NVPL Sparse Management API#

The NVPL Sparse functions for managing the library are described in this section.

nvpl_sparse_create()#

nvpl_sparse_status_t
nvpl_sparse_create(nvpl_sparse_handle_t *handle)

This function initializes the NVPL Sparse library and creates a handle on the NVPL Sparse context. It must be called before any other NVPL Sparse API function is invoked.

Param.	In/out	Meaning
`handle`	IN	The pointer to the handle to the NVPL Sparse context

See nvpl_sparse_status_t for the description of the return status

nvpl_sparse_destroy()#

nvpl_sparse_status_t
nvpl_sparse_destroy(nvpl_sparse_handle_t handle)

This function releases CPU resources used by the NVPL Sparse library.

Param.	In/out	Meaning
`handle`	IN	The handle to the NVPL Sparse context

See nvpl_sparse_status_t for the description of the return status

nvpl_sparse_get_version()#

nvpl_sparse_status_t
nvpl_sparse_get_version(nvpl_sparse_handle_t handle,
                        int*                 version)

This function returns the version number of the NVPL Sparse library.

Param.	In/out	Meaning
`handle`	IN	NVPL Sparse handle
`version`	OUT	The version number of the library

See nvpl_sparse_status_t for the description of the return status

nvpl_sparse_get_max_threads()#

int
nvpl_sparse_get_max_threads(void)

Returns the maximum number of threads that NVPL Sparse may use.

nvpl_sparse_get_num_threads()#

int
nvpl_sparse_get_num_threads(void)

Returns the number of threads in the current OpenMP team when called inside a parallel region.

Notes

When called inside an OpenMP parallel region, this reflects the actual team size (for example, the result of omp_get_num_threads()).
When called outside a parallel region, this function returns 1.

nvpl_sparse_set_num_threads()#

void
nvpl_sparse_set_num_threads(const int numthreads)

Sets the desired number of threads to be used by subsequent OpenMP parallel regions in NVPL Sparse.

Param.	In/out	Meaning
`numthreads`	IN	Desired number of threads for future regions

Notes

This forwards to the underlying runtime (for example, omp_set_num_threads(numthreads)).

nvpl_sparse_get_thread_num()#

int
nvpl_sparse_get_thread_num(void)

Returns the zero-based thread index of the calling thread within the current OpenMP team when called inside a parallel region.

Notes

When called inside an OpenMP parallel region, this reflects the calling thread’s index (for example, the result of omp_get_thread_num()).
When called outside a parallel region, this function returns 0.