This package provides Julia bindings to the low level routines provided by the Intel (R) MKL Pardiso direct sparse solver interface. This package can only be used if a licensed installation of the Intel (R) MKL Pardiso is available in the system. This package is available free of charge and in no way replaces or alters any functionality of the linked libraries.
The goal of this package is to provide very thin wrappers to the low level Pardiso FORTRAN routines. The signatures of the Julia wrappers use only built-in types and are almost one-to-one to the signatures of the wrapped FORTRAN subroutines. This allows advanced users to have fine control over the Pardiso solver. If you are looking for a more user-friendly way to interface with Pardiso, then use the Pardiso.jl package.
- Set the
MKLROOT
environment variable. See the MKL getting started manual for a thorough guide how to set this variable correctly, typically done by executing something likesource /opt/intel/bin/compilervars.sh intel64
. - Make sure that
gcc
is installed and that can be found via the$PATH
environment variable. - Install
PardisoMKL
as any other Julia package.
- Most of the functions exported by
PardisoMKL
have the names of the underlying FORTRAN routines appended with an exclamation sign (!). - The Julia wrappers return an integer error code if the underlying FORTRAN subroutine accept an integer error code in the last argument.
- The arguments in the Julia bindings are named as in the FORTRAN subroutines. See the Intel (R) MKL Pardiso manual for the particular details.
pardisoinit!(pt::Vector{Int},mtype::Integer,iparm::Vector{Int32})
Initializes Intel MKL PARDISO with default parameters depending on the matrix type.
pardiso!( pt::Vector{Int}, maxfct::Integer, mnum::Integer, mtype::Integer,
phase::Integer, n::Integer, a::Vector{T}, ia::Vector{Int32},
ja::Vector{Int32}, perm::Vector{Int32}, nrhs::Integer, iparm::Vector{Int32},
msglvl::Integer, b::Vector{T}, x::Vector{T})::Int where T
Calculates the solution of a set of sparse linear equations with single or multiple right-hand sides and returns an integer error code. The type T
has to match the chosen matrix type mtype
and the provided options iparm
. The valid type T
can be retrieved with the pardiso_data_type
function (see below).
pardiso_64!(pt::Vector{Int}, maxfct::Integer, mnum::Integer, mtype::Integer,
phase::Integer, n::Integer, a::Vector{T}, ia::Vector{Int64},
ja::Vector{Int64}, perm::Vector{Int64}, nrhs::Integer, iparm::Vector{Int64},
msglvl::Integer, b::Vector{T}, x::Vector{T})::Int where T
64-bit integer version of pardiso!
pardiso_getdiag!(pt::Vector{Int}, df::Vector{T}, da::Vector{T}, mnum::Integer)::Int where T
Writes the diagonal elements of initial and factorized matrix in the provided vectors and returns an integer error code. The type T
is as previously detailed for pardiso!
.
pardiso_getdiag!(pt::Vector{Int}, df::Vector{T}, da::Vector{T}, mnum::Integer,
mtype::Integer, iparm::Vector{<:Integer})::Int where T
Safer version of pardiso_getdiag!
that checks if the type T
matches the given mtype
and iparm
.
new_pardiso_handle()::Vector{Int}
Allocates and returns a vector of zeros that can be used as the pt
argument in previous functions.
new_iparm()::Vector{Int32}
Allocates and returns a vector of zeros that can be used as the iparm
argument in previous functions.
new_iparm_64()::Vector{Int64}
64-bit version of new_iparm
. To be used in pardiso_64!
pardiso_data_type(mtype::Integer,iparm::Vector{<:Integer})::DataType
Returns the data type associated with the matrix type mtype
and the options in iparm
.