rocSOLVER 3.26.0 for ROCm 6.2.1

rocSOLVER code for ROCm 6.2.1 did not change. The library was rebuilt for the updated ROCm 6.2.1 stack.

rocSOLVER 3.26.0 for ROCm 6.2.0

Added

• 64-bit APIs for existing functions:
  • GETF2_64 (with batched and strided_batched versions)
  • GETRF_64 (with batched and strided_batched versions)
  • GETRS_64 (with batched and strided_batched versions)
• Added gfx900 to default build targets.
• Partial eigenvalue decomposition routines for symmetric/hermitian matrices using Divide & Conquer and Bisection:
  • SYEVDX (with batched and strided_batched versions)
  • HEEVDX (with batched and strided_batched versions)
• Partial generalized symmetric/hermitian-definite eigenvalue decomposition using Divide & Conquer and Bisection:
  • SYGVDX (with batched and strided_batched versions)
  • HEGVDX (with batched and strided_batched versions)

Optimized

• Improved performance of Cholesky factorization.
• Improved performance of splitlu to extract the L and U triangular matrices from the result of sparse factorization matrix M, where M = (L - eye) + U.

Changed

• Renamed install script arguments of the form *_dir to *-path. Arguments of the form *_dir remain functional for
  backwards compatibility.
• Functions working with arrays of size n - 1 can now accept null pointers when n = 1.

Fixed

• Fixed potential accuracy degradation in SYEVJ/HEEVJ for inputs with small eigenvalues.

rocSOLVER 3.25.0 for ROCm 6.1.2

rocSOLVER code for ROCm 6.1.2 did not change. The library was rebuilt for the updated ROCm 6.1.2 stack.

rocSOLVER 3.25.0 for ROCm 6.1.1

rocSOLVER code for ROCm 6.1.1 did not change. The library was rebuilt for the updated ROCm 6.1.1 stack.

rocSOLVER 3.25.0 for ROCm 6.1.0

Added

• Eigensolver routines for symmetric/hermitian matrices using Divide & Conquer and Jacobi algorithm:
  • SYEVDJ (with batched and strided_batched versions)
  • HEEVDJ (with batched and strided_batched versions)
• Generalized symmetric/hermitian-definite eigensolvers using Divide & Conquer and Jacobi algorithm:
  • SYGVDJ (with batched and strided_batched versions)
  • HEGVDJ (with batched and strided_batched versions)

Changed

• Relaxed array length requirements for GESVDX with rocblas_srange_index.

Removed

• Removed gfx803 and gfx900 from default build targets.

Fixed

• Corrected singular vector normalization in BDSVDX and GESVDX
• Fixed potential memory access fault in STEIN, SYEVX/HEEVX, SYGVX/HEGVX, BDSVDX and GESVDX

rocSOLVER 3.24.0 for ROCm 6.0.2

rocSOLVER code for ROCm 6.0.2 did not change. The library was rebuilt for the updated ROCm 6.0.2 stack.

rocSOLVER 3.24.0 for ROCm 6.0.0

Added

• Cholesky refactorization for sparse matrices
  • CSRRF_REFACTCHOL
• Added rocsolver_rfinfo_mode and the ability to specify the desired refactorization routine (see rocsolver_set_rfinfo_mode).

Changed

• CSRRF_ANALYSIS and CSRRF_SOLVE now support sparse Cholesky factorization

rocSOLVER 3.23.0 for ROCm 5.7.1

rocSOLVER code for ROCm 5.7.1 did not change. The library was rebuilt for the updated ROCm 5.7.1 stack.

rocSOLVER 3.23.0 for ROCm 5.7.0

Added

• LU factorization without pivoting for block tridiagonal matrices:
  • GEBLTTRF_NPVT now supports interleaved_batched format
• Linear system solver without pivoting for block tridiagonal matrices:
  • GEBLTTRS_NPVT now supports interleaved_batched format

Fixed

• Fixed stack overflow in sparse tests on Windows

Changed

• Changed rocsolver-test sparse input data search paths to be relative to the test executable
• Changed build scripts to default to compressed debug symbols in Debug builds

rocSOLVER 3.22.0 for ROCm 5.6.1

rocSOLVER code for ROCm 5.6.1 did not change. The library was rebuilt for the updated ROCm 5.6.1 stack.