winXP 1.1.4 slow as hell

[packaleks]

hey guys,

just wanted to share my personal experience with the latest version of the game. cant help to say that though i followed all advice to get it running after all i was quite disappointed to see it not running smoothly after all. the cursor lags behind. the AI moves slower and slower each version.

how can such a small program run so badly on a PIV with 3ghz and 1gig ram?

its getting unplayable this way. if u call this a fix, i definetely want to know if u test those before they get uploaded.

ah yes.. and i would like to comment on the joke i read in another topic before. quote: "its only a developers version.. just stick to the stable one". this could be an option if the official servers and such wouldnt be updated with every new buggy version thats released.

sometimes it really seems like u guys are completely forgetting how many windows users actually participate in this project. sorry for the harsh words but im really fed up with some linux-users arrogance posts.

[freim]

hey guys,
ah yes.. and i would like to comment on the joke i read in another topic before. quote: "its only a developers version.. just stick to the stable one". this could be an option if the official servers and such wouldnt be updated with every new buggy version thats released.

sometimes it really seems like u guys are completely forgetting how many windows users actually participate in this project. sorry for the harsh words but im really fed up with some linux-users arrogance posts.

We run 3 servers:
Stable server for 1.0.x
Development server for 1.1.x
Bleeding edge server for latest svn trunk

So I advice you to get your facts straight before throwing around insult.

As for the speed, it has gotten a lot slower unfortunately. I always compile with optimizations for decent speed on Linux. Maybe Win has additional problems we don't see, although we do have developers using win also, they will have to answer for that part.

[packaleks]

well, apology about the tone again, anyhow it does not matter, to have 3 servers if more than half of our friends play the newest wesnoth version anyway.

so what do u gain if you have a nice lil server on the old version, and half of your buddies with linux systems play around on the new one.
u only get a split community ..

this post is not meant to critizie minor bugs which are a normal process of developement. i just dont get it into my head that lately everyone put up and sells stuff that just doesnt work. people spend hours and days coding and working their asses off. (and they usually really do a nice job here). so why always the hurry to upload something which isnt even tested wheter it will at least start? sorry, but i dont get this.

anyhow. the performance problem is still there, no matter wheter u agree with my post or not.

[freim]

anyhow. the performance problem is still there, no matter wheter u agree with my post or not.

I just edited my post with a reply to this, and I basicly agree that performance isn't optimal. There are work done to try to optimize, but it isn't trivial work.

[torangan]

There are two problems at play here.
1) The developers don't notice a huge speed decrease for some reason. We've got one person working on a 500 Mhz Athlon stating that the speed of the latest version is reasonable again. On your 3000 Mhz P4 it's not. Rather strange, isn't it?
2) Wesnoth is getting more complex. It's not only the visibile stuff like new animations but also more features for WML. This comes at a price and fixing those performance problems is going to take some rather serious code changes. Parts of the engine were written with assumptions that no longer hold. Rewriting core parts of the engine is very likely to make Wesnoth unstable again, so you might see the problem?

[Ailurus]

One idea I have reading this. Do you have the accelerated speed option on? If you are comparing to 1.1.2a, you should note that in that the windows binary for that one had the accelerated speed option was on by default, but if I recall correctly it isn't in the windows installer for 1.1.4.

If its a problem with slow animations and such (like a fight taking 5 seconds or so to resolve), then the accelerated speed option might be the culprit. However, if its something like the AI taking a long time to decide where to move a unit, then it probably isn't.

(I'm running the windows binary for 1.1.4 on winxp, P4 3.02 ghz, 512 ram and it works fine for me. There is no noticible lag for any computation, the only thing is the time for the animations, which, with accelerated speed on, is pretty short in my opinion).

(Edit - I've only played around with this on Single Player, so I don't know if there are performance issues in MP)

[Sapient]

One of the reasons it appears slower in multiplayer is because it is autosaving at the beginning of each turn instead of the end, and the size of savefiles has increased.

There is some discussion going on about how to remedy this.

[Prometheus]

I run on Windows, and I just finished testing the first scenario in Heir to the Throne, on both 1.02 and 1.14. The options, e.g. accelerated speed were the same. (I compiled 1.02 myself and used the downloadable binaries for 1.14)

Both games lasted 10 turns and were generally very similar. The 1.02 version clocked in at 9 minutes 45 seconds. The 1.14 version clocked in at 18 minutes 12 seconds.

I'm not trying to insult anyone, but this performance is unacceptable, in the sense that I prefer to use 1.02. In my opinion this is a VERY SERIOUS ISSUE which the developers really need to focus on, to the exclusion of other things actually. If you can't get the speed up, you are in my opinion making things worse not better. I'm sorry to say so, because I know you are working hard. But Wesnoth wouldn't be the first game to get progressively worse with additional releases.

more features for WML.

That confirms a suspicion I've had- the slowdown is related to WML. A lot of WML features involve checking each unit against some condition whenever a particular event occurs. With a lot of units on the board, and a lot of events occuring, these checks are going to be processor intensive no matter what you do. Frankly I was a little surprised 1.02 WML could handle the events it did without slowing down, which is why I thought the complaints about speed were always a little misguided.

[ILikeProgramming]

more features for WML

I doubt this is the major problem. I play UTBS without accelerated speed, and it's OK.

...but, whne I try to save or load, the game takes F O R E V E R.

There is also a puse at the beginning of the turn when I'm about to move, but the cursor won't budge on fullscreen. This is probably autosaving (imagine autosaveing every turn...)

[Prometheus]

Maybe it's not so noticeable without acceleration. The moveto events I suppose take up proportionately less time. For sure something is slowing it down.

[torangan]

It seems like there's a chance. Xan made a build using GCC for windows which is nearly twice as fast as the official release. Unfortunately that has it's own set of problems but with luck we'll either find out how to get the same speed into the official release or sorted out how to build Wesnoth (completely) correct with GCC.

[packaleks]

ah thanks guys for all the replies and advices, anyhow i already had the speedup turned on, so i dont have more options than try to compile myself and make things worse.. lol

[freim]

I'm not trying to insult anyone, but this performance is unacceptable, in the sense that I prefer to use 1.02. In my opinion this is a VERY SERIOUS ISSUE which the developers really need to focus on, to the exclusion of other things actually. If you can't get the speed up, you are in my opinion making things worse not better. I'm sorry to say so, because I know you are working hard. But Wesnoth wouldn't be the first game to get progressively worse with additional releases.

There seem to be a severe misconception on these forums regarding the development branch. The dev branch is under heavy development, large changes are done to the codebase which can lead to bugs, temporary regressions in functinality, visuals and/or performance.

The way this works between dev releases is roughly:
1. New functionality gets implemented
2. Bugs and regressions are found and fixed
3. Dev Release
4. Bugs are reported in and fixed if valid
5. Goto 1

Before a stable release there will be a (seperate) freeze period on new features, content and translation string. This period is used for an extended step 2. Toward the end of this period is where optimization work will be most likely to be done, since premature optimization is not a good practice.

So the bottom line is: you should expect bugs and regressions in the development releases, their sole purpose from a developers perspective are for testing. Lift your eyes and look towards the actual goal, the next stable release.

PS: Making "demands" on what the devs should do usually don't go down that well, so unless you plan to start paying us I suggest trying a different approach. I've still to see someone write anything starting with "I'm not trying to insult anyone" on the web and not actually doing exactly that.

[Prometheus]

Premature optimization is when code is optimized without knowing whether it is a performance bottleneck or not. Obviously there are performance issues in 1.1.4, so premature optimization is not the issue.

When performance issues appear, they should be dealt with immediately, otherwise you will have to go back and rewrite who knows how much code, which might prove effectively impossible.

[Jew unit]

to those who say the problem doesnt exist or that its related to accerated speed... your wrong running on my xp+ 3000 with a gig of ram with or withour accelerated speed on it gets laggy... HttT past the first few level (aside from the lack of the elf dude and dalfador) are slow as hell... and autosaving takes forever...