ATI slow performance in 0.94.x versions

[Floris]

And 94 IS FASTER.

uhhhmmm.....
I'm not sure under what circumstances that is but it sure doesnt feel that way.

Now 8v8 DSD games regularly get 0.66 gamespeed when having 1500+ units already. (using ASIM GML OMP) And this is giving me like 8-15 fps or so which is very low considering 0.66 gamespeed. 2000+ units games get nearly unplayable (yes you can still manage stuff but much more carelessly).

My mobile quadcore 1.6ghz i7 (q720) with hd4670 did better with older spring versions.

[zerver]

No, it is not faster, and finally even jK has realized that:
https://github.com/spring/spring/commit ... ba93a6a087

But when you order them to patrol an area so they collide with each other, fps drops from 140 to like 10.

Other devs will kill me for so called "propaganda", but Spring MT is incredibly much faster for the situation you outlined. Did you try it?

[jK]

No, it is not faster, and finally even jK has realized that:
https://github.com/spring/spring/commit ... ba93a6a087

But when you order them to patrol an area so they collide with each other, fps drops from 140 to like 10.

Other devs will kill me for so called "propaganda", but Spring MT is incredibly much faster for the situation you outlined. Did you try it?

Another propaganda post.
PLEASE STOP THIS SHIT.

[abma]

Other devs will kill me for so called "propaganda", but Spring MT is incredibly much faster for the situation you outlined. Did you try it?

wow, you really want a conflict. makes me sad, now i have removed your commit rights / rights on the buildbot.

[PicassoCT]

I know it annoys the crap out of you, but wouldnt it be enough, if the buildbot added a warning to every mt installer? No need to chop a working mans head off..

[zerver]

if the buildbot added a warning

Thanks for the support Picasso, but warning for what?

Warning: Higher performance may lead to more enjoyable games!

[Cheesecan]

If only there were charts and benchmarks..

[zerver]

Do you have a quad core?

Poor man's benchmark:

• "/cheat"
• "/give all" four times.
• Give a move order to 25% of the units.
• Look at the CPU load in player list (/info)

25% is my estimate for the percentage of units on the move in a typical game (towards end of game, when it really matters).

Using this method you will probably find out that Spring MT is at least twice as fast as Spring. In a real game there is also additional action (projectiles, map deformation and such) that benefit a lot from MT.

I'm planning to make some benchmark "demos", but only by comparing many real games (game duration vs total sim cpu time) do you get the whole picture.

[Jools]

[*]Give a move order to 25% of the units.

Standardised benchmarks are nice. However, as we wrote earlier, it matters a lot how the units move, and especially for xta it was the collissions that slowed down the game.

[zerver]

for xta it was the collissions that slowed down the game

Good, because the collisions are multithreaded as well, and even scale very nicely as you add additional CPUs.

[Cheesecan]

For starters numbers from a single 1h BA FFA game with 14+ players would be interesting enough to many users..

[zerver]

https://github.com/spring/spring/commit ... 08f8e394de

... and that's what she said!

[albator]

94.x might be faster, but (94.11 + BA )in ffa is definitely slower (except MT which was ... tut tut tut, RIP alba)