diff --git a/CMakeLists.txt b/CMakeLists.txt
index 80d834a..bbace04 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -8,6 +8,7 @@ ENDIF()
 
 PROJECT(openEMS CXX C)
 cmake_minimum_required(VERSION 2.8)
+cmake_policy(SET CMP0074 NEW)
 
 # default
 set(LIB_VERSION_MAJOR 0)
@@ -21,6 +22,8 @@ IF(EXISTS ${PROJECT_SOURCE_DIR}/localConfig.cmake)
    include(${PROJECT_SOURCE_DIR}/localConfig.cmake)
 ENDIF()
 
+#ADD_DEFINITIONS( -D__SSE2__ )
+
 set(VERSION "v0.0.35")
 
 # add git revision
@@ -106,11 +109,18 @@ INCLUDE_DIRECTORIES( ${CSXCAD_INCLUDE_DIR} )
 # TinyXML module from https://github.com/ros/cmake_modules
 find_package(TinyXML REQUIRED)
 ADD_DEFINITIONS( -DTIXML_USE_STL )
+message(STATUS "TinyXML_INCLUDE_DIR: ${TinyXML_INCLUDE_DIR}")
+message(STATUS "TinyXML_LIBRARY: ${TinyXML_LIBRARY}")
+
+INCLUDE_DIRECTORIES( ${TinyXML_INCLUDE_DIR} )
 
 # hdf5
 find_package(HDF5 1.8 COMPONENTS C HL REQUIRED)
 INCLUDE_DIRECTORIES (${HDF5_INCLUDE_DIRS})
 link_directories(${HDF5_LIBRARIES})
+# hdf5 compat
+#ADD_DEFINITIONS( -DH5_USE_16_API )
+ADD_DEFINITIONS( -DH5_BUILT_AS_DYNAMIC_LIB )
 
 # boost
 find_package(Boost 1.46 COMPONENTS
@@ -119,7 +129,11 @@ find_package(Boost 1.46 COMPONENTS
   date_time
   serialization
   chrono
+  REQUIRED
 )
+message(STATUS "Boost_INCLUDE_DIR: ${Boost_INCLUDE_DIR}")
+message(STATUS "Boost_LIBRARIES: ${Boost_LIBRARIES}")
+INCLUDE_DIRECTORIES (${Boost_INCLUDE_DIR})
 
 # vtk
 find_package(VTK COMPONENTS vtkIOXML vtkIOGeometry vtkIOLegacy vtkIOPLY NO_MODULE REQUIRED)
@@ -193,11 +207,12 @@ if (${MPI_CXX_FOUND})
 endif()
 
 add_library( openEMS SHARED ${SOURCES})
-set_target_properties(openEMS PROPERTIES VERSION ${LIB_VERSION_STRING} SOVERSION ${LIB_VERSION_MAJOR})
+#ADD_EXECUTABLE( openEMS main.cpp ${SOURCES})
+set_target_properties(openEMS PROPERTIES VERSION ${LIB_VERSION_STRING} SOVERSION ${LIB_VERSION_MAJOR} )
 TARGET_LINK_LIBRARIES( openEMS
   ${CSXCAD_LIBRARIES}
   ${fparser_LIBRARIES}
-  tinyxml
+  ${TinyXML_LIBRARY}
   ${HDF5_LIBRARIES}
   ${HDF5_HL_LIBRARIES}
   ${Boost_LIBRARIES}
@@ -205,13 +220,22 @@ TARGET_LINK_LIBRARIES( openEMS
   ${MPI_LIBRARIES}
 )
 
+if (WIN32)
+    # make sure "BUILD_OPENEMS_LIB" is only set for openEMS (dll) not openEMS_bin (exe)
+    target_compile_definitions(openEMS PRIVATE -DBUILD_OPENEMS_LIB )
+endif (WIN32)
+
 # main program
 ADD_EXECUTABLE( openEMS_bin main.cpp )
 SET_TARGET_PROPERTIES(openEMS_bin PROPERTIES OUTPUT_NAME openEMS)
 TARGET_LINK_LIBRARIES(openEMS_bin openEMS)
  
-INSTALL(TARGETS openEMS DESTINATION lib${LIB_SUFFIX})
-INSTALL(TARGETS openEMS_bin DESTINATION bin )
+if (WIN32)
+    INSTALL(TARGETS openEMS DESTINATION bin)
+else()
+    INSTALL(TARGETS openEMS DESTINATION lib${LIB_SUFFIX})
+endif()
+INSTALL(TARGETS openEMS_bin DESTINATION bin)
 
 if (UNIX)
     INSTALL( FILES openEMS.sh 
diff --git a/Common/processing.cpp b/Common/processing.cpp
index a444d9d..9e291ec 100644
--- a/Common/processing.cpp
+++ b/Common/processing.cpp
@@ -203,6 +203,11 @@ void Processing::ShowSnappedCoords()
 		 << stop[0] << "," << stop[1] << "," << stop[2] << "]" << endl;
 }
 
+void Processing::SetProcessInterval(unsigned int interval)
+{
+	ProcessInterval=std::max((unsigned int)1,interval);
+}
+
 void Processing::SetProcessStartStopTime(double start, double stop)
 {
 	double dT = Op->GetTimestep();
diff --git a/Common/processing.h b/Common/processing.h
index 2042706..ab20f78 100644
--- a/Common/processing.h
+++ b/Common/processing.h
@@ -18,6 +18,10 @@
 #ifndef PROCESSING_H
 #define PROCESSING_H
 
+#ifndef __GNUC__ // not GCC
+#include <emmintrin.h>
+#endif
+
 #include <complex>
 typedef std::complex<double> double_complex;
 #define _I double_complex(0.0,1.0)
@@ -31,6 +35,8 @@ typedef std::complex<double> double_complex;
 #include <string>
 #include <vector>
 
+#define _USE_MATH_DEFINES
+
 #include "Common/engine_interface_base.h"
 
 class Operator_Base;
@@ -59,7 +65,7 @@ public:
 
 	virtual void ShowSnappedCoords();
 
-	void SetProcessInterval(unsigned int interval) {ProcessInterval=std::max((unsigned int)1,interval);}
+	void SetProcessInterval(unsigned int interval);
 	void SetProcessStartStopTime(double start, double stop);
 
 	void AddStep(unsigned int step);
diff --git a/FDTD/engine_multithread.h b/FDTD/engine_multithread.h
index 20e7073..dce3b48 100644
--- a/FDTD/engine_multithread.h
+++ b/FDTD/engine_multithread.h
@@ -26,11 +26,13 @@
 #include <boost/fusion/container/list/list_fwd.hpp>
 #include <boost/fusion/include/list_fwd.hpp>
 
-//#ifdef WIN32
-//#include <Winsock2.h> // for struct timeval
-//#endif
-
+#include "tools/useful.h"
+#ifndef __GNUC__
+#include <Winsock2.h> // for struct timeval
+#else
 #include <sys/time.h>
+#endif
+
 
 #ifdef MPI_SUPPORT
 	#define ENGINE_MULTITHREAD_BASE Engine_MPI
diff --git a/nf2ff/CMakeLists.txt b/nf2ff/CMakeLists.txt
index 86811b4..ecc8a27 100644
--- a/nf2ff/CMakeLists.txt
+++ b/nf2ff/CMakeLists.txt
@@ -33,9 +33,12 @@ set(HEADERS
 
 add_library( nf2ff SHARED ${SOURCES})
 set_target_properties(nf2ff PROPERTIES VERSION ${LIB_VERSION_STRING} SOVERSION ${LIB_VERSION_MAJOR})
+if (WIN32)
+    target_compile_definitions(nf2ff PRIVATE -DBUILD_NF2FF_LIB )
+endif (WIN32)
 
 TARGET_LINK_LIBRARIES( nf2ff
-  tinyxml
+  ${TinyXML_LIBRARY}
   ${HDF5_LIBRARIES}
   ${Boost_LIBRARIES}
   ${MPI_LIBRARIES}
@@ -46,7 +49,11 @@ SET_TARGET_PROPERTIES(nf2ff_bin PROPERTIES OUTPUT_NAME nf2ff)
 TARGET_LINK_LIBRARIES(nf2ff_bin nf2ff)
 
 INSTALL(TARGETS nf2ff_bin DESTINATION bin)
-INSTALL(TARGETS nf2ff     DESTINATION lib${LIB_SUFFIX})
+if (WIN32)
+    INSTALL(TARGETS nf2ff     DESTINATION bin)
+else()
+    INSTALL(TARGETS nf2ff     DESTINATION lib${LIB_SUFFIX})
+endif()
 
 INSTALL(FILES ${HEADERS} DESTINATION include/openEMS)
 
diff --git a/nf2ff/nf2ff.h b/nf2ff/nf2ff.h
index 06823af..634989b 100644
--- a/nf2ff/nf2ff.h
+++ b/nf2ff/nf2ff.h
@@ -27,9 +27,20 @@
 
 using namespace std;
 
+#if defined(WIN32)
+	#ifdef BUILD_NF2FF_LIB
+	#define NF2FF_EXPORT __declspec(dllexport)
+	#else
+	#define NF2FF_EXPORT __declspec(dllimport)
+	#endif
+#else
+#define NF2FF_EXPORT
+#endif
+
+
 class TiXmlElement;
 
-class nf2ff
+class NF2FF_EXPORT nf2ff
 {
 public:
 	nf2ff(vector<float> freq, vector<float> theta, vector<float> phi, vector<float> center, unsigned int numThreads=0);
diff --git a/nf2ff/nf2ff_calc.cpp b/nf2ff/nf2ff_calc.cpp
index 9c2aa4f..a5ede5e 100644
--- a/nf2ff/nf2ff_calc.cpp
+++ b/nf2ff/nf2ff_calc.cpp
@@ -376,13 +376,13 @@ bool nf2ff_calc::AddSinglePlane(float **lines, unsigned int* numLines, complex<f
 		normDir[ny]=-1;
 	unsigned int pos[3];
 
-	float edge_length_P[numLines[nP]];
+	float* edge_length_P = new float[numLines[nP]];
 	for (unsigned int n=1;n<numLines[nP]-1;++n)
 		edge_length_P[n]=0.5*fabs(lines[nP][n+1]-lines[nP][n-1]);
 	edge_length_P[0]=0.5*fabs(lines[nP][1]-lines[nP][0]);
 	edge_length_P[numLines[nP]-1]=0.5*fabs(lines[nP][numLines[nP]-1]-lines[nP][numLines[nP]-2]);
 
-	float edge_length_PP[numLines[nPP]];
+	float* edge_length_PP = new float[numLines[nPP]];
 	for (unsigned int n=1;n<numLines[nPP]-1;++n)
 		edge_length_PP[n]=0.5*fabs(lines[nPP][n+1]-lines[nPP][n-1]);
 	edge_length_PP[0]=0.5*fabs(lines[nPP][1]-lines[nPP][0]);
@@ -422,7 +422,7 @@ bool nf2ff_calc::AddSinglePlane(float **lines, unsigned int* numLines, complex<f
 	// setup multi-threading jobs
 	vector<unsigned int> jpt = AssignJobs2Threads(numLines[nP], m_numThreads, true);
 	m_numThreads = jpt.size();
-	nf2ff_data thread_data[m_numThreads];
+	nf2ff_data* thread_data = new nf2ff_data[m_numThreads];
 	m_Barrier = new boost::barrier(m_numThreads+1); // numThread workers + 1 controller
 	unsigned int start=0;
 	unsigned int stop=jpt.at(0)-1;
@@ -518,6 +518,9 @@ bool nf2ff_calc::AddSinglePlane(float **lines, unsigned int* numLines, complex<f
 	Delete2DArray(Np,numAngles);
 	Delete2DArray(Lt,numAngles);
 	Delete2DArray(Lp,numAngles);
+	delete[] edge_length_P; edge_length_P=NULL;
+	delete[] edge_length_PP; edge_length_PP=NULL;
+	delete[] thread_data; thread_data=NULL;
 
 	m_maxDir = 4*M_PI*P_max / m_radPower;
 
diff --git a/nf2ff/nf2ff_calc.h b/nf2ff/nf2ff_calc.h
index ab46e5e..ff8c999 100644
--- a/nf2ff/nf2ff_calc.h
+++ b/nf2ff/nf2ff_calc.h
@@ -24,6 +24,7 @@
 #include <cmath>
 #include <complex>
 #include <boost/thread.hpp>
+#define _USE_MATH_DEFINES
 
 class nf2ff_calc;
 
diff --git a/openems.h b/openems.h
index 6600e06..2de8118 100644
--- a/openems.h
+++ b/openems.h
@@ -19,10 +19,16 @@
 #define OPENEMS_H
 
 #include <sstream>
+#ifdef WIN32
+#include <Winsock2.h> // for struct timeval
+#else
 #include <sys/time.h>
+#endif
 #include <time.h>
 #include <vector>
 
+#include "tools/global.h"
+
 #define __OPENEMS_STAT_FILE__ "openEMS_stats.txt"
 #define __OPENEMS_RUN_STAT_FILE__ "openEMS_run_stats.txt"
 
@@ -39,7 +45,7 @@ class Engine_Ext_SteadyState;
 double CalcDiffTime(timeval t1, timeval t2);
 std::string FormatTime(int sec);
 
-class openEMS
+class OPENEMS_EXPORT openEMS
 {
 public:
 	openEMS();
diff --git a/tools/array_ops.cpp b/tools/array_ops.cpp
index 8710b3c..0c50618 100644
--- a/tools/array_ops.cpp
+++ b/tools/array_ops.cpp
@@ -21,11 +21,9 @@
 using namespace std;
 
 #ifdef WIN32
-#define __MSVCRT_VERSION__ 0x0700
 #include <malloc.h>
-//(void**)&array, 16, sizeof(typeof(f4vector**))*numLines[0]
-#define MEMALIGN( array, alignment, size ) !(*array = _aligned_malloc( size, alignment ))
-#define FREE( array ) _aligned_free( array )
+#define MEMALIGN( array, alignment, size ) !(*array = _mm_malloc( size, alignment ))
+#define FREE( array ) _mm_free( array )
 #else
 #define MEMALIGN( array, alignment, size ) posix_memalign( array, alignment, size )
 #define FREE( array ) free( array )
@@ -70,7 +68,7 @@ void Delete_N_3DArray_v4sf(f4vector**** array, const unsigned int* numLines)
 f4vector* Create1DArray_v4sf(const unsigned int numLines)
 {
 	f4vector* array=NULL;
-	if (MEMALIGN( (void**)&array, 16, sizeof(typeof(f4vector))*numLines ))
+	if (MEMALIGN( (void**)&array, 16, F4VECTOR_SIZE*numLines ))
 	{
 		cerr << "cannot allocate aligned memory" << endl;
 		exit(3);
@@ -92,7 +90,7 @@ f4vector*** Create3DArray_v4sf(const unsigned int* numLines)
 
 	f4vector*** array=NULL;
 	unsigned int pos[3];
-	if (MEMALIGN( (void**)&array, 16, sizeof(typeof(f4vector**))*numLines[0] ))
+	if (MEMALIGN( (void**)&array, 16, F4VECTOR_SIZE*numLines[0] ))
 	{
 		cerr << "cannot allocate aligned memory" << endl;
 		exit(3);
@@ -100,7 +98,7 @@ f4vector*** Create3DArray_v4sf(const unsigned int* numLines)
 	//array = new f4vector**[numLines[0]];
 	for (pos[0]=0; pos[0]<numLines[0]; ++pos[0])
 	{
-		if (MEMALIGN( (void**)&array[pos[0]], 16, sizeof(typeof(f4vector*))*numLines[1] ))
+		if (MEMALIGN( (void**)&array[pos[0]], 16, F4VECTOR_SIZE*numLines[1] ))
 		{
 			cerr << "cannot allocate aligned memory" << endl;
 			exit(3);
@@ -108,7 +106,7 @@ f4vector*** Create3DArray_v4sf(const unsigned int* numLines)
 		//array[pos[0]] = new f4vector*[numLines[1]];
 		for (pos[1]=0; pos[1]<numLines[1]; ++pos[1])
 		{
-			if (MEMALIGN( (void**)&array[pos[0]][pos[1]], 16, sizeof(typeof(f4vector))*numZ ))
+			if (MEMALIGN( (void**)&array[pos[0]][pos[1]], 16, F4VECTOR_SIZE*numZ ))
 			{
 				cerr << "cannot allocate aligned memory" << endl;
 				exit(3);
@@ -129,7 +127,7 @@ f4vector*** Create3DArray_v4sf(const unsigned int* numLines)
 f4vector**** Create_N_3DArray_v4sf(const unsigned int* numLines)
 {
 	f4vector**** array=NULL;
-	if (MEMALIGN( (void**)&array, 16, sizeof(typeof(f4vector***))*3 ))
+	if (MEMALIGN( (void**)&array, 16, F4VECTOR_SIZE*3 ))
 	{
 		cerr << "cannot allocate aligned memory" << endl;
 		exit(3);
diff --git a/tools/array_ops.h b/tools/array_ops.h
index 5c2fd7f..60153d4 100644
--- a/tools/array_ops.h
+++ b/tools/array_ops.h
@@ -1,5 +1,5 @@
 /*
-*	Copyright (C) 2010 Thorsten Liebig (Thorsten.Liebig@gmx.de)
+*	Copyright (C) 2010,2019 Thorsten Liebig (Thorsten.Liebig@gmx.de)
 *
 *	This program is free software: you can redistribute it and/or modify
 *	it under the terms of the GNU General Public License as published by
@@ -31,14 +31,32 @@
 #include <math.h>
 #include "constants.h"
 
-typedef float v4sf __attribute__ ((vector_size (16))); // vector of four single floats
-typedef   int v4si __attribute__ ((vector_size (4*sizeof(int)))); // vector of four single ints
+#define F4VECTOR_SIZE 16 // sizeof(typeid(f4vector))
 
+#ifdef __GNUC__ // GCC
+typedef float v4sf __attribute__ ((vector_size (F4VECTOR_SIZE))); // vector of four single floats
 union f4vector
 {
 	v4sf v;
 	float f[4];
 };
+#else // MSVC
+#include <emmintrin.h>
+union f4vector
+{
+	__m128 v;
+	float f[4];
+};
+inline __m128 operator + (__m128 a, __m128 b) {return _mm_add_ps(a, b);}
+inline __m128 operator - (__m128 a, __m128 b) {return _mm_sub_ps(a, b);}
+inline __m128 operator * (__m128 a, __m128 b) {return _mm_mul_ps(a, b);}
+inline __m128 operator / (__m128 a, __m128 b) {return _mm_div_ps(a, b);}
+
+inline __m128 & operator += (__m128 & a, __m128 b){a = a + b; return a;}
+inline __m128 & operator -= (__m128 & a, __m128 b){a = a - b; return a;}
+inline __m128 & operator *= (__m128 & a, __m128 b){a = a * b; return a;}
+inline __m128 & operator /= (__m128 & a, __m128 b){a = a / b; return a;}
+#endif
 
 void Delete1DArray_v4sf(f4vector* array);
 void Delete3DArray_v4sf(f4vector*** array, const unsigned int* numLines);
diff --git a/tools/global.h b/tools/global.h
index c3bb524..bbbd8ae 100644
--- a/tools/global.h
+++ b/tools/global.h
@@ -19,10 +19,23 @@
 #define GLOBAL_H
 
 #include <sstream>
+#define _USE_MATH_DEFINES
 
+#include "tools/array_ops.h"
+#if defined(WIN32)
+	#ifdef BUILD_OPENEMS_LIB
+	#define OPENEMS_EXPORT __declspec(dllexport)
+	#else
+	#define OPENEMS_EXPORT __declspec(dllimport)
+	#endif
+#else
+#define OPENEMS_EXPORT
+#endif
+
+// declare a parameter as unused
 #define UNUSED(x) (void)(x);
 
-class Global
+class OPENEMS_EXPORT Global
 {
 public:
 	Global();
@@ -57,9 +70,6 @@ protected:
 	int m_SavedVerboseLevel;
 };
 
-extern Global g_settings;
-
-// declare a parameter as unused
-#define UNUSED(x) (void)(x);
+OPENEMS_EXPORT extern Global g_settings;
 
 #endif // GLOBAL_H
diff --git a/tools/hdf5_file_reader.h b/tools/hdf5_file_reader.h
index f190240..4f51caa 100644
--- a/tools/hdf5_file_reader.h
+++ b/tools/hdf5_file_reader.h
@@ -22,6 +22,7 @@
 #include <vector>
 #include <complex>
 #include <hdf5.h>
+#define _USE_MATH_DEFINES
 
 class HDF5_File_Reader
 {
diff --git a/tools/hdf5_file_writer.cpp b/tools/hdf5_file_writer.cpp
index e3a2113..56d3214 100644
--- a/tools/hdf5_file_writer.cpp
+++ b/tools/hdf5_file_writer.cpp
@@ -404,7 +404,7 @@ bool HDF5_File_Writer::WriteData(std::string dataSetName,  hid_t mem_type, void
 		return false;
 	}
 
-	hsize_t dims[dim];
+	hsize_t* dims = new hsize_t[dim];
 	for (size_t n=0;n<dim;++n)
 		dims[n]=datasize[n];
 	hid_t space = H5Screate_simple(dim, dims, NULL);
@@ -422,6 +422,7 @@ bool HDF5_File_Writer::WriteData(std::string dataSetName,  hid_t mem_type, void
 	H5Sclose(space);
 	H5Gclose(group);
 	H5Fclose(hdf5_file);
+	delete[] dims; dims=NULL;
 	return true;
 }
 
@@ -490,18 +491,22 @@ bool HDF5_File_Writer::WriteAtrribute(std::string locName, std::string attr_name
 
 bool HDF5_File_Writer::WriteAtrribute(std::string locName, std::string attr_name, vector<float> values)
 {
-	float val[values.size()];
+	float* val = new float[values.size()];
 	for (size_t n=0;n<values.size();++n)
 		val[n]=values.at(n);
-	return HDF5_File_Writer::WriteAtrribute(locName, attr_name,val,values.size(),H5T_NATIVE_FLOAT);
+	bool ok = HDF5_File_Writer::WriteAtrribute(locName, attr_name,val,values.size(),H5T_NATIVE_FLOAT);
+	delete[] val; val=NULL;
+	return ok;
 }
 
 bool HDF5_File_Writer::WriteAtrribute(std::string locName, std::string attr_name, vector<double> values)
 {
-	double val[values.size()];
+	double* val = new double[values.size()];
 	for (size_t n=0;n<values.size();++n)
 		val[n]=values.at(n);
-	return HDF5_File_Writer::WriteAtrribute(locName, attr_name, val, values.size(), H5T_NATIVE_DOUBLE);
+	bool ok =  HDF5_File_Writer::WriteAtrribute(locName, attr_name, val, values.size(), H5T_NATIVE_DOUBLE);
+	delete[] val; val=NULL;
+	return ok;
 }
 
 bool HDF5_File_Writer::WriteAtrribute(std::string locName, std::string attr_name, float value)
diff --git a/tools/sar_calculation.cpp b/tools/sar_calculation.cpp
index e135aca..791e842 100644
--- a/tools/sar_calculation.cpp
+++ b/tools/sar_calculation.cpp
@@ -16,6 +16,7 @@
 */
 
 
+#include <algorithm>
 #include "sar_calculation.h"
 #include "cfloat"
 #include "array_ops.h"
diff --git a/tools/useful.cpp b/tools/useful.cpp
index 50aacfb..209781c 100644
--- a/tools/useful.cpp
+++ b/tools/useful.cpp
@@ -183,3 +183,19 @@ int LinePlaneIntersection(const double *p0, const double *p1, const double *p2,
 
 	return 0;
 }
+
+#ifndef __GNUC__
+#include <chrono>
+#include <Winsock2.h> // for struct timeval
+
+int gettimeofday(struct timeval* tp, struct timezone* tzp) {
+  namespace sc = std::chrono;
+  sc::system_clock::duration d = sc::system_clock::now().time_since_epoch();
+  sc::seconds s = sc::duration_cast<sc::seconds>(d);
+  tp->tv_sec = s.count();
+  tp->tv_usec = sc::duration_cast<sc::microseconds>(d - s).count();
+
+  return 0;
+}
+
+#endif // _WIN32
\ No newline at end of file
diff --git a/tools/useful.h b/tools/useful.h
index dec13ab..ee4f1b3 100644
--- a/tools/useful.h
+++ b/tools/useful.h
@@ -41,4 +41,8 @@ double* Invert(const double* in, double* out);
 
 int LinePlaneIntersection(const double *p0, const double* p1, const double* p2, const double* l_start, const double* l_stop, double* is_point, double &dist);
 
+#ifndef __GNUC__
+int gettimeofday(struct timeval* tp, struct timezone* tzp);
+#endif // _WIN32
+
 #endif // USEFUL_H