Crate cubecl

Re-exports§

benchmark: Module for benchmarking any executable part
channel: Compute channel module.
client: Compute client module.
codegen
compute
frontend: Cube Frontend Types.
future: Future utils with a compatible API for native, non-std and wasm environments.
ir: Cube Language Internal Representation.
prelude
server: Compute server module.
tune: Autotune module

comment: Insert a literal comment into the kernel source code.
comptime: Mark the contents of this macro as compile time values, turning off all expansion for this code and using it verbatim
cpa: Cube Pseudo Assembly.
debug_print: Print a formatted message using the target’s debug print facilities. The format string is target specific, but Vulkan and CUDA both use the C++ conventions. WGSL isn’t currently supported.
debug_print_expand: Print a formatted message using the target’s debug print facilities. The format string is target specific, but Vulkan and CUDA both use the C++ conventions. WGSL isn’t currently supported.
unexpanded

CubeDim
Execution
InplaceMapping: Simply indicate the output that can be replaced by the input.
KernelExpansion: The information necessary to compile a kernel definition.
KernelId: Kernel unique identifier.
KernelIntegrator: The kernel integrator allows you to create a kernel definition based on kernel expansion and kernel settings.
KernelSettings
Metadata: Helper to calculate metadata offsets based on buffer count and position
MetadataBuilder: Builder for a serialized metadata struct
flex32: A floating point type with relaxed precision, minimum f16, max [f32].
tf32: A 19-bit floating point type implementing the tfloat32 format.

AtomicFeature
CubeCount: Specifieds the number of cubes to be dispatched for a kernel.
CubeCountSettings: The position of the input or output to calculate the number of cubes to launch.
ExecutionMode: The kind of execution to be performed.
Feature: Every feature that can be supported by a cube runtime.
InputInfo: Information related to an input.
MemoryConfiguration: High level configuration of memory management.
OutputInfo: Information related to an output.

Compiler: Compiles the representation into its own representation that can be formatted into tokens.
CompilerRepresentation: Trait for compiled code representation
CubeElement: The base element trait for the jit backend.
Kernel: Implement this trait to create a kernel definition.
Runtime: Runtime for the CubeCL.

bool_elem
calculate_cube_count_elemwise: Calculate the number of cubes required to execute an operation where one cube unit is assigned to one element.
calculate_num_elems_dyn_rank
tensor_line_size
tensor_line_size_parallel: Find the maximum line size usable for parallel vectorization along the given axis from the supported line sizes or return 1 if vectorization is impossible.
tensor_line_size_perpendicular: Find the maximum line size usable for perpendicular vectorization along the given axis from the supported line sizes or return 1 if vectorization is impossible.
tensor_vectorization_factor

cube: Mark a cube function, trait or implementation for expansion.
tune: Crates a tuning set with a specific signature. Should return a tuple of benchmark inputs.

AutotuneKey: Implements display and initialization for autotune keys.
CubeLaunch: Derive macro to define a cube type that is launched with a kernel
CubeType: Derive macro to define a cube type that is not launched