ljc
New Member
Posts: 4
|
Post by ljc on May 12, 2015 19:49:35 GMT
Hi
I am using v1.24.4 to run --mlma. My analysis fails to complete and I get the error message:
terminate called after throwing an instance of 'std::bad_alloc' what(): std::bad_alloc
Or in some cases the job just seems to get killed with no specific error message being printed.
The error seems to be intermittent as I have got the analysis to complete for the odd chr a few times.
I thought it was related to insufficient memory but I have managed to recreate it with a small test file (only 10 individuals and <20 SNPs).
Also it doesn't seem to make any difference whether or not I employ the --thread-num option or not.
I am submitting the analysis as a job to a cluster using PBS to submit a chr at a time (and specifying the number of cores to match the specified thread-num.
I tried v1.24.3 and found I got a segmentation fault instead.
Any suggestions about what the issue might be here would be very gratefully received!
BW
Laura
|
|
|
Post by Zhihong Zhu on May 12, 2015 23:51:17 GMT
Hi Laura,
Usually it is because of insufficient memory. Can you please double check the memory usage by typing "qstat -fx jobid"? And usually I didn't specify the number of CPUs by "--thread-num" when submitting jobs to clusters.
e.g. #PBS -N mkgrm #PBS -l ncpus=20 #PBS -l mem=15G #PBS -l walltime=2:00:00
gcta --bfile test --make-grm --out test > test.mkgrm.log 2>&1
Cheers, Zhihong
|
|
ljc
New Member
Posts: 4
|
Post by ljc on May 13, 2015 9:04:56 GMT
Many thanks, this seems to have sorted it.
|
|