seismo

This started as a place to keep my asteroseismology-related stuff

Note

I just started this but the aim is to have a small set of tools that will allow me to extract frequencies from an unevenly sampled dataset e.g. Kepler lightcurve. This requires a LOT of processing power, which is why there are multiple parallel implementations of the O(N^2) Deeming periodogram, including one that runs on my GPU via opencl.

Installation

Make sure you have pip installed.

On OSX I use homebrew. If you have not, install python from there first then install pip.

(optional) brew install python
brew install pip

On Debian/Ubuntu

sudo apt-get install python-pip

Once pip is installed, install seismo with:

pip install -r requirements.txt
pip install .

in this folder.

You may need sudo for the last command

Running the unit tests

To see if the installation completed successfully, run:

py.test tests

This should give some output and all the tests should pass.

Notes about OpenCL and GPU calculations

To use the OpenCL extensions, you need an OpenCL driver installed for each different device. On OSX this is installed already, on Linux it is not. For NVIDIA users, one is installed with the CUDA SDK, I do not know what is needed for AMD cards and CPUs. NOTE You do NOT need the CUDA SDK for the opencl driver.

Modules

seismo contains the following modules:

timeseries - contains some DFT implementations
fitting - will contain frequency fitting and extraction routines

Goals

Kepler datasets can be very long and contain many datapoints. These take a long time to calculate via a naive O(N^2) algorithm (but needed because I don't have a better algorithm for unevenly sampled data that I'm happy with). For this reason I'm trying to create something that can run more or less with only some user input, so that I can let it run by itself and save each step in the output.

Each DFT for the particular dataset I'm interested in takes about 50min to run on my GTX 760, and about 16 hours on my MBP's i5 with 2 threads, and I may need to do 40 - 50 of them to extract all the frequencies. (Donations for a Nvidia 980 welcome)

Changes:

Version 0.1.2: * Added deeming function that wraps the dft implementations. * Updated examples * added tests

ezietsman/seismo