iron.kernel.BaseKernel Class Reference

Inheritance diagram for iron.kernel.BaseKernel:

Collaboration diagram for iron.kernel.BaseKernel:

Public Member Functions
	__init__ (self, str name, list[type[np.ndarray]|np.dtype]|None arg_types=None)
tuple[int,...]	arg_shape (self, int arg_index=0)
	arg_dtype (self, int arg_index=0)
int	tile_size (self, int arg_index=0)
list	arg_types (self)
	__call__ (self, *args, **kwargs)
Public Member Functions inherited from iron.resolvable.Resolvable
None	resolve (self, ir.Location|None loc=None, ir.InsertionPoint|None ip=None)
list	tiles (self)

Protected Member Functions
	_resolve_arg (self, int arg_index)

Protected Attributes
	_name
	_arg_types
	_op

Detailed Description

Base class for AIE core functions that resolve to a func.func declaration.

Subclasses:
    Kernel: wraps a pre-compiled object file.
    ExternalFunction: compiles C/C++ source at JIT time.

Constructor & Destructor Documentation

__init__()

iron.kernel.BaseKernel.__init__	(		self,
		str	name,
		list[type[np.ndarray] | np.dtype] | None	arg_types = None
	)

Args:
    name: Symbol name of the function.
    arg_types: Type signature of the function arguments.  Defaults to None (empty list).

Reimplemented in iron.kernel.Kernel, and iron.kernel.ExternalFunction.

Member Function Documentation

__call__()

iron.kernel.BaseKernel.__call__	(		self,
		*	args,
		**	kwargs
	)

Emit a func.call to this kernel, validating argument count.

Each argument is passed through :func:`_maybe_collapse_to_match`
before the call.  This silently inserts a ``memref.collapse_shape``
when an N-D contiguous memref arg is being fed into a 1-D kernel
signature with the same element count and dtype — the typical case
when an iron design holds 2-D ObjectFifo elements but the
``aie.iron.kernels.X`` helper declares a flat 1-D arg.  See that
helper's docstring for the full set of conditions.  Real shape /
dtype mismatches still fail at MLIR verification time.

``**kwargs`` are forwarded to the underlying ``func.call`` builder
(typically ``loc=``, ``ip=`` for MLIR location / insertion point).

Reimplemented in iron.kernel.ExternalFunction.

_resolve_arg()

iron.kernel.BaseKernel._resolve_arg (   self, int  arg_index  )

protected

Validate ``arg_index`` and return the underlying type entry.

arg_dtype()

iron.kernel.BaseKernel.arg_dtype	(		self,
		int	arg_index = 0
	)

Return the numpy dtype of the array argument at ``arg_index``.

Args:
    arg_index: Index into ``arg_types``.  Defaults to 0.

Raises:
    ValueError: When ``arg_index`` is out of range or the
        argument at that index is not an array type.

arg_shape()

tuple[int, ...] iron.kernel.BaseKernel.arg_shape	(		self,
		int	arg_index = 0
	)

Return the shape tuple of the array argument at ``arg_index``.

Works for both ``np.ndarray[(...,), np.dtype[T]]`` parameterized
types (the canonical iron kernel signature) and MLIR MemRefType
operands.

Args:
    arg_index: Index into ``arg_types``.  Defaults to 0.

Raises:
    ValueError: When ``arg_index`` is out of range or the
        argument at that index is not an array type.

arg_types()

list iron.kernel.BaseKernel.arg_types

(

self

)

Return a copy of the argument type list.

tile_size()

int iron.kernel.BaseKernel.tile_size	(		self,
		int	arg_index = 0
	)

Return the first dimension of the array argument at ``arg_index``.

Convenience wrapper over :meth:`arg_shape` for the common case of
a 1-D buffer argument.  ``tile_size(i)`` is equivalent to
``arg_shape(i)[0]``.

Args:
    arg_index: Index into ``arg_types``.  Defaults to 0.

Member Data Documentation

_arg_types

iron.kernel.BaseKernel._arg_types

protected

_name

iron.kernel.BaseKernel._name

protected

_op

iron.kernel.BaseKernel._op

protected

---

The documentation for this class was generated from the following file:

• iron/kernel.py