NN Utils Functions

Utils functions are miscellaneous auxiliary tools. More...

Functions
static void	hpm_nn_reshape_s8 (const int8_t *in_tensor, int8_t *out_tensor, const uint32_t size)
	This function turns the input tensor into another tensor with the same data but in a different shape. More...
static int32_t	hpm_nn_top_k_s8 (q7_t *in_vec, uint32_t size, uint32_t k, q7_t *val, uint32_t *idx)
	This function finds the k largest values and their indices from the signed 8-bit integer input vector. More...

Detailed Description

Utils functions are miscellaneous auxiliary tools.

Function Documentation

hpm_nn_reshape_s8()

static void hpm_nn_reshape_s8 ( const int8_t *  in_tensor, int8_t *  out_tensor, const uint32_t  size  )

inlinestatic

#include </home/docs/checkouts/readthedocs.org/user_builds/hpm-sdk/checkouts/latest/middleware/hpm_math/hpm_math.h>

This function turns the input tensor into another tensor with the same data but in a different shape.

Parameters

[in]	in_tensor	pointer of the input tensor
[out]	out_tensor	pointer of the output tensor
[in]	size	size, in bytes, of total input tensor

Example:

#define SIZE 1024
int8_t in_tensor[SIZE] = {...};
int8_t out_tensor[SIZE];
 
hpm_nn_reshape_s8(in_tensor, out_tensor, SIZE);

hpm_nn_top_k_s8()

static int32_t hpm_nn_top_k_s8 ( q7_t *  in_vec, uint32_t  size, uint32_t  k, q7_t *  val, uint32_t *  idx  )

inlinestatic

#include </home/docs/checkouts/readthedocs.org/user_builds/hpm-sdk/checkouts/latest/middleware/hpm_math/hpm_math.h>

This function finds the k largest values and their indices from the signed 8-bit integer input vector.

Parameters

[in]	in_vec	pointer of the input vector
[in]	size	number of elements in the input vector
[in]	k	the number of the largest values to be searched
[out]	val	the k largest values in the input vector
[out]	idx	the indices of the k largest values in the input vector

Returns

This function only returns 0.

Note

• If there is a number of elements with the same value, the element with smaller index will be selected with high priority.
• The k largest values will be sorted from largest to smallest and stored in "val" output vector. If there is a number of elements with the same value, the elements will be sorted from smallest index to largest index.