1. 01 Apr, 2016 1 commit
  2. 27 Mar, 2016 2 commits
  3. 20 Mar, 2016 1 commit
  4. 15 Mar, 2016 1 commit
  5. 13 Mar, 2016 1 commit
  6. 02 Mar, 2016 1 commit
  7. 25 Feb, 2016 1 commit
  8. 19 Feb, 2016 2 commits
  9. 11 Feb, 2016 2 commits
  10. 09 Feb, 2016 1 commit
  11. 04 Feb, 2016 1 commit
  12. 03 Feb, 2016 1 commit
  13. 01 Feb, 2016 2 commits
  14. 23 Jan, 2016 1 commit
  15. 07 Jan, 2016 1 commit
  16. 06 Jan, 2016 1 commit
  17. 03 Jan, 2016 1 commit
    • knopp's avatar
      added avx2 optimized turbo decoder for 16-bit LLR. This decoder parallelizes... · 8d4405bd
      knopp authored
      added avx2 optimized turbo decoder for 16-bit LLR.  This decoder parallelizes by decoding 2 code segments concurrently. requires updates dlsch_decoding.c to identify when new parallel version can be used. other minor changes related to memory allocations for future avx2 optimizations (32-byte alignment).
      8d4405bd
  18. 17 Dec, 2015 1 commit
  19. 14 Dec, 2015 1 commit
    • knopp's avatar
      conversion of some sse4 code to avx2 for turbo encoder. No noticeable... · 1e08469e
      knopp authored
      conversion of some sse4 code to avx2 for turbo encoder.  No noticeable improvement (output expansion for encoder doesn't benefit much because of unpack instruction difference on AVX2.
      
      maxh adjustement in dlsch_demodulation (to be checked further for other mcs)
      
      exit criterion in dlsim to allow estimation of lower BLER.
      1e08469e
  20. 13 Dec, 2015 1 commit
  21. 05 Dec, 2015 1 commit
  22. 04 Dec, 2015 1 commit
  23. 29 Nov, 2015 1 commit
  24. 25 Nov, 2015 1 commit
  25. 24 Nov, 2015 3 commits
  26. 23 Nov, 2015 1 commit
    • Cedric Roux's avatar
      Manage free downlink HARQ processes with a circular list. · f14a401f
      Cedric Roux authored
      The oldest PID (the PID that was used the longest
      TTI ago) is at the head and is the next one that
      will be used.
      
      No test is done to check that when a process is
      allocated it is at least N TTIs old, but some quick
      thinking strongly leads to the 'conclusion' that N
      won't be smaller than 8, which seems to be what
      cat3 UEs require to function properly in the downlink
      direction (from the eNodeB to the UE).
      f14a401f
  27. 18 Nov, 2015 1 commit
  28. 28 Oct, 2015 1 commit
  29. 20 Oct, 2015 1 commit
  30. 19 Oct, 2015 1 commit
  31. 15 Oct, 2015 1 commit
  32. 13 Oct, 2015 2 commits
  33. 12 Oct, 2015 1 commit