rsb.hpp(3)

Classes \fCRsbLib and \fCRsbMatrix provide native C++ access to \fClibrsb.

Section 3 librsb-dev bookworm source

Description

rsblib/rsb.hpp

NAME

librsb - rsblib/rsb.hpp - Classes RsbLib and RsbMatrix provide native C++ access to librsb.

DESCRIPTION

SYNOPSIS

Data Structures

class RsbLib
Class initializing/finalizing librsb state.
class RsbMatrix< NT >
Represent a sparse matrix in RSB format by means of librsb.

Macros

#define RSBP_DEPRECATED
Internal attribute specifier for deprecated member functions.
#define RSBP_NODISCARD
Internal attribute.
#define RSBP_WANT_REV 0
If this is defined to 1 before including <rsb.hpp>, rsb_err_t is the default return type. Otherwise the default is void.
#define RSBP_RVT template <typename Err_t=void>
No return type.
#define RSBP_MSLVRV 10201
Minimal supported librsb version (value of RSB_LIBRSB_VER, defined via rsb.h)

Detailed Description

Classes RsbLib and RsbMatrix provide native C++ access to librsb.

Most of the librsb functionality is available via C++ classes RsbLib and RsbMatrix.
These classes are defined in header file <rsb.hpp>, which wraps functionality of librsb’s C interface <rsb.h>.
The RsbMatrix class can manipulate sparse matrices of several numerical types (same ones as librsb: matrix_supported_numerical_types_section).
Before using RsbMatrix, the library must be initialized by having a RsbLib object.
To avoid problems when including this header, don’t define preprocessor macros prefixed with RSB_ or RSBP_.

For a quick start, check out examples/example.cpp or other examples in its directory.

This file is part of librsb.

librsb is free software; you can redistribute it and/or modify it
under the terms of the GNU Lesser General Public License as published
by the Free Software Foundation; either version 3 of the License, or
(at your option) any later version.

librsb is distributed in the hope that it will be useful, but WITHOUT
ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public
License for more details.

You should have received a copy of the GNU Lesser General Public
License along with librsb; see the file COPYING.
If not, see <http://www.gnu.org/licenses/>.

*/
/*!
ingroup rsb::examples

@author Michele Martone

@brief C++ example based on <rsb.hpp> using RsbMatrix.spmm().

Using a librsb program via rsblib does not differ conceptually much librsb.

Errors caught by librsb shall not go unnoticed and trigger an exception instead.

Memory management of matrices and the library state itself follow the usual C++ RAII rules:
the then librsb is finalized via RsbLib()’s destructor .

include example.cpp
*/
#include <rsb.hpp>
using namespace rsb;
#ifdef RSB_NUMERICAL_TYPE_DOUBLE

#if RSBP_WANT_CPP20
#include <vector>
#include <array>

/* If your compiler is C++20 compatible, the std::span-based interface is available like: */
auto main() -> int {
RsbLib rsblib;
const rsb_nnz_idx_t nnzA { 7 };
const rsb_coo_idx_t nrA { 6 }, ncA { 6 }, nrhs { 2 };
const rsb_coo_idx_t IA [nnzA] {0,1,2,3,4,5,1};
std::array<rsb_coo_idx_t,7> JA {0,1,2,3,4,5,0};
std::vector<double> VA {1,1,1,1,1,1,2}, X(nrhs*ncA,1);
std::vector<double> Y(nrhs*nrA,0);
const double alpha {2}, beta {1};
rsb_int_t tn {0};
rsb_real_t sf {0}; // speedup factor (tune_spmm output)
const rsb_flags_t order {RSB_FLAG_WANT_COLUMN_MAJOR_ORDER};

// IA,JA,VA are respectively from a C array, std::vector, std::array;
// here using the C++20’s std::span interface:
RsbMatrix<double> mtx(IA,JA,VA);

mtx.file_save(); // rsb_file_mtx_save
mtx.tune_spmm(sf,RSB_TRANSPOSITION_N,alpha,nrhs,order,X,beta,Y); // rsb_tune_spmm
mtx.spmm(RSB_TRANSPOSITION_N,alpha,nrhs,order,X,beta,Y); // rsb_spmv
}
#else
#include <vector>
/* The pointer-based interface is available as well: */
auto main() -> int {
RsbLib rsblib;
const rsb_nnz_idx_t nnzA { 7 };
const rsb_coo_idx_t nrA { 6 }, ncA { 6 }, nrhs { 1 };
const std::vector<rsb_coo_idx_t> IA {0,1,2,3,4,5,1}, JA {0,1,2,3,4,5,0};
const std::vector<double> VA {1,1,1,1,1,1,2}, X(ncA,1);
std::vector<double> Y(nrA,0);
const double alpha {2}, beta {1};
rsb_int_t tn {0};
rsb_real_t sf {0}; // speedup factor (tune_spmm output)

RsbMatrix<double> mtx(IA.data(),JA.data(),VA.data(),nnzA);

mtx.file_save(nullptr); // rsb_file_mtx_save
mtx.tune_spmm(&sf,&tn,0,0.0,RSB_TRANSPOSITION_N,alpha,nrhs,RSB_FLAG_WANT_COLUMN_MAJOR_ORDER,X.data(),ncA,beta,Y.data(),nrA); // rsb_tune_spmm
mtx.spmv(RSB_TRANSPOSITION_N, alpha, X.data(), beta, Y.data()); // rsb_spmv
}
#endif
#else
auto main() { }
#endif

Author

Michele Martone

Macro Definition Documentation

#define RSBP_DEPRECATED

Internal attribute specifier for deprecated member functions. Header-only (rsb.hpp) wrapper classes for librsb (http://librsb.sf.net).

#define RSBP_MSLVRV 10201

Minimal supported librsb version (value of RSB_LIBRSB_VER, defined via rsb.h)

#define RSBP_NODISCARD

Internal attribute.

#define RSBP_RVT template <typename Err_t=void>

No return type.

#define RSBP_WANT_REV 0

If this is defined to 1 before including <rsb.hpp>, rsb_err_t is the default return type. Otherwise the default is void.

Author

librsb was written by Michele Martone; this documentation has been generated by Doxygen.