rola

Rola (Rust Optimized Linear Algebra) is (will be) a Rust linear algebra library aiming to be fast and robust with the intent of use in numerical computing.

The scope of the project is to create fast, robust matrix and vector operations, and to solve Ax = b as quickly and efficiently as possible.

Current Progress

Just trying to get things to work. We'll worry about optimization after we have a working implementation.

Goals

1. Idiomatic, clean Rust
2. Mathematically sound
3. Comparable speed to predecessors

Potentially look into GPU acceleration:

• Vulkan (currently unstable)
• OpenCL
  • https://github.com/luqmana/rust-opencl
  • https://github.com/cogciprocate/ocl

Research

References

• LAPACK FORTRAN, tried and true
• Blaze C++, cutting edge and fast
• Eigen Pure C++, clean and robust
• PLASMA C, makes use of cores and hardware

Study Material & Theory

• Numerical Analysis/Methods
  • Numerical Computation of Matrix Functions (Smith)
  • Fundamental Numerical Methods and Data Analysis (Collins)
  • Numerical Analysis (watermarked & difficult to read) (Ipsen)
• Computation of Eigenvalues
  • Olver (UMN)
  • gatech
• Matrix Inversion Algorithms
  • Fast Algorithms for Sparse Matrix Inverse Computations (Li)
• Texts
  • Numerical Analysis (Burden, Faires)
  • Numerical Analysis (Scott)
  • Lectures in Basic Computation Numerical Analysis (McDonough)
• arXiv.org
  • Numerical Analysis Category