split all CUDA-specific code off from CNTKMath into a separate project CNTKMathCUDA, with the final goal that the Math project can compile without the (1 GB large) CUDA SDK installed;

renamed GPUDummy.cpp to NoGPU.cpp;
TODO: two test projects currently fail, and CNTKMath still depends on CUDA, to be fixed next

CNTKSolution/CNTKSolution.sln

@@ -4,6 +4,9 @@ Microsoft Visual Studio Solution File, Format Version 12.00
 VisualStudioVersion = 12.0.21005.1
 MinimumVisualStudioVersion = 10.0.40219.1
 Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "CNTKMath", "..\Math\Math\Math.vcxproj", "{60BDB847-D0C4-4FD3-A947-0C15C08BCDB5}"
+	ProjectSection(ProjectDependencies) = postProject
+		{B3DD765E-694E-4494-BAD7-37BBF2942517} = {B3DD765E-694E-4494-BAD7-37BBF2942517}
+	EndProjectSection
 EndProject
 Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "cn", "..\MachineLearning\cn\cn.vcxproj", "{E6F26F9A-FF64-4F0A-B749-CD309EE357EE}"
 	ProjectSection(ProjectDependencies) = postProject
@@ -62,6 +65,8 @@ Project("{2150E333-8FDC-42A3-9474-1A3956D46DE8}") = "Reader Plugins", "Reader Pl
 EndProject
 Project("{2150E333-8FDC-42A3-9474-1A3956D46DE8}") = "CNTK", "CNTK", "{DD043083-71A4-409A-AA91-F9C548DCF7EC}"
 EndProject
+Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "CNTKMathCUDA", "..\Math\Math\CNTKMathCUDA.vcxproj", "{B3DD765E-694E-4494-BAD7-37BBF2942517}"
+EndProject
 Global
 	GlobalSection(SolutionConfigurationPlatforms) = preSolution
 		Debug|x64 = Debug|x64
@@ -109,6 +114,10 @@ Global
 		{0F30EBCF-09F3-4EED-BF54-4214BCE53FEC}.Debug|x64.ActiveCfg = Debug|x64
 		{0F30EBCF-09F3-4EED-BF54-4214BCE53FEC}.Debug|x64.Build.0 = Debug|x64
 		{0F30EBCF-09F3-4EED-BF54-4214BCE53FEC}.Release|x64.ActiveCfg = Release|x64
+		{B3DD765E-694E-4494-BAD7-37BBF2942517}.Debug|x64.ActiveCfg = Debug|x64
+		{B3DD765E-694E-4494-BAD7-37BBF2942517}.Debug|x64.Build.0 = Debug|x64
+		{B3DD765E-694E-4494-BAD7-37BBF2942517}.Release|x64.ActiveCfg = Release|x64
+		{B3DD765E-694E-4494-BAD7-37BBF2942517}.Release|x64.Build.0 = Release|x64
 	EndGlobalSection
 	GlobalSection(SolutionProperties) = preSolution
 		HideSolutionNode = FALSE
@@ -117,6 +126,7 @@ Global
 		{E6F26F9A-FF64-4F0A-B749-CD309EE357EE} = {DD043083-71A4-409A-AA91-F9C548DCF7EC}
 		{482999D1-B7E2-466E-9F8D-2119F93EAFD9} = {DD043083-71A4-409A-AA91-F9C548DCF7EC}
 		{60BDB847-D0C4-4FD3-A947-0C15C08BCDB5} = {DD043083-71A4-409A-AA91-F9C548DCF7EC}
+		{B3DD765E-694E-4494-BAD7-37BBF2942517} = {DD043083-71A4-409A-AA91-F9C548DCF7EC}
 		{6CEE834A-8104-46A8-8902-64C81BD7928F} = {D45DF403-6781-444E-B654-A96868C5BE68}
 		{668BEED5-AC07-4F35-B3AE-EE65A7F9C976} = {D45DF403-6781-444E-B654-A96868C5BE68}
 		{0F30EBCF-09F3-4EED-BF54-4214BCE53FEC} = {D45DF403-6781-444E-B654-A96868C5BE68}

Math/Math/GPUMatrix.h

@@ -14,6 +14,10 @@
 #include "Helpers.h"
 #include "CommonMatrix.h"
 
+#ifndef CPUONLY
+#pragma comment (lib, "CNTKMathCUDA.lib")   // built by CNTKMathCUDA
+#endif
+
 // predeclare cublasHandle_t
 struct cublasContext;
 typedef struct cublasContext *cublasHandle_t;
@@ -376,4 +380,3 @@ namespace Microsoft { namespace MSR { namespace CNTK {
     typedef GPUMatrix<float> GPUSingleMatrix;
 
 }}}
-

Math/Math/Math.vcxproj

@@ -178,23 +178,12 @@
       </PrecompiledHeader>
     </ClCompile>
     <ClCompile Include="CPUMatrix.cpp" />
-    <CudaCompile Include="GPUSparseMatrix.cu" />
-    <CudaCompile Include="GPUWatcher.cu">
-      <FileType>CppCode</FileType>
-    </CudaCompile>
-    <ClCompile Include="GPUDummy.cpp" />
+    <ClCompile Include="NoGPU.cpp" />
     <ClCompile Include="Matrix.cpp" />
     <ClCompile Include="stdafx.cpp">
       <PrecompiledHeader>Create</PrecompiledHeader>
     </ClCompile>
   </ItemGroup>
-  <ItemGroup>
-    <CudaCompile Include="GPUMatrix.cu" />
-    <CudaCompile Include="GPUMatrixCUDAKernels.cu">
-      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">true</ExcludedFromBuild>
-      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|x64'">true</ExcludedFromBuild>
-    </CudaCompile>
-  </ItemGroup>
   <ItemGroup>
     <None Include="GPUMatrix.h" />
   </ItemGroup>

Math/Math/Math.vcxproj.filters

@@ -1,19 +1,5 @@
 <?xml version="1.0" encoding="utf-8"?>
 <Project ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
-  <ItemGroup>
-    <CudaCompile Include="GPUMatrix.cu">
-      <Filter>GPU</Filter>
-    </CudaCompile>
-    <CudaCompile Include="GPUMatrixCUDAKernels.cu">
-      <Filter>GPU</Filter>
-    </CudaCompile>
-    <CudaCompile Include="GPUSparseMatrix.cu">
-      <Filter>GPU</Filter>
-    </CudaCompile>
-    <CudaCompile Include="GPUWatcher.cu">
-      <Filter>GPU</Filter>
-    </CudaCompile>
-  </ItemGroup>
   <ItemGroup>
     <ClCompile Include="dllmain.cpp" />
     <ClCompile Include="Matrix.cpp" />
@@ -30,7 +16,7 @@
     <ClCompile Include="CPUSparseMatrix.cpp">
       <Filter>CPU</Filter>
     </ClCompile>
-    <ClCompile Include="GPUDummy.cpp">
+    <ClCompile Include="NoGPU.cpp">
       <Filter>GPU</Filter>
     </ClCompile>
   </ItemGroup>

Math/Math/NoGPU.cpp

@@ -1,5 +1,5 @@
 //
-// <copyright file="GPUDummy.cu" company="Microsoft">
+// <copyright file="NoGPU.cpp" company="Microsoft">
 //     Copyright (c) Microsoft Corporation.  All rights reserved.
 // </copyright>
 //