HSF student pages for GSoC 2024
Any collection in Data Model Explorer
Braulio Rivas
Incorporated new features for eede (EDM4hep Event Data Explorer). Added a whole new set of collections that eede can visualize (according to the EDM4hep eventdata model) like Clusters, ParticleId or Tracks. Included visualization of relations between them and the previously developed MCParticle collection. Built a menu that allows users to filter out these objects according to their properties and tested using data from FCC simulations.
CernVM-FS - Integration of FUSE-T library for macOS laptop clients + Leveraging modern Overlay-FS features
Yuriy Belikov
Worked on integration of additional overlay FS features in CVMFS server part and replacement of macFUSE kernel extension with FUSE-T user-space library for macOS laptop clients.At this point a backbone for metadata-only copying and zero-copy directory renames is implemented for Linux server part and still goes through a refinement and adjustment due to various differences in overlay FS behaviour discovered during the course of the project. FUSE-T support for macOS clients is primarily done apart several issues that weclassified as critical for CVMFS client stability. I helped with creation of GitHub Actions CI pipeline for macOS clients, prepared a table of FUSE-T issues that were encountered during this project. Currently, I am continuing my work on adding an ability for users to switch back to macFUSE kext.
Containerisation of CBACK backup system
Yash Raj Singh
This project aimed to enhance the flexibility and scalability of CBACK by introducing dynamic agent spawning through Kubernetes-based scheduling. The project was responsible for developing a new “scheduler” agent type responsible for monitoring the internal database and initiating agent creation based on service demand. Additionally, the project also had containerisation as one of its primary goals to enable dynamic spawning of agents by the scheduler in a kubernetes environment. And to limit the scope of changes CSI drivers were also integrated along with cephx keys allowing for the restriction of file mount scopes on containers running job agents.
Implementing the Differentiation of the Kokkos Framework in Clad
Atell Yehor Krasnopolski
Hi! I’m Atell, and this is my GSoC 2024 blog post, where I describe my experience and journey with Clad and the Compiler Research Team in general.
Inference Code Generation for Deep Learning models
Vedant Mehra
Implemented new Operator classes for SOFIE (System for Optimized Fast Inference code Emit) along with respective operator parsers, tests and onnx models.Worked on adapting existing operator classes to be able to parse parametrised shapes with dimension values of input tensors not specified at compile time.
Lossless compression of raw data for the ATLAS experiment at CERN
Ishan Darji
Investigated lossless data compression for ATLAS experiment raw data to replace the currently used zlib compression. Investigated alternatives such as zstd, brotli, xz, and lzlib. Compared tradeoffs between libraries in metrics such as compression speed and compression ratio, placing additional importance on compression ratio for this project because incremental saves add up at exabyte scale. Implemented example of compression and decompression of ATLAS event data that can read and write compressed data. |
The journey of enabling reverse-mode automatic differentiation of GPU kernels using Clad
Christina Koutsou
A not-very-short description of the challenges faced and my overall experience
Wrapping up GSoC'24: Improving performance of BioDynaMo using ROOT C++ Modules
Isaac Morales Santana
This project, part of Google Summer of Code 2024, aims to reduce the header parsing in BioDynaMo using the ROOT C++ Modules