Re: [MaxImDL] Replace With Masters: Memory Depletion/Leak


Nov 16, 2011

 


----------------------------

#47687 Nov 16, 2011

Hello Doug,



I finally captured the last of my faux-flats (a pain -- my EL panel was

too bright for 2x2 binning, but too dark after being dimmed with Rubylith).



So, I finally clicked "Replace W/ Masters" after an Auto-Calibration run.



It did its thing. It scanned over 5400 files in 135 groups, spanning

four cameras at various binnings, exposures, etc. Most groups had 40

darks or bias. A few had 80. Most flats had 20 each.



It created all 135 Masters, and placed them in the parent directory, as

expected.



When it was done, MaxIm 5.15 continued to consume 1.8 GB of RAM, even

after I'd clicked "OK" and closed out the "Set Calibration" dialog. I

have a screen capture of this if you want it. The memory drain did NOT

occur prior to "Replace w/ Masters" -- the scan and creation of the

groups did not allocate much RAM at all. It seems like the "Replace W/

Masters" button allocates a ton of RAM to keep things straight, and does

not free it upon completion. The more groups you have, the worse it is

(as there are more combinations to keep straight).



Just FYI. Nothing a restart of MaxIM doesn't fix, and I won't be

clicking "Replace w/ Masters" often..... but as a former programmer, I

detest memory leaks. I think they drove me from the industry in the

90's. :)



Jeff W

Reading, PA



----------------------------

#47694 Nov 17, 2011

On 2011-11-17 12:49 AM, Jeff Woods wrote: > Hello Doug,

>

> I finally captured the last of my faux-flats (a pain -- my EL panel was

> too bright for 2x2 binning, but too dark after being dimmed with Rubylith).

>

> So, I finally clicked "Replace W/ Masters" after an Auto-Calibration run.

>

> It did its thing. It scanned over 5400 files in 135 groups, spanning

> four cameras at various binnings, exposures, etc. Most groups had 40

> darks or bias. A few had 80. Most flats had 20 each.

>

> It created all 135 Masters, and placed them in the parent directory, as

> expected.

>

> When it was done, MaxIm 5.15 continued to consume 1.8 GB of RAM, even

> after I'd clicked "OK" and closed out the "Set Calibration" dialog. I

> have a screen capture of this if you want it. The memory drain did NOT

> occur prior to "Replace w/ Masters" -- the scan and creation of the

> groups did not allocate much RAM at all. It seems like the "Replace W/

> Masters" button allocates a ton of RAM to keep things straight, and does

> not free it upon completion. The more groups you have, the worse it is

> (as there are more combinations to keep straight).

>

> Just FYI. Nothing a restart of MaxIM doesn't fix, and I won't be

> clicking "Replace w/ Masters" often..... but as a former programmer, I

> detest memory leaks. I think they drove me from the industry in the

> 90's. :)

>

> Jeff W

> Reading, PA



That's not a leak, though it appears to be an unintended side-effect. Whenever

you load a calibration group MaxIm DL keeps it available in memory for future

use. Normally it only loads the groups you actually use, but it appears "use"

includes generating masters.



That's not a terrible thing to do most of the time, as most users don't use

zillions of calibration groups. We'll have to review how it works given your

use case.



By the way, you *don't* have to restart MaxIm DL to clear the buffers. Just

open Set Calibration again and click OK. That will keep your master frames but

flush all the buffers as they now have not been "used" since the settings were

"changed" (even though they weren't changed). After that it will only load the

specific masters it needs to process images you calibrate.



One more thing - MaxIm DL manages its memory utilization and swaps buffers that

haven't been recently used out to disk if needed. So it might hog system memory

in this case but it won't run out either.



Doug



--



Doug George

dgeorge@...



Diffraction Limited

Makers of Cyanogen Imaging Products

www.cyanogen.com/



100 Craig Henry Dr., Suite 202

Ottawa, Ontario,

Canada, K2G 5W3



Phone: (613) 225-2732

Fax: (613) 225-9688



----------------------------

#47699 Nov 17, 2011

On 2011-11-17 1:54 PM, Douglas B. George wrote:

> That's not a terrible thing to do most of the time, as most users don't use

> zillions of calibration groups. We'll have to review how it works given your

> use case.



We've reviewed this and have decided to toss calibration memory during the

"Replace w/Masters" operation. There's absolutely no reason to hold onto them.

We can reload them from disk when they are needed.



(We do need to hold memory if the master frames aren't generated; otherwise we'd

have to reprocess the stack for every image processed.)



By the way, *this* is far easier a feature request than that other thread. ;-)



Doug



--



Doug George

dgeorge@...



Diffraction Limited

Makers of Cyanogen Imaging Products

www.cyanogen.com/



100 Craig Henry Dr., Suite 202

Ottawa, Ontario,

Canada, K2G 5W3



Phone: (613) 225-2732

Fax: (613) 225-9688







Contact Us
This Site's Privacy Policy
Google's privacy policies

S
e
n
i
o
r
T
u
b
e
.
o
r
g