* Revert disabling of context tracking for Radeon
Original commit df3b7e47
`Disable context tracking for the current version. (#1839)`
* Add env variable for disabling of context tracking for Radeon
`export NCCL_DISABLE_CONTEXT_TRACKING=1` to force disable of context tracking
* Update docs/how-to/rccl-usage-tips.rst
Fix grammar, thanks @amd-jnovotny
Co-authored-by: Jeffrey Novotny <jnovotny@amd.com>
* Rename NCCL_DISABLE_CONTEXT_TRACKING -> RCCL_DISABLE_CONTEXT_TRACKING
* Revert changes in includes and rename util function
---------
Co-authored-by: Jeffrey Novotny <jnovotny@amd.com>
[ROCm/rccl commit: 07925ec027]
* Add tests on slurm cluster
* Integrate slurm.
* Add flags.
* Added dynamic selection of runners for tests and cleanup for slurm reservation
* Revert "Added dynamic selection of runners for tests and cleanup for slurm reservation"
This reverts commit fdd5a6cc968c764d3d1039f0897fb11f11422928.
* Refactor so tests run on both architectures.
* continue on error
* fail fast false on matrix
* remove scancel
* skip all single node tests
* fix pattern matching for pytest
* switch to always skip github job
* Update to latest allocation.
* Clean up workflows and update docker image.
* Updated container image published from PR #1517
* Switch back to TheRock main branch sha.
---------
Co-authored-by: arravikum <arravikum@amd.com>
[ROCm/rccl commit: 01d16d4139]
* Moved new functions to the bottom of the function table to maintain backward compatibility
* Added ordering fixes to api_trace.cc
[ROCm/rccl commit: d86cf78810]
* Batch P2P operations (2 per CU/channel) and update channel-part mapping
- Revert bitreversal and fix channel mapping to be compatible with P2P batching and avoid hangs
- P2P batching is only used for more than 2 nodes to avoid aggregating intra-node traffic when it is dominant for less than 2 nodes
* Address single node regression and channel per net peer
* Add batching threshold
* Add enable switch for batching
* Update CHANGELOG.md
* Add minor comment change
* Update CHANGELOG.md
Co-authored-by: Jeffrey Novotny <jnovotny@amd.com>
* Update CHANGELOG.md
Co-authored-by: Jeffrey Novotny <jnovotny@amd.com>
[ROCm/rccl commit: c1e1f2faeb]
Minimum required cmake version of test/CMakeList.txt is bumped from 2.8
to 3.16. This alignes with the version used in CMakeList.txt and will
enable building with cmake 4.
[ROCm/rccl commit: 0f6fec1553]
* [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
[ROCm/rccl commit: 361d596229]
* Use ROCm 6.4.1 for testing
* Extend RCCL-Tests to multi-node
* Add HSA_NO_SCRATCH_RECLAIM to UT runs
* Limit to single-node rccl-tests for now
[ROCm/rccl commit: d73cee7588]
* 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>
[ROCm/rccl commit: 277747c199]