Decrement context refcount */

Checks if a string contains a given suffix * * @param[in] str Input string * @param[in] suffix Suffix to check for * * @return True if the str

Error function. \param arg function argument \return error function value of \a arg template<typename T> typename enable<expr,T>::type erf(T arg) { re

Get eta. */

Get the data from *ptr after casting according to @p data_type and then convert the data to double. * * @param[in] ptr Pointer to value. * @p

Look up the format corresponding to a channel. * * @param[in] channel Channel type. * * @return Format that contains the given channel. */

Method to return the depthwise convolution configuration function based on data type * * @param[in] data_type Input data type * *

Method to return the matmul native configuration function based on data type * * @param[in] data_type Input data type * * @return

Method to return the direct convolution configuration function based on data type * * @param[in] data_type Input data type * * @re

Given a texel coordinates this function will return the following array of coordinates: * [ P, right neighbour, below neighbour, below right neighbou

Return an iterator that iterates over the record.

Get the preferred dimension in which the scheduler splits the work into multiple jobs. * * @return The split dimension hint. */

Interface implementation - Compulsory functions

Compute the amount of working space required for a single thread. */

Increment context refcount */

Check if normalization is cross map */

Check if a given data type is of asymmetric quantized signed type * * @param[in] dt Input data type. * * @return True if data type is of asymmetri

Check if normalization is not cross map */

Checks if a pixel has valid coordinates * * @param x X coordinate * @param y Y coordinate * @param width Width of the

Get the kappa value */

Limit this border size. * * @param[in] limit Border size to limit this border size to. */

Get the NMS */

Get nms threshold. */

Get the padding. * * @note This should only be used when the padding is symmetric. * * @return a pair: padding left/right, padding

Check that the provided Requantize32 is compatible with the "symmetric" hybrid kernels. These don't include row sums, so the 'b_offset' has to be zer

Multiply by power of two. \param arg number to modify \param exp power of two to multiply with \return \a arg multplied by 2 raised to \a exp template

Scales a given vector by 1/255. * * @note This does not work for all cases. e.g. for float of 0.49999999999999994 and large floats. * * @param in

Return the scaling factor of the normalization function. * * If is_scaled is set to false then [Krichevksy 2012] normalization scaling is pe

If do in-place calculation, then need to use the new output and inherit original output's accessor

combined JPEG 'receive' and JPEG 'extend', since baseline always extends everything it receives.

returns 0..31 for the highest set bit

extract an arbitrarily-aligned N-bit value (N=bits) from v, and then make it 8-bits long and fractionally extend it to full full range.

returns STBI_rgb or whatever, 0 on error

Get the stride. * * @return a pair: stride x, stride y. */