Convolution Functions

Content
	GetBufferSizeNNConv

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

Functions
arm_cmsis_nn_status	arm_depthwise_nhwc_conv_f32 (const cmsis_nn_context *ctx, const cmsis_nn_dw_conv_params_f32 *dw_conv_params, const cmsis_nn_dims *input_dims, const float32_t *input, const cmsis_nn_dims *filter_dims, const float32_t *kernel, const cmsis_nn_dims *bias_dims, const float32_t *bias, const cmsis_nn_dims *output_dims, float32_t *output)
	Depthwise convolution, NHWC layout.
arm_cmsis_nn_status	arm_depthwise_conv_f32 (const cmsis_nn_context *ctx, const cmsis_nn_dw_conv_params_f32 *dw_conv_params, const cmsis_nn_dims *input_dims, const float32_t *input, const cmsis_nn_dims *filter_dims, const float32_t *kernel, const cmsis_nn_dims *bias_dims, const float32_t *bias, const cmsis_nn_dims *output_dims, float32_t *output, arm_nn_tensor_layout layout)
	Depthwise convolution, dispatch by layout.
arm_cmsis_nn_status	arm_depthwise_conv_wrapper_f32 (const cmsis_nn_context *ctx, const cmsis_nn_dw_conv_params_f32 *dw_conv_params, const cmsis_nn_dims *input_dims, const float32_t *input, const cmsis_nn_dims *filter_dims, const float32_t *kernel, const cmsis_nn_dims *bias_dims, const float32_t *bias, const cmsis_nn_dims *output_dims, float32_t *output)
	Depthwise convolution wrapper using the CMSIS-NN baseline path.
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.
arm_cmsis_nn_status	arm_convolve_nhwc_f32 (const cmsis_nn_context *ctx, const cmsis_nn_conv_params_f32 *conv_params, const cmsis_nn_dims *input_dims, const float32_t *input_data, const cmsis_nn_dims *filter_dims, const float32_t *filter_data, const cmsis_nn_dims *bias_dims, const float32_t *bias_data, const cmsis_nn_dims *output_dims, float32_t *output_data)
	Convolution, NHWC layout.
arm_cmsis_nn_status	arm_convolve_f32 (const cmsis_nn_context *ctx, const cmsis_nn_conv_params_f32 *conv_params, const cmsis_nn_dims *input_dims, const float32_t *input_data, const cmsis_nn_dims *filter_dims, const float32_t *filter_data, const cmsis_nn_dims *bias_dims, const float32_t *bias_data, const cmsis_nn_dims *output_dims, float32_t *output_data, arm_nn_tensor_layout layout)
	Convolution, dispatch by layout.
arm_cmsis_nn_status	arm_convolve_wrapper_f32 (const cmsis_nn_context *ctx, const cmsis_nn_conv_params_f32 *conv_params, const cmsis_nn_dims *input_dims, const float32_t *input_data, const cmsis_nn_dims *filter_dims, const float32_t *filter_data, const cmsis_nn_dims *bias_dims, const float32_t *bias_data, const cmsis_nn_dims *output_dims, float32_t *output_data)
	Convolution wrapper using the CMSIS-NN baseline path.
arm_cmsis_nn_status	arm_convolve_1x1_nhwc_f32 (const cmsis_nn_context *ctx, const cmsis_nn_conv_params_f32 *conv_params, const cmsis_nn_dims *input_dims, const float32_t *input_data, const cmsis_nn_dims *filter_dims, const float32_t *filter_data, const cmsis_nn_dims *bias_dims, const float32_t *bias_data, const cmsis_nn_dims *output_dims, float32_t *output_data)
	1x1 convolution, NHWC layout.
arm_cmsis_nn_status	arm_convolve_1x1_f32 (const cmsis_nn_context *ctx, const cmsis_nn_conv_params_f32 *conv_params, const cmsis_nn_dims *input_dims, const float32_t *input_data, const cmsis_nn_dims *filter_dims, const float32_t *filter_data, const cmsis_nn_dims *bias_dims, const float32_t *bias_data, const cmsis_nn_dims *output_dims, float32_t *output_data, arm_nn_tensor_layout layout)
	1x1 convolution, dispatch by layout.
arm_cmsis_nn_status	arm_convolve_1_x_n_nhwc_f32 (const cmsis_nn_context *ctx, const cmsis_nn_conv_params_f32 *conv_params, const cmsis_nn_dims *input_dims, const float32_t *input_data, const cmsis_nn_dims *filter_dims, const float32_t *filter_data, const cmsis_nn_dims *bias_dims, const float32_t *bias_data, const cmsis_nn_dims *output_dims, float32_t *output_data)
	1xN convolution, NHWC layout.
arm_cmsis_nn_status	arm_convolve_1_x_n_f32 (const cmsis_nn_context *ctx, const cmsis_nn_conv_params_f32 *conv_params, const cmsis_nn_dims *input_dims, const float32_t *input_data, const cmsis_nn_dims *filter_dims, const float32_t *filter_data, const cmsis_nn_dims *bias_dims, const float32_t *bias_data, const cmsis_nn_dims *output_dims, float32_t *output_data, arm_nn_tensor_layout layout)
	1xN convolution, dispatch by layout.
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.
arm_cmsis_nn_status	arm_transpose_conv_wrapper_f32 (const cmsis_nn_context *ctx, const cmsis_nn_context *output_ctx, const cmsis_nn_transpose_conv_params_f32 *transpose_conv_params, const cmsis_nn_dims *input_dims, const float32_t *input_data, const cmsis_nn_dims *filter_dims, const float32_t *filter_data, const cmsis_nn_dims *bias_dims, const float32_t *bias_data, const cmsis_nn_dims *output_dims, float32_t *output_data, arm_nn_tensor_layout layout)
	Transpose convolution wrapper using the CMSIS-NN baseline path.
arm_cmsis_nn_status	arm_transpose_conv_nhwc_f32 (const cmsis_nn_context *ctx, const cmsis_nn_context *output_ctx, const cmsis_nn_transpose_conv_params_f32 *transpose_conv_params, const cmsis_nn_dims *input_dims, const float32_t *input_data, const cmsis_nn_dims *filter_dims, const float32_t *filter_data, const cmsis_nn_dims *bias_dims, const float32_t *bias_data, const cmsis_nn_dims *output_dims, float32_t *output_data)
	Transpose convolution, NHWC layout.
arm_cmsis_nn_status	arm_transpose_conv_f32 (const cmsis_nn_context *ctx, const cmsis_nn_context *output_ctx, const cmsis_nn_transpose_conv_params_f32 *transpose_conv_params, const cmsis_nn_dims *input_dims, const float32_t *input_data, const cmsis_nn_dims *filter_dims, const float32_t *filter_data, const cmsis_nn_dims *bias_dims, const float32_t *bias_data, const cmsis_nn_dims *output_dims, float32_t *output_data, arm_nn_tensor_layout layout)
	Transpose convolution, dispatch by layout.
int32_t	arm_transpose_conv_f32_get_buffer_size (const cmsis_nn_transpose_conv_params_f32 *transpose_conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *out_dims)
	Get the temporary buffer size required by transpose convolution.
int32_t	arm_transpose_conv_f32_get_reverse_conv_buffer_size (const cmsis_nn_transpose_conv_params_f32 *transpose_conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims)
	Get the reverse-convolution workspace size used by transpose convolution helpers.
arm_cmsis_nn_status	arm_depthwise_nhwc_conv_f16 (const cmsis_nn_context *ctx, const cmsis_nn_dw_conv_params_f16 *dw_conv_params, const cmsis_nn_dims *input_dims, const float16_t *input, const cmsis_nn_dims *filter_dims, const float16_t *kernel, const cmsis_nn_dims *bias_dims, const float16_t *bias, const cmsis_nn_dims *output_dims, float16_t *output)
	Depthwise convolution, NHWC layout.
arm_cmsis_nn_status	arm_depthwise_conv_f16 (const cmsis_nn_context *ctx, const cmsis_nn_dw_conv_params_f16 *dw_conv_params, const cmsis_nn_dims *input_dims, const float16_t *input, const cmsis_nn_dims *filter_dims, const float16_t *kernel, const cmsis_nn_dims *bias_dims, const float16_t *bias, const cmsis_nn_dims *output_dims, float16_t *output, arm_nn_tensor_layout layout)
	Depthwise convolution, dispatch by layout.
arm_cmsis_nn_status	arm_depthwise_conv_wrapper_f16 (const cmsis_nn_context *ctx, const cmsis_nn_dw_conv_params_f16 *dw_conv_params, const cmsis_nn_dims *input_dims, const float16_t *input, const cmsis_nn_dims *filter_dims, const float16_t *kernel, const cmsis_nn_dims *bias_dims, const float16_t *bias, const cmsis_nn_dims *output_dims, float16_t *output)
	Depthwise convolution wrapper using the CMSIS-NN baseline path.
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.
arm_cmsis_nn_status	arm_convolve_nhwc_f16 (const cmsis_nn_context *ctx, const cmsis_nn_conv_params_f16 *conv_params, const cmsis_nn_dims *input_dims, const float16_t *input_data, const cmsis_nn_dims *filter_dims, const float16_t *filter_data, const cmsis_nn_dims *bias_dims, const float16_t *bias_data, const cmsis_nn_dims *output_dims, float16_t *output_data)
	Convolution, NHWC layout.
arm_cmsis_nn_status	arm_convolve_f16 (const cmsis_nn_context *ctx, const cmsis_nn_conv_params_f16 *conv_params, const cmsis_nn_dims *input_dims, const float16_t *input_data, const cmsis_nn_dims *filter_dims, const float16_t *filter_data, const cmsis_nn_dims *bias_dims, const float16_t *bias_data, const cmsis_nn_dims *output_dims, float16_t *output_data, arm_nn_tensor_layout layout)
	Convolution, dispatch by layout.
arm_cmsis_nn_status	arm_convolve_wrapper_f16 (const cmsis_nn_context *ctx, const cmsis_nn_conv_params_f16 *conv_params, const cmsis_nn_dims *input_dims, const float16_t *input_data, const cmsis_nn_dims *filter_dims, const float16_t *filter_data, const cmsis_nn_dims *bias_dims, const float16_t *bias_data, const cmsis_nn_dims *output_dims, float16_t *output_data)
	Convolution wrapper using the CMSIS-NN baseline path.
arm_cmsis_nn_status	arm_convolve_1x1_nhwc_f16 (const cmsis_nn_context *ctx, const cmsis_nn_conv_params_f16 *conv_params, const cmsis_nn_dims *input_dims, const float16_t *input_data, const cmsis_nn_dims *filter_dims, const float16_t *filter_data, const cmsis_nn_dims *bias_dims, const float16_t *bias_data, const cmsis_nn_dims *output_dims, float16_t *output_data)
	1x1 convolution, NHWC layout.
arm_cmsis_nn_status	arm_convolve_1x1_f16 (const cmsis_nn_context *ctx, const cmsis_nn_conv_params_f16 *conv_params, const cmsis_nn_dims *input_dims, const float16_t *input_data, const cmsis_nn_dims *filter_dims, const float16_t *filter_data, const cmsis_nn_dims *bias_dims, const float16_t *bias_data, const cmsis_nn_dims *output_dims, float16_t *output_data, arm_nn_tensor_layout layout)
	1x1 convolution, dispatch by layout.
arm_cmsis_nn_status	arm_convolve_1_x_n_nhwc_f16 (const cmsis_nn_context *ctx, const cmsis_nn_conv_params_f16 *conv_params, const cmsis_nn_dims *input_dims, const float16_t *input_data, const cmsis_nn_dims *filter_dims, const float16_t *filter_data, const cmsis_nn_dims *bias_dims, const float16_t *bias_data, const cmsis_nn_dims *output_dims, float16_t *output_data)
	1xN convolution, NHWC layout.
arm_cmsis_nn_status	arm_convolve_1_x_n_f16 (const cmsis_nn_context *ctx, const cmsis_nn_conv_params_f16 *conv_params, const cmsis_nn_dims *input_dims, const float16_t *input_data, const cmsis_nn_dims *filter_dims, const float16_t *filter_data, const cmsis_nn_dims *bias_dims, const float16_t *bias_data, const cmsis_nn_dims *output_dims, float16_t *output_data, arm_nn_tensor_layout layout)
	1xN convolution, dispatch by layout.
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.
arm_cmsis_nn_status	arm_transpose_conv_wrapper_f16 (const cmsis_nn_context *ctx, const cmsis_nn_context *output_ctx, const cmsis_nn_transpose_conv_params_f16 *transpose_conv_params, const cmsis_nn_dims *input_dims, const float16_t *input_data, const cmsis_nn_dims *filter_dims, const float16_t *filter_data, const cmsis_nn_dims *bias_dims, const float16_t *bias_data, const cmsis_nn_dims *output_dims, float16_t *output_data, arm_nn_tensor_layout layout)
	Transpose convolution wrapper using the CMSIS-NN baseline path.
arm_cmsis_nn_status	arm_transpose_conv_nhwc_f16 (const cmsis_nn_context *ctx, const cmsis_nn_context *output_ctx, const cmsis_nn_transpose_conv_params_f16 *transpose_conv_params, const cmsis_nn_dims *input_dims, const float16_t *input_data, const cmsis_nn_dims *filter_dims, const float16_t *filter_data, const cmsis_nn_dims *bias_dims, const float16_t *bias_data, const cmsis_nn_dims *output_dims, float16_t *output_data)
	Transpose convolution, NHWC layout.
arm_cmsis_nn_status	arm_transpose_conv_f16 (const cmsis_nn_context *ctx, const cmsis_nn_context *output_ctx, const cmsis_nn_transpose_conv_params_f16 *transpose_conv_params, const cmsis_nn_dims *input_dims, const float16_t *input_data, const cmsis_nn_dims *filter_dims, const float16_t *filter_data, const cmsis_nn_dims *bias_dims, const float16_t *bias_data, const cmsis_nn_dims *output_dims, float16_t *output_data, arm_nn_tensor_layout layout)
	Transpose convolution, dispatch by layout.
int32_t	arm_transpose_conv_f16_get_buffer_size (const cmsis_nn_transpose_conv_params_f16 *transpose_conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims, const cmsis_nn_dims *out_dims)
	Get the temporary buffer size required by transpose convolution.
int32_t	arm_transpose_conv_f16_get_reverse_conv_buffer_size (const cmsis_nn_transpose_conv_params_f16 *transpose_conv_params, const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims)
	Get the reverse-convolution workspace size used by transpose convolution helpers.
arm_cmsis_nn_status	arm_convolve_1_x_n_s4 (const cmsis_nn_context *ctx, const cmsis_nn_conv_params *conv_params, const cmsis_nn_per_channel_quant_params *quant_params, const cmsis_nn_dims *input_dims, const int8_t *input_data, const cmsis_nn_dims *filter_dims, const int8_t *filter_data, const cmsis_nn_dims *bias_dims, const int32_t *bias_data, const cmsis_nn_dims *output_dims, int8_t *output_data)
	1xn convolution for s4 weights
arm_cmsis_nn_status	arm_convolve_1_x_n_s8 (const cmsis_nn_context *ctx, const cmsis_nn_conv_params *conv_params, const cmsis_nn_per_channel_quant_params *quant_params, const cmsis_nn_dims *input_dims, const int8_t *input_data, const cmsis_nn_dims *filter_dims, const int8_t *filter_data, const cmsis_nn_dims *bias_dims, const int32_t *bias_data, const cmsis_nn_dims *output_dims, int8_t *output_data)
	1xn convolution
arm_cmsis_nn_status	arm_convolve_1x1_s4 (const cmsis_nn_context *ctx, const cmsis_nn_conv_params *conv_params, const cmsis_nn_per_channel_quant_params *quant_params, const cmsis_nn_dims *input_dims, const int8_t *input_data, const cmsis_nn_dims *filter_dims, const int8_t *filter_data, const cmsis_nn_dims *bias_dims, const int32_t *bias_data, const cmsis_nn_dims *output_dims, int8_t *output_data)
	s4 version for 1x1 convolution with support for non-unity stride values
arm_cmsis_nn_status	arm_convolve_1x1_s4_fast (const cmsis_nn_context *ctx, const cmsis_nn_conv_params *conv_params, const cmsis_nn_per_channel_quant_params *quant_params, const cmsis_nn_dims *input_dims, const int8_t *input_data, const cmsis_nn_dims *filter_dims, const int8_t *filter_data, const cmsis_nn_dims *bias_dims, const int32_t *bias_data, const cmsis_nn_dims *output_dims, int8_t *output_data)
	Fast s4 version for 1x1 convolution (non-square shape)
arm_cmsis_nn_status	arm_convolve_1x1_s8 (const cmsis_nn_context *ctx, const cmsis_nn_conv_params *conv_params, const cmsis_nn_per_channel_quant_params *quant_params, const cmsis_nn_dims *input_dims, const int8_t *input_data, const cmsis_nn_dims *filter_dims, const int8_t *filter_data, const cmsis_nn_dims *bias_dims, const int32_t *bias_data, const cmsis_nn_dims *output_dims, int8_t *output_data)
	s8 version for 1x1 convolution with support for non-unity stride values
arm_cmsis_nn_status	arm_convolve_1x1_s8_fast (const cmsis_nn_context *ctx, const cmsis_nn_conv_params *conv_params, const cmsis_nn_per_channel_quant_params *quant_params, const cmsis_nn_dims *input_dims, const int8_t *input_data, const cmsis_nn_dims *filter_dims, const int8_t *filter_data, const cmsis_nn_dims *bias_dims, const int32_t *bias_data, const cmsis_nn_dims *output_dims, int8_t *output_data)
	Fast s8 version for 1x1 convolution (non-square shape)
arm_cmsis_nn_status	arm_convolve_even_s4 (const cmsis_nn_context *ctx, const cmsis_nn_conv_params *conv_params, const cmsis_nn_per_channel_quant_params *quant_params, const cmsis_nn_dims *input_dims, const int8_t *input_data, const cmsis_nn_dims *filter_dims, const int8_t *packed_filter_data, const cmsis_nn_dims *bias_dims, const int32_t *bias_data, const cmsis_nn_dims *output_dims, int8_t *output_data)
	Basic s4 convolution function with a requirement of even number of kernels.
arm_cmsis_nn_status	arm_convolve_s16 (const cmsis_nn_context *ctx, const cmsis_nn_conv_params *conv_params, const cmsis_nn_per_channel_quant_params *quant_params, const cmsis_nn_dims *input_dims, const int16_t *input_data, const cmsis_nn_dims *filter_dims, const int8_t *filter_data, const cmsis_nn_dims *bias_dims, const cmsis_nn_bias_data *bias_data, const cmsis_nn_dims *output_dims, int16_t *output_data)
	Basic s16 convolution function.
arm_cmsis_nn_status	arm_convolve_s4 (const cmsis_nn_context *ctx, const cmsis_nn_conv_params *conv_params, const cmsis_nn_per_channel_quant_params *quant_params, const cmsis_nn_dims *input_dims, const int8_t *input_data, const cmsis_nn_dims *filter_dims, const int8_t *packed_filter_data, const cmsis_nn_dims *bias_dims, const int32_t *bias_data, const cmsis_nn_dims *output_dims, int8_t *output_data)
	Basic s4 convolution function.
arm_cmsis_nn_status	arm_convolve_s8 (const cmsis_nn_context *ctx, const cmsis_nn_conv_params *conv_params, const cmsis_nn_per_channel_quant_params *quant_params, const cmsis_nn_dims *input_dims, const int8_t *input_data, const cmsis_nn_dims *filter_dims, const int8_t *filter_data, const cmsis_nn_dims *bias_dims, const int32_t *bias_data, const cmsis_nn_dims *upscale_dims, const cmsis_nn_dims *output_dims, int8_t *output_data)
	Basic s8 convolution function.
arm_cmsis_nn_status	arm_convolve_wrapper_s16 (const cmsis_nn_context *ctx, const cmsis_nn_conv_params *conv_params, const cmsis_nn_per_channel_quant_params *quant_params, const cmsis_nn_dims *input_dims, const int16_t *input_data, const cmsis_nn_dims *filter_dims, const int8_t *filter_data, const cmsis_nn_dims *bias_dims, const cmsis_nn_bias_data *bias_data, const cmsis_nn_dims *output_dims, int16_t *output_data)
	s16 convolution layer wrapper function with the main purpose to call the optimal kernel available in cmsis-nn to perform the convolution.
arm_cmsis_nn_status	arm_convolve_wrapper_s4 (const cmsis_nn_context *ctx, const cmsis_nn_conv_params *conv_params, const cmsis_nn_per_channel_quant_params *quant_params, const cmsis_nn_dims *input_dims, const int8_t *input_data, const cmsis_nn_dims *filter_dims, const int8_t *filter_data, const cmsis_nn_dims *bias_dims, const int32_t *bias_data, const cmsis_nn_dims *output_dims, int8_t *output_data)
	s4 convolution layer wrapper function with the main purpose to call the optimal kernel available in cmsis-nn to perform the convolution.
arm_cmsis_nn_status	arm_convolve_wrapper_s8 (const cmsis_nn_context *ctx, const cmsis_nn_conv_params *conv_params, const cmsis_nn_per_channel_quant_params *quant_params, const cmsis_nn_dims *input_dims, const int8_t *input_data, const cmsis_nn_dims *filter_dims, const int8_t *filter_data, const cmsis_nn_dims *bias_dims, const int32_t *bias_data, const cmsis_nn_dims *output_dims, int8_t *output_data)
	s8 convolution layer wrapper function with the main purpose to call the optimal kernel available in cmsis-nn to perform the convolution.
arm_cmsis_nn_status	arm_depthwise_conv_3x3_s8 (const cmsis_nn_context *ctx, const cmsis_nn_dw_conv_params *dw_conv_params, const cmsis_nn_per_channel_quant_params *quant_params, const cmsis_nn_dims *input_dims, const int8_t *input, const cmsis_nn_dims *filter_dims, const int8_t *kernel, const cmsis_nn_dims *bias_dims, const int32_t *bias, const cmsis_nn_dims *output_dims, int8_t *output)
	Optimized s8 depthwise convolution function for 3x3 kernel size with some constraints on the input arguments(documented below). Refer arm_depthwise_conv_s8() for function argument details.
arm_cmsis_nn_status	arm_depthwise_conv_fast_s16 (const cmsis_nn_context *ctx, const cmsis_nn_dw_conv_params *dw_conv_params, const cmsis_nn_per_channel_quant_params *quant_params, const cmsis_nn_dims *input_dims, const int16_t *input, const cmsis_nn_dims *filter_dims, const int8_t *kernel, const cmsis_nn_dims *bias_dims, const int64_t *bias, const cmsis_nn_dims *output_dims, int16_t *output)
	Optimized s16 depthwise convolution function with constraint that in_channel equals out_channel. Refer arm_depthwise_conv_s16() for function argument details.
arm_cmsis_nn_status	arm_depthwise_conv_s16 (const cmsis_nn_context *ctx, const cmsis_nn_dw_conv_params *dw_conv_params, const cmsis_nn_per_channel_quant_params *quant_params, const cmsis_nn_dims *input_dims, const int16_t *input, const cmsis_nn_dims *filter_dims, const int8_t *kernel, const cmsis_nn_dims *bias_dims, const int64_t *bias, const cmsis_nn_dims *output_dims, int16_t *output)
	Basic s16 depthwise convolution function that doesn't have any constraints on the input dimensions.
arm_cmsis_nn_status	arm_depthwise_conv_s4 (const cmsis_nn_context *ctx, const cmsis_nn_dw_conv_params *dw_conv_params, const cmsis_nn_per_channel_quant_params *quant_params, const cmsis_nn_dims *input_dims, const int8_t *input, const cmsis_nn_dims *filter_dims, const int8_t *kernel, const cmsis_nn_dims *bias_dims, const int32_t *bias, const cmsis_nn_dims *output_dims, int8_t *output)
	Basic s4 depthwise convolution function that doesn't have any constraints on the input dimensions.
arm_cmsis_nn_status	arm_depthwise_conv_s4_opt (const cmsis_nn_context *ctx, const cmsis_nn_dw_conv_params *dw_conv_params, const cmsis_nn_per_channel_quant_params *quant_params, const cmsis_nn_dims *input_dims, const int8_t *input, const cmsis_nn_dims *filter_dims, const int8_t *kernel, const cmsis_nn_dims *bias_dims, const int32_t *bias, const cmsis_nn_dims *output_dims, int8_t *output)
	Optimized s4 depthwise convolution function with constraint that in_channel equals out_channel. Refer arm_depthwise_conv_s4() for function argument details.
	__attribute__ ((optimize("no-unroll-loops")))
arm_cmsis_nn_status	arm_depthwise_conv_s8 (const cmsis_nn_context *ctx, const cmsis_nn_dw_conv_params *dw_conv_params, const cmsis_nn_per_channel_quant_params *quant_params, const cmsis_nn_dims *input_dims, const int8_t *input, const cmsis_nn_dims *filter_dims, const int8_t *kernel, const cmsis_nn_dims *bias_dims, const int32_t *bias, const cmsis_nn_dims *output_dims, int8_t *output)
	Basic s8 depthwise convolution function that doesn't have any constraints on the input dimensions.
arm_cmsis_nn_status	arm_depthwise_conv_s8_opt (const cmsis_nn_context *ctx, const cmsis_nn_dw_conv_params *dw_conv_params, const cmsis_nn_per_channel_quant_params *quant_params, const cmsis_nn_dims *input_dims, const int8_t *input, const cmsis_nn_dims *filter_dims, const int8_t *kernel, const cmsis_nn_dims *bias_dims, const int32_t *bias, const cmsis_nn_dims *output_dims, int8_t *output)
	Optimized s8 depthwise convolution function with constraint that in_channel equals out_channel. Refer arm_depthwise_conv_s8() for function argument details.
arm_cmsis_nn_status	arm_depthwise_conv_wrapper_s16 (const cmsis_nn_context *ctx, const cmsis_nn_dw_conv_params *dw_conv_params, const cmsis_nn_per_channel_quant_params *quant_params, const cmsis_nn_dims *input_dims, const int16_t *input, const cmsis_nn_dims *filter_dims, const int8_t *filter, const cmsis_nn_dims *bias_dims, const int64_t *bias, const cmsis_nn_dims *output_dims, int16_t *output)
	Wrapper function to pick the right optimized s16 depthwise convolution function.
arm_cmsis_nn_status	arm_depthwise_conv_wrapper_s4 (const cmsis_nn_context *ctx, const cmsis_nn_dw_conv_params *dw_conv_params, const cmsis_nn_per_channel_quant_params *quant_params, const cmsis_nn_dims *input_dims, const int8_t *input, const cmsis_nn_dims *filter_dims, const int8_t *filter, const cmsis_nn_dims *bias_dims, const int32_t *bias, const cmsis_nn_dims *output_dims, int8_t *output)
	Wrapper function to pick the right optimized s4 depthwise convolution function.
arm_cmsis_nn_status	arm_depthwise_conv_wrapper_s8 (const cmsis_nn_context *ctx, const cmsis_nn_dw_conv_params *dw_conv_params, const cmsis_nn_per_channel_quant_params *quant_params, const cmsis_nn_dims *input_dims, const int8_t *input, const cmsis_nn_dims *filter_dims, const int8_t *filter, const cmsis_nn_dims *bias_dims, const int32_t *bias, const cmsis_nn_dims *output_dims, int8_t *output)
	Wrapper function to pick the right optimized s8 depthwise convolution function.
arm_cmsis_nn_status	arm_transpose_conv_s8 (const cmsis_nn_context *ctx, const cmsis_nn_context *output_ctx, const cmsis_nn_transpose_conv_params *transpose_conv_params, const cmsis_nn_per_channel_quant_params *quant_params, const cmsis_nn_dims *input_dims, const int8_t *input_data, const cmsis_nn_dims *filter_dims, const int8_t *filter_data, const cmsis_nn_dims *bias_dims, const int32_t *bias_data, const cmsis_nn_dims *output_dims, int8_t *output_data)
	Basic s8 transpose convolution function.
arm_cmsis_nn_status	arm_transpose_conv_wrapper_s8 (const cmsis_nn_context *ctx, const cmsis_nn_context *reverse_conv_ctx, const cmsis_nn_transpose_conv_params *transpose_conv_params, const cmsis_nn_per_channel_quant_params *quant_params, const cmsis_nn_dims *input_dims, const int8_t *input_data, const cmsis_nn_dims *filter_dims, const int8_t *filter_data, const cmsis_nn_dims *bias_dims, const int32_t *bias_data, const cmsis_nn_dims *output_dims, int8_t *output_data)
	Wrapper to select optimal transposed convolution algorithm depending on parameters.

Description

Collection of convolution, depthwise convolution functions and their variants.

The convolution is implemented in 2 steps: im2col and General Matrix Multiplication(GEMM)

im2col is a process of converting each patch of image data into a column. After im2col, the convolution is computed as matrix-matrix multiplication.

To reduce the memory footprint, the im2col is performed partially. Each iteration, only a few column (i.e., patches) are generated followed by GEMM.

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

__attribute__()

__attribute__

(

(optimize("no-unroll-loops"))

)

arm_convolve_1_x_n_f16()

arm_cmsis_nn_status arm_convolve_1_x_n_f16	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_conv_params_f16 *	conv_params,
		const cmsis_nn_dims *	input_dims,
		const float16_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const float16_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const float16_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		float16_t *	output_data,
		arm_nn_tensor_layout	layout
	)

1xN convolution, dispatch by layout.

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()

arm_cmsis_nn_status arm_convolve_1_x_n_f32	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_conv_params_f32 *	conv_params,
		const cmsis_nn_dims *	input_dims,
		const float32_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const float32_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const float32_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		float32_t *	output_data,
		arm_nn_tensor_layout	layout
	)

1xN convolution, dispatch by layout.

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_nhwc_f16()

arm_cmsis_nn_status arm_convolve_1_x_n_nhwc_f16	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_conv_params_f16 *	conv_params,
		const cmsis_nn_dims *	input_dims,
		const float16_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const float16_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const float16_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		float16_t *	output_data
	)

1xN convolution, NHWC layout.

arm_convolve_1_x_n_nhwc_f32()

arm_cmsis_nn_status arm_convolve_1_x_n_nhwc_f32	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_conv_params_f32 *	conv_params,
		const cmsis_nn_dims *	input_dims,
		const float32_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const float32_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const float32_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		float32_t *	output_data
	)

1xN convolution, NHWC layout.

arm_convolve_1_x_n_s4()

arm_cmsis_nn_status arm_convolve_1_x_n_s4	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_conv_params *	conv_params,
		const cmsis_nn_per_channel_quant_params *	quant_params,
		const cmsis_nn_dims *	input_dims,
		const int8_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const int8_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const int32_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		int8_t *	output_data
	)

1xn convolution for s4 weights

Parameters

[in,out]	ctx	Function context that contains the additional buffer if required by the function. arm_convolve_1_x_n_s4_get_buffer_size will return the buffer_size if required The caller is expected to clear the buffer, if applicable, for security reasons.
[in]	conv_params	Convolution parameters (e.g. strides, dilations, pads,...). Range of conv_params->input_offset : [-127, 128] Range of conv_params->output_offset : [-128, 127]
[in]	quant_params	Per-channel quantization info. It contains the multiplier and shift values to be applied to each output channel
[in]	input_dims	Input (activation) tensor dimensions. Format: [N, H, W, C_IN]
[in]	input_data	Input (activation) data pointer. Data type: int8
[in]	filter_dims	Filter tensor dimensions. Format: [C_OUT, 1, WK, C_IN] where WK is the horizontal spatial filter dimension
[in]	filter_data	Filter data pointer. Data type: int8 as packed int4
[in]	bias_dims	Bias tensor dimensions. Format: [C_OUT]
[in]	bias_data	Optional bias data pointer. Data type: int32
[in]	output_dims	Output tensor dimensions. Format: [N, H, W, C_OUT]
[out]	output_data	Output data pointer. Data type: int8

Returns

The function returns either ARM_CMSIS_NN_ARG_ERROR if argument constraints fail. or, ARM_CMSIS_NN_SUCCESS on successful completion.

• Supported framework : TensorFlow Lite Micro
• The following constrains on the arguments apply
  1. stride.w * input_dims->c is a multiple of 4
  2. Explicit constraints(since it is for 1xN convolution) -## input_dims->h equals 1 -## output_dims->h equals 1 -## filter_dims->h equals 1

     Todo:

     Remove constraint on output_dims->w to make the function generic.

arm_convolve_1_x_n_s8()

arm_cmsis_nn_status arm_convolve_1_x_n_s8	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_conv_params *	conv_params,
		const cmsis_nn_per_channel_quant_params *	quant_params,
		const cmsis_nn_dims *	input_dims,
		const int8_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const int8_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const int32_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		int8_t *	output_data
	)

1xn convolution

Parameters

[in,out]	ctx	Function context that contains the additional buffer if required by the function. arm_convolve_1_x_n_s8_get_buffer_size will return the buffer_size if required The caller is expected to clear the buffer, if applicable, for security reasons.
[in]	conv_params	Convolution parameters (e.g. strides, dilations, pads,...). Range of conv_params->input_offset : [-127, 128] Range of conv_params->output_offset : [-128, 127]
[in]	quant_params	Per-channel quantization info. It contains the multiplier and shift values to be applied to each output channel
[in]	input_dims	Input (activation) tensor dimensions. Format: [N, H, W, C_IN]
[in]	input_data	Input (activation) data pointer. Data type: int8
[in]	filter_dims	Filter tensor dimensions. Format: [C_OUT, 1, WK, C_IN] where WK is the horizontal spatial filter dimension
[in]	filter_data	Filter data pointer. Data type: int8
[in]	bias_dims	Bias tensor dimensions. Format: [C_OUT]
[in]	bias_data	Optional bias data pointer. Data type: int32
[in]	output_dims	Output tensor dimensions. Format: [N, H, W, C_OUT]
[out]	output_data	Output data pointer. Data type: int8

Returns

The function returns either ARM_CMSIS_NN_ARG_ERROR if argument constraints fail. or, ARM_CMSIS_NN_SUCCESS on successful completion.

• Supported framework : TensorFlow Lite Micro
• The following constrains on the arguments apply
  1. input_dims->n equals 1
  2. ouput_dims->w is a multiple of 4
  3. Explicit constraints(since it is for 1xN convolution) -## input_dims->h equals 1 -## output_dims->h equals 1 -## filter_dims->h equals 1

     Todo:

     Remove constraint on output_dims->w to make the function generic.

arm_convolve_1x1_f16()

arm_cmsis_nn_status arm_convolve_1x1_f16	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_conv_params_f16 *	conv_params,
		const cmsis_nn_dims *	input_dims,
		const float16_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const float16_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const float16_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		float16_t *	output_data,
		arm_nn_tensor_layout	layout
	)

1x1 convolution, dispatch by layout.

Note

When conv_params->weight_format is set to ARM_NN_WEIGHT_FORMAT_NT_N_PACKED, the matmul-backed 1x1 convolution paths interpret filter_data as an already prepacked NTxN RHS buffer instead of the standard public filter layout.

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()

arm_cmsis_nn_status arm_convolve_1x1_f32	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_conv_params_f32 *	conv_params,
		const cmsis_nn_dims *	input_dims,
		const float32_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const float32_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const float32_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		float32_t *	output_data,
		arm_nn_tensor_layout	layout
	)

1x1 convolution, dispatch by layout.

Note

When conv_params->weight_format is set to ARM_NN_WEIGHT_FORMAT_NT_N_PACKED, the matmul-backed 1x1 convolution paths interpret filter_data as an already prepacked NTxN RHS buffer instead of the standard public filter layout.

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_nhwc_f16()

arm_cmsis_nn_status arm_convolve_1x1_nhwc_f16	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_conv_params_f16 *	conv_params,
		const cmsis_nn_dims *	input_dims,
		const float16_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const float16_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const float16_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		float16_t *	output_data
	)

1x1 convolution, NHWC layout.

arm_convolve_1x1_nhwc_f32()

arm_cmsis_nn_status arm_convolve_1x1_nhwc_f32	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_conv_params_f32 *	conv_params,
		const cmsis_nn_dims *	input_dims,
		const float32_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const float32_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const float32_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		float32_t *	output_data
	)

1x1 convolution, NHWC layout.

arm_convolve_1x1_s4()

arm_cmsis_nn_status arm_convolve_1x1_s4	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_conv_params *	conv_params,
		const cmsis_nn_per_channel_quant_params *	quant_params,
		const cmsis_nn_dims *	input_dims,
		const int8_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const int8_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const int32_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		int8_t *	output_data
	)

s4 version for 1x1 convolution with support for non-unity stride values

Parameters

[in,out]	ctx	Function context that contains the additional buffer if required by the function. None is required by this function.
[in]	conv_params	Convolution parameters (e.g. strides, dilations, pads,...). Range of conv_params->input_offset : [-127, 128] Range of conv_params->output_offset : [-128, 127]
[in]	quant_params	Per-channel quantization info. It contains the multiplier and shift values to be applied to each output channel
[in]	input_dims	Input (activation) tensor dimensions. Format: [N, H, W, C_IN]
[in]	input_data	Input (activation) data pointer. Data type: int8
[in]	filter_dims	Filter tensor dimensions. Format: [C_OUT, 1, 1, C_IN]
[in]	filter_data	Filter data pointer. Data type: int8 packed with 2x int4
[in]	bias_dims	Bias tensor dimensions. Format: [C_OUT]
[in]	bias_data	Optional bias data pointer. Data type: int32
[in]	output_dims	Output tensor dimensions. Format: [N, H, W, C_OUT]
[out]	output_data	Output data pointer. Data type: int8

Returns

The function returns either ARM_CMSIS_NN_ARG_ERROR if argument constraints fail. or, ARM_CMSIS_NN_SUCCESS on successful completion.

• Supported framework : TensorFlow Lite Micro
• The following constrains on the arguments apply
  1. conv_params->padding.w = conv_params->padding.h = 0

arm_convolve_1x1_s4_fast()

arm_cmsis_nn_status arm_convolve_1x1_s4_fast	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_conv_params *	conv_params,
		const cmsis_nn_per_channel_quant_params *	quant_params,
		const cmsis_nn_dims *	input_dims,
		const int8_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const int8_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const int32_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		int8_t *	output_data
	)

Fast s4 version for 1x1 convolution (non-square shape)

Parameters

[in,out]	ctx	Function context that contains the additional buffer if required by the function. arm_convolve_1x1_s4_fast_get_buffer_size will return the buffer_size if required. The caller is expected to clear the buffer ,if applicable, for security reasons.
[in]	conv_params	Convolution parameters (e.g. strides, dilations, pads,...). Range of conv_params->input_offset : [-127, 128] Range of conv_params->output_offset : [-128, 127]
[in]	quant_params	Per-channel quantization info. It contains the multiplier and shift values to be applied to each output channel
[in]	input_dims	Input (activation) tensor dimensions. Format: [N, H, W, C_IN]
[in]	input_data	Input (activation) data pointer. Data type: int8
[in]	filter_dims	Filter tensor dimensions. Format: [C_OUT, 1, 1, C_IN]
[in]	filter_data	Filter data pointer. Data type: int8 packed with 2x int4
[in]	bias_dims	Bias tensor dimensions. Format: [C_OUT]
[in]	bias_data	Optional bias data pointer. Data type: int32
[in]	output_dims	Output tensor dimensions. Format: [N, H, W, C_OUT]
[out]	output_data	Output data pointer. Data type: int8

Returns

The function returns either ARM_CMSIS_NN_ARG_ERROR if argument constraints fail. or, ARM_CMSIS_NN_SUCCESS on successful completion.

• Supported framework : TensorFlow Lite Micro
• The following constrains on the arguments apply
  1. conv_params->padding.w = conv_params->padding.h = 0
  2. conv_params->stride.w = conv_params->stride.h = 1

arm_convolve_1x1_s8()

arm_cmsis_nn_status arm_convolve_1x1_s8	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_conv_params *	conv_params,
		const cmsis_nn_per_channel_quant_params *	quant_params,
		const cmsis_nn_dims *	input_dims,
		const int8_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const int8_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const int32_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		int8_t *	output_data
	)

s8 version for 1x1 convolution with support for non-unity stride values

Parameters

[in,out]	ctx	Function context that contains the additional buffer if required by the function. None is required by this function.
[in]	conv_params	Convolution parameters (e.g. strides, dilations, pads,...). Range of conv_params->input_offset : [-127, 128] Range of conv_params->output_offset : [-128, 127]
[in]	quant_params	Per-channel quantization info. It contains the multiplier and shift values to be applied to each output channel
[in]	input_dims	Input (activation) tensor dimensions. Format: [N, H, W, C_IN]
[in]	input_data	Input (activation) data pointer. Data type: int8
[in]	filter_dims	Filter tensor dimensions. Format: [C_OUT, 1, 1, C_IN]
[in]	filter_data	Filter data pointer. Data type: int8
[in]	bias_dims	Bias tensor dimensions. Format: [C_OUT]
[in]	bias_data	Optional bias data pointer. Data type: int32
[in]	output_dims	Output tensor dimensions. Format: [N, H, W, C_OUT]
[out]	output_data	Output data pointer. Data type: int8

Returns

The function returns either ARM_CMSIS_NN_ARG_ERROR if argument constraints fail. or, ARM_CMSIS_NN_SUCCESS on successful completion.

• Supported framework : TensorFlow Lite Micro
• The following constrains on the arguments apply
  1. conv_params->padding.w = conv_params->padding.h = 0

arm_convolve_1x1_s8_fast()

arm_cmsis_nn_status arm_convolve_1x1_s8_fast	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_conv_params *	conv_params,
		const cmsis_nn_per_channel_quant_params *	quant_params,
		const cmsis_nn_dims *	input_dims,
		const int8_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const int8_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const int32_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		int8_t *	output_data
	)

Fast s8 version for 1x1 convolution (non-square shape)

Parameters

[in,out]	ctx	Function context that contains the additional buffer if required by the function. arm_convolve_1x1_s8_fast_get_buffer_size will return the buffer_size if required. The caller is expected to clear the buffer, if applicable, for security reasons.
[in]	conv_params	Convolution parameters (e.g. strides, dilations, pads,...). Range of conv_params->input_offset : [-127, 128] Range of conv_params->output_offset : [-128, 127]
[in]	quant_params	Per-channel quantization info. It contains the multiplier and shift values to be applied to each output channel
[in]	input_dims	Input (activation) tensor dimensions. Format: [N, H, W, C_IN]
[in]	input_data	Input (activation) data pointer. Data type: int8
[in]	filter_dims	Filter tensor dimensions. Format: [C_OUT, 1, 1, C_IN]
[in]	filter_data	Filter data pointer. Data type: int8
[in]	bias_dims	Bias tensor dimensions. Format: [C_OUT]
[in]	bias_data	Optional bias data pointer. Data type: int32
[in]	output_dims	Output tensor dimensions. Format: [N, H, W, C_OUT]
[out]	output_data	Output data pointer. Data type: int8

Returns

The function returns either ARM_CMSIS_NN_ARG_ERROR if argument constraints fail. or, ARM_CMSIS_NN_SUCCESS on successful completion.

• Supported framework : TensorFlow Lite Micro
• The following constrains on the arguments apply
  1. conv_params->padding.w = conv_params->padding.h = 0
  2. conv_params->stride.w = conv_params->stride.h = 1

arm_convolve_even_s4()

arm_cmsis_nn_status arm_convolve_even_s4	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_conv_params *	conv_params,
		const cmsis_nn_per_channel_quant_params *	quant_params,
		const cmsis_nn_dims *	input_dims,
		const int8_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const int8_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const int32_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		int8_t *	output_data
	)

Basic s4 convolution function with a requirement of even number of kernels.

Parameters

[in,out]	ctx	Function context that contains the additional buffer if required by the function. arm_convolve_s4_get_buffer_size will return the buffer_size if required. The caller is expected to clear the buffer ,if applicable, for security reasons.
[in]	conv_params	Convolution parameters (e.g. strides, dilations, pads,...). Range of conv_params->input_offset : [-127, 128] Range of conv_params->output_offset : [-128, 127]
[in]	quant_params	Per-channel quantization info. It contains the multiplier and shift values to be applied to each output channel
[in]	input_dims	Input (activation) tensor dimensions. Format: [N, H, W, C_IN]
[in]	input_data	Input (activation) data pointer. Data type: int8
[in]	filter_dims	Filter tensor dimensions. Format: [C_OUT, HK, WK, C_IN] where HK and WK are the spatial filter dimensions. Note the product must be even.
[in]	filter_data	Packed Filter data pointer. Data type: int8 packed with 2x int4
[in]	bias_dims	Bias tensor dimensions. Format: [C_OUT]
[in]	bias_data	Optional bias data pointer. Data type: int32
[in]	output_dims	Output tensor dimensions. Format: [N, H, W, C_OUT]
[out]	output_data	Output data pointer. Data type: int8

Returns

The function returns ARM_CMSIS_NN_SUCCESS if successful or ARM_CMSIS_NN_ARG_ERROR if incorrect arguments or ARM_CMSIS_NN_NO_IMPL_ERROR if not for MVE

1. Supported framework: TensorFlow Lite micro
2. Additional memory is required for optimization. Refer to argument 'ctx' for details.

arm_convolve_f16()

arm_cmsis_nn_status arm_convolve_f16	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_conv_params_f16 *	conv_params,
		const cmsis_nn_dims *	input_dims,
		const float16_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const float16_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const float16_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		float16_t *	output_data,
		arm_nn_tensor_layout	layout
	)

Convolution, dispatch by layout.

Parameters

[in,out]	ctx	Function context that may hold a temporary scratch buffer.
[in]	conv_params	Convolution parameters (stride, padding, dilation and activation clamp).
[in]	input_dims	Input tensor dimensions. Format depends on layout.
[in]	input_data	Pointer to the input tensor data.
[in]	filter_dims	Filter tensor dimensions. Format depends on layout.
[in]	filter_data	Pointer to the filter tensor data.
[in]	bias_dims	Bias tensor dimensions. Format: [C_OUT].
[in]	bias_data	Optional bias tensor data.
[in]	output_dims	Output tensor dimensions. Format depends on layout.
[out]	output_data	Pointer to the output tensor data.
[in]	layout	Tensor layout selector. Current float APIs require ARM_NN_LAYOUT_NHWC.

Note

When conv_params->weight_format is set to ARM_NN_WEIGHT_FORMAT_NT_N_PACKED, the matmul-backed convolution paths interpret filter_data as an already prepacked NTxN RHS buffer instead of the standard public filter layout.

Returns

ARM_CMSIS_NN_SUCCESS on success or ARM_CMSIS_NN_ARG_ERROR on invalid arguments.

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_params	Convolution parameters.
[in]	input_dims	Input tensor dimensions.
[in]	filter_dims	Filter tensor dimensions.
[in]	output_dims	Output tensor dimensions.
[in]	layout	Tensor 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()

arm_cmsis_nn_status arm_convolve_f32	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_conv_params_f32 *	conv_params,
		const cmsis_nn_dims *	input_dims,
		const float32_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const float32_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const float32_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		float32_t *	output_data,
		arm_nn_tensor_layout	layout
	)

Convolution, dispatch by layout.

Parameters

[in,out]	ctx	Function context that may hold a temporary scratch buffer.
[in]	conv_params	Convolution parameters (stride, padding, dilation and activation clamp).
[in]	input_dims	Input tensor dimensions. Format depends on layout.
[in]	input_data	Pointer to the input tensor data.
[in]	filter_dims	Filter tensor dimensions. Format depends on layout.
[in]	filter_data	Pointer to the filter tensor data.
[in]	bias_dims	Bias tensor dimensions. Format: [C_OUT].
[in]	bias_data	Optional bias tensor data.
[in]	output_dims	Output tensor dimensions. Format depends on layout.
[out]	output_data	Pointer to the output tensor data.
[in]	layout	Tensor layout selector. Current float APIs require ARM_NN_LAYOUT_NHWC.

Note

When conv_params->weight_format is set to ARM_NN_WEIGHT_FORMAT_NT_N_PACKED, the matmul-backed convolution paths interpret filter_data as an already prepacked NTxN RHS buffer instead of the standard public filter layout.

Returns

ARM_CMSIS_NN_SUCCESS on success or ARM_CMSIS_NN_ARG_ERROR on invalid arguments.

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_params	Convolution parameters.
[in]	input_dims	Input tensor dimensions.
[in]	filter_dims	Filter tensor dimensions.
[in]	output_dims	Output tensor dimensions.
[in]	layout	Tensor 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_nhwc_f16()

arm_cmsis_nn_status arm_convolve_nhwc_f16	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_conv_params_f16 *	conv_params,
		const cmsis_nn_dims *	input_dims,
		const float16_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const float16_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const float16_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		float16_t *	output_data
	)

Convolution, NHWC layout.

arm_convolve_nhwc_f32()

arm_cmsis_nn_status arm_convolve_nhwc_f32	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_conv_params_f32 *	conv_params,
		const cmsis_nn_dims *	input_dims,
		const float32_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const float32_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const float32_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		float32_t *	output_data
	)

Convolution, NHWC layout.

arm_convolve_s16()

arm_cmsis_nn_status arm_convolve_s16	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_conv_params *	conv_params,
		const cmsis_nn_per_channel_quant_params *	quant_params,
		const cmsis_nn_dims *	input_dims,
		const int16_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const int8_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const cmsis_nn_bias_data *	bias_data,
		const cmsis_nn_dims *	output_dims,
		int16_t *	output_data
	)

Basic s16 convolution function.

Parameters

[in,out]	ctx	Function context that contains the additional buffer if required by the function. arm_convolve_s16_get_buffer_size will return the buffer_size if required. The caller is expected to clear the buffer, if applicable, for security reasons.
[in]	conv_params	Convolution parameters (e.g. strides, dilations, pads,...). conv_params->input_offset : Not used conv_params->output_offset : Not used
[in]	quant_params	Per-channel quantization info. It contains the multiplier and shift values to be applied to each output channel
[in]	input_dims	Input (activation) tensor dimensions. Format: [N, H, W, C_IN]
[in]	input_data	Input (activation) data pointer. Data type: int16
[in]	filter_dims	Filter tensor dimensions. Format: [C_OUT, HK, WK, C_IN] where HK and WK are the spatial filter dimensions
[in]	filter_data	Filter data pointer. Data type: int8
[in]	bias_dims	Bias tensor dimensions. Format: [C_OUT]
[in]	bias_data	Struct with optional bias data pointer. Bias data type can be int64 or int32 depending flag in struct.
[in]	output_dims	Output tensor dimensions. Format: [N, H, W, C_OUT]
[out]	output_data	Output data pointer. Data type: int16

Returns

The function returns ARM_CMSIS_NN_SUCCESS if successful or ARM_CMSIS_NN_ARG_ERROR if incorrect arguments or ARM_CMSIS_NN_NO_IMPL_ERROR

1. Supported framework: TensorFlow Lite micro
2. Additional memory is required for optimization. Refer to argument 'ctx' for details.

arm_convolve_s4()

arm_cmsis_nn_status arm_convolve_s4	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_conv_params *	conv_params,
		const cmsis_nn_per_channel_quant_params *	quant_params,
		const cmsis_nn_dims *	input_dims,
		const int8_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const int8_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const int32_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		int8_t *	output_data
	)

Basic s4 convolution function.

Parameters

[in,out]	ctx	Function context that contains the additional buffer if required by the function. arm_convolve_s4_get_buffer_size will return the buffer_size if required. The caller is expected to clear the buffer ,if applicable, for security reasons.
[in]	conv_params	Convolution parameters (e.g. strides, dilations, pads,...). Range of conv_params->input_offset : [-127, 128] Range of conv_params->output_offset : [-128, 127]
[in]	quant_params	Per-channel quantization info. It contains the multiplier and shift values to be applied to each output channel
[in]	input_dims	Input (activation) tensor dimensions. Format: [N, H, W, C_IN]
[in]	input_data	Input (activation) data pointer. Data type: int8
[in]	filter_dims	Filter tensor dimensions. Format: [C_OUT, HK, WK, C_IN] where HK and WK are the spatial filter dimensions
[in]	filter_data	Packed Filter data pointer. Data type: int8 packed with 2x int4
[in]	bias_dims	Bias tensor dimensions. Format: [C_OUT]
[in]	bias_data	Optional bias data pointer. Data type: int32
[in]	output_dims	Output tensor dimensions. Format: [N, H, W, C_OUT]
[out]	output_data	Output data pointer. Data type: int8

Returns

The function returns ARM_CMSIS_NN_SUCCESS

1. Supported framework: TensorFlow Lite micro
2. Additional memory is required for optimization. Refer to argument 'ctx' for details.

arm_convolve_s8()

arm_cmsis_nn_status arm_convolve_s8	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_conv_params *	conv_params,
		const cmsis_nn_per_channel_quant_params *	quant_params,
		const cmsis_nn_dims *	input_dims,
		const int8_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const int8_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const int32_t *	bias_data,
		const cmsis_nn_dims *	upscale_dims,
		const cmsis_nn_dims *	output_dims,
		int8_t *	output_data
	)

Basic s8 convolution function.

Parameters

[in,out]	ctx	Function context that contains the additional buffer if required by the function. arm_convolve_s8_get_buffer_size will return the buffer_size if required. The caller is expected to clear the buffer, if applicable, for security reasons.
[in]	conv_params	Convolution parameters (e.g. strides, dilations, pads,...). Range of conv_params->input_offset : [-127, 128] Range of conv_params->output_offset : [-128, 127]
[in]	quant_params	Per-channel quantization info. It contains the multiplier and shift values to be applied to each output channel
[in]	input_dims	Input (activation) tensor dimensions. Format: [N, H, W, C_IN]
[in]	input_data	Input (activation) data pointer. Data type: int8
[in]	filter_dims	Filter tensor dimensions. Format: [C_OUT, HK, WK, CK] where HK, WK and CK are the spatial filter dimensions. CK != C_IN is used for grouped convolution, in which case the required conditions are C_IN = N * CK and C_OUT = N * M for N groups of size M.
[in]	filter_data	Filter data pointer. Data type: int8
[in]	bias_dims	Bias tensor dimensions. Format: [C_OUT]
[in]	bias_data	Optional bias data pointer. Data type: int32
[in]	upscale_dims	Inserts zeroes to upscale the input in h/w dimensions if set to 2. This is used for tranposed convolution.
[in]	output_dims	Output tensor dimensions. Format: [N, H, W, C_OUT]
[out]	output_data	Output data pointer. Data type: int8

Returns

The function returns ARM_CMSIS_NN_SUCCESS if successful or ARM_CMSIS_NN_ARG_ERROR if incorrect arguments or ARM_CMSIS_NN_NO_IMPL_ERROR

1. Supported framework: TensorFlow Lite micro
2. Additional memory is required for optimization. Refer to argument 'ctx' for details.

arm_convolve_wrapper_f16()

arm_cmsis_nn_status arm_convolve_wrapper_f16	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_conv_params_f16 *	conv_params,
		const cmsis_nn_dims *	input_dims,
		const float16_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const float16_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const float16_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		float16_t *	output_data
	)

Convolution wrapper using the CMSIS-NN baseline path.

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()

arm_cmsis_nn_status arm_convolve_wrapper_f32	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_conv_params_f32 *	conv_params,
		const cmsis_nn_dims *	input_dims,
		const float32_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const float32_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const float32_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		float32_t *	output_data
	)

Convolution wrapper using the CMSIS-NN baseline path.

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()

arm_cmsis_nn_status arm_convolve_wrapper_s16	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_conv_params *	conv_params,
		const cmsis_nn_per_channel_quant_params *	quant_params,
		const cmsis_nn_dims *	input_dims,
		const int16_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const int8_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const cmsis_nn_bias_data *	bias_data,
		const cmsis_nn_dims *	output_dims,
		int16_t *	output_data
	)

s16 convolution layer wrapper function with the main purpose to call the optimal kernel available in cmsis-nn to perform the convolution.

Parameters

[in,out]	ctx	Function context that contains the additional buffer if required by the function. arm_convolve_wrapper_s8_get_buffer_size will return the buffer_size if required The caller is expected to clear the buffer, if applicable, for security reasons.
[in]	conv_params	Convolution parameters (e.g. strides, dilations, pads,...). conv_params->input_offset : Not used conv_params->output_offset : Not used
[in]	quant_params	Per-channel quantization info. It contains the multiplier and shift values to be applied to each output channel
[in]	input_dims	Input (activation) tensor dimensions. Format: [N, H, W, C_IN]
[in]	input_data	Input (activation) data pointer. Data type: int16
[in]	filter_dims	Filter tensor dimensions. Format: [C_OUT, HK, WK, C_IN] where HK and WK are the spatial filter dimensions
[in]	filter_data	Filter data pointer. Data type: int8
[in]	bias_dims	Bias tensor dimensions. Format: [C_OUT]
[in]	bias_data	Struct with optional bias data pointer. Bias data type can be int64 or int32 depending flag in struct.
[in]	output_dims	Output tensor dimensions. Format: [N, H, W, C_OUT]
[out]	output_data	Output data pointer. Data type: int16

Returns

The function returns either ARM_CMSIS_NN_ARG_ERROR if argument constraints fail. or, ARM_CMSIS_NN_SUCCESS on successful completion.

arm_convolve_wrapper_s4()

arm_cmsis_nn_status arm_convolve_wrapper_s4	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_conv_params *	conv_params,
		const cmsis_nn_per_channel_quant_params *	quant_params,
		const cmsis_nn_dims *	input_dims,
		const int8_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const int8_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const int32_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		int8_t *	output_data
	)

s4 convolution layer wrapper function with the main purpose to call the optimal kernel available in cmsis-nn to perform the convolution.

Parameters

[in,out]	ctx	Function context that contains the additional buffer if required by the function. arm_convolve_wrapper_s4_get_buffer_size will return the buffer_size if required. The caller is expected to clear the buffer ,if applicable, for security reasons.
[in]	conv_params	Convolution parameters (e.g. strides, dilations, pads,...). Range of conv_params->input_offset : [-127, 128] Range of conv_params->output_offset : [-128, 127]
[in]	quant_params	Per-channel quantization info. It contains the multiplier and shift values to be applied to each output channel
[in]	input_dims	Input (activation) tensor dimensions. Format: [N, H, W, C_IN]
[in]	input_data	Input (activation) data pointer. Data type: int8
[in]	filter_dims	Filter tensor dimensions. Format: [C_OUT, HK, WK, C_IN] where HK and WK are the spatial filter dimensions
[in]	filter_data	Filter data pointer. Data type: int8 packed with 2x int4
[in]	bias_dims	Bias tensor dimensions. Format: [C_OUT]
[in]	bias_data	Bias data pointer. Data type: int32
[in]	output_dims	Output tensor dimensions. Format: [N, H, W, C_OUT]
[out]	output_data	Output data pointer. Data type: int8

Returns

The function returns either ARM_CMSIS_NN_ARG_ERROR if argument constraints fail. or, ARM_CMSIS_NN_SUCCESS on successful completion.

arm_convolve_wrapper_s8()

arm_cmsis_nn_status arm_convolve_wrapper_s8	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_conv_params *	conv_params,
		const cmsis_nn_per_channel_quant_params *	quant_params,
		const cmsis_nn_dims *	input_dims,
		const int8_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const int8_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const int32_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		int8_t *	output_data
	)

s8 convolution layer wrapper function with the main purpose to call the optimal kernel available in cmsis-nn to perform the convolution.

Parameters

[in,out]	ctx	Function context that contains the additional buffer if required by the function. arm_convolve_wrapper_s8_get_buffer_size will return the buffer_size if required. The caller is expected to clear the buffer, if applicable, for security reasons.
[in]	conv_params	Convolution parameters (e.g. strides, dilations, pads,...). Range of conv_params->input_offset : [-127, 128] Range of conv_params->output_offset : [-128, 127]
[in]	quant_params	Per-channel quantization info. It contains the multiplier and shift values to be applied to each output channel
[in]	input_dims	Input (activation) tensor dimensions. Format: [N, H, W, C_IN]
[in]	input_data	Input (activation) data pointer. Data type: int8
[in]	filter_dims	Filter tensor dimensions. Format: [C_OUT, HK, WK, C_IN] where HK and WK are the spatial filter dimensions
[in]	filter_data	Filter data pointer. Data type: int8
[in]	bias_dims	Bias tensor dimensions. Format: [C_OUT]
[in]	bias_data	Bias data pointer. Data type: int32
[in]	output_dims	Output tensor dimensions. Format: [N, H, W, C_OUT]
[out]	output_data	Output data pointer. Data type: int8

Returns

The function returns either ARM_CMSIS_NN_ARG_ERROR if argument constraints fail. or, ARM_CMSIS_NN_SUCCESS on successful completion.

arm_depthwise_conv_3x3_s8()

arm_cmsis_nn_status arm_depthwise_conv_3x3_s8	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_dw_conv_params *	dw_conv_params,
		const cmsis_nn_per_channel_quant_params *	quant_params,
		const cmsis_nn_dims *	input_dims,
		const int8_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const int8_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const int32_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		int8_t *	output_data
	)

Optimized s8 depthwise convolution function for 3x3 kernel size with some constraints on the input arguments(documented below). Refer arm_depthwise_conv_s8() for function argument details.

Returns

The function returns one of the following ARM_CMSIS_NN_ARG_ERROR - Unsupported dimension of tensors

• Unsupported pad size along the x axis ARM_CMSIS_NN_SUCCESS - Successful operation

• Supported framework : TensorFlow Lite Micro
• The following constrains on the arguments apply
  1. Number of input channel equals number of output channels
  2. Filter height and width equals 3
  3. Padding along x is either 0 or 1.

arm_depthwise_conv_f16()

arm_cmsis_nn_status arm_depthwise_conv_f16	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_dw_conv_params_f16 *	dw_conv_params,
		const cmsis_nn_dims *	input_dims,
		const float16_t *	input,
		const cmsis_nn_dims *	filter_dims,
		const float16_t *	kernel,
		const cmsis_nn_dims *	bias_dims,
		const float16_t *	bias,
		const cmsis_nn_dims *	output_dims,
		float16_t *	output,
		arm_nn_tensor_layout	layout
	)

Depthwise convolution, dispatch by layout.

Parameters

[in,out]	ctx	Function context that may hold a temporary scratch buffer.
[in]	dw_conv_params	Depthwise convolution parameters (stride, padding, dilation, channel multiplier and activation clamp).
[in]	input_dims	Input tensor dimensions. Format depends on layout.
[in]	input	Pointer to the input tensor data.
[in]	filter_dims	Filter tensor dimensions. Format depends on layout.
[in]	kernel	Pointer to the filter tensor data.
[in]	bias_dims	Bias tensor dimensions. Format: [C_OUT].
[in]	bias	Optional bias tensor data.
[in]	output_dims	Output tensor dimensions. Format depends on layout.
[out]	output	Pointer to the output tensor data.
[in]	layout	Tensor layout selector. Current float APIs require ARM_NN_LAYOUT_NHWC.

Note

When ctx->buf is used for internal kernel repacking, it must be aligned to the element type stored in scratch: at least 4-byte aligned for float32_t and, via at least 2-byte aligned for float16_t.

at least 2-byte aligned for float16_t.

Returns

ARM_CMSIS_NN_SUCCESS on success or ARM_CMSIS_NN_ARG_ERROR on invalid arguments.

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_params	Depthwise convolution parameters.
[in]	input_dims	Input tensor dimensions.
[in]	filter_dims	Filter tensor dimensions.
[in]	output_dims	Output tensor dimensions.
[in]	layout	Tensor layout selector.

Returns

Required buffer size in bytes, or 0 when no scratch buffer is needed.

arm_depthwise_conv_f32()

arm_cmsis_nn_status arm_depthwise_conv_f32	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_dw_conv_params_f32 *	dw_conv_params,
		const cmsis_nn_dims *	input_dims,
		const float32_t *	input,
		const cmsis_nn_dims *	filter_dims,
		const float32_t *	kernel,
		const cmsis_nn_dims *	bias_dims,
		const float32_t *	bias,
		const cmsis_nn_dims *	output_dims,
		float32_t *	output,
		arm_nn_tensor_layout	layout
	)

Depthwise convolution, dispatch by layout.

Parameters

[in,out]	ctx	Function context that may hold a temporary scratch buffer.
[in]	dw_conv_params	Depthwise convolution parameters (stride, padding, dilation, channel multiplier and activation clamp).
[in]	input_dims	Input tensor dimensions. Format depends on layout.
[in]	input	Pointer to the input tensor data.
[in]	filter_dims	Filter tensor dimensions. Format depends on layout.
[in]	kernel	Pointer to the filter tensor data.
[in]	bias_dims	Bias tensor dimensions. Format: [C_OUT].
[in]	bias	Optional bias tensor data.
[in]	output_dims	Output tensor dimensions. Format depends on layout.
[out]	output	Pointer to the output tensor data.
[in]	layout	Tensor layout selector. Current float APIs require ARM_NN_LAYOUT_NHWC.

Note

When ctx->buf is used for internal kernel repacking, it must be aligned to the element type stored in scratch: at least 4-byte aligned for float32_t and, via at least 2-byte aligned for float16_t.

at least 2-byte aligned for float16_t.

Returns

ARM_CMSIS_NN_SUCCESS on success or ARM_CMSIS_NN_ARG_ERROR on invalid arguments.

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_params	Depthwise convolution parameters.
[in]	input_dims	Input tensor dimensions.
[in]	filter_dims	Filter tensor dimensions.
[in]	output_dims	Output tensor dimensions.
[in]	layout	Tensor layout selector.

Returns

Required buffer size in bytes, or 0 when no scratch buffer is needed.

arm_depthwise_conv_fast_s16()

arm_cmsis_nn_status arm_depthwise_conv_fast_s16	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_dw_conv_params *	dw_conv_params,
		const cmsis_nn_per_channel_quant_params *	quant_params,
		const cmsis_nn_dims *	input_dims,
		const int16_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const int8_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const int64_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		int16_t *	output_data
	)

Optimized s16 depthwise convolution function with constraint that in_channel equals out_channel. Refer arm_depthwise_conv_s16() for function argument details.

Returns

The function returns one of the following ARM_CMSIS_NN_ARG_ERROR - ctx-buff == NULL and arm_depthwise_conv_fast_s16_get_buffer_size() > 0 or input channel != output channel or ch_mult != 1

ARM_CMSIS_NN_SUCCESS - Successful operation

• Supported framework: TensorFlow Lite
• The following constrains on the arguments apply
  1. Number of input channel equals number of output channels or ch_mult equals 1
• Reccomended when number of channels is 4 or greater.

arm_depthwise_conv_s16()

arm_cmsis_nn_status arm_depthwise_conv_s16	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_dw_conv_params *	dw_conv_params,
		const cmsis_nn_per_channel_quant_params *	quant_params,
		const cmsis_nn_dims *	input_dims,
		const int16_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const int8_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const int64_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		int16_t *	output_data
	)

Basic s16 depthwise convolution function that doesn't have any constraints on the input dimensions.

Parameters

[in,out]	ctx	Function context (e.g. temporary buffer). Check the function definition file to see if an additional buffer is required. Optional function {API}_get_buffer_size() provides the buffer size if an additional buffer is required. exists if additional memory is. The caller is expected to clear the buffer, if applicable, for security reasons.
[in]	dw_conv_params	Depthwise convolution parameters (e.g. strides, dilations, pads,...) conv_params->input_offset : Not used conv_params->output_offset : Not used
[in]	quant_params	Per-channel quantization info. It contains the multiplier and shift values to be applied to each output channel
[in]	input_dims	Input (activation) tensor dimensions. Format: [N, H, W, C_IN] Batch argument N is not used.
[in]	input_data	Input (activation) data pointer. Data type: int8
[in]	filter_dims	Filter tensor dimensions. Format: [1, H, W, C_OUT]
[in]	filter_data	Filter data pointer. Data type: int8
[in]	bias_dims	Bias tensor dimensions. Format: [C_OUT]
[in]	bias_data	Bias data pointer. Data type: int64
[in]	output_dims	Output tensor dimensions. Format: [N, H, W, C_OUT]
[in,out]	output_data	Output data pointer. Data type: int16

Returns

The function returns ARM_CMSIS_NN_SUCCESS

• Supported framework: TensorFlow Lite

arm_depthwise_conv_s4()

arm_cmsis_nn_status arm_depthwise_conv_s4	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_dw_conv_params *	dw_conv_params,
		const cmsis_nn_per_channel_quant_params *	quant_params,
		const cmsis_nn_dims *	input_dims,
		const int8_t *	input,
		const cmsis_nn_dims *	filter_dims,
		const int8_t *	kernel,
		const cmsis_nn_dims *	bias_dims,
		const int32_t *	bias,
		const cmsis_nn_dims *	output_dims,
		int8_t *	output
	)

Basic s4 depthwise convolution function that doesn't have any constraints on the input dimensions.

Parameters

[in,out]	ctx	Function context (e.g. temporary buffer). Check the function definition file to see if an additional buffer is required. Optional function {API}_get_buffer_size() provides the buffer size if an additional buffer is required exists if additional memory is. The caller is expected to clear the buffer ,if applicable, for security reasons.
[in]	dw_conv_params	Depthwise convolution parameters (e.g. strides, dilations, pads,...) dw_conv_params->dilation is not used. Range of dw_conv_params->input_offset : [-127, 128] Range of dw_conv_params->input_offset : [-128, 127]
[in]	quant_params	Per-channel quantization info. It contains the multiplier and shift values to be applied to each output channel
[in]	input_dims	Input (activation) tensor dimensions. Format: [N, H, W, C_IN] Batch argument N is not used.
[in]	input	Input (activation) data pointer. Data type: int8
[in]	filter_dims	Filter tensor dimensions. Format: [1, H, W, C_OUT]
[in]	kernel	Filter data pointer. Data type: int8_t packed 4-bit weights, e.g four sequential weights [0x1, 0x2, 0x3, 0x4] packed as [0x21, 0x43].
[in]	bias_dims	Bias tensor dimensions. Format: [C_OUT]
[in]	bias	Bias data pointer. Data type: int32
[in]	output_dims	Output tensor dimensions. Format: [N, H, W, C_OUT]
[in,out]	output	Output data pointer. Data type: int8

Returns

The function returns ARM_CMSIS_NN_SUCCESS

• Supported framework: TensorFlow Lite

arm_depthwise_conv_s4_opt()

arm_cmsis_nn_status arm_depthwise_conv_s4_opt	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_dw_conv_params *	dw_conv_params,
		const cmsis_nn_per_channel_quant_params *	quant_params,
		const cmsis_nn_dims *	input_dims,
		const int8_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const int8_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const int32_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		int8_t *	output_data
	)

Optimized s4 depthwise convolution function with constraint that in_channel equals out_channel. Refer arm_depthwise_conv_s4() for function argument details.

Returns

The function returns one of the following ARM_CMSIS_NN_ARG_ERROR - input channel != output channel or ch_mult != 1 ARM_CMSIS_NN_SUCCESS - Successful operation

Note

If number of channels is not a multiple of 4, upto 3 elements outside the boundary will be read out for the following if MVE optimizations(Arm Helium Technology) are used.

• Output shift
• Output multiplier
• Output bias
• kernel

• Supported framework: TensorFlow Lite
• The following constrains on the arguments apply
  1. Number of input channel equals number of output channels or ch_mult equals 1
• Reccomended when number of channels is 4 or greater.

arm_depthwise_conv_s8()

arm_cmsis_nn_status arm_depthwise_conv_s8	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_dw_conv_params *	dw_conv_params,
		const cmsis_nn_per_channel_quant_params *	quant_params,
		const cmsis_nn_dims *	input_dims,
		const int8_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const int8_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const int32_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		int8_t *	output_data
	)

Basic s8 depthwise convolution function that doesn't have any constraints on the input dimensions.

Parameters

[in,out]	ctx	Function context (e.g. temporary buffer). Check the function definition file to see if an additional buffer is required. Optional function {API}_get_buffer_size() provides the buffer size if an additional buffer is required exists if additional memory is. The caller is expected to clear the buffer, if applicable, for security reasons.
[in]	dw_conv_params	Depthwise convolution parameters (e.g. strides, dilations, pads,...) dw_conv_params->dilation is not used. Range of dw_conv_params->input_offset : [-127, 128] Range of dw_conv_params->input_offset : [-128, 127]
[in]	quant_params	Per-channel quantization info. It contains the multiplier and shift values to be applied to each output channel
[in]	input_dims	Input (activation) tensor dimensions. Format: [N, H, W, C_IN] Batch argument N is not used.
[in]	input_data	Input (activation) data pointer. Data type: int8
[in]	filter_dims	Filter tensor dimensions. Format: [1, H, W, C_OUT]
[in]	filter_data	Filter data pointer. Data type: int8
[in]	bias_dims	Bias tensor dimensions. Format: [C_OUT]
[in]	bias_data	Bias data pointer. Data type: int32
[in]	output_dims	Output tensor dimensions. Format: [N, H, W, C_OUT]
[in,out]	output_data	Output data pointer. Data type: int8

Returns

The function returns ARM_CMSIS_NN_SUCCESS

• Supported framework: TensorFlow Lite

arm_depthwise_conv_s8_opt()

arm_cmsis_nn_status arm_depthwise_conv_s8_opt	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_dw_conv_params *	dw_conv_params,
		const cmsis_nn_per_channel_quant_params *	quant_params,
		const cmsis_nn_dims *	input_dims,
		const int8_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const int8_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const int32_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		int8_t *	output_data
	)

Optimized s8 depthwise convolution function with constraint that in_channel equals out_channel. Refer arm_depthwise_conv_s8() for function argument details.

Returns

The function returns one of the following ARM_CMSIS_NN_ARG_ERROR - input channel != output channel or ch_mult != 1 ARM_CMSIS_NN_SUCCESS - Successful operation

Note

If number of channels is not a multiple of 4, upto 3 elements outside the boundary will be read out for the following if MVE optimizations(Arm Helium Technology) are used.

• Output shift
• Output multiplier
• Output bias
• kernel

• Supported framework: TensorFlow Lite
• The following constrains on the arguments apply
  1. Number of input channel equals number of output channels or ch_mult equals 1
• Reccomended when number of channels is 4 or greater.

arm_depthwise_conv_wrapper_f16()

arm_cmsis_nn_status arm_depthwise_conv_wrapper_f16	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_dw_conv_params_f16 *	dw_conv_params,
		const cmsis_nn_dims *	input_dims,
		const float16_t *	input,
		const cmsis_nn_dims *	filter_dims,
		const float16_t *	kernel,
		const cmsis_nn_dims *	bias_dims,
		const float16_t *	bias,
		const cmsis_nn_dims *	output_dims,
		float16_t *	output
	)

Depthwise convolution wrapper using the CMSIS-NN baseline path.

Parameters

[in,out]	ctx	Function context that may hold a temporary scratch buffer.
[in]	dw_conv_params	Depthwise convolution parameters.
[in]	input_dims	Input tensor dimensions.
[in]	input	Pointer to the input tensor data.
[in]	filter_dims	Filter tensor dimensions.
[in]	kernel	Pointer to the filter tensor data.
[in]	bias_dims	Bias tensor dimensions. Format: [C_OUT].
[in]	bias	Optional bias tensor data.
[in]	output_dims	Output tensor dimensions.
[out]	output	Pointer to the output tensor data.

Note

When ctx->buf is used for internal kernel repacking, it must be aligned to the element type stored in scratch: at least 4-byte aligned for float32_t and, via at least 2-byte aligned for float16_t.

at least 2-byte aligned for float16_t.

Returns

ARM_CMSIS_NN_SUCCESS on success or ARM_CMSIS_NN_ARG_ERROR on invalid arguments.

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()

arm_cmsis_nn_status arm_depthwise_conv_wrapper_f32	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_dw_conv_params_f32 *	dw_conv_params,
		const cmsis_nn_dims *	input_dims,
		const float32_t *	input,
		const cmsis_nn_dims *	filter_dims,
		const float32_t *	kernel,
		const cmsis_nn_dims *	bias_dims,
		const float32_t *	bias,
		const cmsis_nn_dims *	output_dims,
		float32_t *	output
	)

Depthwise convolution wrapper using the CMSIS-NN baseline path.

Parameters

[in,out]	ctx	Function context that may hold a temporary scratch buffer.
[in]	dw_conv_params	Depthwise convolution parameters.
[in]	input_dims	Input tensor dimensions.
[in]	input	Pointer to the input tensor data.
[in]	filter_dims	Filter tensor dimensions.
[in]	kernel	Pointer to the filter tensor data.
[in]	bias_dims	Bias tensor dimensions. Format: [C_OUT].
[in]	bias	Optional bias tensor data.
[in]	output_dims	Output tensor dimensions.
[out]	output	Pointer to the output tensor data.

Note

When ctx->buf is used for internal kernel repacking, it must be aligned to the element type stored in scratch: at least 4-byte aligned for float32_t and, via at least 2-byte aligned for float16_t.

at least 2-byte aligned for float16_t.

Returns

ARM_CMSIS_NN_SUCCESS on success or ARM_CMSIS_NN_ARG_ERROR on invalid arguments.

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()

arm_cmsis_nn_status arm_depthwise_conv_wrapper_s16	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_dw_conv_params *	dw_conv_params,
		const cmsis_nn_per_channel_quant_params *	quant_params,
		const cmsis_nn_dims *	input_dims,
		const int16_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const int8_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const int64_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		int16_t *	output_data
	)

Wrapper function to pick the right optimized s16 depthwise convolution function.

Parameters

[in,out]	ctx	Function context (e.g. temporary buffer). Check the function definition file to see if an additional buffer is required. Optional function {API}_get_buffer_size() provides the buffer size if required. The caller is expected to clear the buffer, if applicable, for security reasons.
[in]	dw_conv_params	Depthwise convolution parameters (e.g. strides, dilations, pads,...) dw_conv_params->dilation is not used. Range of dw_conv_params->input_offset : Not used Range of dw_conv_params->output_offset : Not used
[in]	quant_params	Per-channel quantization info. It contains the multiplier and shift values to be applied to each output channel
[in]	input_dims	Input (activation) tensor dimensions. Format: [H, W, C_IN] Batch argument N is not used and assumed to be 1.
[in]	input_data	Input (activation) data pointer. Data type: int16
[in]	filter_dims	Filter tensor dimensions. Format: [1, H, W, C_OUT]
[in]	filter_data	Filter data pointer. Data type: int8
[in]	bias_dims	Bias tensor dimensions. Format: [C_OUT]
[in]	bias_data	Bias data pointer. Data type: int64
[in]	output_dims	Output tensor dimensions. Format: [1, H, W, C_OUT]
[in,out]	output_data	Output data pointer. Data type: int16

Returns

The function returns ARM_CMSIS_NN_SUCCESS - Successful completion.

• Supported framework: TensorFlow Lite
• Picks one of the the following functions
  1. arm_depthwise_conv_s16()
  2. arm_depthwise_conv_fast_s16() - Cortex-M CPUs with DSP extension only

arm_depthwise_conv_wrapper_s4()

arm_cmsis_nn_status arm_depthwise_conv_wrapper_s4	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_dw_conv_params *	dw_conv_params,
		const cmsis_nn_per_channel_quant_params *	quant_params,
		const cmsis_nn_dims *	input_dims,
		const int8_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const int8_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const int32_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		int8_t *	output_data
	)

Wrapper function to pick the right optimized s4 depthwise convolution function.

Parameters

[in,out]	ctx	Function context (e.g. temporary buffer). Check the function definition file to see if an additional buffer is required. Optional function {API}_get_buffer_size() provides the buffer size if required. The caller is expected to clear the buffer ,if applicable, for security reasons.
[in]	dw_conv_params	Depthwise convolution parameters (e.g. strides, dilations, pads,...) dw_conv_params->dilation is not used. Range of dw_conv_params->input_offset : [-127, 128] Range of dw_conv_params->output_offset : [-128, 127]
[in]	quant_params	Per-channel quantization info. It contains the multiplier and shift values to be applied to each output channel
[in]	input_dims	Input (activation) tensor dimensions. Format: [H, W, C_IN] Batch argument N is not used and assumed to be 1.
[in]	input_data	Input (activation) data pointer. Data type: int8
[in]	filter_dims	Filter tensor dimensions. Format: [1, H, W, C_OUT]
[in]	filter_data	Filter data pointer. Data type: int8_t packed 4-bit weights, e.g four sequential weights [0x1, 0x2, 0x3, 0x4] packed as [0x21, 0x43].
[in]	bias_dims	Bias tensor dimensions. Format: [C_OUT]
[in]	bias_data	Bias data pointer. Data type: int32
[in]	output_dims	Output tensor dimensions. Format: [1, H, W, C_OUT]
[in,out]	output_data	Output data pointer. Data type: int8

Returns

The function returns ARM_CMSIS_NN_SUCCESS - Successful completion.

• Supported framework: TensorFlow Lite

arm_depthwise_conv_wrapper_s8()

arm_cmsis_nn_status arm_depthwise_conv_wrapper_s8	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_dw_conv_params *	dw_conv_params,
		const cmsis_nn_per_channel_quant_params *	quant_params,
		const cmsis_nn_dims *	input_dims,
		const int8_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const int8_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const int32_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		int8_t *	output_data
	)

Wrapper function to pick the right optimized s8 depthwise convolution function.

Parameters

[in,out]	ctx	Function context (e.g. temporary buffer). Check the function definition file to see if an additional buffer is required. Optional function {API}_get_buffer_size() provides the buffer size if required. The caller is expected to clear the buffer, if applicable, for security reasons.
[in]	dw_conv_params	Depthwise convolution parameters (e.g. strides, dilations, pads,...) dw_conv_params->dilation is not used. Range of dw_conv_params->input_offset : [-127, 128] Range of dw_conv_params->output_offset : [-128, 127]
[in]	quant_params	Per-channel quantization info. It contains the multiplier and shift values to be applied to each output channel
[in]	input_dims	Input (activation) tensor dimensions. Format: [H, W, C_IN] Batch argument N is not used and assumed to be 1.
[in]	input_data	Input (activation) data pointer. Data type: int8
[in]	filter_dims	Filter tensor dimensions. Format: [1, H, W, C_OUT]
[in]	filter_data	Filter data pointer. Data type: int8
[in]	bias_dims	Bias tensor dimensions. Format: [C_OUT]
[in]	bias_data	Bias data pointer. Data type: int32
[in]	output_dims	Output tensor dimensions. Format: [1, H, W, C_OUT]
[in,out]	output_data	Output data pointer. Data type: int8

Returns

The function returns ARM_CMSIS_NN_SUCCESS - Successful completion.

• Supported framework: TensorFlow Lite
• Picks one of the the following functions
  1. arm_depthwise_conv_s8()
  2. arm_depthwise_conv_3x3_s8() - Cortex-M CPUs with DSP extension only
  3. arm_depthwise_conv_s8_opt()
• Check details of arm_depthwise_conv_s8_opt() for potential data that can be accessed outside of the boundary.

arm_depthwise_nhwc_conv_f16()

arm_cmsis_nn_status arm_depthwise_nhwc_conv_f16	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_dw_conv_params_f16 *	dw_conv_params,
		const cmsis_nn_dims *	input_dims,
		const float16_t *	input,
		const cmsis_nn_dims *	filter_dims,
		const float16_t *	kernel,
		const cmsis_nn_dims *	bias_dims,
		const float16_t *	bias,
		const cmsis_nn_dims *	output_dims,
		float16_t *	output
	)

Depthwise convolution, NHWC layout.

Parameters

[in,out]	ctx	Function context that may hold a temporary scratch buffer.
[in]	dw_conv_params	Depthwise convolution parameters (stride, padding, dilation, channel multiplier and activation clamp).
[in]	input_dims	Input tensor dimensions in NHWC format.
[in]	input	Pointer to the input tensor data.
[in]	filter_dims	Filter tensor dimensions in NHWC-compatible depthwise format.
[in]	kernel	Pointer to the filter tensor data.
[in]	bias_dims	Bias tensor dimensions. Format: [C_OUT].
[in]	bias	Optional bias tensor data.
[in]	output_dims	Output tensor dimensions in NHWC format.
[out]	output	Pointer to the output tensor data.

Note

When ctx->buf is used for internal kernel repacking, it must be aligned to the element type stored in scratch: at least 4-byte aligned for float32_t and, via at least 2-byte aligned for float16_t.

at least 2-byte aligned for float16_t.

Returns

ARM_CMSIS_NN_SUCCESS on success or ARM_CMSIS_NN_ARG_ERROR on invalid arguments.

arm_depthwise_nhwc_conv_f32()

arm_cmsis_nn_status arm_depthwise_nhwc_conv_f32	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_dw_conv_params_f32 *	dw_conv_params,
		const cmsis_nn_dims *	input_dims,
		const float32_t *	input,
		const cmsis_nn_dims *	filter_dims,
		const float32_t *	kernel,
		const cmsis_nn_dims *	bias_dims,
		const float32_t *	bias,
		const cmsis_nn_dims *	output_dims,
		float32_t *	output
	)

Depthwise convolution, NHWC layout.

Parameters

[in,out]	ctx	Function context that may hold a temporary scratch buffer.
[in]	dw_conv_params	Depthwise convolution parameters (stride, padding, dilation, channel multiplier and activation clamp).
[in]	input_dims	Input tensor dimensions in NHWC format.
[in]	input	Pointer to the input tensor data.
[in]	filter_dims	Filter tensor dimensions in NHWC-compatible depthwise format.
[in]	kernel	Pointer to the filter tensor data.
[in]	bias_dims	Bias tensor dimensions. Format: [C_OUT].
[in]	bias	Optional bias tensor data.
[in]	output_dims	Output tensor dimensions in NHWC format.
[out]	output	Pointer to the output tensor data.

Note

When ctx->buf is used for internal kernel repacking, it must be aligned to the element type stored in scratch: at least 4-byte aligned for float32_t and, via at least 2-byte aligned for float16_t.

at least 2-byte aligned for float16_t.

Returns

ARM_CMSIS_NN_SUCCESS on success or ARM_CMSIS_NN_ARG_ERROR on invalid arguments.

arm_transpose_conv_f16()

arm_cmsis_nn_status arm_transpose_conv_f16	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_context *	output_ctx,
		const cmsis_nn_transpose_conv_params_f16 *	transpose_conv_params,
		const cmsis_nn_dims *	input_dims,
		const float16_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const float16_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const float16_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		float16_t *	output_data,
		arm_nn_tensor_layout	layout
	)

Transpose convolution, dispatch by layout.

Parameters

[in,out]	ctx	Function context that may hold a temporary scratch buffer.
[in,out]	output_ctx	Output accumulation context for helper implementations.
[in]	transpose_conv_params	Transpose convolution parameters.
[in]	input_dims	Input tensor dimensions.
[in]	input_data	Pointer to the input tensor data.
[in]	filter_dims	Filter tensor dimensions.
[in]	filter_data	Pointer to the filter tensor data.
[in]	bias_dims	Bias tensor dimensions.
[in]	bias_data	Optional bias tensor data.
[in]	output_dims	Output tensor dimensions.
[out]	output_data	Pointer to the output tensor data.
[in]	layout	Tensor layout selector. Current float APIs require ARM_NN_LAYOUT_NHWC.

Returns

ARM_CMSIS_NN_SUCCESS on success or ARM_CMSIS_NN_ARG_ERROR on invalid arguments.

arm_transpose_conv_f16_get_buffer_size()

int32_t arm_transpose_conv_f16_get_buffer_size	(	const cmsis_nn_transpose_conv_params_f16 *	transpose_conv_params,
		const cmsis_nn_dims *	input_dims,
		const cmsis_nn_dims *	filter_dims,
		const cmsis_nn_dims *	out_dims
	)

Get the temporary buffer size required by transpose convolution.

Parameters

[in]	transpose_conv_params	Transpose convolution parameters.
[in]	input_dims	Input tensor dimensions.
[in]	filter_dims	Filter tensor dimensions.
[in]	out_dims	Output tensor dimensions.

Returns

Required buffer size in bytes, or 0 when no scratch buffer is needed.

arm_transpose_conv_f16_get_reverse_conv_buffer_size()

int32_t arm_transpose_conv_f16_get_reverse_conv_buffer_size	(	const cmsis_nn_transpose_conv_params_f16 *	transpose_conv_params,
		const cmsis_nn_dims *	input_dims,
		const cmsis_nn_dims *	filter_dims
	)

Get the reverse-convolution workspace size used by transpose convolution helpers.

Parameters

[in]	transpose_conv_params	Transpose convolution parameters.
[in]	input_dims	Input tensor dimensions.
[in]	filter_dims	Filter tensor dimensions.

Returns

Required buffer size in bytes, or 0 when no reverse-convolution buffer is needed.

arm_transpose_conv_f32()

arm_cmsis_nn_status arm_transpose_conv_f32	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_context *	output_ctx,
		const cmsis_nn_transpose_conv_params_f32 *	transpose_conv_params,
		const cmsis_nn_dims *	input_dims,
		const float32_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const float32_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const float32_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		float32_t *	output_data,
		arm_nn_tensor_layout	layout
	)

Transpose convolution, dispatch by layout.

Parameters

[in,out]	ctx	Function context that may hold a temporary scratch buffer.
[in,out]	output_ctx	Output accumulation context for helper implementations.
[in]	transpose_conv_params	Transpose convolution parameters.
[in]	input_dims	Input tensor dimensions.
[in]	input_data	Pointer to the input tensor data.
[in]	filter_dims	Filter tensor dimensions.
[in]	filter_data	Pointer to the filter tensor data.
[in]	bias_dims	Bias tensor dimensions.
[in]	bias_data	Optional bias tensor data.
[in]	output_dims	Output tensor dimensions.
[out]	output_data	Pointer to the output tensor data.
[in]	layout	Tensor layout selector. Current float APIs require ARM_NN_LAYOUT_NHWC.

Returns

ARM_CMSIS_NN_SUCCESS on success or ARM_CMSIS_NN_ARG_ERROR on invalid arguments.

arm_transpose_conv_f32_get_buffer_size()

int32_t arm_transpose_conv_f32_get_buffer_size	(	const cmsis_nn_transpose_conv_params_f32 *	transpose_conv_params,
		const cmsis_nn_dims *	input_dims,
		const cmsis_nn_dims *	filter_dims,
		const cmsis_nn_dims *	out_dims
	)

Get the temporary buffer size required by transpose convolution.

Parameters

[in]	transpose_conv_params	Transpose convolution parameters.
[in]	input_dims	Input tensor dimensions.
[in]	filter_dims	Filter tensor dimensions.
[in]	out_dims	Output tensor dimensions.

Returns

Required buffer size in bytes, or 0 when no scratch buffer is needed.

arm_transpose_conv_f32_get_reverse_conv_buffer_size()

int32_t arm_transpose_conv_f32_get_reverse_conv_buffer_size	(	const cmsis_nn_transpose_conv_params_f32 *	transpose_conv_params,
		const cmsis_nn_dims *	input_dims,
		const cmsis_nn_dims *	filter_dims
	)

Get the reverse-convolution workspace size used by transpose convolution helpers.

Parameters

[in]	transpose_conv_params	Transpose convolution parameters.
[in]	input_dims	Input tensor dimensions.
[in]	filter_dims	Filter tensor dimensions.

Returns

Required buffer size in bytes, or 0 when no reverse-convolution buffer is needed.

arm_transpose_conv_nhwc_f16()

arm_cmsis_nn_status arm_transpose_conv_nhwc_f16	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_context *	output_ctx,
		const cmsis_nn_transpose_conv_params_f16 *	transpose_conv_params,
		const cmsis_nn_dims *	input_dims,
		const float16_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const float16_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const float16_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		float16_t *	output_data
	)

Transpose convolution, NHWC layout.

arm_transpose_conv_nhwc_f32()

arm_cmsis_nn_status arm_transpose_conv_nhwc_f32	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_context *	output_ctx,
		const cmsis_nn_transpose_conv_params_f32 *	transpose_conv_params,
		const cmsis_nn_dims *	input_dims,
		const float32_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const float32_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const float32_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		float32_t *	output_data
	)

Transpose convolution, NHWC layout.

arm_transpose_conv_s8()

arm_cmsis_nn_status arm_transpose_conv_s8	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_context *	output_ctx,
		const cmsis_nn_transpose_conv_params *	transpose_conv_params,
		const cmsis_nn_per_channel_quant_params *	quant_params,
		const cmsis_nn_dims *	input_dims,
		const int8_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const int8_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const int32_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		int8_t *	output_data
	)

Basic s8 transpose convolution function.

Parameters

[in,out]	ctx	Function context that contains the additional buffer if required by the function. arm_transpose_conv_s8_get_buffer_size will return the buffer_size if required. The caller is expected to clear the buffer, if applicable, for security reasons.
[in,out]	output_ctx	Temporary scratch buffer. The size required size is: output width * output height * output channel * 4 The caller is expected to clear the buffer, if applicable, for security reasons.
[in]	transpose_conv_params	Convolution parameters (e.g. strides, dilations, pads,...). Range of transpose_conv_params->input_offset : [-127, 128] Range of transpose_conv_params->output_offset : [-128, 127]
[in]	quant_params	Per-channel quantization info. It contains the multiplier and shift values to be applied to each out channel.
[in]	input_dims	Input (activation) tensor dimensions. Format: [N, H, W, C_IN]
[in]	input_data	Input (activation) data pointer. Data type: int8
[in]	filter_dims	Filter tensor dimensions. Format: [C_OUT, HK, WK, C_IN] where HK and WK are the spatial filter dimensions
[in]	filter_data	Filter data pointer. Data type: int8
[in]	bias_dims	Bias tensor dimensions. Format: [C_OUT]
[in]	bias_data	Optional bias data pointer. Data type: int32
[in]	output_dims	Output tensor dimensions. Format: [N, H, W, C_OUT]
[out]	output_data	Output data pointer. Data type: int8

Returns

The function returns either ARM_CMSIS_NN_ARG_ERROR if argument constraints fail. or, ARM_CMSIS_NN_SUCCESS on successful completion.

1. Supported framework: TensorFlow Lite micro
2. Additional memory is required for optimization. Refer to arguments 'ctx' and 'output_ctx' for details.

arm_transpose_conv_wrapper_f16()

arm_cmsis_nn_status arm_transpose_conv_wrapper_f16	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_context *	output_ctx,
		const cmsis_nn_transpose_conv_params_f16 *	transpose_conv_params,
		const cmsis_nn_dims *	input_dims,
		const float16_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const float16_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const float16_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		float16_t *	output_data,
		arm_nn_tensor_layout	layout
	)

Transpose convolution wrapper using the CMSIS-NN baseline path.

arm_transpose_conv_wrapper_f32()

arm_cmsis_nn_status arm_transpose_conv_wrapper_f32	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_context *	output_ctx,
		const cmsis_nn_transpose_conv_params_f32 *	transpose_conv_params,
		const cmsis_nn_dims *	input_dims,
		const float32_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const float32_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const float32_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		float32_t *	output_data,
		arm_nn_tensor_layout	layout
	)

Transpose convolution wrapper using the CMSIS-NN baseline path.

arm_transpose_conv_wrapper_s8()

arm_cmsis_nn_status arm_transpose_conv_wrapper_s8	(	const cmsis_nn_context *	ctx,
		const cmsis_nn_context *	output_ctx,
		const cmsis_nn_transpose_conv_params *	transpose_conv_params,
		const cmsis_nn_per_channel_quant_params *	quant_params,
		const cmsis_nn_dims *	input_dims,
		const int8_t *	input_data,
		const cmsis_nn_dims *	filter_dims,
		const int8_t *	filter_data,
		const cmsis_nn_dims *	bias_dims,
		const int32_t *	bias_data,
		const cmsis_nn_dims *	output_dims,
		int8_t *	output_data
	)

Wrapper to select optimal transposed convolution algorithm depending on parameters.

Parameters

[in,out]	ctx	Function context that contains the additional buffer if required by the function. arm_transpose_conv_s8_get_buffer_size will return the buffer_size if required. The caller is expected to clear the buffer, if applicable, for security reasons.
[in,out]	output_ctx	Temporary scratch buffer. The size required size is: output width * output height * output channel * 4 The caller is expected to clear the buffer, if applicable, for security reasons.
[in]	transpose_conv_params	Convolution parameters (e.g. strides, dilations, pads,...). Range of transpose_conv_params->input_offset : [-127, 128] Range of transpose_conv_params->output_offset : [-128, 127]
[in]	quant_params	Per-channel quantization info. It contains the multiplier and shift values to be applied to each out channel.
[in]	input_dims	Input (activation) tensor dimensions. Format: [N, H, W, C_IN]
[in]	input_data	Input (activation) data pointer. Data type: int8
[in]	filter_dims	Filter tensor dimensions. Format: [C_OUT, HK, WK, C_IN] where HK and WK are the spatial filter dimensions
[in]	filter_data	Filter data pointer. Data type: int8
[in]	bias_dims	Bias tensor dimensions. Format: [C_OUT]
[in]	bias_data	Optional bias data pointer. Data type: int32
[in]	output_dims	Output tensor dimensions. Format: [N, H, W, C_OUT]
[out]	output_data	Output data pointer. Data type: int8

Returns

The function returns either ARM_CMSIS_NN_ARG_ERROR if argument constraints fail. or, ARM_CMSIS_NN_SUCCESS on successful completion.

1. Supported framework: TensorFlow Lite micro
2. Additional memory is required for optimization. Refer to arguments 'ctx' and 'output_ctx' for details.