I can reproduce the issue using the AMD APP SDK v3.0 and v16.1.1 of the Intel OpenCL runtime for Core and Xeon processors. It's unclear, however, what exactly VirtualGL is doing to contribute to the application's delinquency. I'll keep poking at it. The problem seems to be two-fold:
(1) Why isn't the AMD GPU OpenCL device enabled when using VirtualGL? (2) Why does it crash when the Intel provider is also installed? These may be related issues, although it's also important to point out that RHEL 6 and AMD CPUs are not a supported configuration for the Intel OpenCL drivers. On 6/27/16 10:23 AM, Jeff McWilliams wrote: > > Can one use OpenCL when running under VirtualGL? > > > > I noticed that clinfo segfaults if I invoke it with: > > > > “vglrun clinfo” > > > > I’m using the latest VirtualGL 2.5, CentOS 6.7 64-bit, AMD FirePro > W8000 graphics. > > > > Core was generated by `clinfo'. > > Program terminated with signal SIGSEGV, Segmentation fault. > > #0 0x00007f6e9fd77fcd in ?? () from > /opt/intel/opencl-1.2-4.4.0.117/lib64/libintelocl.so > > Missing separate debuginfos, use: debuginfo-install > VirtualGL-2.5-20160215.x86_64 glibc-2.12-1.166.el6_7.7.x86_64 > libX11-1.6.0-6.el6.x86_64 libXau-1.0.6-4.el6.x86_64 > libXext-1.3.2-2.1.el6.x86_64 libXv-1.0.9-2.1.el6.x86_64 > libgcc-4.4.7-16.el6.x86_64 libstdc++-4.4.7-16.el6.x86_64 > libxcb-1.9.1-3.el6.x86_64 mesa-libGL-10.4.3-1.el6.x86_64 > numactl-2.0.9-2.el6.x86_64 opencl-1.2-intel-cpu-4.4.0.117-1.x86_64 > > (gdb) bt > > #0 0x00007f6e9fd77fcd in ?? () from > /opt/intel/opencl-1.2-4.4.0.117/lib64/libintelocl.so > > #1 0x00007f6e9fcf1e81 in clGetPlatformInfo () from > /opt/intel/opencl-1.2-4.4.0.117/lib64/libintelocl.so > > #2 0x000000000040e8be in int > cl::detail::getInfoHelper<cl::detail::GetInfoFunctor0<int > (*)(_cl_platform_id*, unsigned int, unsigned long, void*, unsigned > long*), _cl_platform_id*> >(cl::detail::GetInfoFunctor0<int > (*)(_cl_platform_id*, unsigned int, unsigned long, void*, unsigned > long*), _cl_platform_id*>, unsigned int, std::__1::basic_string<char, > std::__1::char_traits<char>, std::__1::allocator<char> >*, long) > [clone .isra.35] [clone .constprop.213] () > > #3 0x000000000040fd57 in > cl::detail::param_traits<cl::detail::cl_platform_info, > 2308>::param_type cl::Platform::getInfo<2308>(int*) const () > > #4 0x0000000000407fc3 in main () > > > > > > If I rename /etc/OpenCL/vendors/intel64.icd I can then run “vglrun > clinfo”, and it does not crash. I get a list indicating AMD APP is > the provider, with one CPU device. The GPU device is not listed. If > I run clinfo without using vglrun, the AMD APP provider lists both a > CPU device and a GPU device. > > > > > > > > Thanks, > > > > Jeff McWilliams > > Software Development Manager – Altair HyperView >
------------------------------------------------------------------------------
_______________________________________________ VirtualGL-Users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/virtualgl-users
