JuliaGPU/CUDA.jl

Schur decomposition

Open

#2,077 opened on Sep 11, 2023

View on GitHub
 (1 comment) (0 reactions) (0 assignees)Julia (274 forks)batch import
enhancementhelp wanted

Repository metrics

Stars
 (1,408 stars)
PR merge metrics
 (Avg merge 5d 5h) (16 merged PRs in 30d)

Description

Schur decomposition is implemented efficiently on GPU in PyTorch, Jax, and MatLab. It is an essential ingredient for the efficient computation of matrix square root (sqrtm).

It would be ideal to have this decomposition implemented directly in CUDA.jl. The only alternative I see at the moment is full-on diagonalization which is less efficient and less stable.

Contributor guide

Research direction
Study the implementation of existing matrix decompositions in CUDA.jl (e.g., eigendecomposition) to understand the code structure. Then implement Schur decomposition using cuSOLVER's gees function. Reference the PyTorch/JAX implementations for guidance.
Tech stack
None
Domain
backendperformance
Issue type
Feature
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
4
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
Over 1 week
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
Active
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
Clear
Prerequisites
JuliaCUDALinear AlgebraGit
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
20