Removed the extra shell around the Cuda Plugin. Changed include pathes.

Former-commit-id: c7fec9220d

CMakeLists.txt

@@ -330,7 +330,8 @@ target_link_libraries(storm-performance-tests ${Boost_LIBRARIES})
 #############################################################
 if (ENABLE_CUDAFORSTORM)
     message (STATUS "StoRM - Linking with CudaForStorm")
-	include_directories("${PROJECT_SOURCE_DIR}/build/cudaForStorm/include")
+	include_directories("${PROJECT_BINARY_DIR}/cudaForStorm/include")
+	include_directories("${PROJECT_SOURCE_DIR}/resources/cudaForStorm")
     target_link_libraries(storm cudaForStorm)
     target_link_libraries(storm-functional-tests cudaForStorm)
     target_link_libraries(storm-performance-tests cudaForStorm)

resources/cudaForStorm/CMakeLists.txt

@@ -183,7 +183,7 @@ include (GenerateExportHeader)
 ##
 #############################################################
 #set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} --gpu-architecture sm_30)
-cuda_add_library(cudaForStorm
+cuda_add_library(cudaForStorm SHARED
   ${CUDAFORSTORM_CUDA_SOURCES} ${CUDAFORSTORM_CUDA_HEADERS}
   OPTIONS -DSTUFF="" -arch=sm_30
   RELEASE -DNDEBUG
@@ -223,4 +223,4 @@ endif(LINK_LIBCXXABI)
 
 # Install Directive
 install(TARGETS cudaForStorm DESTINATION "${STORM_LIB_INSTALL_DIR}/lib")
-install(FILES "${PROJECT_SOURCE_DIR}/src/cudaForStorm.h" "${PROJECT_BINARY_DIR}/cudaForStorm_Export.h" DESTINATION "${STORM_LIB_INSTALL_DIR}/include")
+install(FILES "${PROJECT_SOURCE_DIR}/srcCuda/cudaForStorm.h" "${PROJECT_BINARY_DIR}/include/cudaForStorm_Export.h" DESTINATION "${STORM_LIB_INSTALL_DIR}/include")

resources/cudaForStorm/src/cudaForStorm.h

@@ -1,15 +0,0 @@
-#ifndef STORM_CUDAFORSTORM_CUDAFORSTORM_H_
-#define STORM_CUDAFORSTORM_CUDAFORSTORM_H_
-
-// Library exports
-#include "cudaForStorm_Export.h"
-
-/*
- * List of exported functions in this library
- */
-
-cudaForStorm_EXPORT int cudaForStormTest(int value);
-
-
-
-#endif // STORM_CUDAFORSTORM_CUDAFORSTORM_H_

resources/cudaForStorm/src/cudaTests.h

@@ -1,124 +0,0 @@
-#include <cuda.h>
-#include "srcCuda/allCudaKernels.h"
-
-#include <iostream>
-#include <chrono>
-#include <random>
-
-void cudaShowDevices() {
-	// Todo
-}
-
-void cudaSimpleAddTest(int a, int b) {
-	std::cout << "Running cudaSimpleAddTest:" << std::endl;
-	std::cout << "a = " << a << ", b = " << b << "" << std::endl;
-
-	int c = cuda_basicAdd(a, b);
-	
-	std::cout << "Result: " << c << "" << std::endl;
-}
-
-void cudaArrayFmaTest(int N) {
-	std::cout << "Running cudaArrayFmaTest:" << std::endl;
-	std::cout << "N is " << N << ", resulting in " << (5 * sizeof(int) * N) << " Bytes of Data." << std::endl;
-
-	std::cout << "Generating random input arrays." << std::endl;
-
-	std::default_random_engine generator;
-	std::uniform_int_distribution<int> distribution(0, INT32_MAX);
-	int dice_roll = distribution(generator);
-
-	auto start_time = std::chrono::high_resolution_clock::now();
-
-	int* arrayA = new int[N];
-	int* arrayB = new int[N];
-	int* arrayC = new int[N];
-	int* arrayD = new int[N];
-	int* arrayD_CPU = new int[N];
-
-	for (int i = 0; i < N; ++i) {
-		//arrayA[i] = distribution(generator);
-		//arrayB[i] = distribution(generator);
-		//arrayC[i] = distribution(generator);
-		arrayA[i] = i * 1000 + 137;
-		arrayB[i] = i * 7000 + 1537;
-		arrayC[i] = i * 15000 + 97;
-		arrayD[i] = 0;
-		arrayD_CPU[i] = 0;
-	}
-
-	auto end_time = std::chrono::high_resolution_clock::now();
-	std::cout << "Array generation took " << std::chrono::duration_cast<std::chrono::microseconds>(end_time - start_time).count() << "micros" << std::endl;
-
-	std::cout << "Running FMA test on CPU." << std::endl;
-
-	start_time = std::chrono::high_resolution_clock::now();	
-	cuda_arrayFmaHelper(arrayA, arrayB, arrayC, arrayD_CPU, N);
-	end_time = std::chrono::high_resolution_clock::now();
-	std::cout << "FMA on CPU took " << std::chrono::duration_cast<std::chrono::microseconds>(end_time - start_time).count() << "micros" << std::endl;
-
-	start_time = std::chrono::high_resolution_clock::now();
-	cuda_arrayFma(arrayA, arrayB, arrayC, arrayD, N);
-	end_time = std::chrono::high_resolution_clock::now();
-	std::cout << "FMA on GPU took " << std::chrono::duration_cast<std::chrono::microseconds>(end_time - start_time).count() << "micros" << std::endl;
-
-	int errors = 0;
-	for (int i = 0; i < N; ++i) {
-		if (arrayD[i] != arrayD_CPU[i]) {
-			std::cout << "Error in Entry " << i << ": GPU has " << arrayD[i] << " but CPU has " << arrayD_CPU[i] << "!" << std::endl;
-			++errors;
-		}
-	}
-	std::cout << "Checked Arrays for Errors: " << errors << " Errors occured." << std::endl;
-}
-
-void cudaArrayFmaOptimizedTest(int N, int M) {
-	std::cout << "Running cudaArrayFmaTest:" << std::endl;
-	std::cout << "N is " << N << ", resulting in " << (4 * sizeof(int) * N) << " Bytes of Data." << std::endl;
-
-	size_t freeCudaMemory = getFreeCudaMemory();
-	size_t totalCudaMemory = getTotalCudaMemory();
-	int freeProzent = static_cast<int>(((double)freeCudaMemory)/((double)totalCudaMemory) * 100);
-
-	std::cout << "CUDA Device has " << freeCudaMemory << " Bytes of " << totalCudaMemory << " Bytes free (" << (freeProzent) << "%)." << std::endl;
-
-	std::cout << "Generating random input arrays." << std::endl;
-
-	std::default_random_engine generator;
-	std::uniform_int_distribution<int> distribution(0, INT32_MAX);
-
-	auto start_time = std::chrono::high_resolution_clock::now();
-
-	int* arrayA = new int[4 * N];
-	int* arrayA_CPU = new int[4 * N];
-
-	for (int i = 0; i < 4*N; ++i) {
-		arrayA[i] = i * 1000 + i + (357854878 % (i+1));
-		arrayA_CPU[i] = arrayA[i];
-	}
-
-	auto end_time = std::chrono::high_resolution_clock::now();
-	std::cout << "Array generation took " << std::chrono::duration_cast<std::chrono::microseconds>(end_time - start_time).count() << "micros" << std::endl;
-
-	start_time = std::chrono::high_resolution_clock::now();	
-	cuda_arrayFmaOptimizedHelper(arrayA_CPU, N);
-	end_time = std::chrono::high_resolution_clock::now();
-	std::cout << "FMA on CPU took " << std::chrono::duration_cast<std::chrono::microseconds>(end_time - start_time).count() << "micros" << std::endl;
-
-	start_time = std::chrono::high_resolution_clock::now();
-	cuda_arrayFmaOptimized(arrayA, N, M);
-	end_time = std::chrono::high_resolution_clock::now();
-	std::cout << "FMA on GPU took " << std::chrono::duration_cast<std::chrono::microseconds>(end_time - start_time).count() << "micros" << std::endl;
-
-	int errors = 0;
-	for (int i = 0; i < N; i+=4) {
-		if (arrayA[i+3] != arrayA_CPU[i+3]) {
-			//std::cout << "Error in Entry " << i << ": GPU has " << arrayA[i+3] << " but CPU has " << arrayA_CPU[i+3] << "!" << std::endl;
-			++errors;
-		}
-	}
-	std::cout << "Checked Arrays for Errors: " << errors << " Errors occured." << std::endl;
-
-	delete[] arrayA;
-	delete[] arrayA_CPU;
-}

resources/cudaForStorm/src/main.cpp

@@ -1,69 +0,0 @@
-#include "cudaForStorm.h"
-
-#include <stdio.h>  
-#include <stdlib.h>
-
-#include <iostream>
-#include <chrono>
-#include <random>
-
-#include "cudaTests.h"
-
-int cudaForStormTest(int value) {
-	return value + 42;
-}
-
-
-int main_Test12345(int argc, char **argv){
-	resetCudaDevice();
-
-	int testNumber = 0;
-	int N = 10000;
-	int M = 402653184;
-	if (argc > 1) {
-		testNumber = atoi(argv[1]);
-		if (argc > 2) {
-			N = atoi(argv[2]);
-			if (argc > 3) {
-				M = atoi(argv[3]);
-			}
-		}
-	}
-
-	switch (testNumber) {
-		case 1:
-			cudaSimpleAddTest(N, M);
-			break;
-		case 2:
-			cudaArrayFmaTest(N);
-			break;
-		case 3:
-			cudaArrayFmaOptimizedTest(N, M);
-			break;
-		case 4:
-			cpp_cuda_bandwidthTest(M, N);
-			break;
-		case 5:
-			kernelSwitchTest(N);
-			break;
-			break;
-		// DEFAULT AND 0
-		case 0:
-		default:
-			std::cout << "Available functions are:" << std::endl;
-			std::cout << "0 - Show this  overview" << std::endl;
-			std::cout << "1 - cuda   simpleAddTest(N, M)" << std::endl;
-			std::cout << "2 - cuda   arrayFmaTest(N)" << std::endl;
-			std::cout << "3 - cuda   arrayFmaOptimizedTest(N, M)" << std::endl;
-			std::cout << "4 - cuda   bandwidthTest(M, N)" << std::endl;
-			std::cout << "5 - cuda   kernelSwitchTest(N)" << std::endl;
-			std::cout << std::endl;
-			std::cout << "Call: " << argv[0] << " Selection [N [M]]" << std::endl;
-			std::cout << "Defaults:" <<std::endl;
-			std::cout << "N: 10000" << std::endl;
-			std::cout << "M: 402653184" << std::endl;
-			break;
-	}
-
-    return 0;
-}

resources/cudaForStorm/srcCuda/basicValueIteration.cu

@@ -11,8 +11,12 @@ __global__ void cuda_kernel_basicValueIteration_mvReduce(int const * const A, in
 	*B = *A;
 }
 
+void cudaForStormTestFunction(int a, int b) {
+	std::cout << "Cuda for Storm: a + b = " << (a+b) << std::endl;
+}
+
 void basicValueIteration_mvReduce(uint_fast64_t const maxIterationCount, std::vector<uint_fast64_t> const& matrixRowIndices, std::vector<uint_fast64_t> const& matrixColumnIndices, std::vector<double> const& matrixValues, std::vector<double>& x, std::vector<double> const& b, std::vector<uint_fast64_t> const& nondeterministicChoiceIndices) {
-	std::cout << "basicValueIteration_mvReduce is implemented for ValueType == double :)" << std::endl;
+	
 }
 
 /*

resources/cudaForStorm/srcCuda/basicValueIteration.h

@@ -1,4 +1,8 @@
 #include <cstdint>
 #include <vector>
 
-void basicValueIteration_mvReduce(uint_fast64_t const maxIterationCount, std::vector<uint_fast64_t> const& matrixRowIndices, std::vector<uint_fast64_t> const& matrixColumnIndices, std::vector<double> const& matrixValues, std::vector<double>& x, std::vector<double> const& b, std::vector<uint_fast64_t> const& nondeterministicChoiceIndices);
+// Library exports
+#include "cudaForStorm_Export.h"
+
+cudaForStorm_EXPORT void cudaForStormTestFunction(int a, int b);
+cudaForStorm_EXPORT void basicValueIteration_mvReduce(uint_fast64_t const maxIterationCount, std::vector<uint_fast64_t> const& matrixRowIndices, std::vector<uint_fast64_t> const& matrixColumnIndices, std::vector<double> const& matrixValues, std::vector<double>& x, std::vector<double> const& b, std::vector<uint_fast64_t> const& nondeterministicChoiceIndices);

resources/cudaForStorm/srcCuda/cudaForStorm.h

@@ -0,0 +1,14 @@
+#ifndef STORM_CUDAFORSTORM_CUDAFORSTORM_H_
+#define STORM_CUDAFORSTORM_CUDAFORSTORM_H_
+
+/*
+ * List of exported functions in this library
+ */
+
+// TopologicalValueIteration
+#include "srcCuda/basicValueIteration.h"
+
+
+
+
+#endif // STORM_CUDAFORSTORM_CUDAFORSTORM_H_

src/storm.cpp

@@ -62,6 +62,8 @@
 
 #include "src/exceptions/InvalidSettingsException.h"
 
+#include "cudaForStorm.h"
+
 #include <iostream>
 #include <iomanip>
 #include <fstream>
@@ -160,6 +162,7 @@ void printHeader(const int argc, const char* argv[]) {
 	}
 	std::cout << "Command line: " << commandStream.str() << std::endl;
 	std::cout << "Current working directory: " << getCurrentWorkingDirectory() << std::endl << std::endl;
+	cudaForStormTestFunction(21, 21);
 }
 
 /*!