cython and fastness

Yesterday I put some stolen time into looking at whether we can measure short-period stellar oscillations in long-cadence Kepler data. The point is that Kepler long-cadence data has 30-min exposure times, but stellar oscillations in G dwarfs have 5-ish-min periods. The aperiodicity of Kepler exposures might save the day: In principle the exposing is drifted relative to a periodic exposing by the light-travel-time variations to the Kepler field induced by the Kepler spacecraft orbit. I wrote code to forward-model this and see if we can infer the very short stellar-oscillation periods. It looks (from preliminary experiments) that we can! If the noise is friendly, that is.

Today I met up with Foreman-Mackey and he showed me how to convert the slow parts of my code over to cython. That sped up my code by a factor of 300. That's why I pay Foreman-Mackey the big bucks! So now I have a functional piece of code that just might be able to do asteroseismology way below the naive "Nyquist limit". Stoked! If we can get this to work, there are thousands (or even tens of thousands) of stars that might present to us their asteroseismological secrets.

1 comment:

  1. It would be interesting to see how your code would perform if translated to Julia