RooBatchCompute Namespace Reference

Namespace for dispatching RooFit computations to various backends. More...

Namespaces
namespace	CUDA
namespace	CudaInterface

Classes
class	AbsBuffer
class	AbsBufferManager
class	Batch
class	Batches
class	BracketAdapter
	Little adapter that gives a bracket operator to types that don't have one. More...
class	BracketAdapterWithMask
class	Config
	Minimal configuration struct to steer the evaluation of a single node with the RooBatchCompute library. More...
struct	ReduceNLLOutput
class	RooBatchComputeInterface
	The interface which should be implemented to provide optimised computation functions for implementations of RooAbsReal::doEval(). More...

Typedefs
typedef std::span< double >	ArgSpan
typedef const double *__restrict	InputArr
typedef std::span< const std::span< const double > >	VarSpan

Enumerations
enum class	Architecture {   AVX512 , AVX2 , AVX , SSE4 ,   GENERIC , CUDA }
enum	Computer {   AddPdf , ArgusBG , BMixDecay , Bernstein ,   BifurGauss , BreitWigner , Bukin , CBShape ,   Chebychev , ChiSquare , DeltaFunction , DstD0BG ,   ExpPoly , Exponential , ExponentialNeg , Gamma ,   GaussModelExpBasis , Gaussian , Identity , Johnson ,   Landau , Lognormal , LognormalStandard , NegativeLogarithms ,   NormalizedPdf , Novosibirsk , Poisson , Polynomial ,   Power , ProdPdf , Ratio , TruthModelExpBasis ,   TruthModelSinBasis , TruthModelCosBasis , TruthModelLinBasis , TruthModelQuadBasis ,   TruthModelSinhBasis , TruthModelCoshBasis , Voigtian }

Functions
void	compute (Config cfg, Computer comp, std::span< double > output, std::initializer_list< std::span< const double > > vars, ArgSpan extraArgs={})
	It is not possible to construct a std::span directly from an initializer list (probably it will be with C++26).
void	compute (Config cfg, Computer comp, std::span< double > output, VarSpan vars, ArgSpan extraArgs={})
Architecture	cpuArchitecture ()
std::string	cpuArchitectureName ()
__roodevice__ double	fast_cos (double x)
__roodevice__ double	fast_exp (double x)
__roodevice__ double	fast_isqrt (double x)
__roodevice__ double	fast_log (double x)
__roodevice__ double	fast_sin (double x)
std::string	getBatchComputeChoice ()
int	initCPU ()
	Inspect hardware capabilities, and load the optimal library for RooFit computations.
int	initCUDA ()
ReduceNLLOutput	reduceNLL (Config cfg, std::span< const double > probas, std::span< const double > weights, std::span< const double > offsetProbas)
double	reduceSum (Config cfg, InputArr input, size_t n)
void	setBatchComputeChoice (std::string const &value)

Variables
constexpr std::size_t	bufferSize = 64
R__EXTERN RooBatchComputeInterface *	dispatchCPU = nullptr
	This dispatch pointer points to an implementation of the compute library, provided one has been loaded.
R__EXTERN RooBatchComputeInterface *	dispatchCUDA = nullptr

Detailed Description

Namespace for dispatching RooFit computations to various backends.

A space to attach TBranches.

This namespace contains an interface for providing high-performance computation functions for use in RooAbsReal::doEval(), see RooBatchComputeInterface.

Furthermore, several implementations of this interface can be created, which reside in RooBatchCompute::RF_ARCH, where RF_ARCH may be replaced by the architecture that this implementation targets, e.g. SSE, AVX, etc.

Using the pointer RooBatchCompute::dispatch, a computation request can be dispatched to the fastest backend that is available on a specific platform.

Typedef Documentation

ArgSpan

typedef std::span<double> RooBatchCompute::ArgSpan

Definition at line 45 of file RooBatchCompute.h.

InputArr

typedef const double* __restrict RooBatchCompute::InputArr

Definition at line 46 of file RooBatchCompute.h.

VarSpan

typedef std::span<const std::span<const double> > RooBatchCompute::VarSpan

Definition at line 44 of file RooBatchCompute.h.

Enumeration Type Documentation

Architecture

enum class RooBatchCompute::Architecture

strong

Enumerator
AVX512
AVX2
AVX
SSE4
GENERIC
CUDA

Definition at line 65 of file RooBatchCompute.h.

Computer

enum RooBatchCompute::Computer

Enumerator
AddPdf
ArgusBG
BMixDecay
Bernstein
BifurGauss
BreitWigner
Bukin
CBShape
Chebychev
ChiSquare
DeltaFunction
DstD0BG
ExpPoly
Exponential
ExponentialNeg
Gamma
GaussModelExpBasis
Gaussian
Identity
Johnson
Landau
Lognormal
LognormalStandard
NegativeLogarithms
NormalizedPdf
Novosibirsk
Poisson
Polynomial
Power
ProdPdf
Ratio
TruthModelExpBasis
TruthModelSinBasis
TruthModelCosBasis
TruthModelLinBasis
TruthModelQuadBasis
TruthModelSinhBasis
TruthModelCoshBasis
Voigtian

Definition at line 74 of file RooBatchCompute.h.

Function Documentation

compute() [1/2]

void RooBatchCompute::compute ( Config cfg, Computer comp, std::span< double > output, std::initializer_list< std::span< const double > > vars, ArgSpan extraArgs = {} )

inline

It is not possible to construct a std::span directly from an initializer list (probably it will be with C++26).

That's why we need an explicit overload for this.

Definition at line 221 of file RooBatchCompute.h.

compute() [2/2]

void RooBatchCompute::compute ( Config cfg, Computer comp, std::span< double > output, VarSpan vars, ArgSpan extraArgs = {} )

inline

Definition at line 212 of file RooBatchCompute.h.

cpuArchitecture()

Architecture RooBatchCompute::cpuArchitecture ( )

inline

Definition at line 202 of file RooBatchCompute.h.

cpuArchitectureName()

std::string RooBatchCompute::cpuArchitectureName ( )

inline

Definition at line 207 of file RooBatchCompute.h.

fast_cos()

__roodevice__ double RooBatchCompute::fast_cos ( double x )

inline

Definition at line 78 of file RooVDTHeaders.h.

fast_exp()

__roodevice__ double RooBatchCompute::fast_exp ( double x )

inline

Definition at line 68 of file RooVDTHeaders.h.

fast_isqrt()

__roodevice__ double RooBatchCompute::fast_isqrt ( double x )

inline

Definition at line 88 of file RooVDTHeaders.h.

fast_log()

__roodevice__ double RooBatchCompute::fast_log ( double x )

inline

Definition at line 83 of file RooVDTHeaders.h.

fast_sin()

__roodevice__ double RooBatchCompute::fast_sin ( double x )

inline

Definition at line 73 of file RooVDTHeaders.h.

getBatchComputeChoice()

std::string RooBatchCompute::getBatchComputeChoice

(

)

Definition at line 106 of file Initialisation.cxx.

initCPU()

int RooBatchCompute::initCPU

(

)

Inspect hardware capabilities, and load the optimal library for RooFit computations.

Definition at line 50 of file Initialisation.cxx.

initCUDA()

int RooBatchCompute::initCUDA

(

)

Definition at line 95 of file Initialisation.cxx.

reduceNLL()

ReduceNLLOutput RooBatchCompute::reduceNLL ( Config cfg, std::span< const double > probas, std::span< const double > weights, std::span< const double > offsetProbas )

inline

Definition at line 233 of file RooBatchCompute.h.

reduceSum()

double RooBatchCompute::reduceSum ( Config cfg, InputArr input, size_t n )

inline

Definition at line 227 of file RooBatchCompute.h.

setBatchComputeChoice()

void RooBatchCompute::setBatchComputeChoice

(

std::string const &

value

)

Definition at line 111 of file Initialisation.cxx.

Variable Documentation

bufferSize

constexpr std::size_t RooBatchCompute::bufferSize = 64

constexpr

Definition at line 48 of file RooBatchCompute.h.

dispatchCPU

RooBatchCompute::RooBatchComputeInterface * RooBatchCompute::dispatchCPU = nullptr

This dispatch pointer points to an implementation of the compute library, provided one has been loaded.

Using a virtual call, computation requests are dispatched to backends with architecture-specific functions such as SSE, AVX, AVX2, etc.

See also

RooBatchComputeInterface, RooBatchComputeClass, RF_ARCH

Definition at line 199 of file RooBatchCompute.h.

dispatchCUDA

RooBatchCompute::RooBatchComputeInterface * RooBatchCompute::dispatchCUDA = nullptr

Definition at line 200 of file RooBatchCompute.h.