Skip to content

CLBlast 1.6.3

Latest
Compare
Choose a tag to compare
@CNugteren CNugteren released this 13 Jun 17:50
2a08197

CLBlast version 1.6.3. Changes since previous release (version 1.6.2):

  • Fixed a bug in the GEMMK=1 kernel (with 2D register tiling) when MWG!=NWG
  • CMake fixes for older versions and for the CUDA backend
  • Added tuned parameters for many devices (see doc/tuning.md)