Maximum number of hops before the packet is discarded on the network. Default value is 16 (2.56ms).Įnables NVTX tracing for use with Nsight systems.Ĭontrols the DC_KEY for userspace RDMA DC targets for WekaFS and GPFS. Minimum RNR value for QP after which the QP will error out with RNR timeout if no Work Request is posted on the remote end. Specifies the default log path, which is the current working directory of the application.Ĭontrols the tracing level and can override the trace level for a specific application without requiring a new configuration file. Sets QOS level on IB device QP for userspace RDMA targets (WekaFS and GPFS).ĬUFILE_LOGFILE_PATH= /etc/log/cufile_$$.logĬontrols the path for cuFile log information. Sets QOS level on RoCEv2 device QP for userspace RDMA targets (WekaFS and GPFS). This can be used for container environments and applications that require different configuration settings from system default configuration at /etc/cufile.json. When set to 1, allows testing with new filesystems that are not yet officially enabled with cuFile.ĬUFILE_ENV_PATH_JSON= /home/user/cufile.jsonĬontrols the path where the cuFile library reads the configuration variables from.
GPU index 0 A100-PCIE-40GB bar:1 bar size (MiB):65536 supports GDSĬompletion queue depth for the DC target.Ĭontrols whether cufile checks for supporting filesystems. Miscellaneous.api_check_aggressive : false Properties.rdma_peer_affinity_policy : RoundRobinįs.generic.posix_unaligned_writes : false Properties.posix_pool_slab_count : 128 64 32 Properties.max_device_pinned_mem_size_kb : 33554432 Properties.max_device_cache_size_kb : 131072 Properties.max_batch_io_timeout_msecs : 5 rdma library : Not Loaded (libcufile_rdma.so) Nvidia_fs version: 2.7 libcufile version: 2.4
0 kommentar(er)
