Amadeus MatMul Solver

Amadeus MatMul Solver (RISC-V Edition)

High-performance matrix multiplication solver for the Amadeus uPoW (Useful Proof of Work) protocol, optimized for Tenstorrent Blackhole p150a (RISC-V) hardware.

🔍 Overview

This solver performs the core Amadeus uPoW computation using deterministic matrix generation and probabilistic verification.

Matrix Specification

• Matrix A: 16 × 50,240 (uint8)

• Matrix B: 50,240 × 16 (int8, signed)

• Result (C): 16 × 16 (int32, little-endian)

Verification

• Freivalds’ Algorithm for fast probabilistic correctness checks

 Features

• Deterministic matrix generation via Blake3 XOF

• Parallel matrix multiplication using Rayon

• Freivalds verification for result validation

• Proof-of-Work mining with configurable difficulty

• Built-in Amadeus testnet API client

• Approx. 347 hashes/sec on a standard CPU

 Quick Start

Build

cargo build --release

Commands

# Test connection to Amadeus testnet./target/release/amadeus-solver test# Compute and validate (no mining)./target/release/amadeus-solver compute# Mine a valid PoW solution./target/release/amadeus-solver mine -d 16 -t 60# Validate an existing Base58 solution./target/release/amadeus-solver validate # Run performance benchmarks./target/release/amadeus-solver benchmark

Environment Variables

VariableDescription

📊 Performance

MetricValue

Solution Format

Total Size: 1264 bytes

1. Seed (240 bytes)

FieldSize

2. Tensor C (1024 bytes)

• 16 × 16 matrix

• int32 values

• Little-endian encoding

🌐 API Endpoints

MethodEndpointDescription

- Validation Response

{  "valid": true,  "valid_math": true}

• valid: PoW hash meets difficulty target

• valid_math: Matrix multiplication is correct

🐳 Docker

Build

docker build -t amadeus-matmul .

Run

docker run amadeus-matmul compute