Logo ROOT   6.14/05
Reference Guide
List of all members | Static Public Member Functions | Static Public Attributes | List of all members
TMVA::DNN::TDevice Class Reference

TDevice.

The TDevice class provides static functions for the generation of CUDA grids for kernel launches and is used to encapsulate the distribution of threads and blocks over the data.

Definition at line 36 of file Device.h.

Static Public Member Functions

static dim3 BlockDims1D ()
 
static dim3 BlockDims2D ()
 
template<typename AFloat >
static dim3 GridDims1D (const TCudaMatrix< AFloat > &A)
 
template<typename AFloat >
static dim3 GridDims2D (const TCudaMatrix< AFloat > &A)
 
template<typename AFloat >
static int NThreads (const TCudaMatrix< AFloat > &A)
 

Static Public Attributes

static constexpr int BlockDimX = 1
 
static constexpr int BlockDimY = 32
 
static constexpr int BlockSize = BlockDimX * BlockDimY
 

#include <TMVA/DNN/Architectures/Cuda/Device.h>

Member Function Documentation

◆ BlockDims1D()

static dim3 TMVA::DNN::TDevice::BlockDims1D ( )
inlinestatic

Definition at line 48 of file Device.h.

◆ BlockDims2D()

static dim3 TMVA::DNN::TDevice::BlockDims2D ( )
inlinestatic

Definition at line 55 of file Device.h.

◆ GridDims1D()

template<typename AFloat >
static dim3 TMVA::DNN::TDevice::GridDims1D ( const TCudaMatrix< AFloat > &  A)
inlinestatic

Definition at line 63 of file Device.h.

◆ GridDims2D()

template<typename AFloat >
static dim3 TMVA::DNN::TDevice::GridDims2D ( const TCudaMatrix< AFloat > &  A)
inlinestatic

Definition at line 75 of file Device.h.

◆ NThreads()

template<typename AFloat >
static int TMVA::DNN::TDevice::NThreads ( const TCudaMatrix< AFloat > &  A)
inlinestatic

Definition at line 88 of file Device.h.

Member Data Documentation

◆ BlockDimX

constexpr int TMVA::DNN::TDevice::BlockDimX = 1
static

Definition at line 40 of file Device.h.

◆ BlockDimY

constexpr int TMVA::DNN::TDevice::BlockDimY = 32
static

Definition at line 42 of file Device.h.

◆ BlockSize

constexpr int TMVA::DNN::TDevice::BlockSize = BlockDimX * BlockDimY
static

Definition at line 44 of file Device.h.


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