#include "stdafx.h"

#define NUM_INTS    32
// includes, CUDA
#include <cuda_runtime.h>
#include <cuda.h>
// includes, Project

typedef struct _CUDAContext_st {
    CUcontext   hcuContext;
    CUmodule    hcuModule;
    CUfunction  hcuFunction;
    CUdeviceptr dptr;
    int            deviceID;
} CUDAContext;//CUDA Context结构体

extern "C" void InitCUDAContext( CUDAContext *pContext, CUdevice hcuDevice, int deviceID );


extern "C" int InitCUDA(CUDAContext * cu_context)
{
    int cDevices;
    int hcuDevice = 0;
    CUresult status;
    //调用其它函数前 必须初始化API
    status = cuInit(0);
    // 获得所有设备数量
    status = cuDeviceGetCount( &cDevices );
    if ( CUDA_SUCCESS != status )
        return 1;
    if ( cDevices == 0 ) {
       return 1;//无硬件设备
    }
    //初始化所有硬件设备（多显卡）因为只有一块显卡 故只执行一次
     for ( int iDevice = 0; iDevice < cDevices; iDevice++ ) {
        char szName[256];
        //获得当前设备句柄返回指针到 hcuDevice
        status = cuDeviceGet( &hcuDevice, iDevice );
        if ( CUDA_SUCCESS != status )
            return 1;
        //返回当前设备名称 最大大小为256
        status = cuDeviceGetName( szName, 256, hcuDevice );
        if ( CUDA_SUCCESS != status )
            return 1;
        // 获得设备的内部参数
        //CUdevprop devProps;
        //CUDA_SUCCESS == cuDeviceGetProperties( &devProps, hcuDevice )


        // 初始化 CUDA Context
       InitCUDAContext( cu_context, hcuDevice, iDevice );
     }
}

//////////////////////////////////////////////////////
//                                                    //
//    pContext 自定义结构体 包括Context                //
//                                                    //
//                                                    //
//  初始化CUDA context                                //
//                                                    //
//                                                    //
//                                                    //
//                                                    //
//                                                    //
//////////////////////////////////////////////////////
void InitCUDAContext( CUDAContext *pContext, CUdevice hcuDevice, int deviceID )
{


    CUcontext  hcuContext  = 0;
    CUmodule   hcuModule   = 0;
    CUfunction hcuFunction = 0;

    CUresult status = cuCtxCreate( &hcuContext, 0, hcuDevice );
    if ( CUDA_SUCCESS != status ) {
        //创建线程失败
    }
    status = CUDA_ERROR_INVALID_IMAGE;

    // 将模块加载到当前Context中
    status = cuModuleLoad(&hcuModule,"iradon_kernel.cubin");
    // 返回当前模块中模块中指向函数的指针
    status = cuModuleGetFunction( &hcuFunction, hcuModule, "CUDA_iradon" );
    // 
    CUdeviceptr dptr;
    status = cuMemAlloc( &dptr, NUM_INTS*sizeof(int) );
    
    // Here we must release the CUDA context from the thread context 
    status = cuCtxPopCurrent( NULL );

    pContext->hcuContext  = hcuContext;
    pContext->hcuModule   = hcuModule;
    pContext->hcuFunction = hcuFunction;
    pContext->dptr        = dptr;
    pContext->deviceID    = deviceID;
}



void ThreadProc(CUDAContext *pParams)
{
    int wrong = 0;
    // 将CUDA context 放入 当前的线程中
    // cuCtxPushCurrent: Attach the caller CUDA context to the thread context. 
    CUresult status = cuCtxPushCurrent( pParams->hcuContext );

    // 设置Context中线程块的维数x,y,z
    cuFuncSetBlockShape( pParams->hcuFunction, 32, 1, 1 );

    // 为kernel相关的Function设置整型参量 
    // cuParamSeti(CUfunction func,int offset,unsigned int value);
    cuParamSeti( pParams->hcuFunction, 0,10 );
    cuParamSeti( pParams->hcuFunction, sizeof(int),10);

    //为函数中的参数设定总参数大小
    cuParamSetSize( pParams->hcuFunction, sizeof(int)*2 );

    // cuLaunch: we kick off the CUDA "kernelFunction"
    // 启动相关函数
    status = cuLaunch( pParams->hcuFunction );
    //if ( CUDA_SUCCESS == cuMemcpyDtoH( pInt, pParams->dptr, NUM_INTS*sizeof(int) ) );
    //cuMemFree( pParams->dptr );
    //销毁当前线程中正在使用的CUDA context
    // cuCtxPopCurrent: Detach the current CUDA context from the calling thread.
    cuCtxPopCurrent( NULL );
}

void InitCUDA_Radon()
{
    CUDAContext cuda_context;
    InitCUDA(&cuda_context);
    ThreadProc(&cuda_context);
}