Scroll to navigation

rsblib/rsb.hpp(3) Library Functions Manual rsblib/rsb.hpp(3)

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.

/*
Copyright (C) 2020-2022 Michele Martone
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 a sblib does not differ conceptually much
Errors caught by ced and trigger an exception instead.
Memory management of matrices and the library state itself follow the usual C++ RAII rules:
the \c mtx object is freed first via RsbMatrix's destructor;
then s 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.

SEE ALSO

rsb-examples rsb-spblas.h rsb.h rsb.hpp

Version 1.3.0.2 librsb