First OpenCL program

Question

So this is my first attempt at writing code targeting the OpenCL platform.

Trying to build a program that will actually execute AES on the OpenCL platform. However their is a segmentation fault upon program execution and I'm not familiar with the UNIX environment failures.

OS: Mac OS X (Snow Leopard)

gcc main.c -o main -framework OpenCL -std=c99

#ifdef __APPLE__
#include <OpenCL/opencl.h>
#else
#include <CL/cl.h>
#endif

#include <stdlib.h>     // warning: implicit declaration of function ‘malloc’
#include <stdio.h>      // error: ‘stderr’ undeclared (first use in this function)

int main(int argc, char **argv)
{


    unsigned char S[16] = { 0x32, 0x43, 0xf6, 0xa8, 0x88, 0x5a, 0x30, 0x8d, 0x31, 0x31, 0x98, 0xa2, 0xe0, 0x37, 0x07, 0x34};

    unsigned char Sbox[] = {
        0x63, 0x7C, 0x77, 0x7B, 0xF2, 0x6B, 0x6F, 0xC5, 0x30, 0x01, 0x67, 0x2B, 0xFE, 0xD7, 0xAB, 0x76,
        0xCA, 0x82, 0xC9, 0x7D, 0xFA, 0x59, 0x47, 0xF0, 0xAD, 0xD4, 0xA2, 0xAF, 0x9C, 0xA4, 0x72, 0xC0,
        0xB7, 0xFD, 0x93, 0x26, 0x36, 0x3F, 0xF7, 0xCC, 0x34, 0xA5, 0xE5, 0xF1, 0x71, 0xD8, 0x31, 0x15,
        0x04, 0xC7, 0x23, 0xC3, 0x18, 0x96, 0x05, 0x9A, 0x07, 0x12, 0x80, 0xE2, 0xEB, 0x27, 0xB2, 0x75,
        0x09, 0x83, 0x2C, 0x1A, 0x1B, 0x6E, 0x5A, 0xA0, 0x52, 0x3B, 0xD6, 0xB3, 0x29, 0xE3, 0x2F, 0x84,
        0x53, 0xD1, 0x00, 0xED, 0x20, 0xFC, 0xB1, 0x5B, 0x6A, 0xCB, 0xBE, 0x39, 0x4A, 0x4C, 0x58, 0xCF,
        0xD0, 0xEF, 0xAA, 0xFB, 0x43, 0x4D, 0x33, 0x85, 0x45, 0xF9, 0x02, 0x7F, 0x50, 0x3C, 0x9F, 0xA8,
        0x51, 0xA3, 0x40, 0x8F, 0x92, 0x9D, 0x38, 0xF5, 0xBC, 0xB6, 0xDA, 0x21, 0x10, 0xFF, 0xF3, 0xD2,
        0xCD, 0x0C, 0x13, 0xEC, 0x5F, 0x97, 0x44, 0x17, 0xC4, 0xA7, 0x7E, 0x3D, 0x64, 0x5D, 0x19, 0x73,
        0x60, 0x81, 0x4F, 0xDC, 0x22, 0x2A, 0x90, 0x88, 0x46, 0xEE, 0xB8, 0x14, 0xDE, 0x5E, 0x0B, 0xDB,
        0xE0, 0x32, 0x3A, 0x0A, 0x49, 0x06, 0x24, 0x5C, 0xC2, 0xD3, 0xAC, 0x62, 0x91, 0x95, 0xE4, 0x79,
        0xE7, 0xC8, 0x37, 0x6D, 0x8D, 0xD5, 0x4E, 0xA9, 0x6C, 0x56, 0xF4, 0xEA, 0x65, 0x7A, 0xAE, 0x08,
        0xBA, 0x78, 0x25, 0x2E, 0x1C, 0xA6, 0xB4, 0xC6, 0xE8, 0xDD, 0x74, 0x1F, 0x4B, 0xBD, 0x8B, 0x8A,
        0x70, 0x3E, 0xB5, 0x66, 0x48, 0x03, 0xF6, 0x0E, 0x61, 0x35, 0x57, 0xB9, 0x86, 0xC1, 0x1D, 0x9E,
        0xE1, 0xF8, 0x98, 0x11, 0x69, 0xD9, 0x8E, 0x94, 0x9B, 0x1E, 0x87, 0xE9, 0xCE, 0x55, 0x28, 0xDF,
        0x8C, 0xA1, 0x89, 0x0D, 0xBF, 0xE6, 0x42, 0x68, 0x41, 0x99, 0x2D, 0x0F, 0xB0, 0x54, 0xBB, 0x16
    };


    /* Retrieve Platforms */

    cl_uint Platforms = 0;

    if (CL_SUCCESS == clGetPlatformIDs ( 0, NULL, &Platforms))
    {

        if (Platforms > 0)
        {

            /* Retrieve Platform ID */

            cl_platform_id *Platform = malloc((sizeof(cl_platform_id) * Platforms));

            clGetPlatformIDs( Platforms, Platform, &Platforms);


            /* Retrieve Devices on Platform */

            cl_uint GPUs = 0;

            clGetDeviceIDs( Platform[0], CL_DEVICE_TYPE_GPU, 0, NULL, &GPUs);

            if (GPUs > 0)
            {

                cl_device_id  *GPU = malloc((sizeof(cl_device_id) * GPUs));

                clGetDeviceIDs( Platform[0], CL_DEVICE_TYPE_GPU, GPUs, GPU, &GPUs);



                cl_context GPUcontext = clCreateContextFromType( NULL, CL_DEVICE_TYPE_GPU, NULL, NULL, NULL);


                const char *program_source[] = {
                    "__kernel void OpenCLSubBytes( __global unsigned char *S, __global unsigned char *Sbox)\n",
                    "{\n",
                    "char thread = get_global_id(0);\n",
                    "S[thread] = Sbox[(S[thread])];\n",
                    "}\n"
                };


                cl_program AES = clCreateProgramWithSource( GPUcontext, sizeof(program_source)/sizeof(*program_source), program_source, NULL, NULL);

                if (clBuildProgram( AES, GPUs, &GPU[0], "", NULL, NULL) != CL_SUCCESS) {
                    char buffer[10240];

                    clGetProgramBuildInfo( AES, GPU[0], CL_PROGRAM_BUILD_LOG, sizeof(buffer), buffer, NULL);

                    fprintf(stderr, "CL Compilation failed:\n%s", buffer);
                    exit(2); // abort();
                }

                /* Since OpenCL compilation failed is due to incomplete code work */


                cl_mem S_OpenCL = clCreateBuffer( GPUcontext, CL_MEM_READ_WRITE, 16, S, NULL);

                cl_mem Sbox_OpenCL = clCreateBuffer( GPUcontext, CL_MEM_READ_ONLY, 256, Sbox, NULL);


                cl_kernel SubBytesKernel = clCreateKernel( AES, "OpenCLSubBytes", NULL);
                clSetKernelArg( SubBytesKernel, 0, 16, &S_OpenCL);      // warning: passing argument 4 discards qualifiers from pointer target type (const)
                clSetKernelArg( SubBytesKernel, 1, 256, &Sbox_OpenCL);

                cl_command_queue ExecutionQueue = clCreateCommandQueue( GPUcontext, GPU[0], 0, NULL);


                for ( int Loop = 0; Loop < 16; Loop++)
                {
                    clEnqueueWriteBuffer( ExecutionQueue, S_OpenCL, CL_TRUE, Loop, 1, &S, 0, NULL, NULL);
                }
                for ( int Loop = 0; Loop < 256; Loop++)
                {
                    clEnqueueWriteBuffer( ExecutionQueue, Sbox_OpenCL, CL_TRUE, Loop, 1, &Sbox, 0, NULL, NULL);
                }               


                cl_event ExecutionComplete;

                size_t global_work_size[1] = { 16 };

                clEnqueueNDRangeKernel( ExecutionQueue, SubBytesKernel, 1, NULL, global_work_size, NULL, 0, NULL, &ExecutionComplete);
                clWaitForEvents( 1, &ExecutionComplete);
                clReleaseEvent( ExecutionComplete);


                printf("S:");
                // for ( char Loop = 0; Loop < 16; Loop++)
                // {
                //  clEnqueueReadBuffer( ExecutionQueue, S_OpenCL, CL_TRUE, Loop, 1, &S, 0, NULL, NULL);

                //  printf(" %02hhx", S[Loop]);
                // }

                printf("\n");

                clReleaseMemObject(S_OpenCL);
                clReleaseMemObject(Sbox_OpenCL);

                clReleaseKernel(SubBytesKernel);
                clReleaseProgram(AES);
                clReleaseContext(GPUcontext);


            }

        }

    }

    return 0;
}

Answer 1

Your calls to clSetKernelArg() are incorrect. The third parameter is the size of the argument, not the size of the buffer that the argument points to. So, these lines:

clSetKernelArg( SubBytesKernel, 0, 16, &S_OpenCL);
clSetKernelArg( SubBytesKernel, 1, 256, &Sbox_OpenCL);

Should be changed to these lines:

clSetKernelArg( SubBytesKernel, 0, sizeof(cl_mem), &S_OpenCL);
clSetKernelArg( SubBytesKernel, 1, sizeof(cl_mem), &Sbox_OpenCL);

This fixed the segmentation fault on my own system (MBP 2013, OS X 10.9).

Answer 2

The most likely cause for the segfault you are observing is due to the way you are calling clCreateContextFromType: you are passing NULL as the first parameter. While with many OpenCL implementations this may work as expected, the OpenCL specification says the following:

properties can be NULL in which case the platform that is selected is implementation-defined

For Nvidia drivers, for example, this implementation-defined behavior consists of not returning any platform if no valid cl_context_properties array is passed. A similar issue was addressed in this post. Other vendor's implementations may exhibit similar behavior. In case of the code snippet you posted, the code continues without checking that a valid platform was actually returned, causing unexpected behavior if no valid platform was indeed returned.

Therefore I recommend always checking all OpenCL API error codes. This will make your code a bit more verbose but will save you time in the long run. For example, in the call to clCreateContextFromType you may want to insert a return parameter as the last argument instead of NULL, and check the value returned. Chances are this value will be CL_INVALID_PLATFORM in your case.