jiri/LA_library
1
0 Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

bugfixes in vec mat smat - mainly missing copyonwrite()s

Browse Source
This commit is contained in:
jiri committed 2026-10-26 15:11:45 +01:00
1 parent eccb3a68f8
commit f7ef134fc5
5 files changed
+31 -10
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

mat.cc
+18 -7
View File
@@ -127,6 +127,7 @@ void NRMat<T>::rowset(const NRVec<T> &r, const int i, int l, int offset) {
if(i < 0 || i >= nn) laerror("illegal index");
if(offset<0||l+offset>mm) laerror("illegal len/offset");
#endif
copyonwrite();
LA_traits<T>::copy(
#ifdef MATPTR
v[i]+offset
@@ -893,6 +894,7 @@ void NRMat<T>::storesubmatrix(const int fromrow, const int fromcol, const NRMat
#endif
SAME_LOC(*this, rhs);
if(!LA_traits<T>::is_plaindata()) laerror("only implemented for plain data");
copyonwrite();
const int m = tocol - fromcol + 1;
for(register int i = fromrow; i <= torow; ++i){
@@ -921,7 +923,8 @@ void NRMat<T>::storesubmatrix(const NRVec<int> &rows, const NRVec<int> &cols, co
const int n = rows.size();
const int m = cols.size();
if(rhs.nrows()!=n || rhs.ncols()!=m) laerror("incompatible dimensions in storesubmatrix");
copyonwrite();
for(int i=0; i<n; ++i)
{
int ii=rows[i];
@@ -942,7 +945,7 @@ void NRMat<T>::storesubmatrix(const NRVec<int> &rows, const NRVec<int> &cols, co
******************************************************************************/
template <typename T>
NRMat<T>& NRMat<T>::transposeme(const int _n) {
const int n = (n <= 0)?nn:_n;//!< transpose the entire matrix
const int n = (_n <= 0)?nn:_n;//!< transpose the entire matrix
#ifdef DEBUG
if (n==nn && nn != mm || n>mm || n>nn ) laerror("NRMat<T>::transposeme() - invalid parameter n. Non-square matrix?");
#endif
@@ -1184,23 +1187,27 @@ void NRMat<T>::fscanf(FILE *f, const char *format) {
int n(0), m(0);
if (::fscanf(f, "%d %d", &n, &m) != 2) laerror("cannot read matrix dimensions");
resize(n, m);
#ifdef CUDALA
if(location == cpu){
p = *this;
#endif
p = *this; //resize detached any shared data
#ifdef CUDALA
}else{
tmp = new NRMat<T>(n, m, this->location);
tmp = new NRMat<T>(n, m, cpu); //read on cpu, then copied to the device
p = *tmp;
}
#endif
resize(n, m);
for(int i=0; i<n; i++){
for(int j=0; j<n; j++){
for(int j=0; j<m; j++){
if(::fscanf(f,format, p++) != 1) laerror("cannot read matrix element");
}
}
#ifdef CUDALA
if(location != cpu){
tmp->moveto(this->location);
*this |= *tmp;
delete tmp;
}
#endif
@@ -1402,6 +1409,7 @@ void NRMat<double>::randomize(const double &x) {
#ifdef CUDALA
if(location == cpu){
#endif
copyonwrite();
for(register int i=0; i<nn; ++i){
for(register int j=0; j<mm; ++j){
(*this)(i,j) = x*RANDDOUBLESIGNED();
@@ -1431,6 +1439,7 @@ void NRMat<std::complex<double> >::randomize(const double &x) {
#ifdef CUDALA
if(location == cpu){
#endif
copyonwrite();
for(register int i=0; i<nn; ++i){
for(register int j=0; j<mm; ++j){
const double re = x*RANDDOUBLESIGNED();
@@ -2641,7 +2650,8 @@ int nnmin= nn<=mm?nn:mm;
template<>
void NRMat<double>::orthonormalize(const bool rowcol, const NRSMat<double> *metric) {
SAME_LOC(*this, *metric);
if(metric) SAME_LOC(*this, *metric);
copyonwrite();
if(metric){
if(rowcol){ //vectors are stored in rows
if((*metric).nrows() != mm) laerror("incompatible metric in NRMat<double>::orthonormalize(rowcol = true)");
@@ -3271,6 +3281,7 @@ template<typename T>
void NRMat<T>::axpy(const T alpha, const NRPerm<int> &p, const bool direction)
{
int n=p.size();
copyonwrite();
for(int i=0; i<n; ++i)
{
if(direction) (*this)(i,p[i+1]-1) +=alpha;