Open SMAC-Zhang opened 6 months ago
then I use
./ns3-dev-AstraSimNetwork-default \
--workload-configuration="${SCRIPT_DIR:?}"/../../inputs/workload/ASTRA-sim-2.0/allReduce \
--system-configuration="${SCRIPT_DIR:?}"/../../inputs/system/Switch.json \
--network-configuration="../../../ns-3/scratch/config/config.txt" \
--remote-memory-configuration="${SCRIPT_DIR:?}"/../../inputs/remote_memory/analytical/no_memory_expansion.json \
--logical-topology-configuration="${SCRIPT_DIR:?}"/../../inputs/network/ns3/sample_8nodes_1D.json \
--comm-group-configuration=\"empty\" \
to run it. And I get following output
ASTRA-sim + NS3
There are 8 npus: 8,
ring of node 0, id: 0 dimension: local total nodes in ring: 8 index in ring: 0 offset: 1total nodes in ring: 8
ring of node 0, id: 0 dimension: local total nodes in ring: 8 index in ring: 0 offset: 1total nodes in ring: 8
ring of node 0, id: 0 dimension: local total nodes in ring: 8 index in ring: 0 offset: 1total nodes in ring: 8
ring of node 0, id: 0 dimension: local total nodes in ring: 8 index in ring: 0 offset: 1total nodes in ring: 8
QP is enabled
maxRtt=2040 maxBdp=102000
sys[0] finished, 0 cycles
sys[1] finished, 0 cycles
sys[2] finished, 0 cycles
sys[3] finished, 0 cycles
sys[4] finished, 0 cycles
sys[5] finished, 0 cycles
sys[6] finished, 0 cycles
sys[7] finished, 0 cycles
Why it happened?
Hi @SMAC-Zhang , I encountered the same problem. Have you found a solution?
Hi @SMAC-Zhang , I encountered the same problem. Have you found a solution?
It seems something was wrong in the chakra.et_converter.et_converter, so you shouldn't convert the tracefile in astra-sim1.0. You can modify et_generator.py according to your needs and execute it to generate collective communication tracefile.
Hello, I want to use astra-sim2.0&ns3 to compare different collective algorithm. I want simulate an allreduce sample first. I use
to obtain workload.