chore(avm): update stats

[fcarreiro]

I realized much of the "proving time" (9 seconds in a 2^19 trace) is actually spent when creating the prover. In particular, when allocating memory for the prover polynomials, so I added some stats to track that.

------- STATS -------
prove/all_ms: 6879
prove/check_circuit_ms: 2681
prove/create_composer_ms: 0
prove/create_prover_ms: 872
prove/create_verifier_ms: 0
prove/execute_log_derivative_inverse_commitments_round_ms: 124
prove/execute_log_derivative_inverse_round_ms: 179
prove/execute_pcs_rounds_ms: 268
prove/execute_relation_check_rounds_ms: 1619
prove/execute_wire_commitments_round_ms: 86
prove/gen_trace_ms: 975

---

As a side note for posterity, I researched on using mmap (ANON) for allocating zero-initialized memory and changing the following in polynomial.cpp

template <typename Fr> std::shared_ptr<Fr[]> _allocate_aligned_memory(const size_t n_elements)
{
    // NOLINTNEXTLINE(cppcoreguidelines-avoid-c-arrays)
    // return std::static_pointer_cast<Fr[]>(get_mem_slab(sizeof(Fr) * n_elements));
    auto size = sizeof(Fr) * n_elements;
    auto* ptr = mmap(0, size, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
    auto shared = std::shared_ptr<void>(ptr, [size](void* p) { munmap(p, size); });
    return std::static_pointer_cast<Fr[]>(shared);
}

plus removing the memset in the constructor, made 8 of those 9 seconds go away. I think even better performance (and memory usage) can be achieved for sparse vectors/polynomials if we never set zeroes. At some point I'll talk w/Charlie and Adam about this (which only applies to native and not wasm).

PS: mmap anon makes the kernel return a page-aligned zero-initialized page, but it will re-use the same one and be copy-on-write. This saves both time and memory until we write, and possibly a lot of memory if we have sparse matrices with many pages fulls of zeroes.

[fcarreiro]

• chore(avm): codegen improvements #7703
• chore(avm): update stats #7701 👈
• 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:

• avm_simulation_time_ms (Token:mint_public): 60.2 (-41%)
• avm_simulation_time_ms (FPC:pay_refund): 79.1 (-37%)

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	5,864 (+2%)	1,579 (+2%)	723 (+1%)	766 (+1%)	778 (+1%)
proof_construction_time_sha256_30_ms	12,012 (+2%)	3,207 (+2%)	1,413	1,457 (+1%)	1,483 (+1%)
proof_construction_time_sha256_100_ms	44,858 (+2%)	12,351 (+1%)	5,709 (+1%)	5,618 (+2%)	5,544 (+1%)
proof_construction_time_poseidon_hash_ms	80.0 (+3%)	35.0 (+3%)	35.0 (+3%)	60.0	89.0 (+1%)
proof_construction_time_poseidon_hash_30_ms	1,559 (+2%)	424 (+1%)	201 (+1%)	222 (+1%)	269
proof_construction_time_poseidon_hash_100_ms	5,743 (+2%)	1,545 (+1%)	681 (+1%)	750 (+5%)	757 (+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	708	708	708
l1_rollup_calldata_gas	6,600	6,596	6,600
l1_rollup_execution_gas	611,442	611,589	611,744
l2_block_processing_time_in_ms	754 (-1%)	1,408	2,699 (-1%)
l2_block_building_time_in_ms	10,636	20,776 (-2%)	41,043 (-1%)
l2_block_rollup_simulation_time_in_ms	10,636	20,775 (-2%)	41,042 (-1%)
l2_block_public_tx_process_time_in_ms	8,885	18,818 (-2%)	39,006 (-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	7,038	9,793 (-1%)
node_database_size_in_bytes	12,730,448	16,932,944 (+1%)
pxe_database_size_in_bytes	16,254	26,813

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	proof_size_in_bytes	num_public_inputs	size_in_gates
private-kernel-init	96.3 (-2%)	410 (+2%)	21,798	53,090	N/A	N/A	N/A	N/A
private-kernel-inner	174 (-3%)	716 (+1%)	80,891	53,266	N/A	N/A	N/A	N/A
private-kernel-tail	424 (-3%)	881 (+3%)	61,159	57,447	N/A	N/A	N/A	N/A
base-parity	6.42	688 (+4%)	160	96.0	2,004 (+3%)	13,188	19.0	131,072
root-parity	113 (+1%)	135 (+5%)	69,084	96.0	31,422 (+2%)	13,188	19.0	4,194,304
base-rollup	3,336 (-1%)	4,965 (+4%)	182,192	632	44,567 (+2%)	13,988	44.0	4,194,304
root-rollup	156 (-7%)	127 (+2%)	54,461	684	29,011 (+2%)	13,956	43.0	4,194,304
public-kernel-setup	98.9 (-1%)	2,751 (+3%)	112,985	80,246	18,790 (+6%)	141,444	4,027	2,097,152
public-kernel-app-logic	113 (-1%)	3,995 (+1%)	112,985	80,246	11,171 (+2%)	141,444	4,027	1,048,576
public-kernel-tail	633 (-2%)	26,430	410,726	10,814	67,503 (+2%)	26,372	431	8,388,608
private-kernel-reset-tiny	229 (-6%)	895 (+1%)	76,585	52,961	N/A	N/A	N/A	N/A
private-kernel-tail-to-public	5,353	1,573 (+1%)	888,733	1,697	N/A	N/A	N/A	N/A
public-kernel-teardown	93.5 (-3%)	4,008 (+3%)	112,985	80,246	20,114 (+2%)	141,444	4,027	2,097,152
merge-rollup	60.5 (-2%)	N/A	35,678	632	N/A	N/A	N/A	N/A
undefined	N/A	N/A	N/A	N/A	157,893 (-1%)	N/A	N/A	N/A

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
ContractClassRegisterer:register	1,312	11,699	346	N/A	N/A
ContractInstanceDeployer:deploy	1,376	11,699	27.6 (+1%)	N/A	N/A
MultiCallEntrypoint:entrypoint	1,888	11,699	607	N/A	N/A
GasToken:deploy	1,344	11,699	450	N/A	N/A
SchnorrAccount:constructor	1,280	11,699	431	N/A	N/A
SchnorrAccount:entrypoint	2,272	11,699	817 (-2%)	N/A	N/A
Token:privately_mint_private_note	1,248	11,699	503	N/A	N/A
FPC:fee_entrypoint_public	1,312	11,699	83.5 (-4%)	N/A	N/A
Token:transfer	1,280	11,699	1,292 (+1%)	N/A	N/A
AuthRegistry:set_authorized (avm)	18,427	N/A	N/A	102,048	1,941 (+2%)
FPC:prepare_fee (avm)	22,894	N/A	N/A	102,112	2,166 (+3%)
Token:transfer_public (avm)	46,817	N/A	N/A	102,112	6,574 (+9%)
AuthRegistry:consume (avm)	26,670	N/A	N/A	102,080	2,539 (+4%)
FPC:pay_refund (avm)	26,163	N/A	N/A	102,080	3,540 (+4%)
Benchmarking:create_note	1,312	11,699	419 (-3%)	N/A	N/A
SchnorrAccount:verify_private_authwit	1,248	11,699	47.3 (+4%)	N/A	N/A
Token:unshield	1,344	11,699	1,162 (+8%)	N/A	N/A
FPC:fee_entrypoint_private	1,344	11,699	1,469 (+7%)	N/A	N/A

AVM Simulation

Time to simulate various public functions in the AVM.

Function	time_ms	bytecode_size_in_bytes
GasToken:_increase_public_balance	97.1 (-2%)	8,139
GasToken:set_portal	11.9 (-6%)	2,362
Token:constructor	120 (-2%)	16,400
FPC:constructor	92.6	9,304
GasToken:mint_public	81.1 (-2%)	6,150
Token:mint_public	⚠️ 60.2 (-41%)	11,720
Token:assert_minter_and_mint	60.3 (+12%)	8,028
AuthRegistry:set_authorized	45.1 (-13%)	4,537
FPC:prepare_fee	239 (+4%)	8,812
Token:transfer_public	38.4 (-26%)	32,641
FPC:pay_refund	⚠️ 79.1 (-37%)	12,114
Benchmarking:increment_balance	1,039 (-1%)	7,450
Token:_increase_public_balance	12.1 (+1%)	8,960
FPC:pay_refund_with_shielded_rebate	144	12,663

Public DB Access

Time to access various public DBs.

Function	time_ms
get-nullifier-index	0.153

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 (-4%)	16.6 (-1%)	N/A	N/A	N/A	N/A	N/A
batch_insert_into_append_only_tree_16_depth_hash_count	16.8	31.7	N/A	N/A	N/A	N/A	N/A
batch_insert_into_append_only_tree_16_depth_hash_ms	0.595 (-4%)	0.510 (-1%)	N/A	N/A	N/A	N/A	N/A
batch_insert_into_append_only_tree_32_depth_ms	N/A	N/A	48.2	75.9 (-1%)	131	243 (-1%)	470
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	0.492	0.466 (-1%)	0.450	0.441 (-1%)	0.439
batch_insert_into_indexed_tree_20_depth_ms	N/A	N/A	59.4 (-1%)	110 (-1%)	182 (-1%)	349 (-1%)	691
batch_insert_into_indexed_tree_20_depth_hash_count	N/A	N/A	109	207	355	691	1,363
batch_insert_into_indexed_tree_20_depth_hash_ms	N/A	N/A	0.502 (-1%)	0.494 (-1%)	0.483	0.473 (-1%)	0.476
batch_insert_into_indexed_tree_40_depth_ms	N/A	N/A	72.8 (-1%)	N/A	N/A	N/A	N/A
batch_insert_into_indexed_tree_40_depth_hash_count	N/A	N/A	133	N/A	N/A	N/A	N/A
batch_insert_into_indexed_tree_40_depth_hash_ms	N/A	N/A	0.517 (-1%)	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	72,017	666,910

Transaction size based on fee payment method

| Metric | |
| - | |

[fcarreiro]

Merge activity

• Jul 31, 2:58 PM EDT: @fcarreiro started a stack merge that includes this pull request via Graphite.
• Jul 31, 2:58 PM EDT: @fcarreiro merged this pull request with Graphite.