文件名称：mpi-nccl-tests:使用GPU Direct RDMA进行MPI + NCCL测试
文件大小：19KB
文件格式：ZIP
更新时间：2024-06-06 01:04:39
distributed-systems mpi gpu-computing Cuda
MPI和NCCL GPU直接RDMA测试 建造 mkdir build && cd build cmake .. && make
【文件预览】：
mpi-nccl-tests-master
----benchmark()
--------common.h(4KB)
--------allreduce.cxx(1KB)
--------CMakeLists.txt(354B)
--------nccl_allreduce.cxx(1KB)
----cmake()
--------FindNCCL.cmake(794B)
----compile.sh(171B)
----experiments()
--------send_recv.cu(2KB)
--------nccl-two-communicators-diff-types-with-threads.cu(7KB)
--------nccl_one_device_per_process.cu(4KB)
--------nccl-two-streams-test.cu(5KB)
--------nccl-two-communicators-diff-types.cu(8KB)
--------CMakeLists.txt(346B)
--------reduce_scatter.cu(5KB)
--------nccl-two-streams-test-diff-streams.cu(5KB)
--------nccl-two-streams-test-diff-types.cu(7KB)
----run.sh(979B)
----.gitignore(270B)
----CMakeLists.txt(586B)
----README.md(100B)