operator

Classes

Conv2DOperator

Conv2DOperator(op: AirOperator, model: AirModel, platform: SocPlatform, prefix: str = 'aot', attributes: dict[str, str] = {})

CONV_2D operator.

This operator performs 2D convolution on the input tensor with the given filter tensor.

Parameters:

• op

  (AirOperator) –

  The AIR operator to wrap.

• model

  (AirModel) –

  The AIR model.

• platform

  (SocPlatform) –

  The target platform for code generation.

• prefix

  (str, default: 'aot' ) –

  Prefix for generated code files. Defaults to "aot".

• attributes

  (dict[str, str], default: {} ) –

  Attributes for template values. Defaults to {}.

Functions

ctx_buf_required

ctx_buf_required() -> bool

Whether the selected convolution kernel dereferences a context buf.

Keyed on the kernel actually chosen in :meth:on_resolve, because CONV_2D's dispatch table mixes kernels that require an im2col buffer, kernels that require only the weight-sum buffer, and kernels that ignore the context entirely. Two independent conditions:

• :data:_CTX_BUF_KERNELS -- the kernel's entry point hard-requires ctx->buf.
• :data:_WEIGHT_SUM_BUF_KERNELS_MVE -- the kernel ignores ctx but forwards weight_sum_ctx->buf into a routine that indexes it unchecked under MVE.

Both are satisfied by the same any-of backing check, since :attr:CTX_BUF_BACKING_TENSOR_NAMES covers both tensor roles.

Returns:

• bool –

  True if the selected kernel needs a non-NULL context buffer.

• bool –

  False when no kernel has been selected yet.

compute_values

compute_values() -> dict[str, str]

Compute the values for the operator.

Returns:

• dict[str, str] –

  dict[str, str]: Computed values for the operator.

emit

emit(save_path: Path)

Generate the source code for the operator.

Parameters:

• save_path

  (Path) –

  Path to save the generated code.

Functions