Taylor green vortex in SPH#

This simple example shows a SPH simulation of a Taylor green vortex

 8 import matplotlib.pyplot as plt
 9 import numpy as np
10
11 import shamrock
12
13 # If we use the shamrock executable to run this script instead of the python interpreter,
14 # we should not initialize the system as the shamrock executable needs to handle specific MPI logic
15 if not shamrock.sys.is_initialized():
16     shamrock.change_loglevel(1)
17     shamrock.sys.init("0:0")

Use shamrock documentation style for matplotlib

21 shamrock.matplotlib.set_shamrock_mpl_style()

Setup parameters

25 gamma = 1.4
26
27 rho_g = 1
28
29 Mach = 0.1
30 P_g = (Mach**-2) * rho_g / gamma
31
32 print(f"Mach number : {1 / np.sqrt(gamma * P_g / rho_g)}")
33
34 u_g = P_g / ((gamma - 1) * rho_g)
35
36 resol_per_green = 128
37 vortex_size = 1
38
39 L_green = vortex_size / (2 * np.pi)
Mach number : 0.10000000000000002

Setup the context, model & sim

 43 ctx = shamrock.Context()
 44 ctx.pdata_layout_new()
 45
 46 model = shamrock.get_Model_SPH(context=ctx, vector_type="f64_3", sph_kernel="M6")
 47
 48 cfg = model.gen_default_config()
 49 cfg.set_artif_viscosity_VaryingCD10(
 50     alpha_min=0.0, alpha_max=1, sigma_decay=0.1, alpha_u=1, beta_AV=2
 51 )
 52 cfg.set_boundary_periodic()
 53 cfg.set_eos_adiabatic(gamma)
 54 # Set the CFL
 55 cfg.set_cfl_cour(0.1)
 56 cfg.set_cfl_force(0.1)
 57
 58 # Set the solver config to be the one stored in cfg
 59 model.set_solver_config(cfg)
 60
 61 # Print the solver config
 62 model.get_current_config().print_status()
 63
 64 # We want the patches to split above 10^8 part and merge if smaller than 1 part (e.g. disable patch)
 65 model.init_scheduler(int(1e8), 1)
 66
 67
 68 resol = resol_per_green
 69 (xs, ys, zs) = model.get_box_dim_fcc_3d(1, resol, resol, resol)
 70 dr = 1 / xs
 71
 72 print("Box size : ", xs, ys, zs)
 73 (xs, ys, zs) = (1.0, 1.0, dr * 12)
 74
 75 model.resize_simulation_box((-xs / 2, -ys / 2, -zs / 2), (xs / 2, ys / 2, zs / 2))
 76
 77
 78 setup = model.get_setup()
 79 gen1 = setup.make_generator_lattice_hcp(dr, (-xs / 2, -ys / 2, -zs / 2), (xs / 2, ys / 2, zs / 2))
 80 setup.apply_setup(gen1)
 81
 82 model.set_value_in_a_box("uint", "f64", u_g, (-xs / 2, -ys / 2, -zs / 2), (xs / 2, ys / 2, zs / 2))
 83
 84
 85 def vel_func(r):
 86     x, y, z = r
 87
 88     x += L_green * np.pi
 89
 90     vx = np.cos(x / L_green) * np.sin(y / L_green)
 91     vy = -np.sin(x / L_green) * np.cos(y / L_green)
 92     vz = 0
 93
 94     return (vx, vy, vz)
 95
 96
 97 model.set_field_value_lambda_f64_3("vxyz", vel_func)
 98
 99
100 vol_b = xs * ys * zs
101
102 totmass = rho_g * vol_b
103
104 print("Total mass :", totmass)
105
106
107 pmass = model.total_mass_to_part_mass(totmass)
108 model.set_particle_mass(pmass)
109 print("Current part mass :", pmass)
110
111
112 dump_folder = "_to_trash"
113 import os
114
115 os.system("mkdir -p " + dump_folder)
116
117 current_fig = None
118
119 cnt_plot = 0
----- SPH Solver configuration -----
[
    {
        "artif_viscosity": {
            "alpha_max": 1.0,
            "alpha_min": 0.0,
            "alpha_u": 1.0,
            "beta_AV": 2.0,
            "sigma_decay": 0.1,
            "type": "varying_cd10"
        },
        "boundary_config": {
            "bc_type": "periodic"
        },
        "cfl_config": {
            "cfl_cour": 0.1,
            "cfl_force": 0.1,
            "cfl_multiplier_stiffness": 2.0,
            "eta_sink": 0.05
        },
        "combined_dtdiv_divcurlv_compute": false,
        "debug_dump_filename": "",
        "do_debug_dump": false,
        "dust_config": {
            "drag_mode": {
                "type": "none"
            },
            "mode": {
                "type": "none"
            }
        },
        "enable_particle_reordering": false,
        "eos_config": {
            "Tvec": "f64_3",
            "eos_type": "adiabatic",
            "gamma": 1.4
        },
        "epsilon_h": 1e-06,
        "ext_force_config": {
            "force_list": []
        },
        "gpart_mass": 0.0,
        "h_iter_per_subcycles": 50,
        "h_max_subcycles_count": 100,
        "htol_up_coarse_cycle": 1.1,
        "htol_up_fine_cycle": 1.1,
        "kernel_id": "M6<f64>",
        "mhd_config": {
            "mhd_type": "none"
        },
        "particle_killing": [],
        "particle_reordering_step_freq": 1000,
        "save_dt_to_fields": false,
        "scheduler_config": {
            "merge_load_value": 0,
            "split_load_value": 0
        },
        "self_grav_config": {
            "softening_length": 1e-09,
            "softening_mode": "plummer",
            "type": "none"
        },
        "show_cfl_detail": false,
        "show_ghost_zone_graph": false,
        "show_neigh_stats": false,
        "smoothing_length_config": {
            "type": "density_based"
        },
        "time_state": {
            "cfl_multiplier": 0.01,
            "dt_sph": 0.0,
            "time": 0.0
        },
        "tree_reduction_level": 3,
        "type_id": "sycl::vec<f64,3>",
        "unit_sys": null,
        "use_two_stage_search": true
    }
]
------------------------------------
Box size :  256.0 221.70250336881628 209.02312471749784
SPH setup: generating particles ...
SPH setup: Nstep = 132224 ( 1.3e+05 ) Ntotal = 132224 ( 1.3e+05 rank min = 7.5e+06 max = 1.3e+05) rate = 1.322240e+05 N.s^-1
SPH setup: the generation step took : 0.027431224 s
SPH setup: final particle count = 132224 beginning injection ...
Info: ---------------------------------------------                   [DataInserterUtility][rank=0]
Info: Compute load ...                                                [DataInserterUtility][rank=0]
Info: run scheduler step ...                                          [DataInserterUtility][rank=0]
Info: Scheduler step timings :                                                  [Scheduler][rank=0]
   metadata sync     : 8.17 us    (64.2%)
Info: Summary (strategy = round robin):                                       [LoadBalance][rank=0]
 - strategy "psweep"      : max = 132224.0 min = 132224.0 factor = 1
 - strategy "round robin" : max = 125612.8 min = 125612.8 factor = 0.95
Info: Loadbalance stats :                                                     [LoadBalance][rank=0]
    npatch = 1
    min = 132224
    max = 132224
    avg = 132224
    efficiency = 100.00%
Info: Scheduler step timings :                                                  [Scheduler][rank=0]
   metadata sync     : 912.00 ns  (0.2%)
   patch tree reduce : 1.36 us    (0.3%)
   gen split merge   : 851.00 ns  (0.2%)
   split / merge op  : 0/0
   apply split merge : 992.00 ns  (0.2%)
   LB compute        : 411.90 us  (97.5%)
   LB move op cnt    : 0
   LB apply          : 3.54 us    (0.8%)
Info: patch count stable after 1 runs npatch = 1                      [DataInserterUtility][rank=0]
Info: ---------------------------------------------                   [DataInserterUtility][rank=0]
SPH setup: injected       132224 / 132224 => 100.0% | ranks with patchs = 1 / 1  <- global loop -> (msg count : 0)
SPH setup: the injection step took : 0.018049718000000003 s
Info: injection perf report:                                                    [SPH setup][rank=0]
+======+====================+=======+=============+=============+=============+
| rank | rank get (sum/max) |  MPI  | alloc d% h% | mem (max) d | mem (max) h |
+======+====================+=======+=============+=============+=============+
| 0    |      0.00s / 0.00s | 0.00s |   0.8% 0.0% |   134.92 MB |   460.80 kB |
+------+--------------------+-------+-------------+-------------+-------------+
SPH setup: the setup took : 0.07498199500000001 s
Total mass : 0.046875
Current part mass : 3.5451204017424977e-07

Make plotting functions

125 def plot():
126     global current_fig
127     if current_fig is not None:
128         plt.close(current_fig)
129
130     pixel_x = 1080
131     pixel_y = 1080
132     radius = 0.5
133     center = (0.0, 0.0, 0.0)
134     aspect = pixel_x / pixel_y
135     pic_range = [-radius * aspect, radius * aspect, -radius, radius]
136     delta_x = (radius * 2 * aspect, 0.0, 0.0)
137     delta_y = (0.0, radius * 2, 0.0)
138
139     arr_vel = model.render_cartesian_slice(
140         "vxyz",
141         "f64_3",
142         center=(0.0, 0.0, 0.0),
143         delta_x=delta_x,
144         delta_y=delta_y,
145         nx=pixel_x,
146         ny=pixel_y,
147     )
148
149     v_norm = np.sqrt(arr_vel[:, :, 0] ** 2 + arr_vel[:, :, 1] ** 2 + arr_vel[:, :, 2] ** 2)
150
151     import copy
152
153     import matplotlib
154
155     my_cmap = copy.copy(matplotlib.colormaps.get_cmap("gist_heat"))  # copy the default cmap
156     my_cmap.set_bad(color="black")
157
158     fig_width = 6
159     fig_height = fig_width / aspect
160     current_fig = plt.figure(figsize=(fig_width, fig_height))
161     res = plt.imshow(v_norm, cmap=my_cmap, origin="lower", extent=pic_range)
162
163     cbar = plt.colorbar(res, extend="both")
164     cbar.set_label(r"$\sqrt{vx^2 + vy^2 + vz^2}$ [code unit]")
165
166     plt.title("t = {:0.3f} [code unit]".format(model.get_time()))
167     plt.xlabel("x")
168     plt.ylabel("y")
169     global cnt_plot
170     plt.savefig(dump_folder + f"/test_{cnt_plot}.png")
171     cnt_plot += 1

Run the simulation

176 model.timestep()
177
178 dt_stop = 0.001
179 for i in range(1):
180     t_target = i * dt_stop
181     # skip if the model is already past the target
182     if model.get_time() > t_target:
183         continue
184
185     model.evolve_until(i * dt_stop)
186
187     # Dump name is "dump_xxxx.sham" where xxxx is the timestep
188     model.do_vtk_dump(dump_folder + "/dump_{:04}.vtk".format(i), True)
189     plot()
t = 0.000 [code unit]
---------------- t = 0, dt = 0 ----------------
Info: Summary (strategy = round robin):                                       [LoadBalance][rank=0]
 - strategy "psweep"      : max = 132224.0 min = 132224.0 factor = 1
 - strategy "round robin" : max = 125612.8 min = 125612.8 factor = 0.95
Info: Loadbalance stats :                                                     [LoadBalance][rank=0]
    npatch = 1
    min = 132224
    max = 132224
    avg = 132224
    efficiency = 100.00%
Info: Scheduler step timings :                                                  [Scheduler][rank=0]
   metadata sync     : 21.85 us   (4.3%)
   patch tree reduce : 3.06 us    (0.6%)
   gen split merge   : 580.00 ns  (0.1%)
   split / merge op  : 0/0
   apply split merge : 1.23 us    (0.2%)
   LB compute        : 460.15 us  (90.6%)
   LB move op cnt    : 0
   LB apply          : 3.94 us    (0.8%)
Info: Scheduler step timings :                                                  [Scheduler][rank=0]
   metadata sync     : 1.81 us    (73.9%)
Warning: High interface/patch volume ratio.                                  [InterfaceGen][rank=0]
    This can lead to high mpi overhead, try to increase the patch split crit
    patch 0 high interf/patch volume: 0.6578911544046468
Warning: smoothing length is not converged, rerunning the iterator ...    [Smoothinglength][rank=0]
     largest h = 0.004296875 unconverged cnt = 132224
Warning: High interface/patch volume ratio.                                  [InterfaceGen][rank=0]
    This can lead to high mpi overhead, try to increase the patch split crit
    patch 0 high interf/patch volume: 0.6578911544046468
Warning: smoothing length is not converged, rerunning the iterator ...    [Smoothinglength][rank=0]
     largest h = 0.004726562500000001 unconverged cnt = 132224
Warning: High interface/patch volume ratio.                                  [InterfaceGen][rank=0]
    This can lead to high mpi overhead, try to increase the patch split crit
    patch 0 high interf/patch volume: 0.6638809898354309
Warning: smoothing length is not converged, rerunning the iterator ...    [Smoothinglength][rank=0]
     largest h = 0.005199218750000002 unconverged cnt = 132224
Warning: High interface/patch volume ratio.                                  [InterfaceGen][rank=0]
    This can lead to high mpi overhead, try to increase the patch split crit
    patch 0 high interf/patch volume: 0.9830817400774442
Warning: smoothing length is not converged, rerunning the iterator ...    [Smoothinglength][rank=0]
     largest h = 0.0057191406250000024 unconverged cnt = 132224
Warning: High interface/patch volume ratio.                                  [InterfaceGen][rank=0]
    This can lead to high mpi overhead, try to increase the patch split crit
    patch 0 high interf/patch volume: 0.9911286907066795
Warning: smoothing length is not converged, rerunning the iterator ...    [Smoothinglength][rank=0]
     largest h = 0.006291054687500003 unconverged cnt = 132224
Warning: High interface/patch volume ratio.                                  [InterfaceGen][rank=0]
    This can lead to high mpi overhead, try to increase the patch split crit
    patch 0 high interf/patch volume: 1.0163283518877058
Warning: smoothing length is not converged, rerunning the iterator ...    [Smoothinglength][rank=0]
     largest h = 0.006920160156250004 unconverged cnt = 132224
Warning: High interface/patch volume ratio.                                  [InterfaceGen][rank=0]
    This can lead to high mpi overhead, try to increase the patch split crit
    patch 0 high interf/patch volume: 1.0243753025169406
Warning: the unit system is not set                                           [sph::Config][rank=0]
Info: conservation infos :                                                     [sph::Model][rank=0]
    sum v = (1.9647884481053865e-19,4.773428345076549e-19,0)
    sum a = (1.367610070459655e-16,-9.401966136903069e-16,-1.780827824668828e-14)
    sum e = 8.382254464285767
    sum de = 3.6917083973131426e-16
Info: cfl dt = 6.796398329179769e-07 cfl multiplier : 0.01                     [sph::Model][rank=0]
Info: Timestep perf report:                                                    [sph::Model][rank=0]
+======+============+========+========+===========+======+=============+=============+=============+
| rank | rate (N/s) |  Nobj  | Npatch |   tstep   | MPI  | alloc d% h% | mem (max) d | mem (max) h |
+======+============+========+========+===========+======+=============+=============+=============+
| 0    | 3.8588e+04 | 132224 |      1 | 3.427e+00 | 0.0% |   0.2% 0.0% |   188.57 MB |     5.29 MB |
+------+------------+--------+--------+-----------+------+-------------+-------------+-------------+
Info: estimated rate : 0 (tsim/hr)                                             [sph::Model][rank=0]
Info: evolve_until (target_time = 0.00s, niter_max = -1, max_walltime = -1.00s)       [SPH][rank=0]
Info: iteration since start : 1                                                       [SPH][rank=0]
Info: time since start : 424.41294305900004 (s)                                       [SPH][rank=0]
Info: dump to _to_trash/dump_0000.vtk                                            [VTK Dump][rank=0]
              - took 8.13 ms, bandwidth = 1.37 GB/s
Info: compute_slice field_name: vxyz, positions count: 1166400       [sph::CartesianRender][rank=0]
Info: compute_slice took 1.95 s                                      [sph::CartesianRender][rank=0]

Total running time of the script: (0 minutes 6.899 seconds)

Estimated memory usage: 345 MB

Gallery generated by Sphinx-Gallery