Merge branch 'master' into tensorrt8

include/tkDNN/Int8BatchStream.h

@@ -39,7 +39,7 @@ public:
 	float *getLabels() { return mLabels.data(); }
 	int getBatchesRead() const { return mBatchCount; }
 	int getBatchSize() const { return mBatchSize; }
-	nvinfer1::Dims4 getDims() const { return mDims; }
+	nvinfer1::DimsNCHW getDims() const { return mDims; }
 	float* getFileBatch() { return &mFileBatch[0]; }
 	float* getFileLabels() { return &mFileLabels[0]; }
 	void readInListFile(const std::string& dataFilePath, std::vector<std::string>& mListIn);
@@ -55,7 +55,7 @@ private:
 	int mFileBatchPos{ 0 };
 	int mImageSize{ 0 };
 
-	nvinfer1::Dims4 mDims;
+	nvinfer1::DimsNCHW mDims;
 	std::vector<float> mBatch;
 	std::vector<float> mLabels;
 	std::vector<float> mFileBatch;

include/tkDNN/Int8Calibrator.h

@@ -30,10 +30,10 @@ public:
 	Int8EntropyCalibrator(BatchStream& stream, int firstBatch, const std::string& calibTableFilePath, 
 							const std::string& inputBlobName, bool readCache = true);
 	virtual ~Int8EntropyCalibrator() { checkCuda(cudaFree(mDeviceInput)); }
-	int getBatchSize() const NOEXCEPT override { return mStream.getBatchSize(); }
-	bool getBatch(void* bindings[], const char* names[], int nbBindings) NOEXCEPT override;
-	const void* readCalibrationCache(size_t& length) NOEXCEPT override;
-	void writeCalibrationCache(const void* cache, size_t length) NOEXCEPT override;
+	int getBatchSize() const override { return mStream.getBatchSize(); }
+	bool getBatch(void* bindings[], const char* names[], int nbBindings) override;
+	const void* readCalibrationCache(size_t& length) override;
+	void writeCalibrationCache(const void* cache, size_t length) override;
 
 private:
 	BatchStream mStream;

include/tkDNN/NetworkRT.h

@@ -33,7 +33,6 @@ namespace tk { namespace dnn {
 using namespace nvinfer1;
 
 
-
 class NetworkRT {
 
 public:
@@ -57,6 +56,7 @@ public:
     dnnType *output;
     cudaStream_t stream;
 
+    PluginFactory *pluginFactory;
 
     NetworkRT(Network *net, const char *name);
     virtual ~NetworkRT();

include/tkDNN/pluginsRT/ActivationReLUCeilingRT.h

@@ -78,4 +78,4 @@ namespace nvinfer1 {
     };
 
     REGISTER_TENSORRT_PLUGIN(ActivationReLUCeilingPluginCreator);
-};
+};

include/tkDNN/pluginsRT/RegionRT.h

@@ -107,4 +107,4 @@ namespace nvinfer1 {
     REGISTER_TENSORRT_PLUGIN(RegionRTPluginCreator);
 };
 
-#endif
+#endif

include/tkDNN/pluginsRT/ReorgRT.h

@@ -94,5 +94,4 @@ namespace nvinfer1 {
     };
 
     REGISTER_TENSORRT_PLUGIN(ReorgRTPluginCreator);
-};
-
+};

include/tkDNN/pluginsRT/ReshapeRT.h

@@ -99,4 +99,4 @@ namespace nvinfer1 {
 
     REGISTER_TENSORRT_PLUGIN(ReshapeRTPluginCreator);
 };
-#endif
+#endif

include/tkDNN/pluginsRT/ResizeLayerRT.h

@@ -101,4 +101,3 @@ namespace nvinfer1 {
 
     REGISTER_TENSORRT_PLUGIN(ResizeLayerRTPluginCreator);
 };
-

include/tkDNN/pluginsRT/YoloRT.h

@@ -121,4 +121,4 @@ namespace nvinfer1 {
 
     REGISTER_TENSORRT_PLUGIN(YoloRTPluginCreator);
 };
-#endif
+#endif

include/tkDNN/utils.h

@@ -13,7 +13,6 @@
 #include "cuda_runtime_api.h"
 #include <cublas_v2.h>
 #include <cudnn.h>
-#include <NvInferVersion.h>
 
 
 #ifdef __linux__
@@ -24,15 +23,6 @@
 #include <chrono>
 
 
-
-
-#if NV_TENSORRT_MAJOR > 7
-#define NOEXCEPT noexcept
-#else
-#define NOEXCEPT
-#endif
-
-
 #define dnnType float
 
 template<typename T> void writeBUF(char*& buffer, const T& val)