buffer_sizes

Classes

Functions

conv_is_1x1

conv_is_1x1(conv_params, input_dims, filter_dims, output_dims) -> bool

Checks if the convolution operator is 1x1.

Parameters:

• conv_params

  (dict) –

  Convolution parameters.

• input_dims

  (list) –

  Input dimensions.

• filter_dims

  (list) –

  Filter dimensions.

• output_dims

  (list) –

  Output dimensions.

Returns:

• bool ( bool ) –

  True if the convolution is 1x1, False otherwise.

conv_is_1x1_fast

conv_is_1x1_fast(conv_params, input_dims, filter_dims, output_dims) -> bool

Checks if the convolution operator is 1x1 and maps to the fast kernel implementation.

Parameters:

• conv_params

  (dict) –

  Convolution parameters.

• input_dims

  (list) –

  Input dimensions.

• filter_dims

  (list) –

  Filter dimensions.

• output_dims

  (list) –

  Output dimensions.

Returns:

• bool ( bool ) –

  True if the convolution is 1x1 and maps to the fast kernel implementation, False otherwise.

conv_is_1xn

conv_is_1xn(conv_params, input_dims, filter_dims, output_dims) -> bool

Checks if the convolution operator is (1D).

Parameters:

• conv_params

  (dict) –

  Convolution parameters.

• input_dims

  (list) –

  Input dimensions.

• filter_dims

  (list) –

  Filter dimensions.

• output_dims

  (list) –

  Output dimensions.

Returns:

• bool ( bool ) –

  True if the convolution is 1xn, False otherwise.

conv_1x1_s8_fast_get_buffer_size_dsp

conv_1x1_s8_fast_get_buffer_size_dsp(conv_params, input_dims, filter_dims, output_dims) -> int

Compute the buffer size for 1x1 convolution with fast kernel implementation.

Parameters:

• conv_params

  (dict) –

  Convolution parameters.

• input_dims

  (list) –

  Input dimensions.

• filter_dims

  (list) –

  Filter dimensions.

• output_dims

  (list) –

  Output dimensions.

Returns:

• int ( int ) –

  Buffer size in bytes.

conv_1xn_s8_get_buffer_size_mve

conv_1xn_s8_get_buffer_size_mve(conv_params, input_dims, filter_dims, output_dims) -> int

Compute the buffer size for 1xn convolution when using MVE.

Parameters:

• conv_params

  (dict) –

  Convolution parameters.

• input_dims

  (list) –

  Input dimensions.

• filter_dims

  (list) –

  Filter dimensions.

• output_dims

  (list) –

  Output dimensions.

Returns:

• int ( int ) –

  Buffer size in bytes.

conv_default_s8_get_buffer_size_dsp

conv_default_s8_get_buffer_size_dsp(conv_params, input_dims, filter_dims, output_dims) -> int

Compute the buffer size for default convolution when using DSP.

Parameters:

• conv_params

  (dict) –

  Convolution parameters.

• input_dims

  (list) –

  Input dimensions.

• filter_dims

  (list) –

  Filter dimensions.

• output_dims

  (list) –

  Output dimensions.

Returns:

• int ( int ) –

  Buffer size in bytes.

conv_1xn_s8_get_buffer_size_dsp

conv_1xn_s8_get_buffer_size_dsp(conv_params, input_dims, filter_dims, output_dims) -> int

Compute the buffer size for 1xn convolution when using DSP.

Parameters:

• conv_params

  (dict) –

  Convolution parameters.

• input_dims

  (list) –

  Input dimensions.

• filter_dims

  (list) –

  Filter dimensions.

• output_dims

  (list) –

  Output dimensions.

Returns:

• int ( int ) –

  Buffer size in bytes.

conv_default_s8_get_buffer_size_mve

conv_default_s8_get_buffer_size_mve(conv_params, input_dims, filter_dims, output_dims) -> int

Compute the buffer size for default convolution when using MVE.

Parameters:

• conv_params

  (dict) –

  Convolution parameters.

• input_dims

  (list) –

  Input dimensions.

• filter_dims

  (list) –

  Filter dimensions.

• output_dims

  (list) –

  Output dimensions.

Returns:

• int ( int ) –

  Buffer size in bytes.

conv_s8_get_buffer_size_mve

conv_s8_get_buffer_size_mve(conv_params, input_dims, filter_dims, output_dims) -> int

Compute the buffer size for convolution when using MVE.

Parameters:

• conv_params

  (dict) –

  Convolution parameters.

• input_dims

  (list) –

  Input dimensions.

• filter_dims

  (list) –

  Filter dimensions.

• output_dims

  (list) –

  Output dimensions.

Returns:

• int ( int ) –

  Buffer size in bytes.

conv_s8_get_buffer_size_dsp

conv_s8_get_buffer_size_dsp(conv_params, input_dims, filter_dims, output_dims) -> int

Compute the buffer size for convolution when using DSP.

Parameters:

• conv_params

  (dict) –

  Convolution parameters.

• input_dims

  (list) –

  Input dimensions.

• filter_dims

  (list) –

  Filter dimensions.

• output_dims

  (list) –

  Output dimensions.

Returns:

• int ( int ) –

  Buffer size in bytes.

conv_s8_get_buffer_size

conv_s8_get_buffer_size(conv_params, input_dims, filter_dims, output_dims, mode: str | None = None) -> int

Compute the buffer size for convolution.

Parameters:

• conv_params

  (dict) –

  Convolution parameters.

• input_dims

  (list) –

  Input dimensions.

• filter_dims

  (list) –

  Filter dimensions.

• output_dims

  (list) –

  Output dimensions.

• mode

  (str | None, default: None ) –

  Mode for buffer size calculation. Defaults to None.

Returns:

• int ( int ) –

  Buffer size in bytes.

dw_conv_s8_one_in_ch_get_buffer_size_mve

dw_conv_s8_one_in_ch_get_buffer_size_mve(dw_conv_params, input_dims, filter_dims, output_dims) -> int

Compute the buffer size for depthwise convolution with one input channel when using MVE.

Parameters:

• dw_conv_params

  (dict) –

  Depthwise convolution parameters.

• input_dims

  (list) –

  Input dimensions.

• filter_dims

  (list) –

  Filter dimensions.

• output_dims

  (list) –

  Output dimensions.

Returns:

• int ( int ) –

  Buffer size in bytes.

dw_conv_s8_opt_get_buffer_size_mve

dw_conv_s8_opt_get_buffer_size_mve(dw_conv_params, input_dims, filter_dims, output_dims) -> int

Compute the buffer size for optimized depthwise convolution when using MVE.

Parameters:

• dw_conv_params

  (dict) –

  Depthwise convolution parameters.

• input_dims

  (list) –

  Input dimensions.

• filter_dims

  (list) –

  Filter dimensions.

• output_dims

  (list) –

  Output dimensions.

Returns:

• int ( int ) –

  Buffer size in bytes.

dw_conv_s8_opt_get_buffer_size_dsp

dw_conv_s8_opt_get_buffer_size_dsp(dw_conv_params, input_dims, filter_dims, output_dims) -> int

Compute the buffer size for optimized depthwise convolution when using DSP.

Parameters:

• dw_conv_params

  (dict) –

  Depthwise convolution parameters.

• input_dims

  (list) –

  Input dimensions.

• filter_dims

  (list) –

  Filter dimensions.

• output_dims

  (list) –

  Output dimensions.

Returns:

• int ( int ) –

  Buffer size in bytes.

dw_conv_s8_get_buffer_size

dw_conv_s8_get_buffer_size(dw_conv_params, input_dims, filter_dims, output_dims, mode: str | None = None) -> int

Compute the buffer size for depthwise convolution.

Parameters:

• dw_conv_params

  (dict) –

  Depthwise convolution parameters.

• input_dims

  (list) –

  Input dimensions.

• filter_dims

  (list) –

  Filter dimensions.

• output_dims

  (list) –

  Output dimensions.

• mode

  (str | None, default: None ) –

  Mode for buffer size calculation. Defaults to None.

Returns:

• int ( int ) –

  Buffer size in bytes.

tp_conv_s8_get_buffer_size

tp_conv_s8_get_buffer_size(tp_conv_params, input_dims, filter_dims, output_dims, mode: str | None = None) -> int

Compute the buffer size for transposed convolution.

Parameters:

• tp_conv_params

  (dict) –

  Transposed convolution parameters.

• input_dims

  (list) –

  Input dimensions.

• filter_dims

  (list) –

  Filter dimensions.

• output_dims

  (list) –

  Output dimensions.

• mode

  (str | None, default: None ) –

  Mode for buffer size calculation. Defaults to None.

Returns:

• int ( int ) –

  Buffer size in bytes.