-16 16:24:49 +01:00
}
} ;
/**
* Serializes either as a CFixedBitSet or CFixedVarIntsBitSet , depending on which would give a smaller size
*/
2021-05-27 17:04:17 +02:00
struct AutoBitSetFormatter
2019-01-16 16:24:49 +01:00
{
template < typename Stream >
2021-05-27 17:04:17 +02:00
void Ser ( Stream & s , const autobitset_t & item ) const
2019-01-16 16:24:49 +01:00
{
2021-05-27 17:04:17 +02:00
WriteAutoBitSet ( s , item ) ;
2019-01-16 16:24:49 +01:00
}
template < typename Stream >
2021-05-27 17:04:17 +02:00
void Unser ( Stream & s , autobitset_t & item )
2019-01-16 16:24:49 +01:00
{
2021-05-27 17:04:17 +02:00
ReadAutoBitSet ( s , item ) ;
2019-01-16 16:24:49 +01:00
}
} ;
2020-01-08 17:56:19 +01:00
/** Serialization wrapper class for integers in VarInt format. */
2020-02-11 07:51:08 +01:00
template < VarIntMode Mode >
2020-01-08 17:56:19 +01:00
struct VarIntFormatter
2012-06-15 14:19:11 +02:00
{
2020-01-08 17:56:19 +01:00
template < typename Stream , typename I > void Ser ( Stream & s , I v )
{
WriteVarInt < Stream , Mode , typename std : : remove_cv < I > : : type > ( s , v ) ;
2012-06-15 14:19:11 +02:00
}
2020-01-08 17:56:19 +01:00
template < typename Stream , typename I > void Unser ( Stream & s , I & v )
{
v = ReadVarInt < Stream , Mode , typename std : : remove_cv < I > : : type > ( s ) ;
2012-06-15 14:19:11 +02:00
}
} ;
2020-05-20 13:30:21 +02:00
/** Serialization wrapper class for custom integers and enums.
*
* It permits specifying the serialized size ( 1 to 8 bytes ) and endianness .
*
* Use the big endian mode for values that are stored in memory in native
* byte order , but serialized in big endian notation . This is only intended
* to implement serializers that are compatible with existing formats , and
* its use is not recommended for new data structures .
*/
template < int Bytes , bool BigEndian = false >
2020-02-04 04:49:10 +01:00
struct CustomUintFormatter
{
static_assert ( Bytes > 0 & & Bytes < = 8 , " CustomUintFormatter Bytes out of range " ) ;
static constexpr uint64_t MAX = 0xffffffffffffffff > > ( 8 * ( 8 - Bytes ) ) ;
template < typename Stream , typename I > void Ser ( Stream & s , I v )
{
if ( v < 0 | | v > MAX ) throw std : : ios_base : : failure ( " CustomUintFormatter value out of range " ) ;
2020-05-20 13:30:21 +02:00
if ( BigEndian ) {
uint64_t raw = htobe64 ( v ) ;
2024-02-24 08:36:25 +01:00
s . write ( { BytePtr ( & raw ) + 8 - Bytes , Bytes } ) ;
2020-05-20 13:30:21 +02:00
} else {
uint64_t raw = htole64 ( v ) ;
2024-02-24 08:36:25 +01:00
s . write ( { BytePtr ( & raw ) , Bytes } ) ;
2020-05-20 13:30:21 +02:00
}
2020-02-04 04:49:10 +01:00
}
template < typename Stream , typename I > void Unser ( Stream & s , I & v )
{
2020-05-20 13:30:21 +02:00
using U = typename std : : conditional < std : : is_enum < I > : : value , std : : underlying_type < I > , std : : common_type < I > > : : type : : type ;
static_assert ( std : : numeric_limits < U > : : max ( ) > = MAX & & std : : numeric_limits < U > : : min ( ) < = 0 , " Assigned type too small " ) ;
2020-02-04 04:49:10 +01:00
uint64_t raw = 0 ;
2020-05-20 13:30:21 +02:00
if ( BigEndian ) {
2024-02-24 08:36:25 +01:00
s . read ( { BytePtr ( & raw ) + 8 - Bytes , Bytes } ) ;
2020-05-20 13:30:21 +02:00
v = static_cast < I > ( be64toh ( raw ) ) ;
} else {
2024-02-24 08:36:25 +01:00
s . read ( { BytePtr ( & raw ) , Bytes } ) ;
2020-05-20 13:30:21 +02:00
v = static_cast < I > ( le64toh ( raw ) ) ;
}
2020-02-04 04:49:10 +01:00
}
} ;
2020-05-20 13:30:21 +02:00
template < int Bytes > using BigEndianFormatter = CustomUintFormatter < Bytes , true > ;
2020-12-17 10:02:44 +01:00
2020-02-04 04:49:10 +01:00
/** Formatter for integers in CompactSize format. */
2021-05-29 22:24:52 +02:00
template < bool RangeCheck >
2020-02-04 04:49:10 +01:00
struct CompactSizeFormatter
2016-05-18 22:11:42 +02:00
{
2020-02-04 04:49:10 +01:00
template < typename Stream , typename I >
void Unser ( Stream & s , I & v )
{
2021-05-29 22:24:52 +02:00
uint64_t n = ReadCompactSize < Stream > ( s , RangeCheck ) ;
2020-02-04 04:49:10 +01:00
if ( n < std : : numeric_limits < I > : : min ( ) | | n > std : : numeric_limits < I > : : max ( ) ) {
throw std : : ios_base : : failure ( " CompactSize exceeds limit of type " ) ;
}
v = n ;
2016-05-18 22:11:42 +02:00
}
2020-02-04 04:49:10 +01:00
template < typename Stream , typename I >
void Ser ( Stream & s , I v )
{
static_assert ( std : : is_unsigned < I > : : value , " CompactSize only supported for unsigned integers " ) ;
static_assert ( std : : numeric_limits < I > : : max ( ) < = std : : numeric_limits < uint64_t > : : max ( ) , " CompactSize only supports 64-bit integers and below " ) ;
2016-05-18 22:11:42 +02:00
2020-02-04 04:49:10 +01:00
WriteCompactSize < Stream > ( s , v ) ;
2016-05-18 22:11:42 +02:00
}
} ;
2014-08-07 23:00:01 +02:00
template < size_t Limit >
2021-05-27 16:53:41 +02:00
struct LimitedStringFormatter
2014-08-07 23:00:01 +02:00
{
template < typename Stream >
2021-05-27 16:53:41 +02:00
void Unser ( Stream & s , std : : string & v )
2014-08-07 23:00:01 +02:00
{
size_t size = ReadCompactSize ( s ) ;
if ( size > Limit ) {
throw std : : ios_base : : failure ( " String length limit exceeded " ) ;
}
2021-05-27 16:53:41 +02:00
v . resize ( size ) ;
2024-02-24 08:36:25 +01:00
if ( size ! = 0 ) s . read ( MakeWritableByteSpan ( v ) ) ;
2014-08-07 23:00:01 +02:00
}
template < typename Stream >
2021-05-27 16:53:41 +02:00
void Ser ( Stream & s , const std : : string & v )
2014-08-07 23:00:01 +02:00
{
2021-05-27 16:53:41 +02:00
s < < v ;
2014-08-07 23:00:01 +02:00
}
} ;
2017-09-06 20:25:03 +02:00
/** Formatter to serialize/deserialize vector elements using another formatter
*
* Example :
* struct X {
* std : : vector < uint64_t > v ;
* SERIALIZE_METHODS ( X , obj ) { READWRITE ( Using < VectorFormatter < VarInt > > ( obj . v ) ) ; }
* } ;
* will define a struct that contains a vector of uint64_t , which is serialized
* as a vector of VarInt - encoded integers .
*
* V is not required to be an std : : vector type . It works for any class that
2020-02-04 04:49:10 +01:00
* exposes a value_type , size , reserve , emplace_back , back , and const iterators .
2017-09-06 20:25:03 +02:00
*/
template < class Formatter >
struct VectorFormatter
{
template < typename Stream , typename V >
void Ser ( Stream & s , const V & v )
{
2020-02-04 04:49:10 +01:00
Formatter formatter ;
2017-09-06 20:25:03 +02:00
WriteCompactSize ( s , v . size ( ) ) ;
for ( const typename V : : value_type & elem : v ) {
2020-02-04 04:49:10 +01:00
formatter . Ser ( s , elem ) ;
2017-09-06 20:25:03 +02:00
}
}
template < typename Stream , typename V >
void Unser ( Stream & s , V & v )
{
2020-02-04 04:49:10 +01:00
Formatter formatter ;
2017-09-06 20:25:03 +02:00
v . clear ( ) ;
size_t size = ReadCompactSize ( s ) ;
size_t allocated = 0 ;
while ( allocated < size ) {
// For DoS prevention, do not blindly allocate as much as the stream claims to contain.
// Instead, allocate in 5MiB batches, so that an attacker actually needs to provide
// X MiB of data to make us allocate X+5 Mib.
static_assert ( sizeof ( typename V : : value_type ) < = MAX_VECTOR_ALLOCATE , " Vector element size too large " ) ;
allocated = std : : min ( size , allocated + MAX_VECTOR_ALLOCATE / sizeof ( typename V : : value_type ) ) ;
v . reserve ( allocated ) ;
while ( v . size ( ) < allocated ) {
2020-02-04 04:49:10 +01:00
v . emplace_back ( ) ;
formatter . Unser ( s , v . back ( ) ) ;
2017-09-06 20:25:03 +02:00
}
}
} ;
} ;
2014-10-31 04:34:30 +01:00
/**
* Forward declarations
*/
2010-08-29 18:58:15 +02:00
2014-10-31 04:34:30 +01:00
/**
* string
*/
2016-11-09 12:32:57 +01:00
template < typename Stream , typename C > void Serialize ( Stream & os , const std : : basic_string < C > & str ) ;
template < typename Stream , typename C > void Unserialize ( Stream & is , std : : basic_string < C > & str ) ;
2010-08-29 18:58:15 +02:00
2015-10-29 07:11:24 +01:00
/**
* prevector
* prevectors of unsigned char are a special case and are intended to be serialized as a single opaque blob .
*/
2016-11-09 12:32:57 +01:00
template < typename Stream , unsigned int N , typename T > void Serialize_impl ( Stream & os , const prevector < N , T > & v , const unsigned char & ) ;
template < typename Stream , unsigned int N , typename T , typename V > void Serialize_impl ( Stream & os , const prevector < N , T > & v , const V & ) ;
template < typename Stream , unsigned int N , typename T > inline void Serialize ( Stream & os , const prevector < N , T > & v ) ;
template < typename Stream , unsigned int N , typename T > void Unserialize_impl ( Stream & is , prevector < N , T > & v , const unsigned char & ) ;
template < typename Stream , unsigned int N , typename T , typename V > void Unserialize_impl ( Stream & is , prevector < N , T > & v , const V & ) ;
template < typename Stream , unsigned int N , typename T > inline void Unserialize ( Stream & is , prevector < N , T > & v ) ;
2015-10-29 07:11:24 +01:00
2014-10-31 04:34:30 +01:00
/**
* vector
* vectors of unsigned char are a special case and are intended to be serialized as a single opaque blob .
*/
2016-11-09 12:32:57 +01:00
template < typename Stream , typename T , typename A > void Serialize_impl ( Stream & os , const std : : vector < T , A > & v , const unsigned char & ) ;
2019-08-26 20:32:47 +02:00
template < typename Stream , typename T , typename A > void Serialize_impl ( Stream & os , const std : : vector < T , A > & v , const bool & ) ;
2016-11-09 12:32:57 +01:00
template < typename Stream , typename T , typename A , typename V > void Serialize_impl ( Stream & os , const std : : vector < T , A > & v , const V & ) ;
template < typename Stream , typename T , typename A > inline void Serialize ( Stream & os , const std : : vector < T , A > & v ) ;
template < typename Stream , typename T , typename A > void Unserialize_impl ( Stream & is , std : : vector < T , A > & v , const unsigned char & ) ;
template < typename Stream , typename T , typename A , typename V > void Unserialize_impl ( Stream & is , std : : vector < T , A > & v , const V & ) ;
template < typename Stream , typename T , typename A > inline void Unserialize ( Stream & is , std : : vector < T , A > & v ) ;
2010-08-29 18:58:15 +02:00
2014-10-31 04:34:30 +01:00
/**
* pair
*/
2016-11-09 12:32:57 +01:00
template < typename Stream , typename K , typename T > void Serialize ( Stream & os , const std : : pair < K , T > & item ) ;
template < typename Stream , typename K , typename T > void Unserialize ( Stream & is , std : : pair < K , T > & item ) ;
2010-08-29 18:58:15 +02:00
2018-09-14 17:53:49 +02:00
/**
* pair
*/
template < typename Stream , typename . . . Elements > void Serialize ( Stream & os , const std : : tuple < Elements . . . > & item ) ;
template < typename Stream , typename . . . Elements > void Unserialize ( Stream & is , std : : tuple < Elements . . . > & item ) ;
2014-10-31 04:34:30 +01:00
/**
* map
*/
2016-11-09 12:32:57 +01:00
template < typename Stream , typename K , typename T , typename Pred , typename A > void Serialize ( Stream & os , const std : : map < K , T , Pred , A > & m ) ;
template < typename Stream , typename K , typename T , typename Pred , typename A > void Unserialize ( Stream & is , std : : map < K , T , Pred , A > & m ) ;
Collection of minor performance optimizations (#2855)
* Merge #13176: Improve CRollingBloomFilter performance: replace modulus with FastMod
9aac9f90d5e56752cc6cbfac48063ad29a01143c replace modulus with FastMod (Martin Ankerl)
Pull request description:
Not sure if this is optimization is necessary, but anyway I have some spare time so here it is. This replaces the slow modulo operation with a much faster 64bit multiplication & shift. This works when the hash is uniformly distributed between 0 and 2^32-1. This speeds up the benchmark by a factor of about 1.3:
```
RollingBloom, 5, 1500000, 3.73733, 4.97569e-07, 4.99002e-07, 4.98372e-07 # before
RollingBloom, 5, 1500000, 2.86842, 3.81630e-07, 3.83730e-07, 3.82473e-07 # FastMod
```
Be aware that this changes the internal data of the filter, so this should probably
not be used for CBloomFilter because of interoperability problems.
Tree-SHA512: 04104f3fb09f56c9d14458a6aad919aeb0a5af944e8ee6a31f00e93c753e22004648c1cd65bf36752b6addec528d19fb665c27b955ce1666a85a928e17afa47a
* Use unordered_map in CSporkManager
In one of my profiling sessions with many InstantSend transactions
happening, calls into CSporkManager added up to about 1% of total CPU time.
This is easily avoidable by using unordered maps.
* Use std::unordered_map instead of std::map in limitedmap
* Use unordered_set for CNode::setAskFor
* Add serialization support for unordered maps and sets
* Use unordered_map for mapArgs and mapMultiArgs
* Let limitedmap prune in batches and use unordered_multimap
Due to the batched pruning, there is no need to maintain an ordered map
of values anymore. Only when nPruneAfterSize, there is a need to create
a temporary ordered vector of values to figure out what can be removed.
* Instead of using a multimap for mapAskFor, use a vector which we sort on demand
CNode::AskFor will now push entries into an initially unordered vector
instead of an ordered multimap. Only when we later want to use vecAskFor in
SendMessages, we sort the vector.
The vector will actually be mostly sorted in most cases as insertion order
usually mimics the desired ordering. Only the last few entries might need
some shuffling around. Doing the sort on-demand should be less wasteful
then trying to maintain correct order all the time.
* Fix compilation of tests
* Fix limitedmap tests
* Rename limitedmap to unordered_limitedmap to ensure backports conflict
This ensures that future backports that depends on limitedmap's ordering
conflict so that we are made aware of needed action.
* Fix compilation error on Travis
2019-04-11 14:42:14 +02:00
template < typename Stream , typename K , typename T , typename Hash , typename Pred , typename A > void Serialize ( Stream & os , const std : : unordered_map < K , T , Hash , Pred , A > & m ) ;
template < typename Stream , typename K , typename T , typename Hash , typename Pred , typename A > void Unserialize ( Stream & is , std : : unordered_map < K , T , Hash , Pred , A > & m ) ;
2010-08-29 18:58:15 +02:00
2014-10-31 04:34:30 +01:00
/**
* set
*/
2016-11-09 12:32:57 +01:00
template < typename Stream , typename K , typename Pred , typename A > void Serialize ( Stream & os , const std : : set < K , Pred , A > & m ) ;
template < typename Stream , typename K , typename Pred , typename A > void Unserialize ( Stream & is , std : : set < K , Pred , A > & m ) ;
Collection of minor performance optimizations (#2855)
* Merge #13176: Improve CRollingBloomFilter performance: replace modulus with FastMod
9aac9f90d5e56752cc6cbfac48063ad29a01143c replace modulus with FastMod (Martin Ankerl)
Pull request description:
Not sure if this is optimization is necessary, but anyway I have some spare time so here it is. This replaces the slow modulo operation with a much faster 64bit multiplication & shift. This works when the hash is uniformly distributed between 0 and 2^32-1. This speeds up the benchmark by a factor of about 1.3:
```
RollingBloom, 5, 1500000, 3.73733, 4.97569e-07, 4.99002e-07, 4.98372e-07 # before
RollingBloom, 5, 1500000, 2.86842, 3.81630e-07, 3.83730e-07, 3.82473e-07 # FastMod
```
Be aware that this changes the internal data of the filter, so this should probably
not be used for CBloomFilter because of interoperability problems.
Tree-SHA512: 04104f3fb09f56c9d14458a6aad919aeb0a5af944e8ee6a31f00e93c753e22004648c1cd65bf36752b6addec528d19fb665c27b955ce1666a85a928e17afa47a
* Use unordered_map in CSporkManager
In one of my profiling sessions with many InstantSend transactions
happening, calls into CSporkManager added up to about 1% of total CPU time.
This is easily avoidable by using unordered maps.
* Use std::unordered_map instead of std::map in limitedmap
* Use unordered_set for CNode::setAskFor
* Add serialization support for unordered maps and sets
* Use unordered_map for mapArgs and mapMultiArgs
* Let limitedmap prune in batches and use unordered_multimap
Due to the batched pruning, there is no need to maintain an ordered map
of values anymore. Only when nPruneAfterSize, there is a need to create
a temporary ordered vector of values to figure out what can be removed.
* Instead of using a multimap for mapAskFor, use a vector which we sort on demand
CNode::AskFor will now push entries into an initially unordered vector
instead of an ordered multimap. Only when we later want to use vecAskFor in
SendMessages, we sort the vector.
The vector will actually be mostly sorted in most cases as insertion order
usually mimics the desired ordering. Only the last few entries might need
some shuffling around. Doing the sort on-demand should be less wasteful
then trying to maintain correct order all the time.
* Fix compilation of tests
* Fix limitedmap tests
* Rename limitedmap to unordered_limitedmap to ensure backports conflict
This ensures that future backports that depends on limitedmap's ordering
conflict so that we are made aware of needed action.
* Fix compilation error on Travis
2019-04-11 14:42:14 +02:00
template < typename Stream , typename K , typename Hash , typename Pred , typename A > void Serialize ( Stream & os , const std : : unordered_set < K , Hash , Pred , A > & m ) ;
template < typename Stream , typename K , typename Hash , typename Pred , typename A > void Unserialize ( Stream & is , std : : unordered_set < K , Hash , Pred , A > & m ) ;
2010-08-29 18:58:15 +02:00
2016-11-21 10:51:32 +01:00
/**
* shared_ptr
*/
2019-01-03 21:08:34 +01:00
template < typename Stream , typename T > void Serialize ( Stream & os , const std : : shared_ptr < T > & p ) ;
template < typename Stream , typename T > void Unserialize ( Stream & os , std : : shared_ptr < T > & p ) ;
2010-08-29 18:58:15 +02:00
2016-11-21 10:51:32 +01:00
/**
* unique_ptr
*/
template < typename Stream , typename T > void Serialize ( Stream & os , const std : : unique_ptr < const T > & p ) ;
template < typename Stream , typename T > void Unserialize ( Stream & os , std : : unique_ptr < const T > & p ) ;
2010-08-29 18:58:15 +02:00
2021-09-30 23:00:52 +02:00
/**
* atomic
*/
template < typename Stream , typename T > void Serialize ( Stream & os , const std : : atomic < T > & a ) ;
template < typename Stream , typename T > void Unserialize ( Stream & is , std : : atomic < T > & a ) ;
2010-08-29 18:58:15 +02:00
2014-10-31 04:34:30 +01:00
/**
2019-05-28 15:15:37 +02:00
* If none of the specialized versions above matched and T is a class , default to calling member function .
2014-10-31 04:34:30 +01:00
*/
2019-05-28 15:15:37 +02:00
template < typename Stream , typename T , typename std : : enable_if < std : : is_class < T > : : value > : : type * = nullptr >
2016-11-09 12:32:57 +01:00
inline void Serialize ( Stream & os , const T & a )
2010-08-29 18:58:15 +02:00
{
2016-11-09 12:32:57 +01:00
a . Serialize ( os ) ;
2010-08-29 18:58:15 +02:00
}
2020-12-17 03:07:07 +01:00
template < typename Stream , typename T , typename std : : enable_if < std : : is_class < std : : remove_reference < T > > : : value > : : type * = nullptr >
inline void Unserialize ( Stream & is , T & & a )
2010-08-29 18:58:15 +02:00
{
2016-11-09 12:32:57 +01:00
a . Unserialize ( is ) ;
2010-08-29 18:58:15 +02:00
}
2019-05-28 15:15:37 +02:00
/**
* If none of the specialized versions above matched and T is an enum , default to calling
* Serialize / Unserialze with the underlying type . This is only allowed when a specialized struct of is_serializable_enum < Enum >
* is found which derives from std : : true_type . This is to ensure that enums are not serialized with the wrong type by
* accident .
*/
template < typename T > struct is_serializable_enum ;
template < typename T > struct is_serializable_enum : std : : false_type { } ;
template < typename Stream , typename T , typename std : : enable_if < std : : is_enum < T > : : value > : : type * = nullptr >
2020-12-17 03:07:07 +01:00
inline void Serialize ( Stream & s , const T & a )
2019-05-28 15:15:37 +02:00
{
// If you ever get into this situation, it usaully means you forgot to declare is_serializable_enum for the desired enum type
2019-05-30 10:00:31 +02:00
static_assert ( is_serializable_enum < T > : : value , " Missing declararion of is_serializable_enum " ) ;
2019-05-28 15:15:37 +02:00
typedef typename std : : underlying_type < T > : : type T2 ;
T2 b = ( T2 ) a ;
Serialize ( s , b ) ;
}
template < typename Stream , typename T , typename std : : enable_if < std : : is_enum < T > : : value > : : type * = nullptr >
inline void Unserialize ( Stream & s , T & a )
{
// If you ever get into this situation, it usaully means you forgot to declare is_serializable_enum for the desired enum type
2019-05-30 10:00:31 +02:00
static_assert ( is_serializable_enum < T > : : value , " Missing declararion of is_serializable_enum " ) ;
2019-05-28 15:15:37 +02:00
typedef typename std : : underlying_type < T > : : type T2 ;
T2 b ;
Unserialize ( s , b ) ;
a = ( T ) b ;
}
2010-08-29 18:58:15 +02:00
2017-09-06 20:25:03 +02:00
/** Default formatter. Serializes objects as themselves.
*
* The vector / prevector serialization code passes this to VectorFormatter
* to enable reusing that logic . It shouldn ' t be needed elsewhere .
*/
struct DefaultFormatter
{
template < typename Stream , typename T >
static void Ser ( Stream & s , const T & t ) { Serialize ( s , t ) ; }
2010-08-29 18:58:15 +02:00
2017-09-06 20:25:03 +02:00
template < typename Stream , typename T >
static void Unser ( Stream & s , T & t ) { Unserialize ( s , t ) ; }
} ;
2010-08-29 18:58:15 +02:00
2014-10-31 04:34:30 +01:00
/**
* string
*/
2010-08-29 18:58:15 +02:00
template < typename Stream , typename C >
2016-11-09 12:32:57 +01:00
void Serialize ( Stream & os , const std : : basic_string < C > & str )
2010-08-29 18:58:15 +02:00
{
WriteCompactSize ( os , str . size ( ) ) ;
if ( ! str . empty ( ) )
2024-02-24 08:36:25 +01:00
os . write ( MakeByteSpan ( str ) ) ;
2010-08-29 18:58:15 +02:00
}
template < typename Stream , typename C >
2016-11-09 12:32:57 +01:00
void Unserialize ( Stream & is , std : : basic_string < C > & str )
2010-08-29 18:58:15 +02:00
{
unsigned int nSize = ReadCompactSize ( is ) ;
str . resize ( nSize ) ;
if ( nSize ! = 0 )
2024-02-24 08:36:25 +01:00
is . read ( MakeWritableByteSpan ( str ) ) ;
2010-08-29 18:58:15 +02:00
}
2021-12-17 18:22:11 +01:00
/**
* string_view
*/
template < typename Stream , typename C >
void Serialize ( Stream & os , const std : : basic_string_view < C > & str )
{
WriteCompactSize ( os , str . size ( ) ) ;
if ( ! str . empty ( ) )
2024-02-24 08:36:25 +01:00
os . write ( AsBytes ( Span { str . data ( ) , str . size ( ) * sizeof ( C ) } ) ) ;
2021-12-17 18:22:11 +01:00
}
template < typename Stream , typename C >
void Unserialize ( Stream & is , std : : basic_string_view < C > & str )
{
unsigned int nSize = ReadCompactSize ( is ) ;
str . resize ( nSize ) ;
if ( nSize ! = 0 )
2024-02-24 08:36:25 +01:00
is . read ( AsWritableBytes ( Span { str . data ( ) , nSize * sizeof ( C ) } ) ) ;
2021-12-17 18:22:11 +01:00
}
2010-08-29 18:58:15 +02:00
2015-10-29 07:11:24 +01:00
/**
* prevector
*/
template < typename Stream , unsigned int N , typename T >
2016-11-09 12:32:57 +01:00
void Serialize_impl ( Stream & os , const prevector < N , T > & v , const unsigned char & )
2015-10-29 07:11:24 +01:00
{
WriteCompactSize ( os , v . size ( ) ) ;
if ( ! v . empty ( ) )
2024-02-24 08:36:25 +01:00
os . write ( MakeByteSpan ( v ) ) ;
2015-10-29 07:11:24 +01:00
}
template < typename Stream , unsigned int N , typename T , typename V >
2016-11-09 12:32:57 +01:00
void Serialize_impl ( Stream & os , const prevector < N , T > & v , const V & )
2015-10-29 07:11:24 +01:00
{
2017-09-06 20:25:03 +02:00
Serialize ( os , Using < VectorFormatter < DefaultFormatter > > ( v ) ) ;
2015-10-29 07:11:24 +01:00
}
template < typename Stream , unsigned int N , typename T >
2016-11-09 12:32:57 +01:00
inline void Serialize ( Stream & os , const prevector < N , T > & v )
2015-10-29 07:11:24 +01:00
{
2016-11-09 12:32:57 +01:00
Serialize_impl ( os , v , T ( ) ) ;
2015-10-29 07:11:24 +01:00
}
template < typename Stream , unsigned int N , typename T >
2016-11-09 12:32:57 +01:00
void Unserialize_impl ( Stream & is , prevector < N , T > & v , const unsigned char & )
2015-10-29 07:11:24 +01:00
{
// Limit size per read so bogus size value won't cause out of memory
v . clear ( ) ;
unsigned int nSize = ReadCompactSize ( is ) ;
unsigned int i = 0 ;
while ( i < nSize )
{
unsigned int blk = std : : min ( nSize - i , ( unsigned int ) ( 1 + 4999999 / sizeof ( T ) ) ) ;
Merge #12324: speed up Unserialize_impl for prevector
86b47fa741408b061ab0bda784b8678bfd7dfa88 speed up Unserialize_impl for prevector (Akio Nakamura)
Pull request description:
The unserializer for prevector uses `resize()` for reserve the area, but it's prefer to use `reserve()` because `resize()` have overhead to call its constructor many times.
However, `reserve()` does not change the value of `_size` (a private member of prevector).
This PR make the logic of read from stream to callback function, and prevector handles initilizing new values with that call-back and ajust the value of `_size`.
The changes are as follows:
1. prevector.h
Add a public member function named 'append'.
This function has 2 params, number of elemenst to append and call-back function that initilizing new appended values.
2. serialize.h
In the following two function:
- `Unserialize_impl(Stream& is, prevector<N, T>& v, const unsigned char&)`
- `Unserialize_impl(Stream& is, prevector<N, T>& v, const V&)`
Make a callback function from each original logic of reading values from stream, and call prevector's `append()`.
3. test/prevector_tests.cpp
Add a test for `append()`.
## A benchmark result is following:
[Machine]
MacBook Pro (macOS 10.13.3/i7 2.2GHz/mem 16GB/SSD)
[result]
DeserializeAndCheckBlockTest => 22% faster
DeserializeBlockTest => 29% faster
[before PR]
# Benchmark, evals, iterations, total, min, max, median
DeserializeAndCheckBlockTest, 60, 160, 94.4901, 0.0094644, 0.0104715, 0.0098339
DeserializeBlockTest, 60, 130, 65.0964, 0.00800362, 0.00895134, 0.00824187
[After PR]
# Benchmark, evals, iterations, total, min, max, median
DeserializeAndCheckBlockTest, 60, 160, 77.1597, 0.00767013, 0.00858959, 0.00805757
DeserializeBlockTest, 60, 130, 49.9443, 0.00613926, 0.00691187, 0.00635527
ACKs for top commit:
laanwj:
utACK 86b47fa741408b061ab0bda784b8678bfd7dfa88
Tree-SHA512: 62ea121ccd45a306fefc67485a1b03a853435af762607dae2426a87b15a3033d802c8556e1923727ddd1023a1837d0e5f6720c2c77b38196907e750e15fbb902
2019-06-18 16:52:11 +02:00
v . resize_uninitialized ( i + blk ) ;
2024-02-24 08:36:25 +01:00
is . read ( AsWritableBytes ( Span { & v [ i ] , blk } ) ) ;
2015-10-29 07:11:24 +01:00
i + = blk ;
}
}
template < typename Stream , unsigned int N , typename T , typename V >
2016-11-09 12:32:57 +01:00
void Unserialize_impl ( Stream & is , prevector < N , T > & v , const V & )
2015-10-29 07:11:24 +01:00
{
2017-09-06 20:25:03 +02:00
Unserialize ( is , Using < VectorFormatter < DefaultFormatter > > ( v ) ) ;
2015-10-29 07:11:24 +01:00
}
template < typename Stream , unsigned int N , typename T >
2016-11-09 12:32:57 +01:00
inline void Unserialize ( Stream & is , prevector < N , T > & v )
2015-10-29 07:11:24 +01:00
{
2016-11-09 12:32:57 +01:00
Unserialize_impl ( is , v , T ( ) ) ;
2015-10-29 07:11:24 +01:00
}
2014-10-31 04:34:30 +01:00
/**
* vector
*/
2010-08-29 18:58:15 +02:00
template < typename Stream , typename T , typename A >
2016-11-09 12:32:57 +01:00
void Serialize_impl ( Stream & os , const std : : vector < T , A > & v , const unsigned char & )
2010-08-29 18:58:15 +02:00
{
WriteCompactSize ( os , v . size ( ) ) ;
if ( ! v . empty ( ) )
2024-02-24 08:36:25 +01:00
os . write ( MakeByteSpan ( v ) ) ;
2010-08-29 18:58:15 +02:00
}
2019-08-26 20:32:47 +02:00
template < typename Stream , typename T , typename A >
void Serialize_impl ( Stream & os , const std : : vector < T , A > & v , const bool & )
{
// A special case for std::vector<bool>, as dereferencing
// std::vector<bool>::const_iterator does not result in a const bool&
// due to std::vector's special casing for bool arguments.
WriteCompactSize ( os , v . size ( ) ) ;
for ( bool elem : v ) {
: : Serialize ( os , elem ) ;
}
}
2014-10-08 20:27:07 +02:00
template < typename Stream , typename T , typename A , typename V >
2016-11-09 12:32:57 +01:00
void Serialize_impl ( Stream & os , const std : : vector < T , A > & v , const V & )
2010-08-29 18:58:15 +02:00
{
2017-09-06 20:25:03 +02:00
Serialize ( os , Using < VectorFormatter < DefaultFormatter > > ( v ) ) ;
2010-08-29 18:58:15 +02:00
}
template < typename Stream , typename T , typename A >
2016-11-09 12:32:57 +01:00
inline void Serialize ( Stream & os , const std : : vector < T , A > & v )
2010-08-29 18:58:15 +02:00
{
2016-11-09 12:32:57 +01:00
Serialize_impl ( os , v , T ( ) ) ;
2010-08-29 18:58:15 +02:00
}
template < typename Stream , typename T , typename A >
2016-11-09 12:32:57 +01:00
void Unserialize_impl ( Stream & is , std : : vector < T , A > & v , const unsigned char & )
2010-08-29 18:58:15 +02:00
{
// Limit size per read so bogus size value won't cause out of memory
v . clear ( ) ;
unsigned int nSize = ReadCompactSize ( is ) ;
unsigned int i = 0 ;
while ( i < nSize )
{
2011-05-15 09:11:04 +02:00
unsigned int blk = std : : min ( nSize - i , ( unsigned int ) ( 1 + 4999999 / sizeof ( T ) ) ) ;
2010-08-29 18:58:15 +02:00
v . resize ( i + blk ) ;
2024-02-24 08:36:25 +01:00
is . read ( AsWritableBytes ( Span { & v [ i ] , blk } ) ) ;
2010-08-29 18:58:15 +02:00
i + = blk ;
}
}
2014-10-08 20:27:07 +02:00
template < typename Stream , typename T , typename A , typename V >
2016-11-09 12:32:57 +01:00
void Unserialize_impl ( Stream & is , std : : vector < T , A > & v , const V & )
2010-08-29 18:58:15 +02:00
{
2017-09-06 20:25:03 +02:00
Unserialize ( is , Using < VectorFormatter < DefaultFormatter > > ( v ) ) ;
2010-08-29 18:58:15 +02:00
}
template < typename Stream , typename T , typename A >
2016-11-09 12:32:57 +01:00
inline void Unserialize ( Stream & is , std : : vector < T , A > & v )
2010-08-29 18:58:15 +02:00
{
2016-11-09 12:32:57 +01:00
Unserialize_impl ( is , v , T ( ) ) ;
2010-08-29 18:58:15 +02:00
}
2014-10-31 04:34:30 +01:00
/**
* pair
*/
2010-08-29 18:58:15 +02:00
template < typename Stream , typename K , typename T >
2016-11-09 12:32:57 +01:00
void Serialize ( Stream & os , const std : : pair < K , T > & item )
2010-08-29 18:58:15 +02:00
{
2016-11-09 12:32:57 +01:00
Serialize ( os , item . first ) ;
Serialize ( os , item . second ) ;
2010-08-29 18:58:15 +02:00
}
template < typename Stream , typename K , typename T >
2016-11-09 12:32:57 +01:00
void Unserialize ( Stream & is , std : : pair < K , T > & item )
2010-08-29 18:58:15 +02:00
{
2016-11-09 12:32:57 +01:00
Unserialize ( is , item . first ) ;
Unserialize ( is , item . second ) ;
2010-08-29 18:58:15 +02:00
}
2018-09-14 17:53:49 +02:00
/**
* tuple
*/
2019-03-22 11:52:37 +01:00
template < typename Stream , int index , typename . . . Ts >
struct SerializeTuple {
2018-09-14 17:53:49 +02:00
void operator ( ) ( Stream & s , std : : tuple < Ts . . . > & t ) {
2019-03-22 11:52:37 +01:00
SerializeTuple < Stream , index - 1 , Ts . . . > { } ( s , t ) ;
s < < std : : get < index > ( t ) ;
2018-09-14 17:53:49 +02:00
}
} ;
2019-03-22 11:52:37 +01:00
template < typename Stream , typename . . . Ts >
struct SerializeTuple < Stream , 0 , Ts . . . > {
2018-09-14 17:53:49 +02:00
void operator ( ) ( Stream & s , std : : tuple < Ts . . . > & t ) {
2019-03-22 11:52:37 +01:00
s < < std : : get < 0 > ( t ) ;
2018-09-14 17:53:49 +02:00
}
} ;
2019-03-22 11:52:37 +01:00
template < typename Stream , int index , typename . . . Ts >
struct DeserializeTuple {
void operator ( ) ( Stream & s , std : : tuple < Ts . . . > & t ) {
DeserializeTuple < Stream , index - 1 , Ts . . . > { } ( s , t ) ;
s > > std : : get < index > ( t ) ;
}
} ;
template < typename Stream , typename . . . Ts >
struct DeserializeTuple < Stream , 0 , Ts . . . > {
void operator ( ) ( Stream & s , std : : tuple < Ts . . . > & t ) {
s > > std : : get < 0 > ( t ) ;
}
} ;
2018-09-14 17:53:49 +02:00
template < typename Stream , typename . . . Elements >
void Serialize ( Stream & os , const std : : tuple < Elements . . . > & item )
{
const auto size = std : : tuple_size < std : : tuple < Elements . . . > > : : value ;
2019-03-22 11:52:37 +01:00
SerializeTuple < Stream , size - 1 , Elements . . . > { } ( os , const_cast < std : : tuple < Elements . . . > & > ( item ) ) ;
2018-09-14 17:53:49 +02:00
}
template < typename Stream , typename . . . Elements >
void Unserialize ( Stream & is , std : : tuple < Elements . . . > & item )
{
const auto size = std : : tuple_size < std : : tuple < Elements . . . > > : : value ;
2019-03-22 11:52:37 +01:00
DeserializeTuple < Stream , size - 1 , Elements . . . > { } ( is , item ) ;
2018-09-14 17:53:49 +02:00
}
2010-08-29 18:58:15 +02:00
2014-10-31 04:34:30 +01:00
/**
* map
*/
Collection of minor performance optimizations (#2855)
* Merge #13176: Improve CRollingBloomFilter performance: replace modulus with FastMod
9aac9f90d5e56752cc6cbfac48063ad29a01143c replace modulus with FastMod (Martin Ankerl)
Pull request description:
Not sure if this is optimization is necessary, but anyway I have some spare time so here it is. This replaces the slow modulo operation with a much faster 64bit multiplication & shift. This works when the hash is uniformly distributed between 0 and 2^32-1. This speeds up the benchmark by a factor of about 1.3:
```
RollingBloom, 5, 1500000, 3.73733, 4.97569e-07, 4.99002e-07, 4.98372e-07 # before
RollingBloom, 5, 1500000, 2.86842, 3.81630e-07, 3.83730e-07, 3.82473e-07 # FastMod
```
Be aware that this changes the internal data of the filter, so this should probably
not be used for CBloomFilter because of interoperability problems.
Tree-SHA512: 04104f3fb09f56c9d14458a6aad919aeb0a5af944e8ee6a31f00e93c753e22004648c1cd65bf36752b6addec528d19fb665c27b955ce1666a85a928e17afa47a
* Use unordered_map in CSporkManager
In one of my profiling sessions with many InstantSend transactions
happening, calls into CSporkManager added up to about 1% of total CPU time.
This is easily avoidable by using unordered maps.
* Use std::unordered_map instead of std::map in limitedmap
* Use unordered_set for CNode::setAskFor
* Add serialization support for unordered maps and sets
* Use unordered_map for mapArgs and mapMultiArgs
* Let limitedmap prune in batches and use unordered_multimap
Due to the batched pruning, there is no need to maintain an ordered map
of values anymore. Only when nPruneAfterSize, there is a need to create
a temporary ordered vector of values to figure out what can be removed.
* Instead of using a multimap for mapAskFor, use a vector which we sort on demand
CNode::AskFor will now push entries into an initially unordered vector
instead of an ordered multimap. Only when we later want to use vecAskFor in
SendMessages, we sort the vector.
The vector will actually be mostly sorted in most cases as insertion order
usually mimics the desired ordering. Only the last few entries might need
some shuffling around. Doing the sort on-demand should be less wasteful
then trying to maintain correct order all the time.
* Fix compilation of tests
* Fix limitedmap tests
* Rename limitedmap to unordered_limitedmap to ensure backports conflict
This ensures that future backports that depends on limitedmap's ordering
conflict so that we are made aware of needed action.
* Fix compilation error on Travis
2019-04-11 14:42:14 +02:00
template < typename Stream , typename Map >
void SerializeMap ( Stream & os , const Map & m )
2010-08-29 18:58:15 +02:00
{
WriteCompactSize ( os , m . size ( ) ) ;
2017-11-30 23:09:44 +01:00
for ( const auto & entry : m )
Serialize ( os , entry ) ;
2010-08-29 18:58:15 +02:00
}
Collection of minor performance optimizations (#2855)
* Merge #13176: Improve CRollingBloomFilter performance: replace modulus with FastMod
9aac9f90d5e56752cc6cbfac48063ad29a01143c replace modulus with FastMod (Martin Ankerl)
Pull request description:
Not sure if this is optimization is necessary, but anyway I have some spare time so here it is. This replaces the slow modulo operation with a much faster 64bit multiplication & shift. This works when the hash is uniformly distributed between 0 and 2^32-1. This speeds up the benchmark by a factor of about 1.3:
```
RollingBloom, 5, 1500000, 3.73733, 4.97569e-07, 4.99002e-07, 4.98372e-07 # before
RollingBloom, 5, 1500000, 2.86842, 3.81630e-07, 3.83730e-07, 3.82473e-07 # FastMod
```
Be aware that this changes the internal data of the filter, so this should probably
not be used for CBloomFilter because of interoperability problems.
Tree-SHA512: 04104f3fb09f56c9d14458a6aad919aeb0a5af944e8ee6a31f00e93c753e22004648c1cd65bf36752b6addec528d19fb665c27b955ce1666a85a928e17afa47a
* Use unordered_map in CSporkManager
In one of my profiling sessions with many InstantSend transactions
happening, calls into CSporkManager added up to about 1% of total CPU time.
This is easily avoidable by using unordered maps.
* Use std::unordered_map instead of std::map in limitedmap
* Use unordered_set for CNode::setAskFor
* Add serialization support for unordered maps and sets
* Use unordered_map for mapArgs and mapMultiArgs
* Let limitedmap prune in batches and use unordered_multimap
Due to the batched pruning, there is no need to maintain an ordered map
of values anymore. Only when nPruneAfterSize, there is a need to create
a temporary ordered vector of values to figure out what can be removed.
* Instead of using a multimap for mapAskFor, use a vector which we sort on demand
CNode::AskFor will now push entries into an initially unordered vector
instead of an ordered multimap. Only when we later want to use vecAskFor in
SendMessages, we sort the vector.
The vector will actually be mostly sorted in most cases as insertion order
usually mimics the desired ordering. Only the last few entries might need
some shuffling around. Doing the sort on-demand should be less wasteful
then trying to maintain correct order all the time.
* Fix compilation of tests
* Fix limitedmap tests
* Rename limitedmap to unordered_limitedmap to ensure backports conflict
This ensures that future backports that depends on limitedmap's ordering
conflict so that we are made aware of needed action.
* Fix compilation error on Travis
2019-04-11 14:42:14 +02:00
template < typename Stream , typename Map >
void UnserializeMap ( Stream & is , Map & m )
2010-08-29 18:58:15 +02:00
{
m . clear ( ) ;
unsigned int nSize = ReadCompactSize ( is ) ;
Collection of minor performance optimizations (#2855)
* Merge #13176: Improve CRollingBloomFilter performance: replace modulus with FastMod
9aac9f90d5e56752cc6cbfac48063ad29a01143c replace modulus with FastMod (Martin Ankerl)
Pull request description:
Not sure if this is optimization is necessary, but anyway I have some spare time so here it is. This replaces the slow modulo operation with a much faster 64bit multiplication & shift. This works when the hash is uniformly distributed between 0 and 2^32-1. This speeds up the benchmark by a factor of about 1.3:
```
RollingBloom, 5, 1500000, 3.73733, 4.97569e-07, 4.99002e-07, 4.98372e-07 # before
RollingBloom, 5, 1500000, 2.86842, 3.81630e-07, 3.83730e-07, 3.82473e-07 # FastMod
```
Be aware that this changes the internal data of the filter, so this should probably
not be used for CBloomFilter because of interoperability problems.
Tree-SHA512: 04104f3fb09f56c9d14458a6aad919aeb0a5af944e8ee6a31f00e93c753e22004648c1cd65bf36752b6addec528d19fb665c27b955ce1666a85a928e17afa47a
* Use unordered_map in CSporkManager
In one of my profiling sessions with many InstantSend transactions
happening, calls into CSporkManager added up to about 1% of total CPU time.
This is easily avoidable by using unordered maps.
* Use std::unordered_map instead of std::map in limitedmap
* Use unordered_set for CNode::setAskFor
* Add serialization support for unordered maps and sets
* Use unordered_map for mapArgs and mapMultiArgs
* Let limitedmap prune in batches and use unordered_multimap
Due to the batched pruning, there is no need to maintain an ordered map
of values anymore. Only when nPruneAfterSize, there is a need to create
a temporary ordered vector of values to figure out what can be removed.
* Instead of using a multimap for mapAskFor, use a vector which we sort on demand
CNode::AskFor will now push entries into an initially unordered vector
instead of an ordered multimap. Only when we later want to use vecAskFor in
SendMessages, we sort the vector.
The vector will actually be mostly sorted in most cases as insertion order
usually mimics the desired ordering. Only the last few entries might need
some shuffling around. Doing the sort on-demand should be less wasteful
then trying to maintain correct order all the time.
* Fix compilation of tests
* Fix limitedmap tests
* Rename limitedmap to unordered_limitedmap to ensure backports conflict
This ensures that future backports that depends on limitedmap's ordering
conflict so that we are made aware of needed action.
* Fix compilation error on Travis
2019-04-11 14:42:14 +02:00
auto mi = m . begin ( ) ;
2010-08-29 18:58:15 +02:00
for ( unsigned int i = 0 ; i < nSize ; i + + )
{
Collection of minor performance optimizations (#2855)
* Merge #13176: Improve CRollingBloomFilter performance: replace modulus with FastMod
9aac9f90d5e56752cc6cbfac48063ad29a01143c replace modulus with FastMod (Martin Ankerl)
Pull request description:
Not sure if this is optimization is necessary, but anyway I have some spare time so here it is. This replaces the slow modulo operation with a much faster 64bit multiplication & shift. This works when the hash is uniformly distributed between 0 and 2^32-1. This speeds up the benchmark by a factor of about 1.3:
```
RollingBloom, 5, 1500000, 3.73733, 4.97569e-07, 4.99002e-07, 4.98372e-07 # before
RollingBloom, 5, 1500000, 2.86842, 3.81630e-07, 3.83730e-07, 3.82473e-07 # FastMod
```
Be aware that this changes the internal data of the filter, so this should probably
not be used for CBloomFilter because of interoperability problems.
Tree-SHA512: 04104f3fb09f56c9d14458a6aad919aeb0a5af944e8ee6a31f00e93c753e22004648c1cd65bf36752b6addec528d19fb665c27b955ce1666a85a928e17afa47a
* Use unordered_map in CSporkManager
In one of my profiling sessions with many InstantSend transactions
happening, calls into CSporkManager added up to about 1% of total CPU time.
This is easily avoidable by using unordered maps.
* Use std::unordered_map instead of std::map in limitedmap
* Use unordered_set for CNode::setAskFor
* Add serialization support for unordered maps and sets
* Use unordered_map for mapArgs and mapMultiArgs
* Let limitedmap prune in batches and use unordered_multimap
Due to the batched pruning, there is no need to maintain an ordered map
of values anymore. Only when nPruneAfterSize, there is a need to create
a temporary ordered vector of values to figure out what can be removed.
* Instead of using a multimap for mapAskFor, use a vector which we sort on demand
CNode::AskFor will now push entries into an initially unordered vector
instead of an ordered multimap. Only when we later want to use vecAskFor in
SendMessages, we sort the vector.
The vector will actually be mostly sorted in most cases as insertion order
usually mimics the desired ordering. Only the last few entries might need
some shuffling around. Doing the sort on-demand should be less wasteful
then trying to maintain correct order all the time.
* Fix compilation of tests
* Fix limitedmap tests
* Rename limitedmap to unordered_limitedmap to ensure backports conflict
This ensures that future backports that depends on limitedmap's ordering
conflict so that we are made aware of needed action.
* Fix compilation error on Travis
2019-04-11 14:42:14 +02:00
std : : pair < typename std : : remove_const < typename Map : : key_type > : : type , typename std : : remove_const < typename Map : : mapped_type > : : type > item ;
2016-11-09 12:32:57 +01:00
Unserialize ( is , item ) ;
2010-08-29 18:58:15 +02:00
mi = m . insert ( mi , item ) ;
}
}
Collection of minor performance optimizations (#2855)
* Merge #13176: Improve CRollingBloomFilter performance: replace modulus with FastMod
9aac9f90d5e56752cc6cbfac48063ad29a01143c replace modulus with FastMod (Martin Ankerl)
Pull request description:
Not sure if this is optimization is necessary, but anyway I have some spare time so here it is. This replaces the slow modulo operation with a much faster 64bit multiplication & shift. This works when the hash is uniformly distributed between 0 and 2^32-1. This speeds up the benchmark by a factor of about 1.3:
```
RollingBloom, 5, 1500000, 3.73733, 4.97569e-07, 4.99002e-07, 4.98372e-07 # before
RollingBloom, 5, 1500000, 2.86842, 3.81630e-07, 3.83730e-07, 3.82473e-07 # FastMod
```
Be aware that this changes the internal data of the filter, so this should probably
not be used for CBloomFilter because of interoperability problems.
Tree-SHA512: 04104f3fb09f56c9d14458a6aad919aeb0a5af944e8ee6a31f00e93c753e22004648c1cd65bf36752b6addec528d19fb665c27b955ce1666a85a928e17afa47a
* Use unordered_map in CSporkManager
In one of my profiling sessions with many InstantSend transactions
happening, calls into CSporkManager added up to about 1% of total CPU time.
This is easily avoidable by using unordered maps.
* Use std::unordered_map instead of std::map in limitedmap
* Use unordered_set for CNode::setAskFor
* Add serialization support for unordered maps and sets
* Use unordered_map for mapArgs and mapMultiArgs
* Let limitedmap prune in batches and use unordered_multimap
Due to the batched pruning, there is no need to maintain an ordered map
of values anymore. Only when nPruneAfterSize, there is a need to create
a temporary ordered vector of values to figure out what can be removed.
* Instead of using a multimap for mapAskFor, use a vector which we sort on demand
CNode::AskFor will now push entries into an initially unordered vector
instead of an ordered multimap. Only when we later want to use vecAskFor in
SendMessages, we sort the vector.
The vector will actually be mostly sorted in most cases as insertion order
usually mimics the desired ordering. Only the last few entries might need
some shuffling around. Doing the sort on-demand should be less wasteful
then trying to maintain correct order all the time.
* Fix compilation of tests
* Fix limitedmap tests
* Rename limitedmap to unordered_limitedmap to ensure backports conflict
This ensures that future backports that depends on limitedmap's ordering
conflict so that we are made aware of needed action.
* Fix compilation error on Travis
2019-04-11 14:42:14 +02:00
template < typename Stream , typename K , typename T , typename Pred , typename A >
void Serialize ( Stream & os , const std : : map < K , T , Pred , A > & m )
{
SerializeMap ( os , m ) ;
}
template < typename Stream , typename K , typename T , typename Pred , typename A >
void Unserialize ( Stream & is , std : : map < K , T , Pred , A > & m )
{
UnserializeMap ( is , m ) ;
}
template < typename Stream , typename K , typename T , typename Hash , typename Pred , typename A >
void Serialize ( Stream & os , const std : : unordered_map < K , T , Hash , Pred , A > & m )
{
SerializeMap ( os , m ) ;
}
2010-08-29 18:58:15 +02:00
Collection of minor performance optimizations (#2855)
* Merge #13176: Improve CRollingBloomFilter performance: replace modulus with FastMod
9aac9f90d5e56752cc6cbfac48063ad29a01143c replace modulus with FastMod (Martin Ankerl)
Pull request description:
Not sure if this is optimization is necessary, but anyway I have some spare time so here it is. This replaces the slow modulo operation with a much faster 64bit multiplication & shift. This works when the hash is uniformly distributed between 0 and 2^32-1. This speeds up the benchmark by a factor of about 1.3:
```
RollingBloom, 5, 1500000, 3.73733, 4.97569e-07, 4.99002e-07, 4.98372e-07 # before
RollingBloom, 5, 1500000, 2.86842, 3.81630e-07, 3.83730e-07, 3.82473e-07 # FastMod
```
Be aware that this changes the internal data of the filter, so this should probably
not be used for CBloomFilter because of interoperability problems.
Tree-SHA512: 04104f3fb09f56c9d14458a6aad919aeb0a5af944e8ee6a31f00e93c753e22004648c1cd65bf36752b6addec528d19fb665c27b955ce1666a85a928e17afa47a
* Use unordered_map in CSporkManager
In one of my profiling sessions with many InstantSend transactions
happening, calls into CSporkManager added up to about 1% of total CPU time.
This is easily avoidable by using unordered maps.
* Use std::unordered_map instead of std::map in limitedmap
* Use unordered_set for CNode::setAskFor
* Add serialization support for unordered maps and sets
* Use unordered_map for mapArgs and mapMultiArgs
* Let limitedmap prune in batches and use unordered_multimap
Due to the batched pruning, there is no need to maintain an ordered map
of values anymore. Only when nPruneAfterSize, there is a need to create
a temporary ordered vector of values to figure out what can be removed.
* Instead of using a multimap for mapAskFor, use a vector which we sort on demand
CNode::AskFor will now push entries into an initially unordered vector
instead of an ordered multimap. Only when we later want to use vecAskFor in
SendMessages, we sort the vector.
The vector will actually be mostly sorted in most cases as insertion order
usually mimics the desired ordering. Only the last few entries might need
some shuffling around. Doing the sort on-demand should be less wasteful
then trying to maintain correct order all the time.
* Fix compilation of tests
* Fix limitedmap tests
* Rename limitedmap to unordered_limitedmap to ensure backports conflict
This ensures that future backports that depends on limitedmap's ordering
conflict so that we are made aware of needed action.
* Fix compilation error on Travis
2019-04-11 14:42:14 +02:00
template < typename Stream , typename K , typename T , typename Hash , typename Pred , typename A >
void Unserialize ( Stream & is , std : : unordered_map < K , T , Hash , Pred , A > & m )
{
UnserializeMap ( is , m ) ;
}
2010-08-29 18:58:15 +02:00
2014-10-31 04:34:30 +01:00
/**
* set
*/
Collection of minor performance optimizations (#2855)
* Merge #13176: Improve CRollingBloomFilter performance: replace modulus with FastMod
9aac9f90d5e56752cc6cbfac48063ad29a01143c replace modulus with FastMod (Martin Ankerl)
Pull request description:
Not sure if this is optimization is necessary, but anyway I have some spare time so here it is. This replaces the slow modulo operation with a much faster 64bit multiplication & shift. This works when the hash is uniformly distributed between 0 and 2^32-1. This speeds up the benchmark by a factor of about 1.3:
```
RollingBloom, 5, 1500000, 3.73733, 4.97569e-07, 4.99002e-07, 4.98372e-07 # before
RollingBloom, 5, 1500000, 2.86842, 3.81630e-07, 3.83730e-07, 3.82473e-07 # FastMod
```
Be aware that this changes the internal data of the filter, so this should probably
not be used for CBloomFilter because of interoperability problems.
Tree-SHA512: 04104f3fb09f56c9d14458a6aad919aeb0a5af944e8ee6a31f00e93c753e22004648c1cd65bf36752b6addec528d19fb665c27b955ce1666a85a928e17afa47a
* Use unordered_map in CSporkManager
In one of my profiling sessions with many InstantSend transactions
happening, calls into CSporkManager added up to about 1% of total CPU time.
This is easily avoidable by using unordered maps.
* Use std::unordered_map instead of std::map in limitedmap
* Use unordered_set for CNode::setAskFor
* Add serialization support for unordered maps and sets
* Use unordered_map for mapArgs and mapMultiArgs
* Let limitedmap prune in batches and use unordered_multimap
Due to the batched pruning, there is no need to maintain an ordered map
of values anymore. Only when nPruneAfterSize, there is a need to create
a temporary ordered vector of values to figure out what can be removed.
* Instead of using a multimap for mapAskFor, use a vector which we sort on demand
CNode::AskFor will now push entries into an initially unordered vector
instead of an ordered multimap. Only when we later want to use vecAskFor in
SendMessages, we sort the vector.
The vector will actually be mostly sorted in most cases as insertion order
usually mimics the desired ordering. Only the last few entries might need
some shuffling around. Doing the sort on-demand should be less wasteful
then trying to maintain correct order all the time.
* Fix compilation of tests
* Fix limitedmap tests
* Rename limitedmap to unordered_limitedmap to ensure backports conflict
This ensures that future backports that depends on limitedmap's ordering
conflict so that we are made aware of needed action.
* Fix compilation error on Travis
2019-04-11 14:42:14 +02:00
template < typename Stream , typename Set >
void SerializeSet ( Stream & os , const Set & m )
2010-08-29 18:58:15 +02:00
{
WriteCompactSize ( os , m . size ( ) ) ;
Collection of minor performance optimizations (#2855)
* Merge #13176: Improve CRollingBloomFilter performance: replace modulus with FastMod
9aac9f90d5e56752cc6cbfac48063ad29a01143c replace modulus with FastMod (Martin Ankerl)
Pull request description:
Not sure if this is optimization is necessary, but anyway I have some spare time so here it is. This replaces the slow modulo operation with a much faster 64bit multiplication & shift. This works when the hash is uniformly distributed between 0 and 2^32-1. This speeds up the benchmark by a factor of about 1.3:
```
RollingBloom, 5, 1500000, 3.73733, 4.97569e-07, 4.99002e-07, 4.98372e-07 # before
RollingBloom, 5, 1500000, 2.86842, 3.81630e-07, 3.83730e-07, 3.82473e-07 # FastMod
```
Be aware that this changes the internal data of the filter, so this should probably
not be used for CBloomFilter because of interoperability problems.
Tree-SHA512: 04104f3fb09f56c9d14458a6aad919aeb0a5af944e8ee6a31f00e93c753e22004648c1cd65bf36752b6addec528d19fb665c27b955ce1666a85a928e17afa47a
* Use unordered_map in CSporkManager
In one of my profiling sessions with many InstantSend transactions
happening, calls into CSporkManager added up to about 1% of total CPU time.
This is easily avoidable by using unordered maps.
* Use std::unordered_map instead of std::map in limitedmap
* Use unordered_set for CNode::setAskFor
* Add serialization support for unordered maps and sets
* Use unordered_map for mapArgs and mapMultiArgs
* Let limitedmap prune in batches and use unordered_multimap
Due to the batched pruning, there is no need to maintain an ordered map
of values anymore. Only when nPruneAfterSize, there is a need to create
a temporary ordered vector of values to figure out what can be removed.
* Instead of using a multimap for mapAskFor, use a vector which we sort on demand
CNode::AskFor will now push entries into an initially unordered vector
instead of an ordered multimap. Only when we later want to use vecAskFor in
SendMessages, we sort the vector.
The vector will actually be mostly sorted in most cases as insertion order
usually mimics the desired ordering. Only the last few entries might need
some shuffling around. Doing the sort on-demand should be less wasteful
then trying to maintain correct order all the time.
* Fix compilation of tests
* Fix limitedmap tests
* Rename limitedmap to unordered_limitedmap to ensure backports conflict
This ensures that future backports that depends on limitedmap's ordering
conflict so that we are made aware of needed action.
* Fix compilation error on Travis
2019-04-11 14:42:14 +02:00
for ( auto it = m . begin ( ) ; it ! = m . end ( ) ; + + it )
2016-11-09 12:32:57 +01:00
Serialize ( os , ( * it ) ) ;
2010-08-29 18:58:15 +02:00
}
Collection of minor performance optimizations (#2855)
* Merge #13176: Improve CRollingBloomFilter performance: replace modulus with FastMod
9aac9f90d5e56752cc6cbfac48063ad29a01143c replace modulus with FastMod (Martin Ankerl)
Pull request description:
Not sure if this is optimization is necessary, but anyway I have some spare time so here it is. This replaces the slow modulo operation with a much faster 64bit multiplication & shift. This works when the hash is uniformly distributed between 0 and 2^32-1. This speeds up the benchmark by a factor of about 1.3:
```
RollingBloom, 5, 1500000, 3.73733, 4.97569e-07, 4.99002e-07, 4.98372e-07 # before
RollingBloom, 5, 1500000, 2.86842, 3.81630e-07, 3.83730e-07, 3.82473e-07 # FastMod
```
Be aware that this changes the internal data of the filter, so this should probably
not be used for CBloomFilter because of interoperability problems.
Tree-SHA512: 04104f3fb09f56c9d14458a6aad919aeb0a5af944e8ee6a31f00e93c753e22004648c1cd65bf36752b6addec528d19fb665c27b955ce1666a85a928e17afa47a
* Use unordered_map in CSporkManager
In one of my profiling sessions with many InstantSend transactions
happening, calls into CSporkManager added up to about 1% of total CPU time.
This is easily avoidable by using unordered maps.
* Use std::unordered_map instead of std::map in limitedmap
* Use unordered_set for CNode::setAskFor
* Add serialization support for unordered maps and sets
* Use unordered_map for mapArgs and mapMultiArgs
* Let limitedmap prune in batches and use unordered_multimap
Due to the batched pruning, there is no need to maintain an ordered map
of values anymore. Only when nPruneAfterSize, there is a need to create
a temporary ordered vector of values to figure out what can be removed.
* Instead of using a multimap for mapAskFor, use a vector which we sort on demand
CNode::AskFor will now push entries into an initially unordered vector
instead of an ordered multimap. Only when we later want to use vecAskFor in
SendMessages, we sort the vector.
The vector will actually be mostly sorted in most cases as insertion order
usually mimics the desired ordering. Only the last few entries might need
some shuffling around. Doing the sort on-demand should be less wasteful
then trying to maintain correct order all the time.
* Fix compilation of tests
* Fix limitedmap tests
* Rename limitedmap to unordered_limitedmap to ensure backports conflict
This ensures that future backports that depends on limitedmap's ordering
conflict so that we are made aware of needed action.
* Fix compilation error on Travis
2019-04-11 14:42:14 +02:00
template < typename Stream , typename Set >
void UnserializeSet ( Stream & is , Set & m )
2010-08-29 18:58:15 +02:00
{
m . clear ( ) ;
unsigned int nSize = ReadCompactSize ( is ) ;
Collection of minor performance optimizations (#2855)
* Merge #13176: Improve CRollingBloomFilter performance: replace modulus with FastMod
9aac9f90d5e56752cc6cbfac48063ad29a01143c replace modulus with FastMod (Martin Ankerl)
Pull request description:
Not sure if this is optimization is necessary, but anyway I have some spare time so here it is. This replaces the slow modulo operation with a much faster 64bit multiplication & shift. This works when the hash is uniformly distributed between 0 and 2^32-1. This speeds up the benchmark by a factor of about 1.3:
```
RollingBloom, 5, 1500000, 3.73733, 4.97569e-07, 4.99002e-07, 4.98372e-07 # before
RollingBloom, 5, 1500000, 2.86842, 3.81630e-07, 3.83730e-07, 3.82473e-07 # FastMod
```
Be aware that this changes the internal data of the filter, so this should probably
not be used for CBloomFilter because of interoperability problems.
Tree-SHA512: 04104f3fb09f56c9d14458a6aad919aeb0a5af944e8ee6a31f00e93c753e22004648c1cd65bf36752b6addec528d19fb665c27b955ce1666a85a928e17afa47a
* Use unordered_map in CSporkManager
In one of my profiling sessions with many InstantSend transactions
happening, calls into CSporkManager added up to about 1% of total CPU time.
This is easily avoidable by using unordered maps.
* Use std::unordered_map instead of std::map in limitedmap
* Use unordered_set for CNode::setAskFor
* Add serialization support for unordered maps and sets
* Use unordered_map for mapArgs and mapMultiArgs
* Let limitedmap prune in batches and use unordered_multimap
Due to the batched pruning, there is no need to maintain an ordered map
of values anymore. Only when nPruneAfterSize, there is a need to create
a temporary ordered vector of values to figure out what can be removed.
* Instead of using a multimap for mapAskFor, use a vector which we sort on demand
CNode::AskFor will now push entries into an initially unordered vector
instead of an ordered multimap. Only when we later want to use vecAskFor in
SendMessages, we sort the vector.
The vector will actually be mostly sorted in most cases as insertion order
usually mimics the desired ordering. Only the last few entries might need
some shuffling around. Doing the sort on-demand should be less wasteful
then trying to maintain correct order all the time.
* Fix compilation of tests
* Fix limitedmap tests
* Rename limitedmap to unordered_limitedmap to ensure backports conflict
This ensures that future backports that depends on limitedmap's ordering
conflict so that we are made aware of needed action.
* Fix compilation error on Travis
2019-04-11 14:42:14 +02:00
auto it = m . begin ( ) ;
2010-08-29 18:58:15 +02:00
for ( unsigned int i = 0 ; i < nSize ; i + + )
{
Collection of minor performance optimizations (#2855)
* Merge #13176: Improve CRollingBloomFilter performance: replace modulus with FastMod
9aac9f90d5e56752cc6cbfac48063ad29a01143c replace modulus with FastMod (Martin Ankerl)
Pull request description:
Not sure if this is optimization is necessary, but anyway I have some spare time so here it is. This replaces the slow modulo operation with a much faster 64bit multiplication & shift. This works when the hash is uniformly distributed between 0 and 2^32-1. This speeds up the benchmark by a factor of about 1.3:
```
RollingBloom, 5, 1500000, 3.73733, 4.97569e-07, 4.99002e-07, 4.98372e-07 # before
RollingBloom, 5, 1500000, 2.86842, 3.81630e-07, 3.83730e-07, 3.82473e-07 # FastMod
```
Be aware that this changes the internal data of the filter, so this should probably
not be used for CBloomFilter because of interoperability problems.
Tree-SHA512: 04104f3fb09f56c9d14458a6aad919aeb0a5af944e8ee6a31f00e93c753e22004648c1cd65bf36752b6addec528d19fb665c27b955ce1666a85a928e17afa47a
* Use unordered_map in CSporkManager
In one of my profiling sessions with many InstantSend transactions
happening, calls into CSporkManager added up to about 1% of total CPU time.
This is easily avoidable by using unordered maps.
* Use std::unordered_map instead of std::map in limitedmap
* Use unordered_set for CNode::setAskFor
* Add serialization support for unordered maps and sets
* Use unordered_map for mapArgs and mapMultiArgs
* Let limitedmap prune in batches and use unordered_multimap
Due to the batched pruning, there is no need to maintain an ordered map
of values anymore. Only when nPruneAfterSize, there is a need to create
a temporary ordered vector of values to figure out what can be removed.
* Instead of using a multimap for mapAskFor, use a vector which we sort on demand
CNode::AskFor will now push entries into an initially unordered vector
instead of an ordered multimap. Only when we later want to use vecAskFor in
SendMessages, we sort the vector.
The vector will actually be mostly sorted in most cases as insertion order
usually mimics the desired ordering. Only the last few entries might need
some shuffling around. Doing the sort on-demand should be less wasteful
then trying to maintain correct order all the time.
* Fix compilation of tests
* Fix limitedmap tests
* Rename limitedmap to unordered_limitedmap to ensure backports conflict
This ensures that future backports that depends on limitedmap's ordering
conflict so that we are made aware of needed action.
* Fix compilation error on Travis
2019-04-11 14:42:14 +02:00
typename std : : remove_const < typename Set : : key_type > : : type key ;
2016-11-09 12:32:57 +01:00
Unserialize ( is , key ) ;
2010-08-29 18:58:15 +02:00
it = m . insert ( it , key ) ;
}
}
Collection of minor performance optimizations (#2855)
* Merge #13176: Improve CRollingBloomFilter performance: replace modulus with FastMod
9aac9f90d5e56752cc6cbfac48063ad29a01143c replace modulus with FastMod (Martin Ankerl)
Pull request description:
Not sure if this is optimization is necessary, but anyway I have some spare time so here it is. This replaces the slow modulo operation with a much faster 64bit multiplication & shift. This works when the hash is uniformly distributed between 0 and 2^32-1. This speeds up the benchmark by a factor of about 1.3:
```
RollingBloom, 5, 1500000, 3.73733, 4.97569e-07, 4.99002e-07, 4.98372e-07 # before
RollingBloom, 5, 1500000, 2.86842, 3.81630e-07, 3.83730e-07, 3.82473e-07 # FastMod
```
Be aware that this changes the internal data of the filter, so this should probably
not be used for CBloomFilter because of interoperability problems.
Tree-SHA512: 04104f3fb09f56c9d14458a6aad919aeb0a5af944e8ee6a31f00e93c753e22004648c1cd65bf36752b6addec528d19fb665c27b955ce1666a85a928e17afa47a
* Use unordered_map in CSporkManager
In one of my profiling sessions with many InstantSend transactions
happening, calls into CSporkManager added up to about 1% of total CPU time.
This is easily avoidable by using unordered maps.
* Use std::unordered_map instead of std::map in limitedmap
* Use unordered_set for CNode::setAskFor
* Add serialization support for unordered maps and sets
* Use unordered_map for mapArgs and mapMultiArgs
* Let limitedmap prune in batches and use unordered_multimap
Due to the batched pruning, there is no need to maintain an ordered map
of values anymore. Only when nPruneAfterSize, there is a need to create
a temporary ordered vector of values to figure out what can be removed.
* Instead of using a multimap for mapAskFor, use a vector which we sort on demand
CNode::AskFor will now push entries into an initially unordered vector
instead of an ordered multimap. Only when we later want to use vecAskFor in
SendMessages, we sort the vector.
The vector will actually be mostly sorted in most cases as insertion order
usually mimics the desired ordering. Only the last few entries might need
some shuffling around. Doing the sort on-demand should be less wasteful
then trying to maintain correct order all the time.
* Fix compilation of tests
* Fix limitedmap tests
* Rename limitedmap to unordered_limitedmap to ensure backports conflict
This ensures that future backports that depends on limitedmap's ordering
conflict so that we are made aware of needed action.
* Fix compilation error on Travis
2019-04-11 14:42:14 +02:00
template < typename Stream , typename K , typename Pred , typename A >
void Serialize ( Stream & os , const std : : set < K , Pred , A > & m )
{
SerializeSet ( os , m ) ;
}
template < typename Stream , typename K , typename Pred , typename A >
void Unserialize ( Stream & is , std : : set < K , Pred , A > & m )
{
UnserializeSet ( is , m ) ;
}
template < typename Stream , typename K , typename Hash , typename Pred , typename A >
void Serialize ( Stream & os , const std : : unordered_set < K , Hash , Pred , A > & m )
{
SerializeSet ( os , m ) ;
}
template < typename Stream , typename K , typename Hash , typename Pred , typename A >
void Unserialize ( Stream & is , std : : unordered_set < K , Hash , Pred , A > & m )
{
UnserializeSet ( is , m ) ;
}
2016-11-13 18:52:34 +01:00
/**
* list
*/
template < typename Stream , typename T , typename A >
2016-11-09 12:32:57 +01:00
void Serialize ( Stream & os , const std : : list < T , A > & l )
2016-11-13 18:52:34 +01:00
{
WriteCompactSize ( os , l . size ( ) ) ;
for ( typename std : : list < T , A > : : const_iterator it = l . begin ( ) ; it ! = l . end ( ) ; + + it )
2016-11-09 12:32:57 +01:00
Serialize ( os , ( * it ) ) ;
2016-11-13 18:52:34 +01:00
}
template < typename Stream , typename T , typename A >
2016-11-09 12:32:57 +01:00
void Unserialize ( Stream & is , std : : list < T , A > & l )
2016-11-13 18:52:34 +01:00
{
l . clear ( ) ;
unsigned int nSize = ReadCompactSize ( is ) ;
for ( unsigned int i = 0 ; i < nSize ; i + + )
{
T val ;
2016-11-09 12:32:57 +01:00
Unserialize ( is , val ) ;
2016-11-13 18:52:34 +01:00
l . push_back ( val ) ;
}
}
2010-08-29 18:58:15 +02:00
2016-11-21 10:51:32 +01:00
/**
* unique_ptr
*/
template < typename Stream , typename T > void
Serialize ( Stream & os , const std : : unique_ptr < const T > & p )
{
Serialize ( os , * p ) ;
}
template < typename Stream , typename T >
void Unserialize ( Stream & is , std : : unique_ptr < const T > & p )
{
p . reset ( new T ( deserialize , is ) ) ;
}
/**
* shared_ptr
*/
template < typename Stream , typename T > void
2019-01-03 21:08:34 +01:00
Serialize ( Stream & os , const std : : shared_ptr < T > & p )
2016-11-21 10:51:32 +01:00
{
Serialize ( os , * p ) ;
}
template < typename Stream , typename T >
2019-01-03 21:08:34 +01:00
void Unserialize ( Stream & is , std : : shared_ptr < T > & p )
2016-11-21 10:51:32 +01:00
{
2019-01-03 21:08:34 +01:00
p = std : : make_shared < T > ( deserialize , is ) ;
2016-11-21 10:51:32 +01:00
}
2021-09-30 23:00:52 +02:00
/**
* atomic
*/
template < typename Stream , typename T >
void Serialize ( Stream & os , const std : : atomic < T > & a )
{
Serialize ( os , a . load ( ) ) ;
}
template < typename Stream , typename T >
void Unserialize ( Stream & is , std : : atomic < T > & a )
{
T val ;
Unserialize ( is , val ) ;
a . store ( val ) ;
}
2014-10-31 04:34:30 +01:00
/**
2021-05-27 16:53:41 +02:00
* Support for SERIALIZE_METHODS and READWRITE macro .
2014-10-31 04:34:30 +01:00
*/
2014-08-21 00:49:32 +02:00
struct CSerActionSerialize
2010-08-29 18:58:15 +02:00
{
2016-11-09 12:32:57 +01:00
constexpr bool ForRead ( ) const { return false ; }
2014-08-21 00:49:32 +02:00
} ;
struct CSerActionUnserialize
{
2016-11-09 12:32:57 +01:00
constexpr bool ForRead ( ) const { return true ; }
2014-08-21 00:49:32 +02:00
} ;
2010-08-29 18:58:15 +02:00
2012-01-03 09:03:07 +01:00
2016-11-09 12:32:57 +01:00
/* ::GetSerializeSize implementations
*
* Computing the serialized size of objects is done through a special stream
* object of type CSizeComputer , which only records the number of bytes written
* to it .
*
* If your Serialize or SerializationOp method has non - trivial overhead for
* serialization , it may be worthwhile to implement a specialized version for
* CSizeComputer , which uses the s . seek ( ) method to record bytes that would
* be written instead .
*/
2014-07-10 20:16:58 +02:00
class CSizeComputer
{
protected :
size_t nSize ;
2016-11-09 12:32:57 +01:00
const int nVersion ;
2014-07-10 20:16:58 +02:00
public :
2018-09-11 09:09:57 +02:00
explicit CSizeComputer ( int nVersionIn ) : nSize ( 0 ) , nVersion ( nVersionIn ) { }
2014-07-10 20:16:58 +02:00
2024-02-24 08:36:25 +01:00
void write ( Span < const std : : byte > src )
2016-11-09 12:32:57 +01:00
{
2024-02-24 08:36:25 +01:00
this - > nSize + = src . size ( ) ;
2016-11-09 12:32:57 +01:00
}
/** Pretend _nSize bytes are written, without specifying them. */
void seek ( size_t _nSize )
2014-07-10 20:16:58 +02:00
{
2016-11-09 12:32:57 +01:00
this - > nSize + = _nSize ;
2014-07-10 20:16:58 +02:00
}
template < typename T >
CSizeComputer & operator < < ( const T & obj )
{
2016-11-09 12:32:57 +01:00
: : Serialize ( * this , obj ) ;
2014-07-10 20:16:58 +02:00
return ( * this ) ;
}
size_t size ( ) const {
return nSize ;
}
2016-11-09 12:32:57 +01:00
int GetVersion ( ) const { return nVersion ; }
2014-07-10 20:16:58 +02:00
} ;
2017-07-27 16:28:05 +02:00
template < typename Stream >
2016-11-09 12:32:57 +01:00
void SerializeMany ( Stream & s )
2017-07-27 16:28:05 +02:00
{
}
template < typename Stream , typename Arg , typename . . . Args >
2020-12-17 03:07:07 +01:00
void SerializeMany ( Stream & s , const Arg & arg , const Args & . . . args )
2017-07-27 16:28:05 +02:00
{
2020-12-17 03:07:07 +01:00
: : Serialize ( s , arg ) ;
: : SerializeMany ( s , args . . . ) ;
2017-07-27 16:28:05 +02:00
}
template < typename Stream >
2016-11-09 12:32:57 +01:00
inline void UnserializeMany ( Stream & s )
2017-07-27 16:28:05 +02:00
{
}
template < typename Stream , typename Arg , typename . . . Args >
2020-12-17 03:07:07 +01:00
inline void UnserializeMany ( Stream & s , Arg & & arg , Args & & . . . args )
2017-07-27 16:28:05 +02:00
{
2016-11-09 12:32:57 +01:00
: : Unserialize ( s , arg ) ;
: : UnserializeMany ( s , args . . . ) ;
2017-07-27 16:28:05 +02:00
}
template < typename Stream , typename . . . Args >
2020-12-17 03:07:07 +01:00
inline void SerReadWriteMany ( Stream & s , CSerActionSerialize ser_action , const Args & . . . args )
2017-07-27 16:28:05 +02:00
{
2020-12-17 03:07:07 +01:00
: : SerializeMany ( s , args . . . ) ;
2017-07-27 16:28:05 +02:00
}
template < typename Stream , typename . . . Args >
2020-12-17 03:07:07 +01:00
inline void SerReadWriteMany ( Stream & s , CSerActionUnserialize ser_action , Args & & . . . args )
2016-11-09 12:32:57 +01:00
{
: : UnserializeMany ( s , args . . . ) ;
}
2020-05-20 13:30:21 +02:00
template < typename Stream , typename Type , typename Fn >
inline void SerRead ( Stream & s , CSerActionSerialize ser_action , Type & & , Fn & & )
{
}
template < typename Stream , typename Type , typename Fn >
inline void SerRead ( Stream & s , CSerActionUnserialize ser_action , Type & & obj , Fn & & fn )
{
fn ( s , std : : forward < Type > ( obj ) ) ;
}
template < typename Stream , typename Type , typename Fn >
inline void SerWrite ( Stream & s , CSerActionSerialize ser_action , Type & & obj , Fn & & fn )
{
fn ( s , std : : forward < Type > ( obj ) ) ;
}
template < typename Stream , typename Type , typename Fn >
inline void SerWrite ( Stream & s , CSerActionUnserialize ser_action , Type & & , Fn & & )
{
}
2016-11-09 12:32:57 +01:00
template < typename I >
inline void WriteVarInt ( CSizeComputer & s , I n )
{
s . seek ( GetSizeOfVarInt < I > ( n ) ) ;
}
inline void WriteCompactSize ( CSizeComputer & s , uint64_t nSize )
{
s . seek ( GetSizeOfCompactSize ( nSize ) ) ;
}
template < typename T >
2018-09-11 09:09:57 +02:00
size_t GetSerializeSize ( const T & t , int nVersion )
2016-11-09 12:32:57 +01:00
{
2018-09-11 09:09:57 +02:00
return ( CSizeComputer ( nVersion ) < < t ) . size ( ) ;
2016-11-09 12:32:57 +01:00
}
2018-09-11 09:09:57 +02:00
template < typename . . . T >
size_t GetSerializeSizeMany ( int nVersion , const T & . . . t )
2017-07-27 16:28:05 +02:00
{
2018-09-11 09:09:57 +02:00
CSizeComputer sc ( nVersion ) ;
2018-07-23 12:42:40 +02:00
SerializeMany ( sc , t . . . ) ;
return sc . size ( ) ;
}
2014-08-28 22:21:03 +02:00
# endif // BITCOIN_SERIALIZE_H