feat: run benchmarks for ACIR proving

[TomAFrench]

Resolves #6060

I'm adding some proving benchmarks so we can track how long it takes to prove various ACIR programs using bb.

I'm just adding a single benchmark for now and can follow up with building out the full suite once this is shown to be working.

[AztecBot]

Benchmark results

Metrics with a significant change:

• protocol_circuit_input_size_in_bytes (public-kernel-setup): 112,309 (-18%)
• tx_pxe_processing_time_ms (0): 26,441 (-29%)
• tx_sequencer_processing_time_ms (0): 1,010 (-23%)
• protocol_circuit_witness_generation_time_in_ms (base-parity): 318 (+41%)
• protocol_circuit_witness_generation_time_in_ms (public-kernel-setup): 331 (-20%)
• protocol_circuit_witness_generation_time_in_ms (root-parity): 193 (+37%)
• protocol_circuit_witness_generation_time_in_ms (base-rollup): 7,113 (+48%)
• protocol_circuit_proving_time_in_ms (base-parity): 8,413 (+105%)
• protocol_circuit_proving_time_in_ms (public-kernel-setup): 624 (-34%)
• protocol_circuit_proving_time_in_ms (public-kernel-app-logic): 623 (-34%)
• protocol_circuit_proving_time_in_ms (root-parity): 98,860 (+160%)
• protocol_circuit_proving_time_in_ms (public-kernel-teardown): 755 (-20%)
• protocol_circuit_proving_time_in_ms (base-rollup): 90,124 (+110%)
• protocol_circuit_proving_time_in_ms (merge-rollup): 2,517 (+21%)
• protocol_circuit_proving_time_in_ms (root-rollup): 11,789 (+42%)
• protocol_circuit_proof_size_in_bytes (public-kernel-setup): 88,580 (-22%)
• protocol_circuit_proof_size_in_bytes (public-kernel-app-logic): 89,203 (-21%)
• protocol_circuit_proof_size_in_bytes (public-kernel-teardown): 93,248 (-18%)

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.

Proof generation

Each column represents the number of threads used in proof generation.

Metric	1 threads	4 threads	16 threads	32 threads	64 threads
proof_construction_time_sha256	5,412	1,550	693	737	755

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	6,680	6,692
l1_rollup_execution_gas	586,853	586,865	586,877
l2_block_processing_time_in_ms	1,605	6,041	12,062
l2_block_building_time_in_ms	35,433 (+1%)	139,305 (+1%)	278,080 (+1%)
l2_block_rollup_simulation_time_in_ms	35,214 (+1%)	138,520 (+1%)	276,549 (+1%)
l2_block_public_tx_process_time_in_ms	16,581 (+1%)	71,061 (+1%)	143,798 (+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	18,125 (+2%)	33,970 (+2%)
node_database_size_in_bytes	21,368,912	37,982,288
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	161	3,611 (+1%)	22,074 (-1%)	20,382	61,571 (-1%)	85,440	2,603	1,048,576
private-kernel-inner	532	4,356	45,555	88,820 (-1%)	61,388 (-2%)	85,440	2,603	2,097,152
private-kernel-reset	987	6,547 (+1%)	87,935 (-1%)	325,902 (-1%)	61,571 (-1%)	85,440	2,603	4,194,304
private-kernel-tail	538 (+2%)	4,809 (-1%)	39,411 (+2%)	87,288	79,451	10,656	266	2,097,152
base-parity	8.70 (+7%)	⚠️ 318 (+41%)	⚠️ 8,413 (+105%)	128	64.0	2,208	2.00	131,072
root-parity	52.0 (+2%)	⚠️ 193 (+37%)	⚠️ 98,860 (+160%)	27,064	64.0	2,720	18.0	2,097,152
base-rollup	925 (+1%)	⚠️ 7,113 (+48%)	⚠️ 90,124 (+110%)	111,124	925	3,104	30.0	2,097,152
root-rollup	72.8 (+1%)	186 (+9%)	⚠️ 11,789 (+42%)	11,454	789	3,424	40.0	524,288
public-kernel-app-logic	193 (+1%)	295 (-3%)	⚠️ 623 (-34%)	96,421	84,613	⚠️ 89,203 (-21%)	3,479	4,096
public-kernel-tail	877 (-1%)	897 (+1%)	1,019 (-9%)	380,722 (-2%)	7,721	10,128	250	512
public-kernel-setup	142 (-14%)	⚠️ 331 (-20%)	⚠️ 624 (-34%)	⚠️ 112,309 (-18%)	75,762 (-11%)	⚠️ 88,580 (-22%)	3,479	4,096
public-kernel-teardown	150 (-11%)	384 (-7%)	⚠️ 755 (-20%)	121,137 (-15%)	77,762 (-8%)	⚠️ 93,248 (-18%)	3,479	4,096
private-kernel-ordering	342	N/A	N/A	61,368	39,490	N/A	N/A	N/A
merge-rollup	6.89 (+2%)	100 (+3%)	⚠️ 2,517 (+21%)	2,696	925	3,104	30.0	65,536
private-kernel-tail-to-public	N/A	10,877 (+6%)	68,776 (+1%)	N/A	N/A	113,984	3,495	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	15,520	48,052 (+1%)	2,097,152	418
Test:emit_nullifier	15,520	2,717 (+7%)	65,536	418
FPC:fee_entrypoint_public	15,520	16,467 (+3%)	524,288	418
Token:transfer	15,520	39,575 (+2%)	2,097,152	418
FPC:fee_entrypoint_private	15,520	9,027 (+1%)	524,288	418
Token:unshield	15,520	37,047 (+1%)	2,097,152	418
SchnorrAccount:spend_private_authwit	15,520	2,650 (+2%)	131,072	418

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	13.2	21.4	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	0.770	0.660	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	61.6 (-1%)	97.3 (+1%)	315	614 (+1%)	1,192	2,367	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 (-1%)	0.601 (+1%)	0.572	0.573 (+1%)	0.565	0.566	N/A
batch_insert_into_indexed_tree_20_depth_ms	N/A	N/A	74.2 (+1%)	144 (+1%)	454	901 (+1%)	1,774	3,547	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	0.646	0.647	0.618 (+1%)	0.619 (+1%)	0.616	0.617	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	81.2 (+1%)
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	0.711

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	83,507 (-1%)	765,712 (+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	⚠️ 26,441 (-29%)	3,690 (+2%)	113,939

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	⚠️ 1,010 (-23%)	2,280 (+1%)	1,660 (-13%)	2,518 (-11%)	1,627 (-11%)

[TomAFrench]

As mentioned above, this PR is set up to track regular Noir programs to benchmark against circom rather than testing performance of proving for the purposes of the Aztec network so it's testing a different case to the other proving benchmark PRs.

This PR is just to show the benchmarks are hooked up correctly, I'll follow up with another PR which fills out the suite of programs we want to test.

[github-actions]

Changes to circuit sizes

Generated at commit: 8c79e4672095106d57378142192ff6f068c04cca, compared to commit: 22ad5a5728afce5dcf32c8e6d8025691081e0de1

🧾 Summary (100% most significant diffs)

Program	ACIR opcodes (+/-)	%	Circuit size (+/-)	%
private_kernel_tail_to_public_simulated	0 ➖	0.00%	+4 ❌	+0.11%
public_kernel_app_logic_simulated	0 ➖	0.00%	+4 ❌	+0.11%
public_kernel_setup_simulated	0 ➖	0.00%	+4 ❌	+0.11%
public_kernel_teardown_simulated	0 ➖	0.00%	+4 ❌	+0.11%
private_kernel_tail_to_public	+1,448 ❌	+0.24%	+1,760 ❌	+0.08%
private_kernel_init_simulated	0 ➖	0.00%	+2 ❌	+0.08%
private_kernel_inner_simulated	0 ➖	0.00%	+2 ❌	+0.08%
private_kernel_reset_simulated	0 ➖	0.00%	+2 ❌	+0.08%
private_kernel_inner	+15 ❌	+0.01%	+940 ❌	+0.06%
public_kernel_app_logic	+133 ❌	+0.05%	+172 ❌	+0.02%
public_kernel_tail	-78 ✅	-0.01%	-78 ✅	-0.00%
public_kernel_teardown	-74 ✅	-0.03%	-81 ✅	-0.01%
public_kernel_setup	-85 ✅	-0.04%	-91 ✅	-0.01%
private_kernel_tail	-202 ✅	-0.10%	-616 ✅	-0.05%
private_kernel_init	+12 ❌	+0.01%	-579 ✅	-0.07%
private_kernel_reset	+255 ❌	+0.10%	-2,444 ✅	-0.10%
rollup_base	-98 ✅	-0.05%	-4,100 ✅	-0.23%

---

Full diff report 👇

Program	ACIR opcodes (+/-)	%	Circuit size (+/-)	%
private_kernel_tail_to_public_simulated	1 (0)	0.00%	3,495 (+4)	+0.11%
public_kernel_app_logic_simulated	1 (0)	0.00%	3,495 (+4)	+0.11%
public_kernel_setup_simulated	1 (0)	0.00%	3,495 (+4)	+0.11%
public_kernel_teardown_simulated	1 (0)	0.00%	3,495 (+4)	+0.11%
private_kernel_tail_to_public	611,589 (+1,448)	+0.24%	2,119,608 (+1,760)	+0.08%
private_kernel_init_simulated	1 (0)	0.00%	2,603 (+2)	+0.08%
private_kernel_inner_simulated	1 (0)	0.00%	2,603 (+2)	+0.08%
private_kernel_reset_simulated	1 (0)	0.00%	2,603 (+2)	+0.08%
private_kernel_inner	259,525 (+15)	+0.01%	1,690,457 (+940)	+0.06%
public_kernel_app_logic	247,647 (+133)	+0.05%	789,364 (+172)	+0.02%
public_kernel_tail	1,029,192 (-78)	-0.01%	3,709,601 (-78)	-0.00%
public_kernel_teardown	220,273 (-74)	-0.03%	661,309 (-81)	-0.01%
public_kernel_setup	220,059 (-85)	-0.04%	661,008 (-91)	-0.01%
private_kernel_tail	194,412 (-202)	-0.10%	1,309,261 (-616)	-0.05%
private_kernel_init	225,424 (+12)	+0.01%	827,164 (-579)	-0.07%
private_kernel_reset	249,885 (+255)	+0.10%	2,486,261 (-2,444)	-0.10%
rollup_base	181,027 (-98)	-0.05%	1,772,488 (-4,100)	-0.23%