Hi Matthias,
first of all: Which OS are you working on and how do you measure how much
memory is used for an image?
This is what I know:
-We render video streams in MITK without increasing memory --> There is no
actual memory leak which gets bigger every render() call
-The build in memory display does not give accurate results and is different
from OS to OS. I think most of time you get the memory which is allocated by
the OS for the mitkWorkbench, which is usually quite a big number, because it
blocks memory just in case.
-The value should also depend on your open plugins etc..
As long as everything is rendered and there is no crash, I would not worry too
much about it. There is an entry somewhere in our documentation explaining the
memory display, but unfortunately I don't know where.
Regards,
Thomas
From: Noll, Matthias [mailto:[email protected]]
Sent: Donnerstag, 17. Oktober 2013 16:11
To: [email protected]
Subject: [mitk-users] Possible rendering memory leak?
Dear list,
I observed a strange behavior of MITK while rendering a lot of images.
First, I have an image that is 3200x2448 pixels large. Also the image is
divided in ~480 tiles. Each tile has a size of 128x128 pixels. Calculating the
size I get ~48kB of raw data per tile. That's 22,4MB of raw data per image
using RGB values. Now if I try to just load a portion of the tiles into MITK,
the memory that MITK utilizes is exploding as soon as MITK renders the images.
I tested this using just image 3 rows with 19 tiles each => 57 tiles total. The
memory requirement of MITK while rendering the images is 940MB. However If I
select a second, preloaded image and don't initialize the render view after
loading the tiles, I just get a total MITK memory usage of 350MB. Starting MITK
had a memory size of 273MB. Further, selecting 6x19 tiles I would have expected
something like 1,8GB with the expensive rendering. But I received a value of
2,5GB MITK memory used. That's nearly triple for twice the images. At which
point I wonder WTF is using all that memory for a 23MB image??
I guess you don't usually load that many tiny images to MITK but I can't
explain this due to OOP overhead.
Best regards,
Matthias
------------------------------------------------------------------------------
October Webinars: Code for Performance
Free Intel webinars can help you accelerate application performance.
Explore tips for MPI, OpenMP, advanced profiling, and more. Get the most from
the latest Intel processors and coprocessors. See abstracts and register >
http://pubads.g.doubleclick.net/gampad/clk?id=60135991&iu=/4140/ostg.clktrk
_______________________________________________
mitk-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/mitk-users
