HCC runtime crashes on amdgpu-pro driver

help me.

I am sorry my pure *English because i am Japaneses.

I have been trying built environment of HCC with R9 390 GPU and Ubuntu 16.04, but it's not fine.

I' m happy if someone give me a few information.

following is my trying.

first , I installed ubuntu 16.04 with selecting no installation of automatic update and third party software.

second installed hcc following below site. https://bitbucket.org/multicoreware/hcc/wiki/Developer%20Information

I executed below commands written in the guide site.

sudo apt-get install cmake git subversion g++ libstdc++-4.8-dev libdwarf-dev libelf-dev libtinfo-dev libc6-dev-i386 gcc-multilib llvm llvm-dev llvm-runtime libc++1 libc++-dev libc++abi1 libc++abi-dev re2c libncurses5-dev

then, I downloaded HCC package from here https://bitbucket.org/multicoreware/hcc/downloads/ I selected the package of "hcc-0.9.16045-c4b0995-ff03947-5a1009a-Linux.deb". and I installed.

I created a PATH to hcc by following command, export PATH=$PATH:/opt/hcc/bin

third, install AMDGPU-PRO driver form below. http://support.amd.com/en-us/kb-articles/Pages/AMD-Radeon-GPU-PRO-Linux-Beta-Driver%E2%80%93Release-Notes.aspx

4th, I compiled test cord with clang++ using the command of,

clang++ `clamp-config --install --cxxflags --ldflags` -o test.out test.cpp

compilation is success. but when it run, crash with kernel error in the case of using parallel_for_each.

*****************OK cord******************
int main()
{
    gAcc = ChoseAccelerator();

    gp_gau = new array<double,1>(1000*1000 , gAcc.get_default_view());

    array_view<double,1> agau = *gp_gau;

    cout << agau[2] << "\n";
    return 0;
}
***********************************************

here is output.

chose accelerator to compute

0 : CPU Device
   memory                   : 0 [GB]
   support double precision : true
   has display              : false
1 : AMD
   memory                   : 4.048 [GB]
   support double precision : true
   has display              : false
2 : Intel
   memory                   : 2.048 [GB]
   support double precision : true
   has display              : false

device num = 1

1, yes sir
-5.48613e+303

*****************Bad cord induced error******************


int main()
{
    gAcc = ChoseAccelerator();

    gp_gau = new array<double,1>(1000*1000 , gAcc.get_default_view());

    array_view<double,1> agau = *gp_gau;

    parallel_for_each(
        gAcc.get_default_view(),
        agau.get_extent(),
        [=](concurrency::index<1> id) restrict(amp)
        {
            agau[id] = agau[id] + 2.5;
        }
    );

    cout << agau[2] << "\n";
    return 0;
}
***********************************************

here is output. Error: Kernel '&__OpenCL_ZZ4mainEN3_EC__019__cxxamp_trampolineEPdiiiiiii_kernel' initialization failed.

Is there any person have knowledge to this problem ? please help me.

**appendix : full cord.****

#include <iostream>

#include <chrono>
#include <vector>
#include <string>
using std::cout;
using std::vector;
using std::string;
using std::cin;
using std::wcout;

#include <amp.h>
#include <amp_math.h>
using concurrency::array;
using concurrency::accelerator;
using concurrency::array_view;
using concurrency::parallel_for_each;
using concurrency::extent;

array<double , 1> *gp_gau;

accelerator ChoseAccelerator();
accelerator gAcc;
int main()
{
    gAcc = ChoseAccelerator();

    gp_gau = new array<double,1>(1000*1000 , gAcc.get_default_view());

    array_view<double,1> agau = *gp_gau;


    parallel_for_each(
        gAcc.get_default_view(),
        agau.get_extent(),
        [=](concurrency::index<1> id) restrict(amp)
        {
            agau[id] = agau[id] + 2.5;
        }
    );

    cout << agau[2] << "\n";
    return 0;
}



accelerator ChoseAccelerator()//chosing accelerator
{
    //get all accelerator on the system
    vector<accelerator> allAcc = accelerator::get_all();


    cout << "chose accelerator to compute" << "\n\n";
    for(int i=0;i<allAcc.size();i++)
    {
        wcout << i <<" : "<< allAcc[i].get_description() << "\n";
        cout << "   memory                   : "
            << (double)allAcc[i].get_dedicated_memory()/1024/1000 << " [GB]\n";
        cout << "   support double precision : ";
        if(allAcc[i].get_supports_double_precision())
            cout << "true\n";
        else
            cout << "false\n";

        cout << "   has display              : ";
        if(allAcc[i].get_has_display())
            cout << "true\n";
        else
            cout << "false\n";
    }

    cin.exceptions(std::ios::failbit);

    int selAcs;
    while(true)
    {
        cout << "\ndevice num = ";
        try{
            cin >> selAcs ;
        }
        catch(...){
            cout << "please input a number\n";
            cin.clear();
            cin.seekg(0);
            continue;
        }
        cout << "\n";

        if(0 <= selAcs && selAcs < allAcc.size())
        {
            cout << selAcs <<", yes sir\n";
            return allAcc[selAcs];
        }

        cout << "please input a number up to" << allAcc.size() -1 << ". "
             <<selAcs<<" dosen't exist\n";
    }


    throw -1;//throw exceptions. this case will never occure.
}

Comments (2)