osu_gather significant slowdown from 8b to 1024b
The following comparison illustrates the issue (on current image):
mpiexec --np 3072 --ppn 96 --cpu-bind verbose,list:1:2:3:4:5:6:7:8:9:10:11:12:13:14:15:16:18:19:20:21:22:23:24:25:26:27:28:29:30:31:32:33:35:36:37:38:39:40:41:42:43:44:45:46:47:48:49:50:53:54:55:56:57:58:59:60:61:62:63:64:65:66:67:68:70:71:72:73:74:75:76:77:78:79:80:81:82:83:84:85:87:88:89:90:91:92:93:94:95:96:97:98:99:100:101:102 --gpu-bind verbose,list:0.0:0.0:0.0:0.0:0.0:0.0:0.0:0.0:0.1:0.1:0.1:0.1:0.1:0.1:0.1:0.1:1.0:1.0:1.0:1.0:1.0:1.0:1.0:1.0:1.1:1.1:1.1:1.1:1.1:1.1:1.1:1.1:2.0:2.0:2.0:2.0:2.0:2.0:2.0:2.0:2.1:2.1:2.1:2.1:2.1:2.1:2.1:2.1:3.0:3.0:3.0:3.0:3.0:3.0:3.0:3.0:3.1:3.1:3.1:3.1:3.1:3.1:3.1:3.1:4.0:4.0:4.0:4.0:4.0:4.0:4.0:4.0:4.1:4.1:4.1:4.1:4.1:4.1:4.1:4.1:5.0:5.0:5.0:5.0:5.0:5.0:5.0:5.0:5.1:5.1:5.1:5.1:5.1:5.1:5.1:5.1 /lus/flare/projects/Aurora_testing/mpi/osu_rfm/run_collective/32/gather-gather_persistent-gatherv-gatherv_persistent/stage/2025-09-12_18-37-26/aurora/compute/PrgEnv-intel/BuildMPIcollective_93bceebc/binaries/osu_gatherv -m 8:8 -i 1000 -x 100 -f -z -d sycl
# OSU MPI-SYCL Gatherv Latency Test v7.5 # Datatype: MPI_CHAR. # Size Avg Latency(us) Min Latency(us) Max Latency(us) Iterations P50 Tail Lat(us) P90 Tail Lat(us) P99 Tail Lat(us) 8 119.12 1.61 124007.70 1000 117.50 123.89 193.58
mpiexec --np 3072 --ppn 96 --cpu-bind verbose,list:1:2:3:4:5:6:7:8:9:10:11:12:13:14:15:16:18:19:20:21:22:23:24:25:26:27:28:29:30:31:32:33:35:36:37:38:39:40:41:42:43:44:45:46:47:48:49:50:53:54:55:56:57:58:59:60:61:62:63:64:65:66:67:68:70:71:72:73:74:75:76:77:78:79:80:81:82:83:84:85:87:88:89:90:91:92:93:94:95:96:97:98:99:100:101:102 --gpu-bind verbose,list:0.0:0.0:0.0:0.0:0.0:0.0:0.0:0.0:0.1:0.1:0.1:0.1:0.1:0.1:0.1:0.1:1.0:1.0:1.0:1.0:1.0:1.0:1.0:1.0:1.1:1.1:1.1:1.1:1.1:1.1:1.1:1.1:2.0:2.0:2.0:2.0:2.0:2.0:2.0:2.0:2.1:2.1:2.1:2.1:2.1:2.1:2.1:2.1:3.0:3.0:3.0:3.0:3.0:3.0:3.0:3.0:3.1:3.1:3.1:3.1:3.1:3.1:3.1:3.1:4.0:4.0:4.0:4.0:4.0:4.0:4.0:4.0:4.1:4.1:4.1:4.1:4.1:4.1:4.1:4.1:5.0:5.0:5.0:5.0:5.0:5.0:5.0:5.0:5.1:5.1:5.1:5.1:5.1:5.1:5.1:5.1 /lus/flare/projects/Aurora_testing/mpi/osu_rfm/run_collective/32/gather-gather_persistent-gatherv-gatherv_persistent/stage/2025-09-12_18-37-26/aurora/compute/PrgEnv-intel/BuildMPIcollective_93bceebc/binaries/osu_gatherv -m 1024:1024 -i 1000 -x 100 -f -z -d sycl
# OSU MPI-SYCL Gatherv Latency Test v7.5 # Datatype: MPI_CHAR. # Size Avg Latency(us) Min Latency(us) Max Latency(us) Iterations P50 Tail Lat(us) P90 Tail Lat(us) P99 Tail Lat(us) 1024 94969.64 678.15 116240.08 1000 88296.68 123401.15 441821.83
mpiexec --np 12288 --ppn 96 --cpu-bind verbose,list:1:2:3:4:5:6:7:8:9:10:11:12:13:14:15:16:18:19:20:21:22:23:24:25:26:27:28:29:30:31:32:33:35:36:37:38:39:40:41:42:43:44:45:46:47:48:49:50:53:54:55:56:57:58:59:60:61:62:63:64:65:66:67:68:70:71:72:73:74:75:76:77:78:79:80:81:82:83:84:85:87:88:89:90:91:92:93:94:95:96:97:98:99:100:101:102 --gpu-bind verbose,list:0.0:0.0:0.0:0.0:0.0:0.0:0.0:0.0:0.1:0.1:0.1:0.1:0.1:0.1:0.1:0.1:1.0:1.0:1.0:1.0:1.0:1.0:1.0:1.0:1.1:1.1:1.1:1.1:1.1:1.1:1.1:1.1:2.0:2.0:2.0:2.0:2.0:2.0:2.0:2.0:2.1:2.1:2.1:2.1:2.1:2.1:2.1:2.1:3.0:3.0:3.0:3.0:3.0:3.0:3.0:3.0:3.1:3.1:3.1:3.1:3.1:3.1:3.1:3.1:4.0:4.0:4.0:4.0:4.0:4.0:4.0:4.0:4.1:4.1:4.1:4.1:4.1:4.1:4.1:4.1:5.0:5.0:5.0:5.0:5.0:5.0:5.0:5.0:5.1:5.1:5.1:5.1:5.1:5.1:5.1:5.1 /lus/flare/projects/Aurora_testing/mpi/osu_rfm/run_collective/128/gather-gather_persistent-gatherv-gatherv_persistent/stage/2025-09-12_18-37-27/aurora/compute/PrgEnv-intel/BuildMPIcollective_93bceebc/binaries/osu_gatherv -m 8:8 -i 1000 -x 100 -f -z -d sycl
# OSU MPI-SYCL Gatherv Latency Test v7.5 # Datatype: MPI_CHAR. # Size Avg Latency(us) Min Latency(us) Max Latency(us) Iterations P50 Tail Lat(us) P90 Tail Lat(us) P99 Tail Lat(us) 8 102.16 1.70 288610.67 1000 89.81 109.57 122.85
mpiexec --np 12288 --ppn 96 --cpu-bind verbose,list:1:2:3:4:5:6:7:8:9:10:11:12:13:14:15:16:18:19:20:21:22:23:24:25:26:27:28:29:30:31:32:33:35:36:37:38:39:40:41:42:43:44:45:46:47:48:49:50:53:54:55:56:57:58:59:60:61:62:63:64:65:66:67:68:70:71:72:73:74:75:76:77:78:79:80:81:82:83:84:85:87:88:89:90:91:92:93:94:95:96:97:98:99:100:101:102 --gpu-bind verbose,list:0.0:0.0:0.0:0.0:0.0:0.0:0.0:0.0:0.1:0.1:0.1:0.1:0.1:0.1:0.1:0.1:1.0:1.0:1.0:1.0:1.0:1.0:1.0:1.0:1.1:1.1:1.1:1.1:1.1:1.1:1.1:1.1:2.0:2.0:2.0:2.0:2.0:2.0:2.0:2.0:2.1:2.1:2.1:2.1:2.1:2.1:2.1:2.1:3.0:3.0:3.0:3.0:3.0:3.0:3.0:3.0:3.1:3.1:3.1:3.1:3.1:3.1:3.1:3.1:4.0:4.0:4.0:4.0:4.0:4.0:4.0:4.0:4.1:4.1:4.1:4.1:4.1:4.1:4.1:4.1:5.0:5.0:5.0:5.0:5.0:5.0:5.0:5.0:5.1:5.1:5.1:5.1:5.1:5.1:5.1:5.1 /lus/flare/projects/Aurora_testing/mpi/osu_rfm/run_collective/128/gather-gather_persistent-gatherv-gatherv_persistent/stage/2025-09-12_18-37-27/aurora/compute/PrgEnv-intel/BuildMPIcollective_93bceebc/binaries/osu_gatherv -m 1024:1024 -i 1000 -x 100 -f -z -d sycl
# OSU MPI-SYCL Gatherv Latency Test v7.5 # Datatype: MPI_CHAR. # Size Avg Latency(us) Min Latency(us) Max Latency(us) Iterations P50 Tail Lat(us) P90 Tail Lat(us) P99 Tail Lat(us) 1024 212708.27 8437.28 360412.57 1000 182955.37 339633.23 535149.98
Please show the complete benchmark results from 1 byte to 4096 bytes
We only tested 8B 1024B 4096B, all results can be found in this file
/lus/flare/projects/Aurora_testing/mpi/osu_rfm/run_collective/32/gather-gather_persistent-gatherv-gatherv_persistent/stage/2025-09-12_18-37-26/aurora/compute/PrgEnv-intel/RunMPIcollective/rfm_job.out, feel free to take a look