I'm using r.contour to extract contours from a DEM raster. Everything works fine until I try to do a region over a certain size. Then r.contour just quits spontaneously. If I run it from the command line, all it says is "Killed".
I'm sure it's a memory/size problem, and the routine take a long time, but I just want to set it to work and wait for the response. Is there anyone who understands r.contour enough who would care to do some debugging? I'd be happy to work together on this.
I'm using r.contour to extract contours from a DEM raster. Everything
works fine until I try to do a region over a certain size. Then
r.contour just quits spontaneously. If I run it from the command line,
all it says is "Killed".
I'm sure it's a memory/size problem, and the routine take a long time,
but I just want to set it to work and wait for the response. Is there
anyone who understands r.contour enough who would care to do some
debugging? I'd be happy to work together on this.
try running "top" in a terminal window along side and watch the memory
grow.
I've watched the memory use climb while the CPU cycles drop to 1 or 2 percent. Presumably it does into a swap frenzy. The same thing happens whether r.contour finishes successfully or not.
How would I measure the size of the areas? I'll look for the area sizes later today. Gotta go to work now...
Michael
Hamish wrote:
I'm using r.contour to extract contours from a DEM raster. Everything works fine until I try to do a region over a certain size. Then r.contour just quits spontaneously. If I run it from the command line,
all it says is "Killed".
I'm sure it's a memory/size problem, and the routine take a long time,
but I just want to set it to work and wait for the response. Is there anyone who understands r.contour enough who would care to do some debugging? I'd be happy to work together on this.
try running "top" in a terminal window along side and watch the memory
grow.
I've watched the memory use climb while the CPU cycles drop to 1 or 2
percent. Presumably it does into a swap frenzy. The same thing happens
whether r.contour finishes successfully or not.
how big does the memory get? ("M" in top will sort by memory if the
process slows down) Is that bigger than what you have installed?
Usually a process gets killed when you run out of memory+swap space and
the OS has to take drastic action to keep the rest of the system from
crashing.. it kills out the worst memory hog.
How would I measure the size of the areas?
I was interested in the raster region resolution.
'g.region -p' will show number of rows and columns.
v.info will show the number of vector entities involved.
Hamish, thanks for this; it looks like you've got it. I can see the memory dissapear towards zero, and then there's a long pause in gui response, and r.contour is no more...
I'll have to see if I've got an old disk lying around to add for more swap. Or an old wallet with money for more RAM.
If it's still of interest, I have 256M RAM, 512M swap.
At
rows: 8400
cols: 7200
r.contour can complete successfully.
At
rows: 8400
cols: 7680
r.contour gets booted off before completing.
Thanks again,
Michael
Hamish wrote:
I've watched the memory use climb while the CPU cycles drop to 1 or 2 percent. Presumably it does into a swap frenzy. The same thing happens
whether r.contour finishes successfully or not.
how big does the memory get? ("M" in top will sort by memory if the
process slows down) Is that bigger than what you have installed?
Usually a process gets killed when you run out of memory+swap space and
the OS has to take drastic action to keep the rest of the system from
crashing.. it kills out the worst memory hog.
How would I measure the size of the areas?
I was interested in the raster region resolution.
'g.region -p' will show number of rows and columns.
v.info will show the number of vector entities involved.
Hamish, thanks for this; it looks like you've got it. I can see the
memory dissapear towards zero, and then there's a long pause in gui
response, and r.contour is no more...
I'll have to see if I've got an old disk lying around to add for more
swap. Or an old wallet with money for more RAM.
If it's still of interest, I have 256M RAM, 512M swap.
At
rows: 8400
cols: 7200
r.contour can complete successfully.
At
rows: 8400
cols: 7680
r.contour gets booted off before completing.
For a region that big you will either want a little more memory (another
512mb chip might do it), add some temporary swap space* and let it run
overnight, or split the processing into a north and south half (or
quadrants, etc.) and then put the contours together with v.patch
(v.append?). I think upgrading your RAM is the best solution if you can
find the cash.