Skip to content

Latest commit

 

History

History
130 lines (73 loc) · 4.07 KB

CHANGELOG.md

File metadata and controls

130 lines (73 loc) · 4.07 KB
Raw

Version 0.9.7

  • Fix #28: incorrect output when out is provided but the matrix multiplication is trivially all zeros
  • Added scipy sparse classes where MKL is used for matrix multiplication

Version 0.9.6

  • Add wrapper for MKL iterative CG and FGMRES solvers in sparse_dot.solvers

Version 0.9.5

  • Add wrapper for MKL pardiso solver in sparse_dot.solvers

Version 0.9.4

  • Replace deprecated scipy spmatrix.A calls with spmatrix.todense()

Version 0.9.3

  • Directly expose service functions mkl_get_max_threads, mkl_set_num_threads, mkl_set_num_threads_local, mkl_get_version, and mkl_get_version_string.

Version 0.9.2

  • Explicit check for interface env MKL_INTERFACE_LAYER and interface selection in python. Will raise a RuntimeWarning if ILP64 is requested but the 64-bit interface layer is unavailable. Note that the MKL_INTERFACE_LAYER env must be set before this package is imported.

Version 0.9.1

  • Support for out parameter with sparse-sparse multiplication when dense=True

Version 0.9.0

  • Support for scipy sparse arrays (introduced in scipy 1.11)
  • Additional fallback support for finding mkl DLLs with OneAPI

Version 0.8.3

  • Explicit error message when complex data is put into the QR solver
  • Fix edge condition memory leak in the QR solver

Version 0.8.2

  • Corrected ImportError on windows with oneMKL by explicitly library searching for mkl_rt.{i}.dll, where i is between 1 and 5. That should last till 2026.

Version 0.8.1

  • cast=True will now cast to compatible complex floats if one array is complex and one is real

Version 0.8.0

  • Added support for complex data types
  • Refactored _mkl_interface.py into a subpackage

Version 0.7.3

  • Fixed a memory leak when a CSC matrix was multiplied by a dense matrix in column-major format
  • Switched continuous integration from Travis to Github Actions

Version 0.7.2

  • Corrected ImportError on windows with oneMKL by explicitly library searching for mkl_rt.1.dll
  • Switched test runner from nose to pytest

Version 0.7.1

  • Fixed a bug in sparse * dense multiplication when one of the output array dimensions is 1

Version 0.7.0

  • Added support for block sparse row (BSR) format matrices
  • Added sparse_dot_mkl.set_debug_mode(True) as a module-level debug mode flag instead of calling debug mode from dot_product_mkl. Added a number of debug messages for troubleshooting during development.

Version 0.6.0

  • Added an out parameter that will add a matrix multiplication to an already-allocated dense array

Version 0.5.4

  • Updated error message for a specific sparse product multiplication failure

Version 0.5.3

  • Fixed a bug that was hanging MKL threads if scikit-learn was imported prior to sparse_dot_mkl and mkl functions were run from a child process

Version 0.5.2

  • Fixed a bug that was not searching $LD_LIBRARY_PATH for mkl_rt
  • Clarified several error messages

Version 0.5.1

  • Added gram_matrix_mkl to access mkl_sparse_syrk
  • Fixed a bug that was not correctly multiplying vectors for outer products

Version 0.5.0

Version 0.4.1

Version 0.4.0

Version 0.3.4

  • Added tests to raise python error when matrix indices overflow MKL interface type

Version 0.3.3

  • Refactored _sparse_dense.py for efficiency and readability

Version 0.3.2

  • Fixed windows library dll name

Version 0.3.1

  • Fixed incorrect return type in certain rare cases

Version 0.3.0

  • Full support for numpy arrays in row-major and column-major format