@cfr This is getting out of my hands ;-( I think I'll post a bugfix release this weekend and then work on performance some more. Slowly. Trying to think about what I do. Argh.
@cfr Errors and still slower than expected. My simple tests show that it's 20-30% faster than pgfmath. But I want to test on a real example, read prooftree. So right now, I'm struggling to make your examples work.