Introduction#

Library overview#

hipSOLVER is an open-source marshalling library for LAPACK routines on the GPU. It sits between a backend library and the user application, marshalling inputs to and outputs from the backend library so that the user application remains unchanged when using different backends. Currently, two backend libraries are supported by hipSOLVER: NVIDIA’s cuSOLVER library and AMD’s open-source rocSOLVER library.

The regular hipSOLVER API is a thin wrapper layer around the different backends. As such, it is not expected to introduce significant overhead. However, its main purpose is portability, so when performance is critical, directly using the library backend corresponding to the given architecture is recommended.

Another purpose of hipSOLVER is to facilitate the translation of cuSOLVER applications to AMD’s open source ROCm platform ecosystem. To that end, hipSOLVER also includes hipsolverDn and hipsolverRf compatibility APIs with method signatures that match exactly those of cuSOLVER. For more details see the section Using hipSOLVER.

Currently implemented functionality#

The hipSOLVER library remains in active development. New features are being continuously added, with new functionality documented at each release of the ROCm platform.

The following tables summarize the wrapper functions that are implemented in the regular API for the different supported precisions in hipSOLVER’s latest release. Most of these functions have a corresponding version in the compatibility API, where applicable.

LAPACK auxiliary functions#

LAPACK main functions#

Table 3 Triangular factorizations#

Function

single

double

single complex

double complex

hipsolverXpotrf_bufferSize

x

x

x

x

hipsolverXpotrf

x

x

x

x

hipsolverXpotrfBatched_bufferSize

x

x

x

x

hipsolverXpotrfBatched

x

x

x

x

hipsolverXgetrf_bufferSize

x

x

x

x

hipsolverXgetrf

x

x

x

x

hipsolverXsytrf_bufferSize

x

x

x

x

hipsolverXsytrf

x

x

x

x

Table 4 Orthogonal factorizations#

Function

single

double

single complex

double complex

hipsolverXgeqrf_bufferSize

x

x

x

x

hipsolverXgeqrf

x

x

x

x

Table 5 Problem and matrix reductions#

Function

single

double

single complex

double complex

hipsolverXsytrd_bufferSize

x

x

hipsolverXsytrd

x

x

hipsolverXhetrd_bufferSize

x

x

hipsolverXhetrd

x

x

hipsolverXgebrd_bufferSize

x

x

x

x

hipsolverXgebrd

x

x

x

x

Table 6 Linear-systems solvers#

Function

single

double

single complex

double complex

hipsolverXpotri_bufferSize

x

x

x

x

hipsolverXpotri

x

x

x

x

hipsolverXpotrs_bufferSize

x

x

x

x

hipsolverXpotrs

x

x

x

x

hipsolverXpotrsBatched_bufferSize

x

x

x

x

hipsolverXpotrsBatched

x

x

x

x

hipsolverXgetrs_bufferSize

x

x

x

x

hipsolverXgetrs

x

x

x

x

hipsolverXXgesv_bufferSize

x

x

x

x

hipsolverXXgesv

x

x

x

x

Table 7 Least-square solvers#

Function

single

double

single complex

double complex

hipsolverXXgels_bufferSize

x

x

x

x

hipsolverXXgels

x

x

x

x

Table 8 Symmetric eigensolvers#

Function

single

double

single complex

double complex

hipsolverXsyevd_bufferSize

x

x

hipsolverXsyevd

x

x

hipsolverXsygvd_bufferSize

x

x

hipsolverXsygvd

x

x

hipsolverXheevd_bufferSize

x

x

hipsolverXheevd

x

x

hipsolverXhegvd_bufferSize

x

x

hipsolverXhegvd

x

x

Table 9 Singular value decomposition#

Function

single

double

single complex

double complex

hipsolverXgesvd_bufferSize

x

x

x

x

hipsolverXgesvd

x

x

x

x

LAPACK-like functions#

Table 11 Singular value decomposition#

Function

single

double

single complex

double complex

hipsolverDnXgesvdj_bufferSize

x

x

x

x

hipsolverDnXgesvdj

x

x

x

x

hipsolverDnXgesvdjBatched_bufferSize

x

x

x

x

hipsolverDnXgesvdjBatched

x

x

x

x

Compatibility-only functions#

The following tables summarize the wrapper functions that are provided only in the compatibility APIs. These wrappers are supported in rocSOLVER but either by equivalent functions that use different algorithmic approaches, or by functionality that is not fully exposed in the public API. For these reasons, at present, the corresponding wrappers are not provided in the regular hipSOLVER API.

Partial eigensolver functions#

Partial eigensolvers have been implemented in rocSOLVER, but at present they do not use a divide-and-conquer approach.

Table 12 Symmetric eigensolvers#

Function

single

double

single complex

double complex

hipsolverDnXsyevdx_bufferSize

x

x

hipsolverDnXsyevdx

x

x

hipsolverDnXsygvdx_bufferSize

x

x

hipsolverDnXsygvdx

x

x

hipsolverDnXheevdx_bufferSize

x

x

hipsolverDnXheevdx

x

x

hipsolverDnXhegvdx_bufferSize

x

x

hipsolverDnXhegvdx

x

x

Partial SVD functions#

Partial SVD has been implemented in rocSOLVER, but at present it does not use an approximate algorithm, nor does it compute the residual norm.

Table 13 Singular value decomposition#

Function

single

double

single complex

double complex

hipsolverDnXgesvdaStridedBatched_bufferSize

x

x

x

x

hipsolverDnXgesvdaStridedBatched

x

x

x

x

Refactorization routines#

Refactorization routines and direct solvers for sparse matrices are in the very earliest stages of development. At present, some of the functionality is not exposed as part of rocSOLVER’s public API and is subject to change. Refer to the hipsolverRf compatibility API for a full listing of supported functions.