Move device lock before the execution instead of tensor gathering #3457
wconstab
approved these changes
on 2022-03-30
miladm
force pushed
from
bcd981fa
to
616dabdd
4 years ago
miladm
force pushed
from
d0a03a2f
to
2dd5ad47
4 years ago
miladm
force pushed
from
d47deeda
to
f4dcf8b5
4 years ago
miladm
force pushed
from
fbee1327
to
56b9be9f
4 years ago
Move device lock before the execution instead of tensor gathering
3482ff2e
Handle OpbyOP Lock
db6dd6ff
moving the barrier into RunPostOrder and making changes to coll.indic…
12065510
added a conditional barrier to runpostorder to reduce the frequency o…
f107a07f
moved TensorCollectionBarrier into TryRunCachedSync instead of callin…
f137d89c
moved the barrier call to ScheduleSyncTensorsGraph and optimized the …
df25112e
nit change
67409a9f
Empty-Commit
88774387
fixing ltc lazy api change
854c75c9
Empty-Commit
520a8dea
Added profiling support for RunPostOder. Added race condition caveat …
780f327f
added a missing device filter to skip calling barrier
0dd33568
linter fix
a2cec8cb
removed barrier_applied
930494a7
run test cleanup
e9860e38
cleaner condition
bf5be920
linter fix
f0e087d2
addressed feedbacks
592b01df
reverted tests
72593a37
updated toString API to new format
e4ac76ce
miladm
force pushed
from
56b9be9f
to
e4ac76ce
4 years ago
miladm
merged
5d1c4210
into master 4 years ago
miladm
deleted the move_sync_lock branch 4 years ago
Labels
enhancement
performance
Login to write a write a comment.
Login via GitHub