tilelang.language.customize¶

Some customized operations frequently used in tensor programming, exposed on the TileLang language surface.

函式¶

`dp4a`(A, B, C)	Perform a 4-element dot product with accumulation (DP4A).
`clamp`(dst, min_val, max_val)	Clamps the input value dst between [min_val, max_val]
`reshape`(src, shape)	Reshapes the input buffer to the specified shape.
`view`(src[, shape, dtype])	Return a Tensor view of the input buffer with an optional new shape and dtype.
`loop_break`()	Break out of the current loop.

Module Contents¶

tilelang.language.customize.dp4a(A, B, C)¶

Perform a 4-element dot product with accumulation (DP4A).

參數:

A (Buffer) -- First input buffer
B (Buffer) -- Second input buffer
C (Buffer) -- Accumulation buffer

回傳:

Handle to the DP4A operation

回傳型別:

PrimExpr

tilelang.language.customize.clamp(dst, min_val, max_val)¶

Clamps the input value dst between [min_val, max_val]

參數:

dst (tvm.tir.PrimExpr) -- Input value to be clamped
min_val (tvm.tir.PrimExpr) -- Minimum value
max_val (tvm.tir.PrimExpr) -- Maximum value

回傳:

Value clamped to the specified range

回傳型別:

tvm.tir.PrimExpr

tilelang.language.customize.reshape(src, shape)¶

Reshapes the input buffer to the specified shape.

參數:

src (Buffer) -- Input buffer to be reshaped
shape (ShapeType) -- New shape for the buffer

回傳:

A new buffer view with the specified shape

回傳型別:

Buffer

tilelang.language.customize.view(src, shape=None, dtype=None)¶

Return a Tensor view of the input buffer with an optional new shape and dtype.

If shape is None the source buffer's shape is used; if dtype is None the source buffer's dtype is used. The returned buffer shares the same underlying data as src (no copy).

參數:

src (tvm.tir.Buffer)
shape (tilelang._typing.ShapeType | None)
dtype (tilelang._typing.DType | None)

回傳型別:

tvm.tir.Buffer

tilelang.language.customize.loop_break()¶

Break out of the current loop.

回傳:: A call to the tl.loop_break intrinsic.
回傳型別:: tir.Call