Skip to content

Optimization using cuFFT Callbacks

Jump to bottom Edit New page
Alex Nitz edited this page Mar 11, 2015 · 15 revisions

Algorithm Summary

We'll consider 3 of the kernels within the inner loop of the PyCBC analysis code.

  1. correlate The kernel multiples a complex template and data segment together to create the 'correlation' vector
  2. ifft Using cuFFT, we take the correlation vector and inverse FFT it to produce an SNR time series.
  3. threshold Peaks in the SNR timeseries are located by windows maximums within a window size.

Callback Uses

Tesla K10 Results

Add a custom footer
Add a custom sidebar
Clone this wiki locally