feat: Add gate profiler for noir circuits

[sirasistant]

This PR:

• Modifies acir format to track constraint => original opcode and to store gates per opcode (only gates, no lookups ATM) when building a circuit
• Modifies the gates command to return this information along with the totals
• Adds a new profiler binary in noir tooling with a command generate a gates flamegraph
• Adds a small script in noir-projects to extract a private function as a regular noir artifact, to make it usable by the profiler. This is needed because transpiling makes our artifacts unreadable by noir tooling.

Example result: (altough github breaks the zoom/search functionality when uploading the svg)

Usage:

./target/release/noir-profiler gates-flamegraph --artifact-path PATH_TO_THE_ARTIFACT_JSON --backend-path BB_PATH  --output OUTPUT_FOLDER 

Example:

./target/release/noir-profiler gates-flamegraph --artifact-path ~/aztec-packages/noir-projects/noir-contracts/target/token_contract-Token-transfer.json --backend-path ~/aztec-packages/barretenberg/cpp/build/bin/bb  --output .

Or you can run it from the code in noir/noir-repo/tooling/profiler via cargo run instead of directly calling the built binary.

To use the small tool to extract private functions:

node extractFunctionAsNoirArtifact.js PATH_TO_CONTRACT_ARTIFACT FUNCTION_NAME

Example:

node extractFunctionAsNoirArtifact.js ./target/token_contract-Token.json transfer

[AztecBot]

Benchmark results

Metrics with a significant change:

• avm_simulation_time_ms (Token:mint_public): 566 (+704%)
• avm_simulation_time_ms (Token:assert_minter_and_mint): 74.5 (-67%)
• protocol_circuit_witness_generation_time_in_ms (private-kernel-inner): 1,231 (+22%)
• l2_block_processing_time_in_ms (16): 2,699 (-32%)

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_ms	5,705 (-1%)	1,547 (-5%)	706 (+2%)	758	770
proof_construction_time_sha256_30_ms	11,664 (+2%)	3,147 (-1%)	1,399 (+2%)	1,432 (+2%)	1,469 (+1%)
proof_construction_time_sha256_100_ms	43,621 (-1%)	11,712 (-1%)	5,430 (-1%)	5,425 (-1%)	5,347
proof_construction_time_poseidon_hash_ms	78.0	34.0	34.0	58.0	89.0 (+2%)
proof_construction_time_poseidon_hash_30_ms	1,515	416	201 (+1%)	219 (-5%)	265 (-3%)
proof_construction_time_poseidon_hash_100_ms	5,726	1,562	718 (-1%)	760 (-2%)	788 (-1%)

L2 block published to L1

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

Metric	4 txs	8 txs	16 txs
l1_rollup_calldata_size_in_bytes	1,412	1,412	1,412
l1_rollup_calldata_gas	9,440	9,448	9,452
l1_rollup_execution_gas	610,261	610,269	610,273
l2_block_processing_time_in_ms	750	1,397 (-1%)	⚠️ 2,699 (-32%)
l2_block_building_time_in_ms	23,012 (-8%)	45,888 (-1%)	90,285 (-9%)
l2_block_rollup_simulation_time_in_ms	22,919 (-8%)	45,601 (-1%)	89,945 (-9%)
l2_block_public_tx_process_time_in_ms	19,652 (-9%)	42,139 (-1%)	86,530 (-10%)

L2 chain processing

Each column represents the number of blocks on the L2 chain where each block has 8 txs.

Metric	3 blocks	5 blocks
node_history_sync_time_in_ms	6,384 (-1%)	9,173 (-2%)
node_database_size_in_bytes	10,559,568	14,397,520
pxe_database_size_in_bytes	9,242	15,127

Circuits stats

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

Circuit	simulation_time_in_ms	witness_generation_time_in_ms	proving_time_in_ms	input_size_in_bytes	output_size_in_bytes	proof_size_in_bytes	num_public_inputs	size_in_gates
private-kernel-init	129 (-2%)	495	12,286	20,634	65,142	90,304	2,755	524,288
private-kernel-inner	390 (-1%)	⚠️ 1,231 (+22%)	45,232	92,854	65,142	90,304	2,755	2,097,152
private-kernel-tail	370 (-1%)	1,871	45,929 (-1%)	97,073	73,629	14,912	399	2,097,152
base-parity	6.61 (+3%)	1,777	2,574	128	64.0	2,208	2.00	131,072
root-parity	50.7	50.4 (-10%)	34,047	27,100	64.0	2,720	18.0	2,097,152
base-rollup	7,715 (-3%)	4,460	72,236	170,330	756	3,648	47.0	4,194,304
root-rollup	113	64.2 (+1%)	19,452 (-2%)	25,309	620	3,456	41.0	1,048,576
public-kernel-setup	703 (-1%)	2,881 (+1%)	38,212 (-1%)	113,321	91,286	123,296	3,786	2,097,152
public-kernel-app-logic	592 (-4%)	3,899 (-1%)	39,537	113,321	91,286	123,296	3,786	2,097,152
public-kernel-tail	1,176 (-2%)	28,498 (+12%)	159,670	412,582	10,014	14,912	399	8,388,608
private-kernel-reset-small	589 (-1%)	2,179 (+5%)	40,842	121,265	65,142	90,304	2,755	2,097,152
public-kernel-teardown	602 (+1%)	3,940 (+2%)	39,735	113,321	91,286	123,296	3,786	2,097,152
merge-rollup	29.4 (-4%)	N/A	N/A	16,542	756	N/A	N/A	N/A
private-kernel-tail-to-public	N/A	7,189	85,280 (-3%)	N/A	N/A	123,296	3,786	4,194,304
avm-circuit	N/A	N/A	3,194 (+1%)	N/A	N/A	86,117	0.00	0.00

Stats on running time collected for app circuits

Function	input_size_in_bytes	output_size_in_bytes	witness_generation_time_in_ms	proof_size_in_bytes	proving_time_in_ms	size_in_gates	num_public_inputs
ContractClassRegisterer:register	1,344	9,944	405	N/A	N/A	N/A	N/A
ContractInstanceDeployer:deploy	1,408	9,944	41.6 (+1%)	N/A	N/A	N/A	N/A
MultiCallEntrypoint:entrypoint	1,920	9,944	2,003 (-3%)	N/A	N/A	N/A	N/A
GasToken:deploy	1,376	9,944	978	N/A	N/A	N/A	N/A
SchnorrAccount:constructor	1,312	9,944	1,465 (+1%)	N/A	N/A	N/A	N/A
SchnorrAccount:entrypoint	2,304	9,944	2,818 (-1%)	16,768	50,525 (+1%)	2,097,152	457
Token:privately_mint_private_note	1,280	9,944	1,751 (+1%)	N/A	N/A	N/A	N/A
FPC:fee_entrypoint_public	1,344	9,944	338 (-3%)	16,768	10,115	524,288	457
Token:transfer	1,312	9,944	4,953 (-1%)	16,768	46,135 (-1%)	2,097,152	457
Benchmarking:create_note	1,344	9,944	1,400	N/A	N/A	N/A	N/A
SchnorrAccount:spend_private_authwit	1,280	9,944	77.7 (-7%)	N/A	N/A	N/A	N/A
Token:unshield	1,376	9,944	4,009	N/A	N/A	N/A	N/A
FPC:fee_entrypoint_private	1,376	9,944	4,979	N/A	N/A	N/A	N/A

AVM Simulation

Time to simulate various public functions in the AVM.

Function	time_ms
GasToken:_increase_public_balance	69.7 (-5%)
GasToken:set_portal	16.6 (-1%)
Token:constructor	94.9 (+1%)
FPC:constructor	63.4 (-1%)
GasToken:mint_public	54.3 (-2%)
Token:mint_public	⚠️ 566 (+704%)
Token:assert_minter_and_mint	⚠️ 74.5 (-67%)
SchnorrAccount:approve_public_authwit	32.4
FPC:prepare_fee	75.8 (-1%)
Token:transfer_public	39.5 (+16%)
FPC:pay_refund	65.8 (-6%)
Benchmarking:increment_balance	2,409 (-3%)
Token:_increase_public_balance	57.7 (-2%)
FPC:pay_refund_with_shielded_rebate	50.6 (-3%)

Public DB Access

Time to access various public DBs.

Function	time_ms
get-nullifier-index	0.179 (+2%)

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	256 leaves	512 leaves	1024 leaves
batch_insert_into_append_only_tree_16_depth_ms	10.3	16.8 (-2%)	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
batch_insert_into_append_only_tree_16_depth_hash_ms	0.600	0.515 (-2%)	N/A	N/A	N/A	N/A	N/A
batch_insert_into_append_only_tree_32_depth_ms	N/A	N/A	48.4 (-1%)	75.7	131	243 (-1%)	470 (+1%)
batch_insert_into_append_only_tree_32_depth_hash_count	N/A	N/A	95.9	159	288	543	1,055
batch_insert_into_append_only_tree_32_depth_hash_ms	N/A	N/A	0.494 (-1%)	0.466	0.449	0.440 (-1%)	0.437
batch_insert_into_indexed_tree_20_depth_ms	N/A	N/A	59.5 (-1%)	111 (-1%)	180	349 (-1%)	693 (-15%)
batch_insert_into_indexed_tree_20_depth_hash_count	N/A	N/A	109	208	355	691	1,363
batch_insert_into_indexed_tree_20_depth_hash_ms	N/A	N/A	0.504 (-2%)	0.498 (-1%)	0.478	0.474 (-1%)	0.475 (-16%)
batch_insert_into_indexed_tree_40_depth_ms	N/A	N/A	74.2 (-6%)	N/A	N/A	N/A	N/A
batch_insert_into_indexed_tree_40_depth_hash_count	N/A	N/A	135	N/A	N/A	N/A	N/A
batch_insert_into_indexed_tree_40_depth_hash_ms	N/A	N/A	0.519 (-6%)	N/A	N/A	N/A	N/A

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	83,854	676,749

Transaction size based on fee payment method

| Metric | |
| - | |

[fcarreiro]

A question also for other reviewers: in general I recommend adding /*param=*/{} for any literal value or when the param name is not clear. This does wonders for readability (e.g. here, reviewing). However, VS already fills in the name for you.

What's the recommendation? (from my side, I think I still prefer adding the comment)

[sirasistant]

I typically don't do it for languages that have language servers that provide this info, but as you guys prefer

[fcarreiro]

Do you need to be explicit on == ?

[sirasistant]

If I didn't do this the parent struct complains about Explicitly defaulted equality comparison operator is implicitly deletedclang(-Wdefaulted-function-deleted)

[sirasistant]

IDK why the parent is explicit though

[fcarreiro]

I'd see if you can avoid all the = {}, they will be default constructed. (same in other files)

[sirasistant]

This is just formatter changes, I'd prefer not generating such a huge diff in this PR :) Will try that in the newly added test though!

[sirasistant]

oops!

    +preset-gcc *failed* | /usr/src/barretenberg/src/barretenberg/dsl/acir_format/honk_recursion_constraint.test.cpp:121:9: error: missing initializer for member 'acir_format::AcirFormat::gates_per_opcode' [-Werror=missing-field-initializers]

[TomAFrench]

Some quick things I've noticed.

[TomAFrench]

LGTM.