search

a-paranjape / sahyadri-sandbox

Sandbox for testing codes and scripts related to Sahyadri simulations at IUCAA/TIFR/IISER-Pune/NCRA
0 stars 0 forks source link

Simulations for gadget scaling study #8

Closed SaeeDhawalikar closed 1 month ago

SaeeDhawalikar commented 7 months ago

Following is a list of simulations that were decided to be run before the pilot study to check the scaling of gadget4: naming convention: LxNy refers to a box size of x Mpc/h with y^3 particles.

Default: L200N256: check scaling with number of cores, with N_cores= 64, 128, 256, (512) Target resolution: L50N512: check scaling with number of cores, with N_cores =64, 128, 256, (512)

L50N256: N_cores=128 L200N1024: N_cores=128

Total: 8 simulations, with 2 simulations each having: a. the same resolution (L50N256 & L200N1024) b. the same number of particles (L50N256 & L200N256) c. the same volume (L200N256& L200N1024, L50N256 & L50N512)

Lenovo: total shared memory available: 193184 MB (6GB per cpu)

Simulation | NCPU | CPU time (hrs) | Walltime (hrs) | Peak memory (MB) | Total memory (GB) | Memory per Node (GB) -- | -- | -- | -- | -- | -- | -- L200N256 | 64 | 239.73 | 3.73 | 115.29 | 7.37856 | 3.68928   | 128 | 232.89 | 1.82 | 57.7113 | 7.3870464 | 1.8467616   | 256 | 263.48 | 1.02 | 28.8973 | 7.3977088 | 0.9247136   | 512 | 292.67 | 0.583 | 14.5148 | 7.4315776 | 0.4644736 L50N512 | 64 | 4871.85 | 77.88 | 922.169 | 59.018816 | 29.509408   | 128 | 4882.61 | 38.15 | 461.149 | 59.027072 | 14.756768   | 256 | 5627.79 | 21.97 | 230.667 | 59.050752 | 7.381344 L50N256 | 128 | 378.58 | 2.97 | 57.7113 | 7.3870464 | 1.8467616   | 256 | 422.67 | 1.65 | 28.8973 | 7.3977088 | 0.9247136 L200N1024 | 128 | 32254.86 | 265.43 | 3688.69 | 472.15232 | 118.03808 L100N128 | 16 | 23.46 | 1.47 | 55.8439 | 0.8935024 | 1.7870048 L158N203 | 64 | 130.47 | 2.03 | 57.4959 | 3.6797376 | 1.8398688 L216N276 | 160 | 305.69 | 1.9 | 57.8785 | 9.26056 | 1.852112 L182N233 | 96 | 203.96 | 2.13 | 58.0198 | 5.5699008 | 1.8566336 Advance: total shared memory available: 257703 MB (8 GB per cpu) Simulation | NCPU | CPU time (hrs) | Walltime (hrs) | Peak memory -- | -- | -- | -- | -- L200N256 | 64 | 174.36 | 2.716 | 115.29 | 2618.88   | 128 | 171.36 | 1.33 | 57.7113 | 2598.3   | 256 | 195.7056 | 0.77 | 28.8973 | 3086.71 L100N128 | 32 | 18.04 | 0.57 | 27.9273 | 2571
a-paranjape commented 5 months ago

@SaeeDhawalikar can you also add columns summarizing memory used in each run? i guess we want to know the peak memory used, both maximum of individual processors and peak total memory. this may be the deciding factor in setting ncpus for sahyadri, and we will need to justify the number in our proposal later this year.