cuda.core.experimental API Reference

All of the APIs listed (or cross-referenced from) below are considered experimental and subject to future changes without deprecation notice. Once stablized they will be moved out of the experimental namespace.

CUDA runtime

Device([device_id])	Represent a GPU and act as an entry point for cuda.core features.
launch(stream, config, kernel, *kernel_args)	Launches a Kernel object with launch-time configuration.
EventOptions([enable_timing, ...])	Customizable Event options.
StreamOptions([nonblocking, priority])	Customizable Stream options.
LaunchConfig([grid, cluster, block, shmem_size])	Customizable launch options.

CUDA compilation toolchain

Program(code, code_type[, options])	Represent a compilation machinery to process programs into ObjectCode.
Linker(*object_codes[, options])	Represent a linking machinery to link one or multiple object codes into ObjectCode with the specified options.
ObjectCode(*args, **kwargs)	Represent a compiled program to be loaded onto the device.
ProgramOptions([arch, ...])	Customizable options for configuring Program.
LinkerOptions([arch, max_register_count, ...])	Customizable Linker options.

CUDA system information

property System.driver_version: Tuple[int, int]

cuda.core.experimental.system.driver_version

Query the CUDA driver version.

Returns

tuple of int

A 2-tuple of (major, minor) version numbers.

property System.num_devices: int

cuda.core.experimental.system.num_devices

Query the number of available GPUs.

Returns

int

The number of available GPU devices.

property System.devices: tuple

cuda.core.experimental.system.devices

Query the available device instances.

Returns

tuple of Device

A tuple containing instances of available devices.

Utility functions

args_viewable_as_strided_memory(...)	Decorator to create proxy objects to StridedMemoryView for the specified positional arguments.
StridedMemoryView([obj, stream_ptr])	A dataclass holding metadata of a strided dense array/tensor.