Lighting Performance Regression versus 1.3b91

[jamztheman]

Lighting Performance Regression versus 1.3b91

[JamzTheMan]

Thanks for that map/campaign! Something bad indeed is going on here! I'll dig into this asap.

[Syndaryl]

I am subscribing to this issue as I'm also having big trouble with it. My map is four large (page sized) PNGs with black for the walls, transparent for the not-walls, and I used the token auto VBL feature to set up my vbl one click. Well, five clicks but still.

When it was one PNG, it was OK.
When I placed a second, it got worse.
When I placed the third and fourth it got impossible.

Turning vision to Day removes the problem, but I was rather appreciating that my players couldn't see the big picture, and didn't realize how big the space they were in really was. Once I set it to Day so we could play they kinda flipped out :)

[JamzTheMan]

@Syndaryl Your issue may be separate as you are using "new" functionality, namely TokenVBL. TokenVBL is more intense as it has to calculate it more often since it can "move".

Try transferring it to the normal VBL layer using a macro like:

[r: transferVBL(1)]
[r: finalVblData = "{'shape':'none'}"]
[r: setTokenVBL(finalVblData)] 

FYI: You can transfer VBL to the token via from the map layer:

[r: transferVBL(0)]

Also, Auto-Generated VBL can be very poor performing if too detailed. Like, if you manually drew around every single pixel, it would also be bad. If it's basically just straight walls, auto-generation should be ok.

FYI: You can drag/drop files right into the "issue" here so feel free to supply the campaign file for testing.

[dracwyrm]

I hadn't had a lot of time as of late, but maybe I should revisit the idea of using faster math operations for VBL? Default math in Java is set to high precision, but that is not needed for this, so the faster math libraries can be used. Trouble is, you can't override the math library with your own.

[JamzTheMan]

I made some improvements to Lights for the next release. This campaign file should be usable under 1.4.5.0 (just tested it) and I'll continue to look into this. There may still be room for additional improvements.

Note: You may want to clear FoW to see the best improvements. The fix involves changing lights from a true circle to a polygon which is 10-100 times faster to calculate the intersection with existing FoW. But if existing FoW already has "curves", performance may be poor. YMMV.

[mlangsdorf]

is there a beta build of 1.4.5 that I could test?

[JamzTheMan]

I just created one for Windows/Mac. Appveyor isn't uploading the deb package for linux for some reason and looking into it...

https://github.com/JamzTheMan/MapTool/releases/tag/1.4.5.0-beta2

[JamzTheMan]

Should be fixed as of 1.4.5.3.