Add more profiling regions

doc/generate_profiling_figures.sh

@@ -30,8 +30,7 @@ if needle not in content:
 content = content.replace(
     needle,
     "env = EnvironmentOptions(\n"
-    "    profiling_activated=True,\n"
-    "    profiling_trace=True,\n"
+    "    deactivate_profiling=False,\n"
     ")",
 )
 open(path, "w").write(content)

doc/sections/userguide.rst

@@ -780,13 +780,52 @@ Struphy's simulation-wide profiler is configured in
 The relevant switches live in :class:`~struphy.EnvironmentOptions`:
 
 1. ``profiling_activated=True`` enables profiling data collection.
-2. ``profiling_trace=True`` additionally records a time trace of profiling
-   regions.
 
 The profiler is set up automatically in ``Simulation.__init__()`` and finalized
 when ``Simulation.run()`` finishes. The simulation code already wraps key work
-inside regions such as ``model.integrate`` via
-``ProfileManager.profile_region(...)``.
+inside regions via ``ProfileManager.profile_region(...)``. Since the profiler is
+active from the end of ``Simulation.__init__()``, the setup phase is covered as
+well, and the following regions are recorded out of the box:
+
+1. Setup: ``setup: allocate`` (total allocation time), with the nested regions
+   ``setup: feec`` (``setup: derham``, ``setup: mass ops``, ``setup: basis ops``,
+   ``setup: projected equil``), ``setup: variables`` (one
+   ``setup var: <species>.<variable>`` region per model variable, so that e.g.
+   marker drawing shows up per particle species), ``setup: propagators`` (one
+   ``setup prop: <PropagatorName>`` region per propagator) and
+   ``setup: helpers``.
+2. Remaining run preparation: ``setup: run metadata``, ``setup: data storage``,
+   ``setup: geometry vtk``, ``setup: plasma params``,
+   ``setup: initial diagnostics``, ``setup: hdf5 datasets`` and, for restarted
+   runs, ``setup: restart``.
+3. Time loop: ``model.integrate``, ``diagnostics``, ``save data`` and
+   ``sort particles``.
+
+Inside ``model.integrate`` the regions nest as follows:
+
+1. ``prop: <PropagatorName>``, one per propagator call (twice per step for the
+   half steps of Strang splitting).
+2. Particle pushing: ``pusher: <kernel_name>`` for a full
+   :class:`~struphy.pic.pushing.pusher.Pusher` call, containing one
+   ``kernel: <kernel_name>`` region per pusher, init and eval kernel call.
+3. Accumulation: ``accum: <kernel_name>`` for a full
+   :class:`~struphy.pic.accumulation.particles_to_grid.Accumulator` call,
+   containing the ``kernel: <kernel_name>`` region of the accumulation kernel
+   and ``accum comm: <kernel_name>`` for the assembly/ghost-region exchange and
+   the inter-clone ``Allreduce``.
+4. Particle bookkeeping and communication, recorded wherever they are called
+   from: ``mpi_sort_markers``, ``apply_kinetic_bc``, ``put_particles_in_boxes``
+   and ``do_sort``.
+5. Linear solves: ``solve: SchurSolver``, ``solve: SchurSolverFull``,
+   ``solve: SchurSolverFull3``, ``solve: SaddlePointSolver``,
+   ``solve: ODEsolverFEEC`` for the shared solver classes, and
+   ``solve: <PropagatorName>`` for propagators that call a
+   ``feectools`` inverse operator directly.
+6. ``update_feec_variables`` for writing back FEEC coefficients (includes the
+   ghost-region update).
+
+Since regions nest, the sum over all regions exceeds the wall-clock time; use
+the flame graph (below) to read the containment.
 
 Example configuration:
 

examples/DriftKineticElectrostaticAdiabatic/cyclone/params_cyclone.py

@@ -76,7 +76,7 @@
 # --------------------------
 
 # Environment options
-env = EnvironmentOptions(sim_folder="sim_1",profiling_activated=True, profiling_trace=True, restart=False)
+env = EnvironmentOptions(sim_folder="sim_1",profiling_activated=True, restart=False)
 
 # Time stepping
 time_opts = Time(dt=0.001, Tend=0.01, split_algo="LieTrotter")

examples/DriftKineticElectrostaticAdiabatic/itg_cylindre/params_drift_kinetic.py

@@ -78,7 +78,7 @@
 # --------------------------
 
 # Environment options
-env = EnvironmentOptions(sim_folder="sim_1", profiling_activated=True, profiling_trace=True, restart=False)
+env = EnvironmentOptions(sim_folder="sim_1", profiling_activated=True, restart=False)
 
 # Time stepping
 time_opts = Time(dt=5.0, Tend=500.0, split_algo="LieTrotter")

examples/ToyGyrokinetic/diocotron_instability/params_diocotron.py

@@ -74,7 +74,7 @@
 # --------------------------
 
 # Environment options
-env = EnvironmentOptions(sim_folder="sim_1", profiling_activated=True, profiling_trace=True, restart=False)
+env = EnvironmentOptions(sim_folder="sim_1", profiling_activated=True, restart=False)
 
 # Time stepping
 time_opts = Time(dt=0.01, Tend=51.0, split_algo="LieTrotter")

profiling/examples/Poisson/cube_strong_scaling/params_poisson.py

@@ -65,7 +65,6 @@
 env = EnvironmentOptions(
     sim_folder=f"sim_{args.id:02d}",
     profiling_activated=True,
-    profiling_trace=True,
     restart=False,
 )
 

profiling/examples/ToyGyrokinetic/diocotron_instability/params_diocotron.py

@@ -85,7 +85,6 @@
 env = EnvironmentOptions(
     sim_folder=f"sim_{args.id:02d}",
     profiling_activated=True,
-    profiling_trace=True,
     restart=False
 )
 

pyproject.toml

@@ -47,7 +47,7 @@ dependencies = [
     "pytest-testmon<=2.2.0",
     "ruff==0.15.0, <=0.16.0",
     "line_profiler<=5.0.2",
-    "scope-profiler==0.2.6, <=0.2.6",
+    "scope-profiler<=0.2.8",
 ]
 
 [project.license]

src/struphy/io/options.py

@@ -315,6 +315,9 @@ class EnvironmentOptions(OptionsBase):
         Folder in ``out_folders/`` for the current simulation (default= ``sim_1/`` ).
         Will create the folder if it does not exist OR cleans the folder for new runs.
 
+    sim_label: str | None, optional
+        Label for the simulation (default=None)
+
     restart : bool
         Whether to restart a run (default=False).
 
@@ -332,20 +335,17 @@ class EnvironmentOptions(OptionsBase):
 
     profiling_activated: bool, optional
         Activate profiling with scope-profiler (default=False)
-
-    profiling_trace: bool, optional
-        Save time-trace of each profiling region (default=False)
     """
 
     out_folders: str = os.getcwd()
     sim_folder: str = "sim_1"
+    sim_label: str | None = None
     restart: bool = False
     max_runtime: int = 300
     save_step: int = 1
     sort_step: int = 0
     num_clones: int = 1
     profiling_activated: bool = False
-    profiling_trace: bool = False
 
     def __post_init__(self):
         self.path_out: str = os.path.join(self.out_folders, self.sim_folder)

src/struphy/linear_algebra/saddle_point.py

@@ -7,6 +7,7 @@
 from feectools.linalg.block import BlockLinearOperator, BlockVector, BlockVectorSpace
 from feectools.linalg.direct_solvers import SparseSolver
 from feectools.linalg.solvers import inverse
+from scope_profiler import ProfileManager
 
 from struphy.linear_algebra.tests.test_saddlepoint_massmatrices import _plot_residual_norms
 
@@ -249,6 +250,7 @@ def Apre(self, a):
         elif self._variant == "Inverse_Solver":
             self._Apre = a
 
+    @ProfileManager.profile("solve: SaddlePointSolver")
     def __call__(self, U_init=None, Ue_init=None, P_init=None, out=None):
         """
         Solves the saddle-point problem using the Uzawa algorithm.

src/struphy/linear_algebra/schur_solver.py

@@ -2,6 +2,7 @@
 from feectools.linalg.block import BlockLinearOperator, BlockVector
 from feectools.linalg.solvers import inverse
 from line_profiler import profile
+from scope_profiler import ProfileManager
 
 from struphy.linear_algebra.solver import SolverParameters
 
@@ -108,6 +109,7 @@ def BC(self, bc):
         self._BC = bc
 
     @profile
+    @ProfileManager.profile("solve: SchurSolver")
     def __call__(self, xn, Byn, dt, out=None):
         """Solves the 2x2 block matrix linear system.
 
@@ -228,6 +230,7 @@ def __init__(self, M, solver_name, **solver_params):
         self._rhs = self._A.codomain.zeros()
 
     @profile
+    @ProfileManager.profile("solve: SchurSolverFull")
     def dot(self, v, out=None):
         """Solves the 2x2 block matrix linear system.
 
@@ -346,6 +349,7 @@ def __init__(self, M, solver_name, **solver_params):
         self._rhs2 = self._A.codomain.zeros()
 
     @profile
+    @ProfileManager.profile("solve: SchurSolverFull3")
     def dot(self, v, out=None):
         """Solves the 3x3 block matrix linear system.
 

src/struphy/models/tests/utils_testing.py

@@ -55,7 +55,6 @@ def call_test(model: StruphyModel, test_profiling: bool = False):
         out_folders=test_folder,
         sim_folder=f"{model_name}",
         profiling_activated=test_profiling,
-        profiling_trace=test_profiling,
     )
 
     # read parameters

src/struphy/ode/solvers.py

@@ -3,6 +3,7 @@
 import cunumpy as xp
 from feectools.linalg.block import BlockVector
 from feectools.linalg.stencil import StencilVector
+from scope_profiler import ProfileManager
 
 from struphy.ode.utils import ButcherTableau
 
@@ -59,6 +60,7 @@ def __init__(
         self._yn = [v.copy() for v in self.y]
         self._ystar = [v.copy() for v in self.y]
 
+    @ProfileManager.profile("solve: ODEsolverFEEC")
     def __call__(self, tn, h):
         a = self.butcher.a
         b = self.butcher.b