diff --git a/bitvector.cc b/bitvector.cc
index 0a2c779..6f6ed57 100644
--- a/bitvector.cc
+++ b/bitvector.cc
@@ -29,9 +29,10 @@ return s;
 
 std::istream & operator>>(std::istream  &s, bitvector &x)
 {
-bool a;
-x.copyonwrite();
-for(unsigned int i=0; i<x.size(); ++i) {s >>a; x.assign(i,a);}
+std::string str;
+s >> str;
+x.resize(str.size());
+for(unsigned int i=0; i<x.size(); ++i) {x.assign(i,str[i]!='0');}
 return s;
 }
 
diff --git a/configure.ac b/configure.ac
index 3920212..a479e13 100644
--- a/configure.ac
+++ b/configure.ac
@@ -17,11 +17,39 @@ AC_PROG_CC
 #AC_PROG_RANLIB obsoleted by libtool
 AC_LANG(C++)
 
+# Check for Intel MKL first
+#user has to e.g. setenv LDFLAGS "-L/opt/intel/composer_xe_2013_sp1.1.106/mkl/lib/intel64/"
+#and              setenv CXXFLAGS "-I/opt/intel/composer_xe_2013_sp1.1.106/mkl/include/"
+MKLLIB=""
+mkl_present=1
+AC_CHECK_LIB([mkl_intel_lp64 -lmkl_sequential -lmkl_core -lpthread], [dgemm_], [
+		echo "MKL LIBRARY FOUND"
+	] , [
+	mkl_present=0
+        ])
+
+AC_CHECK_HEADER(mkl_cblas.h, [echo "MKL INCLUDE FOUND"], [mkl_present=0])
+
+if test $mkl_present == 1
+then
+	MKLLIB="-Wl,--start-group -lmkl_intel_lp64 -lmkl_sequential -lmkl_core -Wl,--end-group -lpthread"
+	AC_SUBST([MKLLIB])
+	CLAPACKOPT=-DNONCLAPACK
+	AC_SUBST([CLAPACKOPT])
+	MKLOPT="-DHAS_MKL"
+	AC_SUBST([MKLOPT])
+else
+	echo "MKL LIBRARY OR HEADER NOT FOUND, will try ATLAS next!"
+fi
+
 
 # Checks for mandatory libraries.
 ATLASLIB=""	
+
+if test $mkl_present == 0
+then
 AC_CHECK_LIB([atlas], [ATL_zgemv],ATLASLIB="-lblas -latlas", [
-        echo "ATLAS not found, I hope you are using some other (more) efficient BLAS!"
+        echo "ATLAS not found, I hope you are using some other efficient BLAS!"
         ])
 AC_SUBST([ATLASLIB])
 
@@ -30,7 +58,7 @@ AC_CHECK_LIB([blas], [dgemm_],BLASLIB="-lblas", [
         exit
         ],[$ATLASLIB])
 AC_SUBST([BLASLIB])
-
+fi
 
 MATPTROPT=""
 AC_ARG_ENABLE([matptr],[  --enable-matptr   switch to double** matrix representation (CUDA incompatible)  [[default=no]]],
@@ -43,6 +71,11 @@ AC_ARG_ENABLE([matptr],[  --enable-matptr   switch to double** matrix representa
 
 
 #cblas and clapack available?
+CBLASLIB=""
+CBLASOPT=""
+
+if test $mkl_present == 0
+then
 AC_CHECK_LIB([cblas], [cblas_ddot], [CBLASLIB="-lcblas"], [CBLASOPT=-DNONCBLAS],[$ATLASLIB])
 AC_CHECK_HEADER([cblas.h],,[CBLASOPT=-DNONCBLAS CBLASLIB=""], AC_INCLUDES_DEFAULT)
 AC_SUBST([CBLASLIB])
@@ -59,6 +92,8 @@ AC_CHECK_LIB([lapack], [clapack_dgesv], , [CLAPACKOPT=-DNONCLAPACK],[$CBLASLIB $
 AC_CHECK_HEADER([clapack.h],,[CLAPACKOPT=-DNONCLAPACK], AC_INCLUDES_DEFAULT)
 AC_SUBST([CLAPACKOPT])
 
+fi
+#MKL end
 
 #CUDA available? link with cublas and avoid cblas and clapack then...
 AC_CHECK_LIB([cublas], [cublasInit], [MATPTROPT="" NVCC=nvcc NVCCFLAGS="-O -arch sm_20" CUDALIBS=-lcublas CUDAOPT=-DCUDALA CBLASOPT=-DNONCBLAS CLAPACKOPT=-DNONCLAPACK CBLASLIB=""], [CUDALIB="" CUDAOPT=""])
@@ -138,7 +173,8 @@ echo "**************************************************************************
 echo "The LA library has now been configured. You may run make; make check; make install"
 echo "Please make sure that the generated Makefile employs a proper version of optimized"
 echo "BLAS/LAPACK library. If not, re-run configure with CXXFLAGS and LDFLAGS options   "
-echo "set to '-I path' and '-L path' for your preferred BLAS/LAPACK library version.    "
+echo "set to '-I/path' and '-L/path' for your preferred BLAS/LAPACK library version.    "
+echo "Either use export or specify these options on the command line ./configure .....  "
 echo "In addition, similarly you might set include and link paths for Nvidia CUBLAS.    "
 echo "Use  ./configure --disable-optimize CXXFLAGS='' LDFLAGS='' for a fast compile.    "
 echo "Use --enable-fotran64int to link with BLAS and LAPACK using 64-bit integers.      "
diff --git a/davidson.h b/davidson.h
index 4cf7572..be7daa7 100644
--- a/davidson.h
+++ b/davidson.h
@@ -34,6 +34,7 @@ namespace LA {
 //Note that for efficiency in a direct CI case the diagonalof() should cache its result
 
 
+//@@@ for large krylov spaces >200 it can occur 'convergence problem in sygv/syev in diagonalize()'
 //@@@options: left eigenvectors by matrix transpose, overridesymmetric (for nrmat)
 //@@@small matrix gdiagonalize - shift complex roots up (option to gdiagonalize?)
 //@@@test gdiagonalize whether it sorts the roots and what for complex ones
@@ -73,7 +74,7 @@ else
 
 int i,j;
 
-if(maxkrylov<maxit) maxit=maxkrylov;
+//NO, we will restart, maxit can be bigger if(maxkrylov<maxit) maxit=maxkrylov;
 if(nroots>=maxkrylov) nroots =maxkrylov-1;
 int  nroot=0;
 int oldnroot;
@@ -139,7 +140,9 @@ if(it>0) //if this is the first iteration just need to diagonalise the matrix
 smallV=smallH;
 NRMat<T> smallSwork=smallS;
 if(bigmat.issymmetric()) 
+	{
 	diagonalize(smallV,r,1,1,krylovsize+1,&smallSwork,1); //for symmetric matrix they have already been sorted to ascending order in lapack
+	}
 else 
 	{
 	NRVec<T> ri(krylovsize+1),beta(krylovsize+1);
@@ -160,7 +163,7 @@ for(int iroot=0; iroot<=std::min(krylovsize,nroots-1); ++iroot)
         if(test>eps) nroot=std::min(nroot,iroot);
         if(verbose && iroot<=std::max(oldnroot,nroot)) 
 		{
-		std::cout <<"Davidson: iter="<<it <<" dim="<<krylovsize<<" root="<<iroot<<" energy="<<r[iroot]<<"\n";
+		std::cout <<"Davidson: iter="<<it <<" dim="<<krylovsize<<" root="<<iroot<<" eigenvalue="<<r[iroot]<<"\n";
 		std::cout.flush();
 		}
 	}
@@ -169,7 +172,7 @@ if(verbose && oldnroot!=nroot) {std::cout <<"root no. "<<oldnroot<<" converged\n
 if (nroot>=nroots) goto converged;
 if (it==maxit-1) break; //not converged
 
-if (krylovsize==maxkrylov) //restart, krylov space exceeded
+if (krylovsize==maxkrylov-1) //restart, krylov space exceeded
 	{
 	if(nroot!=0) {flag=1; goto finished;}
 	smallH=0;
@@ -180,7 +183,7 @@ if (krylovsize==maxkrylov) //restart, krylov space exceeded
 		if(!incore) s0->get(vec2,i);
                 vec1.axpy(smallV(i,0),incore?v0[i]:vec2);
 		}
-	s0->put(vec1,0);
+	if(!incore) s0->put(vec1,0);
 	vec1.normalize();
         krylovsize = 0;
 	continue;
diff --git a/la_traits.h b/la_traits.h
index 23ceab9..6b5c9cb 100644
--- a/la_traits.h
+++ b/la_traits.h
@@ -49,7 +49,11 @@
 #include "noncblas.h"
 #else
 extern "C" {
+#ifdef HAS_MKL
+#include "mkl_cblas.h"
+#else
 #include "cblas.h"
+#endif
 }
 #endif
 
diff --git a/mat.h b/mat.h
index 8a997a4..9fd9fb1 100644
--- a/mat.h
+++ b/mat.h
@@ -45,6 +45,8 @@ public:
 #endif
 	friend class NRVec<T>;
 	friend class NRSMat<T>;
+	friend class NRMat_from1<T>;
+	friend class NRSMat_from1<T>;
 	
 	//! standard destructor
 	~NRMat();
@@ -550,7 +552,7 @@ NRMat<T>::NRMat(const NRSMat<T> &rhs) {
 
 	int i(0), j(0), k(0);
 	nn = mm = rhs.nrows();
-	 count = new int;
+	count = new int;
 	*count = 1;
 #ifdef MATPTR
 	v = new T*[nn];
@@ -1222,6 +1224,7 @@ class NRMat_from1 : public NRMat<T> {
 public:
 	NRMat_from1(): NRMat<T>() {};
 	explicit NRMat_from1(const int n): NRMat<T>(n) {};
+	explicit NRMat_from1(const NRSMat_from1<T> &rhs) : NRMat<T>(rhs) {};
 	NRMat_from1(const NRMat<T> &rhs): NRMat<T>(rhs) {};//!< be able to convert the parent class transparently to this
 	NRMat_from1(const int n, const int m): NRMat<T>(n, m) {};
 	NRMat_from1(const T &a, const int n, const int m): NRMat<T>(a, n, m) {};
@@ -1342,6 +1345,10 @@ void NRMat<T>::moveto(const GPUID dest) {
 }
 #endif
 
+
+
+ 
+
 /***************************************************************************//**
  * generate operators: Mat + a, a + Mat, Mat * a
  * corresponding macro is defined in vec.h
diff --git a/noncblas.cc b/noncblas.cc
index 923be5b..8aa3266 100644
--- a/noncblas.cc
+++ b/noncblas.cc
@@ -188,7 +188,7 @@ void cblas_zdotc_sub(const int N, const void *X, const int incX, const void *Y,
 
 
 extern "C" void FORNAME(dspmv) (const char *uplo, const FINT *n, const double *alpha, const double *ap, const double *x, const FINT *incx, const double *beta, double *y, const FINT *incy);
-void cblas_dspmv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
+void cblas_dspmv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
 		const int N, const double alpha, const double *Ap,
 		const double *X, const int incX,
 		const double beta, double *Y, const int incY)
@@ -208,7 +208,7 @@ void cblas_dspmv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
 
 
 extern "C" void FORNAME(zhpmv) (const char *uplo, const FINT *n, const void *alpha, const void *ap, const void *x, const FINT *incx, const void *beta, void *y, const FINT *incy);
-void cblas_zhpmv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
+void cblas_zhpmv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
 		const int N, const void *alpha, const void *Ap,
 		const void *X, const int incX,
 		const void *beta, void *Y, const int incY)
@@ -230,7 +230,7 @@ void cblas_zhpmv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
 //Level 2 and Level 3 on general matrices - take into account the transposed storage of matrices in Fortran and C
 
 extern "C" void FORNAME(dger) (const FINT *m, const FINT *n, const double *alpha, const double *x, const FINT *incx, const double *y, const FINT *incy, double *a, const FINT *lda);
-void cblas_dger(const enum CBLAS_ORDER Order, const int M, const int N,
+void cblas_dger(const CBLAS_ORDER Order, const int M, const int N,
 		const double alpha, const double *X, const int incX,
 		const double *Y, const int incY, double *A, const int lda)
 {
@@ -249,7 +249,7 @@ void cblas_dger(const enum CBLAS_ORDER Order, const int M, const int N,
 }
 
 extern "C" void FORNAME(zgerc)  (const FINT *m, const FINT *n, const void *alpha, const void *x, const FINT *incx, const void *y, const FINT *incy, void *a, const FINT *lda);
-void cblas_zgerc(const enum CBLAS_ORDER Order, const int M, const int N,
+void cblas_zgerc(const CBLAS_ORDER Order, const int M, const int N,
 		const void *alpha, const void *X, const int incX,
 		const void *Y, const int incY, void *A, const int lda)
 {
@@ -273,7 +273,7 @@ p= (complex<double> *) Y; for(int i=0; i<N; ++i) {p->imag() = -p->imag(); p+= in
 
 
 extern "C" void FORNAME(zgeru)  (const FINT *m, const FINT *n, const void *alpha, const void *x, const FINT *incx, const void *y, const FINT *incy, void *a, const FINT *lda);
-void cblas_zgeru(const enum CBLAS_ORDER Order, const int M, const int N,
+void cblas_zgeru(const CBLAS_ORDER Order, const int M, const int N,
 		const void *alpha, const void *X, const int incX,
 		const void *Y, const int incY, void *A, const int lda)
 {
@@ -292,8 +292,8 @@ void cblas_zgeru(const enum CBLAS_ORDER Order, const int M, const int N,
 
 
 extern "C" void FORNAME(dgemm) (const char *transa, const char *transb, const FINT *m, const FINT *n, const FINT *k, const double *alpha, const double *a, const FINT *lda, const double *b, const FINT *ldb, const double *beta, double *c, const FINT *ldc);
-void cblas_dgemm(const enum CBLAS_ORDER Order, const enum CBLAS_TRANSPOSE TransA,
-		const enum CBLAS_TRANSPOSE TransB, const int M, const int N,
+void cblas_dgemm(const CBLAS_ORDER Order, const CBLAS_TRANSPOSE TransA,
+		const CBLAS_TRANSPOSE TransB, const int M, const int N,
 		const int K, const double alpha, const double *A,
 		const int lda, const double *B, const int ldb,
 		const double beta, double *C, const int ldc)
@@ -318,8 +318,8 @@ void cblas_dgemm(const enum CBLAS_ORDER Order, const enum CBLAS_TRANSPOSE TransA
 }
 
 extern "C" void FORNAME(zgemm) (const char *transa, const char *transb, const FINT *m, const FINT *n, const FINT *k, const void *alpha, const void *a, const FINT *lda, const void *b, const FINT *ldb, const void *beta, void *c, const FINT *ldc);
-void cblas_zgemm(const enum CBLAS_ORDER Order, const enum CBLAS_TRANSPOSE TransA,
-		const enum CBLAS_TRANSPOSE TransB, const int M, const int N,
+void cblas_zgemm(const CBLAS_ORDER Order, const CBLAS_TRANSPOSE TransA,
+		const CBLAS_TRANSPOSE TransB, const int M, const int N,
 		const int K, const void *alpha, const void *A,
 		const int lda, const void *B, const int ldb,
 		const void *beta, void *C, const int ldc)
@@ -345,8 +345,8 @@ void cblas_zgemm(const enum CBLAS_ORDER Order, const enum CBLAS_TRANSPOSE TransA
 
 
 extern "C" void FORNAME(dgemv)  (const char *TRANS, const FINT *M, const FINT *N, const double *ALPHA, const double *A,  const FINT *LDA,  const double *X,  const FINT *INCX, const double *BETA, double *Y, const FINT *INCY);
-void cblas_dgemv(const enum CBLAS_ORDER Order,
-		const enum CBLAS_TRANSPOSE TransA, const int M, const int N,
+void cblas_dgemv(const CBLAS_ORDER Order,
+		const CBLAS_TRANSPOSE TransA, const int M, const int N,
 		const double alpha, const double *A, const int lda,
 		const double *X, const int incX, const double beta,
 		double *Y, const int incY)
@@ -371,8 +371,8 @@ void cblas_dgemv(const enum CBLAS_ORDER Order,
 
 
 extern "C" void FORNAME(zgemv)  (const char *TRANS, const FINT *M, const FINT *N, const void *ALPHA, const void *A,  const FINT *LDA,  const void *X,  const FINT *INCX, const void *BETA, void *Y, const FINT *INCY);
-void cblas_zgemv(const enum CBLAS_ORDER Order,
-		const enum CBLAS_TRANSPOSE TransA, const int M, const int N,
+void cblas_zgemv(const CBLAS_ORDER Order,
+		const CBLAS_TRANSPOSE TransA, const int M, const int N,
 		const void *alpha, const void *A, const int lda,
 		const void *X, const int incX, const void *beta,
 		void *Y, const int incY)
@@ -446,7 +446,7 @@ CBLAS_INDEX cblas_izamin(const int N, const void *X, const int incX) {
 //allocate auxiliary storage and transpose input and output quantities to fortran/C order
 extern "C" void FORNAME(dgesv) (const FINT *N, const FINT *NRHS, double *A, const FINT *LDA, FINT *IPIV, double *B, const FINT *LDB, FINT *INFO);
 
-int clapack_dgesv(const enum CBLAS_ORDER Order, const int N, const int NRHS,
+int clapack_dgesv(const CBLAS_ORDER Order, const int N, const int NRHS,
 		double *A, const int lda, int *ipiv,
 		double *B, const int ldb)
 {
diff --git a/noncblas.h b/noncblas.h
index 2bd3cdc..30ab416 100644
--- a/noncblas.h
+++ b/noncblas.h
@@ -191,197 +191,197 @@ void cblas_zdrot(const int N, void *X, const int incX, void *Y, const int incY,
 /*
  * Routines with standard 4 prefixes (S, D, C, Z)
  */
-void cblas_sgemv(const enum CBLAS_ORDER Order,
-                 const enum CBLAS_TRANSPOSE TransA, const int M, const int N,
+void cblas_sgemv(const CBLAS_ORDER Order,
+                 const CBLAS_TRANSPOSE TransA, const int M, const int N,
                  const float alpha, const float *A, const int lda,
                  const float *X, const int incX, const float beta,
                  float *Y, const int incY);
-void cblas_sgbmv(const enum CBLAS_ORDER Order,
-                 const enum CBLAS_TRANSPOSE TransA, const int M, const int N,
+void cblas_sgbmv(const CBLAS_ORDER Order,
+                 const CBLAS_TRANSPOSE TransA, const int M, const int N,
                  const int KL, const int KU, const float alpha,
                  const float *A, const int lda, const float *X,
                  const int incX, const float beta, float *Y, const int incY);
-void cblas_strmv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                 const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+void cblas_strmv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                 const CBLAS_TRANSPOSE TransA, const CBLAS_DIAG Diag,
                  const int N, const float *A, const int lda,
                  float *X, const int incX);
-void cblas_stbmv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                 const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+void cblas_stbmv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                 const CBLAS_TRANSPOSE TransA, const CBLAS_DIAG Diag,
                  const int N, const int K, const float *A, const int lda,
                  float *X, const int incX);
-void cblas_stpmv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                 const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+void cblas_stpmv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                 const CBLAS_TRANSPOSE TransA, const CBLAS_DIAG Diag,
                  const int N, const float *Ap, float *X, const int incX);
-void cblas_strsv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                 const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+void cblas_strsv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                 const CBLAS_TRANSPOSE TransA, const CBLAS_DIAG Diag,
                  const int N, const float *A, const int lda, float *X,
                  const int incX);
-void cblas_stbsv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                 const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+void cblas_stbsv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                 const CBLAS_TRANSPOSE TransA, const CBLAS_DIAG Diag,
                  const int N, const int K, const float *A, const int lda,
                  float *X, const int incX);
-void cblas_stpsv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                 const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+void cblas_stpsv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                 const CBLAS_TRANSPOSE TransA, const CBLAS_DIAG Diag,
                  const int N, const float *Ap, float *X, const int incX);
 
-void cblas_dgemv(const enum CBLAS_ORDER Order,
-                 const enum CBLAS_TRANSPOSE TransA, const int M, const int N,
+void cblas_dgemv(const CBLAS_ORDER Order,
+                 const CBLAS_TRANSPOSE TransA, const int M, const int N,
                  const double alpha, const double *A, const int lda,
                  const double *X, const int incX, const double beta,
                  double *Y, const int incY);
-void cblas_dgbmv(const enum CBLAS_ORDER Order,
-                 const enum CBLAS_TRANSPOSE TransA, const int M, const int N,
+void cblas_dgbmv(const CBLAS_ORDER Order,
+                 const CBLAS_TRANSPOSE TransA, const int M, const int N,
                  const int KL, const int KU, const double alpha,
                  const double *A, const int lda, const double *X,
                  const int incX, const double beta, double *Y, const int incY);
-void cblas_dtrmv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                 const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+void cblas_dtrmv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                 const CBLAS_TRANSPOSE TransA, const CBLAS_DIAG Diag,
                  const int N, const double *A, const int lda,
                  double *X, const int incX);
-void cblas_dtbmv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                 const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+void cblas_dtbmv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                 const CBLAS_TRANSPOSE TransA, const CBLAS_DIAG Diag,
                  const int N, const int K, const double *A, const int lda,
                  double *X, const int incX);
-void cblas_dtpmv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                 const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+void cblas_dtpmv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                 const CBLAS_TRANSPOSE TransA, const CBLAS_DIAG Diag,
                  const int N, const double *Ap, double *X, const int incX);
-void cblas_dtrsv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                 const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+void cblas_dtrsv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                 const CBLAS_TRANSPOSE TransA, const CBLAS_DIAG Diag,
                  const int N, const double *A, const int lda, double *X,
                  const int incX);
-void cblas_dtbsv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                 const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+void cblas_dtbsv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                 const CBLAS_TRANSPOSE TransA, const CBLAS_DIAG Diag,
                  const int N, const int K, const double *A, const int lda,
                  double *X, const int incX);
-void cblas_dtpsv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                 const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+void cblas_dtpsv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                 const CBLAS_TRANSPOSE TransA, const CBLAS_DIAG Diag,
                  const int N, const double *Ap, double *X, const int incX);
 
-void cblas_cgemv(const enum CBLAS_ORDER Order,
-                 const enum CBLAS_TRANSPOSE TransA, const int M, const int N,
+void cblas_cgemv(const CBLAS_ORDER Order,
+                 const CBLAS_TRANSPOSE TransA, const int M, const int N,
                  const void *alpha, const void *A, const int lda,
                  const void *X, const int incX, const void *beta,
                  void *Y, const int incY);
-void cblas_cgbmv(const enum CBLAS_ORDER Order,
-                 const enum CBLAS_TRANSPOSE TransA, const int M, const int N,
+void cblas_cgbmv(const CBLAS_ORDER Order,
+                 const CBLAS_TRANSPOSE TransA, const int M, const int N,
                  const int KL, const int KU, const void *alpha,
                  const void *A, const int lda, const void *X,
                  const int incX, const void *beta, void *Y, const int incY);
-void cblas_ctrmv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                 const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+void cblas_ctrmv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                 const CBLAS_TRANSPOSE TransA, const CBLAS_DIAG Diag,
                  const int N, const void *A, const int lda,
                  void *X, const int incX);
-void cblas_ctbmv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                 const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+void cblas_ctbmv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                 const CBLAS_TRANSPOSE TransA, const CBLAS_DIAG Diag,
                  const int N, const int K, const void *A, const int lda,
                  void *X, const int incX);
-void cblas_ctpmv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                 const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+void cblas_ctpmv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                 const CBLAS_TRANSPOSE TransA, const CBLAS_DIAG Diag,
                  const int N, const void *Ap, void *X, const int incX);
-void cblas_ctrsv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                 const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+void cblas_ctrsv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                 const CBLAS_TRANSPOSE TransA, const CBLAS_DIAG Diag,
                  const int N, const void *A, const int lda, void *X,
                  const int incX);
-void cblas_ctbsv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                 const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+void cblas_ctbsv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                 const CBLAS_TRANSPOSE TransA, const CBLAS_DIAG Diag,
                  const int N, const int K, const void *A, const int lda,
                  void *X, const int incX);
-void cblas_ctpsv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                 const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+void cblas_ctpsv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                 const CBLAS_TRANSPOSE TransA, const CBLAS_DIAG Diag,
                  const int N, const void *Ap, void *X, const int incX);
 
-void cblas_zgemv(const enum CBLAS_ORDER Order,
-                 const enum CBLAS_TRANSPOSE TransA, const int M, const int N,
+void cblas_zgemv(const CBLAS_ORDER Order,
+                 const CBLAS_TRANSPOSE TransA, const int M, const int N,
                  const void *alpha, const void *A, const int lda,
                  const void *X, const int incX, const void *beta,
                  void *Y, const int incY);
-void cblas_zgbmv(const enum CBLAS_ORDER Order,
-                 const enum CBLAS_TRANSPOSE TransA, const int M, const int N,
+void cblas_zgbmv(const CBLAS_ORDER Order,
+                 const CBLAS_TRANSPOSE TransA, const int M, const int N,
                  const int KL, const int KU, const void *alpha,
                  const void *A, const int lda, const void *X,
                  const int incX, const void *beta, void *Y, const int incY);
-void cblas_ztrmv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                 const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+void cblas_ztrmv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                 const CBLAS_TRANSPOSE TransA, const CBLAS_DIAG Diag,
                  const int N, const void *A, const int lda,
                  void *X, const int incX);
-void cblas_ztbmv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                 const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+void cblas_ztbmv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                 const CBLAS_TRANSPOSE TransA, const CBLAS_DIAG Diag,
                  const int N, const int K, const void *A, const int lda,
                  void *X, const int incX);
-void cblas_ztpmv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                 const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+void cblas_ztpmv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                 const CBLAS_TRANSPOSE TransA, const CBLAS_DIAG Diag,
                  const int N, const void *Ap, void *X, const int incX);
-void cblas_ztrsv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                 const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+void cblas_ztrsv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                 const CBLAS_TRANSPOSE TransA, const CBLAS_DIAG Diag,
                  const int N, const void *A, const int lda, void *X,
                  const int incX);
-void cblas_ztbsv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                 const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+void cblas_ztbsv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                 const CBLAS_TRANSPOSE TransA, const CBLAS_DIAG Diag,
                  const int N, const int K, const void *A, const int lda,
                  void *X, const int incX);
-void cblas_ztpsv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                 const enum CBLAS_TRANSPOSE TransA, const enum CBLAS_DIAG Diag,
+void cblas_ztpsv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                 const CBLAS_TRANSPOSE TransA, const CBLAS_DIAG Diag,
                  const int N, const void *Ap, void *X, const int incX);
 
 
 /*
  * Routines with S and D prefixes only
  */
-void cblas_ssymv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
+void cblas_ssymv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
                  const int N, const float alpha, const float *A,
                  const int lda, const float *X, const int incX,
                  const float beta, float *Y, const int incY);
-void cblas_ssbmv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
+void cblas_ssbmv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
                  const int N, const int K, const float alpha, const float *A,
                  const int lda, const float *X, const int incX,
                  const float beta, float *Y, const int incY);
-void cblas_sspmv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
+void cblas_sspmv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
                  const int N, const float alpha, const float *Ap,
                  const float *X, const int incX,
                  const float beta, float *Y, const int incY);
-void cblas_sger(const enum CBLAS_ORDER Order, const int M, const int N,
+void cblas_sger(const CBLAS_ORDER Order, const int M, const int N,
                 const float alpha, const float *X, const int incX,
                 const float *Y, const int incY, float *A, const int lda);
-void cblas_ssyr(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
+void cblas_ssyr(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
                 const int N, const float alpha, const float *X,
                 const int incX, float *A, const int lda);
-void cblas_sspr(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
+void cblas_sspr(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
                 const int N, const float alpha, const float *X,
                 const int incX, float *Ap);
-void cblas_ssyr2(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
+void cblas_ssyr2(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
                 const int N, const float alpha, const float *X,
                 const int incX, const float *Y, const int incY, float *A,
                 const int lda);
-void cblas_sspr2(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
+void cblas_sspr2(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
                 const int N, const float alpha, const float *X,
                 const int incX, const float *Y, const int incY, float *A);
 
-void cblas_dsymv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
+void cblas_dsymv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
                  const int N, const double alpha, const double *A,
                  const int lda, const double *X, const int incX,
                  const double beta, double *Y, const int incY);
-void cblas_dsbmv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
+void cblas_dsbmv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
                  const int N, const int K, const double alpha, const double *A,
                  const int lda, const double *X, const int incX,
                  const double beta, double *Y, const int incY);
-void cblas_dspmv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
+void cblas_dspmv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
                  const int N, const double alpha, const double *Ap,
                  const double *X, const int incX,
                  const double beta, double *Y, const int incY);
-void cblas_dger(const enum CBLAS_ORDER Order, const int M, const int N,
+void cblas_dger(const CBLAS_ORDER Order, const int M, const int N,
                 const double alpha, const double *X, const int incX,
                 const double *Y, const int incY, double *A, const int lda);
-void cblas_dsyr(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
+void cblas_dsyr(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
                 const int N, const double alpha, const double *X,
                 const int incX, double *A, const int lda);
-void cblas_dspr(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
+void cblas_dspr(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
                 const int N, const double alpha, const double *X,
                 const int incX, double *Ap);
-void cblas_dsyr2(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
+void cblas_dsyr2(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
                 const int N, const double alpha, const double *X,
                 const int incX, const double *Y, const int incY, double *A,
                 const int lda);
-void cblas_dspr2(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
+void cblas_dspr2(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
                 const int N, const double alpha, const double *X,
                 const int incX, const double *Y, const int incY, double *A);
 
@@ -389,65 +389,65 @@ void cblas_dspr2(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
 /*
  * Routines with C and Z prefixes only
  */
-void cblas_chemv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
+void cblas_chemv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
                  const int N, const void *alpha, const void *A,
                  const int lda, const void *X, const int incX,
                  const void *beta, void *Y, const int incY);
-void cblas_chbmv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
+void cblas_chbmv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
                  const int N, const int K, const void *alpha, const void *A,
                  const int lda, const void *X, const int incX,
                  const void *beta, void *Y, const int incY);
-void cblas_chpmv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
+void cblas_chpmv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
                  const int N, const void *alpha, const void *Ap,
                  const void *X, const int incX,
                  const void *beta, void *Y, const int incY);
-void cblas_cgeru(const enum CBLAS_ORDER Order, const int M, const int N,
+void cblas_cgeru(const CBLAS_ORDER Order, const int M, const int N,
                  const void *alpha, const void *X, const int incX,
                  const void *Y, const int incY, void *A, const int lda);
-void cblas_cgerc(const enum CBLAS_ORDER Order, const int M, const int N,
+void cblas_cgerc(const CBLAS_ORDER Order, const int M, const int N,
                  const void *alpha, const void *X, const int incX,
                  const void *Y, const int incY, void *A, const int lda);
-void cblas_cher(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
+void cblas_cher(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
                 const int N, const float alpha, const void *X, const int incX,
                 void *A, const int lda);
-void cblas_chpr(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
+void cblas_chpr(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
                 const int N, const float alpha, const void *X,
                 const int incX, void *A);
-void cblas_cher2(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, const int N,
+void cblas_cher2(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo, const int N,
                 const void *alpha, const void *X, const int incX,
                 const void *Y, const int incY, void *A, const int lda);
-void cblas_chpr2(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, const int N,
+void cblas_chpr2(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo, const int N,
                 const void *alpha, const void *X, const int incX,
                 const void *Y, const int incY, void *Ap);
 
-void cblas_zhemv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
+void cblas_zhemv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
                  const int N, const void *alpha, const void *A,
                  const int lda, const void *X, const int incX,
                  const void *beta, void *Y, const int incY);
-void cblas_zhbmv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
+void cblas_zhbmv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
                  const int N, const int K, const void *alpha, const void *A,
                  const int lda, const void *X, const int incX,
                  const void *beta, void *Y, const int incY);
-void cblas_zhpmv(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
+void cblas_zhpmv(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
                  const int N, const void *alpha, const void *Ap,
                  const void *X, const int incX,
                  const void *beta, void *Y, const int incY);
-void cblas_zgeru(const enum CBLAS_ORDER Order, const int M, const int N,
+void cblas_zgeru(const CBLAS_ORDER Order, const int M, const int N,
                  const void *alpha, const void *X, const int incX,
                  const void *Y, const int incY, void *A, const int lda);
-void cblas_zgerc(const enum CBLAS_ORDER Order, const int M, const int N,
+void cblas_zgerc(const CBLAS_ORDER Order, const int M, const int N,
                  const void *alpha, const void *X, const int incX,
                  const void *Y, const int incY, void *A, const int lda);
-void cblas_zher(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
+void cblas_zher(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
                 const int N, const double alpha, const void *X, const int incX,
                 void *A, const int lda);
-void cblas_zhpr(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
+void cblas_zhpr(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
                 const int N, const double alpha, const void *X,
                 const int incX, void *A);
-void cblas_zher2(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, const int N,
+void cblas_zher2(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo, const int N,
                 const void *alpha, const void *X, const int incX,
                 const void *Y, const int incY, void *A, const int lda);
-void cblas_zhpr2(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, const int N,
+void cblas_zhpr2(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo, const int N,
                 const void *alpha, const void *X, const int incX,
                 const void *Y, const int incY, void *Ap);
 
@@ -460,123 +460,123 @@ void cblas_zhpr2(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo, const
 /*
  * Routines with standard 4 prefixes (S, D, C, Z)
  */
-void cblas_sgemm(const enum CBLAS_ORDER Order, const enum CBLAS_TRANSPOSE TransA,
-                 const enum CBLAS_TRANSPOSE TransB, const int M, const int N,
+void cblas_sgemm(const CBLAS_ORDER Order, const CBLAS_TRANSPOSE TransA,
+                 const CBLAS_TRANSPOSE TransB, const int M, const int N,
                  const int K, const float alpha, const float *A,
                  const int lda, const float *B, const int ldb,
                  const float beta, float *C, const int ldc);
-void cblas_ssymm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side,
-                 const enum CBLAS_UPLO Uplo, const int M, const int N,
+void cblas_ssymm(const CBLAS_ORDER Order, const CBLAS_SIDE Side,
+                 const CBLAS_UPLO Uplo, const int M, const int N,
                  const float alpha, const float *A, const int lda,
                  const float *B, const int ldb, const float beta,
                  float *C, const int ldc);
-void cblas_ssyrk(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                 const enum CBLAS_TRANSPOSE Trans, const int N, const int K,
+void cblas_ssyrk(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                 const CBLAS_TRANSPOSE Trans, const int N, const int K,
                  const float alpha, const float *A, const int lda,
                  const float beta, float *C, const int ldc);
-void cblas_ssyr2k(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                  const enum CBLAS_TRANSPOSE Trans, const int N, const int K,
+void cblas_ssyr2k(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                  const CBLAS_TRANSPOSE Trans, const int N, const int K,
                   const float alpha, const float *A, const int lda,
                   const float *B, const int ldb, const float beta,
                   float *C, const int ldc);
-void cblas_strmm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side,
-                 const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA,
-                 const enum CBLAS_DIAG Diag, const int M, const int N,
+void cblas_strmm(const CBLAS_ORDER Order, const CBLAS_SIDE Side,
+                 const CBLAS_UPLO Uplo, const CBLAS_TRANSPOSE TransA,
+                 const CBLAS_DIAG Diag, const int M, const int N,
                  const float alpha, const float *A, const int lda,
                  float *B, const int ldb);
-void cblas_strsm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side,
-                 const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA,
-                 const enum CBLAS_DIAG Diag, const int M, const int N,
+void cblas_strsm(const CBLAS_ORDER Order, const CBLAS_SIDE Side,
+                 const CBLAS_UPLO Uplo, const CBLAS_TRANSPOSE TransA,
+                 const CBLAS_DIAG Diag, const int M, const int N,
                  const float alpha, const float *A, const int lda,
                  float *B, const int ldb);
 
-void cblas_dgemm(const enum CBLAS_ORDER Order, const enum CBLAS_TRANSPOSE TransA,
-                 const enum CBLAS_TRANSPOSE TransB, const int M, const int N,
+void cblas_dgemm(const CBLAS_ORDER Order, const CBLAS_TRANSPOSE TransA,
+                 const CBLAS_TRANSPOSE TransB, const int M, const int N,
                  const int K, const double alpha, const double *A,
                  const int lda, const double *B, const int ldb,
                  const double beta, double *C, const int ldc);
-void cblas_dsymm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side,
-                 const enum CBLAS_UPLO Uplo, const int M, const int N,
+void cblas_dsymm(const CBLAS_ORDER Order, const CBLAS_SIDE Side,
+                 const CBLAS_UPLO Uplo, const int M, const int N,
                  const double alpha, const double *A, const int lda,
                  const double *B, const int ldb, const double beta,
                  double *C, const int ldc);
-void cblas_dsyrk(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                 const enum CBLAS_TRANSPOSE Trans, const int N, const int K,
+void cblas_dsyrk(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                 const CBLAS_TRANSPOSE Trans, const int N, const int K,
                  const double alpha, const double *A, const int lda,
                  const double beta, double *C, const int ldc);
-void cblas_dsyr2k(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                  const enum CBLAS_TRANSPOSE Trans, const int N, const int K,
+void cblas_dsyr2k(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                  const CBLAS_TRANSPOSE Trans, const int N, const int K,
                   const double alpha, const double *A, const int lda,
                   const double *B, const int ldb, const double beta,
                   double *C, const int ldc);
-void cblas_dtrmm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side,
-                 const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA,
-                 const enum CBLAS_DIAG Diag, const int M, const int N,
+void cblas_dtrmm(const CBLAS_ORDER Order, const CBLAS_SIDE Side,
+                 const CBLAS_UPLO Uplo, const CBLAS_TRANSPOSE TransA,
+                 const CBLAS_DIAG Diag, const int M, const int N,
                  const double alpha, const double *A, const int lda,
                  double *B, const int ldb);
-void cblas_dtrsm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side,
-                 const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA,
-                 const enum CBLAS_DIAG Diag, const int M, const int N,
+void cblas_dtrsm(const CBLAS_ORDER Order, const CBLAS_SIDE Side,
+                 const CBLAS_UPLO Uplo, const CBLAS_TRANSPOSE TransA,
+                 const CBLAS_DIAG Diag, const int M, const int N,
                  const double alpha, const double *A, const int lda,
                  double *B, const int ldb);
 
-void cblas_cgemm(const enum CBLAS_ORDER Order, const enum CBLAS_TRANSPOSE TransA,
-                 const enum CBLAS_TRANSPOSE TransB, const int M, const int N,
+void cblas_cgemm(const CBLAS_ORDER Order, const CBLAS_TRANSPOSE TransA,
+                 const CBLAS_TRANSPOSE TransB, const int M, const int N,
                  const int K, const void *alpha, const void *A,
                  const int lda, const void *B, const int ldb,
                  const void *beta, void *C, const int ldc);
-void cblas_csymm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side,
-                 const enum CBLAS_UPLO Uplo, const int M, const int N,
+void cblas_csymm(const CBLAS_ORDER Order, const CBLAS_SIDE Side,
+                 const CBLAS_UPLO Uplo, const int M, const int N,
                  const void *alpha, const void *A, const int lda,
                  const void *B, const int ldb, const void *beta,
                  void *C, const int ldc);
-void cblas_csyrk(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                 const enum CBLAS_TRANSPOSE Trans, const int N, const int K,
+void cblas_csyrk(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                 const CBLAS_TRANSPOSE Trans, const int N, const int K,
                  const void *alpha, const void *A, const int lda,
                  const void *beta, void *C, const int ldc);
-void cblas_csyr2k(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                  const enum CBLAS_TRANSPOSE Trans, const int N, const int K,
+void cblas_csyr2k(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                  const CBLAS_TRANSPOSE Trans, const int N, const int K,
                   const void *alpha, const void *A, const int lda,
                   const void *B, const int ldb, const void *beta,
                   void *C, const int ldc);
-void cblas_ctrmm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side,
-                 const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA,
-                 const enum CBLAS_DIAG Diag, const int M, const int N,
+void cblas_ctrmm(const CBLAS_ORDER Order, const CBLAS_SIDE Side,
+                 const CBLAS_UPLO Uplo, const CBLAS_TRANSPOSE TransA,
+                 const CBLAS_DIAG Diag, const int M, const int N,
                  const void *alpha, const void *A, const int lda,
                  void *B, const int ldb);
-void cblas_ctrsm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side,
-                 const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA,
-                 const enum CBLAS_DIAG Diag, const int M, const int N,
+void cblas_ctrsm(const CBLAS_ORDER Order, const CBLAS_SIDE Side,
+                 const CBLAS_UPLO Uplo, const CBLAS_TRANSPOSE TransA,
+                 const CBLAS_DIAG Diag, const int M, const int N,
                  const void *alpha, const void *A, const int lda,
                  void *B, const int ldb);
 
-void cblas_zgemm(const enum CBLAS_ORDER Order, const enum CBLAS_TRANSPOSE TransA,
-                 const enum CBLAS_TRANSPOSE TransB, const int M, const int N,
+void cblas_zgemm(const CBLAS_ORDER Order, const CBLAS_TRANSPOSE TransA,
+                 const CBLAS_TRANSPOSE TransB, const int M, const int N,
                  const int K, const void *alpha, const void *A,
                  const int lda, const void *B, const int ldb,
                  const void *beta, void *C, const int ldc);
-void cblas_zsymm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side,
-                 const enum CBLAS_UPLO Uplo, const int M, const int N,
+void cblas_zsymm(const CBLAS_ORDER Order, const CBLAS_SIDE Side,
+                 const CBLAS_UPLO Uplo, const int M, const int N,
                  const void *alpha, const void *A, const int lda,
                  const void *B, const int ldb, const void *beta,
                  void *C, const int ldc);
-void cblas_zsyrk(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                 const enum CBLAS_TRANSPOSE Trans, const int N, const int K,
+void cblas_zsyrk(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                 const CBLAS_TRANSPOSE Trans, const int N, const int K,
                  const void *alpha, const void *A, const int lda,
                  const void *beta, void *C, const int ldc);
-void cblas_zsyr2k(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                  const enum CBLAS_TRANSPOSE Trans, const int N, const int K,
+void cblas_zsyr2k(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                  const CBLAS_TRANSPOSE Trans, const int N, const int K,
                   const void *alpha, const void *A, const int lda,
                   const void *B, const int ldb, const void *beta,
                   void *C, const int ldc);
-void cblas_ztrmm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side,
-                 const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA,
-                 const enum CBLAS_DIAG Diag, const int M, const int N,
+void cblas_ztrmm(const CBLAS_ORDER Order, const CBLAS_SIDE Side,
+                 const CBLAS_UPLO Uplo, const CBLAS_TRANSPOSE TransA,
+                 const CBLAS_DIAG Diag, const int M, const int N,
                  const void *alpha, const void *A, const int lda,
                  void *B, const int ldb);
-void cblas_ztrsm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side,
-                 const enum CBLAS_UPLO Uplo, const enum CBLAS_TRANSPOSE TransA,
-                 const enum CBLAS_DIAG Diag, const int M, const int N,
+void cblas_ztrsm(const CBLAS_ORDER Order, const CBLAS_SIDE Side,
+                 const CBLAS_UPLO Uplo, const CBLAS_TRANSPOSE TransA,
+                 const CBLAS_DIAG Diag, const int M, const int N,
                  const void *alpha, const void *A, const int lda,
                  void *B, const int ldb);
 
@@ -584,31 +584,31 @@ void cblas_ztrsm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side,
 /*
  * Routines with prefixes C and Z only
  */
-void cblas_chemm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side,
-                 const enum CBLAS_UPLO Uplo, const int M, const int N,
+void cblas_chemm(const CBLAS_ORDER Order, const CBLAS_SIDE Side,
+                 const CBLAS_UPLO Uplo, const int M, const int N,
                  const void *alpha, const void *A, const int lda,
                  const void *B, const int ldb, const void *beta,
                  void *C, const int ldc);
-void cblas_cherk(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                 const enum CBLAS_TRANSPOSE Trans, const int N, const int K,
+void cblas_cherk(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                 const CBLAS_TRANSPOSE Trans, const int N, const int K,
                  const float alpha, const void *A, const int lda,
                  const float beta, void *C, const int ldc);
-void cblas_cher2k(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                  const enum CBLAS_TRANSPOSE Trans, const int N, const int K,
+void cblas_cher2k(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                  const CBLAS_TRANSPOSE Trans, const int N, const int K,
                   const void *alpha, const void *A, const int lda,
                   const void *B, const int ldb, const float beta,
                   void *C, const int ldc);
-void cblas_zhemm(const enum CBLAS_ORDER Order, const enum CBLAS_SIDE Side,
-                 const enum CBLAS_UPLO Uplo, const int M, const int N,
+void cblas_zhemm(const CBLAS_ORDER Order, const CBLAS_SIDE Side,
+                 const CBLAS_UPLO Uplo, const int M, const int N,
                  const void *alpha, const void *A, const int lda,
                  const void *B, const int ldb, const void *beta,
                  void *C, const int ldc);
-void cblas_zherk(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                 const enum CBLAS_TRANSPOSE Trans, const int N, const int K,
+void cblas_zherk(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                 const CBLAS_TRANSPOSE Trans, const int N, const int K,
                  const double alpha, const void *A, const int lda,
                  const double beta, void *C, const int ldc);
-void cblas_zher2k(const enum CBLAS_ORDER Order, const enum CBLAS_UPLO Uplo,
-                  const enum CBLAS_TRANSPOSE Trans, const int N, const int K,
+void cblas_zher2k(const CBLAS_ORDER Order, const CBLAS_UPLO Uplo,
+                  const CBLAS_TRANSPOSE Trans, const int N, const int K,
                   const void *alpha, const void *A, const int lda,
                   const void *B, const int ldb, const double beta,
                   void *C, const int ldc);
@@ -678,11 +678,15 @@ int cblas_errprn(int ierr, int info, char *form, ...);
 #define NONCLAPACK_H
 
 #ifndef NONCBLAS
-#include <cblas.h>
+#ifdef HAS_MKL
+#include "mkl_cblas.h"
+#else
+#include "cblas.h"
+#endif
 #endif
 
 /*from clapack.h*/
-int clapack_dgesv(const enum CBLAS_ORDER Order, const int N, const int NRHS,
+int clapack_dgesv(const  CBLAS_ORDER Order, const int N, const int NRHS,
                   double *A, const int lda, int *ipiv,
                   double *B, const int ldb); 
 #endif
diff --git a/nonclass.cc b/nonclass.cc
index 8cbccad..a658401 100644
--- a/nonclass.cc
+++ b/nonclass.cc
@@ -1651,6 +1651,11 @@ double r=0;
 return r;
 }
 
+inline double trace2(const NRMat<double> &a, const NRSMat<double> &b, const bool diagscaled)
+{
+return trace2(b,a,diagscaled);
+}
+
 
 //Cholesky interface
 extern "C" void FORNAME(dpotrf)(const char *UPLO, const FINT *N, double *A, const FINT *LDA, FINT *INFO);
diff --git a/nonclass.h b/nonclass.h
index f9fc244..4e122ad 100644
--- a/nonclass.h
+++ b/nonclass.h
@@ -88,14 +88,14 @@ extern const  NRVec<T> diagofproduct(const NRMat<T> &a, const NRMat<T> &b,\
 extern T trace2(const NRMat<T> &a, const NRMat<T> &b, bool trb=0); \
 extern T trace2(const NRSMat<T> &a, const NRSMat<T> &b, const bool diagscaled=0);\
 extern T trace2(const NRSMat<T> &a, const NRMat<T> &b, const bool diagscaled=0);\
+extern T trace2(const NRMat<T> &a, const NRSMat<T> &b, const bool diagscaled=0);\
 extern void linear_solve(NRMat<T> &a, NRMat<T> *b, T *det=0,int n=0); /*solve Ax^T=b^T (b is nrhs x n) */ \
 extern void linear_solve(NRSMat<T> &a, NRMat<T> *b, T *det=0, int n=0); /*solve Ax^T=b^T (b is nrhs x n) */\
 extern void linear_solve(NRMat<T> &a, NRVec<T> &b, double *det=0, int n=0); \
 extern void linear_solve(NRSMat<T> &a, NRVec<T> &b, double *det=0, int n=0); \
 extern void diagonalize(NRMat<T> &a, NRVec<LA_traits<T>::normtype> &w, const bool eivec=1, const bool corder=1, int n=0, NRMat<T> *b=NULL, const int itype=1); \
 extern void diagonalize(NRSMat<T> &a, NRVec<LA_traits<T>::normtype> &w, NRMat<T> *v, const bool corder=1, int n=0, NRSMat<T> *b=NULL, const int itype=1);\
-extern void singular_decomposition(NRMat<T> &a, NRMat<T> *u, NRVec<T> &s,\
-		NRMat<T> *v, const bool corder=1, int m=0, int n=0);
+extern void singular_decomposition(NRMat<T> &a, NRMat<T> *u, NRVec<LA_traits<T>::normtype> &s, NRMat<T> *v, const bool vnotdagger=0, int m=0, int n=0);
 
 /*NOTE!!! all versions of diagonalize DESTROY A and generalized diagonalize also B matrix */
 
diff --git a/t.cc b/t.cc
index 33ca48a..651841f 100644
--- a/t.cc
+++ b/t.cc
@@ -352,18 +352,33 @@ if(0)
 {
 NRMat<double> a;
 cin >>a ;
-NRMat<double> b=a.transpose();
+NRMat<double> abak=a;
 NRMat<double> u(a.nrows(),a.nrows()),v(a.ncols(),a.ncols());
-NRVec<double>s(a.ncols());
-singular_decomposition(a,&u,s,&v);
-//singular_decomposition(a,NULL,s,NULL); //this does not work when linked with static version of lapack, works with .so.3 version (from suse distrib)
+NRVec<double>s(a.ncols()<a.nrows()?a.ncols():a.nrows());
+singular_decomposition(a,&u,s,&v,0);
 cout <<u;
-cout <<s;
+NRMat<double> sdiag(0., u.ncols(),v.nrows());
+sdiag.diagonalset(s);
+cout <<sdiag;
 cout <<v;
-//singular_decomposition(b,&v,s,&u);
-//cout <<v;
-//cout <<s;
-//cout <<u;
+cout << "Error "<<(u*sdiag*v-abak).norm()<<endl;
+}
+
+if(0)
+{
+NRMat<complex<double> > a;
+cin >>a ;
+NRMat<complex<double> > abak=a;
+NRMat<complex<double> > u(a.nrows(),a.nrows()),v(a.ncols(),a.ncols());
+NRVec<double>s(a.ncols()<a.nrows()?a.ncols():a.nrows());
+singular_decomposition(a,&u,s,&v,0);
+cout <<u;
+NRMat<complex<double> > sdiag(0., u.ncols(),v.nrows());
+NRVec<complex<double> > ss = s;
+sdiag.diagonalset(ss);
+cout <<sdiag;
+cout <<v;
+cout << "Error "<<(u*sdiag*v-abak).norm()<<endl;
 }
 
 if(0)
@@ -1579,7 +1594,7 @@ cin >>n;
 NRMat<complex<double> > a(n,n);
 a.randomize(1);
 for(int i=0; i<n; ++i) for(int j=0; j<i; ++j) {a(i,j)=0.;}
-for(int i=0; i<n; ++i) {a(i,i).imag()=0.; if(a(i,i).real()<0) a(i,i).real() *= -1;}
+for(int i=0; i<n; ++i) {a(i,i).imag(0.); if(a(i,i).real()<0) a(i,i).real(-a(i,i).real());}
 cout <<a;
 NRMat<complex<double> > bb=a.transpose(true)*a;
 NRMat<complex<double> > cc(bb);
@@ -1642,7 +1657,7 @@ cin >>n;
 NRMat<complex<double> > a(n,n);
 a.randomize(1);
 for(int i=0; i<n; ++i) for(int j=0; j<i; ++j) {a(i,j)=0.;}
-for(int i=0; i<n; ++i) {a(i,i).imag() = 0.; if(a(i,i).real()<0) a(i,i).real() *= -10; else a(i,i).real() *= 10.;}
+for(int i=0; i<n; ++i) {a(i,i).imag(0.); if(a(i,i).real()<0) a(i,i).real(-10.*a(i,i).real()); else a(i,i).real(10.*a(i,i).real());}
 if(n<100)cout <<a;
 NRMat<complex<double> > bb=a.transpose(true)*a;
 SparseSMat<complex<double> > cc(bb);
@@ -1760,7 +1775,7 @@ cout <<"error = "<<(r2-NRMat<complex<double> >(r)).norm()<<endl;
 */
 
 
-if(1)
+if(0)
 {
 NRMat<complex<double> > m;
 ifstream f("libormat");
@@ -1825,6 +1840,23 @@ cout <<"eigenvalue error = "<<err<<endl;
 
 }
 
+if(0)
+{
+NRMat<double> a;
+cin >>a ;
+double det=determinant_destroy(a);
+cout << "det= "<<det<<endl;
+}
+
+if(1)
+{
+bitvector v(3);
+v.assign(0,0);
+v.assign(1,1);
+v.assign(2,0);
+cin >>v;
+cout <<v;
+}
 
 
 }