feat: private da gas metering

[just-mitch]

Context

At the end of the private kernel, we either use the PrivateKernelTail or PrivateKernelTailToPublic:
if PrivateKernelTail, we are heading straight to the base rollup because there is no public component and we have a CombinedAccumulatedData struct as part of the public inputs, called end, which has a gas_used (which has L2 and DA gas); if PrivateKernelTailToPublic, there is a public component, so we split the side effects from private execution into end and end_non_revertible, each of which has a gas_used.

Without this change, we were only populating end.gas_used to be equal to the teardown_gas which the used supplied for the transaction.

With this change

In PrivateKernelTail

We compute the amount of "metered gas" in the PrivateAccumulatedDataBuilder. This returns the amount of DA and L2 gas the transaction has consumed as part of private execution. L2 gas is presently zero, but that will likely change (see the warning under the L2 gas section here).

To this metered Gas, it adds the teardown gas and a "tx overhead" Gas, which represents flat, fixed costs associated with every transaction; it "tx overhead" presently just has a DA component.

In PrivateKernelTailToPublic

We now sum "metered" gas associated with side effects while splitting non/revertible side effects. The revertible gas used is initialized with the teardown gas, and the "tx overhead" gas is added to non_revertible.

See spec for more info.

[just-mitch]

• feat: enforce gas limits from private kernels #6105
• feat: private da gas metering #6103 👈
• master

This stack of pull requests is managed by Graphite. Learn more about stacking.

Join @just-mitch and the rest of your teammates on Graphite

[github-actions]

Changes to circuit sizes

Generated at commit: 5f2c96df955b52752eebb93bf809ca7eb23da663, compared to commit: 1a89d1a8bfa05fbac9904adc63b57ba94896c8d6

🧾 Summary (100% most significant diffs)

Program	ACIR opcodes (+/-)	%	Circuit size (+/-)	%
private_kernel_tail	+39,085 ❌	+56.29%	+39,990 ❌	+11.51%
private_kernel_tail_to_public	+3,608 ❌	+1.02%	+4,322 ❌	+0.53%

---

Full diff report 👇

Program	ACIR opcodes (+/-)	%	Circuit size (+/-)	%
private_kernel_tail	108,523 (+39,085)	+56.29%	387,541 (+39,990)	+11.51%
private_kernel_tail_to_public	357,466 (+3,608)	+1.02%	825,275 (+4,322)	+0.53%

[AztecBot]

Benchmark results

Metrics with a significant change:

• tx_sequencer_processing_time_ms (5): 2,599 (+18%)
• tx_sequencer_processing_time_ms (8): 1,310 (+17%)

Detailed results

All benchmarks are run on txs on the Benchmarking contract on the repository. Each tx consists of a batch call to create_note and increment_balance, which guarantees that each tx has a private call, a nested private call, a public call, and a nested public call, as well as an emitted private note, an unencrypted log, and public storage read and write.

This benchmark source data is available in JSON format on S3 here.

L2 block published to L1

Each column represents the number of txs on an L2 block published to L1.

Metric	8 txs	32 txs	64 txs
l1_rollup_calldata_size_in_bytes	740	740	740
l1_rollup_calldata_gas	6,668	6,668	6,692
l1_rollup_execution_gas	586,800	586,800	586,824
l2_block_processing_time_in_ms	1,448 (-3%)	5,387 (-4%)	10,330
l2_block_building_time_in_ms	22,562	88,645 (+1%)	176,859 (+1%)
l2_block_rollup_simulation_time_in_ms	22,358	87,891 (+1%)	175,388 (+1%)
l2_block_public_tx_process_time_in_ms	8,966 (+1%)	34,632 (+1%)	69,298 (+1%)

L2 chain processing

Each column represents the number of blocks on the L2 chain where each block has 16 txs.

Metric	5 blocks	10 blocks
node_history_sync_time_in_ms	17,275 (+6%)	30,435 (+1%)
node_database_size_in_bytes	20,758,608	37,417,040
pxe_database_size_in_bytes	29,868	59,425

Circuits stats

Stats on running time and I/O sizes collected for every circuit run across all benchmarks.

Circuit	circuit_simulation_time_in_ms	circuit_input_size_in_bytes	circuit_output_size_in_bytes
private-kernel-init	174 (-1%)	46,819	28,822
private-kernel-inner	206	76,734	28,822
private-kernel-ordering	229 (+1%)	53,040	42,751
base-parity	69.6 (-2%)	128	64.0
base-rollup	16,361	111,062	925
root-parity	1,681 (+9%)	27,064	64.0
root-rollup	65.7 (+2%)	11,454	789
public-kernel-app-logic	1,977	54,321	22,677
public-kernel-tail	4,831	159,770	3,898
merge-rollup	6,485	2,696	925
public-kernel-teardown	170 (+6%)	54,321	30,042
public-kernel-setup	228 (+3%)	54,321	30,042

Tree insertion stats

The duration to insert a fixed batch of leaves into each tree type.

Metric	1 leaves	16 leaves	64 leaves	128 leaves	512 leaves	1024 leaves	2048 leaves	4096 leaves	32 leaves
batch_insert_into_append_only_tree_16_depth_ms	11.1 (+1%)	17.8	N/A	N/A	N/A	N/A	N/A	N/A	N/A
batch_insert_into_append_only_tree_16_depth_hash_count	16.7	31.8	N/A	N/A	N/A	N/A	N/A	N/A	N/A
batch_insert_into_append_only_tree_16_depth_hash_ms	0.646 (+1%)	0.548	N/A	N/A	N/A	N/A	N/A	N/A	N/A
batch_insert_into_append_only_tree_32_depth_ms	N/A	N/A	52.1 (+1%)	80.0	262 (+1%)	499	984 (-2%)	1,956	N/A
batch_insert_into_append_only_tree_32_depth_hash_count	N/A	N/A	95.9	159	543	1,055	2,079	4,127	N/A
batch_insert_into_append_only_tree_32_depth_hash_ms	N/A	N/A	0.532 (+1%)	0.494	0.477 (+1%)	0.467	0.466 (-2%)	0.466	N/A
batch_insert_into_indexed_tree_20_depth_ms	N/A	N/A	60.8	120 (+1%)	375	740 (+1%)	1,470	2,927	N/A
batch_insert_into_indexed_tree_20_depth_hash_count	N/A	N/A	105	208	692	1,363	2,707	5,395	N/A
batch_insert_into_indexed_tree_20_depth_hash_ms	N/A	N/A	0.536	0.538	0.510	0.510 (+1%)	0.507	0.509	N/A
batch_insert_into_indexed_tree_40_depth_ms	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	66.7
batch_insert_into_indexed_tree_40_depth_hash_count	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	108
batch_insert_into_indexed_tree_40_depth_hash_ms	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	0.586

Miscellaneous

Transaction sizes based on how many contract classes are registered in the tx.

Metric	0 registered classes	1 registered classes
tx_size_in_bytes	44,613	535,433

Transaction size based on fee payment method

Metric	native fee payment method	fpc_public fee payment method	fpc_private fee payment method
tx_with_fee_size_in_bytes	897	1,145	1,345

Transaction processing duration by data writes.

Metric	0 new note hashes	1 new note hashes	2 new note hashes
tx_pxe_processing_time_ms	1,459 (+3%)	1,010	3,662 (+1%)

Metric	1 public data writes	2 public data writes	3 public data writes	4 public data writes	5 public data writes	8 public data writes
tx_sequencer_processing_time_ms	878	805 (+1%)	1,429	939	⚠️ 2,599 (+18%)	⚠️ 1,310 (+17%)

[spalladino]

LGTM! Just added a comment on a potential optimization, can be deferred for later.

[spalladino]

Does this builder have empty items? I'd expect it to contain only non-empty, since it uses BoundedVecs (otherwise, we could well use fixed-length arrays). We should be able to remove the loop and emptiness check, and just check the BoundedVec's length.

[just-mitch]

Yeah I think both this and the way we split/meter the non/revertible sets are candidates for significant optimization passes

[spalladino]

Hah, I hadn't managed to get this to work. Nice!