SEXP dgeMatrix_LU_(SEXP x, Rboolean warn_sing)
{
SEXP val = get_factors(x, "LU");
int *dims, npiv, info;
if (val != R_NilValue) /* nothing to do if it's there in 'factors' slot */
return val;
dims = INTEGER(GET_SLOT(x, Matrix_DimSym));
if (dims[0] < 1 || dims[1] < 1)
error(_("Cannot factor a matrix with zero extents"));
npiv = (dims[0] <dims[1]) ? dims[0] : dims[1];
val = PROTECT(NEW_OBJECT(MAKE_CLASS("denseLU")));
slot_dup(val, x, Matrix_xSym);
slot_dup(val, x, Matrix_DimSym);
F77_CALL(dgetrf)(dims, dims + 1, REAL(GET_SLOT(val, Matrix_xSym)),
dims,
INTEGER(ALLOC_SLOT(val, Matrix_permSym, INTSXP, npiv)),
&info);
if (info < 0)
error(_("Lapack routine %s returned error code %d"), "dgetrf", info);
else if (info > 0 && warn_sing)
warning(_("Exact singularity detected during LU decomposition."));
UNPROTECT(1);
return set_factors(x, val, "LU");
}
SEXP dppMatrix_solve(SEXP x)
{
SEXP Chol = dppMatrix_chol(x);
SEXP val = PROTECT(NEW_OBJECT(MAKE_CLASS("dppMatrix")));
int *dims = INTEGER(GET_SLOT(x, Matrix_DimSym)), info;
slot_dup(val, Chol, Matrix_uploSym);
slot_dup(val, Chol, Matrix_xSym);
slot_dup(val, Chol, Matrix_DimSym);
F77_CALL(dpptri)(uplo_P(val), dims,
REAL(GET_SLOT(val, Matrix_xSym)), &info);
UNPROTECT(1);
return val;
}
SEXP dppMatrix_chol(SEXP x)
{
SEXP val = get_factors(x, "pCholesky"),
dimP = GET_SLOT(x, Matrix_DimSym),
uploP = GET_SLOT(x, Matrix_uploSym);
const char *uplo = CHAR(STRING_ELT(uploP, 0));
int *dims = INTEGER(dimP), info;
if (val != R_NilValue) return val;
dims = INTEGER(dimP);
val = PROTECT(NEW_OBJECT(MAKE_CLASS("pCholesky")));
SET_SLOT(val, Matrix_uploSym, duplicate(uploP));
SET_SLOT(val, Matrix_diagSym, mkString("N"));
SET_SLOT(val, Matrix_DimSym, duplicate(dimP));
slot_dup(val, x, Matrix_xSym);
F77_CALL(dpptrf)(uplo, dims, REAL(GET_SLOT(val, Matrix_xSym)), &info);
if (info) {
if(info > 0) /* e.g. x singular */
error(_("the leading minor of order %d is not positive definite"),
info);
else /* should never happen! */
error(_("Lapack routine %s returned error code %d"), "dpptrf", info);
}
UNPROTECT(1);
return set_factors(x, val, "pCholesky");
}
SEXP dtrMatrix_chol2inv(SEXP a)
{
SEXP val = PROTECT(NEW_OBJECT(MAKE_CLASS("dpoMatrix")));
int info, n;
slot_dup(val, a, Matrix_DimSym);
slot_dup(val, a, Matrix_uploSym);
slot_dup(val, a, Matrix_diagSym);
slot_dup(val, a, Matrix_DimNamesSym);
slot_dup(val, a, Matrix_xSym);
n = *INTEGER(GET_SLOT(val, Matrix_DimSym));
F77_CALL(dpotri)(uplo_P(val), &n,
REAL(GET_SLOT(val, Matrix_xSym)), &n, &info);
UNPROTECT(1);
return val;
}
SEXP dgeMatrix_solve(SEXP a)
{
/* compute the 1-norm of the matrix, which is needed
later for the computation of the reciprocal condition number. */
double aNorm = get_norm(a, "1");
/* the LU decomposition : */
SEXP val = PROTECT(NEW_OBJECT(MAKE_CLASS("dgeMatrix"))),
lu = dgeMatrix_LU_(a, TRUE);
int *dims = INTEGER(GET_SLOT(lu, Matrix_DimSym)),
*pivot = INTEGER(GET_SLOT(lu, Matrix_permSym));
/* prepare variables for the dgetri calls */
double *x, tmp;
int info, lwork = -1;
if (dims[0] != dims[1]) error(_("Solve requires a square matrix"));
slot_dup(val, lu, Matrix_xSym);
x = REAL(GET_SLOT(val, Matrix_xSym));
slot_dup(val, lu, Matrix_DimSym);
if(dims[0]) /* the dimension is not zero */
{
/* is the matrix is *computationally* singular ? */
double rcond;
F77_CALL(dgecon)("1", dims, x, dims, &aNorm, &rcond,
(double *) R_alloc(4*dims[0], sizeof(double)),
(int *) R_alloc(dims[0], sizeof(int)), &info);
if (info)
error(_("error [%d] from Lapack 'dgecon()'"), info);
if(rcond < DOUBLE_EPS)
error(_("Lapack dgecon(): system computationally singular, reciprocal condition number = %g"),
rcond);
/* only now try the inversion and check if the matrix is *exactly* singular: */
F77_CALL(dgetri)(dims, x, dims, pivot, &tmp, &lwork, &info);
lwork = (int) tmp;
F77_CALL(dgetri)(dims, x, dims, pivot,
(double *) R_alloc((size_t) lwork, sizeof(double)),
&lwork, &info);
if (info)
error(_("Lapack routine dgetri: system is exactly singular"));
}
UNPROTECT(1);
return val;
}
SEXP dsyMatrix_solve(SEXP a)
{
SEXP trf = dsyMatrix_trf(a);
SEXP val = PROTECT(NEW_OBJECT_OF_CLASS("dsyMatrix"));
int *dims = INTEGER(GET_SLOT(trf, Matrix_DimSym)), info;
slot_dup(val, trf, Matrix_uploSym);
slot_dup(val, trf, Matrix_xSym);
slot_dup(val, trf, Matrix_DimSym);
F77_CALL(dsytri)(uplo_P(val), dims,
REAL(GET_SLOT(val, Matrix_xSym)), dims,
INTEGER(GET_SLOT(trf, Matrix_permSym)),
(double *) R_alloc((long) dims[0], sizeof(double)),
&info);
UNPROTECT(1);
return val;
}
// n.CMatrix --> [dli].CMatrix (not going through CHM!)
SEXP nz2Csparse(SEXP x, enum x_slot_kind r_kind)
{
const char *cl_x = class_P(x);
if(cl_x[0] != 'n') error(_("not a 'n.CMatrix'"));
if(cl_x[2] != 'C') error(_("not a CsparseMatrix"));
int nnz = LENGTH(GET_SLOT(x, Matrix_iSym));
SEXP ans;
char *ncl = strdup(cl_x);
double *dx_x; int *ix_x;
ncl[0] = (r_kind == x_double ? 'd' :
(r_kind == x_logical ? 'l' :
/* else (for now): r_kind == x_integer : */ 'i'));
PROTECT(ans = NEW_OBJECT(MAKE_CLASS(ncl)));
// create a correct 'x' slot:
switch(r_kind) {
int i;
case x_double: // 'd'
dx_x = REAL(ALLOC_SLOT(ans, Matrix_xSym, REALSXP, nnz));
for (i=0; i < nnz; i++) dx_x[i] = 1.;
break;
case x_logical: // 'l'
ix_x = LOGICAL(ALLOC_SLOT(ans, Matrix_xSym, LGLSXP, nnz));
for (i=0; i < nnz; i++) ix_x[i] = TRUE;
break;
case x_integer: // 'i'
ix_x = INTEGER(ALLOC_SLOT(ans, Matrix_xSym, INTSXP, nnz));
for (i=0; i < nnz; i++) ix_x[i] = 1;
break;
default:
error(_("nz2Csparse(): invalid/non-implemented r_kind = %d"),
r_kind);
}
// now copy all other slots :
slot_dup(ans, x, Matrix_iSym);
slot_dup(ans, x, Matrix_pSym);
slot_dup(ans, x, Matrix_DimSym);
slot_dup(ans, x, Matrix_DimNamesSym);
if(ncl[1] != 'g') { // symmetric or triangular ...
slot_dup_if_has(ans, x, Matrix_uploSym);
slot_dup_if_has(ans, x, Matrix_diagSym);
}
UNPROTECT(1);
return ans;
}
SEXP R_to_CMatrix(SEXP x)
{
SEXP ans, tri = PROTECT(allocVector(LGLSXP, 1));
char *ncl = strdup(class_P(x));
static const char *valid[] = { MATRIX_VALID_Rsparse, ""};
int ctype = R_check_class_etc(x, valid);
int *x_dims = INTEGER(GET_SLOT(x, Matrix_DimSym)), *a_dims;
PROTECT_INDEX ipx;
if (ctype < 0)
error(_("invalid class(x) '%s' in R_to_CMatrix(x)"), ncl);
/* replace 'R' with 'C' : */
ncl[2] = 'C';
PROTECT_WITH_INDEX(ans = NEW_OBJECT(MAKE_CLASS(ncl)), &ipx);
a_dims = INTEGER(ALLOC_SLOT(ans, Matrix_DimSym, INTSXP, 2));
/* reversed dim() since we will transpose: */
a_dims[0] = x_dims[1];
a_dims[1] = x_dims[0];
/* triangular: */ LOGICAL(tri)[0] = 0;
if((ctype / 3) != 2) /* not n..Matrix */
slot_dup(ans, x, Matrix_xSym);
if(ctype % 3) { /* s(ymmetric) or t(riangular) : */
SET_SLOT(ans, Matrix_uploSym,
mkString((*uplo_P(x) == 'U') ? "L" : "U"));
if(ctype % 3 == 2) { /* t(riangular) : */
LOGICAL(tri)[0] = 1;
slot_dup(ans, x, Matrix_diagSym);
}
}
SET_SLOT(ans, Matrix_iSym, duplicate(GET_SLOT(x, Matrix_jSym)));
slot_dup(ans, x, Matrix_pSym);
REPROTECT(ans = Csparse_transpose(ans, tri), ipx);
SET_DimNames(ans, x);
// possibly asymmetric for symmetricMatrix is ok
free(ncl);
UNPROTECT(2);
return ans;
}
/* This and the following R_to_CMatrix() lead to memory-not-mapped seg.faults
* only with {32bit + R-devel + enable-R-shlib} -- no idea why */
SEXP compressed_to_TMatrix(SEXP x, SEXP colP)
{
int col = asLogical(colP); /* 1 if "C"olumn compressed; 0 if "R"ow */
/* however, for Csparse, we now effectively use the cholmod-based
* Csparse_to_Tsparse() in ./Csparse.c ; maybe should simply write
* an as_cholmod_Rsparse() function and then do "as there" ...*/
SEXP indSym = col ? Matrix_iSym : Matrix_jSym,
ans, indP = GET_SLOT(x, indSym),
pP = GET_SLOT(x, Matrix_pSym);
int npt = length(pP) - 1;
char *ncl = strdup(class_P(x));
static const char *valid[] = { MATRIX_VALID_Csparse, MATRIX_VALID_Rsparse, ""};
int ctype = R_check_class_etc(x, valid);
if (ctype < 0)
error(_("invalid class(x) '%s' in compressed_to_TMatrix(x)"), ncl);
/* replace 'C' or 'R' with 'T' :*/
ncl[2] = 'T';
ans = PROTECT(NEW_OBJECT(MAKE_CLASS(ncl)));
slot_dup(ans, x, Matrix_DimSym);
if((ctype / 3) % 4 != 2) /* not n..Matrix */
slot_dup(ans, x, Matrix_xSym);
if(ctype % 3) { /* s(ymmetric) or t(riangular) : */
slot_dup(ans, x, Matrix_uploSym);
if(ctype % 3 == 2) /* t(riangular) : */
slot_dup(ans, x, Matrix_diagSym);
}
SET_DimNames(ans, x);
// possibly asymmetric for symmetricMatrix is ok
SET_SLOT(ans, indSym, duplicate(indP));
expand_cmprPt(npt, INTEGER(pP),
INTEGER(ALLOC_SLOT(ans, col ? Matrix_jSym : Matrix_iSym,
INTSXP, length(indP))));
free(ncl);
UNPROTECT(1);
return ans;
}
SEXP dgeMatrix_solve(SEXP a)
{
SEXP val = PROTECT(NEW_OBJECT(MAKE_CLASS("dgeMatrix"))),
lu = dgeMatrix_LU_(a, TRUE);
int *dims = INTEGER(GET_SLOT(lu, Matrix_DimSym)),
*pivot = INTEGER(GET_SLOT(lu, Matrix_permSym));
double *x, tmp;
int info, lwork = -1;
if (dims[0] != dims[1]) error(_("Solve requires a square matrix"));
slot_dup(val, lu, Matrix_xSym);
x = REAL(GET_SLOT(val, Matrix_xSym));
slot_dup(val, lu, Matrix_DimSym);
F77_CALL(dgetri)(dims, x, dims, pivot, &tmp, &lwork, &info);
lwork = (int) tmp;
F77_CALL(dgetri)(dims, x, dims, pivot,
(double *) R_alloc((size_t) lwork, sizeof(double)),
&lwork, &info);
if (info)
error(_("Lapack routine dgetri: system is exactly singular"));
UNPROTECT(1);
return val;
}
SEXP dtCMatrix_sparse_solve(SEXP a, SEXP b)
{
SEXP ans = PROTECT(NEW_OBJECT(MAKE_CLASS("dgCMatrix")));
CSP A = AS_CSP(a), B = AS_CSP(b);
int *xp = INTEGER(ALLOC_SLOT(ans, Matrix_pSym, INTSXP, (B->n) + 1)),
xnz = 10 * B->p[B->n]; /* initial estimate of nnz in x */
int *ti = Calloc(xnz, int), k, lo = uplo_P(a)[0] == 'L', pos = 0;
double *tx = Calloc(xnz, double);
double *wrk = Alloca(A->n, double);
int *xi = Alloca(2*A->n, int); /* for cs_reach */
R_CheckStack();
if (A->m != A->n || B->n < 1 || A->n < 1 || A->n != B->m)
error(_("Dimensions of system to be solved are inconsistent"));
slot_dup(ans, b, Matrix_DimSym);
SET_DimNames(ans, b);
xp[0] = 0;
for (k = 0; k < B->n; k++) {
int top = cs_spsolve (A, B, k, xi, wrk, (int *)NULL, lo);
int nz = A->n - top, p;
xp[k + 1] = nz + xp[k];
if (xp[k + 1] > xnz) {
while (xp[k + 1] > xnz) xnz *= 2;
ti = Realloc(ti, xnz, int);
tx = Realloc(tx, xnz, double);
}
if (lo) /* increasing row order */
for(p = top; p < A->n; p++, pos++) {
ti[pos] = xi[p];
tx[pos] = wrk[xi[p]];
}
else /* decreasing order, reverse copy */
for(p = A->n - 1; p >= top; p--, pos++) {
ti[pos] = xi[p];
tx[pos] = wrk[xi[p]];
}
}
xnz = xp[B->n];
Memcpy(INTEGER(ALLOC_SLOT(ans, Matrix_iSym, INTSXP, xnz)), ti, xnz);
Memcpy( REAL(ALLOC_SLOT(ans, Matrix_xSym, REALSXP, xnz)), tx, xnz);
Free(ti); Free(tx);
UNPROTECT(1);
return ans;
}
/**
* Return a SuiteSparse QR factorization of the sparse matrix A
*
* @param Ap (pointer to) a [m x n] dgCMatrix
* @param ordering integer SEXP specifying the ordering strategy to be used
* see SPQR/Include/SuiteSparseQR_definitions.h
* @param econ integer SEXP ("economy"): number of rows of R and columns of Q
* to return. The default is m. Using n gives the standard economy form.
* A value less than the estimated rank r is set to r, so econ=0 gives the
* "rank-sized" factorization, where nrow(R)==nnz(diag(R))==r.
* @param tol double SEXP: if tol <= -2 use SPQR's default,
* if -2 < tol < 0, then no tol is used; otherwise,
* tol > 0, use as tolerance: columns with 2-norm <= tol treated as 0
*
*
* @return SEXP "SPQR" object with slots (Q, R, p, rank, Dim):
* Q: dgCMatrix; R: dgCMatrix [subject to change to dtCMatrix FIXME ?]
* p: integer: 0-based permutation (or length 0 <=> identity);
* rank: integer, the "revealed" rank Dim: integer, original matrix dim.
*/
SEXP dgCMatrix_SPQR(SEXP Ap, SEXP ordering, SEXP econ, SEXP tol)
{
/* SEXP ans = PROTECT(allocVector(VECSXP, 4)); */
SEXP ans = PROTECT(NEW_OBJECT(MAKE_CLASS("SPQR")));
CHM_SP A = AS_CHM_SP(Ap), Q, R;
SuiteSparse_long *E, rank;/* not always = int FIXME (Windows_64 ?) */
if ((rank = SuiteSparseQR_C_QR(asInteger(ordering),
asReal(tol),/* originally had SPQR_DEFAULT_TOL */
(SuiteSparse_long)asInteger(econ),/* originally had 0 */
A, &Q, &R, &E, &cl)) == -1)
error(_("SuiteSparseQR_C_QR returned an error code"));
slot_dup(ans, Ap, Matrix_DimSym);
/* SET_VECTOR_ELT(ans, 0, */
/* chm_sparse_to_SEXP(Q, 0, 0, 0, "", R_NilValue)); */
SET_SLOT(ans, install("Q"),
chm_sparse_to_SEXP(Q, 0, 0, 0, "", R_NilValue));
/* Also gives a dgCMatrix (not a dtC* *triangular*) :
* may make sense if to be used in the "spqr_solve" routines .. ?? */
/* SET_VECTOR_ELT(ans, 1, */
/* chm_sparse_to_SEXP(R, 0, 0, 0, "", R_NilValue)); */
SET_SLOT(ans, install("R"),
chm_sparse_to_SEXP(R, 0, 0, 0, "", R_NilValue));
cholmod_free_sparse(&Al, &cl);
cholmod_free_sparse(&R, &cl);
cholmod_free_sparse(&Q, &cl);
if (E) {
int *Er;
SET_VECTOR_ELT(ans, 2, allocVector(INTSXP, A->ncol));
Er = INTEGER(VECTOR_ELT(ans, 2));
for (int i = 0; i < A->ncol; i++) Er[i] = (int) E[i];
Free(E);
} else SET_VECTOR_ELT(ans, 2, allocVector(INTSXP, 0));
SET_VECTOR_ELT(ans, 3, ScalarInteger((int)rank));
UNPROTECT(1);
return ans;
}
SEXP magma_dgeMatrix_LU_(SEXP x, Rboolean warn_sing)
{
#ifdef HIPLAR_WITH_MAGMA
SEXP val = get_factors(x, "LU");
int *dims, npiv, info;
if (val != R_NilValue) {
// R_ShowMessage("already in slot"); /* nothing to do if it's there in 'factors' slot */
return val;
}
dims = INTEGER(GET_SLOT(x, Matrix_DimSym));
if (dims[0] < 1 || dims[1] < 1)
error(_("Cannot factor a matrix with zero extents"));
npiv = (dims[0] < dims[1]) ? dims[0] : dims[1];
val = PROTECT(NEW_OBJECT(MAKE_CLASS("denseLU")));
slot_dup(val, x, Matrix_xSym);
slot_dup(val, x, Matrix_DimSym);
double *h_R = REAL(GET_SLOT(val, Matrix_xSym));
int *ipiv = INTEGER(ALLOC_SLOT(val, Matrix_permSym, INTSXP, npiv));
if(GPUFlag == 0){
#ifdef HIPLAR_DBG
R_ShowMessage("DBG: LU decomposition using dgetrf;");
#endif
F77_CALL(dgetrf)(dims, dims + 1, h_R,
dims,
ipiv,
&info);
}
else if(GPUFlag == 1 && Interface == 0){
#ifdef HIPLAR_DBG
R_ShowMessage("DBG: LU decomposition using magma_dgetrf;");
#endif
magma_dgetrf(dims[0], dims[1], h_R, dims[0], ipiv, &info);
}
else if(GPUFlag == 1 && Interface == 1) {
#ifdef HIPLAR_DBG
R_ShowMessage("DBG: LU decomposition using magma_dgetrf_gpu;");
#endif
double *d_A;
int N2 = dims[0] * dims[1];
cublasStatus retStatus;
cublasAlloc( N2 , sizeof(double), (void**)&d_A);
/* Error Checking */
retStatus = cublasGetError ();
if (retStatus != CUBLAS_STATUS_SUCCESS)
error(_("CUBLAS: Error in Memory Allocation"));
/********************************************/
cublasSetVector(N2, sizeof(double), h_R, 1, d_A, 1);
/* Error Checking */
retStatus = cublasGetError ();
if (retStatus != CUBLAS_STATUS_SUCCESS)
error(_("CUBLAS: Error in Date Transfer to Device"));
/********************************************/
magma_dgetrf_gpu(dims[0],dims[1], d_A, dims[0], ipiv, &info);
cublasGetVector( N2, sizeof(double), d_A, 1, h_R, 1);
/* Error Checking */
retStatus = cublasGetError ();
if (retStatus != CUBLAS_STATUS_SUCCESS)
error(_("CUBLAS: Error in Date Transfer from Device"));
/********************************************/
cublasFree(d_A);
/* Error Checking */
retStatus = cublasGetError ();
if (retStatus != CUBLAS_STATUS_SUCCESS)
error(_("CUBLAS: Error freeing data"));
/********************************************/
}
else
error(_("MAGMA/LAPACK/Interface Flag not defined correctly"));
if (info < 0)
error(_("Lapack routine %s returned error code %d"), "dgetrf", info);
else if (info > 0 && warn_sing)
warning(_("Exact singularity detected during LU decomposition: %s, i=%d."),
"U[i,i]=0", info);
UNPROTECT(1);
return set_factors(x, val, "LU");
#endif
return R_NilValue;
}
SEXP magma_dgeMatrix_solve(SEXP a)
{
#ifdef HIPLAR_WITH_MAGMA
/* compute the 1-norm of the matrix, which is needed
later for the computation of the reciprocal condition number. */
double aNorm = magma_get_norm(a, "1");
/* the LU decomposition : */
/* Given that we may be performing this operation
* on the GPU we may put in an optimisation here
* where if we call the LU solver we, we do not require
* the decomposition to be transferred back to CPU. This is TODO
*/
SEXP val = PROTECT(NEW_OBJECT(MAKE_CLASS("dgeMatrix"))),
lu = magma_dgeMatrix_LU_(a, TRUE);
int *dims = INTEGER(GET_SLOT(lu, Matrix_DimSym)),
*pivot = INTEGER(GET_SLOT(lu, Matrix_permSym));
/* prepare variables for the dgetri calls */
double *x, tmp;
int info, lwork = -1;
if (dims[0] != dims[1]) error(_("Solve requires a square matrix"));
slot_dup(val, lu, Matrix_xSym);
x = REAL(GET_SLOT(val, Matrix_xSym));
slot_dup(val, lu, Matrix_DimSym);
int N2 = dims[0] * dims[0];
if(dims[0]) /* the dimension is not zero */
{
/* is the matrix is *computationally* singular ? */
double rcond;
F77_CALL(dgecon)("1", dims, x, dims, &aNorm, &rcond,
(double *) R_alloc(4*dims[0], sizeof(double)),
(int *) R_alloc(dims[0], sizeof(int)), &info);
if (info)
error(_("error [%d] from Lapack 'dgecon()'"), info);
if(rcond < DOUBLE_EPS)
error(_("Lapack dgecon(): system computationally singular, reciprocal condition number = %g"),
rcond);
/* only now try the inversion and check if the matrix is *exactly* singular: */
// This is also a work space query. This is not an option in magma
F77_CALL(dgetri)(dims, x, dims, pivot, &tmp, &lwork, &info);
lwork = (int) tmp;
if( GPUFlag == 0){
F77_CALL(dgetri)(dims, x, dims, pivot,
(double *) R_alloc((size_t) lwork, sizeof(double)),
&lwork, &info);
#ifdef HIPLAR_DBG
R_ShowMessage("DBG: Solve using LU using dgetri;");
#endif
}
else if(GPUFlag == 1) {
double *d_x, *dwork;
cublasStatus retStatus;
#ifdef HIPLAR_DBG
R_ShowMessage("Solve using LU using magma_dgetri;");
#endif
cublasAlloc(N2, sizeof(double), (void**)&d_x);
//cublasAlloc(N2 , sizeof(double), (void**)&dtmp);
/* Error Checking */
retStatus = cublasGetError ();
if (retStatus != CUBLAS_STATUS_SUCCESS)
error(_("CUBLAS: Error in Memory Allocation on Device"));
/********************************************/
cublasSetVector( N2, sizeof(double), x, 1, d_x, 1);
/* Error Checking */
retStatus = cublasGetError ();
if (retStatus != CUBLAS_STATUS_SUCCESS)
error(_("CUBLAS: Error in Data Transfer to Device"));
/********************************************/
lwork = dims[0] * magma_get_dgetri_nb( dims[0] );
cublasAlloc(lwork, sizeof(double), (void**)&dwork);
/* Error Checking */
retStatus = cublasGetError ();
if (retStatus != CUBLAS_STATUS_SUCCESS)
error(_("CUBLAS: Error in Memory Allocation on Device"));
/********************************************/
magma_dgetri_gpu(dims[0], d_x, dims[0], pivot, dwork , lwork, &info);
cublasGetVector(N2, sizeof(double), d_x, 1, x, 1);
/* Error Checking */
retStatus = cublasGetError ();
if (retStatus != CUBLAS_STATUS_SUCCESS)
error(_("CUBLAS: Error in Data From to Device"));
/********************************************/
cublasFree(dwork);
cublasFree(d_x);
/* Error Checking */
retStatus = cublasGetError ();
if (retStatus != CUBLAS_STATUS_SUCCESS)
error(_("CUBLAS: Error freeing memory"));
/********************************************/
}
else
error(_("GPUFlag not set correctly"));
if (info)
error(_("Lapack routine dgetri: system is exactly singular"));
}
UNPROTECT(1);
return val;
#endif
return R_NilValue;
}
SEXP magma_dpoMatrix_dgeMatrix_solve(SEXP a, SEXP b)
{
#ifdef HIPLAR_WITH_MAGMA
SEXP Chol = magma_dpoMatrix_chol(a),
val = PROTECT(NEW_OBJECT(MAKE_CLASS("dgeMatrix")));
int *adims = INTEGER(GET_SLOT(a, Matrix_DimSym)),
*bdims = INTEGER(GET_SLOT(b, Matrix_DimSym)),
info;
/* Checking Matrix Dimensions */
if (adims[1] != bdims[0])
error(_("Dimensions of system to be solved are inconsistent"));
if (adims[0] < 1 || bdims[1] < 1)
error(_("Cannot solve() for matrices with zero extents"));
/* ****************************************** */
SET_SLOT(val, Matrix_factorSym, allocVector(VECSXP, 0));
slot_dup(val, b, Matrix_DimSym);
slot_dup(val, b, Matrix_xSym);
double *A = REAL(GET_SLOT(Chol, Matrix_xSym));
double *B = REAL(GET_SLOT(val, Matrix_xSym));
if(GPUFlag == 1) {
#ifdef HIPLAR_DBG
R_ShowMessage("DBG: Solving system of Ax = b, A = dpo, b = dge, using dpotrs_gpu;");
#endif
double *d_A, *d_B;
const char *uplo = uplo_P(Chol);
magma_int_t NRHS = bdims[1];
magma_int_t lda = adims[1];
magma_int_t ldb = bdims[0];
magma_int_t N = adims[0];
cublasStatus retStatus;
/*if(uplo == "U")
uplo = MagmaUpperStr;
else if(uplo == "L")
uplo = MagmaLowerStr;
else
uplo = MagmaUpperStr;
*/
cublasAlloc(N * lda, sizeof(double), (void**)&d_A);
/* Error Checking */
retStatus = cublasGetError ();
if (retStatus != CUBLAS_STATUS_SUCCESS)
error(_("CUBLAS: Error in Memory Allocation"));
/********************************************/
cublasAlloc(N * NRHS, sizeof(double), (void**)&d_B);
/* Error Checking */
retStatus = cublasGetError ();
if (retStatus != CUBLAS_STATUS_SUCCESS)
error(_("CUBLAS: Error in Memory Allocation"));
/********************************************/
cublasSetVector( N * lda , sizeof(double), A, 1, d_A, 1);
/* Error Checking */
retStatus = cublasGetError ();
if (retStatus != CUBLAS_STATUS_SUCCESS)
error(_("CUBLAS: Error in Data Transfer to Device"));
/********************************************/
cublasSetVector( ldb * NRHS, sizeof(double), B, 1, d_B, 1 );
/* Error Checking */
retStatus = cublasGetError ();
if (retStatus != CUBLAS_STATUS_SUCCESS)
error(_("CUBLAS: Error in Data Transfer to Device"));
/********************************************/
magma_dpotrs_gpu(uplo[0], N, NRHS , d_A, lda, d_B, ldb, &info);
cublasGetVector( ldb * NRHS, sizeof(double), d_B, 1, B, 1);
/* Error Checking */
retStatus = cublasGetError ();
if (retStatus != CUBLAS_STATUS_SUCCESS)
error(_("CUBLAS: Error in Data Transfer from Device"));
/********************************************/
cublasFree(d_A);
cublasFree(d_B);
}
else {
#ifdef HIPLAR_DBG
R_ShowMessage("DBG: Solving system of Ax = b, A = dpo, b = dge, using dpotrs;");
#endif
F77_CALL(dpotrs)(uplo_P(Chol), adims, bdims + 1, A , adims, B , bdims, &info);
}
if (info) {
if(info > 0)
error(_("the leading minor of order %d is not positive definite"),
info);
else /* should never happen! */
error(_("Lapack routine %s returned error code %d"), "dpotrf", info);
}
UNPROTECT(1);
return val;
#endif
return R_NilValue;
}
SEXP magma_dpoMatrix_solve(SEXP x)
{
#ifdef HIPLAR_WITH_MAGMA
SEXP Chol = magma_dpoMatrix_chol(x);
SEXP val = PROTECT(NEW_OBJECT(MAKE_CLASS("dpoMatrix")));
int *dims = INTEGER(GET_SLOT(x, Matrix_DimSym)), info;
SET_SLOT(val, Matrix_factorSym, allocVector(VECSXP, 0));
slot_dup(val, Chol, Matrix_uploSym);
slot_dup(val, Chol, Matrix_xSym);
slot_dup(val, Chol, Matrix_DimSym);
SET_SLOT(val, Matrix_DimNamesSym,
duplicate(GET_SLOT(x, Matrix_DimNamesSym)));
double *A = REAL(GET_SLOT(val, Matrix_xSym));
int N = *dims;
int lda = N;
const char *uplo = uplo_P(val);
if(GPUFlag == 0) {
F77_CALL(dpotri)(uplo_P(val), dims, A, dims, &info);
}
else if(GPUFlag == 1 && Interface == 0) {
#ifdef HIPLAR_DBG
R_ShowMessage("DBG: Solving using magma_dpotri");
#endif
magma_dpotri(uplo[0], N, A, lda, &info);
}
else if(GPUFlag == 1 && Interface == 1){
double *d_A;
cublasStatus retStatus;
cublasAlloc( N * lda , sizeof(double), (void**)&d_A);
#ifdef HIPLAR_DBG
R_ShowMessage("DBG: Solving using magma_dpotri_gpu");
#endif
/* Error Checking */
retStatus = cublasGetError ();
if (retStatus != CUBLAS_STATUS_SUCCESS)
error(_("CUBLAS: Error in Memory Allocation"));
/********************************************/
cublasSetVector( N * lda, sizeof(double), A, 1, d_A, 1);
/* Error Checking */
retStatus = cublasGetError ();
if (retStatus != CUBLAS_STATUS_SUCCESS)
error(_("CUBLAS: Error in Data Transfer to Device"));
/********************************************/
magma_dpotri_gpu(uplo[0], N, d_A, lda, &info);
cublasGetVector(N * lda, sizeof(double), d_A, 1, val, 1);
/* Error Checking */
retStatus = cublasGetError ();
if (retStatus != CUBLAS_STATUS_SUCCESS)
error(_("CUBLAS: Error in Data Transfer from Device"));
/********************************************/
cublasFree(d_A);
}
else
error(_("MAGMA/LAPACK/Interface Flag not defined correctly"));
if (info) {
if(info > 0)
error(_("the leading minor of order %d is not positive definite"),
info);
else /* should never happen! */
error(_("Lapack routine %s returned error code %d"), "dpotrf", info);
}
UNPROTECT(1);
return val;
#endif
return R_NilValue;
}
SEXP Tsparse_diagU2N(SEXP x)
{
static const char *valid[] = {
"dtTMatrix", /* 0 */
"ltTMatrix", /* 1 */
"ntTMatrix", /* 2 : no x slot */
"ztTMatrix", /* 3 */
""};
/* #define xSXP(iTyp) ((iTyp == 0) ? REALSXP : ((iTyp == 1) ? LGLSXP : /\* else *\/ CPLXSXP)); */
/* #define xTYPE(iTyp) ((iTyp == 0) ? double : ((iTyp == 1) ? int : /\* else *\/ Rcomplex)); */
int ctype = Matrix_check_class_etc(x, valid);
if (ctype < 0 || *diag_P(x) != 'U') {
/* "trivially fast" when not triangular (<==> no 'diag' slot),
or not *unit* triangular */
return (x);
}
else { /* instead of going to Csparse -> Cholmod -> Csparse -> Tsparse, work directly: */
int i, n = INTEGER(GET_SLOT(x, Matrix_DimSym))[0],
nnz = length(GET_SLOT(x, Matrix_iSym)),
new_n = nnz + n;
SEXP ans = PROTECT(NEW_OBJECT(MAKE_CLASS(class_P(x))));
int *islot = INTEGER(ALLOC_SLOT(ans, Matrix_iSym, INTSXP, new_n)),
*jslot = INTEGER(ALLOC_SLOT(ans, Matrix_jSym, INTSXP, new_n));
slot_dup(ans, x, Matrix_DimSym);
SET_DimNames(ans, x);
slot_dup(ans, x, Matrix_uploSym);
SET_SLOT(ans, Matrix_diagSym, mkString("N"));
/* Build the new i- and j- slots : first copy the current : */
Memcpy(islot, INTEGER(GET_SLOT(x, Matrix_iSym)), nnz);
Memcpy(jslot, INTEGER(GET_SLOT(x, Matrix_jSym)), nnz);
/* then, add the new (i,j) slot entries: */
for(i = 0; i < n; i++) {
islot[i + nnz] = i;
jslot[i + nnz] = i;
}
/* build the new x-slot : */
switch(ctype) {
case 0: { /* "d" */
double *x_new = REAL(ALLOC_SLOT(ans, Matrix_xSym,
REALSXP, new_n));
Memcpy(x_new, REAL(GET_SLOT(x, Matrix_xSym)), nnz);
for(i = 0; i < n; i++) /* add x[i,i] = 1. */
x_new[i + nnz] = 1.;
break;
}
case 1: { /* "l" */
int *x_new = LOGICAL(ALLOC_SLOT(ans, Matrix_xSym,
LGLSXP, new_n));
Memcpy(x_new, LOGICAL(GET_SLOT(x, Matrix_xSym)), nnz);
for(i = 0; i < n; i++) /* add x[i,i] = 1 (= TRUE) */
x_new[i + nnz] = 1;
break;
}
case 2: /* "n" */
/* nothing to do here */
break;
case 3: { /* "z" */
Rcomplex *x_new = COMPLEX(ALLOC_SLOT(ans, Matrix_xSym,
CPLXSXP, new_n));
Memcpy(x_new, COMPLEX(GET_SLOT(x, Matrix_xSym)), nnz);
for(i = 0; i < n; i++) /* add x[i,i] = 1 (= TRUE) */
x_new[i + nnz] = (Rcomplex) {1., 0.};
break;
}
}/* switch() */
UNPROTECT(1);
return ans;
}
}
