rapidsai/distributed-join

Issues

• Switch to CMake instead of using Makefile directly

  #26 opened 3 years ago by gaohao95
  1

• Investigate device memory usage outside memory pool

  #51 opened 4 years ago by gaohao95
  1

• Memory leak in UCXBufferCommunicator

  #8 opened 5 years ago by nsakharnykh
  2

• Replace `cudaStreamDefault`

  #46 opened 4 years ago by gaohao95
  2

• Code formatting with clang-format

  #40 opened 4 years ago by gaohao95
  0

• Register communication buffers in UCXBufferCommunicator directly

  #38 opened 4 years ago by gaohao95
  0

• Register the whole RMM memory pool upfront

  #33 opened 4 years ago by gaohao95
  0

• Add NVTX range to make Nsight System profiling results easier to understand

  #37 opened 4 years ago by gaohao95
  0

• Docker container

  #20 opened 4 years ago by gaohao95
  0

• Investigate memory usage

  #31 opened 4 years ago by gaohao95
  0

• Improve overlapping efficiency

  #32 opened 4 years ago by gaohao95
  0

• Query parameters from command line

  #4 opened 4 years ago by nsakharnykh
  0

• Use RAII to manage communicator

  #21 opened 4 years ago by gaohao95
  0

• Algorithm description

  #23 opened 4 years ago by mavam
  2

• Move to cuDF/rmm 0.14

  #16 opened 4 years ago by nsakharnykh
  0

• Use cuDF's error checking utilities

  #19 opened 5 years ago by gaohao95
  0

• Test current instructions using older custom cuDF branch

  #1 opened 5 years ago by nsakharnykh
  1

• Update the code to work with cuDF 0.11+ (libcudf++)

  #6 opened 5 years ago by nsakharnykh
  1

• Use all gather instead of broadcast loop for communicating worker addresses

  #3 opened 5 years ago by nsakharnykh
  0

• Remove batch ID from the tag

  #2 opened 5 years ago by nsakharnykh
  0

• Coordinate buffer communicator and default UCX communicator

  #7 opened 5 years ago by nsakharnykh
  0

• Eliminate small custom kernels in favor of lambda functions

  #5 opened 5 years ago by nsakharnykh
  0