why is 4.2.2 so slow

[Sky]

Nick .. do you think it's normal that it takes me 5 to 10 seconds to dig a piece of rock ? Or the same time to rest?

Because this is what is happening. I'm near a vault - but really, any situation where a lot of mobs need to be accounted for - and whenever i do any *takes X turns* action, i have to see them bounce from side to side, probably trying to calculate a path to me that does not exist.

It's SO slow, it can take me 3/4 MINUTES to dig enough rock to get to the vault door, with 18/120 STR.

Nothing like this happens in 4.1.3, nothing like this has existed in ANY nightly build pre-4.2


Am i the only one here experiencing this lag ?

[MattB]

I am certainly experiencing annoying flicker, but nothing to the level you describe. What is your base delay set to?

[Nick]

This is something that gets reported from time to time, and it's hard to pin down the problem. Sometimes someone (occasionally me, more often someone else) finds a way to speed something up. Sometimes we don't really find anything. Sometimes it's not the game's fault, but something else to do with the user's computer (or Gwarl needs to reboot the angband.live server).

Are other people seeing this?

[Sky]

.. so, if it was my computer, then 4.1 and 4.2 would behave the same - which does not happen.

i have re-downloaded 4.1 and 4.2 both several times, every time 4.1 is fine and 4.2 is bad.

My base delay is at preset - i have not changed this in years.

[Pete Mack]

It is quite mysterious, as that part of the game is pretty stable.

Nick--
when's the last time you ran perf test with a bunch of (awake and mobile) monsters in ESP range, player resting or digging granite, and tiles enabled? It's pretty much the only perf test that matters. Once upon a time there were a number of performance options related to this issue. Compare it to a similar case in 4.1--preferably on both Windows and *Nix.

[Cuboideb]

If the tiles need to be resized, main-win.c does it for every grid, everytime the grid needs to be redrawn. I dont know if the rest/dig commands trigger a lot of redraw operations but in android I had to implement a tile cache (java made that easier).

Also, I noticed (perhaps it's android specific) that the rest command doesn't update the hp counter until fully healed.

[Pete Mack]

That is surely a regression. Or it is a really terrible longstanding bug.
I remember explicitly doing this for osx-carbon, circa 2003. With overhead map in a secondary window, you'd need two different sets, and possibly another for the monster list.

If the tiles need to be resized, main-win.c does it for every grid, everytime the grid needs to be redrawn. I dont know if the rest/dig commands trigger a lot of redraw operations but in android I had to implement a tile cache (java made that easier).

Also, I noticed (perhaps it's android specific) that the rest command doesn't update the hp counter until fully healed.

[Sky]

I have captured a short 1min video to show exactly how slow this version is:

https://youtu.be/QlWosGU6iBE

note that in 4.1.3 none of the mob movement would be visible - it would happen in the blink of an eye. Gothmog trying to path to me would be a blur. Each action here takes around 1second in 4.1.3 and .. well, see for yourself.

[Nick]

I have captured a short 1min video to show exactly how slow this version is:

https://youtu.be/QlWosGU6iBE

note that in 4.1.3 none of the mob movement would be visible - it would happen in the blink of an eye. Gothmog trying to path to me would be a blur. Each action here takes around 1second in 4.1.3 and .. well, see for yourself.

Can I ask you to do a similar test with Gervais tiles, and with ascii?

[Cuboideb]

It happens a lot with double sized tiles. I think that isnt so noticeable with gervais tiles.

[Sky]

Can I ask you to do a similar test with Gervais tiles, and with ascii?

contingent on finding another vault (which shouldn't be a problem), then yes. I'm sure i can wizard a character for testing purposes as well.
The thing is, again, these are the same settings as 4.1.3 so unless you know specifically that only one has changed, it will probably be the same.

it seems quite obvious that this is related to pathing, and just mobs "thinking" too much.

[Sky]

well, i take that back. It seems to work fine with Gervais tiles.

It's really confusing not recognizing any of the tiles but AFAIK the gameplay is as it should be.

Oh yeah it's the tiles. I tried recovering hp in sight of a large vault and it's quick as it should be with Gervais tiles (were they always this ugly?) but substantially slower with Shockbolt.

ASCII work fine as well.

[Nick]

Yes, the problem with the Shockbolt tiles is that the double-height ones mean there's a whole lot extra redrawing to make sure half-tiles aren't being left lying around everywhere.

[Sky]

ok but, why isn't this happening in 4.1.3 ?

Sorry to keep hammering the same nail but, you know, 4.1.3 works. With the same tiles. (and better gameplay)

[wobbly]

That's the problem with new angband. You need a high end rig to deal with the fancy modern graphics.