feat: Dynamic proving by PhilWindle · Pull Request #5346 · AztecProtocol/aztec-packages

PhilWindle · 2024-03-20T20:20:40Z

This PR introduces a new proving orhestration system enabling a block to be proven concurrently with simulation. It also creates the prover-client package and moves some of the proving functionality to this package from sequencer-client.

…s into pw/proving-2

AztecBot · 2024-03-20T23:07:13Z

Docs Preview

Hey there! 👋 You can check your preview at https://65fd6a36bacba41b19487787--aztec-docs-dev.netlify.app

AztecBot · 2024-03-20T23:23:15Z

Benchmark results

Metrics with a significant change:

circuit_simulation_time_in_ms (base-parity): 4,413 (+142%)
circuit_simulation_time_in_ms (base-rollup): 18,069 (+2459%)

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 8434d2f3 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	64 txs
l1_rollup_calldata_size_in_bytes	676	676	676
l1_rollup_calldata_gas	6,412	6,424	6,424
l1_rollup_execution_gas	585,745	585,757	585,757
l2_block_processing_time_in_ms	1,318 (-2%)	4,644 (-2%)	9,595 (+2%)
note_successful_decrypting_time_in_ms	188 (-1%)	509 (-4%)	981 (+2%)
note_trial_decrypting_time_in_ms	96.9 (+3%)	25.9 (-43%)	110 (+13%)
l2_block_building_time_in_ms	17,247	65,380 (+1%)	129,023
l2_block_rollup_simulation_time_in_ms	7,942	28,445 (+1%)	55,537
l2_block_public_tx_process_time_in_ms	9,286	36,880 (+1%)	73,390

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	14,346 (+4%)	29,327 (-5%)
note_history_successful_decrypting_time_in_ms	1,224 (-3%)	2,459 (+1%)
note_history_trial_decrypting_time_in_ms	61.7 (-36%)	119 (-22%)
node_database_size_in_bytes	18,649,168	35,053,648
pxe_database_size_in_bytes	29,859	59,414

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	255 (+7%)	44,344	28,181
private-kernel-ordering	228 (+6%)	52,847	14,263
base-parity	⚠️ 4,413 (+142%)	128	311
root-parity	1,636 (+4%)	1,244	311
base-rollup	⚠️ 18,069 (+2459%)	165,727	861
root-rollup	50.1	4,359	725
private-kernel-inner	645 (+9%)	73,728	28,181
public-kernel-app-logic	397 (-1%)	35,185	28,151
public-kernel-tail	171 (-1%)	40,862	28,151
merge-rollup	7.73 (+9%)	2,568	861

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	9.94 (-2%)	16.1 (+1%)	N/A	N/A	N/A	N/A	N/A	N/A	N/A
batch_insert_into_append_only_tree_16_depth_hash_count	16.8	31.6	N/A	N/A	N/A	N/A	N/A	N/A	N/A
batch_insert_into_append_only_tree_16_depth_hash_ms	0.580 (-2%)	0.500 (+1%)	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	45.6	72.0 (-1%)	232 (+1%)	452	866	1,867 (+8%)	N/A
batch_insert_into_append_only_tree_32_depth_hash_count	N/A	N/A	96.0	159	543	1,055	2,079	4,127	N/A
batch_insert_into_append_only_tree_32_depth_hash_ms	N/A	N/A	0.468	0.445 (-1%)	0.423 (+1%)	0.422	0.413	0.448 (+9%)	N/A
batch_insert_into_indexed_tree_20_depth_ms	N/A	N/A	53.6 (-2%)	106	333	659 (-1%)	1,305	2,632 (+1%)	N/A
batch_insert_into_indexed_tree_20_depth_hash_count	N/A	N/A	104	207	691	1,363	2,707	5,395	N/A
batch_insert_into_indexed_tree_20_depth_hash_ms	N/A	N/A	0.477 (-2%)	0.480	0.454	0.454 (-1%)	0.454	0.454	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	60.9
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
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	0.533

Miscellaneous

Transaction sizes based on how many contract classes are registered in the tx.

Metric	0 registered classes
tx_size_in_bytes	21,991

Transaction processing duration by data writes.

Metric	0 new note hashes	1 new note hashes
tx_pxe_processing_time_ms	3,089	1,743 (+6%)

Metric	0 public data writes	1 public data writes
tx_sequencer_processing_time_ms	10.9 (-9%)	1,140 (-1%)

just-mitch · 2024-03-21T14:08:00Z

yarn-project/prover-client/src/orchestrator/proving-state.ts

+ * Captures resolve and reject callbacks to provide a promise base interface to the consumer of our proving.
+ */
+export class ProvingState {
+  private _stateIdentifier: string;


nit: I'm not keen on the underscore prefix. I don't know what it "means" in our codebase.

spalladino · 2024-03-21T12:48:39Z

yarn-project/prover-client/src/dummy-prover.ts

+  ): Promise<ProvingTicket> {
+    const result: ProvingResult = {
+      proof: makeEmptyProof(),
+      block: L2Block.random(1),


Let's set an empty block rather than a random one

spalladino · 2024-03-21T12:53:09Z

yarn-project/prover-client/src/tx-prover/tx-prover.ts

+async function getSimulationProvider(config: ProverConfig): Promise<SimulationProvider> {
+  if (config.acvmBinaryPath && config.acvmWorkingDirectory) {
+    try {
+      await fs.access(config.acvmBinaryPath, fs.constants.R_OK);
+      await fs.mkdir(config.acvmWorkingDirectory, { recursive: true });
+      logger(`Using native ACVM at ${config.acvmBinaryPath}`);
+      return new NativeACVMSimulator(config.acvmWorkingDirectory, config.acvmBinaryPath);
+    } catch {
+      logger(`Failed to access ACVM at ${config.acvmBinaryPath}, falling back to WASM`);
+    }
+  }
+  logger('Using WASM ACVM simulation');
+  return new WASMSimulator();
+}


Remove duplicate code from sequencer-client

spalladino · 2024-03-21T12:54:38Z

yarn-project/sequencer-client/src/sequencer/abstract_phase_manager.ts

    const output = await this.getKernelCircuitOutput(previousOutput, previousProof, callData);
-    const proof = await this.publicProver.getPublicKernelCircuitProof(output);
-    return [output, proof];
+    return [output, makeEmptyProof()];


Is the idea to change this later to a call to the actual prover, or to not return a proof here and do it elsewhere?

Yes, eventually all proving inside sequencer will be gone.

spalladino · 2024-03-21T13:33:01Z

yarn-project/end-to-end/src/integration_l1_publisher.test.ts

      );
-      const [block] = await builder.buildL2Block(globalVariables, txs, l1ToL2Messages);
+      const blockTicket = await buildBlock(globalVariables, txs, l1ToL2Messages, makeEmptyProcessedTx());
+      const result = await blockTicket.provingPromise;


Replace this with an awaitable wait() method that returns a Result-like (as discussed on the standup)

spalladino · 2024-03-21T13:33:35Z

yarn-project/prover-client/src/orchestrator/block-building-helpers.ts

I understand these are all copied directly from the old solo-block-builder, right? Any significant changes worth reviewing?

No, nothing really to look at

spalladino · 2024-03-21T13:56:20Z