Classes
struct	mechanism_memory
	This struct is used to store memory for the CUDA RHS and Jacobian evaluation. Along with the solver_memory struct, this must be initialized and passed to the solvers the run. More...

Functions
__device__ void	dydt (const double t, const double mu, const double __restrict__ y, double __restrict__ dy, const mechanism_memory *__restrict__ d_mem)
	An implementation of the RHS of the van der Pol equation. More...

void	gpuAssert (cudaError_t code, const char *file, int line, bool abort=true)

size_t	required_mechanism_size ()
	Calculates and returns the total memory size (in bytes) required by an individual thread for the mechanism_memory struct. More...

void	initialize_gpu_memory (int padded, mechanism_memory h_mem, mechanism_memory d_mem)
	Initializes the host and device mechanism_memory structs. This is required in order to enable passing the struct to CUDA. More...

void	free_gpu_memory (mechanism_memory h_mem, mechanism_memory d_mem)
	Frees the host and device mechanism_memory structs. More...

void	set_same_initial_conditions (int NUM, double y_host, double var_host)
	Set same ICs for all problems. More...

void	apply_mask (double *y_host)
	Not needed for van der Pol. More...

void	apply_reverse_mask (double *y_host)
	Not needed for van der Pol. More...

__device__ void	eval_jacob (const double t, const double mu, const double __restrict__ y, double __restrict__ jac, const mechanism_memory *__restrict__ d_mem)
	An implementation of the van der Pol jacobian. More...

__device__ void	sparse_multiplier (const double A, const double Vm, double *w)
	Implements Jacobian \ vector multiplication in sparse (or unrolled) form. More...

Function Documentation

◆ apply_mask()

void van_der_pol_cu::apply_mask ( double * y_host )

Not needed for van der Pol.

In pyJac, these are used to transform the input/output vectors to deal with moving the last species mass fraction

Definition at line 46 of file ics.cu.

◆ apply_reverse_mask()

void van_der_pol_cu::apply_reverse_mask ( double * y_host )

Not needed for van der Pol.

In pyJac, these are used to transform the input/output vectors to deal with moving the last species mass fraction

Definition at line 53 of file ics.cu.

◆ dydt()

__device__ void van_der_pol_cu::dydt	(	const double	t,
		const double	mu,
		const double *__restrict__	y,
		double *__restrict__	dy,
		const mechanism_memory *__restrict__	d_mem
	)

An implementation of the RHS of the van der Pol equation.

The y and dy vectors supplied here are the global state vectors. Hence the vector accesses must be done with the global thread ID. The gpu_macros.cuh file defines some useful macros to simplify indexing

See also: gpu_macros.cuh

Parameters

[in]	t	The current system time
[in]	mu	The van der Pol parameter
[in]	y	The state vector
[out]	dy	The output RHS (dydt) vector
[in]	d_mem	The mechanism_memory struct. In future versions, this will be used to access the \(\mu\) parameter to have a consistent interface.

Definition at line 24 of file dydt.cu.

◆ eval_jacob()

__device__ void van_der_pol_cu::eval_jacob	(	const double	t,
		const double	mu,
		const double *__restrict__	y,
		double *__restrict__	jac,
		const mechanism_memory *__restrict__	d_mem
	)

An implementation of the van der Pol jacobian.

Parameters

[in]	t	The current system time
[in]	mu	The van der Pol parameter
[in]	y	The state vector at time t
[out]	jac	The jacobian to populate
[in]	d_mem	The mechanism_memory struct. In future versions, this will be used to access the \(\mu\) parameter to have a consistent interface.

The Jacobian is in Column-major (Fortran) order. As with dydt(), this function operates directly on global state vector and jacobian. Hence we use the INDEX macro defined in gpu_macros.cuh here.

See also: dydt()

Parameters

[in]	t	The current system time
[in]	mu	The van der Pol parameter
[in]	y	The state vector at time t
[out]	jac	The jacobian to populate
[in]	d_mem	The mechanism_memory struct. In future versions, this will be used to access the \(\mu\) parameter to have a consistent interface.

jac[0, 0] = \(\frac{\partial \dot{y_1}}{\partial y_1}\)

jac[0, 1] = \(\frac{\partial \dot{y_2}}{\partial y_1}\)

jac[1, 0] = \(\frac{\partial \dot{y_1}}{\partial y_2}\)

jac[1, 1] = \(\frac{\partial \dot{y_2}}{\partial y_2}\)

Definition at line 31 of file jacob.cu.

◆ free_gpu_memory()

void van_der_pol_cu::free_gpu_memory	(	mechanism_memory **	h_mem,
		mechanism_memory **	d_mem
	)

Frees the host and device mechanism_memory structs.

Parameters

[in,out]	h_mem	The host version of the mechanism_memory struct.
[in,out]	d_mem	The device version of the mechanism_memory struct.

Definition at line 47 of file gpu_memory.cu.

◆ gpuAssert()

void van_der_pol_cu::gpuAssert	(	cudaError_t	code,
		const char *	file,
		int	line,
		bool	abort = `true`
	)

inline

Definition at line 27 of file gpu_macros.cuh.

◆ initialize_gpu_memory()

void van_der_pol_cu::initialize_gpu_memory	(	int	padded,
		mechanism_memory **	h_mem,
		mechanism_memory **	d_mem
	)

Initializes the host and device mechanism_memory structs. This is required in order to enable passing the struct to CUDA.

Parameters

[in]	padded	The padded number of threads to be used by the CUDA solver
[in,out]	h_mem	The host version of the mechanism_memory struct to initialize.
[in,out]	d_mem	The device version of the mechanism_memory struct to copy the resulting host mechanism_memory struct to.

Definition at line 30 of file gpu_memory.cu.

◆ required_mechanism_size()

size_t van_der_pol_cu::required_mechanism_size ( )

Calculates and returns the total memory size (in bytes) required by an individual thread for the mechanism_memory struct.

Definition at line 15 of file gpu_memory.cu.

◆ set_same_initial_conditions()

void van_der_pol_cu::set_same_initial_conditions	(	int	NUM,
		double **	y_host,
		double **	var_host
	)

Set same ICs for all problems.

Parameters

NUM	The number of initial value problems
y_host	The state vectors to initialize
var_host	The vector of \(mu\) parameters for the van der Pol equation
NUM	The number of initial value problems
y_host	The state vectors to initialize
var_host	The vector of \(\mu\) parameters for the van der Pol equation

Definition at line 25 of file ics.cu.

◆ sparse_multiplier()

__device__ void van_der_pol_cu::sparse_multiplier	(	const double *	A,
		const double *	Vm,
		double *	w
	)

Implements Jacobian \ vector multiplication in sparse (or unrolled) form.

Parameters

[in]	A	The (NSP x NSP) Jacobian matrix, see eval_jacob() for details on layout
[in]	Vm	The (NSP x 1) vector to multiply by
[out]	w	The (NSP x 1) vector to store the result in, \(w := A * Vm\)

Definition at line 22 of file sparse_multiplier.cu.

Classes

Functions

Function Documentation

◆ apply_mask()

◆ apply_reverse_mask()

◆ dydt()

◆ eval_jacob()

◆ free_gpu_memory()

◆ gpuAssert()

◆ initialize_gpu_memory()

◆ required_mechanism_size()

◆ set_same_initial_conditions()

◆ sparse_multiplier()