Added abstraction layer for boundary condition and the capability to add profiles to boundary conditions #86

mehdiataei · 2024-10-31T16:18:43Z

Contributing Guidelines

I have read and understood the CONTRIBUTING.md guidelines

Description

Added abstraction layer for boundary condition efficient encoding/decoding of auxiliary data
Added the capability to add profiles to boundary conditions
Fixes Non-uniform boundary velocity profile #74

Type of change

Bug fix (non-breaking change which fixes an issue)
New feature (non-breaking change which adds functionality)
Breaking change (fix or feature that would cause existing functionality to not work as expected)
Documentation update

How Has This Been Tested?

All pytest tests pass

Linting and Code Formatting

Make sure the code follows the project's linting and formatting standards. This project uses Ruff for linting.

To run Ruff, execute the following command from the root of the repository:

ruff check .

Ruff passes

…add profiles to boundary conditions

hsalehipour

Thanks for the PR. This is very helpful. I have made some comments which I think will improve this work.

hsalehipour · 2024-11-01T21:06:15Z

examples/cfd/flow_past_sphere_3d.py

@@ -69,7 +72,7 @@ def define_boundary_indices(self):

    def setup_boundary_conditions(self):
        inlet, outlet, walls, sphere = self.define_boundary_indices()
-        bc_left = RegularizedBC("velocity", (0.04, 0.0, 0.0), indices=inlet)


Can we have a 1-to-1 association between a bc_profile function and a bc object ? Right now, it is not straightforward for example to have two BC's with different profiles. If we add functional to the BC constructor then we would need another operator (like boundary masker) which goes over all BC's and initializes them similar to what initialize_bc_aux_data does right now. This could be done as the first step of the stepper (if iteration is zero). This way we can also remove initilize_bc_aux_data from examples.

Also with these changes, it is kind of difficult to assign constant values to a BC. Is there a simpler way to pass in constant values to BCs without having to define a bc_prfile with jax and warp implementations?

hsalehipour · 2024-11-01T21:07:04Z

examples/cfd/flow_past_sphere_3d.py

    def run(self, num_steps, post_process_interval=100):
        start_time = time.time()
+        self.initialize_bc_aux_data()


I think with some minor changes you will be able to make this a stand alone operator (like boundary_masker).

We can either call that operator inside the stepper stepper (if iteration=0) or leave it to the user to call it in the example.

hsalehipour · 2024-11-01T21:08:19Z

examples/cfd/windtunnel_3d.py

@@ -95,7 +95,6 @@ def define_boundary_indices(self):
    def setup_boundary_conditions(self):
        inlet, outlet, walls, car = self.define_boundary_indices()
        bc_left = EquilibriumBC(rho=1.0, u=(self.wind_speed, 0.0, 0.0), indices=inlet)
-        # bc_left = RegularizedBC('velocity', (self.wind_speed, 0.0, 0.0), indices=inlet)


It would be really nice if we could define BCs like before (simply as constant inputs) as well as with a function as an input to the bc constructor: eg u = (0.04, 0, 0) or u = bc_prfile(). Something like this. I think it could be done in the operator I am proposing for initializing BC aux data.

hsalehipour · 2024-11-01T21:12:41Z

xlb/operator/boundary_condition/bc_regularized.py

@@ -222,6 +216,15 @@ def functional_velocity(
            # Find normal vector
            normals = get_normal_vectors(missing_mask)

+            # Find the value of u from the missing directions


Is there not a performance hit here? The BC is not dynamic like Extrapolation Outflow where its information change in time. Here we have the same value of velocity or pressure being recomputed at each iteration over and over again.

We're not recomputing here we're simply reading it. I didn't notice any performance hit before and after the PR.

An improvement for this could be that we can store the missing direction and avoid the for loop for slightly improved performance. I'll try to do that.

(this only works for linear/planar BCs like zouhe)

hsalehipour · 2024-11-01T21:18:58Z

xlb/operator/boundary_condition/bc_zouhe.py

@@ -158,32 +181,29 @@ def bounceback_nonequilibrium(self, fpop, feq, missing_mask):

    @Operator.register_backend(ComputeBackend.JAX)
    @partial(jit, static_argnums=(0))
-    def jax_implementation(self, f_pre, f_post, bc_mask, missing_mask):
+    def jax_implementation(self, f_0, f_1, bc_mask, missing_mask):


Please make sure this is consistent naming for all other jax implementations of all BCs.

hsalehipour · 2024-11-01T22:15:05Z

xlb/operator/boundary_condition/bc_zouhe.py

+        # Get the density and velocity from the f_1
+        if self.bc_type == "velocity":
+            vel = self.prescribed_values
+        elif self.bc_type == "pressure":
+            rho = self.prescribed_values
+            vel = self.calculate_vel(f_1, rho, missing_mask)
+        else:
+            raise ValueError(f"type = {self.bc_type} not supported! Use 'pressure' or 'velocity'.")


this is redundant.

hsalehipour · 2024-11-04T21:10:24Z

examples/cfd/windtunnel_3d.py

@@ -95,7 +95,6 @@ def define_boundary_indices(self):
    def setup_boundary_conditions(self):
        inlet, outlet, walls, car = self.define_boundary_indices()
        bc_left = EquilibriumBC(rho=1.0, u=(self.wind_speed, 0.0, 0.0), indices=inlet)


what if velocity here is also not constant and has a profile. All these boundary conditions should be set up similarly. It would be nice if we could input the profile to the BC constrcutor.

hsalehipour · 2024-11-04T21:14:29Z

examples/cfd/flow_past_sphere_3d.py

@@ -69,7 +72,7 @@ def define_boundary_indices(self):

    def setup_boundary_conditions(self):
        inlet, outlet, walls, sphere = self.define_boundary_indices()
-        bc_left = RegularizedBC("velocity", (0.04, 0.0, 0.0), indices=inlet)
+        bc_left = RegularizedBC("velocity", indices=inlet)
        # bc_left = EquilibriumBC(rho = 1, u=(0.04, 0.0, 0.0), indices=inlet)


Equilibrium BC should also be able to accept variable values. Can you add that as well ?

hsalehipour · 2024-11-04T21:17:19Z

examples/cfd/lid_driven_cavity_2d.py

@@ -49,7 +50,7 @@ def define_boundary_indices(self):

    def setup_boundary_conditions(self):
        lid, walls = self.define_boundary_indices()
-        bc_top = EquilibriumBC(rho=1.0, u=(0.02, 0.0), indices=lid)
+        bc_top = EquilibriumBC(rho=1.0, u=(self.prescribed_vel, 0.0), indices=lid)


Same comment for EquilibriumBC in all examples.

hsalehipour · 2024-11-04T21:20:47Z

xlb/operator/boundary_condition/bc_extrapolation_outflow.py

        # find and store the normal vector using indices
        self._get_normal_vec(indices)

        # Unpack the two warp functionals needed for this BC!
        if self.compute_backend == ComputeBackend.WARP:
-            self.warp_functional, self.prepare_bc_auxilary_data = self.warp_functional
+            self.warp_functional, self.update_bc_auxilary_data = self.warp_functional


I like this renaming.

mehdiataei · 2024-11-04T23:46:08Z

Thanks for the PR. This is very helpful. I have made some comments which I think will improve this work.

Thanks for the comments. I'll take a look.

mehdiataei marked this pull request as ready for review October 31, 2024 16:56

mehdiataei requested review from massimim and hsalehipour October 31, 2024 17:34

mehdiataei changed the title ~~Profile inlet~~ Added abstraction layer for boundary condition and the capability to add profiles to boundary conditions Oct 31, 2024

mehdiataei force-pushed the profile_inlet branch 5 times, most recently from fb3fdb1 to f0f81c3 Compare November 4, 2024 01:39

Added abstraction layer for boundary condition and the capability to …

d07713d

…add profiles to boundary conditions

mehdiataei force-pushed the profile_inlet branch from f0f81c3 to d07713d Compare November 4, 2024 01:39

hsalehipour requested changes Nov 4, 2024

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Added abstraction layer for boundary condition and the capability to add profiles to boundary conditions #86

Added abstraction layer for boundary condition and the capability to add profiles to boundary conditions #86

mehdiataei commented Oct 31, 2024 •

edited

Loading

hsalehipour left a comment •

edited

Loading

hsalehipour Nov 1, 2024

hsalehipour Nov 1, 2024

hsalehipour Nov 1, 2024

hsalehipour Nov 1, 2024

mehdiataei Nov 4, 2024

mehdiataei Nov 4, 2024

hsalehipour Nov 1, 2024

hsalehipour Nov 1, 2024

hsalehipour Nov 4, 2024

hsalehipour Nov 4, 2024

hsalehipour Nov 4, 2024

hsalehipour Nov 4, 2024

mehdiataei commented Nov 4, 2024

Added abstraction layer for boundary condition and the capability to add profiles to boundary conditions #86

Are you sure you want to change the base?

Added abstraction layer for boundary condition and the capability to add profiles to boundary conditions #86

Conversation

mehdiataei commented Oct 31, 2024 • edited Loading

Contributing Guidelines

Description

Type of change

How Has This Been Tested?

Linting and Code Formatting

hsalehipour left a comment • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mehdiataei commented Nov 4, 2024

mehdiataei commented Oct 31, 2024 •

edited

Loading

hsalehipour left a comment •

edited

Loading