[mpich-discuss] [SPAM] Re: Fw: FAIL of test_primitives in installing of mpich-3.1.2

[yangyi at lzu.edu.cn]

Dear  Rajeev,
Thanks for your reply.
The time that listed is from runing a forecast of test with WRF(Weather Research and Forecasting ) model.
I wonder the FAIL of test_primitives may cause  the performance is very poor.
Thanks!

Yi
> -----原始邮件-----
> 发件人: "Rajeev Thakur" <thakur at mcs.anl.gov>
> 发送时间: 2014-08-28 08:05:47 (星期四)
> 收件人: discuss at mpich.org
> 抄送: 
> 主题: [SPAM] Re: [mpich-discuss] Fw: FAIL of test_primitives in installing of mpich-3.1.2
> 
> If you run "make testing" in the top-level mpich directory, it will run the entire test suite in test/mpi. But those tests don't use as many processes as the number of cores you have listed below. (The number of processes for each test is in the testlist file in each directory under test/mpi.) So what is it that you are running with the different number of cores below?
> 
> Rajeev
> 
> 
> On Aug 27, 2014, at 6:14 PM, <yangyi at lzu.edu.cn>
>  wrote:
> 
> > Dear sir,
> > I installed the mpich-3.1.2 on a machine of IBM X3850 X5 which with 4 Xeon E7-4820 (8 cores, 16 threads) CPU and 64G  memory.
> > It has problem of  test_primitives  in make check.
> > 
> > It uses the mininum amount of time with 8 cores. The following is the time with differernt cores:
> > 60 cores   no resuls are out although the executable program is running
> > 56 cores   EXIT CODE: 11 Segmentation fault (signal 11)
> > 48 cores   1hour18min
> > 40 cores   1h11m
> > 32 cores   1h9m
> > 24 cores   1h9m
> > 16 cores   1h10m
> > 8  cores   55m
> > 4 cors     1h20m
> > 
> > The log files of inatlling are attached.
> > I'm looking forward to your reply at your earliest convenience.
> > 
> > 
> > Sincerely,
> > Yi 
> > 
> > 
> > <config.log><configure.out><make.out><make_check.out><make_install.out>_______________________________________________
> > discuss mailing list     discuss at mpich.org
> > To manage subscription options or unsubscribe:
> > https://lists.mpich.org/mailman/listinfo/discuss
> 
> _______________________________________________
> discuss mailing list     discuss at mpich.org
> To manage subscription options or unsubscribe:
> https://lists.mpich.org/mailman/listinfo/discuss