C++ Libraries for Scientific Software Development
by
Version Now Available
We are thrilled to announce the release of version , now available for download. This release brings substantial improvements across both and modules. Key enhancements to include expanded CSR matrix support with Pardiso integration, along with optimized template-based validation checks for enhanced compile-time performance. The module features major additions including complete eigenvalue decomposition (GEEV) capabilities, a sophisticated virtual expression system enabling lazy evaluation for GPU operations, and an extensive functional algebra API supporting inner products, outer products, and symmetric/Hermitian matrix operations. Additional improvements include enhanced sparse matrix operations with CSR and CSC format support, comprehensive cuBLAS handler enhancements, and numerous bug fixes across both modules. Complete documentation and updated API references are now available to help developers leverage these advanced computational capabilities.
About
is a suite of C++11 libraries engineered for advanced scientific and computational development. While traditional functional programming frameworks provide the high-performance capabilities necessary for production-level code, they often present a steep learning curve characterized by exhaustive documentation and overly complex API argument structures. is designed to bridge the gap between these established legacy frameworks and modern object-oriented paradigms, delivering a streamlined development experience without compromising on efficiency, robustness, or performance.
Features
The objective of this project is to develop a suite of dedicated, parallel, and portable C++ libraries featuring a streamlined and intuitive API to simplify the development process. The project targets the following computational areas:
- Matrix Algebra General-purpose dense, sparse, and low-rank matrix algebra
- Linear Systems High-performance linear solvers and Eigen/QR/SVD decompositions
- Architectural Design Minimal operator design utilizing a virtual operation layer
- Distributed Computing Message Passing Interface (MPI) utilities
- GPU Accelerated Computing GPU acceleration Interface over
Software
(Compact Linear Algebra Parallel Portable Package) a high-performance linear algebra library that leverages advanced object-oriented programming (OOP) design patterns. Its intuitive API is engineered to facilitate the development of high-quality code with minimal overhead and effort.
is a lightweight C++ interface built upon and libraries, engineered to simplify linear algebra GPU computations on accelerators. The current release includes fundamental memory management, vector and matrix handling on GPU architectures, and now features eigenvalue decomposition capabilities, a sophisticated virtual expression system for lazy evaluation, and an extensive functional algebra API supporting inner products, outer products, and symmetric/Hermitian matrix operations. As an add-on to , adheres to the same design patterns, enabling streamlined GPU development while maintaining the intuitive interface and robust numerical framework provided by the underlying library.
(Compact Message Passing Interface) a modern C++ interface engineered to simplify distributed computing and inter-node communication significantly. Currently in the development phase, is scheduled for release as an integral component of the suite.
Download
The source code for is hosted on GitHub, where you can also find comprehensive build and installation instructions.
