feat: improved ClientIvc

[ledwards2225]

The goal is to make ClientIvc more flexible and user friendly, primarily through making recursive folding verification an internal component to the class. E.g. given an arbitrary collection of circuits in container circuits, I can now simply do this:

ClientIvc ivc;
for (auto circuit : circuits) {
    ivc.accumulate(circuit);
}
proof = ivc.prove();

where proof, if verified, establishes knowledge of a witness satisfying all 10 circuits. Previously this was not the case unless the required recursive folding verifiers were baked into the circuits being accumulated (as was the case with the kernel). Now, accumulate automatically appends a recursive folding verifier to each circuit (beginning with the third call to accumulate since prior to that no fold proof exists). Note that this is already how we were handling the recursive merge verification.

This has the advantage of removing details of the IVC scheme from the kernel itself, generalizing ClientIvc for use on an arbitrary set of circuits, and improving the interface in all cases.

As expected, the benchmark does not change in any substantial way. (To the extent it does change, its because the gate counts in the circuits have changed slightly due to this new structure).

--------------------------------------------------------------------------------
Benchmark                      Time             CPU   Iterations UserCounters...
--------------------------------------------------------------------------------
ClientIVCBench/Full/6      17908 ms        12848 ms            1

[AztecBot]

Benchmark results

Metrics with a significant change:

• protocol_circuit_input_size_in_bytes (public-kernel-setup): 109,385 (-20%)
• protocol_circuit_input_size_in_bytes (public-kernel-teardown): 118,792 (-16%)
• tx_pxe_processing_time_ms (0): 26,285 (-29%)
• tx_sequencer_processing_time_ms (0): 1,016 (-22%)
• protocol_circuit_witness_generation_time_in_ms (base-parity): 587 (-23%)
• protocol_circuit_witness_generation_time_in_ms (public-kernel-setup): 329 (-20%)
• protocol_circuit_witness_generation_time_in_ms (root-parity): 185 (+47%)
• protocol_circuit_witness_generation_time_in_ms (base-rollup): 7,173 (+49%)
• protocol_circuit_witness_generation_time_in_ms (public-kernel-tail): 708 (-24%)
• protocol_circuit_witness_generation_time_in_ms (public-kernel-teardown): 368 (-21%)
• protocol_circuit_witness_generation_time_in_ms (merge-rollup): 117 (+34%)
• protocol_circuit_proving_time_in_ms (base-parity): 8,545 (+135%)
• protocol_circuit_proving_time_in_ms (public-kernel-app-logic): 525 (-48%)
• protocol_circuit_proving_time_in_ms (public-kernel-setup): 578 (-38%)
• protocol_circuit_proving_time_in_ms (root-parity): 98,487 (+148%)
• protocol_circuit_proving_time_in_ms (public-kernel-tail): 755 (-30%)
• protocol_circuit_proving_time_in_ms (public-kernel-teardown): 631 (-40%)
• protocol_circuit_proving_time_in_ms (base-rollup): 91,297 (+110%)
• protocol_circuit_proving_time_in_ms (merge-rollup): 2,531 (+56%)
• protocol_circuit_proving_time_in_ms (root-rollup): 12,030 (+44%)
• protocol_circuit_proof_size_in_bytes (public-kernel-app-logic): 84,544 (-25%)
• protocol_circuit_proof_size_in_bytes (public-kernel-setup): 83,805 (-26%)
• protocol_circuit_proof_size_in_bytes (public-kernel-teardown): 89,344 (-21%)

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	6,680	6,680	6,692
l1_rollup_execution_gas	586,865	586,865	586,877
l2_block_processing_time_in_ms	1,601 (-1%)	6,017 (-2%)	11,962 (-1%)
l2_block_building_time_in_ms	35,350 (-1%)	138,436	275,644 (-1%)
l2_block_rollup_simulation_time_in_ms	35,137 (-1%)	137,663	274,062 (-1%)
l2_block_public_tx_process_time_in_ms	16,648	70,617	142,424 (-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	17,851 (-1%)	33,261 (-1%)
node_database_size_in_bytes	21,352,528	38,039,632
pxe_database_size_in_bytes	29,868	59,425

Circuits stats

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

Circuit	protocol_circuit_simulation_time_in_ms	protocol_circuit_witness_generation_time_in_ms	protocol_circuit_proving_time_in_ms	protocol_circuit_input_size_in_bytes	protocol_circuit_output_size_in_bytes	protocol_circuit_proof_size_in_bytes	protocol_circuit_num_public_inputs	protocol_circuit_size_in_gates
private-kernel-init	159 (-3%)	3,616 (+1%)	23,118 (+3%)	20,261 (-1%)	61,193 (-2%)	85,376	2,601	1,048,576
private-kernel-inner	522 (-3%)	4,706 (+2%)	45,491 (+1%)	88,242 (-2%)	60,943 (-2%)	85,376	2,601	2,097,152
private-kernel-reset	972 (-3%)	6,326 (-9%)	88,861 (+2%)	325,444 (-2%)	61,193 (-2%)	85,376	2,601	4,194,304
private-kernel-tail	531 (-1%)	4,943	38,021	87,224	79,333	10,656	266	2,097,152
base-parity	8.70 (+5%)	⚠️ 587 (-23%)	⚠️ 8,545 (+135%)	128	64.0	2,208	2.00	131,072
root-parity	51.4	⚠️ 185 (+47%)	⚠️ 98,487 (+148%)	27,064	64.0	2,720	18.0	2,097,152
base-rollup	908 (-1%)	⚠️ 7,173 (+49%)	⚠️ 91,297 (+110%)	111,124	925	3,104	30.0	2,097,152
root-rollup	72.6	180 (+7%)	⚠️ 12,030 (+44%)	11,454	789	3,424	40.0	524,288
public-kernel-app-logic	190 (-2%)	354 (+8%)	⚠️ 525 (-48%)	96,120 (-1%)	84,344 (-1%)	⚠️ 84,544 (-25%)	3,475	4,096
public-kernel-tail	871 (-2%)	⚠️ 708 (-24%)	⚠️ 755 (-30%)	380,330 (-2%)	7,721	10,128	250	512
public-kernel-setup	141 (-12%)	⚠️ 329 (-20%)	⚠️ 578 (-38%)	⚠️ 109,385 (-20%)	72,888 (-14%)	⚠️ 83,805 (-26%)	3,475	4,096
public-kernel-teardown	146 (-15%)	⚠️ 368 (-21%)	⚠️ 631 (-40%)	⚠️ 118,792 (-16%)	75,476 (-11%)	⚠️ 89,344 (-21%)	3,475	4,096
private-kernel-ordering	314	N/A	N/A	53,688	34,764	N/A	N/A	N/A
merge-rollup	6.59 (-4%)	⚠️ 117 (+34%)	⚠️ 2,531 (+56%)	2,696	925	3,104	30.0	65,536
private-kernel-tail-to-public	N/A	10,172 (-2%)	69,338 (+1%)	N/A	N/A	113,856	3,491	4,194,304

Stats on running time collected for app circuits

Function	app_circuit_proof_size_in_bytes	app_circuit_proving_time_in_ms	app_circuit_size_in_gates	app_circuit_num_public_inputs
SchnorrAccount:entrypoint	15,456	48,060 (-1%)	2,097,152	416
Test:emit_nullifier	15,456	2,426 (-2%)	65,536	416
FPC:fee_entrypoint_public	15,456	8,728 (+2%)	524,288	416
Token:transfer	15,456	39,694	2,097,152	416
FPC:fee_entrypoint_private	15,456	9,017	524,288	416
Token:unshield	15,456	37,198	2,097,152	416
SchnorrAccount:spend_private_authwit	15,456	2,682 (+2%)	131,072	416

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	13.1 (-1%)	21.4 (-1%)	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	0.768	0.659 (-1%)	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	61.5	96.7 (-1%)	314 (-1%)	607	1,192 (-1%)	2,360 (-1%)	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	0.632	0.599 (-1%)	0.572 (-1%)	0.569	0.566 (-1%)	0.565 (-1%)	N/A
batch_insert_into_indexed_tree_20_depth_ms	N/A	N/A	74.0	143 (-1%)	454 (-1%)	889	1,764 (-2%)	3,525 (-2%)	N/A
batch_insert_into_indexed_tree_20_depth_hash_count	N/A	N/A	106	208	692	1,363	2,707	5,395	N/A
batch_insert_into_indexed_tree_20_depth_hash_ms	N/A	N/A	0.645	0.644 (-1%)	0.619 (-1%)	0.615	0.613 (-1%)	0.615 (-1%)	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	80.7
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	109 (+1%)
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	0.708

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	83,255 (-1%)	764,759 (+14%)

Transaction size based on fee payment method

| Metric | |
| - | |

Transaction processing duration by data writes.

Metric	0 new note hashes	1 new note hashes	2 new note hashes
tx_pxe_processing_time_ms	⚠️ 26,285 (-29%)	3,615 (-2%)	114,588 (+2%)

Metric	0 public data writes	1 public data writes	2 public data writes	4 public data writes	8 public data writes
tx_sequencer_processing_time_ms	⚠️ 1,016 (-22%)	2,237 (-1%)	1,658 (-13%)	2,508 (-12%)	1,622 (-10%)

[codygunton]

LGTM

[codygunton]

You're intentionally not passing by reference here, right?

[ledwards2225]

Yes, added an extra comment to ensure this is clear