Minimize the performance impact of the device kernel profiling support when
the profiler plugin is not loaded.
Reduce the overheads of CUDA graph capturing, which increased in NCCL
2.26.2 for large graphs.
Fix the exchange of enhanced connection establishment (ECE) options to
address potential slowdowns on networks utilizing RoCE.
Test if cuMem host allocations work and if not, disable them. Enabled by
default since NCCL 2.24 if the CUDA driver version is at least 12.6, such
allocations rely on NUMA support, which is by default not available under
Docker. We recommend invoking Docker with "--cap-add SYS_NICE" to enable
it.
Fix an initialization error when running with NCCL_NET_GDR_C2C=1 on
multiple MNNVL domains with non-uniform network configurations across
nodes.
Fix the printing of sub-seconds in the debug log when using a custom
NCCL_DEBUG_TIMESTAMP_FORMAT setting.
[ROCm/rccl commit: 0524aef7a0]
w/o the header, experience following errors:
```
error: use of undeclared identifier 'RTLD_NOW'
error: use of undeclared identifier 'RTLD_LOCAL'
error: use of undeclared identifier 'dlerror'
error: use of undeclared identifier 'dlsym'
error: use of undeclared identifier 'dlclose'
```
[ROCm/rccl commit: 1786c0268b]
* Enabling LL128 by default on MI300
* Add missing CUDACHECK
* Adjust BW correction factors to fix the Tree->Ring switching point
* Refactor and add ll128 AR logarithmic factor to tuning models
* Move RCCL tuning changes to a separate file
* Use enum for tunable indexing
* Use explicit indexing in tuning models to avoid mismatch issues
* Place rcclGetSizePerRank in a function
* Remove HIP ifdef for rccl-only call
---------
Co-authored-by: Mustafa Abduljabbar <mustafa.abduljabbar@amd.com>
[ROCm/rccl commit: e40ff4f84a]
* [SRC] Enable unroll=1 for gfx950
* Fix typo from rebase in generate.py
* Support for unroll=1 and gfx90a when building for all GPU targets
---------
Signed-off-by: nileshnegi <Nilesh.Negi@amd.com>
[ROCm/rccl commit: 307bc10781]
* Add fault injection of starting warps with random variations
This is done by inserting randomly delays after __syncthreads().
The feature can be turned off by FAULT_INJECTION=OFF in cmake.
* Remove manually introduced bug for demo purpose
* Use only one thread per warp for checking wall clock
[ROCm/rccl commit: 90ad586d94]
* removed gfx940 and gfx941
* removed gfx940 and gfx941
* Update "gfx94" to "gfx942" in init.cc
* Updated remaining "gfx94" updates to "gfx942"
* Update filenames and variables from gfx940 to gfx942
---------
Co-authored-by: akolliasAMD <akollias@amd.com>
[ROCm/rccl commit: 6505639cf4]
Profiler improvements
* Add events for CUDA kernel start and end.
* Allow network plugins to generate profiling events
* Enable profiling on a per-operation basis, rather than per-communicator.
* Add support for graph capturing.
Add implicit launch order
* Allow to prevent deadlocks when using multiple NCCL communicators per
device by implicitly ordering NCCL operations using the host program
order. Disabled by default, set NCCL_LAUNCH_ORDER_IMPLICIT=1 to enable.
* Add a complementary mechanism to detect host threads racing to launch
to the same device. Enabled by default, set NCCL_LAUNCH_RACE_FATAL=0 to
disable.
Optimize the PAT algorithm
* Separate the computation and execution of PAT steps on different warps,
allowing to run up to 16 PAT steps in parallel to significantly
accelerate PAT and reduce its linear part.
Add support for setting QoS per communicator
* Add a new trafficClass field to the communicator configuration, to
allow the application to select a particular traffic class for a
given communicator. The meaning of the traffic class is
network-specific and should be set in accordance with the network
configuration.
* For the IB/RoCE plugin, existing config variables such as NCCL_IB_SL
and NCCL_IB_TC take precedence.
Allow to enable GPU Direct RDMA specifically on C2C platforms
* Disabled by default, set NCCL_NET_GDR_C2C=1 to enable.
Do not disable user buffer registration unless PXN is really used
* Only disable UB when a communicator has more than one rank per
node on any node.
RAS subsystem improvements
* Report operation counts separately for each collective operation type.
* Provide details about missing communicator ranks and reliably
distinguish ranks that are no longer a given communicator's members
(now reported as NOCOMM) from those that failed to respond.
Add support for timestamps to NCCL diagnostic messages
* On by default for WARN messages; NCCL_DEBUG_TIMESTAMP_LEVELS can be
used to enable them for other debug levels as well.
* The format can be changed using the NCCL_DEBUG_TIMESTAMP_FORMAT config
variable.
Reduce the memory usage with NVLink SHARP (NVLS)
* Potentially save hundreds of MBs of device memory, considering the
multicast buffer size granularity separately from the address alignment.
Update performance tuning for recent Intel CPUs
* Improve algorithm/protocol selection on recent CPUs such as Emerald
Rapids and Sapphire Rapids.
Improve channel scheduling when mixing LL and Simple operations.
* Make LL operations account for 4x more traffic to ensure LL and simple
operations complete at the same time.
Refactor the plugin code
* Clean up and harmonize the support code across the network, tuner,
and profiler plugins.
Add support for comment lines (starting with #) in the nccl.conf file
* Issue #1540.
Make user buffer registration problems print an INFO instead of a WARN.
Drop support for network plugin interface version 5.
Fix a race condition with split-shared communicators
* NCCL could hang during connection setup if multiple communicators
were grouped together that share resources.
Fix a performance regression when using NCCL_CROSS_NIC=1
* NCCL would unnecessarily alternate rings, breaking the GPU-NIC
associations.
Make GID index detection code more resilient
* Dynamic GID detection code was giving up too soon if the
detected index was not available (e.g., wasn't mapped to the
container's sysfs).
* Issues #1538, #1573.
Fix a race condition with non-blocking operation
* Fix issue when creating a non-blocking communicator after a non-
blocking collective operation on another communicator.
Fix shared memory usage on recent Blackwell GPUs.
* Issues NVIDIA/nccl-tests#287, NVIDIA/nccl-tests#291, #1637.
Fix an error with NIC fusion and IB SHARP when recreating communicators
* Disable the unloading of network plugins
Make the auto-merge failures in the NIC fusion non-fatal
* This could happen when trying to merge IB and RoCE devices.
Fixes to ncclCommAbort
* Fix hangs due to the progress thread spinning indefinitely on the
network progress.
* Reduce the abort time by up to two orders of magnitude.
Fix a crash when libnccl.so was dynamically unloaded
* The RAS subsystem was missing a clean-up handler.
Fix a hang if the network plugin's test() call returns an error.
Fix a hang on heterogeneous architectures
* Ensure we harmonize the tuning to avoid different tuning choices,
causing a hang.
Fix double-free on failed ncclCommInitRank and ncclCommFinalize.
Fix a potential list traversal bug during a group launch of multiple
communicators
* Issue #1599.
Unify the handling of NCCL configuration variables
* Under rare circumstances, some variables specified in the config file
could be ignored.
[ROCm/rccl commit: f44ac759fe]
* Limit P2P channels per peer to not exceeding max channels
* [UT] test single GPU cases for all collectives
* [UT] fix out of range root value
[ROCm/rccl commit: 4237caad69]
* misc/msccl: read graphCaptureStatus for every collective call
* fix a bug in checking whether UBR is enabled in MSCCLPP
* cmake: Fix patch reversal order
* misc/msccl: add logging
[ROCm/rccl commit: 23c0b7bd84]
* Improve RDMA flushing by write dummy payload with RO=0
* Rename env var for disabling this change to RCCL_GDR_FLUSH_GPU_MEM_NO_RELAXED_ORDERING
[ROCm/rccl commit: 60c1264d27]
* Enable GDRCopy only on gfx94x
* Use cudaFree instead of hipFree
* Add warning if failed to get device property
* Remove extra return
[ROCm/rccl commit: 32dc7ef47c]
* Print KL/CL/KE events for all warps
* Fix count off-by-one issue
* Fix opCount in KE and restore CPU thread option
* Simplify count calculation
[ROCm/rccl commit: ebf7e2305e]
* erase handle key from mscclpp communicator during deregistration
* remove check on buffer size being a multiple of 32 from registration/deregistration routines since these checks are applied during enqueue
* add check for greater than zero buffer size in mscclpp registration
[ROCm/rccl commit: 19105206f6]
Kamil Iskra
Bertan Dogancay
Tim
Mustafa Abduljabbar
Nusrat Islam
AbandiGa
Dingming Wu
Arm Patinyasakdikul
Pedram Alizadeh
Nilesh M Negi
BertanDogancay
isaki001
gilbertlee-amd
Wenkai Du
corey-derochie-amd
Avinash
akolliasAMD
mberenjk
Vijay Srinivasan