Hi Guys,
I am trying to install and use OpenCL on my local cluster, of which the node have a CPU and 3 MIC cards. I have installed it successfully and running clinfo will return the right information (as expected). However, when running a simple example, e.g., vector addition, only device0 can run and return normally. The other two devices (device1 and device2) can create context, kernels normally, but hangs after that.
When strack the program, it gives the following message. Do anybody have an idea of the reasons?
-----------------------------------------------------------------------------
clone(child_stack=0x7f2552bfcfd0, flags=CLONE_VM|CLONE_FS|CLONE_FILES|CLONE_SIGHAND|CLONE_THREAD|CLONE_SYSVSEM|CLONE_SETTLS|CLONE_PARENT_SETTID|CLONE_CHILD_CLEARTID, parent_tidptr=0x7f2552bfd9d0, tls=0x7f2552bfd700, child_tidptr=0x7f2552bfd9d0) = 7981
futex(0x7f255bfeb618, FUTEX_WAKE_PRIVATE, 1) = 1
futex(0x7f255bfeb618, FUTEX_WAIT_PRIVATE, 2, NULL) = -1 EAGAIN (Resource temporarily unavailable)
futex(0x7f255bfeb618, FUTEX_WAKE_PRIVATE, 1) = 0
futex(0x7f255bfeb644, FUTEX_CMP_REQUEUE_PRIVATE, 1, 2147483647, 0x7f255bfeb618, 94) = 3
futex(0x7f255bfeb618, FUTEX_WAKE_PRIVATE, 1) = 1
futex(0x7f255bfeb644, FUTEX_CMP_REQUEUE_PRIVATE, 1, 2147483647, 0x7f255bfeb618, 100) = 3
futex(0x7f255bfeb618, FUTEX_WAKE_PRIVATE, 1) = 1
futex(0x7f255d734e80, FUTEX_WAIT_PRIVATE, 2, NULL) = -1 EAGAIN (Resource temporarily unavailable)
brk(0x295a000) = 0x295a000
futex(0x7f255d734e80, FUTEX_WAKE_PRIVATE, 1) = 1
futex(0x7f255bfeb644, FUTEX_CMP_REQUEUE_PRIVATE, 1, 2147483647, 0x7f255bfeb618, 106) = 3
futex(0x7f255bfeb644, FUTEX_WAIT_PRIVATE, 110, NULL
-----------------------------------------------------------------------------
Jianbin
連結已複製
1 回應
Hi Jianbin,
Could you please provide more details?
1) What are the versions of OS, OpenCL SDK, MPSS?
2) What is the output of 'micinfo' command?
3) What is the output of 'micctrl -s' command? You need to perform it as a root. The expected output is:
mic0: online
mic1: online
mic2: online
4) Do you have only one OpenCL implementation/platform (Intel's one) installed?
5) What do you mean by device0, device1, device2? Which device is CPU and which is MIC? And I guess in total you should have 4 devices available (1 CPU + 3 MIC).
6) Could you please provide a minimal reproducer for this issue?
Thanks,
Yuri
