create_plan#

nvmath.bindings.cutensor.create_plan( intptr_t handle, intptr_t desc, intptr_t pref, uint64_t workspace_size_limit, ) → intptr_t[source]#

This function allocates a cutensorPlan_t object, selects an appropriate kernel for a given operation (encoded by desc) and prepares a plan that encodes the execution.

Parameters:

handle (intptr_t) – Opaque handle holding cuTENSOR’s library context.
desc (intptr_t) – This opaque struct encodes the given operation (see cutensorCreateContraction, cutensorCreateReduction, cutensorCreatePermutation, cutensorCreateElementwiseBinary, cutensorCreateElementwiseTrinary, or cutensorCreateContractionTrinary).
pref (intptr_t) – This opaque struct is used to restrict the space of applicable candidates/kernels (see cutensorCreatePlanPreference or cutensorPlanPreferenceAttribute_t). May be nullptr, in that case default choices are assumed. Block-sparse contractions currently only support these default settings and ignore other supplied preferences.
workspace_size_limit (uint64_t) – Denotes the maximal workspace that the corresponding operation is allowed to use (see cutensorEstimateWorkspaceSize).

Returns:

Pointer to the data structure created by this function that holds all information (e.g., selected kernel) necessary to perform the desired operation.

Return type:

intptr_t