v0.3.1

CLA3P Module

New Features

• LapackSyev Eigensolver: Symmetric/Hermitian eigenvalue decomposition on CPU
  • Added syev, heev, and geev LAPACK proxies
  • Implemented LapackSyev class for symmetric and Hermitian eigenproblems
• VirtualStrided Support: Dense matrix diagonal can now be referenced via VirtualStrided class

Bug Fixes

• Fixed multiplication dimension checks in mult_dim_check() for matrix-vector operations
• Fixed update operators for both dense and sparse cases

Improvements

• Header Include Restructuring: All internal include paths reorganized into dedicated subdirectories; all includes now use angle-bracket form
• Renamed dns?D file family to dns?d for consistent naming conventions
• Replaced cla3p::svdPolicy_t::Limited with Economy throughout the codebase

---

cuLite Module

New Features

• Dense Object Fill: Added fill() member function and operator=(Scalar) for dense vectors and matrices
• Extended cuSOLVER Support: Additional cuSOLVER operations exposed through the handler interface
• SVD and GEEV Improvements:
  • Re-implemented GESVD fix for S-size bug on CUDA
  • Stabilized DefaultSVD and LapackGeev constructors
  • Added additional members to LapackGeev
  • Fixed complex SVD implementation
  • Added examples and tutorial/module documentation for geev and gesvd
• New Utility Class: GridXD for GPU kernel launch configuration

Bug Fixes

• Fixed complex SVD implementation for correctness
• Fixed complex_t class arithmetic operators (+, -)
• Fixed update operators for both dense and sparse cases

Improvements

• Implementation Split: Moved CuBlasHandler, CuSolverHandler, CuSparseHandler, and cuSPARSE proxy template instantiations from header-only to .cpp files with explicit specializations, reducing compile times
• Header Include Restructuring: All internal include paths reorganized into dedicated subdirectories; all includes now use angle-bracket form

v0.3.0

SimuliCore v0.3.0 (2026-03-26)

CLA3P Module

New Features

• Enhanced Pardiso Integration:
  • CSR matrices now supported in Full and Upper storage formats
  • Improved sparse linear system solver capabilities
• Template-Based Validation: Outer checks modernized with meta classes for compile-time optimization

Improvements

• Code Quality:
  • Type aliases used consistently throughout
  • Removed obsolete comments
  • Enhanced virtual expression system maintenance
  • Additional validation in outer product operations
  • Updated namespace usage and syntax
• Documentation:
  • Fixed bind parameter documentation
  • Updated syntax documentation
  • Improved example programs clarity

Bug Fixes

• Enhanced dimension checking using meta classes in mult_dim_check()
• Improved property validation code

---

cuLite Module

New Features

• Eigenvalue Decomposition (GEEV):

  • Full eigenvalue/eigenvector computation for general matrices
  • Support for eigenvalues-only and full eigenpair computation
  • LapackGeev proxy functions for CPU-based operations
  • Dedicated cuSolver proxy file for eigenvalue operations

• Virtual Expression System:

  • Lazy evaluation for GPU matrices and vectors
  • Virtual row vector, conjugate, transpose operations
  • Virtual arithmetic operations (+, -, *,scaling)
  • Outer product virtual expressions
  • Constructor support for virtual expressions
  • Transpose and conjugate-transpose members for sparse matrices

• Functional Algebra Operations (via culite::ops namespace):

  • Inner product (dot, dotc)
  • Vector addition and scaling (axpy)
  • Matrix-vector/matrix-matrix multiplication (symmetric/Hermitian support)
  • Outer product operations (ger, gerc, syr, her)
  • Rank-1 updates for symmetric/Hermitian matrices
  • Functional update operations for dense and sparse matrices
  • Diagonal matrix-vector multiplication for CSX matrices

• Sparse Matrix Operations:

  • Custom CSX addition kernels
  • Host-device copy operations for CSR matrices
  • Inner and outer algebra operators for sparse matrices

• cuBLAS Handler Enhancements:

  • copy(), dotc(), axpy() members
  • ger(), gerc(), syr(), her() for rank-1 updates

Improvements

• Documentation:
  • Comprehensive Doxygen documentation for modules
  • Enhanced update operations documentation
  • Fixed various documentation references
• Code Organization:
  • Extracted legacy cuSPARSE proxies to separate file
  • Improved visibility control for virtual expressions
  • Enhanced dns1D code structure
• Utilities:
  • Added memSetZero() utility for GPU memory initialization
  • Complex number arithmetic operators (+, -)

Bug Fixes

• Fixed compilation error in operator>>() for stream operations
• Fixed CuSolver GEEV implementation issues
• Added const casts to fix cuSPARSE proxy compilation issues
• Resolved virtual expression conflicts
• Fixed argument handling in update operations

API Changes

• Modified add() function argument lists for consistency
• Enhanced operator support for update operations (+=, -=, *=)
• Added multiplication operators for matrix-vector and matrix-matrix operations

v0.2.0

We are pleased to announce the release of SimuliCore 0.2.0. This update moves beyond the initial architectural framework to introduce functional capabilities and improved system stability:

• CLA3P: Added support for the CSR (Compressed Sparse Row) format, enabling efficient storage and computation for sparse linear algebra operations.
• cuLite: The first functional release of our GPU-acceleration layer, providing the baseline tools for high-performance parallel computing.
• System Robustness: Included various bug fixes and stability enhancements across the entire suite to ensure a more reliable development experience.

For detailed technical notes and implementation details, please visit the SimuliCore Portal.

v0.1.0

We are pleased to announce the initial release of the SimuliCore suite. This foundational version introduces our core computational framework, featuring:

• CLA3P: The primary high-performance linear algebra library.
• CMPI: The initial skeletal framework for message-passing interface integration (empty).
• cuLite: The baseline architecture for GPU-accelerated computing (empty).

For technical documentation and integration guides, please visit the SimuliCore Documentation Portal