Device API (Experimental)
* Introduces device-side APIs to integrate NCCL communication directly into application kernels.
* Supports LSA (Load/Store Access) for CUDA P2P communication over NVLink and some PCIe platforms.
* Supports Multimem for hardware multicast using NVLink SHARP.
* Adds initial framework for GIN (GPU-Initiated Networking), currently under development.
* Introduces device communicators created using ncclDevCommCreate.
* Enables device-side communication operations with synchronization (ncclLsaBarrierSession) and memory accessors (ncclGetLsaPointer, ncclGetLsaMultimemPointer).
* Experimental APIs - signatures and functionality may evolve in future releases.
* No ABI compatibility is guaranteed — applications must be recompiled with each new NCCL release.
Symmetric memory improvements
* Support for aggregating symmetric operations using ncclGroupStart/End APIs.
* Reimplement symmetric kernels using device API.
New Host APIs
* Introduce new host collective APIs: ncclAlltoAll, ncclScatter, ncclGather.
CE (Copy Engine) Collectives
* Reduce SM utilization for alltoall, scatter, gather, and allgather within a single (MN)NVL domain.
* Free up SM capacity for the application to do computation at the same time.
* To enable the feature for ncclAllGather, ncclAlltoAll, ncclGather, ncclScatter, register buffers into symmetric windows and use the NCCL_CTA_POLICY_ZERO flag in the communicator config_t.
NCCL Inspector Plugin
* Introduces an Inspector plugin for always-on performance monitoring.
* Produces structured JSON output with metadata, execution time, bandwidth, and optional event traces for each NCCL operation.
* Enables integration with analysis tools such as Performance Exporter to visualize NCCL performance bottlenecks.
* Lightweight to enable via environment variables NCCL_PROFILER_PLUGIN and NCCL_INSPECTOR_ENABLE.
CMake support (Experiemental)
* Adds a CMake build system as an alternative to existing Makefiles.
* Known issues: pkg.build and Device API currently do not work with CMake.
* The known issues will be addressed in a future release.
Decreased max CTA count from 32 to 16 on Blackwell
* SM overhead is decreased by 50% with this improvement.
* This may cause some perf drop on Blackwell because of the reduced SM usage.
* If the extra SM capacity is not desired, two options are available to restore to previous behavior: 1) Setting NCCL_MIN_CTAS=32 NCCL_MAX_CTAS=32 environment variables; 2) setting communicator config to over-write max CTA count to 32.
* Based on community feedback, future versions may consider different trade-offs between performance and SM overhead.
Plugins
* Network
* App-aware Network plugin. NCCL passes information about communication operations to be executed on the network end point. This allows for better tuning of network end points and their use in the plugins.
* Improve handling of physical and virtual network devices and load/unload.
* Network plugin version 11 - add explicit context and communication ID support for per communicator init/finalize.
* Add Multi-Request Net API. Using this will help NCCL to anticipate multiple send/recv requests and optimize for it. See maxMultiRequestSize field in ncclNetProperties_v11_t.
* Profiler
* Add support for API events (group, collective, and p2p) and for tracking kernel launches in the profiler plugin.
* Add Inspector Profiler Plugin (see section above).
* Add a hook to Google’s CoMMA profiler on github.
* Tuner
* Expose NCCL tuning constants at tuner initialization via ncclTunerConstants_v5_t.
* Add NVL Domain Information API.
* Support multiple plugin types from a single shared object.
New Parameterization and ncclConfig changes:
* Add new option NCCL_MNNVL_CLIQUE_ID=-2 which will use rack serial number to partition the MNNVL clique. This will limit NVLink domains to GPUs within a single rack.
* Add NCCL_NETDEVS_POLICY to control how NET devices are assigned to GPUs. The default (AUTO) is the policy used in previous versions.
* Add NCCL_SINGLE_PROC_MEM_REG_ENABLE control variable to enable NVLS UB registration in the “one process, multiple ranks” case as opt in.
* Move nChannelsPerNetPeer into ncclConfig. NCCL_NCHANNELS_PER_NET_PEER can override the value in ncclConfig.
* Enable PxN over C2C by default
* PxN over C2C will improve performance for Grace-Blackwell platforms by allowing NCCL to leverage the NIC attached to a peer GPU over NVLINK, C2C, and PCIe.
* This behavior can be overridden by setting NCCL_PXN_C2C=0.
Other Improvements:
* Allow FP8 support for non-reductive operations on pre sm90 devices. (See https://github.com/pytorch/pytorch/pull/151594#discussion_r2135777776)
* Fix NVLS+CollNet and temporarily disables COLLNET_CHAIN for >8 GPUs.
* Only consider running interfaces for socket traffic. NCCL will not attempt to use interfaces that do not have the IFF_RUNNING bit. (https://github.com/NVIDIA/nccl/issues/1798)
* Modernize mutex management. Convert to std::mutex and std::lock_guard.
* Remove sm35 and sm50 GENCODE targets which have long been deprecated and were causing issues with the latest NCCL release builds.
* Improved NVLS/NVLSTree tuning prediction to improve algorithm and protocol selection.
* NVLSTree Tuning Fixes. Update tuning data for H100, GB200-NV72.
* Respond better to RoCE link flaps. Instead of reporting an “unknown event” it will now report “GID table changed”.
* Move libvirt bridge interface to the end of possible interfaces so that they are considered last. These interfaces are usually virtual bridges to relay traffic to containers running on the host and cannot be used for traffic to a remote node and are therefore unsuitable.
* [rocm_regression] Return errors when HSA_NO_SCRATCH_RECLAIM=1 even for rocm >= 6.4.0
* [rocm_regression] Check firmware version
* [rocm_regression] Resolve review comments
* [rocm_regression] Move hsa env checking into init once func
* [rocm_regression] Prevent hot fix version in firmware
* [rocm_regression] Improve unit tests
* Support pipelining codegen and template specialization
* Support ReduceCopy pipelining for AllReduce, ReduceScatter, and Reduce (currently enabled for bfloat16)
* Remove need for FUNC_INDEX_TOTAL
* Add pipeline field to device function key construction logic
* Avoid unneeded codegen for LL/LL64 kernels
* Modify conditions and add pipeline dtypes env
* Optimize selection for both gfx942 and gfx950
* Increase pipeline bitfield width
* Use __forceinline__ for all device functions
* Realign reduceCopy with original form
* Add opt-out option to enable perf debugs
* Remove force-reduce-pipelining option from README
* Update CHANGELOG.md
---------
Co-authored-by: Jeffrey Novotny <jnovotny@amd.com>
* Support gfx950 in topo_expl
* Fix dependencies and fetch fmt from sources
* Remove third_party folder in make clean
* Add empty target when fmt is found
* Add MI350 example
* Update README.md
---------
Co-authored-by: isaki001 <ioannissakiotis@gmail.com>
* add direct allgather algorithm
* minor fix
* add debug print for memory allocation tracker
* add message size threshold for direct allgather
* scatter transfers across ranks
* update changelog
* minor fix
* Update CHANGELOG.md
Co-authored-by: Jeffrey Novotny <jnovotny@amd.com>
* enable direct AG when pxn is ON on MI300X or MI350
---------
Co-authored-by: Jeffrey Novotny <jnovotny@amd.com>
Changed `TestBedChild` protocol to send the result code before the return value to avoid hanging if the call fails. Switched `TestBedChild::GetUniqueId` to use this.
* First commit for rccl multi node test workflow
* Adding workflow dispatch
* Added branch based pull trigger
* Changed typo in branch name
* Add input variables to push
* Removed input variables to push
* Added self hosted runner for Vultr cloud
* Skipping build and only running test
* Changed test runner label name
* Made changes to executable paths in test script
* Made changes to run
* Made changes to cd into cvs dir
* This is a dummy commit
* Added cmake options
* Modified build options
* Commiting build changes
* Adding rccl and rccl-tests
* Re-ordering rccl and rccl-tests
* adding --global command
* modified cmake command
* modified script paths
* Testing OIDC for rccl repo
* Testing OIDC for rccl repo
* Testing build and upload workflow
* use default env variable for AMDGPU families on push workflow trigger
* Adding cleanup and correct role
* Adding additional yml files
* Fixing typo';
* Adding new sha
* Adding correct gpu target
* Adding back venv bin activate
* Adding workflow dispatch for tests
* Testing
* Adding cat
* Adding cat
* Adding rocm dir change
* Adding checkout
* cat with sudo
* rccl checkout
* correcting branch
* removing sudo
* trying to adjust correct path'
* Adding output dir path
* Use docker container with pre-installed MPI
* Adding back build steps
* Fixing SHA
* Adding exclusion logic:
* Adding test
* Adding CI check
* Removing testing
* Limit to build only rccl, rccl-tests and required dependencies
* Adding test
* Removing test
* Removing quote
* Reverting test
* PR comments
---------
Co-authored-by: arravikum <arravikum@amd.com>
Co-authored-by: Marius Brehler <marius.brehler@amd.com>
- Updated ncclDevFuncId to use a hash-based lookup with std::unordered_map.
- Keys are now 64-bit integers, which pack coll, algo, proto, devRedOp, and type fields.
- Improved flexibility and maintainability by moving away from row-based indexing.
- Added error handling for missing keys in the hash map.
- Aligned key generation logic with generate.py and updated generate.py.
* File containing test for comm.h
* Update CommTest.cpp
Added gtest API for assert
* Update CommTest.cpp
Adding copyright
* Update CommTest.cpp
Removing info and tested as not required.
* Update and rename CommTest.cpp to CommTests.cpp
* Update CMakeLists.txt
* change gfx950 algo/proto selection for multinode allreduce, allgather, reduceScatter
* gfx950 tuning: enable tuning for broadcast, allreduce starts LL128 earlier and switches to ring earlier, change LL128 start for allgather and reduceScatter
* lower LL128 threshold
* update reduceScatter LL128 min to match LL max for consistency
* enable multinode PXN and increase chunksize for gfx950
* change LL128 start to 128KB, adjust ring-start according to node-count
* disable code-path for fused-AR on LL128 for gfx950
* use LL128 starting from 1KB for multinode allgather on gfx950
* start LL128 earlier for multinode reduceScatter on gfx950
* start LL128 earlier for multinode broadcast on gfx950
* set multinode allreduce to start simple on 64MB for gfx950
* start LL128 from 1KB for multinode broadcast on gfx950
* setting multinode AR to use tree instead of ring at 16MB, 64MB, 128MB
* set multinode broadcast to use LL for up to 256KB depending on node-count for gfx950
* adjust algo for 32MB multinode allreduce on gfx950
* make 32MB tree LL128 for multinode AR on gfx950
* make sure ring is not picked on 2N allreduce on small sizes
Leverages the traits of extended-scope fine-grain memory to get rid of a device-scope acquire-release fence. This improves throughput for single node workloads on gfx942 and gfx950 for some input sizes (e.g., ~32 MiB to about 256 MiB) when using the simple protocol. Multinode workloads on MI300X see a smaller but statistically significant uplift for some message sizes. Runtime disablement is supported via setting the environment variable RCCL_GFX942_CHEAP_FENCE_ON to 0.
* Added useAcc as a template parameter to address the 2% performance regression in allreduceWithBias
---------
Co-authored-by: Marzieh Berenjkoub <mberenjk@amd.com>
Weile
Karthik Ganesan
Kapil S. Pawar
Bertan Dogancay
ycui1984
Arm Patinyasakdikul
mberenjk
Wenkai Du
isaki001
Mustafa Abduljabbar
Mark Santesson
ycui1984
BertanDogancay
Avinash
Nilesh M Negi
jonatluu
Geo Min
Nusrat Islam
ishkool
Marius Brehler
Jeffrey Novotny
corey-derochie-amd
awelling2801
Atul Kulkarni
alex-breslow-amd
Adel Johar
Karthikeyan Arumugam