Developer Reference

Contents

LAPACK Functionality

All of the DPC++ LAPACK computational routines have a corresponding
*_scratchpad_size
function for calculating the required amount of scratchpad space.
LU Factorization Routines
Functionality
CPU
OpenMP Offload Intel GPU
getrf
Yes
Yes
getrs
Yes
Yes
getri
Yes
Yes
Cholesky Factorization Routines
Functionality
CPU
OpenMP Offload Intel GPU
potrf
Yes
Yes
potrs
Yes
Yes
potri
Yes
Yes
Orthogonal Factorization Routines
Functionality
CPU
OpenMP Offload Intel GPU
geqrf
Yes
Yes
{or,un}gqr
Yes
Yes
{or,un}mqr
Yes
Yes
gerqf
Yes
No
{or,un}mrq
Yes
No
Other Linear Equation Routines
Functionality
CPU
OpenMP Offload Intel GPU
trtrs
Yes
Yes
{sy,he}trf
Yes
No
Symmetric Eigenvalue Routines
Functionality
CPU
OpenMP Offload Intel GPU
{sy,he}ev
Yes
Yes
{sy,he}evd
Yes
Yes
{sy,he}evx
Yes
Yes
{sy,he}trd
Yes
Yes
{or,un}gtr
Yes
No
{or,un}mtr
Yes
No
steqr
Yes
Yes
Generalized Symmetric Eigenvalue Routines
Functionality
CPU
OpenMP Offload Intel GPU
{sy,he}gvd
Yes
Yes
{sy,he}gvx
Yes
Yes
Singular Value Routines
Functionality
CPU
OpenMP Offload Intel GPU
gesvd
Yes
Yes
gebrd
Yes
Yes
{or,un}gbr
Yes
No
Batched LAPACK Routines
Functionality
CPU
OpenMP Offload Intel GPU (ILP64 Interface)
getrf_batch
Strided
Strided
getrfnp_batch
Strided
Strided
getrs_batch
Strided
Strided
getrsnp_batch
Strided
Strided
getri_batch
No
No
potrf_batch
No
No
potrs_batch
No
No
geqrf_batch
No
No
{or,un}gqr_batch
No
No
Other LAPACK Routines
CPU
OpenMP Offload Intel GPU
Yes
No

Product and Performance Information

1

Performance varies by use, configuration and other factors. Learn more at www.Intel.com/PerformanceIndex.