Basic functions used in MD simulations
[Tools]

Functions
__device__ void	gpuCumulateVec (FLOAT_ARRAY_TYPE *dVec, FLOAT_ARRAY_TYPE *dVecToAdd)
	Cumulate vector components.
void __global__	gpuRemoveDriftKernel (int N, FLOAT_ARRAY_TYPE *dVel, FLOAT_ARRAY_TYPE *compSum)
	System drift removal kernel.
__device__ int	gpuRnd (float b)
	Round float to integer.
void	removeDrift (int N, int numInterData, int numThreads, int maxThreads, int maxBlocks, FLOAT_ARRAY_TYPE *dVel, FLOAT_ARRAY_TYPE *compSum)
	Remove system drift.

---

Detailed Description

This module contains some basic inline functions for convenient programming.

---

Function Documentation

__device__ void gpuCumulateVec	(	FLOAT_ARRAY_TYPE *	dVec,
		FLOAT_ARRAY_TYPE *	dVecToAdd
	)			[inline]

Cumulate vector components.

Parameters:

[in,out]	dVec	Cumulative vector
[in]	dVecToAdd	Vector to add to /p vec

Returns:

void

Definition at line 111 of file gpu_tools.cuh.

Referenced by gpuLJForces(), gpuLJForcesCut(), gpuLJForcesCutRSmem(), gpuLJForcesCutSmem(), gpuLJForcesCutSmemSmem(), gpuLJForcesCutVar1(), gpuLJForcesSmem(), gpuLJForcesSmemSmem(), and gpuLJForcesVerletList().

void __global__ gpuRemoveDriftKernel	(	int	N,
		FLOAT_ARRAY_TYPE *	dVel,
		FLOAT_ARRAY_TYPE *	compSum
	)

The GPU kernel applies the drift removal factor to all velocities. The kernel is called from removeDrift .

Parameters:

[in]	N	Number of particles
[in,out]	dVel	Velocity array
[in]	compSum	component-wise summed velocites

Returns:

Returns CUDA error information

Definition at line 223 of file gpu_tools.cu.

References FLOAT_ARRAY_TYPE.

__device__ int gpuRnd

(

float

b

)

[inline]

Given b, the nearest integer is returned.

Parameters:

[in]

b

Floating point number

Returns:

Returns rounded integer of b .

Definition at line 98 of file gpu_tools.cuh.

Referenced by gpuDistPBC().

void removeDrift	(	int	N,
		int	numInterData,
		int	numThreads,
		int	maxThreads,
		int	maxBlocks,
		FLOAT_ARRAY_TYPE *	dVel,
		FLOAT_ARRAY_TYPE *	compSum
	)

To ged rid of the system drift, the component-wise total velocity is computed and substracted from the whole system.

Parameters:

[in]	N	Number of particles
[in]	numInterData	Intermediate data size for the reduction
[in]	numThreads	Number of threads used for reduction
[in]	maxThreads	Maximum number of allowed threads
[in]	maxBlocks	Maximum number of allowed blocks
[in,out]	dVel	Velocity array
[in]	compSum	Helper variable, content will be overwritten

Returns:

No return value.

Definition at line 213 of file gpu_tools.cu.

References checkCudaError(), and componentSumReduction().

Referenced by main().