feat: Improve ivc bench

[ledwards2225]

Makes the goblin bench more representative of the real aztec architecture.

• General structure of the bench suite has been updated for clarity
• Each "accumulation round" consists of a function and kernel
• The first function is size 2^19 and subsequent function circuits are 2^17
• Every kernel is size 2^17
• Mock circuit logic now consists of "big" primitive operations, e.g. sha256, merkle membership checks etc. rather than simple arithmetic gates. (Methods for constructing these individual circuits have been moved to appropriate locations in the stdlib as discussed).
• The "nominal" bench has 6 iterations resulting in 12 UGH circuits/proofs

[AztecBot]

Benchmark results

Metrics with a significant change:

• note_history_trial_decrypting_time_in_ms (5): 165 (+84%)

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.

Values are compared against data from master at commit 0ed41261 and shown if the difference exceeds 1%.

L2 block published to L1

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

Metric	8 txs	32 txs	128 txs
l1_rollup_calldata_size_in_bytes	45,412	179,524	716,068
l1_rollup_calldata_gas	221,680	866,392	3,448,432
l1_rollup_execution_gas	314,167	984,132	3,668,089
l2_block_processing_time_in_ms	1,150	4,339	17,182
note_successful_decrypting_time_in_ms	324 (-3%)	871	3,560 (+1%)
note_trial_decrypting_time_in_ms	36.2 (-15%)	139 (+36%)	144 (+1%)
l2_block_building_time_in_ms	18,035	71,773 (+1%)	287,374
l2_block_rollup_simulation_time_in_ms	12,950	51,496	206,856
l2_block_public_tx_process_time_in_ms	5,059 (-1%)	20,147 (+1%)	80,272 (+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	13,312	25,724
note_history_successful_decrypting_time_in_ms	2,298 (-8%)	4,519 (+1%)
note_history_trial_decrypting_time_in_ms	⚠️ 165 (+84%)	202 (+4%)
node_database_size_in_bytes	17,285,200	33,259,600
pxe_database_size_in_bytes	29,923	59,478

Circuits stats

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

Circuit	circuit_simulation_time_in_ms	circuit_input_size_in_bytes	circuit_output_size_in_bytes
private-kernel-init	315	44,897	26,193
private-kernel-ordering	192	44,513	14,929
base-rollup	1,396	128,970	881
root-rollup	81.7	4,088	677
private-kernel-inner	421 (-1%)	72,056	26,193
public-kernel-private-input	241	32,615	26,193
public-kernel-non-first-iteration	240	32,657	26,193
merge-rollup	7.59	2,608	881

Tree insertion stats

The duration to insert a fixed batch of leaves into each tree type.

Metric	1 leaves	2 leaves	8 leaves	16 leaves	32 leaves	128 leaves	64 leaves	512 leaves	1024 leaves	2048 leaves	8192 leaves
batch_insert_into_append_only_tree_16_depth_ms	10.3 (+1%)	10.9 (-19%)	12.7	16.8	22.9	64.5 (+1%)	N/A	N/A	N/A	N/A	N/A
batch_insert_into_append_only_tree_16_depth_hash_count	16.9	17.5	23.0	31.6	47.0	143	N/A	N/A	N/A	N/A	N/A
batch_insert_into_append_only_tree_16_depth_hash_ms	0.596 (+1%)	0.609 (-19%)	0.540	0.522	0.479	0.445 (+1%)	N/A	N/A	N/A	N/A	N/A
batch_insert_into_append_only_tree_32_depth_ms	N/A	N/A	N/A	N/A	N/A	74.1 (-6%)	47.6	239	461	898	3,548
batch_insert_into_append_only_tree_32_depth_hash_count	N/A	N/A	N/A	N/A	N/A	159	96.0	543	1,055	2,079	8,223
batch_insert_into_append_only_tree_32_depth_hash_ms	N/A	N/A	N/A	N/A	N/A	0.459 (-6%)	0.488	0.436	0.431	0.428	0.426
batch_insert_into_indexed_tree_20_depth_ms	N/A	N/A	N/A	N/A	N/A	103	56.6	350 (+1%)	687	1,363 (-1%)	5,399
batch_insert_into_indexed_tree_20_depth_hash_count	N/A	N/A	N/A	N/A	N/A	197	104	691	1,363	2,707	10,771
batch_insert_into_indexed_tree_20_depth_hash_ms	N/A	N/A	N/A	N/A	N/A	0.492	0.495	0.474	0.473	0.473 (-1%)	0.470
batch_insert_into_indexed_tree_40_depth_ms	N/A	N/A	N/A	56.4	N/A	N/A	N/A	N/A	N/A	N/A	N/A
batch_insert_into_indexed_tree_40_depth_hash_count	N/A	N/A	N/A	94.1	N/A	N/A	N/A	N/A	N/A	N/A	N/A
batch_insert_into_indexed_tree_40_depth_hash_ms	N/A	N/A	N/A	0.577	N/A	N/A	N/A	N/A	N/A	N/A	N/A

Miscellaneous

Transaction sizes based on how many contracts are deployed in the tx.

Metric	0 deployed contracts	1 deployed contracts
tx_size_in_bytes	15,575	36,576

Transaction processing duration by data writes.

Metric	0 new commitments	1 new commitments
tx_pxe_processing_time_ms	581	1,359

Metric	1 public data writes
tx_sequencer_processing_time_ms	624

[ledwards2225]

The inclusion of the stdlib components is only necessary now because the full goblin recursion tests (which now use these stdlib primitives in the kernel) live in the goblin module. We don't have an established pattern for avoiding this because this is the only place that we use stdlib primitives outside the stdlib (except for certain bench suites). One option is just to have a goblin_test module that is separate from goblin. Let's discuss

[ludamad]

If they're only in headers, this isn't technically needed. Also, might signal that this is a module boundary break vs having them in a testing module

[ledwards2225]

The restructuring of this suite is substantial enough that it may be easier to review in isolation rather than via the diff.

[codygunton]

Looks good EDIT: requesting changes

[codygunton]

Issues I've raised with you after approving: witness construction time depends on the circuit size and not just it's upper power of 2 magnitude. You said the app circuits are close to the lower end but the kernel circuits are closer to the upper end.

[codygunton]

Also: we ran the benchmarks together in a call, but actually we only ran one, and on further inspection I see that there is a segfault when you try to run the full suite.

[codygunton]

Looked at your updates, look good, thanks.