experiment=ch15_channels_chunks_buffers run_id=20260710T191306Z timestamp_utc=2026-07-10T19:17:35.366714+00:00 hostname=sg-fuxc2-260708-12652-default0-0 world_size=4 algorithm=Ring protocol=Simple cycles=10 iterations=20 process_configs=44 measurement_rows=1060 nsys_profiles=7 private_nsys_dir=/root/nccl-learning/private/ch15_pipeline/20260710T191306Z nccl_tests_commit=5bcd45d2ee8365e37895e118c9d11b0ebc15aa69 nccl_source_commit=178b6b759074597777ce13438efb0e0ba625e429 command_01=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=1 NCCL_MAX_NCHANNELS=1 NCCL_BUFFSIZE=4194304 NCCL_NTHREADS=512 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 4K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_02=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=2 NCCL_MAX_NCHANNELS=2 NCCL_BUFFSIZE=4194304 NCCL_NTHREADS=512 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 4K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_03=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=4 NCCL_MAX_NCHANNELS=4 NCCL_BUFFSIZE=4194304 NCCL_NTHREADS=512 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 4K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_04=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=8 NCCL_MAX_NCHANNELS=8 NCCL_BUFFSIZE=4194304 NCCL_NTHREADS=512 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 4K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 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 NCCL_BUFFSIZE=4194304 NCCL_NTHREADS=512 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 4K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_06=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 NCCL_BUFFSIZE=65536 NCCL_NTHREADS=512 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 512K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_07=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 NCCL_BUFFSIZE=262144 NCCL_NTHREADS=512 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 512K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 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 NCCL_BUFFSIZE=1048576 NCCL_NTHREADS=512 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 512K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_09=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 NCCL_BUFFSIZE=4194304 NCCL_NTHREADS=512 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 512K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 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 NCCL_BUFFSIZE=16777216 NCCL_NTHREADS=512 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 512K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_11=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 NCCL_BUFFSIZE=4194304 NCCL_NTHREADS=64 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 4K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 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 NCCL_BUFFSIZE=4194304 NCCL_NTHREADS=128 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 4K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_13=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 NCCL_BUFFSIZE=4194304 NCCL_NTHREADS=256 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 4K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_14=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 NCCL_BUFFSIZE=4194304 NCCL_NTHREADS=512 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 4K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_15=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=2 NCCL_MAX_NCHANNELS=2 NCCL_BUFFSIZE=262144 NCCL_NTHREADS=128 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 512K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_16=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=2 NCCL_MAX_NCHANNELS=2 NCCL_BUFFSIZE=262144 NCCL_NTHREADS=512 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 512K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_17=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=2 NCCL_MAX_NCHANNELS=2 NCCL_BUFFSIZE=4194304 NCCL_NTHREADS=128 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 512K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_18=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=2 NCCL_MAX_NCHANNELS=2 NCCL_BUFFSIZE=4194304 NCCL_NTHREADS=512 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 512K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_19=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 NCCL_BUFFSIZE=262144 NCCL_NTHREADS=128 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 512K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_20=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 NCCL_BUFFSIZE=262144 NCCL_NTHREADS=512 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 512K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_21=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 NCCL_BUFFSIZE=4194304 NCCL_NTHREADS=128 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 512K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_22=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 NCCL_BUFFSIZE=4194304 NCCL_NTHREADS=512 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 512K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_23=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 NCCL_BUFFSIZE=4194304 NCCL_NTHREADS=512 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 512K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_24=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 NCCL_BUFFSIZE=4194304 NCCL_NTHREADS=128 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 512K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_25=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 NCCL_BUFFSIZE=262144 NCCL_NTHREADS=512 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 512K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_26=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 NCCL_BUFFSIZE=262144 NCCL_NTHREADS=128 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 512K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_27=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=2 NCCL_MAX_NCHANNELS=2 NCCL_BUFFSIZE=4194304 NCCL_NTHREADS=512 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 512K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_28=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=2 NCCL_MAX_NCHANNELS=2 NCCL_BUFFSIZE=4194304 NCCL_NTHREADS=128 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 512K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_29=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=2 NCCL_MAX_NCHANNELS=2 NCCL_BUFFSIZE=262144 NCCL_NTHREADS=512 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 512K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_30=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=2 NCCL_MAX_NCHANNELS=2 NCCL_BUFFSIZE=262144 NCCL_NTHREADS=128 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 512K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_31=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 NCCL_BUFFSIZE=4194304 NCCL_NTHREADS=512 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 4K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_32=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 NCCL_BUFFSIZE=4194304 NCCL_NTHREADS=256 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 4K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_33=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 NCCL_BUFFSIZE=4194304 NCCL_NTHREADS=128 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 4K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_34=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 NCCL_BUFFSIZE=4194304 NCCL_NTHREADS=64 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 4K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_35=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 NCCL_BUFFSIZE=16777216 NCCL_NTHREADS=512 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 512K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_36=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 NCCL_BUFFSIZE=4194304 NCCL_NTHREADS=512 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 512K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_37=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 NCCL_BUFFSIZE=1048576 NCCL_NTHREADS=512 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 512K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_38=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 NCCL_BUFFSIZE=262144 NCCL_NTHREADS=512 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 512K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_39=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 NCCL_BUFFSIZE=65536 NCCL_NTHREADS=512 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 512K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_40=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=12 NCCL_MAX_NCHANNELS=12 NCCL_BUFFSIZE=4194304 NCCL_NTHREADS=512 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 4K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_41=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=8 NCCL_MAX_NCHANNELS=8 NCCL_BUFFSIZE=4194304 NCCL_NTHREADS=512 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 4K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_42=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=4 NCCL_MAX_NCHANNELS=4 NCCL_BUFFSIZE=4194304 NCCL_NTHREADS=512 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 4K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_43=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=2 NCCL_MAX_NCHANNELS=2 NCCL_BUFFSIZE=4194304 NCCL_NTHREADS=512 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 4K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum command_44=NCCL_ALGO=Ring NCCL_PROTO=Simple NCCL_MIN_NCHANNELS=1 NCCL_MAX_NCHANNELS=1 NCCL_BUFFSIZE=4194304 NCCL_NTHREADS=512 /root/nccl-learning/third_party/nccl-tests/build/all_reduce_perf -b 4K -e 64M -f 128 -g 4 -w 5 -n 20 -N 10 -c 1 -I 0 -z 0 -u 0 -C 0 -a 3 -d float -o sum