search

firemodels / fds

Fire Dynamics Simulator
https://pages.nist.gov/fds-smv/
Other
663 stars 622 forks source link

Segmentation Fault on Intel MacPro #215

Closed gforney closed 9 years ago

gforney commented 9 years ago 
Please complete the following lines...

Application Version: 5.0.2 Serial
SVN Revision Number: 977
Compile Date: Mon, 12 Nov 2007
Operating System: Mac OS X 10.5.1 (9B18), Darwin 9.1.0

Describe details of the issue below: FDS input file (1.4 million cells)
generates a segmentation fault and will not run. The same input file does
run using 5.0.2 Serial WinXP version.

Hardware Overview:

  Model Name:               Mac Pro
  Model Identifier:     MacPro1,1
  Processor Name:       Dual-Core Intel Xeon
  Processor Speed:      3 GHz
  Number Of Processors:     2
  Total Number Of Cores:    4
  L2 Cache (per processor): 4 MB
  Memory:                   12 GB
  Bus Speed:                1.33 GHz

Original issue reported on code.google.com by john.e.brooker@nasa.gov on 2007-12-03 20:40:48

gforney commented 9 years ago 
Please attach the input file to this issue.  You can do so by going to the issue 
using the link in the email update you received and clicking in the add a comment 
text box.  You will then get a link to attach a file.

Original issue reported on code.google.com by drjfloyd on 2007-12-03 21:11:56

gforney commented 9 years ago 
Please try the attached binary...

Original issue reported on code.google.com by bryanwklein on 2007-12-03 21:55:00

gforney commented 9 years ago 
Fire Dynamics Simulator (fds5_osx_intel_64)

 Compilation Date : Thu, 29 Nov 2007
 Version          : 5.0.2 Serial
 SVN Revision No. : 1037
 Operating System : Mac OS X 10.5.1 (9B18), Darwin 9.1.0

I tried the new 64-bit binary over the weekend using 3 variants of my input file.

1) 2,764,800 cells (single mesh), still running (timestep 14500)

2) 1,382,400 cells (mesh1), 1,382,400 cells (mesh2), segmentation fault after
timestamp 1100

3) 777,600 cells (mesh1), 1,536,000 cells (mesh2), segmentation fault after timestamp
900

For all cases, Activity Monitor showed 1 thread, ~2 GB real memory, ~5GB virtual memory.

Original issue reported on code.google.com by john.e.brooker@nasa.gov on 2007-12-10 16:17:23

gforney commented 9 years ago 
Try attached 32bit versions.

Original issue reported on code.google.com by bryanwklein on 2007-12-10 22:29:25

gforney commented 9 years ago 
Fire Dynamics Simulator (fds_osx_intel)

 Compilation Date : Wed, 05 Dec 2007
 Version          : 5.0.3 Serial
 SVN Revision No. : 1068

The new 32 bit build, with the updated Intel compiler, appears to be working for case
1 from above (time step 3800). I will need your help compiling MPI in order to test
the 32 bit parallel build.

Original issue reported on code.google.com by john.e.brooker@nasa.gov on 2007-12-11 20:18:32

gforney commented 9 years ago 
Some interesting stats comparing 64 bit (5.0.2 Serial, 1037) and 32 bit (5.0.3
Serial, 1068) for case1.

 Time Step      1   December  7, 2007  17:25:05
       ----------------------------------------------
       CPU/step:    11.782 s, Total CPU:     11.78 s
       Time step:  0.02011 s, Total time:     0.02 s
       Max CFL number:  0.19E-01 at ( 76, 87, 65)
       Max divergence:  0.67E-15 at ( 70, 54,159)
       Min divergence: -0.67E-15 at ( 72,136, 14)

Time Step   7200   December  9, 2007  02:19:42
       ----------------------------------------------
       CPU/step:    12.047 s, Total CPU:     24.68 hr
       Time step:  0.00135 s, Total time:    10.33 s
       Max CFL number:  0.98E+00 at ( 66, 44,130)
       Max divergence:  0.55E+01 at ( 24,119, 72)
       Min divergence: -0.10E+02 at ( 19,122, 72)
       No. of Lagrangian Particles:           818
       Total Heat Release Rate:              2.003 kW
       Radiation Loss to Boundaries:         1.189 kW

Time Step      1   December 10, 2007  18:16:25
       ----------------------------------------------
       CPU/step:    15.834 s, Total CPU:     15.83 s
       Time step:  0.02011 s, Total time:     0.02 s
       Max CFL number:  0.19E-01 at ( 76, 87, 65)
       Max divergence:  0.67E-15 at ( 70, 54,159)
       Min divergence: -0.67E-15 at ( 72,136, 14)

Time Step   7200   December 12, 2007  10:06:54
       ----------------------------------------------
       CPU/step:    15.575 s, Total CPU:     30.96 hr
       Time step:  0.00135 s, Total time:    10.33 s
       Max CFL number:  0.98E+00 at ( 79, 40,110)
       Max divergence:  0.12E+02 at ( 16,126, 72)
       Min divergence: -0.53E+01 at ( 16,131, 72)
       No. of Lagrangian Particles:           833
       Total Heat Release Rate:              2.552 kW
       Radiation Loss to Boundaries:         1.512 kW

Original issue reported on code.google.com by john.e.brooker@nasa.gov on 2007-12-12 20:34:12

gforney commented 9 years ago 
Please try the new 5.1.0 release, it includes both 32 and 64 bit versions.
I am not sure how well 64bit MPI version is working, please start a seperate thread
if it does not work.

Original issue reported on code.google.com by bryanwklein on 2008-01-08 22:49:04

gforney commented 9 years ago 
I would like to try the 32/64 bit MPI versions, although I am not entirely clear how
to do this on a multi-core 
machine. Please advise.

Original issue reported on code.google.com by john.e.brooker@nasa.gov on 2008-01-09 21:45:58

gforney commented 9 years ago 
I will put together a write up for this process and post it as a Wiki when it is
finished.  I will also notify you through this issue posting.

Original issue reported on code.google.com by bryanwklein on 2008-01-15 12:31:28

gforney commented 9 years ago 
A follow-up to Comment 7. I tried the new 32/64 bit 5.1.0 release using case 1 (see
Comment 3) with the 
following results.

Fire Dynamics Simulator (fds_osx_intel_32)
  Compilation Date : Tue, 08 Jan 2008
  Version                  : 5.1.0 Serial
  SVN Revision No. : 1162
Time step 6979, 10s, successful

Fire Dynamics Simulator (fds_osx_intel_64)
  Compilation Date : Tue, 08 Jan 2008
  Version                  : 5.1.0 Serial
  SVN Revision No. : 1162
Time step 600, 1.41s, error (174) segmentation fault

Not sure why the previous 64 bit build worked (see comment 3) and this one didn't.

Original issue reported on code.google.com by john.e.brooker@nasa.gov on 2008-01-24 16:49:57

gforney commented 9 years ago 
Have you tried the 5.1.1 version posted yesterday?
Please confirm that you get the Seg Fault with this one as well.

-Bryan

Original issue reported on code.google.com by ZivBK1 on 2008-01-24 21:08:38

gforney commented 9 years ago 
I tried the new 64 bit 5.1.1 release using case 1 (see Comment 3) with the following
results.

Fire Dynamics Simulator (fds_osx_intel_64)
  Compilation Date : Mon, 21 Jan 2008
  Version                  : 5.1.1 Serial
  SVN Revision No. : 1217
Time step 600, 1.41s, error (174) segmentation fault

Original issue reported on code.google.com by john.e.brooker@nasa.gov on 2008-01-28 16:33:22

gforney commented 9 years ago 
What is the status of this thread?

Original issue reported on code.google.com by mcgratta on 2008-07-22 20:44:04

gforney commented 9 years ago 
We will retry with 5.2 release and latest versions of Intel compilers.

Original issue reported on code.google.com by ZivBK1 on 2008-07-22 21:12:27

gforney commented 9 years ago 
(No text was entered with this change)

Original issue reported on code.google.com by bryanwklein on 2008-08-28 17:06:34

gforney commented 9 years ago 
I got similar errors on my Linux-Workstation (2*QuadCore) with a self-compiled
version of fds (releases of 5.1.6 to present). As far as I now now, the compiler
options are the key to this error... I recompiled fds 5.2.1 with additional options
and one case is running for much longer than ever.

Original issue reported on code.google.com by simon.frey@hbi.ch on 2008-10-14 06:16:49

gforney commented 9 years ago 
I got similar errors on my Linux-Workstation (2*QuadCore) with a self-compiled
version of fds (releases of 5.1.6 to present). As far as I now now, the compiler
options are the key to this error... I recompiled fds with additional compiler
options and the first test-case is still running...

Please see http://code.google.com/p/fds-smv/issues/detail?id=474

Original issue reported on code.google.com by simon.frey@hbi.ch on 2008-10-14 09:17:44

gforney commented 9 years ago 
Thank you for the tip... I am testing this flag out with the 64bit OS X builds and
hopefully it will help there too.

Original issue reported on code.google.com by bryanwklein on 2008-10-22 00:02:09

gforney commented 9 years ago 
Could you manage to fix your problem?

Original issue reported on code.google.com by simon.frey@hbi.ch on 2008-10-24 11:41:22

gforney commented 9 years ago 
bryan will respond when he gets back to the office which should be sometime the week
of november 3rd

Original issue reported on code.google.com by gforney on 2008-10-24 11:53:54

gforney commented 9 years ago 
Bryan and I worked on this issue Wednesday (10/22/08) from our homes using iChat. He
guided me through the process of compiling OpenMP with the inclusion of gfortran. We
still had more to do, but it was late so we stopped for the night. I noticed an email
(home address) this morning that had an attachment which included the remaining
libraries and binaries necessary for 64 bit parallel fds. He said it worked for him
using my input file with 4 meshes. I will try it out this weekend and report back.

Original issue reported on code.google.com by john.e.brooker@nasa.gov on 2008-10-24 18:21:33

gforney commented 9 years ago 
I am back again and will be looking into the newest errors.

John, can you post them here just to keep track for others who may encounter this issue.

-Bryan

Original issue reported on code.google.com by bryanwklein on 2008-11-03 21:32:29

gforney commented 9 years ago 
Will contact again for further troubleshooting.

Original issue reported on code.google.com by bryan%solvercore.com@gtempaccount.com on 2009-01-26 22:03:38

gforney commented 9 years ago 
Closing this Issue based on our conversation.

Original issue reported on code.google.com by bryan%solvercore.com@gtempaccount.com on 2009-02-05 16:41:23