Returns the version tag.

Get trait constraint message by name

Convert an enum value to the corresponding ComparisonOrderingAttr.

Get all supported MMA configurations

Get all supported MMA Scaled configurations

Convert an enum value to SymbolVisibilityAttr.

Create a cuda_tile.global in the enclosing cuda_tile.module.

Helper function to return if step is equal to one

Return True if splitting loop for current branch seems profitable for performance

Check if an cuda_tile.if condition is a cmpi with induction variable. Collect all branches with the same predicate into `ifOps` vector

Loads/produces a ModuleOp into `outMod` based on cfg.input.kind. Returns success() on success, failure() on any error.

Creates a PartitionViewType from a tensor view MLIR type, a tile shape, the type of the indices to use within the view, a dimension mapping a

Creates a StridedViewType from a tensor view MLIR type, a tile shape, traversal strides, the type of the indices to use within the view, a

Creates a TensorViewType from an element, a shape and strides. Args: el_type: Element type - can be a type wrapper (Int32, Float32, etc.)

===----------------------------------------------------------------------===// CUDA Tile IR -> CUDA Tile IR optimization pipeline ===-----------------

Computes the negative of the source tile element-wise.

Computes the arithmetic inverse of the source integer tile element-wise.

Normalize a comparison to always be "iv <op> value" Return false if comparison is not with induction variable Return false if comparison signedness do

----------------------------------------------------------------------------- optimizeTileIR - calls: 1) loadInputModule - from file or buffer: Byteco