CMSIS-NN  
CMSIS NN Software Library
 
Loading...
Searching...
No Matches
Macros | Functions
GetBufferSizeNNConv
Public » Convolution Functions

Macros

#define ARM_NN_DW_NT_T_F16_TILE_ROWS   (4)
 
#define ARM_NN_DW_NT_T_F32_TILE_ROWS   (4)
 

Functions

int32_t arm_convolve_s16_get_buffer_size (const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims)
 Get the required buffer size for s16 convolution function.
 
int32_t arm_convolve_wrapper_s16_get_buffer_size (const cmsis_nn_conv_params *conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims)
 Get the required buffer size for arm_convolve_wrapper_s16.
 
int32_t arm_convolve_wrapper_s16_get_buffer_size_dsp (const cmsis_nn_conv_params *conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims)
 Get the required buffer size for arm_convolve_wrapper_s16 for for processors with DSP extension. Refer to arm_convolve_wrapper_s16_get_buffer_size() for function argument details.
 
int32_t arm_convolve_wrapper_s16_get_buffer_size_mve (const cmsis_nn_conv_params *conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims)
 Get the required buffer size for arm_convolve_wrapper_s16 for Arm(R) Helium Architecture case. Refer to arm_convolve_wrapper_s16_get_buffer_size() for function argument details.
 
int32_t arm_convolve_s4_get_buffer_size (const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims)
 Get the required buffer size for s4 convolution function.
 
int32_t arm_convolve_1_x_n_s4_get_buffer_size (const cmsis_nn_conv_params *conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims)
 Get the required additional buffer size for 1xn convolution.
 
int32_t arm_convolve_1x1_s4_fast_get_buffer_size (const cmsis_nn_dims *input_dims)
 Get the required buffer size for arm_convolve_1x1_s4_fast.
 
int32_t arm_convolve_wrapper_s4_get_buffer_size (const cmsis_nn_conv_params *conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims)
 Get the required buffer size for arm_convolve_wrapper_s4.
 
int32_t arm_convolve_wrapper_s4_get_buffer_size_mve (const cmsis_nn_conv_params *conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims)
 Get the required buffer size for arm_convolve_wrapper_s4 for Arm(R) Helium Architecture case. Refer to arm_convolve_wrapper_s4_get_buffer_size() for function argument details.
 
int32_t arm_convolve_wrapper_s4_get_buffer_size_dsp (const cmsis_nn_conv_params *conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims)
 Get the required buffer size for arm_convolve_wrapper_s4 for processors with DSP extension. Refer to arm_convolve_wrapper_s4_get_buffer_size() for function argument details.
 
int32_t arm_convolve_s8_get_buffer_size (const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims)
 Get the required buffer size for s8 convolution function.
 
int32_t arm_convolve_s8_get_buffer_size_mve (const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims)
 Get the required buffer size for optimized s8 convolution. This is for processors with MVE extension. Refer to arm_convolve_s8_get_buffer_size() for function argument details.
 
int32_t arm_convolve_1_x_n_s8_get_buffer_size (const cmsis_nn_conv_params *conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims)
 Get the required additional buffer size for 1xn convolution.
 
int32_t arm_convolve_1x1_s8_fast_get_buffer_size (const cmsis_nn_dims *input_dims)
 Get the required buffer size for arm_convolve_1x1_s8_fast.
 
int32_t arm_convolve_wrapper_s8_get_buffer_size (const cmsis_nn_conv_params *conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims)
 Get the required buffer size for arm_convolve_wrapper_s8.
 
int32_t arm_convolve_wrapper_s8_get_buffer_size_mve (const cmsis_nn_conv_params *conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims)
 Get the required buffer size for arm_convolve_wrapper_s8 for Arm(R) Helium Architecture case. Refer to arm_convolve_wrapper_s8_get_buffer_size() for function argument details.
 
int32_t arm_convolve_wrapper_s8_get_buffer_size_dsp (const cmsis_nn_conv_params *conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims)
 Get the required buffer size for arm_convolve_wrapper_s8 for processors with DSP extension. Refer to arm_convolve_wrapper_s8_get_buffer_size() for function argument details.
 
int32_t arm_depthwise_conv_fast_s16_get_buffer_size (const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims)
 Get the required buffer size for optimized s16 depthwise convolution function with constraint that in_channel equals out_channel.
 
int32_t arm_depthwise_conv_wrapper_s16_get_buffer_size (const cmsis_nn_dw_conv_params *dw_conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims)
 Get size of additional buffer required by arm_depthwise_conv_wrapper_s16()
 
int32_t arm_depthwise_conv_wrapper_s16_get_buffer_size_mve (const cmsis_nn_dw_conv_params *dw_conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims)
 Get size of additional buffer required by arm_depthwise_conv_wrapper_s16() for Arm(R) Helium Architecture case. Refer to arm_depthwise_conv_wrapper_s16_get_buffer_size() for function argument details.
 
int32_t arm_depthwise_conv_wrapper_s16_get_buffer_size_dsp (const cmsis_nn_dw_conv_params *dw_conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims)
 Get size of additional buffer required by arm_depthwise_conv_wrapper_s16() for processors with DSP extension. Refer to arm_depthwise_conv_wrapper_s16_get_buffer_size() for function argument details.
 
int32_t arm_depthwise_conv_s4_opt_get_buffer_size (const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims)
 Get the required buffer size for optimized s4 depthwise convolution function with constraint that in_channel equals out_channel.
 
int32_t arm_depthwise_conv_wrapper_s4_get_buffer_size (const cmsis_nn_dw_conv_params *dw_conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims)
 Get size of additional buffer required by arm_depthwise_conv_wrapper_s4()
 
int32_t arm_depthwise_conv_wrapper_s4_get_buffer_size_dsp (const cmsis_nn_dw_conv_params *dw_conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims)
 Get size of additional buffer required by arm_depthwise_conv_wrapper_s4() for processors with DSP extension. Refer to arm_depthwise_conv_wrapper_s4_get_buffer_size() for function argument details.
 
int32_t arm_depthwise_conv_wrapper_s4_get_buffer_size_mve (const cmsis_nn_dw_conv_params *dw_conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims)
 Get size of additional buffer required by arm_depthwise_conv_wrapper_s4() for Arm(R) Helium Architecture case. Refer to arm_depthwise_conv_wrapper_s4_get_buffer_size() for function argument details.
 
int32_t arm_depthwise_conv_s8_opt_get_buffer_size_mve (const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims)
 Get the required buffer size for optimized s8 depthwise convolution function with constraint that in_channel equals out_channel. This is for processors with MVE extension. Refer to arm_depthwise_conv_s8_opt_get_buffer_size() for function argument details.
 
int32_t arm_depthwise_conv_s8_opt_get_buffer_size_dsp (const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims)
 Get the required buffer size for optimized s8 depthwise convolution function with constraint that in_channel equals out_channel. This is for processors with DSP extension. Refer to arm_depthwise_conv_s8_opt_get_buffer_size() for function argument details.
 
int32_t arm_depthwise_conv_s8_opt_get_buffer_size (const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims)
 Get the required buffer size for optimized s8 depthwise convolution function with constraint that in_channel equals out_channel.
 
int32_t arm_depthwise_conv_wrapper_s8_get_buffer_size (const cmsis_nn_dw_conv_params *dw_conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims)
 Get size of additional buffer required by arm_depthwise_conv_wrapper_s8()
 
int32_t arm_depthwise_conv_wrapper_s8_get_buffer_size_dsp (const cmsis_nn_dw_conv_params *dw_conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims)
 Get size of additional buffer required by arm_depthwise_conv_wrapper_s8() for processors with DSP extension. Refer to arm_depthwise_conv_wrapper_s8_get_buffer_size() for function argument details.
 
int32_t arm_depthwise_conv_wrapper_s8_get_buffer_size_mve (const cmsis_nn_dw_conv_params *dw_conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims)
 Get size of additional buffer required by arm_depthwise_conv_wrapper_s8() for Arm(R) Helium Architecture case. Refer to arm_depthwise_conv_wrapper_s8_get_buffer_size() for function argument details.
 
int32_t arm_transpose_conv_s8_get_buffer_size (const cmsis_nn_transpose_conv_params *transpose_conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *out_dims)
 Get the required buffer size for ctx in s8 transpose conv function.
 
int32_t arm_transpose_conv_s8_get_buffer_size_mve (const cmsis_nn_transpose_conv_params *transpose_conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *out_dims)
 Get size of additional buffer required by arm_transpose_conv_s8() for Arm(R) Helium Architecture case. Refer to arm_transpose_conv_s8_get_buffer_size() for function argument details.
 
int32_t arm_transpose_conv_s8_get_reverse_conv_buffer_size (const cmsis_nn_transpose_conv_params *transpose_conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims)
 Get the required buffer size for output_ctx in s8 transpose conv function.
 
int32_t arm_depthwise_conv_f16_get_buffer_size (const cmsis_nn_dw_conv_params_f16 *dw_conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims, arm_nn_tensor_layout layout)
 Get the temporary buffer size required by depthwise convolution.
 
int32_t arm_depthwise_conv_wrapper_f16_get_buffer_size (const cmsis_nn_dw_conv_params_f16 *dw_conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims)
 Get the buffer size required by the depthwise convolution wrapper.
 
int32_t arm_convolve_f16_get_buffer_size (const cmsis_nn_conv_params_f16 *conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims, arm_nn_tensor_layout layout)
 Get the temporary buffer size required by convolution.
 
int32_t arm_convolve_wrapper_f16_get_buffer_size (const cmsis_nn_conv_params_f16 *conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims)
 Get the buffer size required by the convolution wrapper.
 
int32_t arm_convolve_1x1_f16_get_buffer_size (const cmsis_nn_conv_params_f16 *conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims, arm_nn_tensor_layout layout)
 Get the buffer size required by 1x1 convolution.
 
int32_t arm_convolve_1_x_n_f16_get_buffer_size (const cmsis_nn_conv_params_f16 *conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims, arm_nn_tensor_layout layout)
 Get the buffer size required by 1xN convolution.
 
int32_t arm_fully_connected_f16_get_buffer_size (const cmsis_nn_fc_params_f16 *fc_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims, arm_nn_tensor_layout layout)
 Get the temporary buffer size required by the fully connected layer.
 
int32_t arm_batch_matmul_f16_get_buffer_size (const cmsis_nn_bmm_params_f16 *bmm_params, const cmsis_nn_dims *input_lhs_dims, const cmsis_nn_dims *input_rhs_dims, const cmsis_nn_dims *output_dims)
 Get the temporary buffer size required by batched matrix multiplication.
 
int32_t arm_depthwise_conv_f32_get_buffer_size (const cmsis_nn_dw_conv_params_f32 *dw_conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims, arm_nn_tensor_layout layout)
 Get the temporary buffer size required by depthwise convolution.
 
int32_t arm_depthwise_conv_wrapper_f32_get_buffer_size (const cmsis_nn_dw_conv_params_f32 *dw_conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims)
 Get the buffer size required by the depthwise convolution wrapper.
 
int32_t arm_convolve_f32_get_buffer_size (const cmsis_nn_conv_params_f32 *conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims, arm_nn_tensor_layout layout)
 Get the temporary buffer size required by convolution.
 
int32_t arm_convolve_wrapper_f32_get_buffer_size (const cmsis_nn_conv_params_f32 *conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims)
 Get the buffer size required by the convolution wrapper.
 
int32_t arm_convolve_1x1_f32_get_buffer_size (const cmsis_nn_conv_params_f32 *conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims, arm_nn_tensor_layout layout)
 Get the buffer size required by 1x1 convolution.
 
int32_t arm_convolve_1_x_n_f32_get_buffer_size (const cmsis_nn_conv_params_f32 *conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims, arm_nn_tensor_layout layout)
 Get the buffer size required by 1xN convolution.
 
int32_t arm_fully_connected_f32_get_buffer_size (const cmsis_nn_fc_params_f32 *fc_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *output_dims, arm_nn_tensor_layout layout)
 Get the temporary buffer size required by the fully connected layer.
 
int32_t arm_batch_matmul_f32_get_buffer_size (const cmsis_nn_bmm_params_f32 *bmm_params, const cmsis_nn_dims *input_lhs_dims, const cmsis_nn_dims *input_rhs_dims, const cmsis_nn_dims *output_dims)
 Get the temporary buffer size required by batched matrix multiplication.
 

Description

Macro Definition Documentation

◆ ARM_NN_DW_NT_T_F16_TILE_ROWS

#define ARM_NN_DW_NT_T_F16_TILE_ROWS   (4)

◆ ARM_NN_DW_NT_T_F32_TILE_ROWS

#define ARM_NN_DW_NT_T_F32_TILE_ROWS   (4)

Function Documentation

◆ arm_batch_matmul_f16_get_buffer_size()

int32_t arm_batch_matmul_f16_get_buffer_size ( const cmsis_nn_bmm_params_f16 *  bmm_params,
const cmsis_nn_dims *  input_lhs_dims,
const cmsis_nn_dims *  input_rhs_dims,
const cmsis_nn_dims *  output_dims 
)

Get the temporary buffer size required by batched matrix multiplication.

Parameters
[in]bmm_paramsBatch matmul parameters.
[in]input_lhs_dimsLeft-hand-side input tensor dimensions.
[in]input_rhs_dimsRight-hand-side input tensor dimensions.
[in]output_dimsOutput tensor dimensions.
Returns
Required buffer size in bytes, or 0 when no scratch buffer is needed.

◆ arm_batch_matmul_f32_get_buffer_size()

int32_t arm_batch_matmul_f32_get_buffer_size ( const cmsis_nn_bmm_params_f32 *  bmm_params,
const cmsis_nn_dims *  input_lhs_dims,
const cmsis_nn_dims *  input_rhs_dims,
const cmsis_nn_dims *  output_dims 
)

Get the temporary buffer size required by batched matrix multiplication.

Parameters
[in]bmm_paramsBatch matmul parameters.
[in]input_lhs_dimsLeft-hand-side input tensor dimensions.
[in]input_rhs_dimsRight-hand-side input tensor dimensions.
[in]output_dimsOutput tensor dimensions.
Returns
Required buffer size in bytes, or 0 when no scratch buffer is needed.

◆ arm_convolve_1_x_n_f16_get_buffer_size()

int32_t arm_convolve_1_x_n_f16_get_buffer_size ( const cmsis_nn_conv_params_f16 *  conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims,
arm_nn_tensor_layout  layout 
)

Get the buffer size required by 1xN convolution.

◆ arm_convolve_1_x_n_f32_get_buffer_size()

int32_t arm_convolve_1_x_n_f32_get_buffer_size ( const cmsis_nn_conv_params_f32 *  conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims,
arm_nn_tensor_layout  layout 
)

Get the buffer size required by 1xN convolution.

◆ arm_convolve_1_x_n_s4_get_buffer_size()

int32_t arm_convolve_1_x_n_s4_get_buffer_size ( const cmsis_nn_conv_params *  conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims 
)

Get the required additional buffer size for 1xn convolution.

Parameters
[in]conv_paramsConvolution parameters (e.g. strides, dilations, pads,...). Range of conv_params->input_offset : [-127, 128] Range of conv_params->output_offset : [-128, 127]
[in]input_dimsInput (activation) tensor dimensions. Format: [N, H, W, C_IN]
[in]filter_dimsFilter tensor dimensions. Format: [C_OUT, 1, WK, C_IN] where WK is the horizontal spatial filter dimension
[in]output_dimsOutput tensor dimensions. Format: [N, H, W, C_OUT]
Returns
The function returns required buffer size(bytes)

◆ arm_convolve_1_x_n_s8_get_buffer_size()

int32_t arm_convolve_1_x_n_s8_get_buffer_size ( const cmsis_nn_conv_params *  conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims 
)

Get the required additional buffer size for 1xn convolution.

Parameters
[in]conv_paramsConvolution parameters (e.g. strides, dilations, pads,...). Range of conv_params->input_offset : [-127, 128] Range of conv_params->output_offset : [-128, 127]
[in]input_dimsInput (activation) tensor dimensions. Format: [N, H, W, C_IN]
[in]filter_dimsFilter tensor dimensions. Format: [C_OUT, 1, WK, C_IN] where WK is the horizontal spatial filter dimension
[in]output_dimsOutput tensor dimensions. Format: [N, H, W, C_OUT]
Returns
The function returns required buffer size(bytes)

◆ arm_convolve_1x1_f16_get_buffer_size()

int32_t arm_convolve_1x1_f16_get_buffer_size ( const cmsis_nn_conv_params_f16 *  conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims,
arm_nn_tensor_layout  layout 
)

Get the buffer size required by 1x1 convolution.

Note
Returns 0 for the unity-stride no-pack path. For non-unity-stride NHWC 1x1 convolution, the returned scratch size enables the packed-tile + GEMM path. When conv_params->weight_format is ARM_NN_WEIGHT_FORMAT_NT_N_PACKED, this still excludes the offline-packed filter storage itself.

◆ arm_convolve_1x1_f32_get_buffer_size()

int32_t arm_convolve_1x1_f32_get_buffer_size ( const cmsis_nn_conv_params_f32 *  conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims,
arm_nn_tensor_layout  layout 
)

Get the buffer size required by 1x1 convolution.

Note
Returns 0 for the unity-stride no-pack path. For non-unity-stride NHWC 1x1 convolution, the returned scratch size enables the packed-tile + GEMM path. When conv_params->weight_format is ARM_NN_WEIGHT_FORMAT_NT_N_PACKED, this still excludes the offline-packed filter storage itself.

◆ arm_convolve_1x1_s4_fast_get_buffer_size()

int32_t arm_convolve_1x1_s4_fast_get_buffer_size ( const cmsis_nn_dims *  input_dims)

Get the required buffer size for arm_convolve_1x1_s4_fast.

Parameters
[in]input_dimsInput (activation) dimensions
Returns
The function returns the required buffer size in bytes

◆ arm_convolve_1x1_s8_fast_get_buffer_size()

int32_t arm_convolve_1x1_s8_fast_get_buffer_size ( const cmsis_nn_dims *  input_dims)

Get the required buffer size for arm_convolve_1x1_s8_fast.

Parameters
[in]input_dimsInput (activation) dimensions
Returns
The function returns the required buffer size in bytes

◆ arm_convolve_f16_get_buffer_size()

int32_t arm_convolve_f16_get_buffer_size ( const cmsis_nn_conv_params_f16 *  conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims,
arm_nn_tensor_layout  layout 
)

Get the temporary buffer size required by convolution.

Parameters
[in]conv_paramsConvolution parameters.
[in]input_dimsInput tensor dimensions.
[in]filter_dimsFilter tensor dimensions.
[in]output_dimsOutput tensor dimensions.
[in]layoutTensor layout selector.
Note
When conv_params->weight_format is ARM_NN_WEIGHT_FORMAT_NT_N_PACKED, this still reports only the temporary input/im2col scratch requirement. Any offline-packed filter storage is expected to be provided by the caller.
Returns
Required buffer size in bytes, or 0 when no scratch buffer is needed.

◆ arm_convolve_f32_get_buffer_size()

int32_t arm_convolve_f32_get_buffer_size ( const cmsis_nn_conv_params_f32 *  conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims,
arm_nn_tensor_layout  layout 
)

Get the temporary buffer size required by convolution.

Parameters
[in]conv_paramsConvolution parameters.
[in]input_dimsInput tensor dimensions.
[in]filter_dimsFilter tensor dimensions.
[in]output_dimsOutput tensor dimensions.
[in]layoutTensor layout selector.
Note
When conv_params->weight_format is ARM_NN_WEIGHT_FORMAT_NT_N_PACKED, this still reports only the temporary input/im2col scratch requirement. Any offline-packed filter storage is expected to be provided by the caller.
Returns
Required buffer size in bytes, or 0 when no scratch buffer is needed.

◆ arm_convolve_s16_get_buffer_size()

int32_t arm_convolve_s16_get_buffer_size ( const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims 
)

Get the required buffer size for s16 convolution function.

Parameters
[in]input_dimsInput (activation) tensor dimensions. Format: [N, H, W, C_IN]
[in]filter_dimsFilter tensor dimensions. Format: [C_OUT, HK, WK, C_IN] where HK and WK are the spatial filter dimensions
Returns
The function returns required buffer size(bytes)

◆ arm_convolve_s4_get_buffer_size()

int32_t arm_convolve_s4_get_buffer_size ( const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims 
)

Get the required buffer size for s4 convolution function.

Parameters
[in]input_dimsInput (activation) tensor dimensions. Format: [N, H, W, C_IN]
[in]filter_dimsFilter tensor dimensions. Format: [C_OUT, HK, WK, C_IN] where HK and WK are the spatial filter dimensions
Returns
The function returns required buffer size(bytes)

◆ arm_convolve_s8_get_buffer_size()

int32_t arm_convolve_s8_get_buffer_size ( const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims 
)

Get the required buffer size for s8 convolution function.

Parameters
[in]input_dimsInput (activation) tensor dimensions. Format: [N, H, W, C_IN]
[in]filter_dimsFilter tensor dimensions. Format: [C_OUT, HK, WK, C_IN] where HK and WK are the spatial filter dimensions
Returns
The function returns required buffer size(bytes)

◆ arm_convolve_s8_get_buffer_size_mve()

int32_t arm_convolve_s8_get_buffer_size_mve ( const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims 
)

Get the required buffer size for optimized s8 convolution. This is for processors with MVE extension. Refer to arm_convolve_s8_get_buffer_size() for function argument details.

Note
Intended for compilation on Host. If compiling for an Arm target, use arm_convolve_s8_get_buffer_size(). Note also this is a support function, so not recommended to call directly even on Host.

◆ arm_convolve_wrapper_f16_get_buffer_size()

int32_t arm_convolve_wrapper_f16_get_buffer_size ( const cmsis_nn_conv_params_f16 *  conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims 
)

Get the buffer size required by the convolution wrapper.

◆ arm_convolve_wrapper_f32_get_buffer_size()

int32_t arm_convolve_wrapper_f32_get_buffer_size ( const cmsis_nn_conv_params_f32 *  conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims 
)

Get the buffer size required by the convolution wrapper.

◆ arm_convolve_wrapper_s16_get_buffer_size()

int32_t arm_convolve_wrapper_s16_get_buffer_size ( const cmsis_nn_conv_params *  conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims 
)

Get the required buffer size for arm_convolve_wrapper_s16.

Parameters
[in]conv_paramsConvolution parameters (e.g. strides, dilations, pads,...). conv_params->input_offset : Not used conv_params->output_offset : Not used
[in]input_dimsInput (activation) dimensions. Format: [N, H, W, C_IN]
[in]filter_dimsFilter dimensions. Format: [C_OUT, HK, WK, C_IN] where HK and WK are the spatial filter dimensions
[in]output_dimsOutput tensor dimensions. Format: [N, H, W, C_OUT]
Returns
The function returns required buffer size(bytes)

◆ arm_convolve_wrapper_s16_get_buffer_size_dsp()

int32_t arm_convolve_wrapper_s16_get_buffer_size_dsp ( const cmsis_nn_conv_params *  conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims 
)

Get the required buffer size for arm_convolve_wrapper_s16 for for processors with DSP extension. Refer to arm_convolve_wrapper_s16_get_buffer_size() for function argument details.

Note
Intended for compilation on Host. If compiling for an Arm target, use arm_convolve_wrapper_s16_get_buffer_size().

◆ arm_convolve_wrapper_s16_get_buffer_size_mve()

int32_t arm_convolve_wrapper_s16_get_buffer_size_mve ( const cmsis_nn_conv_params *  conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims 
)

Get the required buffer size for arm_convolve_wrapper_s16 for Arm(R) Helium Architecture case. Refer to arm_convolve_wrapper_s16_get_buffer_size() for function argument details.

Note
Intended for compilation on Host. If compiling for an Arm target, use arm_convolve_wrapper_s16_get_buffer_size().

◆ arm_convolve_wrapper_s4_get_buffer_size()

int32_t arm_convolve_wrapper_s4_get_buffer_size ( const cmsis_nn_conv_params *  conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims 
)

Get the required buffer size for arm_convolve_wrapper_s4.

Parameters
[in]conv_paramsConvolution parameters (e.g. strides, dilations, pads,...). Range of conv_params->input_offset : [-127, 128] Range of conv_params->output_offset : [-128, 127]
[in]input_dimsInput (activation) dimensions. Format: [N, H, W, C_IN]
[in]filter_dimsFilter dimensions. Format: [C_OUT, HK, WK, C_IN] where HK and WK are the spatial filter dimensions
[in]output_dimsOutput tensor dimensions. Format: [N, H, W, C_OUT]
Returns
The function returns required buffer size(bytes)

◆ arm_convolve_wrapper_s4_get_buffer_size_dsp()

int32_t arm_convolve_wrapper_s4_get_buffer_size_dsp ( const cmsis_nn_conv_params *  conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims 
)

Get the required buffer size for arm_convolve_wrapper_s4 for processors with DSP extension. Refer to arm_convolve_wrapper_s4_get_buffer_size() for function argument details.

Note
Intended for compilation on Host. If compiling for an Arm target, use arm_convolve_wrapper_s4_get_buffer_size().

◆ arm_convolve_wrapper_s4_get_buffer_size_mve()

int32_t arm_convolve_wrapper_s4_get_buffer_size_mve ( const cmsis_nn_conv_params *  conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims 
)

Get the required buffer size for arm_convolve_wrapper_s4 for Arm(R) Helium Architecture case. Refer to arm_convolve_wrapper_s4_get_buffer_size() for function argument details.

Note
Intended for compilation on Host. If compiling for an Arm target, use arm_convolve_wrapper_s4_get_buffer_size(). Currently this operator does not have an mve implementation, so dsp will be used.

◆ arm_convolve_wrapper_s8_get_buffer_size()

int32_t arm_convolve_wrapper_s8_get_buffer_size ( const cmsis_nn_conv_params *  conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims 
)

Get the required buffer size for arm_convolve_wrapper_s8.

Parameters
[in]conv_paramsConvolution parameters (e.g. strides, dilations, pads,...). Range of conv_params->input_offset : [-127, 128] Range of conv_params->output_offset : [-128, 127]
[in]input_dimsInput (activation) dimensions. Format: [N, H, W, C_IN]
[in]filter_dimsFilter dimensions. Format: [C_OUT, HK, WK, C_IN] where HK and WK are the spatial filter dimensions
[in]output_dimsOutput tensor dimensions. Format: [N, H, W, C_OUT]
Returns
The function returns required buffer size(bytes)

◆ arm_convolve_wrapper_s8_get_buffer_size_dsp()

int32_t arm_convolve_wrapper_s8_get_buffer_size_dsp ( const cmsis_nn_conv_params *  conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims 
)

Get the required buffer size for arm_convolve_wrapper_s8 for processors with DSP extension. Refer to arm_convolve_wrapper_s8_get_buffer_size() for function argument details.

Note
Intended for compilation on Host. If compiling for an Arm target, use arm_convolve_wrapper_s8_get_buffer_size().

◆ arm_convolve_wrapper_s8_get_buffer_size_mve()

int32_t arm_convolve_wrapper_s8_get_buffer_size_mve ( const cmsis_nn_conv_params *  conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims 
)

Get the required buffer size for arm_convolve_wrapper_s8 for Arm(R) Helium Architecture case. Refer to arm_convolve_wrapper_s8_get_buffer_size() for function argument details.

Note
Intended for compilation on Host. If compiling for an Arm target, use arm_convolve_wrapper_s8_get_buffer_size().

◆ arm_depthwise_conv_f16_get_buffer_size()

int32_t arm_depthwise_conv_f16_get_buffer_size ( const cmsis_nn_dw_conv_params_f16 *  dw_conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims,
arm_nn_tensor_layout  layout 
)

Get the temporary buffer size required by depthwise convolution.

Parameters
[in]dw_conv_paramsDepthwise convolution parameters.
[in]input_dimsInput tensor dimensions.
[in]filter_dimsFilter tensor dimensions.
[in]output_dimsOutput tensor dimensions.
[in]layoutTensor layout selector.
Returns
Required buffer size in bytes, or 0 when no scratch buffer is needed.

◆ arm_depthwise_conv_f32_get_buffer_size()

int32_t arm_depthwise_conv_f32_get_buffer_size ( const cmsis_nn_dw_conv_params_f32 *  dw_conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims,
arm_nn_tensor_layout  layout 
)

Get the temporary buffer size required by depthwise convolution.

Parameters
[in]dw_conv_paramsDepthwise convolution parameters.
[in]input_dimsInput tensor dimensions.
[in]filter_dimsFilter tensor dimensions.
[in]output_dimsOutput tensor dimensions.
[in]layoutTensor layout selector.
Returns
Required buffer size in bytes, or 0 when no scratch buffer is needed.

◆ arm_depthwise_conv_fast_s16_get_buffer_size()

int32_t arm_depthwise_conv_fast_s16_get_buffer_size ( const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims 
)

Get the required buffer size for optimized s16 depthwise convolution function with constraint that in_channel equals out_channel.

Parameters
[in]input_dimsInput (activation) tensor dimensions. Format: [1, H, W, C_IN] Batch argument N is not used.
[in]filter_dimsFilter tensor dimensions. Format: [1, H, W, C_OUT]
Returns
The function returns required buffer size in bytes

◆ arm_depthwise_conv_s4_opt_get_buffer_size()

int32_t arm_depthwise_conv_s4_opt_get_buffer_size ( const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims 
)

Get the required buffer size for optimized s4 depthwise convolution function with constraint that in_channel equals out_channel.

Parameters
[in]input_dimsInput (activation) tensor dimensions. Format: [1, H, W, C_IN] Batch argument N is not used.
[in]filter_dimsFilter tensor dimensions. Format: [1, H, W, C_OUT]
Returns
The function returns required buffer size in bytes

◆ arm_depthwise_conv_s8_opt_get_buffer_size()

int32_t arm_depthwise_conv_s8_opt_get_buffer_size ( const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims 
)

Get the required buffer size for optimized s8 depthwise convolution function with constraint that in_channel equals out_channel.

Parameters
[in]input_dimsInput (activation) tensor dimensions. Format: [1, H, W, C_IN] Batch argument N is not used.
[in]filter_dimsFilter tensor dimensions. Format: [1, H, W, C_OUT]
Returns
The function returns required buffer size in bytes

◆ arm_depthwise_conv_s8_opt_get_buffer_size_dsp()

int32_t arm_depthwise_conv_s8_opt_get_buffer_size_dsp ( const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims 
)

Get the required buffer size for optimized s8 depthwise convolution function with constraint that in_channel equals out_channel. This is for processors with DSP extension. Refer to arm_depthwise_conv_s8_opt_get_buffer_size() for function argument details.

Note
Intended for compilation on Host. If compiling for an Arm target, use arm_depthwise_conv_s8_opt_get_buffer_size(). Note also this is a support function, so not recommended to call directly even on Host.

◆ arm_depthwise_conv_s8_opt_get_buffer_size_mve()

int32_t arm_depthwise_conv_s8_opt_get_buffer_size_mve ( const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims 
)

Get the required buffer size for optimized s8 depthwise convolution function with constraint that in_channel equals out_channel. This is for processors with MVE extension. Refer to arm_depthwise_conv_s8_opt_get_buffer_size() for function argument details.

Note
Intended for compilation on Host. If compiling for an Arm target, use arm_depthwise_conv_s8_opt_get_buffer_size(). Note also this is a support function, so not recommended to call directly even on Host.

◆ arm_depthwise_conv_wrapper_f16_get_buffer_size()

int32_t arm_depthwise_conv_wrapper_f16_get_buffer_size ( const cmsis_nn_dw_conv_params_f16 *  dw_conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims 
)

Get the buffer size required by the depthwise convolution wrapper.

◆ arm_depthwise_conv_wrapper_f32_get_buffer_size()

int32_t arm_depthwise_conv_wrapper_f32_get_buffer_size ( const cmsis_nn_dw_conv_params_f32 *  dw_conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims 
)

Get the buffer size required by the depthwise convolution wrapper.

◆ arm_depthwise_conv_wrapper_s16_get_buffer_size()

int32_t arm_depthwise_conv_wrapper_s16_get_buffer_size ( const cmsis_nn_dw_conv_params *  dw_conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims 
)

Get size of additional buffer required by arm_depthwise_conv_wrapper_s16()

Parameters
[in]dw_conv_paramsDepthwise convolution parameters (e.g. strides, dilations, pads,...) Range of dw_conv_params->input_offset : Not used Range of dw_conv_params->input_offset : Not used
[in]input_dimsInput (activation) tensor dimensions. Format: [H, W, C_IN] Batch argument N is not used and assumed to be 1.
[in]filter_dimsFilter tensor dimensions. Format: [1, H, W, C_OUT]
[in]output_dimsOutput tensor dimensions. Format: [1, H, W, C_OUT]
Returns
Size of additional memory required for optimizations in bytes.

◆ arm_depthwise_conv_wrapper_s16_get_buffer_size_dsp()

int32_t arm_depthwise_conv_wrapper_s16_get_buffer_size_dsp ( const cmsis_nn_dw_conv_params *  dw_conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims 
)

Get size of additional buffer required by arm_depthwise_conv_wrapper_s16() for processors with DSP extension. Refer to arm_depthwise_conv_wrapper_s16_get_buffer_size() for function argument details.

Note
Intended for compilation on Host. If compiling for an Arm target, use arm_depthwise_conv_wrapper_s16_get_buffer_size().

◆ arm_depthwise_conv_wrapper_s16_get_buffer_size_mve()

int32_t arm_depthwise_conv_wrapper_s16_get_buffer_size_mve ( const cmsis_nn_dw_conv_params *  dw_conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims 
)

Get size of additional buffer required by arm_depthwise_conv_wrapper_s16() for Arm(R) Helium Architecture case. Refer to arm_depthwise_conv_wrapper_s16_get_buffer_size() for function argument details.

Note
Intended for compilation on Host. If compiling for an Arm target, use arm_depthwise_conv_wrapper_s16_get_buffer_size().

◆ arm_depthwise_conv_wrapper_s4_get_buffer_size()

int32_t arm_depthwise_conv_wrapper_s4_get_buffer_size ( const cmsis_nn_dw_conv_params *  dw_conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims 
)

Get size of additional buffer required by arm_depthwise_conv_wrapper_s4()

Parameters
[in]dw_conv_paramsDepthwise convolution parameters (e.g. strides, dilations, pads,...) Range of dw_conv_params->input_offset : [-127, 128] Range of dw_conv_params->input_offset : [-128, 127]
[in]input_dimsInput (activation) tensor dimensions. Format: [H, W, C_IN] Batch argument N is not used and assumed to be 1.
[in]filter_dimsFilter tensor dimensions. Format: [1, H, W, C_OUT]
[in]output_dimsOutput tensor dimensions. Format: [1, H, W, C_OUT]
Returns
Size of additional memory required for optimizations in bytes.

◆ arm_depthwise_conv_wrapper_s4_get_buffer_size_dsp()

int32_t arm_depthwise_conv_wrapper_s4_get_buffer_size_dsp ( const cmsis_nn_dw_conv_params *  dw_conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims 
)

Get size of additional buffer required by arm_depthwise_conv_wrapper_s4() for processors with DSP extension. Refer to arm_depthwise_conv_wrapper_s4_get_buffer_size() for function argument details.

Note
Intended for compilation on Host. If compiling for an Arm target, use arm_depthwise_conv_wrapper_s4_get_buffer_size().

◆ arm_depthwise_conv_wrapper_s4_get_buffer_size_mve()

int32_t arm_depthwise_conv_wrapper_s4_get_buffer_size_mve ( const cmsis_nn_dw_conv_params *  dw_conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims 
)

Get size of additional buffer required by arm_depthwise_conv_wrapper_s4() for Arm(R) Helium Architecture case. Refer to arm_depthwise_conv_wrapper_s4_get_buffer_size() for function argument details.

Note
Intended for compilation on Host. If compiling for an Arm target, use arm_depthwise_conv_wrapper_s4_get_buffer_size().

◆ arm_depthwise_conv_wrapper_s8_get_buffer_size()

int32_t arm_depthwise_conv_wrapper_s8_get_buffer_size ( const cmsis_nn_dw_conv_params *  dw_conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims 
)

Get size of additional buffer required by arm_depthwise_conv_wrapper_s8()

Parameters
[in]dw_conv_paramsDepthwise convolution parameters (e.g. strides, dilations, pads,...) Range of dw_conv_params->input_offset : [-127, 128] Range of dw_conv_params->input_offset : [-128, 127]
[in]input_dimsInput (activation) tensor dimensions. Format: [H, W, C_IN] Batch argument N is not used and assumed to be 1.
[in]filter_dimsFilter tensor dimensions. Format: [1, H, W, C_OUT]
[in]output_dimsOutput tensor dimensions. Format: [1, H, W, C_OUT]
Returns
Size of additional memory required for optimizations in bytes.

◆ arm_depthwise_conv_wrapper_s8_get_buffer_size_dsp()

int32_t arm_depthwise_conv_wrapper_s8_get_buffer_size_dsp ( const cmsis_nn_dw_conv_params *  dw_conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims 
)

Get size of additional buffer required by arm_depthwise_conv_wrapper_s8() for processors with DSP extension. Refer to arm_depthwise_conv_wrapper_s8_get_buffer_size() for function argument details.

Note
Intended for compilation on Host. If compiling for an Arm target, use arm_depthwise_conv_wrapper_s8_get_buffer_size().

◆ arm_depthwise_conv_wrapper_s8_get_buffer_size_mve()

int32_t arm_depthwise_conv_wrapper_s8_get_buffer_size_mve ( const cmsis_nn_dw_conv_params *  dw_conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims 
)

Get size of additional buffer required by arm_depthwise_conv_wrapper_s8() for Arm(R) Helium Architecture case. Refer to arm_depthwise_conv_wrapper_s8_get_buffer_size() for function argument details.

Note
Intended for compilation on Host. If compiling for an Arm target, use arm_depthwise_conv_wrapper_s8_get_buffer_size().

◆ arm_fully_connected_f16_get_buffer_size()

int32_t arm_fully_connected_f16_get_buffer_size ( const cmsis_nn_fc_params_f16 *  fc_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims,
arm_nn_tensor_layout  layout 
)

Get the temporary buffer size required by the fully connected layer.

Parameters
[in]fc_paramsFully connected parameters.
[in]input_dimsInput tensor dimensions.
[in]filter_dimsFilter tensor dimensions.
[in]output_dimsOutput tensor dimensions.
[in]layoutTensor layout selector.
Returns
Required buffer size in bytes, or 0 when no scratch buffer is needed.

◆ arm_fully_connected_f32_get_buffer_size()

int32_t arm_fully_connected_f32_get_buffer_size ( const cmsis_nn_fc_params_f32 *  fc_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  output_dims,
arm_nn_tensor_layout  layout 
)

Get the temporary buffer size required by the fully connected layer.

Parameters
[in]fc_paramsFully connected parameters.
[in]input_dimsInput tensor dimensions.
[in]filter_dimsFilter tensor dimensions.
[in]output_dimsOutput tensor dimensions.
[in]layoutTensor layout selector.
Returns
Required buffer size in bytes, or 0 when no scratch buffer is needed.

◆ arm_transpose_conv_s8_get_buffer_size()

int32_t arm_transpose_conv_s8_get_buffer_size ( const cmsis_nn_transpose_conv_params *  transposed_conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  out_dims 
)

Get the required buffer size for ctx in s8 transpose conv function.

Parameters
[in]transposed_conv_paramsTransposed convolution parameters
[in]input_dimsInput (activation) tensor dimensions. Format: [N, H, W, C_IN]
[in]filter_dimsFilter tensor dimensions. Format: [C_OUT, HK, WK, C_IN] where HK and WK are the spatial filter dimensions
[in]out_dimsOutput tensor dimensions. Format: [N, H, W, C_OUT]
Returns
The function returns required buffer size(bytes)

◆ arm_transpose_conv_s8_get_buffer_size_mve()

int32_t arm_transpose_conv_s8_get_buffer_size_mve ( const cmsis_nn_transpose_conv_params *  transposed_conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims,
const cmsis_nn_dims *  out_dims 
)

Get size of additional buffer required by arm_transpose_conv_s8() for Arm(R) Helium Architecture case. Refer to arm_transpose_conv_s8_get_buffer_size() for function argument details.

Note
Intended for compilation on Host. If compiling for an Arm target, use arm_transpose_conv_s8_get_buffer_size().

◆ arm_transpose_conv_s8_get_reverse_conv_buffer_size()

int32_t arm_transpose_conv_s8_get_reverse_conv_buffer_size ( const cmsis_nn_transpose_conv_params *  transposed_conv_params,
const cmsis_nn_dims *  input_dims,
const cmsis_nn_dims *  filter_dims 
)

Get the required buffer size for output_ctx in s8 transpose conv function.

Parameters
[in]transposed_conv_paramsTransposed convolution parameters
[in]input_dimsInput (activation) tensor dimensions. Format: [N, H, W, C_IN]
[in]filter_dimsFilter tensor dimensions. Format: [C_OUT, HK, WK, C_IN] where HK and WK are the spatial filter dimensions
Returns
The function returns required buffer size(bytes)