[Eg6_analysis] user_ana on 64 bit machines
Lamiaa El Fassi
lamiomar at gmail.com
Wed Jan 8 16:45:34 EST 2014
Hi Stepan, Hi all,
I managed to run successfully user_ana in a batch farm after implementing
some
changes in one of recsis routines. There is still some issues with for
example a Fortran
"IF" statement which is not acting as it should be, and other things which
need more investigation. However, I am thinking if at least the RTPC
response is similar to what
we have in Pass1_v1, and if that what only matters in the drift path
calibration, can I
go ahead with processing the data for this purpose and don't hold this work
while doing
the other tests? I may need the RTPC experts to give me their words on that
after
having a look to the outputs of this batch cooking located at:
/volatile/clas/claseg6/tgemRHL6_v1/1p2gev/
I processed in this test 10 files of a run 61448 with the RTPC calibration
bosbanks.
Please ignore the outputs of a file 61448.A03 which crashed for a different
issue!
Thank you,
Lamiaa
On Sat, Dec 21, 2013 at 7:55 AM, Stepan Stepanyan <stepanya at jlab.org> wrote:
> Hello Nathan,
>
> Thanks for looking into this. Although we now managed to run the code
> on 64-bin machines, there is still this issue with FORTRAN "IF" statement.
> That could be one of sources of difference.
>
> But I agree with you, more cooking and comparison has to be done.
>
> Regards, Stepan
>
> On 12/20/13 8:10 PM, Nathan Baltzell wrote:
>
> Dear EG6ers,
>
> Here's an h10 comparison of this new 64bit-compiled cooking versus
> pass1v1.
> The only significant differences I noticed are trigger time and neutral
> yields.
>
> Similarities:
> - same number of reconstructed events
> - same number of good TPC tracks
> - same number of IC clusters
>
> Differences: (only in CLAS)
> - few-% more charged tracks for 64bit
> - few-% more electrons and protons
> - 40% more neutrons, but same number of gammas (!?)
>
> Looking a little closer, event-by-event:
> - trigger time is ALWAYS different (by hundreds of ps up to ~1 ns) (!?)
> - beta for neutrals differs due to trigger time
> - reconstructed SC time, EC energy, EC time, CC nphe are identical
> - DC momentum is very similar (discrepancy is usually sub-MeV)
> - reconstructed TPC and IC banks are practically identical
>
> Note this is based on an early 1.2 GeV file, maybe we should check a 6
> GeV run too.
>
> Regards,
> Nathan
>
>
>
> P.S. The files I compared are here:
>
> new 64bit:
> /w/hallb/claseg6/top_dir/test_64compil/ntp_61448_05.hbook
> /w/hallb/claseg6/baltzell/test_64compil/hroot_61448_05.root
>
> pass1v1:
>
> /cache/mss/clas/eg6/production/pass1_v1/1p2gev/PAW/ntp_61448_05_pass1v1.hbook
>
> /cache/mss/clas/eg6/production/pass1_v1/1p2gev/HROOT/hroot_61448_05_pass1v1.root
>
>
>
>
> On Thu, 19 Dec 2013 20:57:19 -0600, Lamiaa El Fassi <lamiomar at gmail.com><lamiomar at gmail.com>wrote:
>
> Hi Stepan, Hi All,
>
> I succeeded today to build the appropriate claseg6 32 bit libraries and
> binaries in the
> new 64 operating machine. Many thanks to Hovanes, who showed me how to do
> that!
>
> The new user_ana executable under ~claseg6/bin/LinuxRHEL6/ runs
> successfully in the interactive mode. From the printout below, it seems
> that all HBT and TBT conditions are fulfilled:
> ........
> TRK_STATI L: Cut# Description Value Passed %
> TRK_STATI L: 1 trk_evnt: Ntrig (Nevt) > 0.0 50000 100.0
> TRK_STATI L: 2 ana_segm: Nevt w. segm = 5.0 49645 99.3
> TRK_STATI L: 3 trk_link: Nevt w. link = 5.0 47862 95.7
> 100.0
> TRK_STATI L: 4 ana_prfit: Nevt PR chi2 < 70.0 47119 94.2
> 98.4
> TRK_STATI L: 5 SEB_eid: Nevt SEB = 0.0 46394 92.8
> 96.9
> TRK_STATI L: 6 trk_dcam: Nevt w. LRseg= 4.0 45968 91.9
> 96.0
> TRK_STATI L: 7 trk_dcam: Nevt w. LRseg= 5.0 44345 88.7
> 92.7
> TRK_STATI L: 8 ana_fit: Nevt FF1 chi2 < 70.0 43367 86.7
> 90.6
> TRK_STATI L: 9 ana_fit: Nevt in TBT > 0.0 43367 86.7
> 90.6
> ..............
> The outputs of this test run, in case anyone want to have a look at
> them, are saved
> under this directory: /w/hallb/claseg6/top_dir/test_64compil. Files
> produced with RTPC bosbanks and 50K events.
>
> On the other hand, running user_ana on batch mode is still giving the I/O
> errors while reading the tcl flags on "retcl_yield.F" routine. I am
> attaching an example of the two farm
> output files; example.err & example.txt.
>
> In order to know if this error is happening at the end of a tcl file or
> while reading it, I
> put some write/print statements around the line causing this issue, as
> follow:
>
>
> -----------------------------------retcl_yield.F----------------------------------------------
> ...........................
> if (nbyt .gt. 0) then
> write(6,*) "RETCL YIELD READING FILE" ! Added 1st write
> read(*,13, end=999) cline ! line 101
> print *, cline !
> Added print
> write(6,*) "END RETCL YIELD READING FILE" ! Added 2nd write
> 13 format(a200)
> cline = spaces(cline,1)
> iend = lenocc(cline)
> .......
> 999 continue
> ............
>
> --------------------------------------------------------------------------------------------------------
>
> As you can see in .txt file, only the 1st write statement was printed out
> (twice), and
> nothing after reading the tcl flag!
>
> Any comments/suggestions on how to deal with this issue will be
> appreciated!
>
> Best regards,
>
> Lamiaa
>
>
>
> On Thu, Dec 19, 2013 at 5:34 PM, Stepan Stepanyan <stepanya at jlab.org>wrote:
>
>> Hello everyone,
>>
>> With Hovanes's help, I managed to compile user_ana on ifarm1102 with
>> -m32 flag.
>> More, it runs and produced reasonable results. My build with all libs is
>> in
>> /home/stepanya/eg6/test
>> Appropriate files in cms directory have been modified to take into
>> account -m32
>> and other related changes.
>>
>> Have a nice holidays, Stepan
>>
>> _______________________________________________
>> Eg6_analysis mailing list
>> Eg6_analysis at jlab.org
>> https://mailman.jlab.org/mailman/listinfo/eg6_analysis
>>
>
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://mailman.jlab.org/pipermail/eg6_analysis/attachments/20140108/71a3debb/attachment.html
More information about the Eg6_analysis
mailing list