Integer ceiling division operation.

Decorator to check if the specified tile has the expected data type.

Decorator to check if the specified tile has the expected data type.

Decorator to check if lhs and rhs have the same tile type.

Implementation of `classof` that supports all of the potential terminator operations.

Performs element-wise comparison of two tiles.

Computes the cosine of the source tile element-wise.

Computes the hyperbolic cosine of the source tile.

===----------------------------------------------------------------------===// Helper for Operation Creation and Result Handling ===------------------

Helper function that inserts a new cuda_tile.global in the enclosing module and a cuda_tile.get_global at the current insertion point.

Creates a kernel entry in the CUDA Tile IR module. Args: kernel_name (str): The name of the kernel entry. ar

Creates a CUDA Tile IR module.

This is a decorator that needs to be used in each cuda_tile OP to manage pre-generation things. Currently, it only generate source locati

Dispatch to the correct op writer. Returns the number of results that were serialized.

Raises e to the power of source.

Raises 2 to the power of source.

Extracts a slice from the source tile at the specified indices.

Computes the floor of the source tile element-wise.

Constructs a for loop with the provided body. The body is a function taking as argument the iteration variables and building the operations w

Provide custom aliasing for debug info attributes.

Helper function to retrieve an attribute from SM-specific or default entry. Returns the attribute if found, otherwise std::nullopt.

Look up a constant by attribute without adding it

Returns a CudaTile location for a given location attribute.

This method gets or creates an index for a debug info attribute.

Returns a subprogram attribute for a given local scope attribute.

Get number of tile blocks.

Get the ID of the current tile block.

Returns the MLIR index type for this tensor view.