I meant there are solutions to work around this problem in code 
Also, as I don’t know the code of that filter, I don’t know whether the case I described happens there. Can also be something completely different.
I had this case in past a lot of times and for that reason I wanted to mention it.
I got some news from Samj/Gimpchat who is regularly compiling G’MIC and GIMP on Windows, and she told me that her version was way faster than mine.
Problem is that setting the compiling environment on Windows looks like real pain.
So maybe I’ll stop compiling my own executable on Windows in the future, and rely on samj’s compilations instead. Waiting for her answer about that
I guess I’ll be waiting too, David.
Any word from samj yet, David? Just curious.
Yes Lyle, samj has actually worked hard to send me a (huge) new environment compilation for Windows, including g++ 6.2.0 (quite recent).
I’m currently testing the compilation on my VM Windows, and if all is going well, I’ll be able to post new pre-release binaries and installers for Windows, maybe tonight or tomorrow at least.
1 Like
Thanks for the update, David. Regardless, it was cool to see my Ubuntu VM run the image in 5 to 10 seconds so hopefully samj’s new environment will result in a compile that will give similar results. 3 minutes was quite a bit of time. lol
So, @lylejk and @gubach, I’m happy to say there are new pre-release binaries/installers available for testing on the G’MIC web site : Index of /files/prerelease (dated from today).
These new Windows binaries/installers have been compiled with g++, version 6.2.0, so a quite recent one, and apparently this makes a huge difference
Particularly, the filter Deformations / Conformal maps now takes only a few seconds to run, in my VM Windows.
I’d be really interested by your tests with those new binaries.
Also, I’d like to thank SamJ (Sylvie Alexandre) for providing me with this new compiling environment !
And to be sure you are running the very latest version, check out the ENSICAEN logo, it has been changed this morning.
Also @lyle, please note that there are a bunch of new DLLs that come along with the new binaries, so be sure you discard the old ones !
1 Like
Installation without problem. (win 10 64 bit) Till now no errors. Thank you David. Yes, it is faster with this filter. Sylvie, also thanks.
David, is it possible that this also improved ( a lot) performance for Color mask [interactive]. Because I think it does. Or did you improve that filter before?
Thanks @iarga for your feedback.
Yes, actually there are chances all filters get some speed-up, particularly those using the math parser heavily (and yes, Color mask [interactive] does).
That’s why it’s quite cool and important to get feedback about this, just to be sure
The difference with the “Color mask [interactive]” filter is big. Now it is really usable. I love it. Thank you very much again. Next week I hope to have some time to investigate the “Colorize lineart [smart coloring]” filter.
With “Conformal maps” less time is needed. But with Anti-aliasing “on” it is still a time consuming filter on my pc. Nevertheless it is a cool effect.
edit: also huge difference with Smooth [patch-pca].
Wow; preview’s near instantaneous now (before I had to disable it since it all but would not complete) and I did a 1200px by 1200px with the custom formula and it did so in around 7 seconds. This build’s a keeper, David. Going to duplicate this message at GIMPChat. Thank you David and samj.
Thank you Lyle for your wish to have this in G’MIC. It brings more than only a cool effect. 
My only question is, will David recompile 1.7.9 since 2.0.0 is not released yet. Some of Brian’s custom python plugins won’t work with 2.0.0 unfortunately and I don’t think Brian’s going to update them unfortunately.
I haven’t planed to compile version 1.7.9 with this new compiling environment, but if there is really a demand, why not 
Happy to see everything is working as expected. I’m sure a lot of other filters also benefit from this optimization. I couldn’t imagine this could make such a big difference, only by upgrading the g++ version from 5.x to 6.2.
That’s again some Windows’ “magic”. Definitely not a trusty system for developers
Yeah; unfortunately, WIndows’s a beast; my bet is MACs are too but they do use OpenBSD so maybe not so bad.
Only reason I would like 1.7.9 is Brian’s flamboyance python plugin; don’t think he will do the changes for 2.0.0, but maybe I’ll try to ask him to do so once you make 2.0.0 final. Then, 1.7.9 won’t be needed.
Again, thanks to you and samj for making Windows G’MIC plugin great.