Selected work

Articrafts

Papers, code, reports, and experiments

Each piece is kept with a short note, its versions, and the files available alongside it—and can keep growing through later revisions.

Showing all 5 works.

Work / cinder-cuda-tensor-library

Cinder: Design and Implementation of a Compact CUDA Tensor Library

A design record for a compact Python/C++/CUDA dense tensor library, covering tensor semantics, shape planning, device-memory ownership, kernel protocols, and extension risks.

Last activity
Files
Papers
Release
v1.0.0
  • CUDA
  • tensor library
  • C++
  • Python
  • systems design

Work / jacobi-svd-locality-kernel-policy

Before MPI: Locality and Kernel Policy in One-Sided Jacobi SVD

A 420-configuration factorial study of one-sided Jacobi SVD across six matrix layouts, five kernel policies, and fourteen matrix families, separating locality, convergence, and parallel-wrapper overhead.

Last activity
Files
Papers
Release
v1.0.0
  • MPI
  • Jacobi SVD
  • locality
  • numerical computing
  • benchmarking

Work / chronosys-modern-cpp-hpc

ChronoSys: Modern C++ High-Performance Programming Through Domain-Driven Design

A parallel-programming report on matrix column dot products and sum reduction, covering domain-driven layering, cache/CUDA/superscalar strategies, performance regression analysis, and agent collaboration.

Last activity
Files
Papers
Release
v1.0.0
  • C++
  • HPC
  • CUDA
  • domain-driven design
  • performance