Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[freeformschooler]

And the complaints were getting more and more common as more updates came out with new content but no new fixes. 

Thats false, each of the release this year and last after the initial .31 release has had bug fixes. With an average of 18 or so per release.

Yeah, I completely forgot that like, every other update was almost exclusively bug-fixing.  I don't know how, since I always waited for the bug-patches to get an update.  But it did seem to me like people were getting more vocal over time, and I'm pretty sure that was just because the hospital bugs staying out of every patch was bugging lots of people.  I expect things to die down a little bit for a little while.

To be fair, mostly the bugs being fixed by those releases were the higher priority bugs introduced by the release that came just behind it. That's why they could come out so quickly. For this release Toady is focusing a lot on the older bugs that have been sitting around the bugtracker for some time.

[Orkel]

This would be amazing.

http://www.youtube.com/watch?v=rVhsIGGg3QE

[Urist Imiknorris]

That would be beautiful. I doubt it's possible in the near future though.

[Kogut]

   (*) Stopped massive lag from certain ghosts
What caused this bug? I am really interested in it.

[Urist Imiknorris]

I am too. Was it pathing, dealing with the removal of the ghost from the unit list, or something else that I can't think of?

[Dante]

My guess would have to be that some ghosts were using the normal A*-based pathfinding code, but with ALL tiles allowed (including hidden ones), making it incredibly expensive to recalculate.

[G-Flex]

My guess would have to be that some ghosts were using the normal A*-based pathfinding code, but with ALL tiles allowed (including hidden ones), making it incredibly expensive to recalculate.

Why would that make it more expensive? It's not computationally expensive to "pathfind" a straight line. Unless that path fails, obviously.

[The Merchant Of Menace]

Don't they still have to evaluate every possible square, in order to determine that they can go in a straight line to their target?

[G-Flex]

By "all tiles allowed" I interpreted that as meaning that all tiles are passable, in which case it obviously wouldn't be an issue. I don't know how ghosts move around in DF, though; do they travel through walls and fluids and such?

[The Merchant Of Menace]

I don't think they can, as far as I recall there was a case of a ghost throwing itself off a cliff repeatedly in an attempt to path to it's body once

[NW_Kohaku]

A* is a pathfinding method that tries to travel in a straight line directly to its destination unless it hits an obstacle, at which point it then tries to find a way around those obstacles.  In a situation with a wide open area, A* will simply pathfind directly to the destination and be extremely efficient.  In fact, wide open spaces with no obstacles is an ideal condition for A* pathfinding.

A* is only really problem if there are plenty of dead-ends for the pathfinding to go down, realize it's a dead end, then have to path back out, go down another dead end, and path around until it finally finds its way to the destination by trial-and-error. 

Ghost pathfinding (and possibly flying HFS pathfinding) probably did a flood-fill type of pathfinding, instead, which would contribute to the massive lag, since that WOULD involve checking every tile.

[Dante]

Yeah, that's what I meant I think. All tiles allowed to be checked for a route, instead of just the immediate adjacent non-fluid-covered non-walled non-empty-space ones.

I don't think they can, as far as I recall there was a case of a ghost throwing itself off a cliff repeatedly in an attempt to path to it's body once

Ghosts do travel through walls, floors, fluids, etc. You can see them hanging out inside your walls sometimes, or travelling in a straight line through a mountain.

[EmeraldWind]

Heck, I had a ghost path straight up through solid stone...

[G-Flex]

Yeah, that's what I meant I think. All tiles allowed to be checked for a route, instead of just the immediate adjacent non-fluid-covered non-walled non-empty-space ones.

That is not the difference we're talking about. NW_Kohaku is talking about the difference between something like A* and a basic flood-fill algorithm, although I don't know why anything would use the latter. Hell, something that can travel in a straight line doesn't really need pathfinding at all.

I don't think they can, as far as I recall there was a case of a ghost throwing itself off a cliff repeatedly in an attempt to path to it's body once

Ghosts do travel through walls, floors, fluids, etc. You can see them hanging out inside your walls sometimes, or travelling in a straight line through a mountain.
[/quote]

[NW_Kohaku]

That is not the difference we're talking about. NW_Kohaku is talking about the difference between something like A* and a basic flood-fill algorithm, although I don't know why anything would use the latter. Hell, something that can travel in a straight line doesn't really need pathfinding at all.

I can't remember where, but I remember in my searching for material on Pathfinding seeing a Toady quote that said that there is no connection map for fliers in the game.  This is why fliers don't work so well in fortress mode, except for wild animals. Toady said he let them do flood-fill pathfinding and that he just kept their searches short-range to prevent that from doing dire things to the FPS.  HFS pathfinding is probably very slow and causes dramatic spikes in lag specifically because they are using flood-fill pathing constantly.

It's possible Toady just made ghosts flood-fill pathfind entire maps until they found their destination to cut down on how much he'd have to code or take up memory for a secondary connectivity map that includes every tile, or something.  It would certainly explain the massive lag ghosts cause.