Re: new simplex-method constraint solver

On Sat, 2004-04-03 at 19:40 -0800, Adam Megacz wrote:
> Also, I think incremental solving (preloading the basis matrix with
> the correct solution from the last round of layout) would help a lot
> too, especially in situations where a box has 1,000 children and only
> two of them change...

Perhaps but you're not being fair to the current reflow concept because
there are a large number of inefficiencies built into the current
implementation.

I'm not certain we'll be gaining as much as you think we will, if at
all, and as such I may be tempted to speed up the reflow implementation
just so we have something fair to compare against.

Also, and I'm quite serious when I say this, you had _better_ test this
because I'm not going to spend another 2 weeks fixing it as I've just
had to with the current implementation.  Which is why I don't want you
committing it before we are certain it works.  Given the flexibility we
have using Darcs, it'd be unforgivable to commit this prematurely.

(And, frankly, it's very, very annoying to have my _tested_ and
_correct_ fixes completely ignored.  Especially when I know the chances
are you just won't test this properly, yet probably commit it anyway.)
-- 
- Charlie

Charles Goodwin <[email protected]>
Online @ http://www.charlietech.com