Inheritance diagram for quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >:

Collaboration diagram for quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >:

[legend]

Public Member Functions
	BlasCuda (SpinorX &X, SpinorY &Y, SpinorZ &Z, SpinorW &W, SpinorV &V, Functor &f, ColorSpinorField &x, ColorSpinorField &y, ColorSpinorField &z, ColorSpinorField &w, ColorSpinorField &v, int length)

virtual	~BlasCuda ()

TuneKey	tuneKey () const

void	apply (const cudaStream_t &stream)

void	preTune ()

void	postTune ()

void	initTuneParam (TuneParam &param) const

void	defaultTuneParam (TuneParam &param) const

long long	flops () const

long long	bytes () const

int	tuningIter () const

Public Member Functions inherited from quda::Tunable
	Tunable ()

virtual	~Tunable ()

virtual std::string	paramString (const TuneParam &param) const

virtual std::string	perfString (float time) const

virtual bool	advanceTuneParam (TuneParam &param) const

void	checkLaunchParam (TuneParam &param)

CUresult	jitifyError () const

CUresult &	jitifyError ()

Private Member Functions
unsigned int	sharedBytesPerThread () const

unsigned int	sharedBytesPerBlock (const TuneParam &param) const

virtual bool	advanceSharedBytes (TuneParam &param) const

Private Attributes
const int	nParity

BlasArg< SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >	arg

const ColorSpinorField &	x

const ColorSpinorField &	y

const ColorSpinorField &	z

const ColorSpinorField &	w

const ColorSpinorField &	v

char *	X_h

char *	Y_h

char *	Z_h

char *	W_h

char *	V_h

char *	Xnorm_h

char *	Ynorm_h

char *	Znorm_h

char *	Wnorm_h

char *	Vnorm_h

Additional Inherited Members
Protected Member Functions inherited from quda::Tunable
virtual unsigned int	minThreads () const

virtual bool	tuneGridDim () const

virtual bool	tuneAuxDim () const

virtual bool	tuneSharedBytes () const

virtual bool	advanceGridDim (TuneParam &param) const

virtual unsigned int	maxBlockSize (const TuneParam &param) const

virtual unsigned int	maxGridSize () const

virtual unsigned int	minGridSize () const

virtual int	gridStep () const
	gridStep sets the step size when iterating the grid size in advanceGridDim. More...

virtual int	blockStep () const

virtual int	blockMin () const

virtual void	resetBlockDim (TuneParam &param) const

virtual bool	advanceBlockDim (TuneParam &param) const

unsigned int	maxBlocksPerSM () const
	For some reason this can't be queried from the device properties, so here we set set this. Based on Table 14 of the CUDA Programming Guide 10.0 (Technical Specifications per Compute Capability) More...

template<typename F >
void	setMaxDynamicSharedBytesPerBlock (F *func) const
	Enable the maximum dynamic shared bytes for the kernel "func" (values given by maxDynamicSharedBytesPerBlock()). More...

unsigned int	maxDynamicSharedBytesPerBlock () const
	This can't be correctly queried in CUDA for all architectures so here we set set this. Based on Table 14 of the CUDA Programming Guide 10.0 (Technical Specifications per Compute Capability). More...

virtual unsigned int	maxSharedBytesPerBlock () const
	The maximum shared memory that a CUDA thread block can use in the autotuner. This isn't necessarily the same as maxDynamicSharedMemoryPerBlock since that may need explicit opt in to enable (by calling setMaxDynamicSharedBytes for the kernel in question). If the CUDA kernel in question does this opt in then this function can be overloaded to return maxDynamicSharedBytesPerBlock. More...

virtual bool	advanceAux (TuneParam &param) const

int	writeAuxString (const char *format,...)

Protected Attributes inherited from quda::Tunable
char	aux [TuneKey::aux_n]

CUresult	jitify_error

Detailed Description

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>
class quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >

Definition at line 29 of file blas_quda.cu.

Constructor & Destructor Documentation

◆ BlasCuda()

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>

quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::BlasCuda	(	SpinorX &	X,
		SpinorY &	Y,
		SpinorZ &	Z,
		SpinorW &	W,
		SpinorV &	V,
		Functor &	f,
		ColorSpinorField &	x,
		ColorSpinorField &	y,
		ColorSpinorField &	z,
		ColorSpinorField &	w,
		ColorSpinorField &	v,
		int	length
	)

inline

Definition at line 58 of file blas_quda.cu.

References quda::Tunable::aux, quda::LatticeField::AuxString(), and quda::LatticeField::Precision().

Here is the call graph for this function:

◆ ~BlasCuda()

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>

virtual quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::~BlasCuda ( )

inlinevirtual

Definition at line 89 of file blas_quda.cu.

Member Function Documentation

◆ advanceSharedBytes()

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>

virtual bool quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::advanceSharedBytes ( TuneParam & param ) const

inlineprivatevirtual

The goal here is to throttle the number of thread blocks per SM by over-allocating shared memory (in order to improve L2 utilization, etc.). We thus request the smallest amount of dynamic shared memory that guarantees throttling to a given number of blocks, in order to allow some extra leeway.

Reimplemented from quda::Tunable.

Definition at line 46 of file blas_quda.cu.

References quda::Tunable::advanceBlockDim(), quda::TuneParam::block, quda::TuneParam::shared_bytes, quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::sharedBytesPerBlock(), and quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::sharedBytesPerThread().

Here is the call graph for this function:

◆ apply()

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>

void quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::apply ( const cudaStream_t & stream )

inlinevirtual

Implements quda::Tunable.

Definition at line 93 of file blas_quda.cu.

References quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::arg, quda::TuneParam::block, getTuning(), getVerbosity(), quda::TuneParam::grid, quda::Tunable::jitify_error, quda::TuneParam::shared_bytes, and quda::tuneLaunch().

Referenced by quda::blas::nativeBlas().

Here is the call graph for this function:

Here is the caller graph for this function:

◆ bytes()

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>

long long quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::bytes ( ) const

inlinevirtual

Reimplemented from quda::Tunable.

Definition at line 138 of file blas_quda.cu.

References quda::ColorSpinorField::Bytes(), and quda::blas::BlasArg< SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::f.

Referenced by quda::blas::nativeBlas().

Here is the call graph for this function:

Here is the caller graph for this function:

◆ defaultTuneParam()

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>

void quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::defaultTuneParam ( TuneParam & param ) const

inlinevirtual

sets default values for when tuning is disabled

Reimplemented from quda::Tunable.

Definition at line 131 of file blas_quda.cu.

References quda::TuneParam::grid, quda::Tunable::initTuneParam(), and quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::nParity.

Here is the call graph for this function:

◆ flops()

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>

long long quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::flops ( ) const

inlinevirtual

Implements quda::Tunable.

Definition at line 137 of file blas_quda.cu.

References quda::blas::BlasArg< SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::f, and quda::blas::BlasArg< SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::length.

Referenced by quda::blas::nativeBlas().

Here is the caller graph for this function:

◆ initTuneParam()

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>

void quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::initTuneParam ( TuneParam & param ) const

inlinevirtual

Reimplemented from quda::Tunable.

Definition at line 125 of file blas_quda.cu.

References quda::TuneParam::grid, quda::Tunable::initTuneParam(), and quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::nParity.

Here is the call graph for this function:

◆ postTune()

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>

void quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::postTune ( )

inlinevirtual

Reimplemented from quda::Tunable.

Definition at line 116 of file blas_quda.cu.

References quda::ColorSpinorField::Bytes(), quda::ColorSpinorField::NormBytes(), quda::blas::BlasArg< SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::V, quda::blas::BlasArg< SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::W, quda::blas::BlasArg< SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::X, quda::blas::BlasArg< SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::Y, and quda::blas::BlasArg< SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::Z.

Here is the call graph for this function:

◆ preTune()

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>

void quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::preTune ( )

inlinevirtual

Reimplemented from quda::Tunable.

Definition at line 107 of file blas_quda.cu.

References quda::ColorSpinorField::Bytes(), quda::ColorSpinorField::NormBytes(), quda::blas::BlasArg< SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::V, quda::blas::BlasArg< SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::W, quda::blas::BlasArg< SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::X, quda::blas::BlasArg< SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::Y, and quda::blas::BlasArg< SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::Z.

Here is the call graph for this function:

◆ sharedBytesPerBlock()

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>

unsigned int quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::sharedBytesPerBlock ( const TuneParam & param ) const

inlineprivatevirtual

Implements quda::Tunable.

Definition at line 44 of file blas_quda.cu.

Referenced by quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::advanceSharedBytes().

Here is the caller graph for this function:

◆ sharedBytesPerThread()

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>

unsigned int quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::sharedBytesPerThread ( ) const

inlineprivatevirtual

Implements quda::Tunable.

Definition at line 43 of file blas_quda.cu.

Referenced by quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::advanceSharedBytes().

Here is the caller graph for this function:

◆ tuneKey()

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>

TuneKey quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::tuneKey ( ) const

inlinevirtual

Implements quda::Tunable.

Definition at line 91 of file blas_quda.cu.

References quda::Tunable::aux, quda::blas::BlasArg< SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::f, and quda::LatticeField::VolString().

Here is the call graph for this function:

◆ tuningIter()

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>

int quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::tuningIter ( ) const

inlinevirtual

Reimplemented from quda::Tunable.

Definition at line 144 of file blas_quda.cu.

Member Data Documentation

◆ arg

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>

BlasArg<SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor> quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::arg

mutableprivate

Definition at line 34 of file blas_quda.cu.

Referenced by quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::apply().

◆ nParity

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>

const int quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::nParity

private

Definition at line 33 of file blas_quda.cu.

Referenced by quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::defaultTuneParam(), and quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::initTuneParam().

◆ v

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>

const ColorSpinorField & quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::v

private

Definition at line 36 of file blas_quda.cu.

Referenced by quda::blas::mixed_blas(), and quda::blas::uni_blas().

◆ V_h

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>

char * quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::V_h

private

Definition at line 40 of file blas_quda.cu.

◆ Vnorm_h

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>

char * quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::Vnorm_h

private

Definition at line 41 of file blas_quda.cu.

◆ w

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>

const ColorSpinorField & quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::w

private

Definition at line 36 of file blas_quda.cu.

Referenced by quda::blas::caxpbypczw(), quda::blas::caxpbypzYmbw(), quda::blas::mixed_blas(), quda::blas::tripleCGUpdate(), and quda::blas::uni_blas().

◆ W_h

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>

char * quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::W_h

private

Definition at line 40 of file blas_quda.cu.

◆ Wnorm_h

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>

char * quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::Wnorm_h

private

Definition at line 41 of file blas_quda.cu.

◆ x

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>

const ColorSpinorField& quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::x

private

◆ X_h

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>

char* quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::X_h

private

Definition at line 40 of file blas_quda.cu.

◆ Xnorm_h

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>

char* quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::Xnorm_h

private

Definition at line 41 of file blas_quda.cu.

◆ y

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>

const ColorSpinorField & quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::y

private

◆ Y_h

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>

char * quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::Y_h

private

Definition at line 40 of file blas_quda.cu.

◆ Ynorm_h

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>

char * quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::Ynorm_h

private

Definition at line 41 of file blas_quda.cu.

◆ z

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>

const ColorSpinorField & quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::z

private

Definition at line 36 of file blas_quda.cu.

Referenced by quda::blas::axpbyz(), quda::blas::axpyBzpcx(), quda::blas::axpyZpbx(), quda::blas::caxpbypczw(), quda::blas::caxpbypzYmbw(), quda::blas::caxpyBxpz(), quda::blas::caxpyBzpx(), quda::blas::caxpyXmaz(), quda::blas::caxpyXmazMR(), quda::blas::cxpaypbz(), quda::blas::doubleCG3Init(), quda::blas::doubleCG3Update(), quda::blas::mixed_blas(), quda::blas::tripleCGUpdate(), and quda::blas::uni_blas().

◆ Z_h

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>

char * quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::Z_h

private

Definition at line 40 of file blas_quda.cu.

◆ Znorm_h

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>

char * quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >::Znorm_h

private

Definition at line 41 of file blas_quda.cu.

The documentation for this class was generated from the following file:

lib/blas_quda.cu

Public Member Functions

Private Member Functions

Private Attributes

Additional Inherited Members

Detailed Description

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor> class quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >

Constructor & Destructor Documentation

◆ BlasCuda()

◆ ~BlasCuda()

Member Function Documentation

◆ advanceSharedBytes()

◆ apply()

◆ bytes()

◆ defaultTuneParam()

◆ flops()

◆ initTuneParam()

◆ postTune()

◆ preTune()

◆ sharedBytesPerBlock()

◆ sharedBytesPerThread()

◆ tuneKey()

◆ tuningIter()

Member Data Documentation

◆ arg

◆ nParity

◆ v

◆ V_h

◆ Vnorm_h

◆ w

◆ W_h

◆ Wnorm_h

◆ x

◆ X_h

◆ Xnorm_h

◆ y

◆ Y_h

◆ Ynorm_h

◆ z

◆ Z_h

◆ Znorm_h

template<typename FloatN, int M, typename SpinorX, typename SpinorY, typename SpinorZ, typename SpinorW, typename SpinorV, typename Functor>
class quda::blas::BlasCuda< FloatN, M, SpinorX, SpinorY, SpinorZ, SpinorW, SpinorV, Functor >