Coherent Imaging: GPU-accelerated data analysis using PyNX
Coherent imaging techniques will highly benefit from the Extremely Brilliant Source (ESRF-EBS) upgrade. However, one of the challenges will be to handle the increased amount of data, and to enable users to analyse data as much as possible during their experiments.
The PyNX software currently developed at ESRF (http://ftp.esrf.fr/pub/scisoft/PyNX/doc/) systematically uses Graphical Processing Units (GPU) to provide 1 to 2 orders of magnitude acceleration in data analysis throughput.
This seminar will focus on:
- the tools already available with the choices in algorithms
- how the python library can be easily extended to new algorithms using an Operator-based approach (without any knowledge in GPU programming)
- a few details about the underlying computing approach (allowing to use OpenCL or CUDA)
Examples will include Coherent Diffraction Imaging (in the small-angle and Bragg regimes), Ptychography (near field and far field, small angle and Bragg), and ongoing developments (Phase Contrast Imaging/Holo-Tomography) will also be discussed.
We will also give an outlook on remaining challenges: coping for the increased intensity, massive parallelisation for near-field tomography techniques,...
Visitors from off-site please contact Anaïs Fernandes tel +33 (0)438 88 19 92 to arrange for a gate pass.
Requests made by e-mail will be confirmed.
If you do not receive a confirmation e-mail, please contact us by phone.
Requests made by e-mail will be confirmed.
If you do not receive a confirmation e-mail, please contact us by phone.