First commit OpenFPM_devices

configure.ac

+#                                               -*- Autoconf -*-
+# Process this file with autoconf to produce a configure script.
+
+
+AC_PREREQ(2.59)
+AC_INIT(FULL-PACKAGE-NAME, VERSION, BUG-REPORT-ADDRESS)
+AC_CANONICAL_SYSTEM
+AC_CONFIG_SRCDIR([src/main.cpp])
+AM_INIT_AUTOMAKE
+AC_CONFIG_HEADER([src/config.h])
+m4_ifdef([MYSQL_FOUND],,[m4_include([m4/ax_lib_mysql.m4])])
+m4_ifdef([AX_CHECK_COMPILER_FLAGS],,[m4_include([m4/ax_check_compiler_flags.m4])])
+m4_ifdef([ACX_PTHREAD],,[m4_include([m4/acx_pthread.m4])])
+m4_ifdef([AX_CHECK_CL],,[m4_include([m4/ax_opencl.m4])])
+m4_ifdef([AX_BOOST_BASE],,[m4_include([m4/ax_boost_base.m4])])
+m4_ifdef([AX_BOOST_PROGRAM_OPTIONS],,[m4_include([m4/ax_boost_program_options.m4])])
+m4_ifdef([AX_BOOST_THREAD],,[m4_include([m4/ax_boost_thread.m4])])
+m4_ifdef([ACX_MPI],,[m4_include([m4/acx_mpi.m4])])
+m4_ifdef([AX_OPENMP],,[m4_include([m4/ax_openmp.m4])])
+m4_ifdef([AX_GCC_X86_CPUID],,[m4_include([m4/ax_gcc_x86_cpuid.m4])])
+m4_ifdef([AX_GCC_ARCHFLAG],,[m4_include([m4/ax_gcc_archflag.m4])])
+m4_ifdef([AX_CUDA],,[m4_include([m4/ax_cuda.m4])])
+
+CXXFLAGS+=" --std=c++11 "
+NVCCFLAGS=" "
+
+# Checks for programs.
+AC_PROG_CXX
+
+# Checks g++ flags
+
+AC_CANONICAL_HOST
+
+# Check target architetture
+
+AX_GCC_ARCHFLAG([], [CXXFLAGS="$CXXFLAGS $ax_cv_gcc_archflag"], [])
+
+###### Check for debug compilation
+
+AC_MSG_CHECKING(whether to build with debug information)
+debuger=no
+AC_ARG_ENABLE(debug,
+	AC_HELP_STRING(
+	    [--enable-debug],
+	    [enable debug data generation (def=no)]
+	),
+  	debuger="$enableval"
+)
+
+
+
+AC_MSG_RESULT($debuger)
+if test x"$debuger" = x"yes"; then
+	AC_DEFINE([DEBUG_MODE],[],[Debug])
+	AC_DEFINE([DEBUG],[],[Debug])
+	CXXFLAGS="$CXXFLAGS -g3  -Wall -O0 "
+	NVCCFLAGS+="$NVCCFLAGS -g -O0 "
+else
+	CXXFLAGS="$CXXFLAGS -Wall -O3 -g3 "
+	NVCCFLAGS+="$NVCCFLAGS -O3 "
+fi
+
+####### Checking for GPU support
+
+AX_CUDA
+
+AC_MSG_CHECKING(whether to build with GPU support)
+gpu_support=no
+AC_ARG_ENABLE(gpu,
+	AC_HELP_STRING(
+	    [--enable-gpu],
+	    [enable gpu support]
+	),
+  	gpu_support="$enableval"
+)
+
+
+
+AC_MSG_RESULT($gpu_support)
+if test x"$gpu_support" = x"yes"; then
+	AC_DEFINE([GPU],[],[GPU support])
+fi
+
+
+# Set this conditional if cuda is wanted
+
+AM_CONDITIONAL(BUILDCUDA, test ! x$NVCC = x"no")
+
+###########################
+
+no_avx=no
+no_sse42=no
+no_sse41=no
+no_sse3=no
+no_sse2=no
+no_sse=no
+no_mmx=no
+AX_CHECK_COMPILER_FLAGS([-msse4.2],[CXXFLAGS="$CXXFLAGS -mavx"],[no_avx=yes])
+AX_CHECK_COMPILER_FLAGS([-msse4.2],[CXXFLAGS="$CXXFLAGS -msse4.2"],[no_sse42=yes])
+AX_CHECK_COMPILER_FLAGS([-msse4.1],[CXXFLAGS="$CXXFLAGS -msse4.1"],[no_sse41=yes])
+AX_CHECK_COMPILER_FLAGS([-msse3],[CXXFLAGS="$CXXFLAGS -msse3"],[no_sse3=yes])
+AX_CHECK_COMPILER_FLAGS([-msse2],[CXXFLAGS="$CXXFLAGS -msse2"],[no_sse2=yes])
+AX_CHECK_COMPILER_FLAGS([-msse],[CXXFLAGS="$CXXFLAGS -msse"],[no_sse=yes])
+AX_CHECK_COMPILER_FLAGS([-mmmx],[CXXFLAGS="$CXXFLAGS -mmmx"],[no_mmx=yes])
+AX_CHECK_COMPILER_FLAGS([-Wno-unused-but-set-variable],[CXXFLAGS="$CXXFLAGS -Wno-unused-but-set-variable"],[])
+
+AC_SUBST(NVCCFLAGS)
+
+# Checks for typedefs, structures, and compiler characteristics.
+
+# Checks for library functions.
+
+AC_CONFIG_FILES([Makefile src/Makefile])
+AC_OUTPUT
+echo ""
+echo   "***********************************"
+echo   "*                                 *"
+arch_str="${ax_cv_gcc_archflag#-march=#-mtune=}"
+arch_str="${arch_str#-mtune=}"
+n_arch_str=${#arch_str}
+for (( X=0; X<23-n_arch_str; X++ ))
+do
+   arch_str="$arch_str "
+done
+echo "*    arch: $arch_str*"
+if [ test x"$no_sse42" = x"no" ]; then
+    echo "*    sse4.2: yes                  *"
+else
+    echo "*    sse4.2: no                   *"
+fi
+if [ test x"$no_sse41" = x"no" ]; then
+    echo "*    sse4.1: yes                  *"
+else
+    echo "*    sse4.1: no                   *"
+fi
+if [ test x"$no_sse3" = x"no" ]; then
+    echo "*    sse3: yes                    *"
+else
+    echo "*    sse3: no                     *"
+fi
+if [ test x"$no_sse2" = x"no" ]; then
+    echo "*    sse2: yes                    *"
+else
+    echo "*    sse2: no                     *"
+fi
+if [ test x"$no_sse" = x"no" ]; then
+    echo "*    sse: yes                     *"
+else
+    echo "*    sse: no                      *"
+fi
+if [ test x"$no_mmx" = x"no" ]; then
+  echo "*    mmx: yes                     *"
+else
+  echo "*    mmx: no                      *"
+fi
+if [ test x"$profiler" = x"yes" ]; then
+  echo "*    profiler: yes                *"
+else
+  echo "*    profiler: no                 *"
+fi
+if [ test x"$debuger" = x"yes" ]; then
+  echo "*    debug: yes                   *"
+else
+  echo "*    debug: no                    *"
+fi
+if [ test x"$no_64" = x"no" ]; then
+  echo "*    64 bit: yes                  *"
+else
+  echo "*    64 bit: no                   *"
+fi
+if [ test x"$gpu_support" = x"no" ]; then
+  echo "*    gpu: no                      *"
+else
+  echo "*    gpu: yes                     *"
+fi
+echo   "*                                 *"
+echo   "***********************************"
+

install_MPI.sh

+#!/bin/bash 
+
+mkdir ${HOME}/MPI
+wget http://www.open-mpi.de/software/ompi/v1.8/downloads/openmpi-1.8.1.tar.bz2
+tar -xvf openmpi-1.8.1.tar.bz2
+cd openmpi-1.8.1
+sh ./configure --prefix=${HOME}/MPI --enable-opal-multi-threads --enable-mpi-f90
+make
+make install

src/Makefile.am

+
+LINKLIBS = $(PTHREAD_LIBS) $(OPT_LIBS) $(BOOST_LDFLAGS) $(BOOST_PROGRAM_OPTIONS_LIB) $(CUDA_LIBS) $(BOOST_THREAD_LIB)
+
+bin_PROGRAMS = ppline
+ppline_SOURCES = main.cpp
+ppline_CXXFLAGS = $(CUDA_CFLAGS)
+ppline_CFLAGS = $(CUDA_CFLAGS)
+ppline_LDADD = $(LINKLIBS) -L/usr/lib64/nvidia-bumblebee/
+
+.cu.o :
+	$(NVCC) $(NVCCFLAGS) -o $@ -c $<

src/cuda_macro.h

+/*
+ * cuda_macro.h
+ *
+ *  Created on: Aug 17, 2014
+ *      Author: Pietro Incardona
+ */
+
+#include <iostream>
+
+#define CUDA_SAFE_CALL(call) {\
+	cudaError_t err = call;\
+	if (cudaSuccess != err) {\
+		std::cerr << "Cuda error in file "<< __FILE__ << " in line " << __LINE__ <<  ": " << cudaGetErrorString(err);\
+	}\
+}
+
+

src/instructions/memory/CudaMemory.cu

+/*
+ * OpenFPMwdeviceCudaMemory.cu
+ *
+ *  Created on: Aug 11, 2014
+ *      Author: Pietro Incardona
+ */
+
+#include <cstddef>
+#include <cuda_runtime.h>
+#include "CudaMemory.cuh"
+
+/*! \brief Allocate a chunk of memory
+ *
+ * Allocate a chunk of memory
+ *
+ * \param sz size of the chunk of memory to allocate in byte
+ *
+ */
+bool CudaMemory::allocate(size_t sz)
+{
+	//! Allocate the device memory
+	if (dm == NULL)
+	{dv = new boost::shared_ptr<void>(new thrust::device_vector<void>(sz));}
+}
+
+void CudaMemory::destroy()
+{
+	dv = NULL;
+}
+
+void CudaMemory::copyFromPointer(ThreadWorker t)
+{
+	// check if we have a host buffer, if not allocate it
+
+	// put on queue a copy from device to host
+
+	t.call();
+
+	// put on queue a memory copy from pointers
+}
+
+void CudaMemory::copyDeviceToDevice(ThreadWorker t)
+{
+	// put on queue a copy from device to device
+
+	t.call();
+}
+
+bool CudaMemory::copy(memory m, ThreadWorker t)
+{
+	//! Here we try to cast memory into OpenFPMwdeviceCudaMemory
+	CudaMemory * ofpm = dynamic_cast<CudaMemory>(m);
+
+	//! if we fail we get the pointer and simply copy from the pointer
+
+	if (ofpm == NULL)
+	{
+		// copy the memory from device to host and from host to device
+
+		copyFromPointer(t);
+	}
+	else
+	{
+		// they are the same memory type, use cuda/thrust buffer copy
+
+		copyDeviceToDevice();
+	}
+}
+
+bool CudaMemory::copy(OpenFPMwdeviceCudaMemory m)
+{
+	// they are the same type of memory so copy from device to device
+
+	copyDeviceToDevice();
+}
+
+size_t CudaMemory::size()
+{
+	dv->size();
+}
+
+bool CudaMemory::resize(size_t sz)
+{
+	//! Allocate the device memory
+	if (dv == NULL)
+	{dv = new boost::shared_ptr<void>(new thrust::device_vector<void>());}
+	else
+	{dv.get()->resize(sz);}
+}

src/instructions/memory/CudaThrustMemory.cuh

+/*
+ * OpenFPMwdeviceCudaMemory.h
+ *
+ *  Created on: Aug 8, 2014
+ *      Author: Pietro Incardona
+ */
+
+#include <thrust/host_vector.h>
+#include <thrust/device_vector.h>
+#include <thrust/generate.h>
+#include <thrust/sort.h>
+#include <thrust/copy.h>
+#include <boost/shared_ptr>
+
+/**
+ * \brief This class create instructions to allocate, and destroy GPU memory
+ * 
+ * This class create instructions to allocate, destroy, resize GPU buffer, 
+ * eventually if direct, comunication is not supported, it can instruction
+ * to create an Host Pinned memory.
+ * 
+ * Usage:
+ * 
+ * TO DO
+ * 
+ * This class in general is used by OpenFPM_data project to basically
+ * record all the operation made and generate a set of instructions
+ * 
+ */
+
+class CudaMemory : public memory
+{
+	//! 
+	
+	//! device memory
+	void * dm;
+	
+	//! allocate memory
+	virtual bool allocate(size_t sz);
+	//! destroy memory
+	virtual void destroy();
+	//! copy memory
+	virtual bool copy(memory m);
+	//! the the size of the allocated memory
+	virtual size_t size();
+	//! resize the momory allocated
+	virtual bool resize(size_t sz);
+	
+};
+
+

src/main.cpp

+#include <iostream.h>
+
+int main()
+{
+
+}

src/memory/CudaMemory.cu

+#include <cstddef>
+#include <cuda_runtime.h>
+#include "CudaMemory.cuh"
+#include "cuda_macro.h"
+
+/*! \brief Allocate a chunk of memory
+ *
+ * Allocate a chunk of memory
+ *
+ * \param sz size of the chunk of memory to allocate in byte
+ *
+ */
+bool CudaMemory::allocate(size_t sz)
+{
+	//! Allocate the device memory
+	if (dm == NULL)
+	{CUDA_SAFE_CALL(cudaAlloc(&dv,sz,cudaAlloc));}
+}
+
+/*! \brief destroy a chunk of memory
+ *
+ * Destroy a chunk of memory
+ *
+ */
+void CudaMemory::destroy()
+{
+	CUDA_SAFE_CALL(cudaDestroy(dv));
+}
+
+/*! \brief copy the device memory to a pointer
+ *
+ * copy the device memory to a pointer
+ *
+ */
+void CudaMemory::copyToPointer()
+{
+	// check if we have a host buffer, if not allocate it
+
+	if (hm == NULL)
+	{CUDA_SAFE_CALL(cudaHostAlloc(&hm,sz*sizeof(T),cudaHostAllocMapped))}
+
+	// put on queue a copy from device to host
+
+	t.call();
+
+	// put on queue a memory copy from pointers
+}
+
+void CudaMemory::copyDeviceToDevice()
+{
+	// put on queue a copy from device to device
+
+	t.call();
+}
+
+bool CudaMemory::copy(memory m)
+{
+	//! Here we try to cast memory into OpenFPMwdeviceCudaMemory
+	CudaMemory * ofpm = dynamic_cast<CudaMemory>(m);
+
+	//! if we fail we get the pointer and simply copy from the pointer
+
+	if (ofpm == NULL)
+	{
+		// copy the memory from device to host and from host to device
+
+		copyFromPointer(t);
+	}
+	else
+	{
+		// they are the same memory type, use cuda/thrust buffer copy
+
+		copyDeviceToDevice();
+	}
+}
+
+bool CudaMemory::copy(OpenFPMwdeviceCudaMemory m)
+{
+	// they are the same type of memory so copy from device to device
+
+	copyDeviceToDevice();
+}
+
+size_t CudaMemory::size()
+{
+	dv->size();
+}
+
+bool CudaMemory::resize(size_t sz)
+{
+	//! Allocate the device memory
+	if (dv == NULL)
+	{dv = new boost::shared_ptr<void>(new thrust::device_vector<void>());}
+	else
+	{dv.get()->resize(sz);}
+}

src/memory/CudaMemory.cuh

+/*
+ * CudaMemory.cu
+ *
+ *  Created on: Aug 17, 2014
+ *      Author: Pietro Incardona
+ */
+
+/**
+ * \brief This class create instructions to allocate, and destroy GPU memory
+ * 
+ * This class create instructions to allocate, destroy, resize GPU buffer, 
+ * eventually if direct, comunication is not supported, it can instruction
+ * to create an Host Pinned memory.
+ * 
+ * Usage:
+ * 
+ * TO DO
+ * 
+ * This class in general is used by OpenFPM_data project to basically
+ * record all the operation made and generate a set of instructions
+ * 
+ */
+
+class CudaMemory : public memory
+{
+	//! 
+	
+	//! device memory
+	void * dm;
+	
+	//! host memory
+	void * hm;
+	
+	//! allocate memory
+	virtual bool allocate(size_t sz);
+	//! destroy memory
+	virtual void destroy();
+	//! copy memory
+	virtual bool copy(memory m);
+	//! the the size of the allocated memory
+	virtual size_t size();
+	//! resize the momory allocated
+	virtual bool resize(size_t sz);
+	
+};
+

src/memory/HeapMemory.cpp

+/*
+ * HeapMemory.cpp
+ *
+ *  Created on: Aug 17, 2014
+ *      Author: Pietro Incardona
+ */
+
+#include "HeapMemory.hpp"
+
+

src/memory/HeapMempory.hpp

+/*
+ * HeapMempory.hpp
+ *
+ *  Created on: Aug 17, 2014
+ *      Author: Pietro Incardona
+ */
+
+class HeapMemory
+{
+
+};
+