Network user buffer support for collectives
* Leverage user buffer registration to achieve zero-copy
inter-node communications for Ring, NVLS and Collnet
Add RAS subsystem
* Create a RAS thread keeping track of all NCCL communicators.
* Add a ncclras tool contacting the RAS thread and getting a
report.
Add fp8 support
* Add support for e5m2 and e4m3 8-bit floating point operations.
* Use Tree/PAT algorithms when possible for better numerical
stability.
Add NIC fusion
* Add a NET API to ask the network plugin to fuse a set of
interfaces together.
* Fuse multiple NICs under the same PCI switch as a single,
larger NIC.
Socket connection failure retry
* Retry in case of socket connection failure (unreachable host)
* Avoid "Software caused connection abort" errors on retries
QP connection failure retry
* Retry in case of IB QP connection failure during ibv_modify_qp.
NET API improvements
* Allow plugins to force a flush in case data and completion
ordering is not guaranteed.
* Indicate when completion is not needed (e.g. for the LL128
protocol), allowing plugins to skip generating a completion.
* Allow for full offload of allgather operations when using one
GPU per node.
NCCL_ALGO/NCCL_PROTO strict enforcement
* Extend NCCL_ALGO/NCCL_PROTO syntax to be able to specify
ALGO/PROTO filters for each collective operation.
* Strictly enforce the ALGO/PROTO filters, no longer fall back
on the ring algorithm when the filtering leaves no option and
error out instead.
Enable CUMEM host allocations
* Use cumem functions for host memory allocation by default.
Improved profiler plugin API
* Avoid dependencies with NCCL includes.
* Add information on whether the buffer is registered or not
Adjust PAT tuning
* Improve transition between PAT and ring at scale.
Fix hangs when running with different CPU architectures
* Detect when we use a mix of GPU architectures
* Ensure Algo/Proto decisions are made based on that unified
state.
Fix FD leak in UDS
* Fix a leak when mapping buffers intra-node with cumem IPCs.
Fix crash when mixing buffer registration and graph buffer registration.
* Separate local and graph registration to avoid crashes when we free
buffers.
Fix user buffer registration with dmabuf
* Make ncclSend/ncclRecv communication with buffer registration functional
on network plugins relying on dmabuf for buffer registration.
Fix crash in IB code caused by uninitialized fields.
Fix non-blocking ncclSend/ncclRecv
* Fix case where ncclSend/ncclRecv would return ncclSuccess in non-blocking
mode even though the operation was not enqueued onto the stream.
* Issue #1495
Various compiler tweaks and fixes
* PR #758
Fix typo in ncclTopoPrintGraph
* Issue #1468
[ROCm/rccl commit: 6aae379278]
It seems like here wants to check xgmi_node instead. If checks node for "nvlink", it will verify the link_info everytime.
If checks node for "xgmi", when get yes answer, it won't need check vsmi topo interface.
[ROCm/rccl commit: f2ee8d9132]
* Switched calls to `cudaMemcpyAsync` to be `cudaMemcpy` in `ncclTransportP2pSetup` to avoid race condition with `cudaIpcOpenMemHandle` inside p2p `connect`. See `ncclP2pImportShareableBuffer`.
* Moved synchronize outside of the loop, as it isn't necessary to sync between every iteration of the loop.
[ROCm/rccl commit: c158d3a9b4]
* Initializing all ranks to the same value to avoid failure of UT AllReduce for FP8 type
Co-authored-by: Marzieh Berenjkoub <mberenjk@amd.com>
[ROCm/rccl commit: 39483c55f8]
Certain CMake functions deduplicates arguments by default. For example, if we
have two `target_link_options` with both `-Xoffload-linker -opt-A` and then
`-Xoffload-linker -opt-B`, the final link command would be `-Xoffload-linker
-opt-A -opt-B`, which is not what we want.
[ROCm/rccl commit: 7386fac64a]
* Add RCCL debugging guide
* Changes from external review
* More edits from internal review
* Additional edits
* Minor correction
* More changes after external review
* Integrate index and ToC changes with incoming merge changes
* Integrate feedback from management review
* Minor edits from the internal review
[ROCm/rccl commit: 6d34fb7632]
* Add Topologies for 16-GPU gfx942 SuperNode
- Add GigaIO topologies to tools/topo_expl for dev and testing
- Add GigaIO Columba 16 GPU romeModel and adjust topology
matching algorithm in rome_models for 16 GPU system
- Fix bug which failed to match Rome Model when using subsets
of system resources (i.e. ROCR_VISIBLE_DEVICES is set)
- Fixes for topo_expl
* Fix bug w/ 1H16P
[ROCm/rccl commit: a05329bd0d]
* Refactor RCCL install guide into several pages
* Changes from code review and new docker guide
* Add missing entries to ToC
* Minor fixes
* Fix help strings
* Edits after review and remove extra white space
[ROCm/rccl commit: bf7c130631]
* mapping devices wrt pci
* Gpu allocation by using pci mapping
* Passing gpuPriorityOrder in as an argument rather than making the functions non-static.
* Removing redundant testBed instance calling
[ROCm/rccl commit: 69b2b712ab]
* Changing C-strings to be const.
* Changed variable-length arrays to std::vector to avoid warnings. VLA is a compiler extension.
* Changed `#define` inside functions into `constexpr int` to preserve scoping and avoid macro redefinition warnings.
* Disabled warnings for modifying `CMAKE_CXX_FLAGS` caused by `check_symbol_exists`, which temporarily modifies the flag to do a compile check.
* Fixed VLA in rccl UT.
[ROCm/rccl commit: 1c45962273]
To get the improved performance for TP=4, the user needs to use
RCCL_MSCCL_FORCE_ENABLE=1 and MSCCLPP_READ_ALLRED=1. For TP=8, the
user should use MSCCLPP_HIERARCHICAL_ALLRED=1.
[ROCm/rccl commit: 0fb3b5eba9]
Nilesh M Negi
Luna
Jeffrey Novotny
JhaShweta1
Sylvain Jeaugey
qiwei_ji
Xeonacid
dependabot[bot]
David Addison
WoosungMyung
corey-derochie-amd
Mustafa Abduljabbar
mberenjk
Ziyue Yang
akolliasAMD
Shilei Tian
Hujingbo
Nusrat Islam
Benjamin Kitor
gilbertlee-amd
Bertan Dogancay
darren-amd
saurabhAMD
Abhishek Kulkarni
Avinash
Tim