scan.h

Go to the documentation of this file.

#ifndef SCAN_H
#define SCAN_H



#include <starpu.h>
#ifdef SKEPU_OPENCL 
    #include <starpu_opencl.h>
#endif  



#ifdef SKEPU_OPENCL
#include <string>
#include <vector>
#ifdef USE_MAC_OPENCL
#include <OpenCL/opencl.h>
#else
#include <CL/cl.h>
#endif
#include "src/device_cl.h"
#endif

#include "src/environment.h"
#include "skepu/vector.h"
#include "skepu/task.h"
#include "src/operator_macros.h"
#include "src/exec_plan.h"

#ifdef SKEPU_CUDA
#include "src/device_mem_pointer_cu.h"
#include "src/device_cu.h"

#endif

namespace skepu
{

enum ScanType
{
    INCLUSIVE,
    EXCLUSIVE
};

template <typename ScanFunc, typename T>
class Scan: public Task
{

typedef struct
{
    Scan<ScanFunc,T> *pScan;
    bool unaryOperand;
    ScanType type;
    T initValue;
}ArgumentBuffer;

public:
    ScanType type;
    T initValue;
    Vector<T> *input;
    Vector<T> *output;

    Scan(ScanFunc* scanFunc);
    
    Scan(ScanFunc* scanFunc, Vector<T> *in, Vector<T> *out, ScanType type, T init = T());

    ~Scan();
    
    Vector<T>& getOutput()
    {
        return *output;
    } 
    
    void *get_result()
    {
        return output;
    }    

    void setInput(Vector<T> *in)
    {
        input = in;
    }
    
    void setOutput(Vector<T> *out)
    {
        output = out;
    } 
    
    void setInitValue(T _init)
    {
        initValue = _init;
    } 
    
    void run_async();

    void finishAll() {m_environment->finishAll();}
    void setExecPlan(ExecPlan& plan) {m_execPlan = plan;}

public:
    Environment<int>* m_environment;
    ScanFunc* m_scanFunc;
    ExecPlan m_execPlan;
    
    struct starpu_codelet *codelet;
    struct starpu_codelet *codelet_two;
    
    bool unaryOperand;

#ifdef USE_STARPU_HISTORY_MODEL    
    starpu_perfmodel *scan1_perf_model;
    starpu_perfmodel *scan2_perf_model;      
#endif    

public:
    void operator()(Vector<T>& input, ScanType type, T init = T());

    void operator()(Vector<T>& input, Vector<T>& output, ScanType type, T init = T());

public:
    static void cpu_func(void *buffers[], void *arg);

#ifdef SKEPU_OPENMP
public:
    static void omp_func(void *buffers[], void *arg);
#endif

#ifdef SKEPU_CUDA
public:
    static void cuda_func(void *buffers[], void *arg);

    T scanLargeVectorRecursively_CU(T* input, T* output, std::vector<DeviceMemPointer_CU<T>*>& blockSums, unsigned int numElements, int level, ScanType type, T init, int deviceID);
#endif

#ifdef SKEPU_OPENCL
public:
    static void opencl_func(void *buffers[], void *arg);

    T scanLargeVectorRecursively_CL(T* input, T* output, std::vector<DeviceMemPointer_CL<T>*>& blockSums, unsigned int numElements, int level, ScanType type, T init, Device_CL *deviceCL);
    
public:
    struct starpu_opencl_program opencl_codelet;
    char kernelNameScan[1024];
    char kernelNameScanUpdate[1024];
    char kernelNameScanAdd[1024];
    
    void replaceText(std::string& text, std::string find, std::string replace);
    void createOpenCLProgram();
#endif

    std::string perfmodel1_str;
    std::string perfmodel2_str;

};

}

#include "src/scan.inl"

#include "src/scan_cpu.inl"

#ifdef SKEPU_OPENMP
#include "src/scan_omp.inl"
#endif

#ifdef SKEPU_OPENCL
#include "src/scan_cl.inl"
#endif

#ifdef SKEPU_CUDA
#include "src/scan_cu.inl"
#endif

#endif