openucx/xccl issues and pull requests

#135 - Fix block source address

Pull Request - State: closed - Opened by lappazos over 3 years ago - 2 comments

#134 - Add non-inline UMR

Pull Request - State: closed - Opened by lappazos over 3 years ago

#133 - CI: ubuntu update

Pull Request - State: closed - Opened by vspetrov over 3 years ago

#132 - CI: ubuntu update

Pull Request - State: closed - Opened by vspetrov over 3 years ago

#131 - Cherry-picking commits from master

Pull Request - State: closed - Opened by pavanbalaji over 3 years ago - 3 comments

#130 - TEAM/NCCL: fix datarace in collective post

Pull Request - State: closed - Opened by Sergei-Lebedev over 3 years ago

#129 - Multi-cluster collectives

Pull Request - State: closed - Opened by pavanbalaji over 3 years ago - 9 comments

#128 - Bytes_Skip limit

Pull Request - State: closed - Opened by lappazos over 3 years ago

#127 - TEST: add cuda cppflags for tests build

Pull Request - State: closed - Opened by Sergei-Lebedev over 3 years ago

#126 - Add DC QPs

Pull Request - State: closed - Opened by lappazos over 3 years ago

#125 - CUDA: set ucs cflags for nvcc

Pull Request - State: closed - Opened by Sergei-Lebedev over 3 years ago

#124 - xccl/mc: Added a new memcpy_async operation

Pull Request - State: closed - Opened by pavanbalaji over 3 years ago - 5 comments

#123 - xccl/ucx: fix allgather for IN_PLACE buffers

Pull Request - State: closed - Opened by pavanbalaji over 3 years ago

#122 - xccl/core: Bug fix in the OOB allgather code

Pull Request - State: closed - Opened by pavanbalaji over 3 years ago - 4 comments

#121 - Add block size limit

Pull Request - State: closed - Opened by lappazos over 3 years ago

#120 - Fix fanin race in first op

Pull Request - State: closed - Opened by lappazos almost 4 years ago

#119 - Remove HW limitations choice

Pull Request - State: closed - Opened by lappazos almost 4 years ago

#118 - Change UMR Layout

Pull Request - State: closed - Opened by lappazos almost 4 years ago

#117 - NVCC cannot find UCS header files

Issue - State: closed - Opened by kingchc almost 4 years ago - 3 comments

#116 - CI: change sockaddr setting according to the changes in ucx master

Pull Request - State: closed - Opened by vspetrov almost 4 years ago

#115 - Cyclic Shift

Pull Request - State: closed - Opened by lappazos almost 4 years ago

#114 - Fix block size

Pull Request - State: closed - Opened by lappazos almost 4 years ago

#113 - TEAM/MHBA: Adds mkey cache

Pull Request - State: closed - Opened by vspetrov almost 4 years ago

#112 - TEAM/UCX: don't use event query with dynamic kernel

Pull Request - State: open - Opened by Sergei-Lebedev almost 4 years ago

#111 - NCCL: Add nccl datatype for Byte(Uint8) and Char(Int8)

Pull Request - State: closed - Opened by bureddy almost 4 years ago

#110 - Sync stream on entry

Pull Request - State: closed - Opened by Sergei-Lebedev almost 4 years ago

#109 - TEAM/UCX: switch to nbx api

Pull Request - State: closed - Opened by Sergei-Lebedev almost 4 years ago

#108 - Cyclic send blocks

Pull Request - State: closed - Opened by lappazos almost 4 years ago

#107 - UCX: Option to pre-register communication buffer to populate the rcache.

Pull Request - State: closed - Opened by bureddy almost 4 years ago - 1 comment

#106 - Adding the XCCL DPU team, and DPU daemon

Pull Request - State: open - Opened by janjust almost 4 years ago - 4 comments

#105 - Replace inter node barrier

Pull Request - State: open - Opened by lappazos almost 4 years ago

#104 - Filter UMR Calls

Pull Request - State: closed - Opened by lappazos almost 4 years ago

#103 - Remove un-needed barriers in team creation

Pull Request - State: closed - Opened by lappazos almost 4 years ago

#102 - Multiple transpose buff

Pull Request - State: open - Opened by lappazos almost 4 years ago

#101 - Add Net CQs

Pull Request - State: open - Opened by lappazos almost 4 years ago

#100 - Rcache

Pull Request - State: closed - Opened by lappazos almost 4 years ago - 2 comments

#99 - Add Block_size cfg

Pull Request - State: closed - Opened by lappazos almost 4 years ago - 1 comment

#98 - API: added stream as coll argument

Pull Request - State: closed - Opened by Sergei-Lebedev almost 4 years ago

#97 - Create nccl stream with highest priority

Pull Request - State: closed - Opened by srinivas212 almost 4 years ago - 1 comment

#96 - Refactoring and cleanup

Pull Request - State: closed - Opened by vspetrov almost 4 years ago

#95 - TEAM/NCCL: broadcast

Pull Request - State: closed - Opened by Sergei-Lebedev almost 4 years ago

#94 - Transpose Data

Pull Request - State: closed - Opened by lappazos almost 4 years ago

#93 - mhba: ASR rank_map

Pull Request - State: closed - Opened by vspetrov almost 4 years ago

#92 - Full Send Recv

Pull Request - State: closed - Opened by lappazos almost 4 years ago

#91 - Full Send & Recv

Pull Request - State: closed - Opened by lappazos almost 4 years ago

#90 - team/mhba: fix devname setting

Pull Request - State: closed - Opened by vspetrov almost 4 years ago

#89 - team/mhba: fix devname setting

Pull Request - State: closed - Opened by vspetrov almost 4 years ago

#88 - GH: disable use of rdma_cm in the GH actions tests

Pull Request - State: closed - Opened by vspetrov almost 4 years ago

#87 - GH: disable use of rdma_cm in the GH actions tests

Pull Request - State: closed - Opened by vspetrov almost 4 years ago

#86 - debug [don't merge]

Pull Request - State: closed - Opened by vspetrov almost 4 years ago

#85 - dummy pr for debug

Pull Request - State: closed - Opened by vspetrov almost 4 years ago

#84 - TEAM_LIB: vmc -> hmc

Pull Request - State: closed - Opened by vspetrov almost 4 years ago

#83 - TEAM_LIB/UCX: set estimated num ppn for ucp context

Pull Request - State: closed - Opened by vspetrov almost 4 years ago

#82 - Shared PD

Pull Request - State: closed - Opened by lappazos almost 4 years ago - 2 comments

#81 - BUILD: fix cuda ldflags

Pull Request - State: closed - Opened by Sergei-Lebedev almost 4 years ago

#81 - BUILD: fix cuda ldflags

Pull Request - State: closed - Opened by Sergei-Lebedev almost 4 years ago

#80 - TEAM/UCX: pairwise alltoall with barrier

Pull Request - State: closed - Opened by Sergei-Lebedev almost 4 years ago

#79 - TEAM/MHBA: Add team mkey

Pull Request - State: closed - Opened by lappazos about 4 years ago

#78 - CORE: pass config list to table register

Pull Request - State: closed - Opened by Sergei-Lebedev about 4 years ago

#77 - TEAM/MRAIL: disable by default

Pull Request - State: closed - Opened by Sergei-Lebedev about 4 years ago

#76 - team/mhba: add umr and mkey api

Pull Request - State: closed - Opened by lappazos about 4 years ago

#75 - TEAM/NCCL: add allgather

Pull Request - State: closed - Opened by Sergei-Lebedev about 4 years ago

#74 - team_lib/mhba: adds ucx team creation and barrier

Pull Request - State: closed - Opened by vspetrov about 4 years ago

#73 - Fix nonbound bcast in heir team

Pull Request - State: closed - Opened by Sergei-Lebedev about 4 years ago

#72 - team_lib/mhba: RC QPs for ASRs subgroup

Pull Request - State: closed - Opened by vspetrov about 4 years ago

#71 - team_lib/mhba: schedule/tasks for the new A2A

Pull Request - State: closed - Opened by vspetrov about 4 years ago

#70 - core/progress: call progress queue even if no progress fn of tl_ctx

Pull Request - State: closed - Opened by vspetrov about 4 years ago

#69 - team_lib/mhba: node fanin/fanout using shm

Pull Request - State: closed - Opened by vspetrov about 4 years ago

#68 - team_lib/mhba: allocate shared memory storage

Pull Request - State: closed - Opened by vspetrov about 4 years ago

#67 - xccl/sbgp: sbgp oob collectives

Pull Request - State: closed - Opened by vspetrov about 4 years ago

#66 - Sbgp to common

Pull Request - State: closed - Opened by vspetrov about 4 years ago

#65 - TEAM/HIER: 2lvl alltoall

Pull Request - State: closed - Opened by Sergei-Lebedev about 4 years ago

#64 - Feature request: ability to set priority for CUDA streams

Issue - State: closed - Opened by srinivas212 about 4 years ago - 2 comments
Labels: enhancement

#63 - Team/hier common task progress/complete handling

Pull Request - State: closed - Opened by vspetrov about 4 years ago

#62 - TEAM/NCCL: add alltoall

Pull Request - State: closed - Opened by Sergei-Lebedev about 4 years ago

#61 - XCCL: fix ucx configure warning and add min required version check

Pull Request - State: closed - Opened by Sergei-Lebedev about 4 years ago

#60 - TEAM/NCCL: add nccl team library

Pull Request - State: closed - Opened by Sergei-Lebedev about 4 years ago

#59 - CI debug print [DON'T MERGE]

Pull Request - State: closed - Opened by vspetrov about 4 years ago
Labels: WIP-DNM

#58 - Hier schedule fix no socket

Pull Request - State: closed - Opened by vspetrov about 4 years ago - 3 comments

#57 - Poor performance with NVLink

Issue - State: open - Opened by froody about 4 years ago - 2 comments

#56 - Segfault creating process group with 1 member

Issue - State: closed - Opened by froody about 4 years ago - 3 comments

#55 - fixes to build with ucx + cuda

Pull Request - State: closed - Opened by froody about 4 years ago - 1 comment

#54 - TEAM/UCX: set ucp worker thread mode according to context thread mode

Pull Request - State: closed - Opened by Sergei-Lebedev about 4 years ago

#53 - mem component cache

Pull Request - State: closed - Opened by Sergei-Lebedev about 4 years ago

#52 - remove torch ucc

Pull Request - State: closed - Opened by Sergei-Lebedev about 4 years ago

#51 - Centralized progress per task

Pull Request - State: closed - Opened by lappazos about 4 years ago

#50 - TEAM/UCX: use multireduce in sra

Pull Request - State: closed - Opened by Sergei-Lebedev about 4 years ago

#49 - DO NOT MERGE: debug alltoallv

Pull Request - State: closed - Opened by Sergei-Lebedev about 4 years ago

#48 - TEAM/UCX: knomial sra allreduce

Pull Request - State: closed - Opened by Sergei-Lebedev over 4 years ago - 1 comment

#47 - TORCH: add alltoall benchmark

Pull Request - State: closed - Opened by Sergei-Lebedev over 4 years ago

#46 - XCCL: fix build with nonstandard cuda path

Pull Request - State: closed - Opened by Sergei-Lebedev over 4 years ago

#45 - TEAM/UCX: add ring allgather

Pull Request - State: closed - Opened by Sergei-Lebedev over 4 years ago

#44 - TORCH: add alltoallv support

Pull Request - State: closed - Opened by Sergei-Lebedev over 4 years ago

#43 - TEAM/UCX: add pairwise alltoallv

Pull Request - State: closed - Opened by Sergei-Lebedev over 4 years ago

#42 - TORCH: add torch ddp plugin

Pull Request - State: closed - Opened by Sergei-Lebedev over 4 years ago

#41 - TEAM/UCX: inplace alltoall

Pull Request - State: closed - Opened by Sergei-Lebedev over 4 years ago

#40 - ompi_coll_xccl patch update

Pull Request - State: closed - Opened by vspetrov over 4 years ago

#39 - team/ucx: alltoall pairwise exchange

Pull Request - State: closed - Opened by vspetrov over 4 years ago

#38 - ompi/coll/xccl: adds XCCL_REDUCE

Pull Request - State: closed - Opened by vspetrov over 4 years ago

#37 - xccl.h: context query

Pull Request - State: closed - Opened by vspetrov over 4 years ago

GitHub / openucx/xccl issues and pull requests