issues
search
openucx
/
xccl
Other
22
stars
14
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
dummy pr for debug
#85
vspetrov
closed
3 years ago
0
TEAM_LIB: vmc -> hmc
#84
vspetrov
closed
3 years ago
0
TEAM_LIB/UCX: set estimated num ppn for ucp context
#83
vspetrov
closed
3 years ago
0
Shared PD
#82
lappazos
closed
3 years ago
2
BUILD: fix cuda ldflags
#81
Sergei-Lebedev
closed
3 years ago
0
TEAM/UCX: pairwise alltoall with barrier
#80
Sergei-Lebedev
closed
3 years ago
0
TEAM/MHBA: Add team mkey
#79
lappazos
closed
3 years ago
0
CORE: pass config list to table register
#78
Sergei-Lebedev
closed
4 years ago
0
TEAM/MRAIL: disable by default
#77
Sergei-Lebedev
closed
4 years ago
0
team/mhba: add umr and mkey api
#76
lappazos
closed
4 years ago
0
TEAM/NCCL: add allgather
#75
Sergei-Lebedev
closed
4 years ago
0
team_lib/mhba: adds ucx team creation and barrier
#74
vspetrov
closed
4 years ago
0
Fix nonbound bcast in heir team
#73
Sergei-Lebedev
closed
4 years ago
0
team_lib/mhba: RC QPs for ASRs subgroup
#72
vspetrov
closed
4 years ago
0
team_lib/mhba: schedule/tasks for the new A2A
#71
vspetrov
closed
4 years ago
0
core/progress: call progress queue even if no progress fn of tl_ctx
#70
vspetrov
closed
4 years ago
0
team_lib/mhba: node fanin/fanout using shm
#69
vspetrov
closed
4 years ago
0
team_lib/mhba: allocate shared memory storage
#68
vspetrov
closed
4 years ago
0
xccl/sbgp: sbgp oob collectives
#67
vspetrov
closed
4 years ago
0
Sbgp to common
#66
vspetrov
closed
4 years ago
0
TEAM/HIER: 2lvl alltoall
#65
Sergei-Lebedev
closed
4 years ago
0
Feature request: ability to set priority for CUDA streams
#64
srinivas212
closed
3 years ago
2
Team/hier common task progress/complete handling
#63
vspetrov
closed
4 years ago
0
TEAM/NCCL: add alltoall
#62
Sergei-Lebedev
closed
4 years ago
0
XCCL: fix ucx configure warning and add min required version check
#61
Sergei-Lebedev
closed
4 years ago
0
TEAM/NCCL: add nccl team library
#60
Sergei-Lebedev
closed
4 years ago
0
CI debug print [DON'T MERGE]
#59
vspetrov
closed
4 years ago
0
Hier schedule fix no socket
#58
vspetrov
closed
4 years ago
3
Poor performance with NVLink
#57
froody
opened
4 years ago
2
Segfault creating process group with 1 member
#56
froody
closed
4 years ago
3
fixes to build with ucx + cuda
#55
froody
closed
1 year ago
1
TEAM/UCX: set ucp worker thread mode according to context thread mode
#54
Sergei-Lebedev
closed
4 years ago
0
mem component cache
#53
Sergei-Lebedev
closed
4 years ago
0
remove torch ucc
#52
Sergei-Lebedev
closed
4 years ago
0
Centralized progress per task
#51
lappazos
closed
4 years ago
0
TEAM/UCX: use multireduce in sra
#50
Sergei-Lebedev
closed
4 years ago
0
DO NOT MERGE: debug alltoallv
#49
Sergei-Lebedev
closed
4 years ago
0
TEAM/UCX: knomial sra allreduce
#48
Sergei-Lebedev
closed
4 years ago
1
TORCH: add alltoall benchmark
#47
Sergei-Lebedev
closed
4 years ago
0
XCCL: fix build with nonstandard cuda path
#46
Sergei-Lebedev
closed
4 years ago
0
TEAM/UCX: add ring allgather
#45
Sergei-Lebedev
closed
4 years ago
0
TORCH: add alltoallv support
#44
Sergei-Lebedev
closed
4 years ago
0
TEAM/UCX: add pairwise alltoallv
#43
Sergei-Lebedev
closed
4 years ago
0
TORCH: add torch ddp plugin
#42
Sergei-Lebedev
closed
4 years ago
0
TEAM/UCX: inplace alltoall
#41
Sergei-Lebedev
closed
4 years ago
0
ompi_coll_xccl patch update
#40
vspetrov
closed
4 years ago
0
team/ucx: alltoall pairwise exchange
#39
vspetrov
closed
4 years ago
0
ompi/coll/xccl: adds XCCL_REDUCE
#38
vspetrov
closed
4 years ago
0
xccl.h: context query
#37
vspetrov
closed
4 years ago
0
Team ucx reduce switch to multi
#36
vspetrov
closed
4 years ago
1
Previous
Next