《cuda by example》 book.h 头文件解析

cpu_bitmap.h 头函数解析

               blog.csdn.net/ghost129/article/details/4409565

都是一些函数，等用到的时候再过来解析，哈哈！

#ifndef __BOOK_H__
#define __BOOK_H__
#include <stdio.h>

/* cuda检错程序 */

static void HandleError( cudaError_t err,const char *file,int line )

{

    if (err != cudaSuccess)

    {

        printf( "%s in %s at line %d\n", cudaGetErrorString( err ),
                file, line );
        exit( EXIT_FAILURE );
    }
}
#define HANDLE_ERROR( err ) (HandleError( err, __FILE__, __LINE__ ))


#define HANDLE_NULL( a ) {if (a == NULL) { \
                            printf( "Host memory failed in %s at line %d\n", \
                                    __FILE__, __LINE__ ); \
                            exit( EXIT_FAILURE );}}

template< typename T >
void swap( T& a, T& b ) {
    T t = a;
    a = b;
    b = t;
}


void* big_random_block( int size ) {
    unsigned char *data = (unsigned char*)malloc( size );
    HANDLE_NULL( data );
    for (int i=0; i<size; i++)
        data[i] = rand();

    return data;
}

int* big_random_block_int( int size ) {
    int *data = (int*)malloc( size * sizeof(int) );
    HANDLE_NULL( data );
    for (int i=0; i<size; i++)
        data[i] = rand();

    return data;
}


// a place for common kernels - starts here

__device__ unsigned char value( float n1, float n2, int hue ) {
    if (hue > 360)      hue -= 360;
    else if (hue < 0)   hue += 360;

    if (hue < 60)
        return (unsigned char)(255 * (n1 + (n2-n1)*hue/60));
    if (hue < 180)
        return (unsigned char)(255 * n2);
    if (hue < 240)
        return (unsigned char)(255 * (n1 + (n2-n1)*(240-hue)/60));
    return (unsigned char)(255 * n1);
}

__global__ void float_to_color( unsigned char *optr,
                              const float *outSrc ) {
    // map from threadIdx/BlockIdx to pixel position
    int x = threadIdx.x + blockIdx.x * blockDim.x;
    int y = threadIdx.y + blockIdx.y * blockDim.y;
    int offset = x + y * blockDim.x * gridDim.x;

    float l = outSrc[offset];
    float s = 1;
    int h = (180 + (int)(360.0f * outSrc[offset])) % 360;
    float m1, m2;

    if (l <= 0.5f)
        m2 = l * (1 + s);
    else
        m2 = l + s - l * s;
    m1 = 2 * l - m2;

    optr[offset*4 + 0] = value( m1, m2, h+120 );
    optr[offset*4 + 1] = value( m1, m2, h );
    optr[offset*4 + 2] = value( m1, m2, h -120 );
    optr[offset*4 + 3] = 255;
}

__global__ void float_to_color( uchar4 *optr,
                              const float *outSrc ) {
    // map from threadIdx/BlockIdx to pixel position
    int x = threadIdx.x + blockIdx.x * blockDim.x;
    int y = threadIdx.y + blockIdx.y * blockDim.y;
    int offset = x + y * blockDim.x * gridDim.x;

    float l = outSrc[offset];
    float s = 1;
    int h = (180 + (int)(360.0f * outSrc[offset])) % 360;
    float m1, m2;

    if (l <= 0.5f)
        m2 = l * (1 + s);
    else
        m2 = l + s - l * s;
    m1 = 2 * l - m2;

    optr[offset].x = value( m1, m2, h+120 );
    optr[offset].y = value( m1, m2, h );
    optr[offset].z = value( m1, m2, h -120 );
    optr[offset].w = 255;
}


#if _WIN32
    //Windows threads.
    #include <windows.h>

    typedef HANDLE CUTThread;
    typedef unsigned (WINAPI *CUT_THREADROUTINE)(void *);

    #define CUT_THREADPROC unsigned WINAPI
    #define  CUT_THREADEND return 0

#else
    //POSIX threads.
    #include <pthread.h>

    typedef pthread_t CUTThread;
    typedef void *(*CUT_THREADROUTINE)(void *);

    #define CUT_THREADPROC void
    #define  CUT_THREADEND
#endif

//Create thread.
CUTThread start_thread( CUT_THREADROUTINE, void *data );

//Wait for thread to finish.
void end_thread( CUTThread thread );

//Destroy thread.
void destroy_thread( CUTThread thread );

//Wait for multiple threads.
void wait_for_threads( const CUTThread *threads, int num );

#if _WIN32
    //Create thread
    CUTThread start_thread(CUT_THREADROUTINE func, void *data){
        return CreateThread(NULL, 0, (LPTHREAD_START_ROUTINE)func, data, 0, NULL);
    }

    //Wait for thread to finish
    void end_thread(CUTThread thread){
        WaitForSingleObject(thread, INFINITE);
        CloseHandle(thread);
    }

    //Destroy thread
    void destroy_thread( CUTThread thread ){
        TerminateThread(thread, 0);
        CloseHandle(thread);
    }

    //Wait for multiple threads
    void wait_for_threads(const CUTThread * threads, int num){
        WaitForMultipleObjects(num, threads, true, INFINITE);

        for(int i = 0; i < num; i++)
            CloseHandle(threads[i]);
    }

#else
    //Create thread
    CUTThread start_thread(CUT_THREADROUTINE func, void * data){
        pthread_t thread;
        pthread_create(&thread, NULL, func, data);
        return thread;
    }

    //Wait for thread to finish
    void end_thread(CUTThread thread){
        pthread_join(thread, NULL);
    }

    //Destroy thread
    void destroy_thread( CUTThread thread ){
        pthread_cancel(thread);
    }

    //Wait for multiple threads
    void wait_for_threads(const CUTThread * threads, int num){
        for(int i = 0; i < num; i++)
            end_thread( threads[i] );
    }

#endif




#endif  // __BOOK_H__

参考书籍 ：《cuda by example》

参考链接 ： http://code.google.com/p/snp-gpgpu/source/browse/trunk/f/common/book.h?r=4