We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
There was an error while loading. Please reload this page.
NCCL_PROTO=^LL128
1 parent 720fd50 commit 92689ceCopy full SHA for 92689ce
docs/software/communication/nccl_env_vars
@@ -10,6 +10,9 @@ export NCCL_NET="AWS Libfabric"
10
# https://docs.nvidia.com/deeplearning/nccl/user-guide/docs/env.html#nccl-net-gdr-level-formerly-nccl-ib-gdr-level
11
export NCCL_NET_GDR_LEVEL=PHB
12
export NCCL_CROSS_NIC=1
13
+# Starting with nccl 2.27 a new protocol (LL128) was enabled by default, which
14
+# typically performs worse on Slingshot. The following disables that protocol.
15
+export NCCL_PROTO=^LL128
16
# These `FI` (libfabric) environment variables have been found to give the best
17
# performance on the Alps network across a wide range of applications. Specific
18
# applications may perform better with other values.
0 commit comments