* 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>
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.
- Introduced double-buffering to reduce copy overhead and overlap BF16 arithmetic with data prefetching.
- Aimed to improve performance of reduction-based collectives by up to 10%.
- Implemented based on recommendations from Guennadi Riguer (AMD)
- Added --force-reduce-pipeline option to install.sh to activate this optimization for BF16 reductions.
- Feature is disabled by default to prevent regressions with large messages until auto-tuning logic is upstreamed.
---------
Co-authored-by: Jeffrey Novotny <jnovotny@amd.com>
Co-authored-by: Pedram Alizadeh <pmohamma@amd.com>
* Support fused all reduce and elementwise operations
Add additional "acc" parameter to RCCL Replayer logs
Add flag which indicates availability of new API
* Fix Recorder json parsing
* Remove unreachable code
* Remove extra acc pointer check
* .
* Revert "[DEVICE] Adding ability to choose unroll factor at runtime (#1734)"
This reverts commit 9d72be7b2f.
* Use noinline to reduce kernels linking time
* Don't use noinline for gfx942 and gfx950 to avoid perf regression
---------
Co-authored-by: AtlantaPepsi <timhu102@amd.com>
Co-authored-by: BertanDogancay <bertan.dogancay@gmail.com>
Boosts single node bfloat16 allreduce performance by up to 20% for some data sizes and provides gating with the RCCL_GFX942_CHEAP_FENCE_OFF environment variable
* Increased max stack size to 640
* Added new binary for executing unit tests
Added new unit tests for argcheck.cc and alt_rsmi.cc files
Modified the method to execute unit tests to cover static methods
by using a bash script to convert static to non-static functions
and variables on the fly restricted to debug build type.
* Add support for extended fine-grained system memory pool
* Use hipHostRegisterUncached
* Add "sc0 sc1" flags for LL store on gfx950
* Update after HIP flag is changed to hipExtHostRegisterUncached
This feature tracks the proxy events and status of each send/recv op. ProxyTrace keeps a fixed number of active ops in host mem and dumps the status of each op when the program crashes or hangs.
* Make offload-compress the default
* Add guard for --offload-compress since it was introduced in ROCm 6.2
* Address some of Nilesh's feedback.
* Reorganize for code cleanliness
* Improve comment
* Compress gpu code at link and compile time
* 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
* 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>
* 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
COMPILING_TARGETS is not actually used for --offload-arch option,
instead GPU_TARGETS is being used implicitly when we call
find_package(hip REQUIRED) (See hip-config-amd.cmake).
* Accept an EXPLICIT_ROCM_VERSION and use that vs inspecting the environment if provided.
* Use CMake's built in file reading support vs execute_process (without error checking) to avoid silent but deadly later failures.
* Properly quote some comparisons to avoid syntax errors if they happen to have an empty string.
* Guard against ROCM_PATH being an empty string, avoiding stray path extensions to root directories, etc.
Co-authored-by: Stella Laurenzo <stellaraccident@gmail.com>
Arm Patinyasakdikul
alex-breslow-amd
Bertan Dogancay
BertanDogancay
Avinash
jonatluu
Mustafa Abduljabbar
Rahul Vaidya
awelling2801
ycui1984
Atul Kulkarni
Wenkai Du
isaki001
Nilesh M Negi
Dingming Wu
Richard Barnes
Atul Kulkarni
Tim
Pedram Alizadeh
gilbertlee-amd
Sohaib Nadeem
rahulc1984