* Fix build on additional architectures
Instead of directly wrapping a platform-specific operation with a
preprocessor check against a gfx macro, it can be more flexible to
check a macro that can be overriden by the user. The gfx macro can then
just provide the default value for the macro, resulting in the same
default behaviour as if the gfx macro was checked directly but with
more control at build-time.
For example, to build rccl without using buffer_wbinvl1_vol on
gfx902, but still use the default on other archs, a user could
export CXXFLAGS='-Xarch_gfx902 -DRCCL_USE_WBINVL1_VOL=1' before
configuring the build. This flexibility isn't always necessary, but
it's nicer to have it and not need it than to need it and not have it.
* Define WARP_SIZE using warpSize builtin
[ROCm/rccl commit: b1a65afd58]
* add DMA_BUF support
* remove unused libraries in src/init.cc
* change NCCL_ALL to NCCL_INIT
* remove extra pointer functions in transport/net.cc
[ROCm/rccl commit: d52b6c0d24]
The default value of unroll factor is 2. Changing the unroll
factor to 4 provides better performance for most of the collectives.
[ROCm/rccl commit: 4d1cfb17c8]
Added a RCCL_MSCCL_ENABLE_DONE_EVENT env var, set it be 0 by default.
The env var is to control whether to use doneEvent when invoking MSCCL
kernels.
Skipping doneEvent would cause the firmware to skip L2 cache flush,
resulting in overall performance improvement.
[ROCm/rccl commit: 12dba425de]
* Revert "Ensure memory copy integrity during transport setup (#731)"
This reverts commit 0c0c35927a.
Add stream synchronization in ncclStrongStreamRelease.
* Use event record and wait
[ROCm/rccl commit: 4ca7742c61]
Add support for IB SHARP to NVLS (NVLink SHARP algorithm).
Add NVLS+Tree algorithm.
Add support for memory management using cuMem* functions.
Use all NICs for Send/Receive operations on systems with more than
one NIC per GPU (#804).
Add ncclCommSplit primitive, with resource sharing option in config.
Fix alltoallv hang (#788)
Increase number of channels on H100 when we're not limited by NVLink.
Improve error reporting in case of IB failure, printing local and
remote ID (#779).
Add build option to allow compilation against RDMA includes instead
of dynamically loading IB verbs symbols (#802).
Fix context creation for progress thread (#803).
NET/IB: add option to use multiple QPs in round-robin mode.
Fix tree performance issue when NVB is disabled on HCM topologies.
[ROCm/rccl commit: d97a32fac8]
* Refactoring unit tests to improve performance
* Spawning child processes during InitComms instead of on TestBed construction
* Temporarily disabling graph unit tests
[ROCm/rccl commit: 27e0cb43c2]
* remove ref to old script in static analysis
* update README with doc build instructions
update gitignore with doc artifacts
[ROCm/rccl commit: 8c56e6c892]
* Fix XGMI detection
* Increase stack size
* Temporarily disable signal hangler in CI
[Process: 17281] Inside handler function signal: Segmentation fault (11)
BFD: DWARF error: section .debug_info is larger than its filesize! (0x93ef57 vs 0x530ea0)
BFD: DWARF error: section .debug_info is larger than its filesize! (0x93ef57 vs 0x530ea0)
[ROCm/rccl commit: 22b81fbaae]
dependabot[bot]
gilbertlee-amd
Cory Bloor
Pedram Alizadeh
Wenkai Du
Bertan Dogancay
Nusrat Islam
akolliasAMD
Ziyue Yang
Sam Wu
Wen-Heng (Jack) Chung
Wen-Heng (Jack) Chung
Wenkai Du
Sylvain Jeaugey
Saad Rahim
Kaiming Ouyang
Ziyue Yang