NERVOUS Program Irradiation Dataset

Dataset is derived from recordings of internal signals from a RISC-V processor during execution of various workloads, along with spike trains generated from the same, created to investigate the application of spiking neural networks (SNNs) to the task of detecting faults in program execution. The workloads are: a Quicksort algorithm (Quicksort), using the C standard library implementation, which sorts a list of randomly generated 16-bit numbers with values between 0 and 32767; a prime number finder (Prime Finder), using an implementation of the sieve of Erathosthenes, which locates all prime numbers up 1000; a matrix multiplier (Matrix Multi.), which multiplies a randomly generated 10x10 matrix (each value has a range of 0 to 50) with a fixed 10x10 diagonal matrix; a Fibonacci sequence generator (Fibonacci), which produces all numbers in the Fibonacci sequence that are less than 400 Million; a fast Fourier transform (FFT), which performs a time to frequency domain transform on a 256 time step long signal composed of randomly generated values between -1.0 and 1.0; and a cellular automata (Cellular Automata), specifically therule set for Conway\u2019s Game of Life, running on a 40x40 grid that randomly initalised with a 30% chance for a given cell to be alive. Each recording is 131,072 time steps long, consisting of 131 channels encompassing: the program counter; instruction memory output; instruction word; post-instruction decoder control signals for the registers, ALU and memory; and the CPU status signals.