Integrate alrezni/quick_fix into master

Makefile

@@ -1188,6 +1188,7 @@ UNITTEST_MATH_SRC = \
 	$(SOURCEDIR)/../Tests/UnitTests/MathTests/MatrixLearnerTests.cpp \
 	$(SOURCEDIR)/../Tests/UnitTests/MathTests/stdafx.cpp \
 
+UNITTEST_MATH_SRC += $(CNTK_COMMON_SRC)
 UNITTEST_MATH_OBJ := $(patsubst %.cpp, $(OBJDIR)/%.o, $(UNITTEST_MATH_SRC))
 
 UNITTEST_MATH := $(BINDIR)/mathtests

Source/Common/Common.vcxproj

@@ -34,7 +34,9 @@
   <Import Project="$(SolutionDir)\CNTK.Cpp.props" />
   <Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
   <Import Project="$(VCTargetsPath)\Microsoft.Cpp.props" />
-  <ImportGroup Label="ExtensionSettings" />
+  <ImportGroup Condition="$(GpuBuild)" Label="ExtensionSettings">
+    <Import Project="$(VCTargetsPath)\BuildCustomizations\CUDA $(CudaVersion).props" />
+  </ImportGroup>
   <ImportGroup Label="PropertySheets">
     <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
   </ImportGroup>
@@ -56,6 +58,15 @@
       <AdditionalOptions>/d2Zi+ %(AdditionalOptions)</AdditionalOptions>
     </ClCompile>
   </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="$(GpuBuild)">
+    <ClCompile>
+      <AdditionalIncludeDirectories>%(AdditionalIncludeDirectories);$(CudaInclude)</AdditionalIncludeDirectories>
+    </ClCompile>
+    <Link>
+      <AdditionalLibraryDirectories>%(AdditionalLibraryDirectories);$(CudaLibPath)</AdditionalLibraryDirectories>
+      <DelayLoadDLLs>%(DelayLoadDLLs);nvml.dll;$(CudaRuntimeDll)</DelayLoadDLLs>
+    </Link>
+  </ItemDefinitionGroup>
   <ItemDefinitionGroup Condition="$(CpuOnlyBuild)">
     <ClCompile>
       <PreprocessorDefinitions>%(PreprocessorDefinitions);CPUONLY</PreprocessorDefinitions>
@@ -63,6 +74,7 @@
   </ItemDefinitionGroup>
   <ItemGroup>
     <ClCompile Include="..\Math\NcclComm.cpp" />
+    <ClCompile Include="BestGpu.cpp" />
     <ClCompile Include="Config.cpp" />
     <ClCompile Include="DataReader.cpp" />
     <ClCompile Include="DataWriter.cpp" />
@@ -76,5 +88,7 @@
     <ClCompile Include="TimerUtility.cpp" />
   </ItemGroup>
   <Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
-  <ImportGroup Label="ExtensionTargets" />
+  <ImportGroup Condition="$(GpuBuild)" Label="ExtensionTargets">
+    <Import Project="$(VCTargetsPath)\BuildCustomizations\CUDA $(CudaVersion).targets" />
+  </ImportGroup>
 </Project>

Source/Common/Include/BestGpu.h

@@ -15,8 +15,6 @@ namespace Microsoft { namespace MSR { namespace ScriptableObjects { struct IConf
 
 namespace Microsoft { namespace MSR { namespace CNTK {
 
-using namespace std;
-
 #ifndef CPUONLY
 enum class GpuValidity
 {

Tests/UnitTests/MathTests/GPUMatrixTests.cpp

@@ -7,6 +7,7 @@
 #include "stdafx.h"
 #include "../../../Source/Math/GPUMatrix.h"
 #include "../../../Source/Math/Matrix.h"
+#include "BestGpu.h"
 
 using namespace Microsoft::MSR::CNTK;
 
@@ -16,18 +17,13 @@ BOOST_AUTO_TEST_SUITE(GPUMatrixSuite)
 
 BOOST_FIXTURE_TEST_CASE(MatrixCopyAssignAccrossDevices, RandomSeedFixture)
 {
+    bool hasTwoGpus = false;
+#ifndef CPUONLY
+    auto gpus = GetAllGpusData();
+    hasTwoGpus = (gpus.size() > 1);
+#endif
     std::array<float, 6> array = { 1, 2, 3, 4, 5, 6 };
 
-    bool hasTwoGpus = true;
-    try
-    {
-        Matrix<float> m_gpu(2, 3, array.data(), c_deviceIdZero + 1, matrixFlagNormal);
-    }
-    catch (...)
-    {
-        hasTwoGpus = false;
-    }
-
     {
         Matrix<float> m_gpu(2, 3, array.data(), c_deviceIdZero, matrixFlagNormal);
         Matrix<float> m_copy_gpu_0(m_gpu, c_deviceIdZero);