Code

CUDA support#35

Code

Draft

vchuravy wants to merge 16 commits intovc/legacyJuliaParallel/UCX.jl:vc/legacyfrom vc/cudaJuliaParallel/UCX.jl:vc/cudaCopy head branch name to clipboard

+1,311-38

Commits

Commits on Mar 4, 2021

• Add initial support for Distributed.jl rpc over UCX AM

  vchuravy

  committed
  987298f

  Copy full SHA for 987298f

• fix UCX benchmarks

  vchuravy

  committed
  e61e3e3

  Copy full SHA for e61e3e3

• Fix benchmark plotting

  vchuravy

  committed
  a26e5ec

  Copy full SHA for a26e5ec

• synchronous am handler for small messages

  vchuravy

  committed
  c41341f

  Copy full SHA for c41341f

• short-circuit self

  vchuravy

  committed
  64c41a3

  Copy full SHA for 64c41a3

• refactore deserialization_msg

  vchuravy

  committed
  de644c3

  Copy full SHA for de644c3

• add send_arg

  vchuravy

  committed
  578ef72

  Copy full SHA for 578ef72

• send arguments early and without serialization if possible

  vchuravy

  committed
  387ebc5

  Copy full SHA for 387ebc5

• improve worker wait

  vchuravy

  committed
  a01a6dc

  Copy full SHA for a01a6dc

• fast view

  vchuravy

  committed
  64b9525

  Copy full SHA for 64b9525

• improve plotting

  vchuravy

  committed
  ce692a6

  Copy full SHA for ce692a6

• add distributed.jl to tested examples

  vchuravy

  committed
  f255fe1

  Copy full SHA for f255fe1

• don't leak memory

  vchuravy

  committed
  4b72a7d

  Copy full SHA for 4b72a7d

• protoypical cuda support

  vchuravy

  committed
  79114f5

  Copy full SHA for 79114f5

• add latency measurments from satori

  vchuravy

  committed
  9d1d589

  Copy full SHA for 9d1d589

• prettier plots

  vchuravy

  committed
  977e68a

  Copy full SHA for 977e68a