Error mesage when running Intel® Optimized MP LINPACK Benchmark for Clusters on Intel Phi cards.

[Tinway_Chen]

Hi

I refer to below instruction of Intel® Optimized MP LINPACK Benchmark for Clusters, I am trying offload computations to Intel Xeon Phi coprocessors by running runme_offload_intel64 script file.

(http://software.intel.com/sites/products/documentation/doclib/mkl_sa/11/mkl_userguide_lnx/GUID-10DFCB17-3953-47C6-9971-8C455A925BFE.htm)

My test environment :

1. AIC Sandy Bridge EP-4S server system with Sandy Bridge EP-4S *4 + 98GB memory
2. Intel Xeon Phi : 3 pcs of 3110 and 4 pcs of 3115
3. OS: Redhat Enterprise Linux 6.2 x64
4. Xeon Phi MPSS: KNC_gold_update_2-2.1.5889-16-rhel-6.2.tar
5. Intel Composer XE : l_ccompxe_2013.3.163.tgz
6. Intel MPI : l_mpi_p_4.1.0.024.tgz or l_mpi_p_4.1.0.030.tgz

I found the below error messages by unning dmesg command in Phi card Linux OS. Does it impact the performance?

[ 1321.750417] INFO: task kworker/u:0:5 blocked for more than 120 seconds.

[ 1321.750445] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.

[ 1321.750467] kworker/u:0     D ffff8800fce0d7b0  6136     5      2 0x00000000

[ 1321.750506]  ffff8801738b7cf0 0000000000000046 ffffffff8171c040 ffff8801ffffffff

[ 1321.750531]  ffff8801738a1e00 ffff8801738b7fd8 ffff8801738b7fd8 0000000000011940

[ 1321.750554]  0000000000011940 ffff8801738a1e00 0000000000000000 0000000100000286

[ 1321.750578] Call Trace:

[ 1321.750614]  [<ffffffff81355403>] schedule_timeout+0x9d/0xc4

[ 1321.750641]  [<ffffffff8103f07f>] ? process_timeout+0x0/0x10

[ 1321.750663]  [<ffffffff81355175>] wait_for_common+0x9e/0x10c

[ 1321.750688]  [<ffffffff8102d16a>] ? default_wake_function+0x0/0x14

[ 1321.750711]  [<ffffffff81355280>] wait_for_completion_timeout+0x13/0x15

[ 1321.750736]  [<ffffffff81253874>] mic_idlewake_handler+0x27/0xd4

[ 1321.750761]  [<ffffffff81048965>] process_one_work+0x163/0x289

[ 1321.750781]  [<ffffffff8125384d>] ? mic_idlewake_handler+0x0/0xd4

[ 1321.750805]  [<ffffffff810499e9>] worker_thread+0xfb/0x17f

[ 1321.750827]  [<ffffffff810498ee>] ? worker_thread+0x0/0x17f

[ 1321.750851]  [<ffffffff8104ce99>] kthread+0x84/0x8c

[ 1321.750877]  [<ffffffff81003854>] kernel_thread_helper+0x4/0x10

[ 1321.750901]  [<ffffffff8104ce15>] ? kthread+0x0/0x8c

[ 1321.750922]  [<ffffffff81003850>] ? kernel_thread_helper+0x0/0x10

[Loc_N_Intel]

Hi Tinway Chen,

Let me investigate this issue and get back to you. Thanks.

[Tinway_Chen]

Hi Mr.Nguyen

Thanks for watching this issue.

[Loc_N_Intel]

Hello Tinway,

dmesg messages are genereated by Intel(R) Xeon Phi(TM) kernel, usually at boot time. They are not harmful and don't have any impact on  the coprocessor performance at all. Hope this helps.

[Tinway_Chen]

Hi Mr.Nguyen

Thanks.

However, the messages were generated after I ran Intel® Optimized MP LINPACK. There are not these messages before run MP LINPACK.

[Loc_N_Intel]

Hello Tinway,

Can you confirm that you get these dmesg messages repeatly everytime you ran Intel(R) Optimized MP LINPACK on your system? Can you run Intel(R) Optimized MP LINPACK three times and look at the dmesg messages? Thank you.

[Loc_N_Intel]

Hi Tinway,

I verified with MKL folks about your dmesg messages, they said these messages are generated if the workload is too heavy but they are not bad for performance at all. Please upgrade the new MPSS Gold Update 3 at http://software.intel.com/en-us/articles/intel-manycore-platform-software-stack-mpss 

Thank you.

[Tinway_Chen]

Hi Mr.Nguyen

Thanks. I will ignore this issue.

I have updated MPSS Gold Update 3 and will perform my further testing.

Appreciated your kindly supoort.