I made the multithreaded version of ZK mostly as a proof-of-concept, and although there is too much Lua in ZK to get kickass performance, it still performs considerably better than the regular version - about twice as fast simulation and much higher rendering FPS. http://www.file-upload.net/download-8180008/zk_1196_mt.sdz.htmlAs a BA player, I would rate ZK MT as "almost playable" (performance wise). Testers are welcome, ba.chezmorgen.fr:8200 or rtsmp dot com:8200. Because this ZK version is not endorsed by the ZK dev team, it will not be hosted on the spring lobby server. Thanks.
+6 / -1
|
Have you got some sort of changelog to make merging easier if this turns out to be worthwhile?
+0 / -0
|
Because Zero-K MT will desync with spring-multithreaded, as correctly pointed out already by many ZK players who tried to run MT in the past, I have locked ZK MT so it can only run on my latest MT build (RTSMP 94-20131015). This effectively stops all bug reports related to spring-multithreaded. However, I can provide a changelog and/or updated files that run with Spring also, if you would like to merge it in at some point.
+1 / -0
|
I don't understand so what needs to be done? Can you commit your changes to ZK repo to separate branch? Regarding engine will/are your changes going to be merged with the spring?
+0 / -0
|
This looks very interesting. Would you upload the spring MT build it runs on or provide a download link?
+0 / -0
|
@Zerver: you are still working on a community split? rtsmp still has no source code -> gpl violation! As long as the "issues" of the rtsmp-code are not fixed, there is no interest of merging it with spring.
+0 / -0
|
Licho you should read Spring forums for some background on this.
+0 / -0
|
Basically it seems that that the game files and mt-engine need a lot of babying to work together properly. So just about every new release of zero-k would need a new release of the engine, updated by hand by someone who knows what he is doing.
+0 / -0
|
Best of luck to you, but as long as you refuse to release source to your fork of Spring I suspect you will mainly be alone in keeping your modification of ZK up to date. There are many people who would love to see a more effective version of multithreading with ZK/Spring, but keeping a closed source violates GPL (as has been discussed many times) and really ties the hands of developers here. I will try it out though and investigate performance
+1 / -0
|
ZKL support multiple versions of engine, so there could be a auto-host/room with that engine and ZK MT fork? "Because this ZK version is not endorsed by the ZK dev team, it will not be hosted on the spring lobby server." We have a auto-host/room for test like "BLEEDING EDGE of spring engine" Iridium2; "TEST: latest game version" Oxygen7; the "fork" "Zero-K DOTA" Rhodium1, so we can have a TEST MT room, Licho and GoogleFrog seem interested in it...
+0 / -0
|
@[DOOP]fortaleza Everyone is interested in having a working multi-threaded engine. The problem is that it only works when one particular person tweaks a black box enough for it to work for one particular release of zero-k.
+2 / -0
|
The Spring engine is already very brittle without Zerver's MT. "Whoops, innocuous change broke everything in a completely counter-intuitive way and now we have crazy sync/pathing/performance bugs" is a near-constant event on every Spring game. This would be even more so?
+4 / -0
|
It is brittle but not primarily because game development is exposing new bugs. There is more of an issue with devs not working on the same level and making conflicting changes that result in bugs. I made this crappy vid that gives a hint of how much faster MT is, with BA. Like already said, due to the singlethreadedness of Lua, you can't expect such kickass speed with ZK, but still a nice boost. First half of the vid is Spring. Game speed and FPS indicators are visible in the upper right corner.
+3 / -0
|
Its too fast. Are you really honest with that video or is it a fake? 0_0
+0 / -0
|
quote:
Its too fast. Are you really honest with that video or is it a fake? 0_0 |
Just test it, huh.
+0 / -0
|
Comparing vs 91.0 for Zero-K was harder because the demos are incompatible.
+2 / -0
|
Are there any big bugs/desyncs? If not, i see no reason not to use MT as default engine. Could Licho set-up ZK MT room like it is with bleeding edge?
+0 / -0
|
quote:
If not, i see no reason not to use MT as default engine. |
As far as i know RTSMP still has no source available. Please correct me if this isn't so. If it is, a violation of GPL is quite a hardcore reason to not use it as default engine for an open-source game. Additionally, that would essentially be making zk a hostage to zerver's cause, whatever it is. I'm not up to date on the causes and latest developements in the conflict between zerver and other spring devs, but getting dragged into that would almost certainly be a bad thing for zk.
+4 / -0
|
AFAIK, the GPL issue wasn't the main engine distro, but the modifications he made to SpringLobby. The reason his SL modifications weren't opened is that they were done on the binary with a hex editor (which is spectacularly bad practice, but whatever) and therefor there is no "source". But I could be misunderstanding the details of this absurd pile of drama.
+0 / -0
|
Well i couldn't find sources for latest release of rtsmp on rtsmp.com. All there is are two links to github, all with last commit being 6 months ago by Kloot. Latest rtsmp release zerver put up was 8 days ago as described on rtsmp.com, so it's not the same.
+0 / -0
|
