code_saturne User's Forum

Hello Yvan,

Thank you so much for your reply. I've installed the code with Metis-5.0.2. I followed your instructions and the new logged "configure" command is now the following:

r01ken0/Code_Saturne/v2.0.5/installer/ecs-2.0.2/configure --prefix=/users/r01ken0/Code_Saturne/v2.0.5/cs-2.0 --with-bft=/users/r01ken0/Code_Saturne/v2.0.5/cs-2.0 --with-hdf5=/users/r01ken0/Code_Saturne/opt --with-cgns=/users/r01ken0/Code_Saturne/opt --with-med=/users/r01ken0/Code_Saturne/opt --with-metis=/users/r01ken0/Code_Saturne/opt CC=gcc --enable-long-int


However, this has still not solved the problem. When I run "code_saturne mesh_check -m meshname.cgns" I obtain the following error in the terminal and in an error file:

***************************************************************

Local case configuration:

Date: Mon 11 Nov 2013 15:25:45 GMT
System: Linux 2.6.32-279.14.1.el6.x86_64
Machine: vhmem01
Processor: Intel(R) Xeon(R) CPU E7- 4820 @ 2.00GHz
Memory: 1033945 MB
User: r01ken0
Directory: /users/r01ken0/simulations/60degree_bundle_OLD/MESH
MPI ranks: 1

===============================================================

CALCULATION PREPARATION
=======================

===========================================================


Reading file: preprocessor_output

/users/r01ken0/Code_Saturne/v2.0.5/installer/bft-1.1.5/src/bft_file.c:794: Fatal error.

Error opening file "preprocessor_output":

No such file or directory

Call stack:
1: 0x2b3bfe7cd177 <bft_file_open_stream+0xa7> (libbft.so.1)
2: 0x2b3bfe7cd3a6 <bft_file_open+0xa6> (libbft.so.1)
3: 0x2b3bfd69a020 <fvm_file_open+0x140> (libfvm.so.0)
4: 0x2b3bfc35133b <+0xb033b> (libsaturne.so.0)
5: 0x2b3bfc352323 <cs_io_initialize+0x93> (libsaturne.so.0)
6: 0x2b3bfc3eeec9 <ledevi_+0x79> (libsaturne.so.0)
7: 0x2b3bfc43b557 <iniini_+0xce3> (libsaturne.so.0)
8: 0x2b3bfc43e2ae <initi1_+0x16> (libsaturne.so.0)
9: 0x2b3bfc2c5688 <cs_run+0xa8> (libsaturne.so.0)
10: 0x2b3bfc2c6105 <main+0x1f5> (libsaturne.so.0)
11: 0x34d101ecdd <__libc_start_main+0xfd> (libc.so.6)
12: 0x400739 <> (cs_solver)
End of stack

Traceback (most recent call last):
File "/users/r01ken0/Code_Saturne/v2.0.5/cs-2.0/bin/code_saturne", line 127, in <module>
main(sys.argv[1:])
File "/users/r01ken0/Code_Saturne/v2.0.5/cs-2.0/bin/code_saturne", line 122, in main
commands[command](argv[1:])
File "/users/r01ken0/Code_Saturne/v2.0.5/cs-2.0/bin/code_saturne", line 71, in check_mesh
cs_check_mesh.main(argv)
File "/users/r01ken0/Code_Saturne/v2.0.5/cs-2.0/lib/python2.6/site-packages/ncs/cs_check_mesh.py", line 112, in main
retcode = run_check(argv)
File "/users/r01ken0/Code_Saturne/v2.0.5/cs-2.0/lib/python2.6/site-packages/ncs/cs_check_mesh.py", line 91, in run_check
dir_files = os.listdir('chr.ensight')
OSError: [Errno 2] No such file or directory: 'chr.ensight'


In the .log file (attached) it mentions a memory allocation error. I've made sure I have memory quota on the cluster and I don't really know what else to do. Any other suggestions?

Thank you,

Kristin

Hello,

It seems that u do not have enough memory :

Maximum program memory measure: 3647288 kB
Current program memory measure: 3581756 kB

Error in ECS execution
======================

System error: Cannot allocate memory

/users/r01ken0/Code_Saturne/v2.0.5/installer/ecs-2.0.2/src/base/ecs_vec_int.c:273: Fatal error.

Failure to allocate "this_vec_int->val_tab" (4478332224 bytes)


Call stack
1: 0x2ae5ae156836 <+0x5836> (libbft.so.1)
2: 0x2ae5ae15722e <bft_mem_malloc+0x13e> (libbft.so.1)
3: 0x47ee0e <> (cs_preprocess)
4: 0x45d497 <> (cs_preprocess)
5: 0x44e424 <> (cs_preprocess)
6: 0x455085 <> (cs_preprocess)
7: 0x45aa20 <> (cs_preprocess)
8: 0x40a1a5 <> (cs_preprocess)
9: 0x34d101ecdd <__libc_start_main+0xfd> (libc.so.6)
10: 0x403ea9 <> (cs_preprocess)
End of stack

How much free memory is available before launching the computation ?

Hello,

You are right, I did not have enough memory. I made 49GB available and no longer obtain the ECS error mentioned in my previous post, however, I now get the following error:

Memory allocation summary
-------------------------

Theoretical current allocated memory: 11866038 kB
Theoretical maximum allocated memory: 11866038 kB
Maximum program memory measure: 12042380 kB
Current program memory measure: 12041392 kB


System error: Cannot allocate memory

/users/r01ken0/Code_Saturne/v2.0.5/installer/ncs-2.0.5/src/base/cs_base.c:1098: Fatal error.

Failure to allocate "*ra" (18446744061741474368 bytes)


Call stack:
1: 0x2b968ad1122e <bft_mem_malloc+0x13e> (libbft.so.1)
2: 0x2b968884d2a7 <cs_base_mem_init_work+0x67> (libsaturne.so.0)
3: 0x2b9688808c6a <cs_run+0x68a> (libsaturne.so.0)
4: 0x2b9688809105 <main+0x1f5> (libsaturne.so.0)
5: 0x34d101ecdd <__libc_start_main+0xfd> (libc.so.6)
6: 0x400739 <> (cs_solver)
End of stack

According to the log file attached, the preprocessor finishes, but the 'calculation preparation' part fails. I have 50GB available in my home space on the cluster. Am I making another obvious mistake?

Thank you,

Kristin

Hello,

On how many processors did the solver run ? The solver data is distributed, and if you run it on a too small number of processors (I would guess around 2 million cells per processor, but this can be mesh dependent), things will fail...

Also, with version 2.0, memory management is not fully automatic (if I remember correctly), so if the multipliers you define for memory management are too high, things may fail earlier.

Regards,

Yvan Fournier