* mscclpp patch apply clip patch and set allreduce8 blocks from 512 to 1024
* add compilation flag for enabling/disabling clipping in mscclpp
* change flag name for consistency, set flag to OFF
* add compilation flag in rccl for enabling clipping in mscclpp
* set 1024 threads for mscclpp allreduce8 only for bfloat16
* fix improper description for ENABLE_MSCCLPP_CLIP flag
* Revert "Merge branch 'clip-patch' of https://github.com/isaki001/rccl into clip-patch"
This reverts commit 6e31857a9db98314b8a748eb024f2c3699ebe2d5, reversing
changes made to 193f4caa8ffa78b4e056893212fd8344aa14e937.
* update clip remove-clip.patch for rebase
[ROCm/rccl commit: 8145c4f3b8]
* Internal RCCL/NCCL functionality exposed when RCCL_EXPOSE_STATIC is enabled
* Algo/protocol/max channels can be obtained with the new RCCL API
* Introduce rccl_static and rccl_static_inline macros to work around invisible functions in core source files like enqueue.cc
* Add usage example in topo-explorer tool
[ROCm/rccl commit: 82afb2bcfe]
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]
* added rccl version using rccl-tests
* Added function to get rccl version from rccl-tests
* removed whitespace
* Added rccl version
* Updated readme and fixed formatting
* removed debug prints
[ROCm/rccl commit: 3dc0478722]
* Fix compiler issues due to broken compatibility
* Fix segfault and pass rank instead of busid and add a pointer to cover a new algorithm
[ROCm/rccl commit: aace4e27f8]
* [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]
Bertan Dogancay
Mustafa Abduljabbar
Siu Chi Chan
Nilesh M Negi
deeksha-amd
isaki001
BertanDogancay
Tim
Richard Barnes
Jeffrey Novotny
Kamil Iskra
gilbertlee-amd
Nusrat Islam
AbandiGa
Giuseppe Congiu
Dingming Wu
Arm Patinyasakdikul
Pedram Alizadeh
dependabot[bot]
Nikhil-Nunna
Istvan Kiss
Wenkai Du