LLNL/Aluminum

Issues

• Check buffers in debug mode

  #190 opened a year ago by ndryden
  0

• Request: Flag that sets all streams to the default stream

  #214 opened a year ago by tbennun
  0

• [1.4.1] Tests crash

  #211 opened a year ago by yurivict
  1

• Tests use invalid integer types in std::uniform_int_distribution

  #209 opened a year ago by yurivict
  2

• I'm a newbie. Can I ask you a question

  #205 opened a year ago by lxz12
  2

• No refenrence to 'hwloc_linux_parse_cpumap_file'

  #203 opened a year ago by hanfluid
  4

• Performance by OpenMPI vs MPICH

  #204 opened a year ago by hanfluid
  1

• Update examples

  #189 opened a year ago by ndryden
  1

• In-place MPI scatter segfaults on one processor

  #90 opened 2 years ago by ndryden
  2

• Tests crash

  #179 opened 2 years ago by yurivict
  2

• Threads linkage isn't propagated to CMake export

  #184 opened 2 years ago by benson31
  0

• Hang in progress engine binding

  #182 opened 2 years ago by ndryden
  0

• Traces are not thread-safe

  #177 opened 2 years ago by ndryden
  1

• Support bfloat16

  #170 opened 2 years ago by ndryden
  0

• Support half in MPI and Host-Transfer

  #171 opened 2 years ago by ndryden
  0

• Testing half

  #104 opened 2 years ago by ndryden
  0

• Optimize NCCL collectives

  #115 opened 4 years ago by ndryden
  0

• Support CMake components in export

  #158 opened 2 years ago by benson31
  1

• Support no progress engine

  #118 opened 2 years ago by ndryden
  0

• Progress engine cleanup

  #119 opened 2 years ago by ndryden
  0

• Support HIP stream memory operations

  #156 opened 2 years ago by ndryden
  0

• Progress engine binding on HIP/AMD

  #85 opened 2 years ago by ndryden
  1

• RCCL @ ROCm 3.6.0 doesn't support point-to-point

  #79 opened 2 years ago by benson31
  1

• Fix cpuset discovery with rocm smi

  #149 opened 2 years ago by dmcdougall
  0

• Thread safety

  #117 opened 3 years ago by ndryden
  2

• Early exit for trivial NCCL collectives

  #50 opened 3 years ago by ndryden
  0

• Reorganize MPI backend allreduces

  #76 opened 3 years ago by ndryden
  0

• MPSC queue debug checks

  #132 opened 3 years ago by ndryden
  0

• Host-transfer multi-threading issue

  #133 opened 3 years ago by ndryden
  0

• Better memory pools

  #116 opened 4 years ago by ndryden
  0

• Serialized MPI

  #122 opened 4 years ago by ndryden
  1

• MPI communicator should be wrapped with shared_ptr

  #18 opened 4 years ago by naoyam
  1

• NCCL inplace reduce-scatterv trashes rank 0 buffer

  #110 opened 4 years ago by ndryden
  0

• Host-transfer allgather fails

  #106 opened 4 years ago by ndryden
  0

• Changing a CUDA stream of a CUDA communicator should be disabled

  #19 opened 4 years ago by naoyam
  0

• CUDA Communicator's copy should duplicate a CUDA stream.

  #20 opened 4 years ago by naoyam
  0

• Rooted in-place NCCL calls wrong

  #74 opened 4 years ago by ndryden
  1

• Host-transfer allreduce on one processor fails

  #89 opened 4 years ago by ndryden
  0

• Could not get NUMA node

  #45 opened 4 years ago by mxmlnkn
  1

• Improve progress thread binding

  #64 opened 4 years ago by ndryden
  0

• GPU memory pool

  #77 opened 4 years ago by ndryden
  0

• NCCL group limits

  #75 opened 4 years ago by ndryden
  0

• Fix coding style

  #58 opened 5 years ago by ndryden
  0

• Consistent issues with the RingMPICUDA initialization

  #5 opened 5 years ago by benson31
  5

• Build breaks because of missing 'cuda.h' when ALUMINUM_ENABLE_CUDA=OFF

  #43 opened 6 years ago by yurivict
  0

• Add note and CMake requirement for minimum CUDA version

  #39 opened 6 years ago by mxmlnkn
  1

• Algorithm types for different collectives

  #4 opened 6 years ago by ndryden
  0

• In-place NCCL allgather is incorrect

  #22 opened 6 years ago by ndryden
  0

• In-place NCCL reduce-scatter is not in-place

  #21 opened 6 years ago by ndryden
  0