Particle Container Management (MC)

The following kernels manage the particle container, including initialization, repartitioning, merging, and defragmentation.

Type	Name	Policy	Brief Description	Number of Calls
reduction	mc_init_first	range(size)	Spawns particles by setting positions, calling model initialization, and calculating total mass.	1
for	get_repartition	range(size)	Counts the number of particles per compartment (if multiple compartments exist).	n_export
for	insert_merge	TeamPolicy(n_add_item, Kokkos::AUTO, Model::n_var)	Back-inserts to merge the main container and buffer.	n_step
scan	find_and_fill_gap	range(size)	Finds non-idle particles and replaces them with new ones (defragmentation).	If n_non_idle > threshold

Iteration Cycle Kernels

Each iteration cycle launches the following kernels:

Follow "team" policy is actually a way to deal with continous chunk of particle per team, the size of chunk is defined at compile time depending on the selected backend but can be overwritten with env variable.

Type	Name	Policy	Brief Description	Number of Calls
for	cycle_move	team	Moves particles based on the flowmap (if n_compartment > 1).	n_step
reduce	cycle_move_leave	range:	Returns the number of particles leaving (if continuous reactor with feed != 0).	n_step
reduce	cycle_model	team	Updates the model, handles division, and returns the number of particles leaving and waiting for allocation.	n_step
reduce	cycle_model_contribs	team	Scatters particle contributions (general case)	n_step
reduce	cycle_model_contribs_0d	team	Scatters particle contributions (1D only)	n_step

Data Export Kernels

The following kernel is used for exporting data:

Type	Name	Policy	Brief Description	Number of Calls
for	get_properties	range(size)	Deep copies the particle container to the host in a view format for HDF5 export. Skips non-idle particles.	n_export