Merzbild.jl benchmarks
Various benchmarks and comparisons to other open-source codes are provided here for reference test cases.
Couette flow, serial
Comparison with SPARTA are provided for a single-species (argon) Couette flow test case with 50000 particles and 50 cells (averaging over 36k timesteps after t>14000). The computation is serial. Timing in Merzbild.jl providedd by TimerOutputs.jl, timing in SPARTA provided by the inbuilt timers.
Merzbild.jl version 0.6.2, run with --check-bounds=no -O3.
SPARTA version 4Sep2024, compiled with -O3.
Intel Core i9-13900K, 128 GB RAM
Ubuntu 22.04.5, Julia version 1.11.2, gcc version 11.4.0.
Merzbild.jl
──────────────────────────────────────────────────────────────────────────
Time Allocations
─────────────────────── ────────────────────────
Tot / % measured: 30.1s / 95.8% 697MiB / 0.4%
Section ncalls time %tot avg alloc %tot avg
──────────────────────────────────────────────────────────────────────────
sort 50.0k 10.7s 37.0% 213μs 0.00B 0.0% 0.00B
convect 50.0k 6.61s 22.9% 132μs 0.00B 0.0% 0.00B
collide 2.50M 6.33s 22.0% 2.53μs 0.00B 0.0% 0.00B
props compute 36.0k 5.19s 18.0% 144μs 0.00B 0.0% 0.00B
sampling 1 2.61ms 0.0% 2.61ms 3.05MiB 100.0% 3.05MiB
I/O 1 369μs 0.0% 369μs 240B 0.0% 240B
──────────────────────────────────────────────────────────────────────────SPARTA
Loop time of 31.1462 on 1 procs for 50000 steps with 50000 particles
MPI task timing breakdown:
Section | min time | avg time | max time |%varavg| %total
---------------------------------------------------------------
Move | 8.1848 | 8.1848 | 8.1848 | 0.0 | 26.28
Coll | 11.155 | 11.155 | 11.155 | 0.0 | 35.81
Sort | 2.8435 | 2.8435 | 2.8435 | 0.0 | 9.13
Comm | 0.0032952 | 0.0032952 | 0.0032952 | 0.0 | 0.01
Modify | 8.9578 | 8.9578 | 8.9578 | 0.0 | 28.76
Output | 0.00046062 | 0.00046062 | 0.00046062 | 0.0 | 0.00
Other | | 0.001451 | | | 0.00M1 Pro (Macbook Pro), 32 GB RAM
MacOS 12.7.5, Julia version 1.11.2, Apple clang version 14.0.0.
Merzbild.jl
──────────────────────────────────────────────────────────────────────────
Time Allocations
─────────────────────── ────────────────────────
Tot / % measured: 33.8s / 96.3% 693MiB / 0.4%
Section ncalls time %tot avg alloc %tot avg
──────────────────────────────────────────────────────────────────────────
sort 50.0k 11.6s 35.7% 233μs 0.00B 0.0% 0.00B
collide 2.50M 8.57s 26.3% 3.43μs 0.00B 0.0% 0.00B
convect 50.0k 6.86s 21.0% 137μs 0.00B 0.0% 0.00B
props compute 36.0k 5.51s 16.9% 153μs 0.00B 0.0% 0.00B
sampling 1 2.64ms 0.0% 2.64ms 3.05MiB 100.0% 3.05MiB
I/O 1 139μs 0.0% 139μs 240B 0.0% 240B
──────────────────────────────────────────────────────────────────────────SPARTA
Loop time of 51.1924 on 1 procs for 50000 steps with 50000 particles
MPI task timing breakdown:
Section | min time | avg time | max time |%varavg| %total
---------------------------------------------------------------
Move | 13.622 | 13.622 | 13.622 | 0.0 | 26.61
Coll | 13.132 | 13.132 | 13.132 | 0.0 | 25.65
Sort | 2.8252 | 2.8252 | 2.8252 | 0.0 | 5.52
Comm | 0.002234 | 0.002234 | 0.002234 | 0.0 | 0.00
Modify | 21.607 | 21.607 | 21.607 | 0.0 | 42.21
Output | 0.0025806 | 0.0025806 | 0.0025806 | 0.0 | 0.01
Other | | 0.0009735 | | | 0.00