[codex] add evm-only executor load test harness

[codchen]

Summary

Adds a standalone evmonly-loadtest command that feeds generated EVM-only blocks into the EVM-only executor with generated genesis state, configurable result sinks, and Prometheus/stdout throughput metrics.

The executor owns the result-sink boundary. giga/evmonly exposes ResultSink and WithResultSink, and executor completion invokes the sink for the produced StateChangeSet and receipts before returning. The loadtest harness implements discard/file sink modes through that interface.

This branch also pipelines stateless sender recovery ahead of execution:

• Executor.PrepareBlock decodes raw tx RLP and recovers senders into PreparedBlock.
• Executor.ExecutePreparedBlock executes an already prepared block.
• ExecuteBlock remains the convenience prepare-then-execute path.
• evmonly-loadtest now has --prepare-workers; raw blocks flow through parallel prepare workers into an ordered prepared-block queue consumed by the single external executor worker.

The harness supports optimistic no-overlap native transfers and ERC20 transfers. It defaults to unique senders and recipients, supports --prebuild-blocks, and includes a lightweight async file sink via --result-sink=file --persist-dir=<dir>. Persistence records are append-only RLP records for changesets and receipts, and the sink removes files on normal completion, execution errors, and SIGINT/SIGTERM.

Metrics include block input/prepared/finished throughput, prepared tx/s, executed tx/s, gas/s, OCC attempts/fallbacks/conflicts, result-sink queue depth, enqueue wait, write time, bytes written, and record counts. sink_enqueue_wait is the backpressure signal: if it rises, executor workers are blocked on persistence queue capacity.

Validation

• go test ./giga/evmonly/...
• Native smoke:

go run ./giga/evmonly/cmd/evmonly-loadtest \
  --metrics-addr= \
  --report-interval=0 \
  --prebuild-blocks \
  --blocks=30 \
  --txs-per-block=1000 \
  --builders=8 \
  --prepare-workers=8 \
  --workers=1 \
  --executor-workers=12 \
  --gas-price-wei=0 \
  --min-gas-price-wei=0 \
  --queue-size=64

Observed locally: 197,429 TPS, prepare_errors=0, errors=0.

• ERC20 smoke:

go run ./giga/evmonly/cmd/evmonly-loadtest \
  --metrics-addr= \
  --report-interval=0 \
  --prebuild-blocks \
  --blocks=30 \
  --txs-per-block=1000 \
  --builders=8 \
  --prepare-workers=8 \
  --workers=1 \
  --executor-workers=12 \
  --workload=erc20-transfer \
  --gas-price-wei=0 \
  --min-gas-price-wei=0 \
  --queue-size=64

Observed locally: 189,483 TPS, prepare_errors=0, errors=0.

EC2 Persistent Runs

Run on commit e312a845b8439fe28824e4bc3aef85cd298ecdd2 using a temporary c8i.48xlarge in us-east-1a, SMT disabled with CoreCount=96,ThreadsPerCore=1, Go 1.25.6, GOMAXPROCS=96, GOGC=400, one external executor worker, prebuilt raw blocks, unique senders/recipients, zero gas price/min gas price, and --result-sink=file. The temporary instance and security group were deleted after collecting logs.

Native transfer tuned command:

GOMAXPROCS=96 GOGC=400 /tmp/evmonly-loadtest \
  --metrics-addr= \
  --blocks=1000 \
  --txs-per-block=5000 \
  --prebuild-blocks \
  --builders=96 \
  --prepare-workers=24 \
  --workers=1 \
  --executor-workers=56 \
  --gas-price-wei=0 \
  --min-gas-price-wei=0 \
  --report-interval=5s \
  --queue-size=64 \
  --result-sink=file \
  --persist-dir="$DIR"

Observed:

prebuild complete elapsed=8.965s blocks=1000 txs=5000000 build_blocks/s=111.55 build_tx/s=557745.01
complete elapsed=27.043s input_blocks=1000 prepared_blocks=1000 prepared_txs=5000000 finished_blocks=1000 txs=5000000 gas=105000000000 prepare_errors=0 errors=0 occ_attempts=1000 occ_fallbacks=0 occ_conflicts=0 sink_queue=1 sink_enqueued=2000 sink_written=1998 sink_bytes=1727519751 sink_enqueue_wait=0s sink_enqueue_wait_events=0 sink_write=7.209933s avg_input_blocks/s=36.98 avg_prepared_blocks/s=36.98 avg_prepared_tx/s=184893.42 avg_finished_blocks/s=36.98 avg_tx/s=184893.42 avg_gas/s=3882761797.16
result sink close elapsed=97ms sink_queue=0 sink_enqueued=2000 sink_written=2000 sink_bytes=1729249000 sink_enqueue_wait=0s sink_enqueue_wait_events=0 sink_write=7.212607s

ERC20 transfer tuned command:

GOMAXPROCS=96 GOGC=400 /tmp/evmonly-loadtest \
  --metrics-addr= \
  --blocks=1000 \
  --txs-per-block=5000 \
  --prebuild-blocks \
  --builders=96 \
  --prepare-workers=24 \
  --workers=1 \
  --executor-workers=64 \
  --workload=erc20-transfer \
  --gas-price-wei=0 \
  --min-gas-price-wei=0 \
  --report-interval=5s \
  --queue-size=64 \
  --result-sink=file \
  --persist-dir="$DIR"

Observed:

prebuild complete elapsed=19.506s blocks=1000 txs=5000000 build_blocks/s=51.27 build_tx/s=256327.90
complete elapsed=29.693s input_blocks=1000 prepared_blocks=1000 prepared_txs=5000000 finished_blocks=1000 txs=5000000 gas=229960303044 prepare_errors=0 errors=0 occ_attempts=1000 occ_fallbacks=0 occ_conflicts=0 sink_queue=1 sink_enqueued=2000 sink_written=1998 sink_bytes=3146535315 sink_enqueue_wait=0s sink_enqueue_wait_events=0 sink_write=10.447174s avg_input_blocks/s=33.68 avg_prepared_blocks/s=33.68 avg_prepared_tx/s=168389.38 avg_finished_blocks/s=33.68 avg_tx/s=168389.38 avg_gas/s=7744574591.71
result sink close elapsed=167ms sink_queue=0 sink_enqueued=2000 sink_written=2000 sink_bytes=3149685000 sink_enqueue_wait=0s sink_enqueue_wait_events=0 sink_write=10.451667s

Tuning notes from the same EC2 instance:

• --queue-size=64 beat 512 by reducing heap/RSS; native improved from 170.2k to 178.1k TPS before worker tuning.
• Lowering --prepare-workers from 96 to 24 helped because recovery is no longer the steady-state bottleneck and otherwise competes with execution.
• Native short sweep best: --prepare-workers=24 --executor-workers=56, 194.0k TPS over 2.5M tx.
• ERC20 short sweep best: --prepare-workers=24 --executor-workers=64, 170.5k TPS over 2.5M tx.
• No persistent sink backpressure was observed in the tuned full runs: sink_enqueue_wait=0s.

Historical Non-Persistent 199.3k Repro

The earlier 199.3k EC2 benchmark was run before persistent sink and prepared pipeline changes, from commit 4ec8da52c, on a c8i.48xlarge in us-east-1a with SMT disabled, Go 1.25.6, one external executor worker, prebuilt blocks, unique senders/recipients, and zero gas price/min gas price.

GOMAXPROCS=96 GOGC=200 /tmp/evmonly-loadtest \
  --metrics-addr= \
  --blocks=1000 \
  --txs-per-block=5000 \
  --prebuild-blocks \
  --builders=96 \
  --workers=1 \
  --executor-workers=96 \
  --gas-price-wei=0 \
  --min-gas-price-wei=0 \
  --report-interval=5s

Observed:

prebuild complete elapsed=9.366s blocks=1000 txs=5000000 build_blocks/s=106.77 build_tx/s=533873.27
complete elapsed=25.09s input_blocks=1000 finished_blocks=1000 txs=5000000 gas=105000000000 errors=0 avg_input_blocks/s=39.86 avg_finished_blocks/s=39.86 avg_tx/s=199284.61 avg_gas/s=4184976805.58

EC2 Worker Pool / Pinning Follow-up

Commit 63b8662b4bc90936175181b1fff6347c502fb03f adds a persistent OCC worker pool and Linux worker pinning controls. Re-tested on a temporary c8i.48xlarge in us-east-1a, SMT disabled with CoreCount=96,ThreadsPerCore=1, Go 1.25.6, GOMAXPROCS=96, GOGC=400, prebuilt raw blocks, async file persistence, and the same 5M-tx shapes.

Native transfer, persistent OCC pool, unpinned:

complete elapsed=26.794s input_blocks=1000 prepared_blocks=1000 prepared_txs=5000000 finished_blocks=1000 txs=5000000 gas=105000000000 prepare_errors=0 errors=0 occ_attempts=1000 occ_fallbacks=0 occ_conflicts=0 sink_queue=1 sink_enqueued=2000 sink_written=1998 sink_bytes=1727519751 sink_enqueue_wait=0s sink_enqueue_wait_events=0 sink_write=7.416901s avg_input_blocks/s=37.32 avg_prepared_blocks/s=37.32 avg_prepared_tx/s=186607.41 avg_finished_blocks/s=37.32 avg_tx/s=186607.41 avg_gas/s=3918755519.49
Percent of CPU: 2404%

Native transfer, persistent OCC pool, --pin-workers --prepare-cpu-offset=0 --executor-cpu-offset=24:

complete elapsed=26.783s input_blocks=1000 prepared_blocks=1000 prepared_txs=5000000 finished_blocks=1000 txs=5000000 gas=105000000000 prepare_errors=0 errors=0 occ_attempts=1000 occ_fallbacks=0 occ_conflicts=0 sink_queue=1 sink_enqueued=2000 sink_written=1998 sink_bytes=1727519751 sink_enqueue_wait=0s sink_enqueue_wait_events=0 sink_write=7.68812s avg_input_blocks/s=37.34 avg_prepared_blocks/s=37.34 avg_prepared_tx/s=186682.92 avg_finished_blocks/s=37.34 avg_tx/s=186682.92 avg_gas/s=3920341280.53
Percent of CPU: 2283%

ERC20 transfer, persistent OCC pool, unpinned:

complete elapsed=29.325s input_blocks=1000 prepared_blocks=1000 prepared_txs=5000000 finished_blocks=1000 txs=5000000 gas=229960303044 prepare_errors=0 errors=0 occ_attempts=1000 occ_fallbacks=0 occ_conflicts=0 sink_queue=1 sink_enqueued=2000 sink_written=1998 sink_bytes=3146535315 sink_enqueue_wait=0s sink_enqueue_wait_events=0 sink_write=10.314016s avg_input_blocks/s=34.10 avg_prepared_blocks/s=34.10 avg_prepared_tx/s=170501.05 avg_finished_blocks/s=34.10 avg_tx/s=170501.05 avg_gas/s=7841694425.87
Percent of CPU: 2003%

ERC20 transfer, persistent OCC pool, pinned:

complete elapsed=30.25s input_blocks=1000 prepared_blocks=1000 prepared_txs=5000000 finished_blocks=1000 txs=5000000 gas=229960303044 prepare_errors=0 errors=0 occ_attempts=1000 occ_fallbacks=0 occ_conflicts=0 sink_queue=1 sink_enqueued=2000 sink_written=1998 sink_bytes=3146535315 sink_enqueue_wait=0s sink_enqueue_wait_events=0 sink_write=11.436487s avg_input_blocks/s=33.06 avg_prepared_blocks/s=33.06 avg_prepared_tx/s=165290.26 avg_finished_blocks/s=33.06 avg_tx/s=165290.26 avg_gas/s=7602039743.89
Percent of CPU: 2009%

Conclusion: the persistent OCC pool removes per-block worker creation overhead but does not materially increase core utilization. Native moved from ~184.9k to ~186.6k TPS, while CPU remained around 20-24 effective cores on a 96-core instance. Thread pinning did not help on this workload and slightly reduced CPU utilization in the native case.

[github-actions]

The latest Buf updates on your PR. Results from workflow Buf / buf (pull_request).

Build	Format	Lint	Breaking	Updated (UTC)
✅ passed	✅ passed	✅ passed	✅ passed	Jul 2, 2026, 9:54 AM

[codecov]

Codecov Report

❌ Patch coverage is 63.55463% with 527 lines in your changes missing coverage. Please review.
✅ Project coverage is 58.35%. Comparing base (9cb00d6) to head (63b8662).

Files with missing lines	Patch %	Lines
giga/evmonly/cmd/evmonly-loadtest/main.go	63.40%	402 Missing and 91 partials ⚠️
...only/cmd/evmonly-loadtest/worker_affinity_linux.go	0.00%	17 Missing ⚠️
giga/evmonly/occ.go	79.26%	16 Missing and 1 partial ⚠️

Additional details and impacted files

@@                          Coverage Diff                          @@
##           codex/evmonly-staking-dynamic-gas    #3658      +/-   ##
=====================================================================
+ Coverage                              58.15%   58.35%   +0.19%     
=====================================================================
  Files                                   2187     2182       -5     
  Lines                                 178718   178712       -6     
=====================================================================
+ Hits                                  103939   104291     +352     
+ Misses                                 65454    65234     -220     
+ Partials                                9325     9187     -138     

Flag	Coverage Δ
sei-chain-pr	69.07% <63.55%> (+12.21%)	⬆️
sei-db	70.41% <ø> (ø)

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

Files with missing lines	Coverage Δ
...only/cmd/evmonly-loadtest/worker_affinity_linux.go	0.00% <0.00%> (ø)
giga/evmonly/occ.go	78.94% <79.26%> (+6.26%)	⬆️
giga/evmonly/cmd/evmonly-loadtest/main.go	63.40% <63.40%> (ø)

... and 8 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.