Diff4Earth/ige-jaxathon-2025
Mar 2025 – Mar 2025
1 Month active
Languages Used
Jupyter NotebookMarkdownPython
Technical Skills
JAXLinear AlgebraMatplotlibNumPyNumerical AnalysisNumerical Methods
Hugo Frezat
PROFILE
Hugo Frezat
Developed a sparse spectral solver performance analysis framework for the Diff4Earth/ige-jaxathon-2025 repository, focusing on Poisson-like problems using spectral methods. The work involved building a reproducible pipeline in Python and JAX, incorporating matrix generation, GMRES solver testing, and preconditioner exploration, with an option for multi-GPU scaling via jax.pmap. A comprehensive Jupyter notebook was created to demonstrate the solver’s capabilities, including the Segment class, spectral transforms, derivative computations, and matrix operations. Benchmarks comparing dense and sparse solvers were included, providing a foundation for performance analysis in scientific computing with NumPy, SciPy, and Matplotlib integration.
Overall Statistics
Feature vs Bugs
100%Features
Repository Contributions
3Total
Bugs
0
Commits
3
Features
1
Lines of code
959
Activity Months1
Your Network
6 peopleShared Repositories
6
Work History
March 2025
3 Commits • 1 Features
Mar 1, 2025March 2025 monthly work summary for Diff4Earth/ige-jaxathon-2025 focused on delivering a new sparse spectral solver performance analysis framework in JAX, with optional multi-GPU scaling and a demonstration notebook. The work established a reproducible performance analysis pipeline for Poisson-like spectral problems, including problem setup, matrix generation, GMRES testing, and preconditioner exploration, plus an option to scale across GPUs using jax.pmap. A comprehensive Jupyter notebook demonstrates a sparse spectral solver (Segment class, transforms, derivatives, matrix operations) solving Poisson equations with a Galerkin basis and includes dense-vs-sparse solver benchmarks.
3 Commits • 1 Features
Mar 1, 2025March 2025 monthly work summary for Diff4Earth/ige-jaxathon-2025 focused on delivering a new sparse spectral solver performance analysis framework in JAX, with optional multi-GPU scaling and a demonstration notebook. The work established a reproducible performance analysis pipeline for Poisson-like spectral problems, including problem setup, matrix generation, GMRES testing, and preconditioner exploration, plus an option to scale across GPUs using jax.pmap. A comprehensive Jupyter notebook demonstrates a sparse spectral solver (Segment class, transforms, derivatives, matrix operations) solving Poisson equations with a Galerkin basis and includes dense-vs-sparse solver benchmarks.
March 2025
Activity
Loading activity data...
Quality Metrics
Correctness86.6%
Maintainability86.6%
Architecture86.6%
Performance86.6%
AI Usage20.0%
Skills & Technologies
Programming Languages
Jupyter NotebookMarkdownPython
Technical Skills
JAXLinear AlgebraMatplotlibNumPyNumerical AnalysisNumerical MethodsPartial Differential EquationsPerformance AnalysisSciPyScientific ComputingSparse MatricesSpectral Methods
Repositories Contributed To
Overview of all repositories you've contributed to across your timeline