Module cubecl_core::ir

Cube Language Internal Representation.

Structs

• BinaryOperator
• Binding
• CheckedIndex
  Perform a check bound on the index (lhs) of value (rhs)
• CheckedIndexAssign
  Perform a check bound on the index (lhs) of output before assigning the value (rhs)
• ClampOperator
• ConditionalAssign
  Assign value to a variable based on a given condition.
• CubeDim
• FmaOperator
• If
• IfElse
• IndexOffsetGlobalWithLayout
  Calculate the index offset for all tensor variables provided compatible with the given layout.
• InitOperator
• Item
• KernelDefinition
• Loop
• Matrix
• RangeLoop
• ReadGlobal
  Read a global array.
• ReadGlobalOperator
• ReadGlobalWithLayout
  Read a global array with the given layout.
• ReadGlobalWithLayoutOperator
• Scope
  The scope is the main operation and variable container that simplify the process of reading inputs, creating local variables and adding new operations.
• SliceOperator
• UnaryOperator
• UnrolledRangeLoop
• WriteGlobal
  Write to a global array.

Enums

• Branch
  All branching types.
• ConstantScalarValue
  The scalars are stored with the highest precision possible, but they might get reduced during compilation.
• CoopMma
  Cooperative Matrix-Multiply and Accumulate Instruction.
• Elem
• FloatKind
• IntKind
• Location
• MatrixIdent
• MatrixLayout
• Metadata
  All metadata that can be access in a shader.
• Operation
  All operations that can be used in a GPU compute shader.
• Operator
  All operators that can be used in a GPU compute shader.
• Procedure
  Tensor operations that can’t be executed with a simple operator should use a procedure.
• ReadingStrategy
• Subcube
  All subcube operations.
• Synchronization
  All synchronization types.
• Variable
• Visibility

Type Aliases

• Vectorization