Releases ownership of the managed object, if any

Get the strides of the tensor. Extracts the strides (third field) from the LLVM struct value. :return: A tuple of integers represent

Wrapper for binary operations on Numeric types. This wrapper handles type promotion, operation execution, and result type determination for b

Compute element-wise exponential of the input tensor. :param a: Input tensor :type a: Union[TensorSSA, Numeric] :param fastmath: Enable f

Number of iterations per output tile

Issuing the launch_dependents instruction hints a dependent kernel to launch earlier launch_dependents doesn't impact the functionality but the perfor

Attempt to acquire the current buffer without blocking. This method tries to acquire the current buffer stage for producing data with

Lightweight update given a subset of arguments

Create a getelementptr operation. Parameters ---------- ptr : ir.Value The pointer to the element. indice

Constructs a concise readable pretty string.

< Layout function

Creates an AST node representing a qualified attribute access to a function in a module or submodule. :param func_name: The attribute or function

Add an operation to the list of supported kernels

CUTLASS helper to get SMEM pointer

Match function used in get_underlying_impl

Get the shape of the tensor. Extracts the shape (second field) from the LLVM struct value. :return: A tuple of integers representing

Converts ``value`` to the corresponding storage needed for the ctype that will store ``value``.

Align a pointer to a specified byte alignment. :param min_align: The minimum byte alignment requirement. Must be a power of 2.

Returns a lane index in the warp. The threads in warp may not be convergent

Returns the dim3 cluster shape.

Constructs a ``cutlass_cppgen.backend.GemmUniversalOperation`` based on the input parameters and current kernel specification of the

Generate an IfOp with optional else branch and return values. Args: cond: The condition expression then_body: Function to ex

Returns true if the current coordinate is within the filter tensor W

Constructor

< Efficiently disables all accesses guarded by mask

Input A: float_e4m3_t ; Input B: float_e5m2_t

Converts to float

Converts to float

Returns a pointer

Loads a fragment from memory with additional logical offset

Obtains a Coord<2> from GemmCoord

Accesses raw internal state

Dispatch the function to either a TensorSSA or a Numeric(Float). :param func: The function to dispatch :param args: The input tensor or scala

* Find quotient and remainder using device-side intrinsics */

Value

Get all the input nodes sorted by edge weight

Get the value of a boolean environment variable. If the value it not in False, 0, or empty string, it is considered True. Note that the v

Returns the logical number of elements stored in the host tensor

Wait until all TMA descriptor previously issued are safe to be modified after tma_desc_commit_group()

Check a condition and throw an error if false. Parameters ---------- current_block : ir.Block The current block.

Create cute tensor, handling float8 types that don't support dlpack. For float8 types, the storage_tensor should be uint8 (for DLPack compatibili

Returns the pointer to referenced data