proof-aggregation/workflow/BENCHMARKS.md

Benchmarks

In here we show the preliminary benchmarks of codex storage proofs circuits.

Running Benchmarks

To run the benchmarks for safe merkle tree circuit, you can use the following command:

cargo bench --bench safe_circuit

To run the benchmarks for sampling circuit, you can use the following command: Note: make sure to adjust the parameters as need in ....

cargo bench --bench sample_cells

The following operations were benchmarked:

• Build Circuit: Time taken to construct the circuit for the specified params.
• Prove Circuit: Time taken to generate a proof for the constructed circuit.
• Verify Circuit: Time taken to verify the generated proof.

Bench Results

The following is the result of running the codex storage proof circuit (sample_cells). The bench uses the Goldilocks field and Poseidon2 Hash. All results were run on Mac Mini with M2 Pro and 16GB RAM.

Bench params

The benchmark runs with the default params which are the following:

export MAXDEPTH=32        # Maximum depth of the slot tree
export MAXSLOTS=256       # Maximum number of slots
export CELLSIZE=2048      # Cell size in bytes
export BLOCKSIZE=65536    # Block size in bytes
export NSAMPLES=5         # Number of samples to prove

export ENTROPY=1234567    # External randomness
export SEED=12345         # Seed for creating fake data

export NSLOTS=11          # Number of slots in the dataset
export SLOTINDEX=3        # Which slot to prove (0..NSLOTS-1)
export NCELLS=512         # Number of cells in this slot

Build Time

Build time for plonky2 circuits is 39.644 ms. Baseline Groth16 with same params: 61 seconds for the circuit specific setup.

Prove Time

Prove time for plonky2 circuits is 53.940 ms. Baseline Groth16 with same params: 4.56 seconds using snarkjs

improvement: approx 80x

Verify Time

To be done once recursion is added to the codebase.

Proof Size

Plonky Proof size: 116008 bytes This is without recursion or Groth16 wrapper.

Peak Memory Usage

To be done.