Sparse Linear Algebra#
See the latest specification for the sparse domain here.
This page documents implementation specific or backend specific details of the sparse domain.
OneMKL Intel CPU and GPU backends#
Currently known limitations:
All operations’ algorithms except
no_optimize_alg
map to the default algorithm.The required external workspace size is always 0 bytes.
oneapi::mkl::sparse::set_csr_data
andoneapi::mkl::sparse::set_coo_data
functions cannot be used on a handle that has already been used for an operation or its optimize function. Doing so will throw aoneapi::mkl::unimplemented
exception.Using
spsv
with theoneapi::mkl::sparse::spsv_alg::no_optimize_alg
and a sparse matrix that does not have theoneapi::mkl::sparse::matrix_property::sorted
property will throw aoneapi::mkl::unimplemented
exception.Using
spmm
on Intel GPU with a sparse matrix that isoneapi::mkl::transpose::conjtrans
and has theoneapi::mkl::sparse::matrix_property::symmetric
property will throw aoneapi::mkl::unimplemented
exception.Using
spmv
with a sparse matrix that isoneapi::mkl::transpose::conjtrans
with atype_view
matrix_descr::symmetric
ormatrix_descr::hermitian
will throw aoneapi::mkl::unimplemented
exception.Using
spsv
on Intel GPU with a sparse matrix that isoneapi::mkl::transpose::conjtrans
and will throw aoneapi::mkl::unimplemented
exception.Scalar parameters
alpha
andbeta
should be host pointers to prevent synchronizations and copies to the host.
cuSPARSE backend#
Currently known limitations:
The COO format requires the indices to be sorted by row. See the cuSPARSE documentation. Sparse operations using matrices with the COO format without the property
matrix_property::sorted_by_rows
ormatrix_property::sorted
will throw aoneapi::mkl::unimplemented
exception.Using
spmm
with the algorithmspmm_alg::csr_alg3
and anopA
other thantranspose::nontrans
or anopB
transpose::conjtrans
will throw aoneapi::mkl::unimplemented
exception.Using
spmm
with the algorithmspmm_alg::csr_alg3
,opB=transpose::trans
and real fp64 precision will throw aoneapi::mkl::unimplemented
exception. This configuration can fail as of CUDA 12.6.2, see the related issue `here<https://forums.developer.nvidia.com/t/cusparse-spmm-sample-failing-with-misaligned-address/311022>`_.Using
spmv
with atype_view
other thanmatrix_descr::general
will throw aoneapi::mkl::unimplemented
exception.Using
spsv
with the algorithmspsv_alg::no_optimize_alg
may still perform some mandatory preprocessing.oneMKL Interface does not provide a way to use non-default algorithms without calling preprocess functions such as
cusparseSpMM_preprocess
orcusparseSpMV_preprocess
. Feel free to create an issue if this is needed.
Operation algorithms mapping#
The following tables describe how a oneMKL SYCL Interface algorithm maps to the backend’s algorithms. Refer to the backend’s documentation for a more detailed explanation of the algorithms.
Backends with no equivalent algorithms will fallback to the backend’s default behavior.
spmm#
|
MKLCPU/MKLGPU |
cuSPARSE |
---|---|---|
|
none |
|
|
none |
|
|
none |
|
|
none |
|
|
none |
|
|
none |
|
|
none |
|
|
none |
|
|
none |
|
spmv#
|
MKLCPU/MKLGPU |
cuSPARSE |
---|---|---|
|
none |
|
|
none |
|
|
none |
|
|
none |
|
|
none |
|
|
none |
|
|
none |
|
spsv#
|
MKLCPU/MKLGPU |
cuSPARSE |
---|---|---|
|
none |
|
|
none |
|