1. 10 Aug, 2020 1 commit
  2. 06 Apr, 2020 1 commit
    • Andreas Marek's avatar
      AVX kernels should not use FMA · 2b2bb0d2
      Andreas Marek authored
      Due to an error, on systems where AVX-2 (FMA) was supported, also
      the "pure" AVX kernels used FMA instructions. This has been fixed
      now
      2b2bb0d2
  3. 20 Nov, 2019 1 commit
    • Wenzhe Yu's avatar
      Rewrite compute_hh_trafo CUDA kernels · 6cd5a4f1
      Wenzhe Yu authored
      * Switch to a simple non-WY algorithm
      * Unify real and complex cases
      * Update reduction kernel
      * Use __shfl_xor_sync for warp reduce (CUDA 9+)
      * Support 2^n block size, n = 1,2,...,10
      * Use templates when possible
      * Clean up unused CUDA functions
      * Increase default stripe width when using GPU
      6cd5a4f1
  4. 29 May, 2019 2 commits
  5. 07 Mar, 2019 1 commit
  6. 05 Mar, 2019 3 commits
  7. 04 Mar, 2019 1 commit
  8. 01 Mar, 2019 2 commits
  9. 25 Feb, 2019 1 commit
  10. 24 Feb, 2019 1 commit
  11. 20 Feb, 2019 1 commit
  12. 07 May, 2018 1 commit
  13. 14 Dec, 2017 2 commits
  14. 23 Nov, 2017 2 commits
  15. 20 Nov, 2017 2 commits
  16. 18 Nov, 2017 2 commits
  17. 03 Aug, 2017 1 commit
  18. 01 Aug, 2017 1 commit
  19. 30 Jul, 2017 2 commits
  20. 26 May, 2017 1 commit
  21. 23 May, 2017 1 commit
  22. 22 May, 2017 3 commits
  23. 21 May, 2017 1 commit
  24. 18 May, 2017 1 commit
    • Lorenz Huedepohl's avatar
      Some cleanup of the "time_evp*" entries · 768db157
      Lorenz Huedepohl authored
      We decided not to provide this information via the key/value store, but
      rather to use the ftimings module for that.
      
      @loh will take care to implement ftimings in the new API
      @amarek wants to port the legacy API so that the old global module
        variables still contain valid values
      768db157
  25. 16 May, 2017 1 commit
    • Lorenz Huedepohl's avatar
      Working version of ELPA with new API · 3e42d4be
      Lorenz Huedepohl authored
      Still missing is the compatibility layer, currently it only compiles
      when configure is called with
      
        --disable-legacy
      
      Also, a more general solution to parameter passing via environment
      variables would be nice.
      3e42d4be
  26. 20 Apr, 2017 1 commit
  27. 06 Apr, 2017 1 commit
  28. 29 Mar, 2017 1 commit
    • Andreas Marek's avatar
      Rename cuda functions · 22a12154
      Andreas Marek authored
      - the functions now contain the appropiate real/complex in their
        name
      - unused functions have been removed as cleanup
      22a12154
  29. 26 Mar, 2017 1 commit