TMVA::Experimental::SOFIE::UTILITY Namespace Reference

Functions
std::string	Clean_name (std::string input_tensor_name)
template<typename T >
void	Im2col (const T *data_im, const int channels, const int height, const int width, const int kernel_h, const int kernel_w, const int pad_h, const int pad_w, const int stride_h, const int stride_w, const int dilation_h, const int dilation_w, T *data_col)
	im2col : efficient function to re-arrange input data of convolution to a matrix that can be used by BLAS Use trick to loop on each element of filtered region first and follow input data layout By doing this reads and writes are of consecutive data in memory and one gains in efficiency The resulting matrix will be already transposed and can be used directly in BLAS since output will be a matrix : (channels*kernel_h*kernel_w , output_h*output_w) Example: with an input matrix a1 a2 a3 b1 b2 b3 and a 2x2 kernel (k1,k2,k3,k4) and padding 1 : c1 c2 c3 outpout will be a matrix (4 x 16) the routine will follow output order :
template<typename T >
void	Im2col_3d (const T *data_im, const int channels, const int depth, const int height, const int width, const int kernel_d, const int kernel_h, const int kernel_w, const int pad_d, const int pad_h, const int pad_w, const int stride_d, const int stride_h, const int stride_w, const int dilation_d, const int dilation_h, const int dilation_w, T *data_col)
	3d implementation
bool	is_a_ge_zero_and_a_lt_b (int a, int b)
	function to check if a >> 0 and a < MAX using a single comparison / use trick casting to unsigned values so it becomes a single comparison
template<typename T >
T *	Unidirectional_broadcast (const T *original_data, const std::vector< size_t > original_shape, const std::vector< size_t > target_shape)

Function Documentation

Clean_name()

std::string TMVA::Experimental::SOFIE::UTILITY::Clean_name

(

std::string

input_tensor_name

)

Definition at line 138 of file SOFIE_common.cxx.

Im2col()

template<typename T >

void TMVA::Experimental::SOFIE::UTILITY::Im2col	(	const T *	data_im,
		const int	channels,
		const int	height,
		const int	width,
		const int	kernel_h,
		const int	kernel_w,
		const int	pad_h,
		const int	pad_w,
		const int	stride_h,
		const int	stride_w,
		const int	dilation_h,
		const int	dilation_w,
		T *	data_col
	)

im2col : efficient function to re-arrange input data of convolution to a matrix that can be used by BLAS Use trick to loop on each element of filtered region first and follow input data layout By doing this reads and writes are of consecutive data in memory and one gains in efficiency The resulting matrix will be already transposed and can be used directly in BLAS since output will be a matrix : (channels*kernel_h*kernel_w , output_h*output_w) Example: with an input matrix a1 a2 a3 b1 b2 b3 and a 2x2 kernel (k1,k2,k3,k4) and padding 1 : c1 c2 c3
outpout will be a matrix (4 x 16)
the routine will follow output order :

-> ( 0 0 0 0 0 a1 a2 a3 0 b1 b2 b3 0 c1 c2 c3 ) all elements for k1 ( 0 0 0 0 a1 a2 a3 0 b1 b2 b3 0 c1 c2 c3 0 ) for k2 ( 0 a1 a2 a3 0 b1 b2 b3 0 c1 c2 c3 0 0 0 0 ) for k3 ( a1 a2 a3 0 b1 b2 b3 0 c1 c2 c3 0 0 0 0 0 ) for k4

Definition at line 138 of file SOFIE_common.hxx.

Im2col_3d()

template<typename T >

void TMVA::Experimental::SOFIE::UTILITY::Im2col_3d	(	const T *	data_im,
		const int	channels,
		const int	depth,
		const int	height,
		const int	width,
		const int	kernel_d,
		const int	kernel_h,
		const int	kernel_w,
		const int	pad_d,
		const int	pad_h,
		const int	pad_w,
		const int	stride_d,
		const int	stride_h,
		const int	stride_w,
		const int	dilation_d,
		const int	dilation_h,
		const int	dilation_w,
		T *	data_col
	)

3d implementation

Definition at line 174 of file SOFIE_common.hxx.

is_a_ge_zero_and_a_lt_b()

bool TMVA::Experimental::SOFIE::UTILITY::is_a_ge_zero_and_a_lt_b ( int  a, int  b  )

inline

function to check if a >> 0 and a < MAX using a single comparison / use trick casting to unsigned values so it becomes a single comparison

Definition at line 113 of file SOFIE_common.hxx.

Unidirectional_broadcast()

template<typename T >

T * TMVA::Experimental::SOFIE::UTILITY::Unidirectional_broadcast	(	const T *	original_data,
		const std::vector< size_t >	original_shape,
		const std::vector< size_t >	target_shape
	)

Definition at line 72 of file SOFIE_common.cxx.