Is there a faster alternative to `hcat(xx...)`?

General Usage Performance
2

You could use reduce(hcat,xx)' (see How to convert Vector of Vectors to Matrix - #6 by rafael.guerra ), or in Julia 1.9 you can use stack.

To improve speed further, you could maybe post a minimal working example so that people can run the code. (Not visible here, but just a guess: Make sure that you don’t use globals by accident.)

There might also be a way to write the code differently, like

function filter_triangles(triangles, tokeep)
    counts = map(t -> tokeep[t[1]] + tokeep[t[2]] + tokeep[t[3]], triangles)
    keep = counts .> 0
    new_triangles = triangles[keep]
    return (counts = counts[keep], 
            triangles = new_triangles, 
            connec = connect.(new_triangles))  
end

(This uses more memory, but depending on your setting, it might even be faster.)

EDIT: As @fatteneder pointed out, I removed my wrong comment that Int is an abstract type…