Current number of transmissions parsed from MSCCL algorithm is 1-based value,
but when calculating proxy number of chunks, it's taken as 0-based value.
This commit fixes this issue.
With Simple protocol, unroll factor of 4 offers better
performance for most of the collectives (on MI200. MI250, and
MI300) except large message allreduce with Ring algorithm
on MI250 and MI200). This PR changes the default unroll factor
to 4 while adding fine tuning for reduction operations.
Fix data corruption with Tree/LL128 on systems with 1GPU:1NIC.
Fix hang with Collnet on bfloat16 on systems with less than one NIC
per GPU.
Fix long initialization time.
Fix data corruption with Collnet when mixing multi-process and
multi-GPU per process.
Fix crash when shared memory creation fails.
Fix Avg operation with Collnet/Chain.
Fix performance of alltoall at scale with more than one NIC per GPU.
Fix performance for DGX H800.
Fix race condition in connection progress causing a crash.
Fix network flush with Collnet.
Fix performance of aggregated allGather/reduceScatter operations.
Fix PXN operation when CUDA_VISIBLE_DEVICES is set.
Fix NVTX3 compilation issues on Debian 10.
* Fix build on additional architectures
Instead of directly wrapping a platform-specific operation with a
preprocessor check against a gfx macro, it can be more flexible to
check a macro that can be overriden by the user. The gfx macro can then
just provide the default value for the macro, resulting in the same
default behaviour as if the gfx macro was checked directly but with
more control at build-time.
For example, to build rccl without using buffer_wbinvl1_vol on
gfx902, but still use the default on other archs, a user could
export CXXFLAGS='-Xarch_gfx902 -DRCCL_USE_WBINVL1_VOL=1' before
configuring the build. This flexibility isn't always necessary, but
it's nicer to have it and not need it than to need it and not have it.
* Define WARP_SIZE using warpSize builtin
Added a RCCL_MSCCL_ENABLE_DONE_EVENT env var, set it be 0 by default.
The env var is to control whether to use doneEvent when invoking MSCCL
kernels.
Skipping doneEvent would cause the firmware to skip L2 cache flush,
resulting in overall performance improvement.
* Revert "Ensure memory copy integrity during transport setup (#731)"
This reverts commit 36e453c61e.
Add stream synchronization in ncclStrongStreamRelease.
* Use event record and wait
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.
gilbertlee-amd
Wenkai Du
Ziyue Yang
Nusrat Islam
Bertan Dogancay
Wenkai Du
akolliasAMD
Dmitrii Gabor
arvindcheru
Sylvain Jeaugey
Cory Bloor
Wen-Heng (Jack) Chung
Wen-Heng (Jack) Chung
Kaiming Ouyang
Ziyue Yang