|
float32x4x3_t | load_matrix_row (const float *ptr, int weights_offset=0) |
| Loads a 3x3 matrix as a row (float). More...
|
|
qint8x8x3_t | load_matrix_row (const qint8_t *ptr, int weights_offset=0) |
| Loads a 3x3 matrix as a row (qint8_t). More...
|
|
int32x4x3_t | load_matrix_row (const uint8_t *ptr, int weights_offset=0) |
| Loads a 3x3 matrix as a row (uint8_t). More...
|
|
template<unsigned int stridex> |
float32x4x2_t | convolve_3x3 (const float *in_top, const float *in_mid, const float *in_low, const float32x4x3_t &m0, const float32x4x3_t &m1, const float32x4x3_t &m2, int fixed_point_position, int input_offset=0) |
| Perform a convolve3x3 on float32. More...
|
|
template<> |
float32x4x2_t | convolve_3x3< 1 > (const float *in_top, const float *in_mid, const float *in_low, const float32x4x3_t &m0, const float32x4x3_t &m1, const float32x4x3_t &m2, int fixed_point_position, int input_offset) |
|
template<> |
float32x4x2_t | convolve_3x3< 2 > (const float *in_top, const float *in_mid, const float *in_low, const float32x4x3_t &m0, const float32x4x3_t &m1, const float32x4x3_t &m2, int fixed_point_position, int input_offset) |
|
template<> |
float32x4x2_t | convolve_3x3< 3 > (const float *in_top, const float *in_mid, const float *in_low, const float32x4x3_t &m0, const float32x4x3_t &m1, const float32x4x3_t &m2, int fixed_point_position, int input_offset) |
|
template<unsigned int stridex> |
qint16x8x2_t | convolve_3x3 (const qint8_t *in_top, const qint8_t *in_mid, const qint8_t *in_low, const qint8x8x3_t &m0, const qint8x8x3_t &m1, const qint8x8x3_t &m2, int fixed_point_position, int input_offset=0) |
| Perform a convolve3x3 on qint16. More...
|
|
template<> |
qint16x8x2_t | convolve_3x3< 1 > (const qint8_t *in_top, const qint8_t *in_mid, const qint8_t *in_low, const qint8x8x3_t &m0, const qint8x8x3_t &m1, const qint8x8x3_t &m2, int fixed_point_position, int input_offset) |
|
template<> |
qint16x8x2_t | convolve_3x3< 2 > (const qint8_t *in_top, const qint8_t *in_mid, const qint8_t *in_low, const qint8x8x3_t &m0, const qint8x8x3_t &m1, const qint8x8x3_t &m2, int fixed_point_position, int input_offset) |
|
template<> |
qint16x8x2_t | convolve_3x3< 3 > (const qint8_t *in_top, const qint8_t *in_mid, const qint8_t *in_low, const qint8x8x3_t &m0, const qint8x8x3_t &m1, const qint8x8x3_t &m2, int fixed_point_position, int input_offset) |
|
template<unsigned int stridex> |
int32x4x2_t | convolve_3x3 (const uint8_t *in_top, const uint8_t *in_mid, const uint8_t *in_low, const int32x4x3_t &m0, const int32x4x3_t &m1, const int32x4x3_t &m2, int fixed_point_position, int input_offset) |
| Perform a convolve3x3 on uint8_t. More...
|
|
template<> |
int32x4x2_t | convolve_3x3< 1 > (const uint8_t *in_top, const uint8_t *in_mid, const uint8_t *in_low, const int32x4x3_t &m0, const int32x4x3_t &m1, const int32x4x3_t &m2, int fixed_point_position, int input_offset) |
|
template<> |
int32x4x2_t | convolve_3x3< 2 > (const uint8_t *in_top, const uint8_t *in_mid, const uint8_t *in_low, const int32x4x3_t &m0, const int32x4x3_t &m1, const int32x4x3_t &m2, int fixed_point_position, int input_offset) |
|
template<> |
int32x4x2_t | convolve_3x3< 3 > (const uint8_t *in_top, const uint8_t *in_mid, const uint8_t *in_low, const int32x4x3_t &m0, const int32x4x3_t &m1, const int32x4x3_t &m2, int fixed_point_position, int input_offset) |
|
template<unsigned int stridex> |
void | store_results (float *buffer, const float32x4x2_t &values) |
| Stores a float32x4x2_t array into a memory location. More...
|
|
template<> |
void | store_results< 1 > (float *buffer, const float32x4x2_t &values) |
|
template<> |
void | store_results< 2 > (float *buffer, const float32x4x2_t &values) |
|
template<> |
void | store_results< 3 > (float *buffer, const float32x4x2_t &values) |
|
template<unsigned int stridex> |
void | store_results (qint16_t *buffer, const qint16x8x2_t &values) |
| Stores a qint16_t array into a memory location. More...
|
|
template<> |
void | store_results< 1 > (qint16_t *buffer, const qint16x8x2_t &values) |
|
template<> |
void | store_results< 2 > (qint16_t *buffer, const qint16x8x2_t &values) |
|
template<> |
void | store_results< 3 > (qint16_t *buffer, const qint16x8x2_t &values) |
|
template<unsigned int stridex> |
void | store_results (int32_t *buffer, const int32x4x2_t &values) |
| Stores a uint32_t array into a memory location. More...
|
|
template<> |
void | store_results< 1 > (int32_t *buffer, const int32x4x2_t &values) |
|
template<> |
void | store_results< 2 > (int32_t *buffer, const int32x4x2_t &values) |
|
template<> |
void | store_results< 3 > (int32_t *buffer, const int32x4x2_t &values) |
|
template<unsigned int stridex> |
int | get_input_num_elems_processed (unsigned int num_elems_written_per_iteration) |
| Get the number of elements processed on 3x3 convolution. More...
|
|
template<> |
int | get_input_num_elems_processed< 1 > (unsigned int num_elems_written_per_iteration) |
|
template<> |
int | get_input_num_elems_processed< 2 > (unsigned int num_elems_written_per_iteration) |
|
template<> |
int | get_input_num_elems_processed< 3 > (unsigned int num_elems_written_per_iteration) |
|
int | get_input_num_elems_processed (unsigned int num_elems_written_per_iteration, unsigned int stridex) |
|