GPT Sentinel 1 TOPSAR coherence graph: GC overhead limit exceeded

I’m quite new to this so apologies if this is an obvious mistake.
I’m having a memory issue using GPT to process a Graph that works without fault (with the same input data) in the desktop application.
I am receiving the following error messages after the line below is run:

$SNAP_HOME/gpt $GRAPHSDIR/Coherence_graph.xml -Pinput1=$PRODUCT_DIR1/ -Pinput2=$PRODUCT_DIR2/ -Pinput3='IW1' -Pinput4='VV' -Ptarget1=${OUTPUT_FOLDER1}/"${OUTPUT_ROOT_NAME1}_IW1_CC.dim"

GC overhead limit exceeded
org.esa.snap.core.gpf.OperatorException: GC overhead limit exceeded
Error: GC overhead limit exceeded
– org.jblas INFO Deleting /tmp/jblas5199528369946157133/
– org.jblas INFO Deleting /tmp/jblas5199528369946157133/
– org.jblas INFO Deleting /tmp/jblas5199528369946157133

Looking at my system monitor shows that the actual memory limits of my machine are not reached during the process, which cuts out after about 10 seconds. Would altering the memory allocation manually help, and if so how is that done within the command line environment?
The graph xml is : Coherence_graph.xml (4.7 KB)

My version and installation conditions of SNAP are:

SNAP Desktop implementation version: 3.0.3
SNAP Engine implementation version: 3.0.3
JRE: Java™ SE Runtime Environment 1.8.0_60-b27
JVM: Java Hotspot™ 64-Bit Server VM by Oracle Corporation
Memory:8889 MiB
Linux (Ubuntu 16.04 LTS)

Thanks for any advice,


Found a solution at

This is what I did:


compute the best system options (the top compute button) and copy the VM parameters from the top line

gedit $SNAP_HOME_DIRECTORY/gpt.vmoptions

paste each parameter from the top line in snap-conf-optimiser to a new line in gpt.vmoptions

Hope that helps others with same issue.



Very interesting, I will try because some times I have similar issues.


1 Like

Hi James,

I have been facing the same problem while carrying out the TOPSAR coregistration, but in my case the operating system is windows 7. Shall the same steps suffice? Cause the second step, as you have mentioned, is difficult to execute.