std::linalg: Linear Algebra Coming to Standard C++
summary
Proposal claims a C++ standard linear algebra layer that uses mdspan views, maps classic BLAS kernels into idiomatic templates, and already ships two implementations working on C++17+ with vendor dispatch. Examples include an in‑place, recursive Cholesky factorization that returns failure indices, transpose and scalar accessors implemented as view transforms, and backends calling optimized BLAS on Hopper GPUs and Intel CPUs with comparable throughput for large matrices.
AI-written summary · about
chapters (8)
- 0:00std::linalg proposal intro
- 3:36Motivating linear algebra example
- 6:24Linear algebra layers
- 12:23Performance trade-offs
- 23:47Extending parallel algorithms
- 30:53BLAS history
- 47:34Cholesky factorization deep dive
- 57:57Q&A
related talks
01h 18mC++ and Linear Algebra
01h 00mIntroducing a matrix class to the C++ standard library
01h 24mTyped Linear Algebra - How to Not Crash on Mars
57mGraphBLAS: Building a C++ Matrix API for Graph Algorithms
52mC++ Standard Parallelism
32mA Primer (or Refresher) On Linear Algebra for Data Science
32mRandomized Numerical Linear Algebra: Overview
01h 31mPast, Present and Future of Randomized Numerical Linear Algebra II