Module cubecl_core::frontend

Cube Frontend Types.

Modules

• ABSOLUTE_POS
  The position of the working unit in the whole cube kernel, without regards to cubes and axis.
• ABSOLUTE_POS_X
  The index of the working unit in the whole cube kernel along the X axis, without regards to cubes.
• ABSOLUTE_POS_Y
  The index of the working unit in the whole cube kernel along the Y axis, without regards to cubes.
• ABSOLUTE_POS_Z
  The index of the working unit in the whole cube kernel along the Z axis, without regards to cubes.
• CUBE_COUNT
  The number of cubes launched.
• CUBE_COUNT_X
  The number of cubes launched along the X axis.
• CUBE_COUNT_Y
  The number of cubes launched along the Y axis.
• CUBE_COUNT_Z
  The number of cubes launched along the Z axis.
• CUBE_DIM
  The total amount of working units in a cube.
• CUBE_DIM_X
  The dimension of the cube along the X axis.
• CUBE_DIM_Y
  The dimension of the cube along the Y axis.
• CUBE_DIM_Z
  The dimension of the cube along the Z axis.
• CUBE_POS
  The cube position, without regards to axis.
• CUBE_POS_X
  The cube position along the X axis.
• CUBE_POS_Y
  The cube position along the Y axis.
• CUBE_POS_Z
  The cube position along the Z axis.
• SUBCUBE_DIM
  The total amount of working units in a subcube.
• UNIT_POS
  The position of the working unit inside the cube, without regards to axis.
• UNIT_POS_X
  The position of the working unit inside the cube along the X axis.
• UNIT_POS_Y
  The position of the working unit inside the cube along the Y axis.
• UNIT_POS_Z
  The position of the working unit inside the cube along the Z axis.
• add
• add_assign
• add_assign_array_op
• add_assign_op
• and
• assign
• bitand
• bitxor
• branch
• cmma
  This module exposes cooperative matrix-multiply and accumulate operations.
• div
• div_assign_array_op
• div_assign_op
• eq
• ge
• gt
• index
• index_assign
• le
• lt
• mul
• mul_assign_array_op
• mul_assign_op
• ne
• not
• or
• rem
• shl
• shr
• sub
• sub_assign_array_op
• sub_assign_op
• subcube_all
  Module containing the expand function for subcube_all().
• subcube_max
  Module containing the expand function for subcube_max().
• subcube_min
  Module containing the expand function for subcube_min().
• subcube_prod
  Module containing the expand function for subcube_prod().
• subcube_sum
  Module containing the expand function for subcube_sum().
• synchronization

Structs

• Array
  A contiguous array of elements.
• ArrayHandleRef
  Tensor representation with a reference to the server handle.
• AtomicI32
  An unsigned atomic integer. Can only be acted on atomically.
• AtomicI64
  An unsigned atomic integer. Can only be acted on atomically.
• AtomicUInt
  An atomic version of UInt. Can only be acted on atomically. An atomic unsigned int.
• BF16
• Comptime
  Encapsulates a value to signify it must be used at compilation time rather than in the kernel
• CubeContext
• ExpandElementTyped
  Expand type associated with a type.
• F16
• F32
• F64
• I32
• I64
• ScalarArg
• Sequence
  A sequence of cube types that is inlined during compilation.
• SequenceExpand
  Expand type of Sequence.
• SharedMemory
• Slice
  A read-only contiguous list of elements
• SliceMut
  A read-write contiguous list of elements.
• Tensor
  The tensor type is similar to the array type, however it comes with more metadata such as stride and shape.
• TensorHandleRef
  Tensor representation with a reference to the server handle, the strides and the shape.
• UInt
  An unsigned int. Preferred for indexing operations
• VariablePool

Enums

• ArrayArg
• ExpandElement
  Reference to a JIT variable
• TensorArg
  Argument to be used for tensors passed as arguments to kernels.

Constants

• ABSOLUTE_POS
  The position of the working unit in the whole cube kernel, without regards to cubes and axis.
• ABSOLUTE_POS_X
  The index of the working unit in the whole cube kernel along the X axis, without regards to cubes.
• ABSOLUTE_POS_Y
  The index of the working unit in the whole cube kernel along the Y axis, without regards to cubes.
• ABSOLUTE_POS_Z
  The index of the working unit in the whole cube kernel along the Z axis, without regards to cubes.
• CUBE_COUNT
  The number of cubes launched.
• CUBE_COUNT_X
  The number of cubes launched along the X axis.
• CUBE_COUNT_Y
  The number of cubes launched along the Y axis.
• CUBE_COUNT_Z
  The number of cubes launched along the Z axis.
• CUBE_DIM
  The total amount of working units in a cube.
• CUBE_DIM_X
  The dimension of the cube along the X axis.
• CUBE_DIM_Y
  The dimension of the cube along the Y axis.
• CUBE_DIM_Z
  The dimension of the cube along the Z axis.
• CUBE_POS
  The cube position, without regards to axis.
• CUBE_POS_X
  The cube position along the X axis.
• CUBE_POS_Y
  The cube position along the Y axis.
• CUBE_POS_Z
  The cube position along the Z axis.
• SUBCUBE_DIM
  The total amount of working units in a subcube.
• UNIT_POS
  The position of the working unit inside the cube, without regards to axis.
• UNIT_POS_X
  The position of the working unit inside the cube along the X axis.
• UNIT_POS_Y
  The position of the working unit inside the cube along the Y axis.
• UNIT_POS_Z
  The position of the working unit inside the cube along the Z axis.

Traits

• Abs
• ArgSettings
  Defines the argument settings used to launch a kernel.
• Atomic
  An atomic type. Represents an shared value that can be operated on atomically.
• BitCast
  Enables reinterpet-casting/bitcasting from any floating point value to any integer value and vice versa
• BoolOps
  Extension trait for bool.
• Cast
  Enable elegant casting from any to any CubeElem
• Ceil
• Clamp
• ComptimeType
  Type that can be used within Comptime.
• Cos
• CubePrimitive
  Form of CubeType that encapsulates all primitive types: Numeric, UInt, Bool
• CubeType
  Types used in a cube function must implement this trait
• Erf
• Exp
• ExpandElementBaseInit
• Float
  Floating point numbers. Used as input in float kernels
• Floor
• Init
  Trait to be implemented by cube types implementations.
• Int
  Signed integer. Used as input in int kernels
• LaunchArg
  Defines a type that can be used as argument to a kernel.
• LaunchArgExpand
  Defines how a launch argument can be expanded.
• Log
• Log1p
• Max
• Min
• Numeric
  Type that encompasses both (unsigned or signed) integers and floats Used in kernels that should work for both.
• Powf
• Recip
• Remainder
• ScalarArgSettings
  Similar to ArgSettings, however only for scalar types that don’t depend on the Runtime trait.
• Sin
• SliceOperator
• SliceOperatorExpand
• Sqrt
• Tanh
• Vectorized

Functions

• array_assign_binary_op_expand
• fma
  Fused multiply-add A*B+C.
• fma_expand
  Expand method of fma.
• init_expand
• slice_expand
• subcube_all
  Perform a reduce all operation across all units in a subcube.
• subcube_elect
  Returns true if the cube unit has the lowest subcube_unit_id among active unit in the subcube
• subcube_elect_expand
• subcube_max
  Perform a reduce max operation across all units in a subcube.
• subcube_min
  Perform a reduce min operation across all units in a subcube.
• subcube_prod
  Perform a reduce prod operation across all units in a subcube.
• subcube_sum
  Perform a reduce sum operation across all units in a subcube.
• unary_expand

Type Aliases

• Bool
  Boolean type.