GIMP Version: 2.8.4
Operating System: Windows
OS Version: Win 7 - 32 bits
GIMP Experience: Beginner Level

List any relevant plug-ins or scripts:
GMIC - enhancement

List any ERROR messages you received:
failed to allocate memory

---

After having done in the previous days some tests here and there with sample images of limited size, I wanted to try experiment GMIC possibilities on a real photo, and by real I mean having a size which nowadays is quite usual, 6 megabytes, JPG.
I got many times the error indicated above, so I increased in Preferences - Environment - Maximum new image size (step by step) from the standard Gimp value up to (now) 756 Megabytes. I still get that message on GMIC functions like - for instance - Make Up.
So come my questions:
1-do I have to increase that value again? and to which value?
2-do I have to modify some other parameter in Gimp preferences?
3-is it normal that an input JPG image of 6 MB become a Gimp image of about 40 MB?
4-is it normal that some function of GMIC use in the intermediate steps so much memory?
Thanks a lot

How much memory do you have on your computer?

Some explanation: a jpeg file is normally compressed, so uncompressed in memory it will take up a lot more. And then you have the fact G'MIC works in 32 bit float per channel per pixel, which basically means 4 times as much memory usage compared to standard 8bit RGB. The calculation will be something like: height * width * 3 * 4 (bytes). And that's before any operations... if you use a filter that has to duplicate layers you instantly double the memory usage (and sometimes a lot more layers are needed).

Interesting K1TesseraEna, a 20000x10000 float image would be over 2GB and over 3GB with alpha channel.

Many thanks for the detailed info given.
Following K1T indications I modified my GIMP environment to these values:
1-Tile cache = 3 gigabytes (maximum allowed)
2-Image size = 1 gigabyte
3-Colour management = off
Still GMIC MakeUp crashes, with the following message:
"*** Error in ./make_up/edges/gradient_norm/*repeat/*local/ *** [instance(0,0,0,0,00000000,non-shared)] gmic<float>::assign(): Failed to allocate memory (182.3 Mio) for image (4608,3456,1,3)."
(additional info: make-up on a smaller image - 350 kb - works with no problem)
My system is as follows:
- IntelCore i5 750 2.67 GHz
- RAM 4 GB Dual Channel DDR3 668 MHz
- hard drives 2 * 1 Terabyte
Is this system insufficient to run on my photos GIMP + GMIC?
Very bad news, if so.....

I've tried using that filter on images larger than you say and it works for me (ram usage hits about 2GB), so can you give us some more info?

- G'MIC version
- exact size of the starting image (height, width, does it have an alpha channel)
- settings of the make_up filter

then I'll test to see if it goes wrong for me, thanks.

Thanks GC for your support.
Well, first of all, because - based on your questions - I suspected to have some wrong version of GMIC, I removed the gmic files and reinstalled using the installation file provided (gmic_gimp_win32_3.exe), downloaded on January 28, then I updated the definitions once entered into gmis to have the most current ones.
- on the GMIC panel the version shown is 1.5.4.0
- the image size is in pixels 4608 * 3456
- the image is jpg and open as a background layer, thus w/o alpha channel
- setting are the default (space 15, value 4)
I'm hoping you can reproduce the error.
(BTW after reinstalling gmic and updating scripts, the error is still there)

Hmm that's very strange, everything the same as mine except I'm using 64 bit. I'll test with 32 bit command line to see what happens.

*edit*
Good news (well in a way ), I get the same error with gmic 32 bit version from command line but not 64 bit. So we know two things: it's nothing to do with gimp and it must be to do with assigning enough memory in 32 bit (I think the limit will be 2GB of address space). What we do from here I'm not sure though

Thanks GC.
In the meanwhile, I have new infos:
- just doing a very little crop (image reduced to 4264 * 3168) MakeUp works
- even saving the result exporting it as png, reopening it, the script works
There is something magic in the original size?....

The image size must just be very close to the 32bit memory limit for that filter, it's possible the only way around it is to use 64 bit (which of course for you means a new OS!). There are tricks to get around this sort of limit in code but it can be slow and I doubt it will be done when 64bit works.

...also: the reduced (in pixels) image is much bigger than the original (being png, I suppose): 12.6 MB instead of 6.6 MB, which means that the problem is NOT related to MB size but maybe to pixel size ?!?

Yes that's right, the image file size on disk is almost irrelevant because different formats have different compression. All that matters once it's loaded is the dimensions (height, width, channels).

OK! Many thanks for your excellent support, now I'm sure it's not my fault.
I'll try to understand the exact limit in pixels for a 32 bits memory.
Hope some other gimp-er will benefit of the infos you gave me.
Thanks a lot again.
(btw, I don't now how to "close" a topic in gimp-chat...)

No problem, I'll add something to the gmic wiki about this since I've seen a few questions about large images lately.

I think is a more then reasonable result :
it is true that for nowadays pro digital camera size as 20000x10000 are not unusual

But on the other hand is expected a 64bit OS and plenty of Ram to edit image at that size,
for sure can't be done with a 32bit OS limited to 2GB ram max :
of that 2GB about 500MB are needed by the operative system to display itself and what left can't be sufficient to edit at similar sizes

Go to the first post in the thread and edit the subject line and put solved next to the title

The Make Up filter is mine, and I have updated it, to give more options and reduce memory usage.

There are now two options (fast and limit hue range) which are disabled by default, so that the the filter will run with about half the memory requirements. Previously you did not have these options and the were hard coded as enabled, so the filter was using more memory than always necessary.

The main part of the filter is also a bit more memory efficient.

The penalty is that without the fast option, the filter is much slower.

This tells me that the problems is with the 'limit hue range' section of the filter, so if you disable that in the updated filter, hopefully it will work. The results will be slightly different and I suggest set G'MIC to create a new layer for the output. Then you will probably have to add a layer mask in GIMP to have the effect only on the skin tones.