Tracing in 4D data
Project reference: 1719
Europe has several synchrotron facilities including, ESRF, PSI, and MAX-IV. These facilities produce still more data. Especially medical beam lines produce enormous datasets; they collect 3D volumes at a frequency that produce 3D movies. An example is volumes of 2560^3 at a frequency of 15 Hz in a 10 min experiment produce 9.000 volumes with a total size of more than 300 TB. The analysis of this data typically involves tracing one or more objects, such as a heart clap or lung tissue, in 3D over time.
Existing algorithms are typically implemented in Matlab and do not scale to the new data rates. Thus the project involves selecting a single algorithm that should be implemented in a High Performance version. Because of the large datasets, conventional processors using MPI is the most straightforward technology choice for this project.
3D visualization of the flight muscles (from http://www.nature.com/articles/srep08727)
Project Mentor: Brian Vinter
Site Co-ordinator: Brian Vinter
Learn different tracing algorithms and how to parallelize them using MP
Student Prerequisites (compulsory):
Basic linux skills
Elementary knowledge of parallel computing
Student Prerequisites (desirable):
Experience with MPI and OpenMP
Week 1: Training
Week 2-3: Study of tracing algorithms
Week 4-7: Implement a parallel version of the tracing algorithm
Week 8: Finalise Report
Final Product Description:
The final product is a parallelized implementation of 4D tracing
Adapting the Project: Increasing the Difficulty:
The student could explore a hybrid programming approach by implementing a version that both uses shared memory parallelism (OpenMP) and distributed memory parallelism (MPI)
The student will need access to a MPI Cluster, which we will provide.
Niels Bohr Institute University of Copenhagen