Namespaces
	bitonic

	linalg

	linalg2

	polynomial

Classes
class	CudaException
	an exception which carries a standard error code More...

class	CudaExceptionStream
	used to simplify the process of generating an exception message More...

struct	DeviceProp

class	ExceptionStream
	used to simplify the process of generating an exception message More...

struct	FuncAttributes

Typedefs
typedef CudaExceptionStream	cudaEx

typedef ExceptionStream < std::runtime_error >	ex

typedef std::map< std::string, FuncAttributes >	fattrMap_t
	print a formatted report about the specified kernels More...

typedef cudaMemcpyKind	MemcpyKind

Functions
void	checkLastError (const std::string &msg="checkLastError")
	wraps getLastError More...

void	checkResult (cudaError_t result, const std::string &where)
	throws an exception if result is not success More...

void	deviceSynchronize ()
	blocks the host thread until kernels are done executing More...

template<typename T >
T	divideBy2 (const T &x)
	returns x/2 (using bit shift), if x is odd then the result is floor(x/2) More...

template<typename T >
T	dividePow2 (T x, T y)
	returns x/y if x and y are both powers of two, otherwise the result is undefined More...

void	free (void *devPtr)
	wraps cudaFree More...

template<unsigned int I, typename T >
T	getBit (T x)
	returns the value of the specified bit More...

int	getDevice ()
	wraps cudaGetDevice More...

int	getDeviceCount ()
	wraps cudaGetDeviceCount More...

template<class oiter >
void	getDeviceList (oiter &out)
	build a list of device names More...

DeviceProp	getDeviceProperties (int dev)
	wraps cudaGetDeviceProperties More...

template<typename T >
T	intDivideRoundUp (T x, T y)
	integer divide with round up More...

template<typename T >
T	intPow (T x, T p)
	returns `x` to the power of `p` More...

template<typename T >
T	isEven (T x)
	returns true if the number is even More...

template<typename T >
T	isOdd (T x)
	returns true if the number is odd More...

template<typename T >
bool	isPow2 (T x)
	returns true if the parameter is an exact power of two More...

template<typename T >
T	log2 (T x)
	if `x` is a power of two then it returns the log of `x` with base 2 More...

void *	malloc (size_t size)
	wraps cudaMalloc More...

void *	mallocPitch (size_t &pitch, size_t cols, size_t rows)
	wraps cudaMallocPitch More...

template<typename T >
T *	mallocPitchT (size_t &pitch, size_t obsPerRow, size_t cols)
	allocates `nObjs` objects of type T More...

template<typename T >
T *	mallocT (size_t nObjs)
	allocates `nObjs` objects of type T More...

void	memcpy (void dst, const void src, size_t count, MemcpyKind kind)
	wraps cudaMemcpy More...

void	memcpy2D (void dst, size_t dpitch, const void src, size_t spitch, size_t width, size_t height, MemcpyKind kind)
	wraps cudaMemcpy2D More...

template<typename T >
void	memcpy2DT (T dst, size_t dpitchBytes, const T src, size_t spitchBytes, size_t widthObs, size_t height, MemcpyKind kind)
	wraps cudaMemcpy2D More...

template<typename T >
void	memcpyT (T dst, const T src, size_t nObjs, MemcpyKind kind)
	allocates `nObjs` objects of type T More...

void	memset (void *devPtr, int value, size_t count)
	wraps cudaMemset More...

template<typename T >
void	memset (T *devPtr, int value, size_t nObs)
	wraps cudaMemset More...

void	memset2D (void *devPtr, size_t pitch, int value, size_t width, size_t height)
	wraps cudaMemset2D More...

template<typename T >
void	memset2DT (T *devPtr, size_t pitchBytes, int value, size_t widthObjs, size_t height)
	wraps cudaMemset2D More...

template<typename T >
T	nextPow2 (T x)
	returns the smallest power of two that is not less than `x` More...

template<typename T >
T	prevPow2 (T x)
	returns the largest power of two that is not greater than `x` More...

void	printDeviceReport (std::ostream &out, int dev)
	print a formatted report about the specified device More...

void	printDeviceReport (std::ostream &out)

void	printKernelReport (std::ostream &out, const fattrMap_t &attr)

void	setDevice (int dev)
	wraps cudaSetDevice More...

template<typename T >
T	times2 (const T &x)
	returns x*2 (using bit shift) More...

template<typename T >
T	twoPow (T x)
	returns 2 to the power of `x` (2^x) More...

Typedef Documentation

typedef CudaExceptionStream mpblocks::cuda::cudaEx

Definition at line 120 of file ExceptionStream.h.

typedef ExceptionStream<std::runtime_error> mpblocks::cuda::ex

Definition at line 119 of file ExceptionStream.h.

typedef std::map<std::string,FuncAttributes> mpblocks::cuda::fattrMap_t

print a formatted report about the specified kernels

Definition at line 166 of file wrap.h.

typedef cudaMemcpyKind mpblocks::cuda::MemcpyKind

Definition at line 36 of file wrap.h.

Function Documentation

void mpblocks::cuda::checkLastError ( const std::string & msg = "checkLastError" )

wraps getLastError

void mpblocks::cuda::checkResult	(	cudaError_t	result,
		const std::string &	where
	)

throws an exception if result is not success

void mpblocks::cuda::deviceSynchronize ( )

blocks the host thread until kernels are done executing

void mpblocks::cuda::free ( void * devPtr )

wraps cudaFree

int mpblocks::cuda::getDevice ( )

wraps cudaGetDevice

int mpblocks::cuda::getDeviceCount ( )

wraps cudaGetDeviceCount

template<class oiter >

void mpblocks::cuda::getDeviceList ( oiter & out )

build a list of device names

Definition at line 151 of file wrap.h.

DeviceProp mpblocks::cuda::getDeviceProperties ( int dev )

wraps cudaGetDeviceProperties

void* mpblocks::cuda::malloc ( size_t size )

wraps cudaMalloc

void* mpblocks::cuda::mallocPitch	(	size_t &	pitch,
		size_t	cols,
		size_t	rows
	)

wraps cudaMallocPitch

template<typename T >

T * mpblocks::cuda::mallocPitchT	(	size_t &	pitch,
		size_t	obsPerRow,
		size_t	cols
	)

allocates nObjs objects of type T

Note: the pitch returned is in bytes, not in nObjs

Definition at line 49 of file wrap.hpp.

template<typename T >

T * mpblocks::cuda::mallocT ( size_t nObjs )

allocates nObjs objects of type T

Definition at line 42 of file wrap.hpp.

void mpblocks::cuda::memcpy	(	void *	dst,
		const void *	src,
		size_t	count,
		MemcpyKind	kind
	)

wraps cudaMemcpy

void mpblocks::cuda::memcpy2D	(	void *	dst,
		size_t	dpitch,
		const void *	src,
		size_t	spitch,
		size_t	width,
		size_t	height,
		MemcpyKind	kind
	)

wraps cudaMemcpy2D

template<typename T >

void mpblocks::cuda::memcpy2DT	(	T *	dst,
		size_t	dpitchBytes,
		const T *	src,
		size_t	spitchBytes,
		size_t	widthObs,
		size_t	height,
		MemcpyKind	kind
	)

wraps cudaMemcpy2D

Definition at line 62 of file wrap.hpp.

template<typename T >

void mpblocks::cuda::memcpyT	(	T *	dst,
		const T *	src,
		size_t	nObjs,
		MemcpyKind	kind
	)

allocates nObjs objects of type T

Definition at line 56 of file wrap.hpp.

void mpblocks::cuda::memset	(	void *	devPtr,
		int	value,
		size_t	count
	)

wraps cudaMemset

template<typename T >

void mpblocks::cuda::memset	(	T *	devPtr,
		int	value,
		size_t	nObs
	)

wraps cudaMemset

Definition at line 81 of file wrap.hpp.

void mpblocks::cuda::memset2D	(	void *	devPtr,
		size_t	pitch,
		int	value,
		size_t	width,
		size_t	height
	)

wraps cudaMemset2D

template<typename T >

void mpblocks::cuda::memset2DT	(	T *	devPtr,
		size_t	pitchBytes,
		int	value,
		size_t	widthObjs,
		size_t	height
	)

wraps cudaMemset2D

Definition at line 88 of file wrap.hpp.

template<typename T >

T mpblocks::cuda::prevPow2 ( T x )

inline

returns the largest power of two that is not greater than x

If x is a power of two, then it returns x, otherwise, it returns the next lower one

For example:

    x    prev     next
 ------ ------   ------
    0 :     0        0
    1 :     1        1
    2 :     2        2
    3 :     2        4
    4 :     4        4
    5 :     4        8
    6 :     4        8
    7 :     4        8
    8 :     8        8

Definition at line 137 of file powersOfTwo.h.

void mpblocks::cuda::printDeviceReport	(	std::ostream &	out,
		int	dev
	)

print a formatted report about the specified device

void mpblocks::cuda::printDeviceReport ( std::ostream & out )

void mpblocks::cuda::printKernelReport	(	std::ostream &	out,
		const fattrMap_t &	attr
	)

void mpblocks::cuda::setDevice ( int dev )

wraps cudaSetDevice

Namespaces

Classes

Typedefs

Functions

Typedef Documentation

Function Documentation