Issues

• could testcase run on one node?

  gpudirect/libgdsync#108 · Filter by author protoss1235 opened on Jan 6, 2023
  2

• Support spectrum MPI

  enhancement
  gpudirect/libgdsync#94 · Filter by author pakmarkthub opened on Feb 17, 2021

• Optimize gds_mlx5_post_send

  enhancement
  gpudirect/libgdsync#90 · Filter by author pakmarkthub opened on Feb 3, 2020

• gds_prepare_send should support all types of send_flags

  enhancement
  gpudirect/libgdsync#89 · Filter by author pakmarkthub opened on Feb 3, 2020

• gds_prepare_send should support all types of opcode

  enhancement
  gpudirect/libgdsync#88 · Filter by author pakmarkthub opened on Feb 3, 2020

• Code style violation

  enhancement
  gpudirect/libgdsync#83 · Filter by author pakmarkthub opened on May 20, 2019
  

• get rid of _exp_ types in core.h

  enhancement
  gpudirect/libgdsync#79 · Filter by author drossetti opened on May 1, 2019
  1

  

• detect and enable IBV_EXP_PEER_OP_POLL_NOR_DWORD_CAP

  gpudirect/libgdsync#73 · Filter by author drossetti opened on Jun 7, 2018

• fix assumption in gds_ordinal_from_device() implementation

  bug
  gpudirect/libgdsync#67 · Filter by author drossetti opened on May 22, 2018

• provide optimized APIs accepting GPU device handles or ordinals

  enhancement
  gpudirect/libgdsync#66 · Filter by author drossetti opened on May 22, 2018

• in gds_kernel_latency !peersync case does not work

  bug
  gpudirect/libgdsync#64 · Filter by author drossetti opened on May 17, 2018
  1

• hpgmg and gds_kernel_latency errors on DGX TeslaV100

  bug
  gpudirect/libgdsync#61 · Filter by author e-ago opened on May 11, 2018
  4

• registration cache is not correctly maintained

  bug
  gpudirect/libgdsync#54 · Filter by author drossetti opened on Apr 18, 2018

• flusher can be optimized away in certain cases

  enhancement
  gpudirect/libgdsync#53 · Filter by author drossetti opened on Apr 14, 2018
  1

• GPU context bug in gds_fill_poll when GDS_ENABLE_FLUSHER=1

  bug
  gpudirect/libgdsync#52 · Filter by author drossetti opened on Apr 14, 2018
  

• Question about multi streams operate the same QP and CQ

  question
  gpudirect/libgdsync#48 · Filter by author Hannah-xxl opened on Apr 9, 2018
  6

• Improve, in case of wait on stream (SA model), multiple receives from different peers (CQs) using CUDA threads

  enhancement
  gpudirect/libgdsync#39 · Filter by author e-ago opened on Mar 5, 2018

• Improve, in case of send on stream (SA model), multiple sends to different peers (QPs) using CUDA threads

  enhancement
  gpudirect/libgdsync#38 · Filter by author e-ago opened on Mar 5, 2018

• properly fence sends to same QP

  bug
  gpudirect/libgdsync#28 · Filter by author drossetti opened on Jun 16, 2017

• efficiently support COPY_BLOCK for kernel-side API

  enhancement
  gpudirect/libgdsync#24 · Filter by author drossetti opened on May 31, 2017

• provide a service to long-running GPU kernels to flush incoming RDMA traffic

  gpudirect/libgdsync#23 · Filter by author drossetti opened on May 31, 2017
  

• use verbs res_domain to obtain QPs with dedicated blueflame space

  gpudirect/libgdsync#22 · Filter by author drossetti opened on May 31, 2017
  3

• lack of context initialization in case of host memory

  bug
  gpudirect/libgdsync#18 · Filter by author e-ago opened on May 24, 2017

• add APIs to abort outstanding work related to QP and CQ

  enhancement
  gpudirect/libgdsync#12 · Filter by author drossetti opened on May 19, 2017

• Support RoCE protocol in tests

  gpudirect/libgdsync#10 · Filter by author ksang opened on May 17, 2017
  2