Multithreading

[Mariko]

Hi,

i just wanted to ask if there is a plan to re- implement a multi threaded version of spring. We had one a few years ago.. and another one is out there .. giving a performance increase of 5x, which would finally allow all Players to finish the games without lagging out after 20mins and to play the game how its meant to be .. 8vs8 with at least 500 units each.

M.

Removed link to GPL violating material. (Silentwings)

[FLOZi]

Please post your benchmarks.

[jamerlan]

Just test spring 99 release. Maybe it will work fast enough for you

[gajop]

Please post your benchmarks.

This. 5x faster? Definitely not. I concede that it was fast and faster, but it's certainly not going to achieve a 5x speed increase normally.
Also the much more important issues of being GPL infringing, "bad design" (I can't confirm this but that was the consensus among other devs), and not really working well with Lua games AFAIK (and Lua is the future)

[albator]

Tbh, MT was not x5 in average but in some specific case it gave even better enhancement: when catching up a game the simulation was fluid (>30fps) while now I am around 5 fps in the same situation.

[Mariko]

oh... okok now i get why no on is talking about it :) and ok the 5 times perfomance increase is just something i read on the page, i couldn't try it out.. but i remember the MT version being faster like albator said.

what's new in spring 99? I couldn't find anything performance related in the changelog..

and back to the original question .. are you working on a multithreaded version of the simulation at the moment?

[hokomoko]

are you working on a multithreaded version of the simulation at the moment?

I really like these questions, they remind me of the old issue of bosses who tell you both what to do and how to do it.

If your issue is that Spring runs too slow on you machine, lately every release candidate of 99.0 is faster than the previous one, you're more than welcome to test them and give us feedback on your issues if there are any.

Whether the simulation is powered by multiple threads, by a single thread or by 12 grandmas on bicycles is irrelevant.

[Mariko]

I'm neither telling you what to do, nor how to do it, nor I criticised your way of doing thing .. or have I? (if I have i'm sorry)

i realize now that my question might have brought up old issues .. but in the end it was just a simple question. Sure you don't have to answer it but then why answer at all telling me about grandmas?

[hokomoko]

k, currently some of the simulation is multithreaded, some isn't.

Sorry if I was too hostile, I really want to know how the latest engine performs on your machine.

[Forboding Angel]

Sorry if I was too hostile, I really want to know how the latest engine performs on your machine.

You weren't. This discussion is idiotic. Zerver's MT build put itself out there like it was the only thing spring multithreaded, when spring was ALREADY multithreaded. Zerver's implementation was faster in some respects (probably because it mostly tossed lua out the window), but far inferior.

I for one any getting very effing sick and tired of seeing this pop up.

I suggest ripping metal and energy out of the engine and moving it to lua (this isn't a bad idea anyway). That will solve little idiotic threads like this.

[PicassoCT]

Calling a discussion idiotic is not civil.
None the less - both sides had there say. The thread is archived and can be found.
Patch is welcome when it comes to improving lua performance or threadpool implementation.

Its a shame that all of zervers hard work wents to waste, but sometimes thats the run of the world.
As one of the not so clever people i find it rather positive though, to be still able to read and understand the source.

Please do not feel discouraged from contribooting. Many Spring devs are mature, serious people and will happily discuss the pros and cons of the mt engine version with you.

[smoth]

Eh, you fell victim to marketing, unless you have some real proof you are just going by hearsay.