Implement EIP-8038: "State-access gas cost update"

[chfast]

Reprice state access and split it into access and write components.
https://eips.ethereum.org/EIPS/eip-8038

• COLD_ACCOUNT_ACCESS rises 2600 -> 3000; COLD_STORAGE_ACCESS (2100)
  and WARM_ACCESS (100) are unchanged.
• New ACCOUNT_WRITE (9000) prices account mutations: CALL_VALUE becomes
  ACCOUNT_WRITE + CALL_STIPEND (11300), SELFDESTRUCT sending a balance
  to a dead beneficiary pays it, and CREATE/CREATE2 pay CREATE_ACCESS
  (12000) in place of the flat 32000.
• New STORAGE_WRITE (10000) makes the first SSTORE change to a slot
  cost 10100 for both set and reset, its cold surcharge now matches
  SLOAD, and the clear refund becomes 11616.
• EXTCODESIZE and EXTCODECOPY pay an extra WARM_ACCESS for their
  second database read.
• Access-list entries cost 2900 per address and 2000 per storage key.

[codspeed]

Merging this PR will improve performance by 2.71%

⚠️ Different runtime environments detected

Some benchmarks with significant performance changes were compared across different runtime environments,
which may affect the accuracy of the results.

Open the report in CodSpeed to investigate

⚡ 1 improved benchmark
✅ 128 untouched benchmarks

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
⚡	baseline/analyse/main/sha1_shifts	11.5 µs	11.2 µs	+2.71%

Tip

Curious why performance improved? Comment @codspeedbot explain why performance improved on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing amsterdam/8038 (1baca41) with master (4029845)}

[codecov]

Codecov Report

❌ Patch coverage is 95.12195% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 97.98%. Comparing base (4029845) to head (1baca41).

Files with missing lines	Patch %	Lines
lib/evmone/instructions.hpp	71.42%	1 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1731      +/-   ##
==========================================
- Coverage   97.99%   97.98%   -0.01%     
==========================================
  Files         181      182       +1     
  Lines       16759    16778      +19     
  Branches     3827     3832       +5     
==========================================
+ Hits        16423    16440      +17     
- Misses        249      250       +1     
- Partials       87       88       +1     

Flag	Coverage Δ
eest-develop	81.98% <76.92%> (-0.06%)	⬇️
eest-develop-gmp	25.90% <48.78%> (+0.02%)	⬆️
eest-legacy	17.12% <24.39%> (+0.02%)	⬆️
eest-libsecp256k1	28.11% <48.78%> (+0.02%)	⬆️
eest-stable	81.95% <76.92%> (-0.06%)	⬇️
evmone-unittests	94.39% <95.12%> (-0.01%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Components	Coverage Δ
core	95.94% <92.30%> (-0.03%)	⬇️
tooling	94.30% <ø> (ø)
tests	99.81% <100.00%> (+<0.01%)	⬆️

Files with missing lines	Coverage Δ
lib/evmone/instructions_calls.cpp	95.26% <100.00%> (+0.08%)	⬆️
lib/evmone/instructions_storage.cpp	93.93% <ø> (ø)
lib/evmone/instructions_traits.hpp	100.00% <100.00%> (ø)
test/state/state.cpp	98.59% <100.00%> (+0.03%)	⬆️
test/unittests/evm_storage_test.cpp	100.00% <100.00%> (ø)
...ttests/state_transition_eip7778_block_gas_test.cpp	100.00% <100.00%> (ø)
...ttests/state_transition_eip8037_state_gas_test.cpp	100.00% <100.00%> (ø)
test/unittests/state_transition_tx_test.cpp	100.00% <100.00%> (ø)
test/unittests/tooling_t8n_test.cpp	100.00% <100.00%> (ø)
lib/evmone/instructions.hpp	98.83% <71.42%> (-0.47%)	⬇️

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Copilot]

Copilot review overview

🟡 Changes recommended

The block RLP remains stale and Amsterdam authorization processing still uses legacy gas pricing.

Get a fresh assessment by requesting another Copilot review.

Review effort: Balanced
Findings: 1 · 1 · 1

Open (3)

• Update EIP-7702 authorization gas accounting for Amsterdam · New
• Regenerate RLP fixture with updated block header · New
• Add Amsterdam EXTCODE warm and cold gas tests · New

What changed in this PR

Implements Amsterdam’s EIP-8038 state-access gas repricing across opcode execution, storage accounting, transaction costs, and tests.

Changes:

• Adds named account/storage access and write costs.
• Reprices calls, creation, self-destruct, EXTCODE operations, SSTORE, and access lists.
• Updates gas expectations and blockchain fixtures.

File	Description
lib/​evmone/​instructions_traits.hpp	Defines EIP-8038 costs and opcode pricing.
lib/​evmone/​instructions.hpp	Applies account-access and SELFDESTRUCT charges.
lib/​evmone/​instructions_calls.cpp	Reprices calls and delegation access.
lib/​evmone/​instructions_storage.cpp	Updates SSTORE costs and refunds.
test/​state/​state.cpp	Reprices Amsterdam access lists.
test/​unittests/​tooling_t8n_test.cpp	Updates refund accounting expectation.
test/​unittests/​state_transition_tx_test.cpp	Updates access-list tests.
test/​unittests/​state_transition_eip8037_state_gas_test.cpp	Adjusts state-gas scenarios.
test/​unittests/​state_transition_eip7778_block_gas_test.cpp	Updates block-gas refund test.
test/​integration/​evmone-cli/​test/​blockchaintest/​eip7778_block_gas.json	Updates expected block outputs.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

Copilot review overview

🟡 Changes recommended

Several consensus-critical gas schedules and OOG branches lack direct regression coverage.

Get a fresh assessment by requesting another Copilot review.

Review effort: Balanced
Findings: 1 · 3 · 1

Open (5)

• Update EIP-7702 authorization gas accounting for Amsterdam
• Test Amsterdam SELFDESTRUCT gas charging and OOG ordering · New
• Extend storage refund schedule tests for Amsterdam · New
• Add Amsterdam opcode pricing boundary tests · New
• Add Amsterdam EXTCODE warm and cold gas tests

Resolved since last review (1)

• Regenerate RLP fixture with updated block header

[Copilot]

Copilot review overview

🟡 Changes recommended

Several consensus-critical Amsterdam gas paths lack direct boundary tests.

Get a fresh assessment by requesting another Copilot review.

Review effort: Balanced
Findings: 1 · 5 · 1

Open (7)

• Update EIP-7702 authorization gas accounting for Amsterdam
• Cover Amsterdam SELFDESTRUCT charge and beneficiary conditions · New
• Pin Amsterdam gas costs for account reads and value calls · New
• Add Amsterdam opcode pricing boundary tests
• Extend storage refund schedule tests for Amsterdam
• Test Amsterdam SELFDESTRUCT gas charging and OOG ordering
• Add Amsterdam EXTCODE warm and cold gas tests

[chfast]

@codspeedbot explain why performance improved

[codspeed]

@codspeedbot explain why performance improved

Short answer: this one is not caused by your change — it is a runner-environment artifact, and the +2.71% should be ignored.

What the profiles show

Comparing the flamegraphs of baseline/analyse/main/sha1_shifts between master (4029845) and 1baca41:

Frame	BASE	HEAD	Δ
_int_malloc (glibc, self)	2.29 µs	1.93 µs	−0.36 µs
unlink_chunk (glibc, self)	0.29 µs	0.23 µs	−0.06 µs
__memcpy_avx_unaligned_erms (glibc, self)	2.75 µs	2.86 µs	+0.11 µs
analyze_jumpdests (evmone, self)	1.910 µs	1.912 µs	~0
total	11.5 µs	11.2 µs	−0.31 µs

The entire gain comes from glibc's allocator; the evmone code in the benchmark is unchanged. analyze_jumpdests is identical down to its instruction component (829.6 ns vs 829.9 ns), i.e. it executed exactly the same work, and evmone::baseline::analyze self time is flat (86.7 ns vs 90.4 ns). Inside _int_malloc the shift is almost entirely in the memory component (2.1 µs → 1.8 µs), not instructions — a cache/layout effect, not less work.

Why the environment differs

The comparison report flags a toolchain difference between the two runs: the build IDs of libc.so.6, libm.so.6 and ld-linux-x86-64.so.2 changed between base and head, so the two runs linked against different glibc builds. A different glibc gives a different malloc code path and a different memory layout, which in Simulation mode changes the modeled cache behaviour of the allocator — exactly where the 0.3 µs lives. All other baseline/analyse/* benchmarks stayed flat; sha1_shifts is simply the one whose allocation sizes fell on the right side of the noise threshold.

Why the PR cannot affect this benchmark anyway

The diff touches only lib/evmone/instructions.hpp, instructions_calls.cpp, instructions_storage.cpp, instructions_traits.hpp and test/state code. baseline/analyse only exercises evmone::baseline::analyze, which allocates the padded code buffer, copies the code and scans for JUMPDESTs — it never consults the gas cost tables, so repricing state access has no reachable effect on it. Consistently, every baseline/execute/* benchmark is untouched too: none of them hits a state-access opcode.

Nothing to fix here, and no changes were pushed to the branch.

Summary

• Warming up the performance wizard...
• Investigate the sha1_shifts analyse improvement — Compared flamegraphs: the whole delta sits in glibc _int_malloc, user code is bit-for-bit identical
• Check whether the PR can affect code analysis — PR touches no file on the baseline analysis path; analyze() is gas-cost independent
• Reply with the explanation — Reply with the explanation
• Handling comment — Comment handling finished successfully

_{Was this response helpful? React with 👍 or 👎 to help us improve CodSpeedBot.}

[Copilot]

Copilot review overview

🔵 Needs a closer look

Consensus-critical gas accounting changes require final human validation against the EIP specification and execution tests.

Review effort: Balanced
Findings: None

Resolved since last review (7)

• Update EIP-7702 authorization gas accounting for Amsterdam
• Pin Amsterdam gas costs for account reads and value calls
• Cover Amsterdam SELFDESTRUCT charge and beneficiary conditions
• Add Amsterdam opcode pricing boundary tests
• Extend storage refund schedule tests for Amsterdam
• Test Amsterdam SELFDESTRUCT gas charging and OOG ordering
• Add Amsterdam EXTCODE warm and cold gas tests