Fixed compilation errors on Windows

libraries/vecmath/include/neon_mathfun.h

@@ -26,6 +26,7 @@
 */
 
 #include <arm_neon.h>
+#include "openmm/internal/windowsExport.h"
 
 typedef float32x4_t v4sf;  // vector of 4 float
 typedef uint32x4_t v4su;  // vector of 4 uint32
@@ -48,7 +49,7 @@ typedef int32x4_t v4si;  // vector of 4 uint32
 /* natural logarithm computed for 4 simultaneous float 
    return NaN for x <= 0
 */
-v4sf log_ps(v4sf x) {
+OPENMM_EXPORT v4sf log_ps(v4sf x) {
   v4sf one = vdupq_n_f32(1);
 
   x = vmaxq_f32(x, vdupq_n_f32(0)); /* force flush to zero on denormal values */
@@ -133,7 +134,7 @@ v4sf log_ps(v4sf x) {
 #define c_cephes_exp_p5 5.0000001201E-1
 
 /* exp() computed for 4 float at once */
-v4sf exp_ps(v4sf x) {
+OPENMM_EXPORT v4sf exp_ps(v4sf x) {
   v4sf tmp, fx;
 
   v4sf one = vdupq_n_f32(1);
@@ -219,7 +220,7 @@ v4sf exp_ps(v4sf x) {
    almost no extra price so both sin_ps and cos_ps make use of
    sincos_ps..
   */
-void sincos_ps(v4sf x, v4sf *ysin, v4sf *ycos) { // any x
+OPENMM_EXPORT void sincos_ps(v4sf x, v4sf *ysin, v4sf *ycos) { // any x
   v4sf xmm1, xmm2, xmm3, y;
 
   v4su emm2;
@@ -286,13 +287,13 @@ void sincos_ps(v4sf x, v4sf *ysin, v4sf *ycos) { // any x
   *ycos = vbslq_f32(sign_mask_cos, yc, vnegq_f32(yc));
 }
 
-v4sf sin_ps(v4sf x) {
+OPENMM_EXPORT v4sf sin_ps(v4sf x) {
   v4sf ysin, ycos; 
   sincos_ps(x, &ysin, &ycos); 
   return ysin;
 }
 
-v4sf cos_ps(v4sf x) {
+OPENMM_EXPORT v4sf cos_ps(v4sf x) {
   v4sf ysin, ycos; 
   sincos_ps(x, &ysin, &ycos); 
   return ycos;

libraries/vecmath/include/sse_mathfun.h

@@ -30,6 +30,7 @@
 */
 
 #include <xmmintrin.h>
+#include "openmm/internal/windowsExport.h"
 
 /* yes I know, the top of this file is quite ugly */
 
@@ -109,7 +110,7 @@ typedef union xmm_mm_union {
 /* natural logarithm computed for 4 simultaneous float 
    return NaN for x <= 0
 */
-v4sf log_ps(v4sf x) {
+OPENMM_EXPORT v4sf log_ps(v4sf x) {
 #ifdef USE_SSE2
   v4si emm0;
 #else
@@ -211,7 +212,7 @@ _PS_CONST(cephes_exp_p3, 4.1665795894E-2);
 _PS_CONST(cephes_exp_p4, 1.6666665459E-1);
 _PS_CONST(cephes_exp_p5, 5.0000001201E-1);
 
-v4sf exp_ps(v4sf x) {
+OPENMM_EXPORT v4sf exp_ps(v4sf x) {
   v4sf tmp = _mm_setzero_ps(), fx;
 #ifdef USE_SSE2
   v4si emm0;
@@ -329,7 +330,7 @@ _PS_CONST(cephes_FOPI, 1.27323954473516); // 4 / M_PI
    Since it is based on SSE intrinsics, it has to be compiled at -O2 to
    deliver full speed.
 */
-v4sf sin_ps(v4sf x) { // any x
+OPENMM_EXPORT v4sf sin_ps(v4sf x) { // any x
   v4sf xmm1, xmm2 = _mm_setzero_ps(), xmm3, sign_bit, y;
 
 #ifdef USE_SSE2
@@ -446,7 +447,7 @@ v4sf sin_ps(v4sf x) { // any x
 }
 
 /* almost the same as sin_ps */
-v4sf cos_ps(v4sf x) { // any x
+OPENMM_EXPORT v4sf cos_ps(v4sf x) { // any x
   v4sf xmm1, xmm2 = _mm_setzero_ps(), xmm3, y;
 #ifdef USE_SSE2
   v4si emm0, emm2;
@@ -565,7 +566,7 @@ v4sf cos_ps(v4sf x) { // any x
 
 /* since sin_ps and cos_ps are almost identical, sincos_ps could replace both of them..
    it is almost as fast, and gives you a free cosine with your sine */
-void sincos_ps(v4sf x, v4sf *s, v4sf *c) {
+OPENMM_EXPORT void sincos_ps(v4sf x, v4sf *s, v4sf *c) {
   v4sf xmm1, xmm2, xmm3 = _mm_setzero_ps(), sign_bit_sin, y;
 #ifdef USE_SSE2
   v4si emm0, emm2, emm4;

libraries/vecmath/src/vecmath.cpp

@@ -2,6 +2,7 @@
     #include "neon_mathfun.h"
 #else
     #if !defined(__PNACL__)
+        #define USE_SSE2
         #include "sse_mathfun.h"
     #endif
 #endif