diff --git a/la_traits.h b/la_traits.h
index 22e76c2..054f4e3 100644
--- a/la_traits.h
+++ b/la_traits.h
@@ -392,10 +392,10 @@ static inline bool bigger(const  X<C> &x, const X<C> &y) {return x>y;} \
 static inline bool smaller(const  X<C> &x, const X<C> &y) {return x<y;} \
 static inline normtype norm (const X<C> &x) {return x.norm();} \
 static inline void axpy (X<C>&s, const X<C> &x, const C c) {s.axpy(c,x);} \
-static void put(int fd, const X<C> &x, bool dimensions=1, bool transp=0) {x.put(fd,dimensions,transp);} \
-static void get(int fd, X<C> &x, bool dimensions=1, bool transp=0) {x.get(fd,dimensions,transp);} \
-static void multiput(size_t n,int fd, const X<C> *x, bool dimensions=1) {for(size_t i=0; i<n; ++i) x[i].put(fd,dimensions);} \
-static void multiget(size_t n,int fd, X<C> *x, bool dimensions=1) {for(size_t i=0; i<n; ++i) x[i].get(fd,dimensions);} \
+static void put(int fd, const X<C> &x, bool dimensions=1, bool transp=0, bool orcaformat=false) {x.put(fd,dimensions,transp,orcaformat);} \
+static void get(int fd, X<C> &x, bool dimensions=1, bool transp=0, bool orcaformat=false) {x.get(fd,dimensions,transp,orcaformat);} \
+static void multiput(size_t n,int fd, const X<C> *x, bool dimensions=1, bool orcaformat=false) {for(size_t i=0; i<n; ++i) x[i].put(fd,dimensions,false,orcaformat);} \
+static void multiget(size_t n,int fd, X<C> *x, bool dimensions=1, bool orcaformat=false) {for(size_t i=0; i<n; ++i) x[i].get(fd,dimensions,false,orcaformat);} \
 static void copy(X<C> *dest, X<C> *src, size_t n) {for(size_t i=0; i<n; ++i) dest[i]=src[i];} \
 static void clear(X<C> *dest, size_t n) {for(size_t i=0; i<n; ++i) dest[i].clear();}\
 static void copyonwrite(X<C> &x) {x.copyonwrite();}\
@@ -435,10 +435,10 @@ static inline bool bigger(const  C &x, const C &y) {return x>y;} \
 static inline bool smaller(const  C &x, const C &y) {return x<y;} \
 static inline normtype norm (const X<C> &x) {return x.norm();}  \
 static inline void axpy (X<C>&s, const X<C> &x, const C c) {s.axpy(c,x);}  \
-static void put(int fd, const X<C> &x, bool dimensions=1, bool transp=0) {x.put(fd,dimensions);}  \
-static void get(int fd, X<C> &x, bool dimensions=1, bool transp=0) {x.get(fd,dimensions);}  \
-static void multiput(size_t n,int fd, const X<C> *x, bool dimensions=1) {for(size_t i=0; i<n; ++i) x[i].put(fd,dimensions);}  \
-static void multiget(size_t n,int fd, X<C> *x, bool dimensions=1) {for(size_t i=0; i<n; ++i) x[i].get(fd,dimensions);}  \
+static void put(int fd, const X<C> &x, bool dimensions=1, bool transp=0, bool orcaformat=false) {x.put(fd,dimensions,false,orcaformat);}  \
+static void get(int fd, X<C> &x, bool dimensions=1, bool transp=0, bool orcaformat=false) {x.get(fd,dimensions,false,orcaformat);}  \
+static void multiput(size_t n,int fd, const X<C> *x, bool dimensions=1, bool orcaformat=false) {for(size_t i=0; i<n; ++i) x[i].put(fd,dimensions,false,orcaformat);}  \
+static void multiget(size_t n,int fd, X<C> *x, bool dimensions=1, bool orcaformat=false) {for(size_t i=0; i<n; ++i) x[i].get(fd,dimensions,false,orcaformat);}  \
 static void copy(C *dest, C *src, size_t n) {for(size_t i=0; i<n; ++i) dest[i]=src[i];}  \
 static void clear(C *dest, size_t n) {for(size_t i=0; i<n; ++i) dest[i].clear();} \
 static void copyonwrite(X<C> &x) {x.copyonwrite();} \
diff --git a/mat.cc b/mat.cc
index 05cb918..df373e9 100644
--- a/mat.cc
+++ b/mat.cc
@@ -121,12 +121,12 @@ const NRVec<T> NRMat<T>::row(const int i, int l) const {
  * @see NRVec<T>::put()
  ******************************************************************************/
 template <typename T>
-void NRMat<T>::put(int fd, bool dim, bool transp) const {
+void NRMat<T>::put(int fd, bool dim, bool transp, bool orcaformat) const {
 #ifdef CUDALA
 	if(location != cpu) {
 		NRMat<T> tmp = *this;
 		tmp.moveto(cpu);
-		tmp.put(fd, dim, transp);
+		tmp.put(fd, dim, transp, orcaformat);
 		return;
 	}
 #endif
@@ -134,6 +134,11 @@ void NRMat<T>::put(int fd, bool dim, bool transp) const {
 	if(dim){
 		if(sizeof(int) != write(fd,&(transp?mm:nn),sizeof(int))) laerror("write failed");
 		if(sizeof(int) != write(fd,&(transp?nn:mm),sizeof(int))) laerror("write failed");
+		if(orcaformat)
+			{
+			int tmp=sizeof(T);
+			if(sizeof(int) != write(fd,&tmp,sizeof(int))) laerror("write failed");
+			}
 	}
 
 	if(transp){  //not particularly efficient
@@ -167,12 +172,12 @@ void NRMat<T>::put(int fd, bool dim, bool transp) const {
  * @see NRVec<T>::get(), copyonwrite()
  ******************************************************************************/
 template <typename T>
-void NRMat<T>::get(int fd, bool dim, bool transp){
+void NRMat<T>::get(int fd, bool dim, bool transp, bool orcaformat){
 #ifdef CUDALA
 	if(location != cpu){
 		NRMat<T> tmp;
 		tmp.moveto(cpu);
-		tmp.get(fd, dim, transp);
+		tmp.get(fd, dim, transp,orcaformat);
 		tmp.moveto(getlocation());
 		*this = tmp;
 		return;
@@ -183,6 +188,12 @@ void NRMat<T>::get(int fd, bool dim, bool transp){
 	if(dim){
 		if(sizeof(int) != read(fd, &nn0, sizeof(int))) laerror("read failed");
 		if(sizeof(int) != read(fd, &mm0, sizeof(int))) laerror("read failed");
+		if(orcaformat)
+			{
+			int tmp;
+			if(sizeof(int) != read(fd, &tmp, sizeof(int))) laerror("read failed");
+			if(tmp!=sizeof(T)) laerror("mismatch in orca format");
+			}
 		if(transp) resize(mm0, nn0); else resize(nn0, mm0);
 	}else{
 		copyonwrite();
diff --git a/mat.h b/mat.h
index c3e8561..eba5a8e 100644
--- a/mat.h
+++ b/mat.h
@@ -305,9 +305,9 @@ public:
 	inline size_t size() const;
 
 	//! unformatted input
-        void get(int fd, bool dimensions = 1, bool transposed = false);
+        void get(int fd, bool dimensions = 1, bool transposed = false, bool orcaformat=false);
 	//! unformatted output
-        void put(int fd, bool dimensions = 1, bool transposed = false) const;
+        void put(int fd, bool dimensions = 1, bool transposed = false, bool orcaformat=false) const;
 	//! formatted output
 	void fprintf(FILE *f, const char *format, const int modulo) const;
 	//! formatted input
diff --git a/smat.cc b/smat.cc
index aaceec8..7b93aa2 100644
--- a/smat.cc
+++ b/smat.cc
@@ -41,12 +41,12 @@ namespace LA {
  * @see NRMat<T>::get(), NRSMat<T>::copyonwrite()
  ******************************************************************************/
 template <typename T>
-void NRSMat<T>::put(int fd, bool dim, bool transp) const {
+void NRSMat<T>::put(int fd, bool dim, bool transp, bool orcaformat) const {
 #ifdef CUDALA
 	if(location != cpu){
 		NRSMat<T> tmp= *this;
 		tmp.moveto(cpu);
-		tmp.put(fd,dim,transp);
+		tmp.put(fd,dim,transp,orcaformat);
 		return;
 	}
 #endif
@@ -54,6 +54,11 @@ void NRSMat<T>::put(int fd, bool dim, bool transp) const {
 	if(dim){
 		if(sizeof(int) != write(fd,&nn,sizeof(int))) laerror("cannot write");
 		if(sizeof(int) != write(fd,&nn,sizeof(int))) laerror("cannot write");
+		if(orcaformat)
+			{
+			int tmp=sizeof(T);
+			if(sizeof(int) != write(fd,&tmp,sizeof(int))) laerror("cannot write");
+			}
 	}
 	LA_traits<T>::multiput((size_t)nn*(nn+1)/2,fd,v,dim);
 }
@@ -66,12 +71,12 @@ void NRSMat<T>::put(int fd, bool dim, bool transp) const {
  * @see NRSMat<T>::put(), NRSMat<T>::copyonwrite()
  ******************************************************************************/
 template <typename T>
-void NRSMat<T>::get(int fd, bool dim, bool transp) {
+void NRSMat<T>::get(int fd, bool dim, bool transp, bool orcaformat) {
 #ifdef CUDALA
 	if(location != cpu){
 		NRSMat<T> tmp;
 		tmp.moveto(cpu);
-		tmp.get(fd,dim,transp);
+		tmp.get(fd,dim,transp,orcaformat);
 		tmp.moveto(location);
 		*this = tmp;
 		return;
@@ -82,6 +87,12 @@ void NRSMat<T>::get(int fd, bool dim, bool transp) {
 	errno = 0;
 	if(dim){
 		if(2*sizeof(int) != read(fd,&nn0,2*sizeof(int))) laerror("cannot read");
+		if(orcaformat)
+			{
+			int tmp;
+			if(sizeof(int) != read(fd,&tmp,sizeof(int))) laerror("cannot read");
+			if(tmp!=sizeof(T)) laerror("mismatch in orca format");
+			}
 		resize(nn0[0]);
 	}else{
 		copyonwrite();
diff --git a/smat.h b/smat.h
index 96eb517..e795119 100644
--- a/smat.h
+++ b/smat.h
@@ -183,8 +183,8 @@ public:
 
 
 	const T trace() const;
-	void get(int fd, bool dimensions = 1, bool transp = 0);
-        void put(int fd, bool dimensions = 1, bool transp = 0) const;
+	void get(int fd, bool dimensions = 1, bool transp = 0, bool orcaformat=false);
+        void put(int fd, bool dimensions = 1, bool transp = 0, bool orcaformat=false) const;
 
 	void copyonwrite(bool detachonly=false, bool deep=true);
 
diff --git a/vec.h b/vec.h
index 3b8ebbe..2623a29 100644
--- a/vec.h
+++ b/vec.h
@@ -455,12 +455,12 @@ public:
 	//! routine for formatted output 
 	void fprintf(FILE *f, const char *format, const int modulo) const;
 	//! routine for unformatted output 
-        void put(int fd, bool dimensions=1, bool transp=0) const;
+        void put(int fd, bool dimensions=1, bool transp=0, bool orca_format=false) const;
 	
 	//! routine for formatted input 
 	void fscanf(FILE *f, const char *format);
 	//! routine for unformatted input 
-        void get(int fd, bool dimensions=1, bool transp=0);
+        void get(int fd, bool dimensions=1, bool transp=0, bool orca_format=false);
 
 	//! constructor creating vector from sparse matrix
 	explicit NRVec(const SparseMat<T> &rhs);
@@ -2015,12 +2015,12 @@ inline const std::complex<double> NRVec<std::complex<double> >::amin() const {
  * @see NRMat<T>::put()
  ******************************************************************************/
 template <typename T>
-void NRVec<T>::put(int fd, bool dim, bool transp) const {
+void NRVec<T>::put(int fd, bool dim, bool transp, bool orcaformat) const {
 #ifdef CUDALA
 	if(location != cpu){
 		NRVec<T> tmp = *this;
 		tmp.moveto(cpu);
-		tmp.put(fd,dim,transp);
+		tmp.put(fd,dim,transp,orcaformat);
 		return;
 	}
 #endif
@@ -2028,7 +2028,15 @@ void NRVec<T>::put(int fd, bool dim, bool transp) const {
 	int pad(1); //align at least 8-byte
 	if(dim){
 		if(sizeof(int) != write(fd,&nn,sizeof(int))) laerror("write failed");
-		if(sizeof(int) != write(fd,&pad,sizeof(int))) laerror("write failed");
+		if(!orcaformat) 
+			{
+			if(sizeof(int) != write(fd,&pad,sizeof(int))) laerror("write failed");
+			}
+		if(orcaformat)
+			{
+			int tmp=sizeof(T);
+			if(sizeof(int) != write(fd,&tmp,sizeof(int))) laerror("write failed");
+			}
 	}
 	LA_traits<T>::multiput(nn,fd,v,dim);
 }
@@ -2041,12 +2049,12 @@ void NRVec<T>::put(int fd, bool dim, bool transp) const {
  * @see NRMat<T>::get(), copyonwrite()
  ******************************************************************************/
 template <typename T>
-void NRVec<T>::get(int fd, bool dim, bool transp) {
+void NRVec<T>::get(int fd, bool dim, bool transp, bool orcaformat) {
 #ifdef CUDALA
 	if(location != cpu){
 		NRVec<T> tmp;
 		tmp.moveto(cpu);
-		tmp.get(fd,dim,transp);
+		tmp.get(fd,dim,transp,orcaformat);
 		tmp.moveto(location);
 		*this = tmp;
 		return;
@@ -2055,7 +2063,14 @@ void NRVec<T>::get(int fd, bool dim, bool transp) {
 	int nn0[2]; //align at least 8-byte
 	errno = 0;
 	if(dim){
-		if(2*sizeof(int) != read(fd,&nn0,2*sizeof(int))) laerror("read failed");
+		int ndims=orcaformat?1:2;
+		if(ndims*sizeof(int) != read(fd,&nn0,ndims*sizeof(int))) laerror("read failed");
+		if(orcaformat)
+			{
+			int tmp;
+			if(sizeof(int) != read(fd,&tmp,sizeof(int))) laerror("read failed");
+			if(tmp!=sizeof(T)) laerror("mismatch in orca format");
+			}
 		resize(nn0[0]);
 	}else{
 		copyonwrite();