Benchmarks

Here we benchmark the model performance in two Architectures. The number of individuals used in the benchmark are (2^10, 2^12, 2^14, 2^15). And we also use different grid resolutions in 2-Dimensional and 3-Dimensional model setup.

0-Dimensional model

This is a benchmark of a simple 0-Dimensional model setup without advection of Eulerian tracers. However, the advection of individuals still take the same amount of time whether the velocity field is provided or not.

PlanktonIndividuals v0.7.5
Julia Version 1.11.7
Commit f2b3dbda30a (2025-09-08 12:10 UTC)
Platform Info:
  OS: Linux (x86_64-linux-gnu)
  CPU: 56 × Intel(R) Xeon(R) Gold 6132 CPU @ 2.60GHz
  WORD_SIZE: 64
  LLVM: libLLVM-16.0.6 (ORCJIT, skylake-avx512)
  GPU: Quadro GV100 (sm_70, 32.000 GiB available)
  CUDA runtime 12.9, artifact installation
  CUDA driver 565.57.1 for 12.7

Arch	N	min	median	mean	max	memory	allocs	samples
CPU	1024	2.536 ms	2.629 ms	2.696 ms	3.204 ms	463.71 KiB	3673	10
CPU	4096	8.091 ms	8.201 ms	8.252 ms	8.829 ms	632.18 KiB	3673	10
CPU	16384	30.433 ms	30.558 ms	30.745 ms	31.809 ms	1.28 MiB	3595	10
CPU	32768	59.959 ms	60.364 ms	60.354 ms	60.980 ms	2.15 MiB	3595	10
GPU	1024	13.006 ms	13.194 ms	13.322 ms	14.415 ms	2.68 MiB	77257	10
GPU	4096	13.152 ms	13.334 ms	13.386 ms	13.991 ms	2.68 MiB	77257	10
GPU	16384	13.562 ms	13.755 ms	13.800 ms	14.595 ms	2.68 MiB	77260	10
GPU	32768	14.646 ms	14.879 ms	14.948 ms	15.450 ms	2.68 MiB	77263	10

2-Dimensional model

This is the benchmark of a 2-Dimensional model setup with (Ns, 1, Ns) grid cells. Here Ns = [32, 64, 128].

PlanktonIndividuals v0.7.5
Julia Version 1.11.7
Commit f2b3dbda30a (2025-09-08 12:10 UTC)
Platform Info:
  OS: Linux (x86_64-linux-gnu)
  CPU: 56 × Intel(R) Xeon(R) Gold 6132 CPU @ 2.60GHz
  WORD_SIZE: 64
  LLVM: libLLVM-16.0.6 (ORCJIT, skylake-avx512)
  GPU: Quadro GV100 (sm_70, 32.000 GiB available)
  CUDA runtime 12.9, artifact installation
  CUDA driver 565.57.1 for 12.7

Arch	N	Ns	min	median	mean	max	memory	allocs	samples
CPU	1024	32	5.659 ms	5.788 ms	5.874 ms	6.463 ms	1.92 MiB	4317	10
CPU	1024	64	13.497 ms	13.608 ms	13.770 ms	15.069 ms	5.80 MiB	4878	10
CPU	1024	128	45.442 ms	54.734 ms	53.155 ms	61.680 ms	20.65 MiB	6158	10
CPU	4096	32	11.307 ms	11.427 ms	11.519 ms	12.184 ms	2.08 MiB	4238	10
CPU	4096	64	19.120 ms	19.479 ms	19.667 ms	20.766 ms	5.96 MiB	4878	10
CPU	4096	128	51.716 ms	56.439 ms	57.364 ms	65.370 ms	20.82 MiB	6158	10
CPU	16384	32	33.569 ms	33.907 ms	34.243 ms	35.965 ms	2.74 MiB	4239	10
CPU	16384	64	41.597 ms	42.110 ms	42.844 ms	45.371 ms	6.62 MiB	4879	10
CPU	16384	128	75.032 ms	87.652 ms	83.260 ms	89.017 ms	21.47 MiB	6159	10
CPU	32768	32	63.176 ms	63.657 ms	63.717 ms	64.465 ms	3.62 MiB	4239	10
CPU	32768	64	71.786 ms	72.317 ms	73.374 ms	76.535 ms	7.50 MiB	4879	10
CPU	32768	128	106.093 ms	116.530 ms	113.994 ms	120.500 ms	22.35 MiB	6159	10
GPU	1024	32	12.915 ms	13.093 ms	13.219 ms	13.851 ms	2.87 MiB	83564	10
GPU	1024	64	13.688 ms	14.272 ms	14.355 ms	15.455 ms	3.15 MiB	93423	10
GPU	1024	128	15.540 ms	16.113 ms	16.036 ms	16.361 ms	3.92 MiB	117894	10
GPU	4096	32	12.888 ms	13.249 ms	13.541 ms	15.147 ms	2.87 MiB	83564	10
GPU	4096	64	13.768 ms	13.912 ms	14.277 ms	15.145 ms	3.15 MiB	93423	10
GPU	4096	128	15.740 ms	16.697 ms	16.592 ms	17.422 ms	3.92 MiB	117894	10
GPU	16384	32	13.514 ms	13.832 ms	14.139 ms	16.544 ms	2.87 MiB	83565	10
GPU	16384	64	13.956 ms	14.564 ms	14.831 ms	17.842 ms	3.15 MiB	93425	10
GPU	16384	128	15.721 ms	15.844 ms	15.941 ms	16.880 ms	3.92 MiB	117896	10
GPU	32768	32	13.689 ms	13.823 ms	13.926 ms	15.010 ms	2.87 MiB	83568	10
GPU	32768	64	14.460 ms	15.067 ms	15.092 ms	15.700 ms	3.15 MiB	93428	10
GPU	32768	128	16.284 ms	17.469 ms	17.356 ms	18.070 ms	3.92 MiB	117899	10

3-Dimensional model

This is the benchmark of a 3-Dimensional model setup with (Ns, Ns, Ns) grid cells. Here Ns = [32, 64, 128].

PlanktonIndividuals v0.7.5
Julia Version 1.11.7
Commit f2b3dbda30a (2025-09-08 12:10 UTC)
Platform Info:
  OS: Linux (x86_64-linux-gnu)
  CPU: 56 × Intel(R) Xeon(R) Gold 6132 CPU @ 2.60GHz
  WORD_SIZE: 64
  LLVM: libLLVM-16.0.6 (ORCJIT, skylake-avx512)
  GPU: Quadro GV100 (sm_70, 32 GiB available)
  CUDA runtime 12.9, artifact installation
  CUDA driver 565.57.1 for 12.7

Arch	N	Ns	min	median	mean	max	memory	allocs	samples
CPU	1024	32	51.587 ms	52.013 ms	52.390 ms	54.706 ms	1.06 MiB	4305	10
CPU	1024	64	407.569 ms	414.229 ms	413.787 ms	419.221 ms	5.55 MiB	8017	10
CPU	1024	128	3.293 s	3.301 s	3.301 s	3.310 s	40.95 MiB	21585	2
CPU	4096	32	56.841 ms	58.601 ms	58.606 ms	60.820 ms	1.22 MiB	4305	10
CPU	4096	64	422.655 ms	425.630 ms	425.657 ms	428.760 ms	5.72 MiB	8017	10
CPU	4096	128	3.318 s	3.337 s	3.337 s	3.357 s	41.11 MiB	21585	2
CPU	16384	32	81.034 ms	82.007 ms	82.366 ms	84.177 ms	1.88 MiB	4306	10
CPU	16384	64	454.705 ms	457.783 ms	457.526 ms	459.801 ms	6.37 MiB	8018	10
CPU	16384	128	3.336 s	3.374 s	3.374 s	3.413 s	41.77 MiB	21586	2
CPU	32768	32	112.225 ms	113.137 ms	113.290 ms	114.879 ms	2.76 MiB	4306	10
CPU	32768	64	495.213 ms	497.812 ms	498.189 ms	501.221 ms	7.25 MiB	8018	10
CPU	32768	128	3.472 s	3.482 s	3.482 s	3.493 s	42.65 MiB	21586	2
GPU	1024	32	12.288 ms	12.387 ms	12.577 ms	13.428 ms	4.08 MiB	115165	10
GPU	1024	64	20.906 ms	21.227 ms	21.697 ms	24.127 ms	13.06 MiB	353213	10
GPU	1024	128	90.657 ms	110.816 ms	113.920 ms	168.241 ms	83.72 MiB	2211802	10
GPU	4096	32	12.225 ms	12.322 ms	12.434 ms	13.159 ms	4.08 MiB	115165	10
GPU	4096	64	20.767 ms	21.059 ms	21.390 ms	24.283 ms	13.06 MiB	353213	10
GPU	4096	128	90.616 ms	110.673 ms	114.507 ms	170.346 ms	83.72 MiB	2211802	10
GPU	16384	32	12.400 ms	12.531 ms	12.670 ms	13.325 ms	4.08 MiB	115167	10
GPU	16384	64	21.132 ms	21.983 ms	22.558 ms	25.184 ms	13.06 MiB	353215	10
GPU	16384	128	90.849 ms	110.157 ms	114.308 ms	169.462 ms	83.72 MiB	2211804	10
GPU	32768	32	13.077 ms	14.088 ms	13.939 ms	14.860 ms	4.08 MiB	115170	10
GPU	32768	64	21.697 ms	22.645 ms	22.986 ms	25.235 ms	13.06 MiB	353217	10
GPU	32768	128	92.342 ms	110.662 ms	112.529 ms	169.928 ms	83.72 MiB	2211807	10

PlanktonIndividuals.jl now can also run on Apple M-series GPU. Below is a similar benchmark on Apple CPU and GPU.

PlanktonIndividuals v0.7.5
Julia Version 1.11.7

macOS 26.0.0, Darwin 25.0.0

Toolchain:
- Julia: 1.11.7
- LLVM: 16.0.6

Julia packages: 
- Metal.jl: 1.8.0
- GPUArrays: 11.2.3
- GPUCompiler: 1.6.1
- KernelAbstractions: 0.9.38
- ObjectiveC: 3.4.2
- LLVM: 9.4.2
- LLVMDowngrader_jll: 0.6.0+1

1 device:
- Apple M4 Pro 20 GPU cores (64 GiB Unified Memory)

Arch	N	min	median	mean	max	memory	allocs	samples
CPU	1024	718.333 μs	771.458 μs	798.633 μs	1.006 ms	467.94 KiB	3679	10
CPU	4096	2.447 ms	2.518 ms	2.565 ms	2.944 ms	636.72 KiB	3679	10
CPU	16384	9.089 ms	9.216 ms	9.226 ms	9.497 ms	1.28 MiB	3680	10
CPU	65536	37.136 ms	37.409 ms	37.412 ms	37.770 ms	3.92 MiB	3601	10
GPU	1024	70.685 ms	72.641 ms	73.513 ms	79.681 ms	3.47 MiB	115344	10
GPU	4096	77.897 ms	81.464 ms	81.831 ms	89.131 ms	3.48 MiB	115532	10
GPU	16384	80.004 ms	81.610 ms	81.961 ms	88.787 ms	3.47 MiB	115508	10
GPU	65536	81.589 ms	81.993 ms	82.922 ms	91.138 ms	3.48 MiB	115611	10

Arch	N	Ns	min	median	mean	max	memory	allocs	samples
CPU	1024	32	1.650 ms	1.727 ms	1.746 ms	2.009 ms	2.20 MiB	4323	10
CPU	1024	64	4.302 ms	4.335 ms	4.428 ms	4.902 ms	7.53 MiB	4963	10
CPU	1024	128	14.636 ms	14.975 ms	15.051 ms	15.846 ms	21.81 MiB	6164	10
CPU	1024	256	58.705 ms	59.051 ms	59.212 ms	61.558 ms	81.09 MiB	8725	10
CPU	4096	32	3.334 ms	3.425 ms	3.462 ms	3.846 ms	2.36 MiB	4323	10
CPU	4096	64	6.199 ms	6.277 ms	6.349 ms	6.809 ms	7.70 MiB	4963	10
CPU	4096	128	16.575 ms	16.851 ms	16.908 ms	17.764 ms	21.98 MiB	6164	10
CPU	4096	256	60.498 ms	61.014 ms	61.461 ms	65.441 ms	81.26 MiB	8725	10
CPU	16384	32	9.935 ms	10.054 ms	10.088 ms	10.442 ms	3.02 MiB	4245	10
CPU	16384	64	15.174 ms	15.359 ms	15.371 ms	15.721 ms	8.36 MiB	4885	10
CPU	16384	128	26.593 ms	26.860 ms	26.911 ms	27.408 ms	22.64 MiB	6165	10
CPU	16384	256	71.047 ms	71.621 ms	71.659 ms	72.294 ms	81.92 MiB	8726	10
CPU	65536	32	38.325 ms	38.712 ms	38.760 ms	39.282 ms	5.66 MiB	4245	10
CPU	65536	64	49.727 ms	50.438 ms	50.447 ms	50.888 ms	10.99 MiB	4885	10
CPU	65536	128	64.143 ms	64.990 ms	64.915 ms	65.964 ms	25.28 MiB	6165	10
CPU	65536	256	112.613 ms	113.154 ms	113.323 ms	114.930 ms	84.55 MiB	8726	10
GPU	1024	32	52.458 ms	53.845 ms	54.525 ms	61.694 ms	3.58 MiB	118477	10
GPU	1024	64	54.444 ms	55.915 ms	57.301 ms	72.506 ms	3.93 MiB	130398	10
GPU	1024	128	58.973 ms	60.532 ms	62.796 ms	86.138 ms	4.85 MiB	159336	10
GPU	1024	256	75.546 ms	78.564 ms	83.052 ms	111.422 ms	7.59 MiB	241115	10
GPU	4096	32	51.781 ms	53.934 ms	55.887 ms	76.180 ms	3.58 MiB	118592	10
GPU	4096	64	55.067 ms	56.103 ms	56.670 ms	63.328 ms	3.93 MiB	130565	10
GPU	4096	128	54.806 ms	60.063 ms	60.309 ms	72.531 ms	4.85 MiB	159534	10
GPU	4096	256	78.388 ms	79.156 ms	82.094 ms	101.848 ms	7.59 MiB	241281	10
GPU	16384	32	50.982 ms	53.887 ms	54.549 ms	61.776 ms	3.58 MiB	118608	10
GPU	16384	64	55.210 ms	56.055 ms	58.040 ms	77.365 ms	3.93 MiB	130575	10
GPU	16384	128	59.182 ms	60.448 ms	62.112 ms	78.982 ms	4.85 MiB	159542	10
GPU	16384	256	76.740 ms	78.906 ms	81.357 ms	105.150 ms	7.59 MiB	241292	10
GPU	65536	32	52.495 ms	53.323 ms	54.469 ms	63.653 ms	3.58 MiB	118715	10
GPU	65536	64	54.705 ms	56.284 ms	57.580 ms	70.857 ms	3.93 MiB	130660	10
GPU	65536	128	54.231 ms	56.561 ms	57.933 ms	70.151 ms	4.85 MiB	159663	10
GPU	65536	256	76.922 ms	79.406 ms	82.436 ms	110.916 ms	7.60 MiB	241376	10

Arch	N	Ns	min	median	mean	max	memory	allocs	samples
CPU	1024	32	13.701 ms	14.026 ms	14.027 ms	14.352 ms	1.07 MiB	4311	10
CPU	1024	64	104.133 ms	104.601 ms	104.670 ms	105.534 ms	5.55 MiB	8023	10
CPU	1024	128	828.092 ms	829.160 ms	828.977 ms	829.878 ms	43.23 MiB	21591	7
CPU	1024	256	6.707 s	6.707 s	6.707 s	6.707 s	322.40 MiB	73303	1
CPU	4096	32	15.903 ms	16.124 ms	16.141 ms	16.600 ms	1.24 MiB	4311	10
CPU	4096	64	106.679 ms	107.173 ms	107.316 ms	108.028 ms	5.72 MiB	8023	10
CPU	4096	128	820.950 ms	821.295 ms	821.341 ms	821.975 ms	41.12 MiB	21591	7
CPU	4096	256	6.801 s	6.801 s	6.801 s	6.801 s	322.56 MiB	73303	1
CPU	16384	32	25.392 ms	25.647 ms	25.625 ms	25.994 ms	1.90 MiB	4312	10
CPU	16384	64	117.064 ms	117.777 ms	117.660 ms	118.296 ms	6.38 MiB	8024	10
CPU	16384	128	833.843 ms	835.038 ms	834.786 ms	835.743 ms	41.77 MiB	21592	6
CPU	16384	256	6.712 s	6.712 s	6.712 s	6.712 s	323.22 MiB	73304	1
CPU	65536	32	65.352 ms	66.282 ms	66.218 ms	66.924 ms	4.54 MiB	4312	10
CPU	65536	64	165.360 ms	166.149 ms	166.157 ms	167.143 ms	9.02 MiB	8024	10
CPU	65536	128	888.348 ms	889.139 ms	889.124 ms	890.255 ms	44.41 MiB	21592	6
CPU	65536	256	6.776 s	6.776 s	6.776 s	6.776 s	325.86 MiB	73304	1
GPU	1024	32	21.830 ms	22.871 ms	23.228 ms	28.077 ms	4.56 MiB	140410	10
GPU	1024	64	29.849 ms	30.973 ms	32.603 ms	48.207 ms	13.65 MiB	383611	10
GPU	1024	128	119.130 ms	124.133 ms	126.141 ms	150.394 ms	87.53 MiB	2247365	10
GPU	1024	256	932.425 ms	962.754 ms	966.198 ms	1.023 s	658.84 MiB	17009358	6
GPU	4096	32	21.630 ms	22.592 ms	22.975 ms	26.425 ms	4.57 MiB	140587	10
GPU	4096	64	29.367 ms	29.591 ms	32.406 ms	54.415 ms	13.66 MiB	383807	10
GPU	4096	128	118.579 ms	125.494 ms	127.510 ms	154.465 ms	86.75 MiB	2247534	10
GPU	4096	256	915.231 ms	950.850 ms	949.480 ms	969.102 ms	658.84 MiB	17009532	6
GPU	16384	32	21.454 ms	22.460 ms	22.777 ms	26.804 ms	4.57 MiB	140601	10
GPU	16384	64	27.957 ms	29.587 ms	31.845 ms	53.485 ms	13.65 MiB	383729	10
GPU	16384	128	119.955 ms	124.230 ms	127.157 ms	160.412 ms	85.97 MiB	2247538	10
GPU	16384	256	904.653 ms	963.670 ms	948.708 ms	966.893 ms	658.84 MiB	17009526	6
GPU	65536	32	20.600 ms	23.022 ms	23.021 ms	26.879 ms	4.57 MiB	140708	10
GPU	65536	64	26.643 ms	29.079 ms	30.251 ms	44.419 ms	13.66 MiB	383811	10
GPU	65536	128	120.069 ms	121.475 ms	126.132 ms	163.317 ms	87.53 MiB	2247622	10
GPU	65536	256	894.671 ms	943.257 ms	938.207 ms	962.530 ms	658.84 MiB	17009612	6