Scientific Linux Forum.org



  Reply to this topicStart new topicStart Poll

> CERNLIB issue, problem when trying to produce NTUPLES
mattcryo
 Posted: Jul 2 2015, 02:54 PM
Quote Post


SLF Newbie


Group: Members
Posts: 4
Member No.: 3485
Joined: 2-July 15









Hi everyone,
I'm working on my thesis, but I'm kinda stuck upon correctly installing CERNLIB.
I'm using SLC6.6(Carbon) and I'm trying to generate some NTUPLE with a Monte Carlo program, MCFM.
In order to do that, according to the manual, I have to link some libraries of CERNLIB toolkit to MCFM.
After many, many, many tries, the following warnig keeps showing up:

RZOPEN. record length: 8192 > maximum safe value (8191 words).
RZOPEN. You may have problems transferring your file to other systems
or writing it to tape.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
locf_() Warning: changing base from 0 to 7fff00000000!!!
This may result in program crash or incorrect results
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

and the produced output is either corrupted or misfilled (i.e. some variables are always put to 0, and this should not happen at all!)

I've tried to change MCFM version, but nothing changed. I'd like to add that the warning shows up if I run the "single thread" version of the program; when I run the omp one, it just crashes, ending in a segmentation fault.

I think that the problem is in my installation of CERNLIB. I've tried to compile it from source, but I was not able to, so I've used some precompiled versions. Since I've not found any version specifically for SLC6, I've ended up in using this one:

PC Linux Cern x86_64-slc5-gcc43-opt(Cernlib 2006)

Any help will be very appreciated, thank you very much.
PMEmail Poster
^
helikaon
 Posted: Jul 3 2015, 08:30 AM
Quote Post


SLF Administrator
*******

Group: Admins
Posts: 836
Member No.: 4
Joined: 8-April 11









QUOTE (mattcryo @ Jul 2 2015, 02:54 PM)
Hi everyone,
I'm working on my thesis, but I'm kinda stuck upon correctly installing CERNLIB.
...
...
Any help will be very appreciated, thank you very much.


Hi there,
this looks like more like a question best answered by devs - please write this same question SL mailing list, i'm not sure if you get many responses here, as the question is very, very specific and regarding very (IMHO) rarely used SW.

Also, (again IMHO) if you need to get help, please write in more detail what you exactly do command - by command if possible and what you linked where, so ppl that wants to help might duplicate the issue.
Since (as per googling) this applicaiton is used directly for hadron collider calculations, i think not many ppl around globe know it :]

best of luck,
cheers,

--------------------
PMEmail Poster
^
mattcryo
 Posted: Jul 3 2015, 02:04 PM
Quote Post


SLF Newbie


Group: Members
Posts: 4
Member No.: 3485
Joined: 2-July 15









Thank you very much for your reply, I'll send the issue to the ML ASAP, including the detailed procedure in order to replicate the problem.

I'll also post the mail here, just in case smile.gif

Cheers
PMEmail Poster
^
mattcryo
 Posted: Jul 6 2015, 08:26 AM
Quote Post


SLF Newbie


Group: Members
Posts: 4
Member No.: 3485
Joined: 2-July 15









Hi everybody,
I've sent an email to the mailing list, Dr. Schroeder gave me some really precious information and I've managed to solve the problem.
Someone could have my same problem, so I post the solution:

Due to architecture compatibility, CERNLIB needs to be liken statically, so I modified the LIBFLAGS of CERNLIBS in the makefile with the output of the command:

cernlib mathlib (Note: this works only on lxplus)

the output was the following:

-L/usr/lib64/cernlib/2006/lib -Wl, -static -lmathlib -lpacklib -lkernlib -Wl, -dy -llapack -lm -lnsl -lcrypt -ldl

Then, since ZEBRA output mode is somewhat bugged in MCFM, I've set FROOT in the makefile then, since ROOT bin and lib folder are in very different path, I had to modify again the makefile by removing the --prefix directive in ROOTLIB and ROOTINC variable.

With these modifications, it worked just fine, even if the OMP version of MCFM-7.0 crashes (due to an incompatibility between FROOT and OMP, I suppose)

PMEmail Poster
^
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

Topic Options Reply to this topicStart new topicStart Poll