Welcome to Squiggles.jl, a Julia package to accelerate 1D cross correlations of seismic signals using any GPU!

Installation

Open the Julia REPL and type either

using Pkg; Pkg.add(url="git@github.com:gbene/Squiggles.jl.git")

or

] add git@github.com:gbene/Squiggles.jl.git 

To use GPU acceleration you need to install the appropriate JuliaGPU package

• ] add CUDA -- CUDA cards
• ] add Metal -- Apple silicon
• More to come!

To plot use either CairoMakie or GLMakie (for interactive plots)

• ] add CairoMakie
• ] add GLMakie

Quick example

Here is a basic example using CUDA

using Squiggles
using CUDA 

set_GPUbackend()

i = 10          # Number of signals (columns) for in the A matrix
j = 10          # Number of signals (columns) for in the B matrix

n_samples = 128 # Number of samples (rows) for A and B 

A = SignalMatrix(n_samples, 1, i, 1:20, 1, -0.5:0.1:0.5) # Random signals
B = SignalMatrix(n_samples, 1, j, 1:20, 1, -0.5:0.1:0.5)

τ = 128 # Amount of samples to correlate 

threads_per_block = 128 # Amount of threads per block 

correlograms_norm_gpu = norm_correlogram(A, B, τ, threads_per_block) # Normalized correlogram volume

norm_coeffs_gpu, lags_gpu = simplelags(correlograms_norm_gpu, τ) # Correlation coeffs and lags matrices

Check the docs for more examples!

Acknowledgments

Libraries

We use many libraries of the Julia ecosystem and we thank all of them! These are the main ones that made this project possible

• JuliaGPU: https://juliagpu.org/
• KernelAbstractions.jl: https://juliagpu.github.io/KernelAbstractions.jl

People

• Ylse Anna de Vries: https://github.com/ylseanna
• Tom Winder: https://github.com/TomWinder
• Elías Rafn Heimisson: https://github.com/eliasrh