AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

2.18.1-1

Browse Source 
Add support for IB SHARP to NVLS (NVLink SHARP algorithm).
Add NVLS+Tree algorithm.
Add support for memory management using cuMem* functions.
Use all NICs for Send/Receive operations on systems with more than
one NIC per GPU (#804).
Add ncclCommSplit primitive, with resource sharing option in config.
Fix alltoallv hang (#788)
Increase number of channels on H100 when we're not limited by NVLink.
Improve error reporting in case of IB failure, printing local and
remote ID (#779).
Add build option to allow compilation against RDMA includes instead
of dynamically loading IB verbs symbols (#802).
Fix context creation for progress thread (#803).
NET/IB: add option to use multiple QPs in round-robin mode.
Fix tree performance issue when NVB is disabled on HCM topologies.


[ROCm/rccl commit: d97a32fac8]
This commit is contained in:
Sylvain Jeaugey
2023-04-03 05:32:07 -07:00
parent 19e361cc45
commit 902ff02645
64 changed files with 4758 additions and 3131 deletions
Download Patch File Download Diff File Expand all files Collapse all files
projects/rccl/src/Makefile
+1 -1
View File
@@ -10,7 +10,7 @@ include ../makefiles/version.mk
##### src files
INCEXPORTS := nccl.h nccl_net.h
LIBSRCFILES := init.cc init_nvtx.cc channel.cc bootstrap.cc transport.cc enqueue.cc group.cc debug.cc proxy.cc net.cc \
misc/cudawrap.cc misc/nvmlwrap.cc misc/ibvwrap.cc misc/gdrwrap.cc \
misc/cudawrap.cc misc/nvmlwrap.cc misc/ibvsymbols.cc misc/ibvwrap.cc misc/gdrwrap.cc \
misc/utils.cc misc/argcheck.cc misc/socket.cc misc/shmutils.cc misc/profiler.cc misc/param.cc misc/strongstream.cc \
misc/ipcsocket.cc \
transport/p2p.cc transport/shm.cc transport/net.cc transport/net_socket.cc transport/net_ib.cc transport/coll_net.cc transport/nvls.cc \