Bot releases are hidden (Show)
This release is focused on Cancun and EOF.
--trace
in evmone-t8n
#616
riscv32
architecture #700
DATALOAD
, DATALOADN
, DATASIZE
, DATACOPY
#586 #663 #717 #741
DELEGATECALL
from EOF to legacy contracts during execution #588
0x04
#632
RJUMPV
immediate argument meaning has been changed to "max index" #640
JUMPF
instruction and the non-returning functions #644
evmone-statetest
with goevmlab has been improved #658 #757
Published by chfast over 1 year ago
The highlights of this release are support for Shanghai execution specification upgrade and implementation of EOF "v1.0". There are also big enhancements to the EVM testing tools and infrastructure. In particular, we added the t8n command-line utility.
As it is tradition, the EVM performance has been improved as well. Comparing with the previous release using the "main" benchmark suite, the Baseline interpreter is now:
evmone-eofparse
and evmone-eofparsefuzz
tools for testing EOF validation. #568
evmc::Host
, state transition and block finalization for testing purposes. #484 #519 #575 #608 #609
evmone-t8n
— a command line utility for transaction execution and state transition testing. It allows executing and generating tests with cooperation of retesteth or execution-spec-tests. #552 #555 #558 #569 #583 #590 #591 #604 #606 #607 #612
EVMONE_X86_64_ARCH_LEVEL
to set the x86-64 microarchitecture level. On Linux and Windows this is set to x86-64-v2 by default. #548
evmone-statetest
:
evmone-bench
) have been migrated to external evm-benchmarks which use JSON State Test format. #513 #530
evmc::instructions
. #533 #534 #537
Published by chfast about 2 years ago
Published by chfast about 2 years ago
In this release we have been focused on improving performance of the Baseline interpreter. The end result is that the Baseline is 26% faster than in previous version 0.8.0 and 18% faster than the current Advanced interpreter while having over 8x smaller code analysis cost. The Baseline is now the default interpreter because it is simpler and has become better than the Advanced.
The Advanced also has got 4% faster than in the previous version.
All numbers are from running the "main" benchmark suite on 4.0 GHz Intel Haswell CPU, using the Clang 15 compiler.
Moreover, evmone now calculates gas refund and reports it back using EVMC 10 API.
Finally, the options O=2
and O=0
have been replaced by advanced
. See below for details.
PUSH0
instruction implementation (EIP-3855), enabled in Shanghai. #448 #432
O=0
(use Baseline) and O=2
(use Advanced) have been replaced with single option advanced
to use the non-default Advanced interpreter. #500
advanced
option. Reasons are explained in the introduction. #500
cgoto=no
option. #495
evmone-fuzzer
has removed aleth-interpreter as it is not maintained and lacks the latest EVM features. #453
Published by chfast about 3 years ago
Published by chfast about 3 years ago
Published by chfast over 3 years ago
trace
option in Baseline. #325
histogram
option is enabled in Baseline. #323
evmone-bench
now reports time of execution without code analysis under "execute" label. The EVMC-like analysis + execution invocation is reported as "total". #343
evmone-bench
has started utilizing evmc::MockedHost
which allows using state-access (e.g. SLOAD
/SSTORE
) instructions in benchmarks. #319
evmone::baseline
API. #314
ADDMOD
instruction. #345
KECCAK256
implementation. #332
Published by chfast over 3 years ago
JUMPDEST
analysis used by Baseline interpreter. #306 #308
JUMPDEST
analysis. #305
Published by chfast over 3 years ago
O=0
option. #261 #280
Published by chfast over 4 years ago
Published by chfast over 4 years ago
Published by chfast almost 5 years ago
Published by chfast almost 5 years ago
This release of evmone adds changes for the Istanbul EVM revision.
Published by chfast about 5 years ago
This release of evmone is binary compatible with 0.1 and delivers big performance improvements
– both code preprocessing and execution is ~66% faster (needs ~40% less time).
std::vector<intx::uint256>
— -8–16%. #79
Published by chfast about 5 years ago
Published by chfast over 5 years ago