cuda.core._module.KernelAttributes

class cuda.core._module.KernelAttributes(*args, **kwargs)

Read-only view of a kernel’s per-device attributes.

The default view returned by Kernel.attributes is bound to the current device, resolved at attribute-access time. Use kernel.attributes[device] to obtain a view bound to a specific device (an int device ordinal or Device). Per-device views share the underlying cache so a value queried through one view is visible through the others.

Methods

__init__(*args, **kwargs)

Attributes

property binary_version: int

The binary architecture version for which the function was compiled. This attribute is read-only.

Type:

int

property cache_mode_ca: bool

Whether the function has been compiled with user specified option “-Xptxas –dlcm=ca” set. This attribute is read-only.

Type:

bool

property cluster_scheduling_policy_preference: int

The block scheduling policy of a function.

Type:

int

property cluster_size_must_be_set: bool

The kernel must launch with a valid cluster size specified. This attribute is read-only.

Type:

bool

property const_size_bytes: int

The size in bytes of user-allocated constant memory required by this function. This attribute is read-only.

Type:

int

property local_size_bytes: int

The size in bytes of local memory used by each thread of this function. This attribute is read-only.

Type:

int

property max_dynamic_shared_size_bytes: int

The maximum size in bytes of dynamically-allocated shared memory that can be used by this function.

Type:

int

property max_threads_per_block: int

The maximum number of threads per block. This attribute is read-only.

Type:

int

property non_portable_cluster_size_allowed: bool

Whether the function can be launched with non-portable cluster size.

Type:

bool

property num_regs: int

The number of registers used by each thread of this function. This attribute is read-only.

Type:

int

property preferred_shared_memory_carveout: int

The shared memory carveout preference, in percent of the total shared memory.

Type:

int

property ptx_version: int

The PTX virtual architecture version for which the function was compiled. This attribute is read-only.

Type:

int

property required_cluster_depth: int

The required cluster depth in blocks.

Type:

int

property required_cluster_height: int

The required cluster height in blocks.

Type:

int

property required_cluster_width: int

The required cluster width in blocks.

Type:

int

property shared_size_bytes: int

The size in bytes of statically-allocated shared memory required by this function. This attribute is read-only.

Type:

int