diff --git a/libNeonCore/include/Neon/core/tools/io/IODense_imp.h b/libNeonCore/include/Neon/core/tools/io/IODense_imp.h
index fb16dbb2..f8c2a355 100644
--- a/libNeonCore/include/Neon/core/tools/io/IODense_imp.h
+++ b/libNeonCore/include/Neon/core/tools/io/IODense_imp.h
@@ -280,11 +280,14 @@ auto IODense<ExportType, IntType>::maxDiff(const IODense<ExportType, IntType>& a
         const auto newDiff = std::abs(valA - valB);
         const int  threadId = omp_get_thread_num();
         if (newDiff > std::get<0>(max_diff[threadId])) {
+#pragma omp critical
+            if (newDiff > std::get<0>(max_diff[threadId])) {
             Values& md = max_diff[threadId];
             std::get<0>(md) = newDiff;
             std::get<1>(md) = idx;
             std::get<2>(md) = c;
         }
+        }
     },
               b);
 
diff --git a/libNeonCore/include/Neon/core/types/vec/vec3d_integer.timp.h b/libNeonCore/include/Neon/core/types/vec/vec3d_integer.timp.h
index 304c5b37..8d4f30b0 100644
--- a/libNeonCore/include/Neon/core/types/vec/vec3d_integer.timp.h
+++ b/libNeonCore/include/Neon/core/types/vec/vec3d_integer.timp.h
@@ -1,9 +1,3 @@
-
-
-// $Id$
-
-// $Log$
-
 #pragma once
 
 #include <cmath>
@@ -13,6 +7,7 @@
 #include <string>
 #include <type_traits>
 
+#include "Neon/core/types/Macros.h"
 // #include <cuda.h>
 // #include <cuda_runtime_api.h>
 
@@ -686,7 +681,11 @@ void Vec_3d<IntegerType_ta, true, false>::forEach(const self_t&
                                                   std::function<void(Integer idxX, Integer idxY, Integer idxZ)> lambda)
 {
     if constexpr (computeMode_ta == Neon::computeMode_t::par) {
+#ifdef NEON_OS_WINDOWS
+#pragma omp parallel for
+#else
 #pragma omp parallel for simd collapse(3)
+#endif
         for (Integer z = 0; z < len.z; z++) {
             for (Integer y = 0; y < len.y; y++) {
                 for (Integer x = 0; x < len.x; x++) {
@@ -713,7 +712,11 @@ auto Vec_3d<IntegerType_ta, true, false>::forEach(const Lambda& lambda) const
 {
     if constexpr (std::is_invocable_v<Lambda, Integer, Integer, Integer>) {
         if constexpr (computeMode_ta == Neon::computeMode_t::par) {
+#ifdef NEON_OS_WINDOWS
+#pragma omp parallel for
+#else
 #pragma omp parallel for simd collapse(3)
+#endif
             for (Integer zIdx = 0; zIdx < this->z; zIdx++) {
                 for (Integer yIdx = 0; yIdx < this->y; yIdx++) {
                     for (Integer xIdx = 0; xIdx < this->x; xIdx++) {
@@ -734,7 +737,11 @@ auto Vec_3d<IntegerType_ta, true, false>::forEach(const Lambda& lambda) const
     }
     if constexpr (std::is_invocable_v<Lambda, Vec_3d<IntegerType_ta, true, false>>) {
         if constexpr (computeMode_ta == Neon::computeMode_t::par) {
+#ifdef NEON_OS_WINDOWS
+#pragma omp parallel for
+#else
 #pragma omp parallel for simd collapse(3)
+#endif
             for (Integer zIdx = 0; zIdx < this->z; zIdx++) {
                 for (Integer yIdx = 0; yIdx < this->y; yIdx++) {
                     for (Integer xIdx = 0; xIdx < this->x; xIdx++) {
diff --git a/libNeonCore/tests/unit/coreUt_io/src/coreUt_io.dense.cpp b/libNeonCore/tests/unit/coreUt_io/src/coreUt_io.dense.cpp
index 63760473..261f8bc4 100644
--- a/libNeonCore/tests/unit/coreUt_io/src/coreUt_io.dense.cpp
+++ b/libNeonCore/tests/unit/coreUt_io/src/coreUt_io.dense.cpp
@@ -73,4 +73,5 @@ TEST(coreUt_io, denseDiffRandom)
         ASSERT_EQ(maxDiff.z, id.z);
         ASSERT_EQ(maxCard, card);
     }
+        A.ioVtk("A", "A");
 }
\ No newline at end of file