Hmm, something in there is making things slower, not sure what. No time to test today, already spent too much time on this :) I did manage to get a speed bump with #1160 (test suite from 5.0s to 4.9s) but still getting slower on my test model, not sure why.

You've included the commits from #1158 in this. Can you remove those?

And thank you for doing all this!

Sure, guess I can split out a few of mine from here. The idea was that the build system in the original pr wasn't great, and all but one of my eigen changes were little "peephole" optimizations based on hotspots i saw when profiling 4 years ago 😅 we'll see when I get a chance, I did way more Solvespace work the past few days than I was supposed to

Oh, ok, so I got the PR's mixed up in my head. The changes in here are necessarily dependent on at least a few commits from #1158. I did drop most of my (minimal) changes from #432 because I thought that's what you had asked for 🤦 . I also dropped the "remove too-many-constraints" commit for now. In any case, this is now rebased, builds on #1158 and by extension #1155 . Rebasing now in a different branch to drop as much of #1158 as possible.

OK, the version of this with as little of #1158 as possible, is here: master...rpavlik:evilspirit-eigen-over-ryan-buildsys-fewer-solver-changes I could drop the first three commits of #1158 but the rest were built upon by this code.

@rpavlik Looks like the first 9 commits here conflict since they were merged in the previous PR. What's the best way to handle that?

Weird, must be squash merging or something. I'll just rebase this if you don't beat me to it.

OK, rebased!

@rpavlik Lets see how this goes ;-)

@rpavlik question, in testing I ran into "Too many unknowns". The define is still set at 1024, but I thought the code was modified to not have an actual limit. Should we remove some checks, or increase the value?

BTW I made a long chain (snake) of connected rectangles much like evil_spirit showed in a video. This runs dramatically smoother when dragging it around. ;-)

Closes #225

This is an awesome way to start the year!! Happy new year everybody! Thanks @Evil-Spirit and @rpavlik for making this happen!!

Awesome, thanks! The limit is now artificial, but left in to avoid getting to a point of being unresponsive. We can increase the define for Max unknowns as we see fit. Probably double it if we're a lot smoother now. (I do still have a commit here that I'll put in a draft pr maybe that totally removes that error, but it does seem reasonable to have a limit to avoid painting ourselves into a corner)

After the fact I code-reviewed this fairly carefully and it is fine! Thank you very much for cleaning in up, making the integration clean and rebasing it @rpavlik and thank you very much @Evil-Spirit for writing it in the first place back in 2017 :-)