cuda

codegen.cuda

CUDA Adapter for TVBO

Run generated CUDA kernels using PyCUDA.

Usage

from tvbo import SimulationExperiment

exp = SimulationExperiment.from_file(“experiment.yaml”) result = exp.run(‘cuda’)

Functions

Name	Description
compile_cuda	Compile CUDA kernel for experiment.
run_cuda	Run CUDA simulation for experiment.
save_cuda	Save CUDA source to file.

compile_cuda

codegen.cuda.compile_cuda(experiment)

Compile CUDA kernel for experiment.

Parameters

Name	Type	Description	Default
experiment	SimulationExperiment	SimulationExperiment instance	required

Returns

Name	Type	Description
	Tuple[Any, Any]	Tuple of (module, kernel_func)

run_cuda

codegen.cuda.run_cuda(
    experiment,
    n_steps=None,
    dt=None,
    n_work_items=1,
    global_speed=1.0,
    global_coupling=0.1,
    buffer_length=None,
    swept_params=None,
)

Run CUDA simulation for experiment.

All configuration comes from experiment metadata.

Parameters

Name	Type	Description	Default
experiment	SimulationExperiment	SimulationExperiment instance	required
n_steps	Optional[int]	Number of integration steps (default from experiment.integration)	None
dt	Optional[float]	Integration time step in ms (default from experiment.integration)	None
n_work_items	int	Number of parallel parameter configurations	1
global_speed	float	Conduction speed (m/s)	1.0
global_coupling	float	Global coupling strength	0.1
buffer_length	Optional[int]	History buffer length (auto-calculated if None)	None
swept_params	Optional[Dict[str, np.ndarray]]	Dict mapping param names to arrays of values per work item	None

Returns

Name	Type	Description
	Dict[str, np.ndarray]	Dict with ‘tavg’, ‘n_node’, ‘n_steps’, ‘dt’

save_cuda

codegen.cuda.save_cuda(experiment, path)

Save CUDA source to file.

Parameters

Name	Type	Description	Default
experiment	SimulationExperiment	SimulationExperiment instance	required
path	str	Output file path (.cu)	required

Returns

Name	Type	Description
	str	Path to saved file