Support subcycling, add metadata in result csv.

partitioned-heat-conduction/doConvergenceStudy.py

@@ -29,9 +29,12 @@ def render(precice_config_params):
         file.write(precice_config_template.render(precice_config_params))
 
 
-def do_run(dt, error_tol=10e-3, precice_config_params=default_precice_config_params):
+def do_run(dt, n_substeps = 1, error_tol=10e-3, precice_config_params=default_precice_config_params):
+    time_window_size = dt
+    time_step_size = time_window_size / n_substeps
+
     fenics = Path(__file__).parent.absolute() / "fenics"
-    precice_config_params['time_window_size'] = dt
+    precice_config_params['time_window_size'] = time_window_size
     render(precice_config_params)
     print(f"{datetime.datetime.now()}: Start run with parameters {precice_config_params}")
     print("Running...")
@@ -51,21 +54,22 @@ def do_run(dt, error_tol=10e-3, precice_config_params=default_precice_config_par
 
     for participant in participants:
         with open(fenics / participant['logfile'], "w") as outfile:
-            p = subprocess.Popen(["python3", fenics / "heat.py", participant["cmd"], f"-e {error_tol}"], cwd=fenics, stdout=outfile)
+            p = subprocess.Popen(["python3", fenics / "heat.py", participant["cmd"], f"-e {error_tol}", f"-s {n_substeps}"], cwd=fenics, stdout=outfile)
             participant["proc"] = p
 
     for participant in participants:
         participant["proc"].wait()
 
     for participant in participants:
         if participant["proc"].returncode != 0:
-            raise Exception(f'Experiment with dt={dt} failed. See logs {[p["logfile"] for p in participants]}')
+            raise Exception(f'Experiment failed. See logs {[p["logfile"] for p in participants]}')
 
     print("Done.")
     print("Postprocessing...")
-    summary = {"dt":dt}
+    summary = {"time window size": time_window_size}
     for participant in participants:
         df = pd.read_csv(fenics / f"errors-{participant['name']}.csv", comment="#")
+        summary[f"time step size {participant['name']}"] = time_step_size
         summary[f"error {participant['name']}"] = df["errors"].abs().max()
     print("Done.")
 
@@ -76,30 +80,52 @@ def do_run(dt, error_tol=10e-3, precice_config_params=default_precice_config_par
     min_dt = 0.1
     dts = [min_dt * 0.5**i for i in range(5)]
 
-    df = pd.DataFrame(columns=["dt", "error Dirichlet", "error Neumann"])
+    df = pd.DataFrame()
 
     precice_config_params = {
         'max_used_iterations': 10,
         'time_windows_reused': 5,
     }
 
-    summary_file = f"convergence-studies/{uuid.uuid4()}.csv"
+    summary_file = Path("convergence-studies") / f"{uuid.uuid4()}.csv"
 
     for dt in dts:
-        summary = do_run(dt, error_tol=10e10, precice_config_params=precice_config_params)
-        df = pd.concat([df, pd.DataFrame(summary, index=[0])], ignore_index=True)
+        for n in [1]:
+            summary = do_run(dt, n_substeps=n, error_tol=10e10, precice_config_params=precice_config_params)
+            df = pd.concat([df, pd.DataFrame(summary, index=[0])], ignore_index=True)
 
-        print(f"Write preliminary output to {summary_file}")
-        df.to_csv(summary_file)
+            print(f"Write preliminary output to {summary_file}")
+            df.to_csv(summary_file)
 
-        term_size = os.get_terminal_size()
-        print('-' * term_size.columns)
-        print(df)
-        print('-' * term_size.columns)
+            term_size = os.get_terminal_size()
+            print('-' * term_size.columns)
+            print(df)
+            print('-' * term_size.columns)
 
-    df = df.set_index('dt')
+    df = df.set_index(['time window size', 'time step size Dirichlet', 'time step size Neumann'])
     print(f"Write final output to {summary_file}")
-    df.to_csv(summary_file)
+
+    import git
+
+    repo_base = Path(__file__).parent / ".."
+    repo = git.Repo(repo_base)
+    chash = str(repo.head.commit)[:7]
+    if repo.is_dirty():
+        chash += "-dirty"
+
+    metadata={
+        "git repository": repo.remotes.origin.url,
+        "git commit": chash,
+        "precice_config_params": precice_config_params,
+    }
+
+    summary_file.unlink()
+
+    with open(summary_file, 'a') as f:
+        for key, value in metadata.items():
+            f.write(f"# {key}:{value}\n")
+        df.to_csv(f)
+
     print('-' * term_size.columns)
     print(df)
     print('-' * term_size.columns)

partitioned-heat-conduction/precice-config-template.xml

@@ -67,8 +67,8 @@
       <acceleration:IQN-ILS>
         <data name="Temperature" mesh="Neumann-Mesh" />
         <initial-relaxation value="0.1" />
+        <!-- Quasi-Newton can run into errors for small time steps. Commenting out the PRECICE_CHECK can help. See https://github.com/precice/tutorials/issues/385#issuecomment-1779178024 -->
         <max-used-iterations value="{{max_used_iterations}}" />
-        <!-- Don't reuse windows, for g_poly because this can lead to failures, especially for small time steps. Reuse for g_tri -->
         <time-windows-reused value="{{time_windows_reused}}" />
         <filter type="QR2" limit="1e-3" />
       </acceleration:IQN-ILS>