lag in 4.2.1

[Sky]

the lag, i can tell you is not tied to resting, but very likely to mobs sleeping. Because - while the video is of me resting (as i do not know when i'll run into a ton of mobs sleeping, dont want to keep the recording running and then have to trim a 17Gb file) - this mostly happens when there are mobs sleeping nearby. It *also* happens when i rest, but try to enter a vault and you will see. Think, 5 seconds to be able to do 1 move.

[Gwarl]

Originally Posted by Nick

Yes, this is unsurprising. I think there are three main things that account for it, which were deliberate design decisions on my part:

1. More frequent screen updates, so that the player can see monsters moving while they are running or resting;
2. More detailed monster AI, or more precisely less shortcuts in monster AI;
3. Sacrifice of efficiency for readability of code.

All these I deemed to be reasonable on the grounds of improving the game (and its maintainability) at the expense of efficiencies which were no longer required because of increase in computer speeds since Angband was first written.

None of this is to say that efficiencies can't be made (and the noise update one I mentioned is an example), but at a certain point it becomes a trade-off between efficiency and other game features. And I'm a bit reluctant to cut features which improve the quality and immersion of the game for the sake of those who are choosing to use massively inefficient graphical tiles or transmit all their keystrokes halfway around the planet before they have an effect

I'm of the mind to do some purely abstract numerate hacking, if you can specify the inefficient functions I can give it a try. Particularly if we can decouple it for testing.

I know I should be doing various other stuff but an abstract problem like algorithmic efficiency might be relaxing.

[geoff_tewierik]

I noticed lag recently in my game on my laptop and taking note of others post, I wondered what had triggered it.

Once I moved a bit further down the corridor I came across a pack of hounds, so I wonder whether that was the cause? Were they sleeping?

Sorry, no save file.

[Cuboideb]

Originally Posted by Nick

And I'm a bit reluctant to cut features which improve the quality and immersion of the game for the sake of those who are choosing to use massively inefficient graphical tiles or transmit all their keystrokes halfway around the planet before they have an effect

It's okay. The shockbolt tiles are very resource-intensive. For android I had to break the tileset into 8 smaller pages, storing the tiles in a secondary less-recently-used cache to load and resize just what is needed.

I've noticed that the game do more refreshs when the size of the tiles is greater than 1 (more chance to have garbage on the screen). And... shockbolt has double sized tiles. That's a lot of overhead.

[Sideways]

Originally Posted by Nick

All these I deemed to be reasonable on the grounds of improving the game (and its maintainability) at the expense of efficiencies which were no longer required because of increase in computer speeds since Angband was first written.

Old V had many efficiencies, but also many efficiency options for turning them on or off as needed and desired; and while AI or code modifications are a bit hard to just toggle on or off, that solution would still work for screen update frequency.

[Nick]

Originally Posted by Sideways

AI or code modifications are a bit hard to just toggle on or off

Tiles, on the other hand...

[Sky]

if shockbolt tiles are resource intensive, why doesn't 4.1.3 have lag under the same circumstances?
Also, which exactly are these resources? RAM ?

[Nick]

Originally Posted by Sky

if shockbolt tiles are resource intensive, why doesn't 4.1.3 have lag under the same circumstances?

Well, that's the question. There have been nearly 1000 changes made since 4.1.3, so tracking down what did this will take a bit of work.

Originally Posted by Sky

Also, which exactly are these resources? RAM ?

Probably a combination of CPU cycles and graphics memory. I'm not sure how efficiently the Windows routines we use (mostly written pre-2000) use Windows 10.

[Cuboideb]

Memory mostly in mobile. There is a lot of tile resize happening too. That can be improved with tile caches, but they are specific to each port.

[Pete Mack]

Redraw while resting and running is a huge hog. Has anyone compared tiles with no tiles directly?