feat(avm): parallelize polynomial alloc and set

[fcarreiro]

This PR

• Parallelizes allocation and initialization of prover polys
• Removes duplicated initialization
• Removes copy (replaces with move)

Improvements in proving time for a 2^20 trace

• Creation of prover time goes down 90%
• Overall proving time goes down ~40%

2^20 before

>>> prove/create_prover_ms: 53011
>>> prove/all_ms: 95457

prove/create_composer_ms: 0
prove/create_verifier_ms: 322
prove/execute_log_derivative_inverse_commitments_round_ms: 2579
prove/execute_log_derivative_inverse_round_ms: 2294
prove/execute_pcs_rounds_ms: 1915
prove/execute_relation_check_rounds_ms: 12066
prove/execute_wire_commitments_round_ms: 1138
prove/gen_trace_ms: 18690

2^20 after

>>> prove/create_prover_ms: 5797
>>> prove/all_ms: 50688

circuit_builder/init_polys_to_be_shifted_ms: 660
circuit_builder/init_polys_unshifted_ms: 301
circuit_builder/set_polys_shifted_ms: 0
circuit_builder/set_polys_unshifted_ms: 4395
composer/create_prover:commitment_key_ms: 152
composer/create_prover:construct_prover_ms: 55
composer/create_prover:proving_key_ms: 177
composer/create_prover:witness_ms: 5410
prove/create_composer_ms: 0
prove/create_verifier_ms: 326
prove/execute_log_derivative_inverse_commitments_round_ms: 2523
prove/execute_log_derivative_inverse_round_ms: 2185
prove/execute_pcs_rounds_ms: 1766
prove/execute_relation_check_rounds_ms: 11966
prove/execute_wire_commitments_round_ms: 1148
prove/gen_trace_ms: 20765

[fcarreiro]

• feat(avm): parallelize polynomial alloc and set #8520 👈
• master

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

Join @fcarreiro and the rest of your teammates on Graphite

[AztecBot]

Benchmark results

Metrics with a significant change:

• protocol_circuit_simulation_time_in_ms (private-kernel-tail-to-public): 786 (-47%)
• avm_simulation_time_ms (Token:mint_public): 43.9 (-82%)
• avm_simulation_time_ms (Token:assert_minter_and_mint): 540 (+1159%)
• avm_simulation_time_ms (Token:_increase_public_balance): 53.1 (+33%)
• avm_simulation_bytecode_size_in_bytes (FPC:prepare_fee): 1,588 (+16%)
• l2_block_processing_time_in_ms (8): 440 (-45%)
• node_history_sync_time_in_ms (3): 3,114 (-29%)
• node_history_sync_time_in_ms (5): 3,932 (-47%)

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	5,761	1,546	702 (-1%)	756 (-2%)	778
proof_construction_time_sha256_30_ms	11,537 (+1%)	3,062	1,371 (+1%)	1,413	1,463
proof_construction_time_sha256_100_ms	45,239 (-1%)	11,721 (-1%)	5,847	6,216 (+7%)	5,651 (+6%)
proof_construction_time_poseidon_hash_ms	78.0 (-1%)	34.0	34.0	58.0	87.0
proof_construction_time_poseidon_hash_30_ms	1,527	416	199 (+1%)	218 (-2%)	273 (+3%)
proof_construction_time_poseidon_hash_100_ms	5,651	1,500	669	707 (-1%)	746 (+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	4,612	8,260	15,524
l1_rollup_calldata_gas	52,792	97,558	186,860
l1_rollup_execution_gas	819,478	1,555,652	3,346,389
l2_block_processing_time_in_ms	258 (+1%)	⚠️ 440 (-45%)	810 (-2%)
l2_block_building_time_in_ms	9,358	18,556	36,675 (-1%)
l2_block_rollup_simulation_time_in_ms	9,358	18,555	36,674 (-1%)
l2_block_public_tx_process_time_in_ms	7,969	17,137 (+1%)	35,230 (-1%)

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	⚠️ 3,114 (-29%)	⚠️ 3,932 (-47%)
node_database_size_in_bytes	14,205,168	18,194,672 (-1%)
pxe_database_size_in_bytes	16,258	26,818

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	input_size_in_bytes	output_size_in_bytes	proving_time_in_ms
private-kernel-init	71.2 (-6%)	374 (-2%)	20,975	44,933	N/A
private-kernel-inner	140 (-1%)	693 (-1%)	71,787	45,067	N/A
private-kernel-reset-tiny	335 (+2%)	707	64,273	44,922	N/A
private-kernel-tail	152	132 (-1%)	50,551	57,178	N/A
base-parity	5.57 (-9%)	N/A	160	96.0	N/A
root-parity	35.0 (-3%)	N/A	73,964	96.0	N/A
base-rollup	2,143	N/A	189,128	664	N/A
block-root-rollup	41.2 (-3%)	N/A	58,217	2,448	N/A
public-kernel-setup	87.2 (+1%)	N/A	114,433	79,670	N/A
public-kernel-app-logic	99.7	N/A	114,251	79,670	N/A
public-kernel-tail	587 (+1%)	N/A	487,098	16,414	N/A
private-kernel-reset-small	303 (-1%)	N/A	66,345	45,629	N/A
private-kernel-tail-to-public	⚠️ 786 (-47%)	593 (-2%)	448,120	1,825	N/A
public-kernel-teardown	86.7 (+1%)	N/A	114,697	79,670	N/A
merge-rollup	20.0 (-2%)	N/A	38,182	664	N/A
undefined	N/A	N/A	N/A	N/A	103,345 (-2%)

Stats on running time collected for app circuits

Function	input_size_in_bytes	output_size_in_bytes	witness_generation_time_in_ms
ContractClassRegisterer:register	1,344	11,731	350 (+1%)
ContractInstanceDeployer:deploy	1,408	11,731	18.2 (-4%)
MultiCallEntrypoint:entrypoint	1,920	11,731	425
FeeJuice:deploy	1,376	11,731	400
SchnorrAccount:constructor	1,312	11,731	60.9 (-5%)
SchnorrAccount:entrypoint	2,336	11,731	370 (-1%)
FeeJuice:claim	1,344	11,731	39.1 (+1%)
Token:privately_mint_private_note	1,280	11,731	67.7 (-4%)
FPC:fee_entrypoint_public	1,344	11,731	25.1
Token:transfer	1,312	11,731	206 (-2%)
Benchmarking:create_note	1,344	11,731	74.4 (-5%)
SchnorrAccount:verify_private_authwit	1,280	11,731	26.3 (-2%)
Token:unshield	1,376	11,731	485 (-2%)
FPC:fee_entrypoint_private	1,376	11,731	653 (-1%)

AVM Simulation

Time to simulate various public functions in the AVM.

Function	time_ms	bytecode_size_in_bytes
FeeJuice:_increase_public_balance	48.9 (+10%)	1,188
FeeJuice:set_portal	10.1 (-4%)	1,170
Token:constructor	248 (-2%)	10,378 (-1%)
FPC:constructor	73.6 (+3%)	7,877
FeeJuice:check_balance	35.5 (-4%)	995
Token:mint_public	⚠️ 43.9 (-82%)	2,123
Token:assert_minter_and_mint	⚠️ 540 (+1159%)	1,435
AuthRegistry:set_authorized	35.2 (-6%)	515
FPC:prepare_fee	86.7 (+3%)	⚠️ 1,588 (+16%)
Token:transfer_public	16.9 (+27%)	8,603
FPC:pay_refund	37.7 (-3%)	2,090 (+12%)
Benchmarking:increment_balance	1,001	1,516 (-2%)
Token:_increase_public_balance	⚠️ 53.1 (+33%)	1,159
FPC:pay_refund_with_shielded_rebate	56.3 (+9%)	2,090 (+12%)

Public DB Access

Time to access various public DBs.

Function	time_ms
get-nullifier-index	0.163 (-1%)

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	2.15 (-2%)	3.89 (-1%)	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	0.112 (-2%)	0.109 (-1%)	N/A	N/A	N/A	N/A	N/A
batch_insert_into_append_only_tree_32_depth_ms	N/A	N/A	11.2 (-1%)	17.8	31.3 (-1%)	59.9 (-2%)	115 (-1%)
batch_insert_into_append_only_tree_32_depth_hash_count	N/A	N/A	95.9	159	287	543	1,055
batch_insert_into_append_only_tree_32_depth_hash_ms	N/A	N/A	0.108 (-1%)	0.103	0.102 (-1%)	0.103 (-2%)	0.103 (-1%)
batch_insert_into_indexed_tree_20_depth_ms	N/A	N/A	14.3 (+1%)	26.2 (+1%)	44.7 (-1%)	83.7	164
batch_insert_into_indexed_tree_20_depth_hash_count	N/A	N/A	109	207	357	691	1,363
batch_insert_into_indexed_tree_20_depth_hash_ms	N/A	N/A	0.109	0.106 (+1%)	0.104 (-4%)	0.104	0.103 (-1%)
batch_insert_into_indexed_tree_40_depth_ms	N/A	N/A	16.2 (-1%)	N/A	N/A	N/A	N/A
batch_insert_into_indexed_tree_40_depth_hash_count	N/A	N/A	129	N/A	N/A	N/A	N/A
batch_insert_into_indexed_tree_40_depth_hash_ms	N/A	N/A	0.107	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	72,419	668,302

Transaction size based on fee payment method

| Metric | |
| - | |

[ludamad]

We have a similar time tracking mechanism that ties into google benchmark, BB_OP_COUNT_TIME. It will report as a google benchmark counter

[fcarreiro]

Thanks, will look into it!

[ludamad]

Makes sense, we should likely do this too

[fcarreiro]

Merge activity

• Sep 12, 3:31 PM EDT: @fcarreiro merged this pull request with Graphite.