Hendrik/xgboost
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
xgboost/src/collective
History
Jiaming Yuan 8bad677c2f
Update collective implementation. (#10152) 
* Update collective implementation.

- Cleanup resource during `Finalize` to avoid handling threads in destructor.
- Calculate the size for allgather automatically.
- Use simple allgather for small (smaller than the number of worker) allreduce.
2024-03-30 18:57:31 +08:00
..
aggregator.cuh
Remove internal use of gpu_id. (#9568)
2023-09-20 23:29:51 +08:00
aggregator.h
Require context in aggregators. (#10075)
2024-02-28 03:12:42 +08:00
allgather.cc
Update collective implementation. (#10152)
2024-03-30 18:57:31 +08:00
allgather.h
Update collective implementation. (#10152)
2024-03-30 18:57:31 +08:00
allreduce.cc
Update collective implementation. (#10152)
2024-03-30 18:57:31 +08:00
allreduce.h
Serialize expand entry for allgather. (#9702)
2023-10-24 14:33:28 +08:00
broadcast.cc
Cleanup code for distributed training. (#9805)
2023-11-25 09:10:56 +08:00
broadcast.h
[coll] Broadcast. (#9659)
2023-10-14 09:34:37 +08:00
coll.cc
Update collective implementation. (#10152)
2024-03-30 18:57:31 +08:00
coll.cu
Update collective implementation. (#10152)
2024-03-30 18:57:31 +08:00
coll.cuh
Update collective implementation. (#10152)
2024-03-30 18:57:31 +08:00
coll.h
Update collective implementation. (#10152)
2024-03-30 18:57:31 +08:00
comm_group.cc
Update collective implementation. (#10152)
2024-03-30 18:57:31 +08:00
comm_group.h
Update collective implementation. (#10152)
2024-03-30 18:57:31 +08:00
comm.cc
Update collective implementation. (#10152)
2024-03-30 18:57:31 +08:00
comm.cu
Update collective implementation. (#10152)
2024-03-30 18:57:31 +08:00
comm.cuh
Update collective implementation. (#10152)
2024-03-30 18:57:31 +08:00
comm.h
Update collective implementation. (#10152)
2024-03-30 18:57:31 +08:00
communicator-inl.cc
Use UBJSON for serializing splits for vertical data split. (#10059)
2024-02-25 00:18:23 +08:00
communicator-inl.cuh
Support column split in GPU evaluate splits (#9511)
2023-08-23 16:33:43 +08:00
communicator-inl.h
Use UBJSON for serializing splits for vertical data split. (#10059)
2024-02-25 00:18:23 +08:00
communicator.cc
Use dlopen to load NCCL. (#9796)
2023-11-22 19:27:31 +08:00
communicator.cu
Use dlopen to load NCCL. (#9796)
2023-11-22 19:27:31 +08:00
communicator.h
Use dlopen to load NCCL. (#9796)
2023-11-22 19:27:31 +08:00
device_communicator_adapter.cuh
Use std::uint64_t for row index. (#10120)
2024-03-15 18:43:49 +08:00
device_communicator.cuh
Support column split in GPU evaluate splits (#9511)
2023-08-23 16:33:43 +08:00
in_memory_communicator.cc
Add an in-memory collective communicator (#8494)
2022-12-01 00:24:12 +08:00
in_memory_communicator.h
Improve allgather functions (#9649)
2023-10-12 23:31:43 +08:00
in_memory_handler.cc
Improve allgather functions (#9649)
2023-10-12 23:31:43 +08:00
in_memory_handler.h
Update collective implementation. (#10152)
2024-03-30 18:57:31 +08:00
loop.cc
Fix error handling in the event loop. (#9990)
2024-01-17 05:35:35 +08:00
loop.h
Fix error handling in the event loop. (#9990)
2024-01-17 05:35:35 +08:00
nccl_device_communicator.cu
Cleanup set info. (#10139)
2024-03-26 23:26:24 +08:00
nccl_device_communicator.cuh
Cleanup code for distributed training. (#9805)
2023-11-25 09:10:56 +08:00
nccl_stub.cc
Cleanup code for distributed training. (#9805)
2023-11-25 09:10:56 +08:00
nccl_stub.h
Cleanup code for distributed training. (#9805)
2023-11-25 09:10:56 +08:00
noop_communicator.h
Improve allgather functions (#9649)
2023-10-12 23:31:43 +08:00
protocol.h
[coll] Implement a new tracker and a communicator. (#9650)
2023-10-12 12:49:16 +08:00
rabit_communicator.h
Support column-wise data split with in-memory inputs (#9628)
2023-10-17 12:16:39 +08:00
socket.cc
Build a simple event loop for collective. (#9593)
2023-09-20 02:09:07 +08:00
tracker.cc
Work with IPv6 in the new tracker. (#10125)
2024-03-20 05:19:23 +08:00
tracker.h
Work with IPv6 in the new tracker. (#10125)
2024-03-20 05:19:23 +08:00