[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[GEMPAK #BRA-222690]: dcgrib2 memory leak

> Daryl,
> 
> I'll have to take a look at this. The NCWD is the largest grid
> being decoded at the moment (almost). Its 1.6 million points
> and will grow the malloc'd space , so I'll pass things through valgrind
> too watch the memory use. The 1/12 degree SST is 9 million
> points which can really look big in memory as well.
> 
> Steve Chiswell
> Unidata User Support
> 

Daryl,

I have tracked down the memory leak related to the bit masked grids such 
as the NCWD grid. The problem was a missing static declaration for the pointer 
and size 
so that each time the bit mask was realloc'd, it was creating a new block.

I have a source code fix, or can post a binary of just the dcgrib2 decoder for
the 5.10.4 distribution, whichever you would like.

Steve Chiswell
Unidata User Support




> 
> 
> > Hi,
> >
> > I've been noticing a dcgrib memory leak while decoding this:
> >
> > HDS     ^[YZ]..... KKCI.*/mAWC_NCWD
> > PIPE    dcgrib2 -d data/gempak/logs/dcgrib2_AWC_NCWD.log
> > -e GEMTBL=/home/ldm/nawips/gempak/tables
> > data/gempak/model/awc/YYYYMMDD_ncwd.gem
> >
> > I'm rolling with 5.10.3 UPC binaries x86_64 Linux on RHEL5.
> >
> > Here is snapshots of the process in the last 10 minutes
> >
> > ldm      11410  0.1  0.4  53000 16260 ?        S    17:38   0:00 dcgrib2
> >
> > ldm      11410  0.0  0.7  72696 30204 ?        S    17:38   0:00 dcgrib2
> >
> > ldm      11410  0.0  0.9  79256 36768 ?        S    17:38   0:00 dcgrib2
> >
> > It grew to 8 GB in just under 2 days :(
> >
> > Nothing special in the log file.
> >
> > thanks!
> > daryl
> >
> >
> 


Ticket Details
===================
Ticket ID: BRA-222690
Department: Support GEMPAK
Priority: Normal
Status: Closed