# Tandem Benchmarks

This directory contains benchmarks for [Tandem](../README.md), which directly hook into the core engine without any client/server wrappers.

All benchmarks are run using the [`simulate`](tandem/src/simulator.rs) function, which executes a Multi-Party Computation under ideal network conditions, without any latency or bandwidth restrictions.

All benchmarks were implemented with [Criterion.rs](https://github.com/bheisler/criterion.rs).

## `circuits.rs`

This file includes 4 benchmarks with only `AND` gates and 4 benchmarks with only `XOR` gates.

| Function    | Gates     |
| ----------- | --------- |
| `and(10)`    | 10 `AND`  |
| `and(100)`   | 100 `AND` |
| `and(1000)`  | 1k `AND`  |
| `and(10000)` | 10k `AND` |
| `xor(10)`    | 10 `XOR`  |
| `xor(100)`   | 100 `XOR` |
| `xor(1000)`  | 1k `XOR`  |
| `xor(10000)` | 10k `XOR` |

## `credit_scoring.rs`

This file contains a benchmark based on an integration test, simulating a real use case of the Tandem
engine. This example takes a credit scoring algorithm as the contributor's input and a user's data
and credit history as the evaluator's input. Keeping the algorithm and the user informations
private, the Tandem engine calculates the user's credit score.

| Function                   | Gates                           |
| -------------------------- | ------------------------------- |
| `credit_scoring_benchmark` | 24k `XOR`, 10k `NOT`, 33k `AND` |

## `multiplications.rs`

This file contains 2 benchmarks, with both `AND` and `XOR` gates, generated from simple Garble functions that multiply the `contributor`'s input by the `evaluator`'s input 1 and 10 times.

| Function | Gates                  |
| -------- | ---------------------- |
| `mul_1`  | 16k `AND`, 16k `XOR`   |
| `mul_10` | 160k `AND`, 161k `XOR` |