| | |
| | | <IntrinsicFunctions>true</IntrinsicFunctions> |
| | | <SDLCheck>true</SDLCheck> |
| | | <AdditionalIncludeDirectories>C:\opencv_2.4.9\opencv\build\include;..\..\3rdparty\include;%(AdditionalIncludeDirectories);$(CudaToolkitIncludeDir);$(cudnn)\include</AdditionalIncludeDirectories> |
| | | <PreprocessorDefinitions>OPENCV;_TIMESPEC_DEFINED;_CRT_SECURE_NO_WARNINGS;GPU;WIN32;NDEBUG;_CONSOLE;_LIB;%(PreprocessorDefinitions)</PreprocessorDefinitions> |
| | | <PreprocessorDefinitions>CUDNN;OPENCV;_TIMESPEC_DEFINED;_CRT_SECURE_NO_WARNINGS;GPU;WIN32;NDEBUG;_CONSOLE;_LIB;%(PreprocessorDefinitions)</PreprocessorDefinitions> |
| | | <CLanguageStandard>c11</CLanguageStandard> |
| | | <CppLanguageStandard>c++1y</CppLanguageStandard> |
| | | <PrecompiledHeaderCompileAs>CompileAsCpp</PrecompiledHeaderCompileAs> |
| | |
| | | local[id] += (i+id < spatial) ? delta[index] : 0; |
| | | } |
| | | } |
| | | __syncthreads(); |
| | | |
| | | if(id == 0){ |
| | | mean_delta[filter] = 0; |
| | |
| | | local[id] += (i+id < spatial) ? delta[index]*(x[index] - mean[filter]) : 0; |
| | | } |
| | | } |
| | | __syncthreads(); |
| | | |
| | | if(id == 0){ |
| | | variance_delta[filter] = 0; |
| | |
| | | local[id] += (i+id < spatial) ? x[index] : 0; |
| | | } |
| | | } |
| | | __syncthreads(); |
| | | |
| | | if(id == 0){ |
| | | mean[filter] = 0; |
| | |
| | | local[id] += (i+id < spatial) ? pow((x[index] - mean[filter]), 2) : 0; |
| | | } |
| | | } |
| | | __syncthreads(); |
| | | |
| | | if(id == 0){ |
| | | variance[filter] = 0; |
| | |
| | | activate_array_ongpu(l.output_gpu, l.outputs*l.batch, l.activation); |
| | | //if(l.dot > 0) dot_error_gpu(l); |
| | | if(l.binary || l.xnor) swap_binary(&l); |
| | | //cudaDeviceSynchronize(); // for correct profiling of performance |
| | | } |
| | | |
| | | void backward_convolutional_layer_gpu(convolutional_layer l, network_state state) |