microsoft/mscclpp
1
0
Fork 0
mirror of https://github.com/microsoft/mscclpp.git synced 2026-05-11 17:00:22 +00:00
Code Issues Packages Projects Releases Wiki Activity

Manage runtime environments (#452)

Browse Source 
* Add `Env` class that manages all runtime environments.
* Changed `NPKIT_DUMP_DIR` to `MSCCLPP_NPKIT_DUMP_DIR`.
This commit is contained in:
Changho Hwang
2025-01-15 09:44:52 -08:00
committed by GitHub
parent 8ac50dc85d
commit 869cdba00c
19 changed files with 229 additions and 51 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
tools/npkit/build_and_run_npkit.sh
View File

@@ -17,6 +17,6 @@ parallel-ssh -h ${HOSTFILE} "mkdir -p ${NPKIT_RUN_DIR}/npkit_dump"
parallel-ssh -h ${HOSTFILE} "mkdir -p ${NPKIT_RUN_DIR}/npkit_trace"
# --bind-to numa is required because hardware timer from different cores (or core groups) can be non-synchronized.
mpirun --allow-run-as-root -hostfile ${HOSTFILE} -map-by ppr:8:node --bind-to numa -x LD_PRELOAD=${NPKIT_RUN_DIR}/mscclpp/build/lib/libmscclpp.so -x MSCCLPP_DEBUG=WARN -x NPKIT_DUMP_DIR=${NPKIT_RUN_DIR}/npkit_dump ${NPKIT_RUN_DIR}/mscclpp/build/bin/tests/allgather_test -ip_port ${LEADER_IP_PORT} -kernel 0
mpirun --allow-run-as-root -hostfile ${HOSTFILE} -map-by ppr:8:node --bind-to numa -x LD_PRELOAD=${NPKIT_RUN_DIR}/mscclpp/build/lib/libmscclpp.so -x MSCCLPP_DEBUG=WARN -x MSCCLPP_NPKIT_DUMP_DIR=${NPKIT_RUN_DIR}/npkit_dump ${NPKIT_RUN_DIR}/mscclpp/build/bin/tests/allgather_test -ip_port ${LEADER_IP_PORT} -kernel 0
parallel-ssh -h ${HOSTFILE} "cd ${NPKIT_RUN_DIR}/mscclpp/tools/npkit && python npkit_trace_generator.py --npkit_dump_dir ${NPKIT_RUN_DIR}/npkit_dump --npkit_event_header_path ${NPKIT_RUN_DIR}/mscclpp/src/include/npkit/npkit_event.h --output_dir ${NPKIT_RUN_DIR}/npkit_trace"