Pscf::Rpg::VecOpFts Namespace Reference

Element-wise vector operations performed on the GPU for FTS classes. More...

Functions
void	mcftsScale (DeviceArray< Prdc::Cuda::cudaReal > &a, Prdc::Cuda::cudaReal const b)
	Rescale array a from [0,1] to [-b, b], GPU kernel wrapper.
void	fourierMove (DeviceArray< Prdc::Cuda::cudaComplex > &a, DeviceArray< Prdc::Cuda::cudaReal > const &b, DeviceArray< Prdc::Cuda::cudaReal > const &c)
	Add array b to real part of a and array c to imaginary part of a.
void	computeDField (DeviceArray< Prdc::Cuda::cudaReal > &d, DeviceArray< Prdc::Cuda::cudaReal > const &Wc, DeviceArray< Prdc::Cuda::cudaReal > const &Cc, Prdc::Cuda::cudaReal const a, Prdc::Cuda::cudaReal const b, Prdc::Cuda::cudaReal const s)
	Compute d field (functional derivative of H[w])
void	computeForceBias (DeviceArray< Prdc::Cuda::cudaReal > &result, DeviceArray< Prdc::Cuda::cudaReal > const &di, DeviceArray< Prdc::Cuda::cudaReal > const &df, DeviceArray< Prdc::Cuda::cudaReal > const &dwc, Prdc::Cuda::cudaReal mobility)
	Compute force bias.

Detailed Description

Element-wise vector operations performed on the GPU for FTS classes.

CUDA kernels that perform the operations are defined in VecOpFts.cu in an anonymous namespace, so they are not directly accessible. Kernel wrapper functions to be called by the host CPU, which call the kernel internally, are public.

The output (the LHS of the vector operation) will always be the first parameter passed to the function.

Function Documentation

mcftsScale()

void Pscf::Rpg::VecOpFts::mcftsScale	(	DeviceArray< Prdc::Cuda::cudaReal > &	a,
		Prdc::Cuda::cudaReal const	b )

Rescale array a from [0,1] to [-b, b], GPU kernel wrapper.

Parameters

a	array containing data between 0 and 1 (LHS)
b	scalar bound for rescaled array (RHS)

Definition at line 77 of file VecOpFts.cu.

References Pscf::DeviceArray< Data >::capacity(), Pscf::DeviceArray< Data >::cArray(), and Pscf::ThreadArray::setThreadsLogical().

fourierMove()

void Pscf::Rpg::VecOpFts::fourierMove	(	DeviceArray< Prdc::Cuda::cudaComplex > &	a,
		DeviceArray< Prdc::Cuda::cudaReal > const &	b,
		DeviceArray< Prdc::Cuda::cudaReal > const &	c )

Add array b to real part of a and array c to imaginary part of a.

Parameters

a	output array of cudaComplex values
b	input array of cudaReal values
c	input array of cudaReal values

Definition at line 92 of file VecOpFts.cu.

References Pscf::DeviceArray< Data >::capacity(), Pscf::DeviceArray< Data >::cArray(), Pscf::ThreadArray::setThreadsLogical(), and UTIL_CHECK.

computeDField()

void Pscf::Rpg::VecOpFts::computeDField	(	DeviceArray< cudaReal > &	d,
		DeviceArray< cudaReal > const &	Wc,
		DeviceArray< cudaReal > const &	Cc,
		cudaReal const	a,
		cudaReal const	b,
		cudaReal const	s )

Compute d field (functional derivative of H[w])

Definition at line 112 of file VecOpFts.cu.

References Pscf::DeviceArray< Data >::capacity(), Pscf::DeviceArray< Data >::cArray(), Pscf::ThreadArray::setThreadsLogical(), and UTIL_CHECK.

Referenced by Pscf::Rpg::Simulator< D >::computeDc().

computeForceBias()

void Pscf::Rpg::VecOpFts::computeForceBias	(	DeviceArray< cudaReal > &	result,
		DeviceArray< cudaReal > const &	di,
		DeviceArray< cudaReal > const &	df,
		DeviceArray< cudaReal > const &	dwc,
		cudaReal	mobility )

Compute force bias.

Definition at line 133 of file VecOpFts.cu.

References Pscf::DeviceArray< Data >::capacity(), Pscf::DeviceArray< Data >::cArray(), Pscf::ThreadArray::setThreadsLogical(), and UTIL_CHECK.

Referenced by Pscf::Rpg::ForceBiasMove< D >::move().