Memory Panic
mohack wrote on Jul 05, 2009
Hi!
I tried to use imcombine to build a real big mosaic image out of several smaller ones with imcombine. Combining the smaller images worked fine, but when I try to combine the mosaic, I get the error "PANIC in `/usr/iraf/iraf/bin.linux/x_images.e': Memory has been corrupted"
I don't know what to do, already read some posts in here but I can't figure it out. IRAF 2.14 is running on a 64bit machine.
Any help?
I tried to use imcombine to build a real big mosaic image out of several smaller ones with imcombine. Combining the smaller images worked fine, but when I try to combine the mosaic, I get the error "PANIC in `/usr/iraf/iraf/bin.linux/x_images.e': Memory has been corrupted"
I don't know what to do, already read some posts in here but I can't figure it out. IRAF 2.14 is running on a 64bit machine.
Any help?
Mike Fitzpatrick wrote on Jul 05, 2009
How big is a "real big mosaic"? You may just be hitting the 2GB file limit of the 32-bit binary if this means e.g. a 30,000x30,000 pixel image, or if the number of images means the task is working with a lot of memory. If you're certain the image is smaller than 2GB by a fair amount, it could also be a new bug in the task, but we'd need more information such as parameters and perhaps data to reproduce it along with whatever other information you can provide.
-Mike
-Mike
mohack wrote on Jul 05, 2009
Well, the single frames which are to be comined are about 7.5GB (12 sections, each taken with a 4 chip mosaic CCD x 10 frames x 16MB (2048x2048px per Frame)) alltogether, the mosaic after the combination should be about 12 times the 4 CCDs, which would be 12 x 64MB = 768MB. Is there a way, imcombine can handle that? I'm a little confused though, because we combined another mosaic out of this data earlier on another slower single core 32Bit machine (maybe with one or two frames less)...
Mike Fitzpatrick wrote on Jul 05, 2009
In all probability this is one of three things: 1) a symptom of the 32-bit limitation in the system, 2) a limitation of running a 32-bit binary on a 64-bit system, or 3) an unknown bug of some kind (or of course 4, some combination of each).
If you're trying to combine all 7.5GB at once (12 sections times 10 frames) then this is almost certainly #1 above. 120 files is also starting to push the limits of the task. Do smaller mosaics work (e.g. 6 sections or only 5 frames)? Your last statement about this working previously on a 32-bit machine is curious, was it the same 7.5GB of data or a smaller mosaic?
If you're trying to combine all 7.5GB at once (12 sections times 10 frames) then this is almost certainly #1 above. 120 files is also starting to push the limits of the task. Do smaller mosaics work (e.g. 6 sections or only 5 frames)? Your last statement about this working previously on a 32-bit machine is curious, was it the same 7.5GB of data or a smaller mosaic?
mohack wrote on Jul 05, 2009
Yes, I try to combine them at once. If I would combine the single frames first to get 12 master frames, and then combine those, the overlap of the edges would not be as good as the combination of all frames simultaneously beacause of the averaging.
Combining smaller mosaics (4 of the 12 sections) does work.
The ealier combination was the same big mosaic.
Combining smaller mosaics (4 of the 12 sections) does work.
The ealier combination was the same big mosaic.
Francisco Valdes wrote on Jul 05, 2009
If you continue to have problems you could send the output of "lpar imcombine" and header listings (from imhead) for the images. With that I could determine what output imcombine is trying to produce. The header listings might be large so you can send them to me by email at valdes@noao.edu. There should be no problem with 120 images though if they have exceptionally long headers or if you are using masks that could eat up memory.
Good luck,
Frank Valdes
Good luck,
Frank Valdes
Last post on Jul 05, 2009