Improved thrust interop tutorial.

samples/cpp/tutorial_code/gpu/gpu-thrust-interop/Thrust_interop.hpp

@@ -6,20 +6,10 @@
 #include <thrust/iterator/counting_iterator.h>
 #include <thrust/device_ptr.h>
 
-template<typename T> struct
-CV_TYPE
-{
-	static const int DEPTH;
-};
-
-template<> static const int CV_TYPE<float>::DEPTH = CV_32F;
-template<> static const int CV_TYPE<double>::DEPTH = CV_64F;
-template<> static const int CV_TYPE<int>::DEPTH = CV_32S;
-template<> static const int CV_TYPE<uchar>::DEPTH = CV_8U;
-template<> static const int CV_TYPE<char>::DEPTH = CV_8S;
-template<> static const int CV_TYPE<ushort>::DEPTH = CV_16U;
-template<> static const int CV_TYPE<short>::DEPTH = CV_16S;
-
+/*
+	@Brief step_functor is an object to correctly step a thrust iterator according to the stride of a matrix
+*/
+//! [step_functor]
 template<typename T> struct step_functor : public thrust::unary_function<int, int>
 {
 	int columns;
@@ -41,7 +31,8 @@ template<typename T> struct step_functor : public thrust::unary_function<int, in
 		return idx;
 	}
 };
-
+//! [step_functor]
+//! [begin_itr]
 /*
 	@Brief GpuMatBeginItr returns a thrust compatible iterator to the beginning of a GPU mat's memory.  
 	@Param mat is the input matrix
@@ -52,11 +43,13 @@ thrust::permutation_iterator<thrust::device_ptr<T>, thrust::transform_iterator<s
 {
 	if (channel == -1)
 		mat = mat.reshape(1);
-	CV_Assert(mat.depth() == CV_TYPE<T>::DEPTH);
+	CV_Assert(mat.depth() == cv::DataType<T>::depth);
 	CV_Assert(channel < mat.channels());
 	return thrust::make_permutation_iterator(thrust::device_pointer_cast(mat.ptr<T>(0) + channel),
 		thrust::make_transform_iterator(thrust::make_counting_iterator(0), step_functor<T>(mat.cols, mat.step / sizeof(T), mat.channels())));
 }
+//! [begin_itr]
+//! [end_itr]
 /*
 @Brief GpuMatEndItr returns a thrust compatible iterator to the end of a GPU mat's memory.
 @Param mat is the input matrix
@@ -67,8 +60,11 @@ thrust::permutation_iterator<thrust::device_ptr<T>, thrust::transform_iterator<s
 {
 	if (channel == -1)
 		mat = mat.reshape(1);
-	CV_Assert(mat.depth() == CV_TYPE<T>::DEPTH);
+	CV_Assert(mat.depth() == cv::DataType<T>::depth);
 	CV_Assert(channel < mat.channels());
 	return thrust::make_permutation_iterator(thrust::device_pointer_cast(mat.ptr<T>(0) + channel),
 		thrust::make_transform_iterator(thrust::make_counting_iterator(mat.rows*mat.cols), step_functor<T>(mat.cols, mat.step / sizeof(T), mat.channels())));
-}
+}
+//! [end_itr]
+
+

samples/cpp/tutorial_code/gpu/gpu-thrust-interop/main.cu

@@ -5,6 +5,7 @@
 #include <thrust/random.h>
 #include <thrust/sort.h>
 #include <thrust/system/cuda/execution_policy.h>
+//! [prg]
 struct prg
 {
 	float a, b;
@@ -22,36 +23,10 @@ struct prg
 		return dist(rng);
 	}
 };
+//! [prg]
 
-template<typename T> struct pred_eq
-{
-	T value;
-	int channel;
-	__host__ __device__
-		pred_eq(T value_, int channel_ = 0) :value(value_), channel(channel_){}
-	__host__ __device__
-		bool operator()(const T val) const
-	{
-		return val == value;
-	}
-	template<int N>
-	__host__ __device__ bool operator()(const cv::Vec<T, N>& val)
-	{
-		if (channel < N)
-			return val.val[channel] == value;
-		return false;
-	}
 
-	__host__ __device__ bool operator()( const thrust::tuple<T, T, T>& val)
-	{
-		if (channel == 0)
-			return thrust::get<0>(val) == value;
-		if (channel == 1)
-			return thrust::get<1>(val) == value;
-		if (channel == 2)
-			return thrust::get<2>(val) == value;
-	}
-};
+//! [pred_greater]
 template<typename T> struct pred_greater
 {
 	T value;
@@ -61,12 +36,14 @@ template<typename T> struct pred_greater
 		return val > value;
 	}
 };
+//! [pred_greater]
 
 
 int main(void)
 {
 	// Generate a 2 channel row matrix with 100 elements.  Set the first channel to be the element index, and the second to be a randomly
 	// generated value.  Sort by the randomly generated value while maintaining index association.
+	//! [sort]
 	{
 		cv::cuda::GpuMat d_idx(1, 100, CV_32SC2);
 
@@ -82,8 +59,10 @@ int main(void)
 
 		cv::Mat h_idx(d_idx);
 	}
+	//! [sort]
 
 	// Randomly fill a row matrix with 100 elements between -1 and 1
+	//! [random]
 	{
 		cv::cuda::GpuMat d_value(1, 100, CV_32F);
 		auto valueBegin = GpuMatBeginItr<float>(d_value);
@@ -92,8 +71,10 @@ int main(void)
 
 		cv::Mat h_value(d_value);
 	}
+	//! [random]
 
 	// OpenCV has count non zero, but what if you want to count a specific value?
+	//! [count_value]
 	{
 		cv::cuda::GpuMat d_value(1, 100, CV_32S);
 		d_value.setTo(cv::Scalar(0));
@@ -101,18 +82,24 @@ int main(void)
 		auto count = thrust::count(GpuMatBeginItr<int>(d_value), GpuMatEndItr<int>(d_value), 15);
 		std::cout << count << std::endl;
 	}
+	//! [count_value]
+
 	// Randomly fill an array then copy only values greater than 0.  Perform these tasks on a stream.
+	//! [copy_greater]
 	{
 		cv::cuda::GpuMat d_value(1, 100, CV_32F);
 		auto valueBegin = GpuMatBeginItr<float>(d_value);
 		auto valueEnd = GpuMatEndItr<float>(d_value);
 		cv::cuda::Stream stream;
+		//! [random_gen_stream]
 		thrust::transform(thrust::system::cuda::par.on(cv::cuda::StreamAccessor::getStream(stream)), thrust::make_counting_iterator(0), thrust::make_counting_iterator(d_value.cols), valueBegin, prg(-1, 1));
+		//! [random_gen_stream]
 		int count = thrust::count_if(thrust::system::cuda::par.on(cv::cuda::StreamAccessor::getStream(stream)), valueBegin, valueEnd, pred_greater<float>(0.0));
 		cv::cuda::GpuMat d_valueGreater(1, count, CV_32F);
 		thrust::copy_if(thrust::system::cuda::par.on(cv::cuda::StreamAccessor::getStream(stream)), valueBegin, valueEnd, GpuMatBeginItr<float>(d_valueGreater), pred_greater<float>(0.0));
 		cv::Mat h_greater(d_valueGreater);
 	}
+	//! [copy_greater]
 	
 	return 0;
 }