FEAP User Forum
FEAP => FEAPpv => Topic started by: agni_roni on January 25, 2018, 05:51:22 PM
-
Hi,
I have written the attached input file for a square material specimen under patch load. When discretize it into 32 X 32 X 4 solid elements I get some results, but when I discretize into 64 X 64 X 8 the compiler throws me the following error,
Program received signal SIGSEGV: Segmentation fault - invalid memory reference.
Backtrace for this error:
#0 0x10d180579
#1 0x10d17f945
#2 0x7fff6b8f9f59
Segmentation fault: 11
I am stumped.
Thanks for the help in advance.
-
Did you run out of memory? Check the requirements after the mesh is input to see how much memory the solve will require.
In interactive mode issue: SHOW it should tell you how much memory is required for the tangent. Then issue TANG,,1 and it should fail almost immediately probably.
-
Thank you Prof. Taylor.
I issued the 'inter' command right before the batch where the tangent matrix was not formed and then in the interactive mode I input 'tang' only. The program stopped immediately with the error output as before. Any way of knowing the tangent matrix's probable size?
Does this mean I ran out of memory?
-
Also, when I typed in 'show', the number of profile terms were shown as *********.
-
The profile is rather large. But on my computer with 16GB RAM it will work. Though you should really run OPTImize on the profile otherwise it will take forever if you are using the profile solver.
As input:
Number equations = 113292
Number tang terms= 626766876
Average col. ht. = 5533
Command 1> opti
Number equations = 113292
Number tang terms= 240478020
Average col. ht. = 2123
(Took 82 seconds on my laptop)
Note you can also use pardisio (without OPTI); in that case it takes 11 seconds on my laptop.
-
Thanks for all the help Admin.
Just another question, is the pardiso solver available for use with FEAPpv?
-
The solver has only been tested with FEAP.
-
Thank you, Admin. Using the optimize along with the iterative solver I have been able to bring down the computational time to just under 50 s.
Thanks a lot for the help.