Firefly and PC GAMESS-related discussion club



Learn how to ask questions correctly


Re^3: MPICH2 version always runs in a single processor (Linux 64-bit)

Miro Moman
edelmiro.moman@gmail.com


Dear Alex,

Thanks for your advice. However, I am still not able to have Firefly propagating.

I have added  $smp httfix=.f. smppar=.t. $end  to the input file.

This is the command line I am using now:

mpiexec -np 16 /home/mirix/Downloads/firefly/firefly_71g_linux_mpich2_p4/firefly -t /disk/scratch/pcgamess/ -ncores 8 -nthreads 2

I have tried different values for -np, -ncores and -nthreads, including removing those options entirely. I have also tried mpirun.

Each time only one virtual core is used (half of a physical core). Eventually, I was able to have Firefly running on two virtual cores... but that is all.

Best regards,

Miro


On Sun Jun 20 '10 10:39pm, Alex Granovsky wrote
-----------------------------------------------
>Hi,

>you are using wrong command line syntax.
>Firefly requires use of fully qualified filenames
>in the command line. Moreover, if you are using the
>old-style command line syntax passing directories
>list at the end of command line, you should specify
>exactly one directory per each process, and all these
>directories MUST exist prior to Firefly's run.

>So, please check the documentation on command line options.

>One addition comment concerning your run.
>If you are going to run 16 Firefly instances
>on all 16 logical cores of your dual-cpu/eight-core
>system, add the following line to your input:

>

 $smp httfix=.f. smppar=.t. $end      

>Without this line, FF will affinitize (bind to cores)
>itself in such a way so that only half of all logical
>cores will be in use; and thus, you'll effectively run
>16 processes using only 8 logical cores.

>Regards,
>Alex Granovsky
>
>
>
>On Sun Jun 20 '10 8:41pm, Miro Moman wrote
>------------------------------------------
>>More info:

>>If I run:

>>mpiexec -np 16 /home/mirix/Downloads/firefly/firefly_71g_linux_mpich2_p4/firefly -i complex.inp -o output.out /disk/scratch/pcgamess /tmp/pcgamess

>>The program does not run and I obtain:

>> FATAL ERROR INITIALIZING FIREFLY! CHECK YOUR COMMAND LINE OPTIONS!

>> ON MASTER NODE, ERROR CODE IS : 0x00000004
>>    Unable to create output file

>>If I remove the "-o output.out" option, the program runs, but the end of the output file reads:

>>    2000000 WORDS OF MEMORY AVAILABLE

>> Warning: running without fastdiag runtime extension!

>> Warning: HTT is enabled, bitmask of physically unique cores is 0x0000FF00

>> SMT aware parts of program will use              2 threads.

>> Creating thread pool to serve up to            128 threads.

>> Activating Call64 option.

>> Using 64-bit DGEMM by default.
>> **** ERROR, NO $DATA GROUP WAS FOUND

>> CPU        TIME:   STEP =      0.90 ,  TOTAL =        1.3 SECONDS (    0.0 MIN)
>> WALL CLOCK TIME:   STEP =      0.89 ,  TOTAL =        0.9 SECONDS (    0.0 MIN)
>> CPU UTILIZATION:   STEP =    100.74%,  TOTAL =     145.51%

>>           0 WORDS OF    DYNAMIC MEMORY USED
>> EXECUTION OF FIREFLY TERMINATED ABNORMALLY AT 19:34:26 LT  20-JUN-2010
>>
>>
>>
>>
>>
>>On Sun Jun 20 '10 8:04pm, Miro Moman wrote
>>------------------------------------------
>>>Hello,

>>>I am trying to run the MPICH2-Linux version of Firefly (firefly_71g_linux_mpich2_p4) in a local machine with two quad-core Nehalem processors. I have started the mpd daemon as per the MPICH website instructions.

>>>This is the command line I am using:

>>>/home/mirix/Downloads/firefly/firefly_71g_linux_mpich2_p4/firefly -i complex.inp -o output.out /disk/scratch/pcgamess /tmp/pcgamess -p4pg /home/mirix/Downloads/firefly/combo/procgrp

>>>The input file is the example from Combo (optimization of the final complex).

>>>As for the procgrp file I have tried:

>>>local 7 and aeolus 7 /home/mirix/Downloads/firefly/firefly_71g_linux_mpich2_p4/firefly

>>>Firefly correctly detects 8 cores and 16 threads but it only uses one processor.

>>>Best regards,

>>>Miro


[ Previous ] [ Next ] [ Index ]           Mon Jun 21 '10 1:53am
[ Reply ] [ Edit ] [ Delete ]           This message read 1000 times