狮子OpenCL上的Xcode 4.2不断与EXC_BAD_ACCESS暂停

Question

因此，我在Lion上运行Xcode 4.1中的一些简单的Hello World OpenCL代码，并且它不断在clEnqueueTask中断。当我从MacResearch.org OpenCL tutorials运行源文件时，发生同样的事情，该文件在clEnqueueNDRangeKernel处中断。 lldb给出代码1，地址0x30。

下面是代码：

#include <stdio.h> 
#include <stdlib.h> 

#include <OpenCL/opencl.h> 

#define MEM_SIZE (128) 
#define MAX_SOURCE_SIZE (0x100000) 

int main() 
{ 

    char *program_source = "\n"\ 
    "__kernel void hello(__global char* string) \n"\ 
    "{           \n"\ 
    " string[0] = 'H';      \n"\ 
    " string[1] = 'e';      \n"\ 
    " string[2] = 'l';      \n"\ 
    " string[3] = 'l';      \n"\ 
    " string[4] = 'o';      \n"\ 
    " string[5] = ',';      \n"\ 
    " string[6] = ' ';      \n"\ 
    " string[7] = 'w';      \n"\ 
    " string[8] = 'o';      \n"\ 
    " string[9] = 'r';      \n"\ 
    " string[10] = 'l';      \n"\ 
    " string[11] = 'd';      \n"\ 
    " string[12] = '!';      \n"\ 
    " string[13] = '\0';      \n"\ 
    "}           \n"\ 
    "\n"; 

    size_t source_size = sizeof(program_source); 

    cl_device_id device_id   = NULL; 
    cl_context context    = NULL; 
    cl_command_queue command_queue = NULL; 
    cl_mem memobj     = NULL; 
    cl_program program    = NULL; 
    cl_kernel kernel    = NULL; 
    cl_platform_id platform_id  = NULL; 
    cl_uint ret_num_devices; 
    cl_uint ret_num_platforms; 
    cl_int ret; 

    char string[MEM_SIZE]; 


    // get platform and device information 
    ret = clGetPlatformIDs(1, &platform_id, &ret_num_platforms); 
    ret = clGetDeviceIDs(platform_id, CL_DEVICE_TYPE_GPU, 1, &device_id, &ret_num_devices); 

    cl_int err = 0; 
    size_t returned_size = 0; 
    size_t buffer_size; 

    // Get some information about the returned device 
    cl_char vendor_name[1024] = {0}; 
    cl_char device_name[1024] = {0}; 
    err = clGetDeviceInfo(device_id, CL_DEVICE_VENDOR, sizeof(vendor_name), vendor_name, &returned_size); 
    err |= clGetDeviceInfo(device_id, CL_DEVICE_NAME, sizeof(device_name),device_name, &returned_size); 
// assert(err == CL_SUCCESS); 
    printf("Connecting to %s %s...\n", vendor_name, device_name); 

    // create OpenCL context 
    context = clCreateContext(NULL, 1, &device_id, NULL, NULL, &ret); 

    // create command queue 
    command_queue = clCreateCommandQueue(context, device_id, 0, &ret); 

    // create memory buffer 
    memobj = clCreateBuffer(context,CL_MEM_READ_WRITE, MEM_SIZE*sizeof(char), NULL, &ret); 

    // create kernel program from source code 
    program = clCreateProgramWithSource(context, 1, (const char **)&program_source, (const size_t*)&source_size, &ret); 

    // build kernel program 
    ret = clBuildProgram(program, 1, &device_id, NULL, NULL, NULL); 

    // create OpenCL Kernel 
    kernel = clCreateKernel(program, "hello", &ret); 

    // set OpenCL kernel parameters 
    ret = clSetKernelArg(kernel, 0, sizeof(cl_mem), (void *)&memobj); 

    // Execute OpenCL kernel 
    ret = clEnqueueTask(command_queue, kernel, 0, NULL, NULL); 

    // copy results from the memory buffer 
    ret = clEnqueueReadBuffer(command_queue, memobj, CL_TRUE, 0, MEM_SIZE*sizeof(char), string, 0, NULL, NULL); 

    // display results 
    puts(string); 

    // finish up 
    ret = clFlush(command_queue); 
    ret = clFinish(command_queue); 
    ret = clReleaseKernel(kernel); 
    ret = clReleaseProgram(program); 
    ret = clReleaseMemObject(memobj); 
    ret = clReleaseCommandQueue(command_queue); 
    ret = clReleaseContext(context); 

    return 0; 
} 

使用malloc的后卫尝试，得到：

GuardMalloc[OCL_HW-1453]: recording malloc stacks to disk using standard recorder 
GuardMalloc[OCL_HW-1453]: Allocations will be placed on 16 byte boundaries. 
GuardMalloc[OCL_HW-1453]: - Some buffer overruns may not be noticed. 
GuardMalloc[OCL_HW-1453]: - Applications using vector instructions (e.g., SSE) should work. 
GuardMalloc[OCL_HW-1453]: version 24.1 
OCL_HW(1453) malloc: process 1423 no longer exists, stack logs deleted from /tmp/stack-logs.1423.OCL_HW.yL5f5u.index 
OCL_HW(1453) malloc: stack logs being written into /tmp/stack-logs.1453.OCL_HW.pCjTNR.index 
Connecting to NVIDIA GeForce GT 330M... 

我曾与下雪豹和Xcode的3这些代码没有问题，我确信不编译任何.cl文件通过从目标中删除它们，并且'OpenCl.framework'被链接和一切。

我甚至擦干净了我的电脑和干净的狮子和xcode，但仍然是个问题。我很肯定在这一点上，这是愚蠢的。

- 感谢一堆

Answer 1

你是对的 - 这是愚蠢的。您正在向clCreateProgramWithSource的第四个参数传递一个不正确的值。你应该传递你的源字符串的长度，但是你传递的是指针的大小。你可以这样修复它：

size_t source_size = strlen(program_source); 

请注意，我通过检查clBuildProgram的返回值来发现此问题。它是-11，CL_BUILD_PROGRAM_FAILURE，这意味着你的内核编译失败。既然你的内核看起来很正常，我这样做是在命令行：

CL_LOG_ERRORS=stdout ./test 

这引起了苹果OpenCL实现转储编译器生成日志到标准输出。我看到这个：

[CL_BUILD_ERROR] : OpenCL Build Error : Compiler build log: 
<program source>:2:1: error: unknown type name '__kerne' 
__kerne 

<program source>:2:8: error: expected identifier or '(' 
__kerne 

这使我立即想到你的源代码长度参数有问题。

另外请注意，你需要在你的内核可以改变：

string[13] = '\0'; 

到

string[13] = 0; 

做了这些改变之后，我认为这对我的MacBook Pro：

Connecting to AMD ATI Radeon HD 6490M... 
Hello, world!