Olivier Chafik

Looks like this fails to compile w/ Epeck (calling `remesh_planar_patches(Surface_mesh())`): ``` In file included from /usr/local/include/CGAL/Exact_predicates_exact_constructions_kernel.h:27: In file included from /usr/local/include/CGAL/Lazy_kernel.h:18: /usr/local/include/CGAL/Static_filtered_predicate.h:180:40: error: member reference base type 'const double' is...

@sloriot Wow thanks! This code works amazingly and seems to fixe exponential growth of nested unions in [this model](https://twitter.com/braincode/status/1491547378435854338). Are there any caveats to using it? Would be amazing if...

@sloriot oh I see... "just" need to see which of these sibling faces are still standing at the end of the corefinement? Are there any existing helpers to do the...

> In your case, it might be even faster to keep track of the original faces to avoid the re-detection of planar parts. This can be done with the visitor...

@pca006132 after thorough tinkering (see https://github.com/openscad/openscad/pull/4643), I managed to bring down the cost of the toString / fromString routine by using an indexed mesh (which tesselate_faces was creating under the...

> how easy it is for alternate grammars to dramatically inflate the stack. For instance, imagine a grammar that says: > `root ::= [0-9] ("a"? "a"? "a"? "a"? "a"? [0-9])*`...

> ways to speed up the grammar tree navigation even with more poorly written grammars @HanClinto I'd be inclined to detect some easily rewritable grammar cases on the fly *and*...

> I'm also wondering about parse_sequence() in grammar-parser.cpp, and this section of rewriting +*? operators: > `// S* --> S' ::= S S' |` > ... > Makes me wonder...

I've taken a deeper look at how grammar sampling works and besides some possible tactical improvements (e.g. [extra reserve calls](https://github.com/ggerganov/llama.cpp/commit/a2747858d5151bf3d8f00317da2a4dec6adbfd5e) & [reusing stacks vectors](https://github.com/ggerganov/llama.cpp/commit/959d9d34da89dc1fd186b392636ef10ce3143487) combined give over 20% speedup on...

> The first was attempting to modify reject_candidates to modify the candidates array in-place to reduce allocation of vectors, but that still has a bug in it and I haven't...