How to solve the linear system Ax = b in mex file when A is a sparse matrix

2 Ansichten (letzte 30 Tage)
Rylan
Rylan am 8 Jan. 2022
Kommentiert: Rylan am 14 Jan. 2022
I have multiple linear systems Ax = b to be solved in mex function. When A is of small or medium size, A is stored as a full matrix. For this case, this system is solved with the following function.
void EquationSolve(double *dest, ptrdiff_t m, ptrdiff_t n, double *CoeMatrix, double dt, double ImplicitParam, int Nvar, int K3d, int Np)
{
double *TempCoeMatrix = malloc(m*m*sizeof(double));
ptrdiff_t *iPivot = malloc(m*sizeof(ptrdiff_t));
ptrdiff_t info;
ptrdiff_t p = m;
for (int var = 0; var < Nvar; var++){
memcpy(TempCoeMatrix, CoeMatrix + var*m*m, m*m*sizeof(double));
for (int col = 0; col < m; col++){
for (int row = 0; row < m; row++){
TempCoeMatrix[col*m + row] = -1 * dt*ImplicitParam*TempCoeMatrix[col*m + row];
}
TempCoeMatrix[col*m + col] += 1;
}
% This is the core of steps used to solve the linear system.
dgesv(&m, &n, TempCoeMatrix, &m, iPivot, dest + var*K3d*Np, &p, &info);
}
free(iPivot);
free(TempCoeMatrix);
}
However, if A is of large size, and much elements contained in A are zeros. Storing A as a full matrix is not wise. I want to convert A into a sparse one( this is easy to implement, since I already know the structure of A). But how to solve this system in mex function for such case.
I know, in matlab, solving this system can be easily implemented with x = A\b, no matter A is full or sparse.
  2 Kommentare
James Tursa
James Tursa am 14 Jan. 2022
You would need to use a sparse library. I thought Tim Davis had such a library on the FEX, but I can't find it there anymore.
Rylan
Rylan am 14 Jan. 2022
Well, thanks for you warmly reply @James Tursa. Actually, I have sort this out by calling MKL. And I've already confirmed that it can give the right answer. However, I run into another question. The thread control in mex function does not take effect. Since I need to use OPENMP to accelarete the computation in the mex function, the MKL should run in serial. I use the following thread control sentence in my program:
mkl_set_num_threads(1);
However, the thread number for each part of the mkl domain, i.e. MKL_DOMAIN_PARDISO, MKL_DOMAIN_BLAS, remain unchanged.
Do you have any idea about this? thanks.

Melden Sie sich an, um zu kommentieren.

Antworten (0)

Kategorien

Mehr zu Operating on Diagonal Matrices finden Sie in Help Center und File Exchange

Produkte


Version

R2020b

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by