neural_infer.h File Reference

Neural inference operator scaffolding. More...

#include "oakfield/operator_split.h"
#include <stdbool.h>
#include <stddef.h>

Include dependency graph for neural_infer.h:

This graph shows which files directly or indirectly include this file:

Go to the source code of this file.

Classes
struct	SimNeuralInferenceRequest
	Inference request metadata passed to neural callbacks. More...
struct	SimNeuralBaseConfig
	Shared base configuration for neural operators. More...

Macros
#define	SIM_NEURAL_MODEL_ID_MAX   63U

Typedefs
typedef struct SimNeuralInferenceRequest	SimNeuralInferenceRequest
	Inference request metadata passed to neural callbacks.
typedef SimResult(*	SimNeuralInferenceFn) (void *userdata, const SimField *input, SimField *output, const SimNeuralInferenceRequest *request)
	Callback surface for pluggable neural inference backends.
typedef struct SimNeuralBaseConfig	SimNeuralBaseConfig
	Shared base configuration for neural operators.
typedef SimNeuralBaseConfig	SimNeuralInferOperatorConfig
	Configuration for inference-only neural operator.

Functions
SimResult	sim_add_neural_infer_operator (struct SimContext *context, const SimNeuralInferOperatorConfig *config, size_t *out_index)
	Register an inference-only neural operator.
SimResult	sim_neural_infer_config (struct SimContext *context, size_t operator_index, SimNeuralInferOperatorConfig *out_config)
	Copy the current neural inference configuration from a registered operator.
SimResult	sim_neural_infer_update (struct SimContext *context, size_t operator_index, const SimNeuralInferOperatorConfig *config)
	Replace the configuration of a registered neural inference operator.

Detailed Description

Neural inference operator scaffolding.

Macro Definition Documentation

SIM_NEURAL_MODEL_ID_MAX

#define SIM_NEURAL_MODEL_ID_MAX   63U

Maximum model identifier length (excluding null terminator).

Typedef Documentation

SimNeuralInferenceFn

typedef SimResult(* SimNeuralInferenceFn) (void *userdata, const SimField *input, SimField *output, const SimNeuralInferenceRequest *request)

Callback surface for pluggable neural inference backends.

Implementations write model outputs into output using the same shape and element type as configured fields.

Parameters

	userdata	Opaque backend pointer from the operator configuration.
	input	Source field snapshot for the inference request.
[out]	output	Field storage that receives model predictions.
	request	Normalized inference request metadata.

Returns

SIM_RESULT_OK on success or an error code reported by the backend.

Function Documentation

sim_add_neural_infer_operator()

SimResult sim_add_neural_infer_operator	(	struct SimContext *	context,
		const SimNeuralInferOperatorConfig *	config,
		size_t *	out_index
	)

Register an inference-only neural operator.

Parameters

	context	Simulation context that will own the operator.
	config	Optional neural inference configuration; NULL selects normalized defaults.
[out]	out_index	Optional destination for the registered operator index.

Returns

SIM_RESULT_OK on success, or an error code from argument validation, field compatibility checks, allocation, or split registration.

sim_neural_infer_config()

SimResult sim_neural_infer_config	(	struct SimContext *	context,
		size_t	operator_index,
		SimNeuralInferOperatorConfig *	out_config
	)

Copy the current neural inference configuration from a registered operator.

Parameters

	context	Simulation context containing the operator.
	operator_index	Index returned by sim_add_neural_infer_operator().
[out]	out_config	Receives the normalized configuration.

Returns

SIM_RESULT_OK on success, SIM_RESULT_INVALID_ARGUMENT for NULL pointers, SIM_RESULT_NOT_FOUND for a missing operator, or SIM_RESULT_INVALID_STATE when the operator has no neural state.

sim_neural_infer_update()

SimResult sim_neural_infer_update	(	struct SimContext *	context,
		size_t	operator_index,
		const SimNeuralInferOperatorConfig *	config
	)

Replace the configuration of a registered neural inference operator.

config is required. The replacement is normalized and field/shape compatibility is checked before storing it.

Parameters

context	Simulation context containing the operator.
operator_index	Index of the neural inference operator to update.
config	Replacement neural inference configuration.

Returns

SIM_RESULT_OK on success, or an error code from argument validation, lookup, field compatibility checks, or state validation.