Release v0.3.0

Significant redesign of the library based on the experience of the first two versions.

Release v0.2.1

• Removed Apple Metal files
• Fixed build scripts
• Made Sleef optional

v0.2.0

• Integrated SIMD operations.
• Updated to MPI and Lapack routines.
• Many small fixes.

v0.1.1

What's changed

• Minor release fixing generation of documentation

Full Changelog: v0.1.0...v0.1.1

v0.1.0

Features

• n-dimensional array data structures
• All array functionality available for dynamically heap allocated and statically sized stack allocated arrays
• Many array operations support lazy compile-time evaluation, ie an expression like res.fill_from(3 * arr1 + arr2) produces code that writes the result as efficient single for-loop into res.
• Growing support of Lapack functions
• Matrix Multiplication through BLAS
• Users free in their choice of BLAS/Lapack providers using the standard Rust Blas/Lapack infrastructure
• Support for arrays on Apple Metal. So far Metal based matrix multiplication is supported (Cuda in planning as well).
• Sparse CSR and CSC matrices on single nodes and via MPI on multi-nodes (MPI less tested and not fully production ready).
• Umfpack support for the solution of sparse linear system.
• Matrix market import (still early and i/o infrastructure may change)
• A general operator algebra that can wrap arbitrary linear objects
• Conjugate Gradients implemented for the operator algebra with more iterative solvers in the works.

Full Changelog: https://github.com/linalg-rs/rlst/commits/v0.1.0