[Date Prev][Date Next][Thread Prev][Thread Next][Author Index][Date Index][Thread Index]
- To: <xanadu!michael>
- Subject: innovations
- From: Eric Dean Tribble <xanadu!tribble>
- Date: Thu, 1 Mar 90 10:07:29 PST
- Cc: <drexler@evilotto>, <terry@evilotto>, <xtech@evilotto>
- In-reply-to: <Michael>,03 PST <9003011730.AA11882@xanadu>
No, it's not April 1.
When do we get the explanation of how it works?
Certainly after it *works* for real. That should be later today. The
description will probably wait a few days till we wake up....
One amazing result is that a lot of the ent code is simplifying
further!!! The splay algorithm, the copy operations, the combine
operation, etc. all shrink! Further, the amount of space required by
the ent shrinks (we eliminated one pointer per inner loaf). This
makes the general case more space efficient than the special case in
the previous ent. When we really want to squeeze space, we can
probably eliminate one more pointer (with some ugly cleverness).
Incidentally, MarkM and I started implementing the arbitrary
non-overlapping combine. When we finish it, we'll be able to combine
any two orgls with disjoint domains, even if the simpleDomains overlap
(an orgl with odd indices vs and orgl with even indices). The time
would be proportional to the number of pieces we need to break the
orgls into to combine without overlap.