1.5M block import time (VerifyBlocks disabled) on AMD Ryzen 5 1600/16GB/HDD,
before:
real 159m16.551s
user 69m58.279s
sys 7m34.334s
after:
real 139m41.836s
user 67m12.477s
sys 6m19.420s
12% which is even a bit more than could be expected from inputs analysis (that
has around 10% cache hits for a block-wide cache), worth doing.
This change reduces pressure on DB by doing the following things:
* not storing additional KV pair for SpentCoin
* storing Output right in the UnspentCoin, thus eliminating the need to get a
full transaction from DB
At the same time it makes UnspentCoin more fat and hot, but it should probably
worth it.
Also drop `GetUnspentCoinStateOrNew` as it shouldn't ever existed, UTXOs
can't come out of nowhere.
1.5M block import time (VerifyBlocks disabled) on AMD Ryzen 5 1600/16GB/HDD,
before:
real 302m9.895s
user 96m17.200s
sys 13m37.084s
after:
real 159m16.551s
user 69m58.279s
sys 7m34.334s
So it's almost two-fold which is a great improvement.
C# uses ToArray() or UintXXX(bytes) here which interprets hashes as they
should be interpreted (BE, although they always convert to LE when converting
to String just for the fun of it). It leads to state difference for us at
block 2025204 where even though we have the same value for the key, the key
itself differs, ours:
dd2b538e2a0c1db1ae5061c15be14f916bd1e678e512ffcda6d9499d8e7fe97ee71fd6b8004583d9afe09cc4dadbd5deb63d01e061009b7cffdaa674beae0f930ebe6085af900093e5fe56b34a5c220ccdcf6efc336fc5000000000000000000000000000000000010
theirs:
dd2b538e2a0c1db1ae5061c15be14f916bd1e67861e0013db6ded5dbdac49ce0afd9834500b8d61fe77ee97f8e9d49d9a6cdff12e5009b7cffdaa674beae0f930ebe6085af900093e5fe56b34a5c220ccdcf6efc336fc5000000000000000000000000000000000010
In this key there is a tx hash encoded
(e512ffcda6d9499d8e7fe97ee71fd6b84583d9afe09cc4dadbd5deb63d01e061 in LE used
by all the tools like neoscan).
I love Neo.
Both verification and invocation scripts need to
be unmarshaled from hex.
Also fix failing RPC tests: block contains non-pointer
`transaction.Witness` field and (*Witness).MarshalJSON method
is not called.
They have it specified right in the transaction. Unfortunately, this little
change rendered invalid our RPC test chain, but I think it became even better
after it, especially given that chain generation is a nice test by itself, so
it should be running as a regular test.
prevHash == input.PrevHash, so make less DB accesses and more real work. Fix
some bugs along the way:
* spentCoins structure may already be present in the DB when persisting TX,
there is nothing wrong with that and we shouldn't overwrite it
* it's only used for NEO and only to check for claim validity. Thus, when
processing claim tx the corresponding spentCoins should always be present
in the DB
Everywhere in this code prevHash == input.PrevHash, thus we can easily move
some common code out of the loop saving on DB accesses and
serialization/deserialization.
Implement mempool and consensus block creation policies, almost the same as
SimplePolicy plugin for C# node provides with two caveats:
* HighPriorityTxType is not configured and hardcoded to ClaimType
* BlockedAccounts are not supported
Other than that it allows us to run successfuly as testnet CN, previously our
proposals were rejected because we were proposing blocks with oversized
transactions (that are rejected by PoolTx() now).
Mainnet and testnet configuration files are updated accordingly, but privnet
is left as is with no limits.
Configuration is currently attached to the Blockchain and so is the code that
does policying, it may be moved somewhere in the future, but it works for
now.
Both are very useful outside of the core, this change also makes respective
transactions initialize with the package as they don't depend on any kind of
input and it makes no sense recreating them again and again on every use.
It wasn't sorted when all validators were elected. There is also no need to do
`Unique()` on the result because validators are distinguished by the key, so
no two registered validators can have the same key.
- Attribute should have 2 fields (usage, data)
- VOut should have 4 (5) fields (asset, value, address, n)
- Script should have 2 fields (invocation, verification)
As C# node does it. Technically it's only needed for consensus and could be
implemented in the appropriate package, but for better compatibility with C#
node we're better returning it sorted right here.
We were completely lacking ValidatorsCount that is supposed to track the
number of votes with particular count of consensus nodes which in theory can
change the number of active consensus nodes (if it ever to exceed the number
of standby validators), so we were not producing the right count and based on
that not giving the right set of validators.
Fixes#512.
Simple as that: UnregisteredAndHasNoVotes != !RegisteredAndHasVotes
Registered validators should stay in the DB, we might be in the process of
updating votes for them and that starts with subtraction.
Lack of FreeGasLimit in privnet leads to gas limit exceeding in case of transactions with small amount of GAS to be used for invoke operation (< real cost of the transaction). Solution: Fixed constraint in case when FreeGasLimit == 0. So now we are able to perform transactions in privnet with FreeGasLimit = 0 for free.
Tesnet sync failed with:
Feb 07 00:04:19 nodoka neo-go[1747]: 2020-02-07T00:04:19.838+0300 WARN blockQueue: failed adding block into the blockchain {"error": "failed to store notifications: not supported", "blockHeight": 713984, "nextIndex": 713985}
because some (not so) smart contract emitted a notification with an
InteropItem inside.
Fix mempool and chain locking
This allows us easily make 1000 Tx/s in 4-nodes privnet, fixes potential
double spends and improves mempool testing coverage.
Our mempool only contains valid verified transactions all the time, it never
has any unverified ones. Unverified pool made some sense for quick unverifying
after the new block acceptance (and gradual background reverification), but
reverification needs some non-trivial locking between blockchain and mempool
and internal mempool state locking (reverifying tx and moving it between
unverified and verified pools must be atomic). But our current reverification
is fast enough (and has all the appropriate locks), so bothering with
unverified pool makes little sense.
We not only need to remove transactions stored in the block, but also
invalidate some potential double spends caused by these transactions. Usually
new block contains a substantial number of transactions from the pool, so it's
easier to make one pass over it only keeping valid items rather than remove
them one by one and make an additional pass to recheck inputs/witnesses.
It doesn't harm as we have transactions naturally ordered by fee anyway and it
makes managing them a little easier. This also makes slices store item itself
instead of pointers to it which reduces the pressure on the memory subsystem.
They shouldn't depend on the chain state and for the same transaction they
should always produce the same result. Thus, it makes no sense recalculating
them over and over again.
Eliminate races between tx checks and adding them to the mempool, ensure the
chain doesn't change while we're working with the new tx. Ensure only one
block addition attempt could be in progress.
If blockchain is not closed, logging in defer can occur
after test has finished, which will lead to a panic with
"Log in goroutine after Test* has completed".
Error in Seek means something is terribly wrong (e.g. db was not opened) and
error drop is not the right thing to do, because caller
will continue working with the wrong view.
Turns out, our dApps use it a lot and we were going to the DB to get it which
is a useless waste of time. Technically we could also remove blockHeight here,
but not doing it at the moment as it's more involved.
It eliminates this time waste from the pprof graph, but doesn't change 1.4M ->
1.5M 100K mainnet block import test case in any noticeable way.
This solves two problems:
* adds support for shortened SYSCALL form that uses IDs (similar to #434, but
for NEO 2.0, supporting both forms), which is important for compatibility
with C# node and mainnet chain that uses it from some height
* reworks interop plugging to use callbacks rather than appending to the map,
these map mangling functions are clearly visible in the VM profiling
statistics and we want spawning a VM to be fast, so it makes sense
optimizing it. This change moves most of the work to the init() phase
making VM setup cheaper.
Caveats:
* InteropNameToID accepts `[]byte` because that's the thing we have in
SYSCALL processing and that's the most often usecase for it, it leads to
some conversions in other places but that's acceptable because those are
either tests or init()
* three getInterop functions are: `getDefaultVMInterop`, `getSystemInterop`
and `getNeoInterop`
Our 100K (1.4M->1.5M) block import time improves by ~4% with this change.
First of all, it was wrong, it was not checking for inputs really, it compared
tx hashes for some reason, second, when it did compare inputs it compared only
the PrevIndex part of them which is also wrong.
Also, there is absolutely no reason to go through GetVerifiedTransactions()
here, we don't need this copy of pointers and it can also be outdated by the
time we're to finish our check.
Before:
BenchmarkTXPerformanceTest-4
5000 485506 ns/op 65886 B/op 409 allocs/op
ok github.com/CityOfZion/neo-go/integration 3.212s
After:
enchmarkTXPerformanceTest-4
5000 371104 ns/op 44367 B/op 408 allocs/op
ok github.com/CityOfZion/neo-go/integration 2.712s
This simple change improves our BenchmarkTXPerformanceTest by 14%, just
because we don't waste time on reallocations during append().
Before:
10000 439754 ns/op 218859 B/op 428 allocs/op
ok github.com/CityOfZion/neo-go/integration 5.423s
After:
10000 369833 ns/op 87209 B/op 412 allocs/op
ok github.com/CityOfZion/neo-go/integration 4.612s
It's useless work being done before it's actually needed. These (updated with
new values) are going to be written with some kind of Put anyway, so writing
them here is just a waste of time.
We're spending a lot of time here, 100K blocks import starting at 1.4M, before
this patch:
real 4m17,748s
user 6m23,316s
sys 0m37,866s
After:
real 3m54,968s
user 5m56,547s
sys 0m39,398s
9% is quite a substantial improvement to justify this change.
Importing 100K blocks starting at 1.4M, before this patch:
real 6m0,356s
user 8m52,293s
sys 0m47,372s
After this patch:
real 4m17,748s
user 6m23,316s
sys 0m37,866s
Almost 30% better.
It's a getter function and even though it's quite fancy with its transactions
processing (for consensus operation) it shouldn't ever change the state of the
Blockchain. If we're to change anything here these changes may conflict with
the actual block processing later or may lead to broken state (if transactions
won't be approved for some reason).
go vet is not happy about them:
pkg/io/binaryReader.go:92:21: method ReadByte() byte should have signature ReadByte() (byte, error)
pkg/io/binaryWriter.go:75:21: method WriteByte(u8 byte) should have signature WriteByte(byte) error
This seriously improves the serialization/deserialization performance for
several reasons:
* no time spent in `binary` reflection
* no memory allocations being made on every read/write
* uses fast ReadBytes everywhere it's appropriate
It also makes Fixed8 Serializable just for convenience.
add dao which takes care about all CRUD operations on storage
remove blockchain state since everything is stored on change
remove storage operations from structs(entities)
move structs to entities package
This change (closely related to the neo-project/neo#1321 proposal) speeds up
1.4M mainnet blocks import by 30%. Basically, we're eliminating key decoding
for block's multisignature that has the same keys most of the time.
Things I don't like about this patch:
* yet another parameter for verifyHashAgainstScript()
* vm keys are not copied in/out
But it's rather simple and solves the problem for this particular case, so I
think it's worth it.
It makes very little sense having pointers here, these structures MUST have
some kind of key and this key is not gonna be wandering somewhere on its
own. Fixes a part of #519.
It reduces heap pressure a little for these elements as we don't have to
allocate/free them individually. And they're directly tied to transactions or
block, not being shared or anything like that, so it makes little sense for
them to be pointer-based. It only makes building transactions a little easier,
but that's obviously a minor usecase.
Before this patch on block import we could easily be spending more than 6
seconds out of 30 in Uint256 encoding for UnspentBalance, now it's completely
off the radar.
