void blocked_cholesky( int NB, float A[NB][NB] ) {
int i, j, k;
for (k=0; k<NB; k++) {
#pragma omp task depend(inout:A[k][k])
spotrf (A[k][k]) ;
for (i=k+1; i<NT; i++)
#pragma omp task depend(in:A[k][k]) depend(inout:A[k][i])
strsm (A[k][k], A[k][i]);
// update trailing submatrix
for (i=k+1; i<NT; i++) {
for (j=k+1; j<i; j++)
#pragma omp task depend(in:A[k][i],A[k][j]) depend(inout:A[j][i])
sgemm( A[k][i], A[k][j], A[j][i]);
#pragma omp task depend(in:A[k][i]) depend(inout:A[i][i])
ssyrk (A[k][i], A[i][i]);
}
}
}
void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[])
{
mwIndex i, j; /* indeces for populating lower triangle when finished. */
mwSize n; /* Size of the matrix. */
mwSignedIndex info; /* flag for success in dpotrf, spotrf, dpotri, spotri */
char *uplo = "U"; /* upper or lower triangle */
mxClassID type; /* array type */
/* If we don't pass in the correct number of arguments, throw error. */
if (nrhs!=1) {
mexErrMsgIdAndTxt("mexFunction:invChol_mex:numInputs",
"1 input required: A (square matrix)");
}
type = mxGetClassID(prhs[0]); /* get the array type */
n = mxGetM(prhs[0]); /* input matrix dimension */
/* check for symmetric matrix*/
if (n!=mxGetN(prhs[0])) {
mexErrMsgIdAndTxt("MATLAB:invChol_mex:matchdims",
"matrix is not symmetric");
}
/* create output matrix (fortran modifies the input so we need to copy the input to output) */
plhs[0]=mxDuplicateArray(prhs[0]);
/* If we passed in an empty return an empty. */
if (n==0) {
return;
}
/* double precision */
if(type==mxDOUBLE_CLASS)
{
double *B; /* double pointer to input & output matrices*/
B = mxGetPr(plhs[0]); /* output matrix pointer */
dpotrf( uplo, &n, B, &n, &info ); /* Double Cholesky decomposition */
/* check for success */
if (info<0) {
mexErrMsgIdAndTxt("MATLAB:invChol_mex:dpotrf:illegalvalue",
"cholesky decomposition failed: illegal value ");
}
if (info>0) {
mexErrMsgIdAndTxt("MATLAB:invChol_mex:dpotrf:notposdef",
"cholesky decomposition failed: matrix is not positive definite");
}
dpotri( uplo, &n, B, &n, &info ); /* Double Inverse using Cholesky decomposition */
/* check for success */
if (info<0) {
mexErrMsgIdAndTxt("MATLAB:invChol_mex:dpotri:illegalvalue",
"failed to invert: illegal value");
}
if (info>0) {
mexErrMsgIdAndTxt("MATLAB:invChol_mex:dpotri:singularmatrix",
"failed to invert: a diagonal element was 0");
}
/* populate the lower triangle */
for (i=0; i<n; i++) {
for (j=i+1; j<n; j++) {
B[n*i+j]=B[j*n+i];
}
}
} else if (type==mxSINGLE_CLASS) {
float *B; /* float pointer to input and output matrices */
B = (float*) mxGetData(plhs[0]); /* output matrix pointer */
spotrf( uplo, &n, B, &n, &info ); /* Double Cholesky decomposition */
/* check for success */
if (info<0) {
mexErrMsgIdAndTxt("MATLAB:invChol_mex:spotrf:illegalvalue",
"cholesky decomposition failed: illegal value ");
}
if (info>0) {
mexErrMsgIdAndTxt("MATLAB:invChol_mex:spotrf:notposdef",
"cholesky decomposition failed: matrix is not positive definite");
}
spotri( uplo, &n, B, &n, &info ); /* Double Inverse using Cholesky decomposition */
/* check for success */
if (info<0) {
mexErrMsgIdAndTxt("MATLAB:invChol_mex:spotri:illegalvalue",
"failed to invert: illegal value");
}
if (info>0) {
mexErrMsgIdAndTxt("MATLAB:invChol_mex:spotri:singularmatrix",
"failed to invert: a diagonal element was 0");
}
/* populate the lower triangle*/
for (i=0; i<n; ++i)
{
for (j=i+1; j<n; ++j)
{
B[n*i+j]=B[j*n+i];
}
}
} else {
mexErrMsgIdAndTxt("MATLAB:invChol_mex:illegaltype",
"only single or double matrix inputs are allowed");
}
return;
}
