In the current VM opcode dispatcher, a two-level case statement is
generated that first matches the opcode and then uses another nested
case statement to select the actual implementation based on which fork
it is, causing the dispatcher to grow by `O(opcodes) * O(forks)`.
The fork does not change between instructions causing significant
inefficiency for this approach - not only because it repeats the fork
lookup but also because of code size bloat and missed optimizations.
A second source of inefficiency in dispatching is the tracer code which
in the vast majority of cases is disabled but nevertheless sees multiple
conditionals being evaluated for each instruction only to remain
disabled throughout exeuction.
This PR rewrites the opcode dispatcher macro to generate a separate
dispatcher for each fork and tracer setting and goes on to pick the
right one at the start of the computation.
This has many advantages:
* much smaller dispatcher
* easier to compile
* better inlining
* fewer pointlessly repeated instruction
* simplified macro (!)
* slow "low-compiler-memory" dispatcher code can be removed
Net block import improvement at about 4-6% depending on the contract -
synthetic EVM benchmnarks would show an even better result most likely.
The reverse slot hash mechanism causes quite a bit of database traffic
but is broadly not useful except for iterating the storage of an
account, something that a validator never does (it's used by the
tracers).
This flag adds one more thing that is not stored in the database, to be
explored more comprehensively when designing full, validator and archive
modes with different pruning options in the future.
`ldb` says this is 60gb of data (!):
```
ldb --db=. --ignore_unknown_options --column_family=KvtGen approxsize
--hex --from=0x05
--to=0x05ffffffffffffffffffffffffffffffffffffffffffffff
66488353954
```
* Use block number or timestamp to determine fork rules
Avoid confusion raised by `forkGTE` usage where block informations are present.
* Get rid of forkGTE
Introduce a new `StoData` payload type similar to `AccountData`
* slightly more efficient storage format
* typed api
* fewer seqs
* fix encoding docs - it wasn't rlp after all :)
This significantly speeds up block import at the cost of less protection
against invalid data, potentially resulting in an invalid database
getting stored.
The risk is small given that import is used only for validated data -
evaluating the right level of of validation vs performance is left for a
future PR.
A side effect of this approach is that there is no cached stated root in
the database - computing it currently requires a lot of memory since the
intermediate roots get cached in memory in full while the computation is
ongoing - a future PR will need to address this deficiency, for example
by streaming the already-computed hashes directly to the database.
* Normalised storage tree addressing in function prototypes
detail:
Argument list is always `<db> <account-path> <slot-path> ..` with
both path arguments as `openArray[]`
* Remove cruft
* CoreDb internally Use full account paths rather than addresses
* Update API logging
* Use hashed account address only in prototypes
why:
This avoids unnecessary repeated hashing of the same account address.
The burden of doing that is upon the application. In the case here,
the ledger caches all kinds of stuff anyway so it is common sense to
exploit that for account address hashes.
caveat:
Using `openArray[byte]` argument types for hashed accounts is inherently
fragile. In non-release mode, a length verification `doAssert` is
enabled by default.
* No accPath in data record (use `AristoAccount` as `CoreDbAccount`)
* Remove now unused `eAddr` field from ledger `AccountRef` type
why:
Is duplicate of lookup key
* Avoid merging the account record/statement in the ledger twice.
It is common for many accounts to share the same code - at the database
level, code is stored by hash meaning only one copy exists per unique
program but when loaded in memory, a copy is made for each account.
Further, every time we execute the code, it must be scanned for invalid
jump destinations which slows down EVM exeuction.
Finally, the extcodesize call causes code to be loaded even if only the
size is needed.
This PR improves on all these points by introducing a shared
CodeBytesRef type whose code section is immutable and that can be shared
between accounts. Further, a dedicated `len` API call is added so that
the EXTCODESIZE opcode can operate without polluting the GC and code
cache, for cases where only the size is requested - rocksdb will in this
case cache the code itself in the row cache meaning that lookup of the
code itself remains fast when length is asked for first.
With 16k code entries, there's a 90% hit rate which goes up to 99%
during the 2.3M attack - the cache significantly lowers memory
consumption and execution time not only during this event but across the
board.
State lookups potentially trigger expensive re-hashings - this is the
first of several steps to remove the unnecessary ones from the general
flow of block processing
* avoid re-reading parent block header from database when it's already
in memory
* Bump nim-eth, nim-web3, nimbus-eth2
- Replace std.Option with results.Opt
- Fields name changes
* More fixes
* Fix Portal stream async raises and portal testnet Opt usage
* Bump eth + nimbus-eth2 + more fixes related to eth_types changes
* Fix in utp test app and nimbus-eth2 bump
* Fix test_blockchain_json rebase conflict
* Fix EVMC block_timestamp conversion plus commentary
---------
Co-authored-by: kdeme <kim.demey@gmail.com>
`initTable` is obsolete since nim 0.19 and can introduce significant
memory overhead while providing no benefit (since the table will be
grown to the default initial size on first use anyway).
In particular, aristo layers will not necessarily use all tables they
initialize, for exampe when many empty accounts are being created.
This PR consolidates the split header-body sequences into a single EthBlock
sequence and cleans up the fallout from that which significantly reduces
block processing overhead during import thanks to less garbage collection
and fewer copies of things all around.
Notably, since the number of headers must always match the number of bodies,
we also get rid of a pointless degree of freedom that in the future could
introduce unnecessary bugs.
* only read header and body from era file
* avoid several unnecessary copies along the block processing way
* simplify signatures, cleaning up unused arguemnts and returns
* use `stew/assign2` in a few strategic places where the generated
nim assignent is slow and add a few `move` to work around poor
analysis in nim 1.6 (will need to be revisited for 2.0)
```
stats-20240607_2223-a814aa0b.csv vs stats-20240608_0714-21c1d0a9.csv
bps_x bps_y tps_x tps_y bpsd tpsd timed
block_number
(498305, 713245] 1,540.52 1,809.73 2,361.58 2775.340189 17.63% 17.63% -14.92%
(713245, 928185] 730.36 865.26 1,715.90 2028.973852 18.01% 18.01% -15.21%
(928185, 1143126] 663.03 789.10 2,529.26 3032.490771 19.79% 19.79% -16.28%
(1143126, 1358066] 393.46 508.05 2,152.50 2777.578119 29.13% 29.13% -22.50%
(1358066, 1573007] 370.88 440.72 2,351.31 2791.896052 18.81% 18.81% -15.80%
(1573007, 1787947] 283.65 335.11 2,068.93 2441.373402 17.60% 17.60% -14.91%
(1787947, 2002888] 287.29 342.11 2,078.39 2474.179448 18.99% 18.99% -15.91%
(2002888, 2217828] 293.38 343.16 2,208.83 2584.77457 17.16% 17.16% -14.61%
(2217828, 2432769] 140.09 167.86 1,081.87 1296.336926 18.82% 18.82% -15.80%
blocks: 1934464, baseline: 3h13m1s, contender: 2h43m47s
bpsd (mean): 19.55%
tpsd (mean): 19.55%
Time (total): -29m13s, -15.14%
```
* Cleanup unneeded stateless and block witness code. Keeping MultiKeys which is used in the eth_getProofsByBlockNumber RPC endpoint which is needed for the Fluffy state network bridge.
* Rename generateWitness flag to collectWitnessData to better describe what the flag does. We only collect the keys of the touched accounts and storage slots but no block witness generation is supported for now.
* Move remaining stateless code into nimbus directory.
* Add vmstate parameter to ChainRef to fix test.
* Exclude *.in from check copyright year
---------
Co-authored-by: jangko <jangko128@gmail.com>
* Remove exception from evm memory
* Remove exception from gas meter
* Remove exception from stack
* Remove exception from precompiles
* Remove exception from gas_costs
* Remove exception from op handlers
* Remove exception from op dispatcher
* Remove exception from call_evm
* Remove exception from EVM
* Fix tools and tests
* Remove exception from EVMC
* fix evmc
* Fix evmc
* Remove remnants of async evm stuff
* Remove superflous error handling
* Proc to func
* Fix errors detected by CI
* Fix EVM op call stack usage
* REmove exception handling from getVmState
* Better error message instead of just doAssert
* Remove unused validation
* Remove superflous catchRaise
* Use results.expect instead of unsafeValue
`persist` is a hotspot when processing blocks because it is run at least
once per transaction and loops over the entire account cache every time.
Here, we introduce an extra `dirty` map that keeps track of all accounts
that need checking during `persist` which fixes the immediate
inefficiency, though probably this could benefit from a more thorough
review - we also get rid of the unused clearCache flag - we start with
a fresh cache on every fresh vmState.
* avoid unnecessary code hash comparisons
* avoid unnecessary copies when iterating
* use EMPTY_CODE_HASH throughout for code hash comparison
* Attempt to roll back stateless mode implementation in a single PR
why:
+ Stateless mode is not fully working and in the way
+ Single PR should make it feasible to investigate for a possible
re-implementation
* Fix copyright year
* Fix annotation for exception (evmc mode)
Jacek Sieka
andri lim
Jordan Hrycaj
Miran
web3-developer
tersec