smyalygames/FiniteVolumeGPU
1
0
Fork 0
mirror of https://github.com/smyalygames/FiniteVolumeGPU.git synced 2025-10-16 09:17:41 +02:00
Code Issues Actions Packages Projects Releases Wiki Activity

Added benchmarks and updated figures.

Browse Source
This commit is contained in:
Martin Lilleeng Sætra 2022-06-24 14:33:16 +02:00
parent aa693a9a46
commit 528b738dda
4 changed files with 305 additions and 148 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

358
Figures.ipynb
View File

File diff suppressed because one or more lines are too long

2
dgx-2_scaling_benchmark.job
View File

@ -6,7 +6,7 @@
#SBATCH -t 0-00:10 # time (D-HH:MM) #SBATCH -t 0-00:10 # time (D-HH:MM)
#SBATCH -o slurm.%N.%j.out # STDOUT #SBATCH -o slurm.%N.%j.out # STDOUT
#SBATCH -e slurm.%N.%j.err # STDERR #SBATCH -e slurm.%N.%j.err # STDERR
#SBATCH --reservation=martinls_8 #SBATCH --reservation=martinls_17
# For Linux 64, Open MPI is built with CUDA awareness but this support is disabled by default. # For Linux 64, Open MPI is built with CUDA awareness but this support is disabled by default.

61
dgx-2_strong_scaling_benchmark.sh
View File

@ -22,17 +22,52 @@ TIMESTAMP=$(date "+%Y-%m-%dT%H%M%S")
#sbatch --nodes=1 --gpus-per-node=16 --ntasks-per-node=16 --export=ALL,NX=8192,NY=512,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job #sbatch --nodes=1 --gpus-per-node=16 --ntasks-per-node=16 --export=ALL,NX=8192,NY=512,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
# one node: 4-16 GPUs # one node: 4-16 GPUs
sbatch --nodes=1 --gpus-per-node=4 --ntasks-per-node=4 --export=ALL,NX=41984,NY=10496,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job #sbatch --nodes=1 --gpus-per-node=4 --ntasks-per-node=4 --export=ALL,NX=41984,NY=10496,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=5 --ntasks-per-node=5 --export=ALL,NX=41984,NY=8396,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job #sbatch --nodes=1 --gpus-per-node=5 --ntasks-per-node=5 --export=ALL,NX=41984,NY=8396,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=6 --ntasks-per-node=6 --export=ALL,NX=41984,NY=6997,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job #sbatch --nodes=1 --gpus-per-node=6 --ntasks-per-node=6 --export=ALL,NX=41984,NY=6997,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=7 --ntasks-per-node=7 --export=ALL,NX=41984,NY=5997,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job #sbatch --nodes=1 --gpus-per-node=7 --ntasks-per-node=7 --export=ALL,NX=41984,NY=5997,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=8 --ntasks-per-node=8 --export=ALL,NX=41984,NY=5248,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job #sbatch --nodes=1 --gpus-per-node=8 --ntasks-per-node=8 --export=ALL,NX=41984,NY=5248,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
#
#sbatch --nodes=1 --gpus-per-node=9 --ntasks-per-node=9 --export=ALL,NX=41984,NY=4664,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
#sbatch --nodes=1 --gpus-per-node=10 --ntasks-per-node=10 --export=ALL,NX=41984,NY=4198,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
#sbatch --nodes=1 --gpus-per-node=11 --ntasks-per-node=11 --export=ALL,NX=41984,NY=3816,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
#sbatch --nodes=1 --gpus-per-node=12 --ntasks-per-node=12 --export=ALL,NX=41984,NY=3498,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
#sbatch --nodes=1 --gpus-per-node=13 --ntasks-per-node=13 --export=ALL,NX=41984,NY=3229,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
#sbatch --nodes=1 --gpus-per-node=14 --ntasks-per-node=14 --export=ALL,NX=41984,NY=2998,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
#sbatch --nodes=1 --gpus-per-node=15 --ntasks-per-node=15 --export=ALL,NX=41984,NY=2798,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
#sbatch --nodes=1 --gpus-per-node=16 --ntasks-per-node=16 --export=ALL,NX=41984,NY=2624,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=9 --ntasks-per-node=9 --export=ALL,NX=41984,NY=4664,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job # one node: 1-16 GPUs
sbatch --nodes=1 --gpus-per-node=10 --ntasks-per-node=10 --export=ALL,NX=41984,NY=4198,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job sbatch --nodes=1 --gpus-per-node=1 --ntasks-per-node=1 --export=ALL,NX=22528,NY=22528,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=11 --ntasks-per-node=11 --export=ALL,NX=41984,NY=3816,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job sbatch --nodes=1 --gpus-per-node=2 --ntasks-per-node=2 --export=ALL,NX=22528,NY=11264,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=12 --ntasks-per-node=12 --export=ALL,NX=41984,NY=3498,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job sbatch --nodes=1 --gpus-per-node=3 --ntasks-per-node=3 --export=ALL,NX=22528,NY=7509,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=13 --ntasks-per-node=13 --export=ALL,NX=41984,NY=3229,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job sbatch --nodes=1 --gpus-per-node=4 --ntasks-per-node=4 --export=ALL,NX=22528,NY=5632,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=14 --ntasks-per-node=14 --export=ALL,NX=41984,NY=2998,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job sbatch --nodes=1 --gpus-per-node=5 --ntasks-per-node=5 --export=ALL,NX=22528,NY=4505,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=15 --ntasks-per-node=15 --export=ALL,NX=41984,NY=2798,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job sbatch --nodes=1 --gpus-per-node=6 --ntasks-per-node=6 --export=ALL,NX=22528,NY=3754,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=16 --ntasks-per-node=16 --export=ALL,NX=41984,NY=2624,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job sbatch --nodes=1 --gpus-per-node=7 --ntasks-per-node=7 --export=ALL,NX=22528,NY=3218,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=8 --ntasks-per-node=8 --export=ALL,NX=22528,NY=2816,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=9 --ntasks-per-node=9 --export=ALL,NX=22528,NY=2503,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=10 --ntasks-per-node=10 --export=ALL,NX=22528,NY=2252,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=11 --ntasks-per-node=11 --export=ALL,NX=22528,NY=2048,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=12 --ntasks-per-node=12 --export=ALL,NX=22528,NY=1877,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=13 --ntasks-per-node=13 --export=ALL,NX=22528,NY=1732,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=14 --ntasks-per-node=14 --export=ALL,NX=22528,NY=1609,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=15 --ntasks-per-node=15 --export=ALL,NX=22528,NY=1501,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=16 --ntasks-per-node=16 --export=ALL,NX=22528,NY=1408,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
# one node: 4-16 GPUs
sbatch --nodes=1 --gpus-per-node=4 --ntasks-per-node=4 --export=ALL,NX=45056,NY=11264,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=5 --ntasks-per-node=5 --export=ALL,NX=45056,NY=8396,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=6 --ntasks-per-node=6 --export=ALL,NX=45056,NY=6997,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=7 --ntasks-per-node=7 --export=ALL,NX=45056,NY=5997,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=8 --ntasks-per-node=8 --export=ALL,NX=45056,NY=5248,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=9 --ntasks-per-node=9 --export=ALL,NX=45056,NY=4664,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=10 --ntasks-per-node=10 --export=ALL,NX=45056,NY=4198,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=11 --ntasks-per-node=11 --export=ALL,NX=45056,NY=3816,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=12 --ntasks-per-node=12 --export=ALL,NX=45056,NY=3498,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=13 --ntasks-per-node=13 --export=ALL,NX=45056,NY=3229,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=14 --ntasks-per-node=14 --export=ALL,NX=45056,NY=2998,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=15 --ntasks-per-node=15 --export=ALL,NX=45056,NY=2798,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=16 --ntasks-per-node=16 --export=ALL,NX=45056,NY=2624,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job

32
dgx-2_weak_scaling_benchmark.sh
View File

@ -22,20 +22,20 @@ TIMESTAMP=$(date "+%Y-%m-%dT%H%M%S")
#sbatch --nodes=1 --gpus-per-node=16 --ntasks-per-node=16 --export=ALL,NX=8192,NY=8192,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job #sbatch --nodes=1 --gpus-per-node=16 --ntasks-per-node=16 --export=ALL,NX=8192,NY=8192,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
# one node: 1-16 GPUs # one node: 1-16 GPUs
sbatch --nodes=1 --gpus-per-node=1 --ntasks-per-node=1 --export=ALL,NX=41984,NY=41984,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job sbatch --nodes=1 --gpus-per-node=1 --ntasks-per-node=1 --export=ALL,NX=22528,NY=22528,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=2 --ntasks-per-node=2 --export=ALL,NX=41984,NY=41984,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job sbatch --nodes=1 --gpus-per-node=2 --ntasks-per-node=2 --export=ALL,NX=22528,NY=22528,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=3 --ntasks-per-node=3 --export=ALL,NX=41984,NY=41984,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job sbatch --nodes=1 --gpus-per-node=3 --ntasks-per-node=3 --export=ALL,NX=22528,NY=22528,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=4 --ntasks-per-node=4 --export=ALL,NX=41984,NY=41984,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job sbatch --nodes=1 --gpus-per-node=4 --ntasks-per-node=4 --export=ALL,NX=22528,NY=22528,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=5 --ntasks-per-node=5 --export=ALL,NX=41984,NY=41984,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job sbatch --nodes=1 --gpus-per-node=5 --ntasks-per-node=5 --export=ALL,NX=22528,NY=22528,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=6 --ntasks-per-node=6 --export=ALL,NX=41984,NY=41984,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job sbatch --nodes=1 --gpus-per-node=6 --ntasks-per-node=6 --export=ALL,NX=22528,NY=22528,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=7 --ntasks-per-node=7 --export=ALL,NX=41984,NY=41984,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job sbatch --nodes=1 --gpus-per-node=7 --ntasks-per-node=7 --export=ALL,NX=22528,NY=22528,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=8 --ntasks-per-node=8 --export=ALL,NX=41984,NY=41984,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job sbatch --nodes=1 --gpus-per-node=8 --ntasks-per-node=8 --export=ALL,NX=22528,NY=22528,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=9 --ntasks-per-node=9 --export=ALL,NX=41984,NY=41984,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job sbatch --nodes=1 --gpus-per-node=9 --ntasks-per-node=9 --export=ALL,NX=22528,NY=22528,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=10 --ntasks-per-node=10 --export=ALL,NX=41984,NY=41984,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job sbatch --nodes=1 --gpus-per-node=10 --ntasks-per-node=10 --export=ALL,NX=22528,NY=22528,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=11 --ntasks-per-node=11 --export=ALL,NX=41984,NY=41984,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job sbatch --nodes=1 --gpus-per-node=11 --ntasks-per-node=11 --export=ALL,NX=22528,NY=22528,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=12 --ntasks-per-node=12 --export=ALL,NX=41984,NY=41984,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job sbatch --nodes=1 --gpus-per-node=12 --ntasks-per-node=12 --export=ALL,NX=22528,NY=22528,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=13 --ntasks-per-node=13 --export=ALL,NX=41984,NY=41984,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job sbatch --nodes=1 --gpus-per-node=13 --ntasks-per-node=13 --export=ALL,NX=22528,NY=22528,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=14 --ntasks-per-node=14 --export=ALL,NX=41984,NY=41984,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job sbatch --nodes=1 --gpus-per-node=14 --ntasks-per-node=14 --export=ALL,NX=22528,NY=22528,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=15 --ntasks-per-node=15 --export=ALL,NX=41984,NY=41984,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job sbatch --nodes=1 --gpus-per-node=15 --ntasks-per-node=15 --export=ALL,NX=22528,NY=22528,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job
sbatch --nodes=1 --gpus-per-node=16 --ntasks-per-node=16 --export=ALL,NX=41984,NY=41984,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job sbatch --nodes=1 --gpus-per-node=16 --ntasks-per-node=16 --export=ALL,NX=22528,NY=22528,NOW=$TIMESTAMP dgx-2_scaling_benchmark.job