experiment=ch13_tree_ring_hierarchy
run_id=20260710T155522Z
timestamp_utc=2026-07-10T15:59:41.565440+00:00
hostname=sg-fuxc2-260708-12652-default0-0
cycles=5
iterations=20
world_sizes=2,3,4
sizes=4B..1GiB_factor2
algorithms=Ring,Tree
protocol=Simple
channels=12
practical_tie_percent=5.0
measurement_rows=1740
active_tree_rows=108
dtree_rows=136
dtree_compile=g++ -std=c++17 -O2 -I/usr/include -I/usr/local/cuda-12.6/targets/x86_64-linux/include /root/nccl-learning/probes/ch13_dtree_probe.cc /root/nccl-learning/third_party/nccl-2.22.3/src/graph/trees.cc -o /root/nccl-learning/logs/ch13_tree/20260710T155522Z/ch13_dtree_probe
nccl_tests_commit=5bcd45d2ee8365e37895e118c9d11b0ebc15aa69
nccl_source_commit=178b6b759074597777ce13438efb0e0ba625e429
command_01=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 4 -e 1G -f 2 -g 2 -w 5 -n 20 -N 5 -c 1 -I 0 -z 0 -C 0 -a 3 -d float -o sum
command_02=NCCL_ALGO=Tree NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 4 -e 1G -f 2 -g 2 -w 5 -n 20 -N 5 -c 1 -I 0 -z 0 -C 0 -a 3 -d float -o sum
command_03=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 4 -e 1G -f 2 -g 3 -w 5 -n 20 -N 5 -c 1 -I 0 -z 0 -C 0 -a 3 -d float -o sum
command_04=NCCL_ALGO=Tree NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 4 -e 1G -f 2 -g 3 -w 5 -n 20 -N 5 -c 1 -I 0 -z 0 -C 0 -a 3 -d float -o sum
command_05=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 4 -e 1G -f 2 -g 4 -w 5 -n 20 -N 5 -c 1 -I 0 -z 0 -C 0 -a 3 -d float -o sum
command_06=NCCL_ALGO=Tree NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 4 -e 1G -f 2 -g 4 -w 5 -n 20 -N 5 -c 1 -I 0 -z 0 -C 0 -a 3 -d float -o sum
command_07=NCCL_ALGO=Tree NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 4 -e 1G -f 2 -g 4 -w 5 -n 20 -N 5 -c 1 -I 0 -z 0 -C 0 -a 3 -d float -o sum
command_08=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 4 -e 1G -f 2 -g 4 -w 5 -n 20 -N 5 -c 1 -I 0 -z 0 -C 0 -a 3 -d float -o sum
command_09=NCCL_ALGO=Tree NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 4 -e 1G -f 2 -g 3 -w 5 -n 20 -N 5 -c 1 -I 0 -z 0 -C 0 -a 3 -d float -o sum
command_10=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 4 -e 1G -f 2 -g 3 -w 5 -n 20 -N 5 -c 1 -I 0 -z 0 -C 0 -a 3 -d float -o sum
command_11=NCCL_ALGO=Tree NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 4 -e 1G -f 2 -g 2 -w 5 -n 20 -N 5 -c 1 -I 0 -z 0 -C 0 -a 3 -d float -o sum
command_12=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 4 -e 1G -f 2 -g 2 -w 5 -n 20 -N 5 -c 1 -I 0 -z 0 -C 0 -a 3 -d float -o sum