Description
Electron reconstruction at CMS uses a special tracking algorithm called Gaussian Sum Filter (GSF) to account for radiative loss from brehmsstrahlung photons. The GSF algorithm is highly CPU intensive and thus cannot be run over all hits in the tracker. Hence the GSF tracking is run with only those hit patterns (seeds) that are compatible with an electron trajectory. These seeds are identified by matching hits in the pixel detector to the energy deposit of an electron in the electromagnetic calorimeter. Since the pixel matching procedure requires trying out all possible pixel hit combinations in the luminous region, it is currently the most time-consuming step of electron reconstruction. Hence is is important to speed up this step especially for high level triggers in Phase-2, where one will have a very high particle multiplicity (and consequently a huge number of pixel hits). One way of improving the pixel matching speed would be to port the algorithm to GPUs in a parallelizable manner. This project will involve developing and optimizing a GPU-compatible version of the the pixel matching algorithm, and studing its performance under Phase-2 scenarios.
Special Qualifications:
Must: C++ and Python
Great to have: introduction to GPU programming
Nice to Have: basic knowledge of Physics objects
| Field | B2: Data processing (software-oriented) |
|---|---|
| DESY Place | Hamburg |
| DESY Division | FH |
| DESY Group | CMS |
