Multimangos

[KingofBananas]

I personally like the idea of Multiprocessing.

Wyk3d's objections are good, but I think that it always will be a bad time to implement something that big. You will always have big changes going on if you like it or not and that multiprocessing is needed in the future is also a fact.

But it's the devs who are implementing such things so I'll stick with what they are going to do.

[rilex]

Why would multithreading be "big" but not multiprocessing/clustering support?

[Mynt]

Why not go with Simultaneous Multithreads?

Or while your at it since your going to build the frame work. Your improving an already done thread task(s). Start using preemptive threading.

Preemptive loads may take a toll on the performance, but it would also increase render solutions and load support. Right?

-Mynt

[Derex]

Sad, from all talking here, I realy dont see a person who has enough knowlege to join and develop the multi-project with me. Doing it alone in the jungle is kinda wierd and If at some point of time I dissapear there will be nobody to maintain the code, or have any idea of what it does and how it does it.

Cypherjb seems smart, but he probably has another work.

Balrock, if multithreading gets inplemented, you can just forget about multiprocessing, I dont say it wont be possible to do multiprocessing from multithreaded version ( because everything is possible ). BUT doing it will require rewrite of the already written thing, ITS IMPOSSIBLE to reuse the multithreaded code and make multiprocess. ( just forget that any dev will write a bunch of code and then just delete it and write another bunch of code ).

[Cypherjb]

Lots of interesting replies posted since my last one.

First off.

Someone brought up an advantage I neglected to mention. On x86 a single process cannot address more than its pointer size (ignoring AWE which I assume requires modifications to the software), by splitting mangos into several processes you can work around this limitation.

Whether or not this issue can be addressed by improving high-memory features is another issue entirely.

Secondly.

To Wyk3d, if its not an 'if' but a 'when' why not address it now? By doing multi-threading only to have to rewrite it a second time to implement muti-processes you're doing a lot of double handling. I say implement multi-processes now and then its not an issue.

Third.

Derex, unfortunately I am quite busy at the moment, I just finished high school and am starting Uni in 4 months, doing freelance software development at the moment to get some cash (WoW related stuff, writing bots, hacks, etc for private buyers). I may possibly have time in the future but I wouldn't put my eggs into that basket.

[Mynt]

Sad, from all talking here, I realy dont see a person who has enough knowlege to join and develop the multi-project with me. Doing it alone in the jungle is kinda wierd and If at some point of time I dissapear there will be nobody to maintain the code, or have any idea of what it does and how it does it.

Cypherjb seems smart, but he probably has another work.

Balrock, if multithreading gets inplemented, you can just forget about multiprocessing, I dont say it wont be possible to do multiprocessing from multithreaded version ( because everything is possible ). BUT doing it will require rewrite of the already written thing, ITS IMPOSSIBLE to reuse the multithreaded code and make multiprocess. ( just forget that any dev will write a bunch of code and then just delete it and write another bunch of code ).

That was a pretty hard hit to intellegance levels right there. First I would like to personal put out there that when you were talking about multithreading vs multiprocessing, you never stated what you planned on implementing.

Are you going to rewrite the entire core on how processes are going to be transcieved, handled and or moved around internally? Are you working on making the process for outputs or input reading faster? Do you plan on rewriting the entire server? Are you working on character synchronization with maps/vmaps, collision or anything else relating to where the x/y/z axis input controllers are. Are you working on making SQL queries be non redundant. Making them seemlessy intergrate internally making less commands and less repitition.

What I am trying to get at is, what is your plan of attack. Its easy to say, which is better? Which do you prefer working on. You cannot judge someones intellegance or capability of knowledge by asking them which one they thing would be better to intergrate.

Are you planning on making a Multicore support while your at it, make it HT, or ST capable if applicable? Are you going to add quad core support, maybe mutilevel processing with buffer support using pipeline, or streamline data instead of a solid query every strike to the database. What is your plan?

You don't have one. I cannot comment and give you anything that would tell you.. x person knows what hes talking about if you don't give an A solution to your B answer.

I don't mean to come off rude, but honestly if I ask you which is better.. the Pri console debugger or Ollydebug. People are going to have there prefrences, suggestions and ideas of implementation. Now if I said, which one is better for Mangos, since we are working on making this a gaming framework. Obviously, the choice would be OllyDebug, since Pri is for single streamed data chunks. Where as Ollydebug can use data chunks. But then again, I can say your programming knowledge isn't good enough, or your not up to where I am because you have never used Pri. *Mainly because Pri is a program only used internally in game design*.

So, whats your plan?

What are you planning on improving?

What is your philosophy?

Why would changing x to y improve z?

Ask and answer/purpose what you want to do, and why you want to do it. Maybe then you might get what your looking for.

-Mynt

[Cypherjb]

That was a pretty hard hit to intellegance levels right there. First I would like to personal put out there that when you were talking about multithreading vs multiprocessing, you never stated what you planned on implementing.

Are you going to rewrite the entire core on how processes are going to be transcieved, handled and or moved around internally? Are you working on making the process for outputs or input reading faster? Do you plan on rewriting the entire server? Are you working on character synchronization with maps/vmaps, collision or anything else relating to where the x/y/z axis input controllers are. Are you working on making SQL queries be non redundant. Making them seemlessy intergrate internally making less commands and less repitition.

What I am trying to get at is, what is your plan of attack. Its easy to say, which is better? Which do you prefer working on. You cannot judge someones intellegance or capability of knowledge by asking them which one they thing would be better to intergrate.

Are you planning on making a Multicore support while your at it, make it HT, or ST capable if applicable? Are you going to add quad core support, maybe mutilevel processing with buffer support using pipeline, or streamline data instead of a solid query every strike to the database. What is your plan?

You don't have one. I cannot comment and give you anything that would tell you.. x person knows what hes talking about if you don't give an A solution to your B answer.

I don't mean to come off rude, but honestly if I ask you which is better.. the Pri console debugger or Ollydebug. People are going to have there prefrences, suggestions and ideas of implementation. Now if I said, which one is better for Mangos, since we are working on making this a gaming framework. Obviously, the choice would be OllyDebug, since Pri is for single streamed data chunks. Where as Ollydebug can use data chunks. But then again, I can say your programming knowledge isn't good enough, or your not up to where I am because you have never used Pri. *Mainly because Pri is a program only used internally in game design*.

So, whats your plan?

What are you planning on improving?

What is your philosophy?

Why would changing x to y improve z?

Ask and answer/purpose what you want to do, and why you want to do it. Maybe then you might get what your looking for.

-Mynt

You spelled "intelligence" wrong.

For multi-process (which I assume is what is the decision currently) pretty much all parts of the app will need modifications. I think the main idea is to improve stability, speed, and scalability.

Also, the Visual Studio debugger would be better for the Windows platform due to its focus as a 'development' debugger rather than a 'reversing' debugger (ie Its much easier to use symbols and source files in VS than it is in OllyDbg). I know it was just a metaphor but I thought it was worth pointing out.

[rilex]

Sad, from all talking here, I realy dont see a person who has enough knowlege to join and develop the multi-project with me. Doing it alone in the jungle is kinda wierd and If at some point of time I dissapear there will be nobody to maintain the code, or have any idea of what it does and how it does it.

If your plan was just going to be 'insult everyone', why did you bother asking the community what they would like to see?

[jaggex]

If your plan was just going to be 'insult everyone', why did you bother asking the community what they would like to see?

That doesn't necessarily have to be an insult. I think he just talks about the negative sides of letting one person develop this by himself. That's why it's kind of "sad" that no one else is able to offer their support.

Cheers

[rilex]

At least with a normal American infliction, the way it is phrased would be insulting. I suppose we do have to account for the non-native English speakers

[Derex]

I started this thread, because I wanted to put some discussion here, and eventually see somebody who is smart enough to join the "game".

Note that smart is different than educated for me, even if somebody is educated and knows about multithreading, networking or whatever, if he is not smart enough to use the knowlege he has ... he is nobody. So I prefer smart person with small or no expiriance in this area of coding, than some dude who has a bunch of computer books in his head and doesnt know what to do when it comes to doing real thing. Ofc knowing C++ is requirement .

For me its not that of a big deal what I will code ... I may even not code anything, I dont want to start cowboy coding for such big thing.

Its all about fun after all

[rilex]

Ah okay, it wasn't clear from your original post that you wanted help in something. In my opinion, you should just have said "hey, I want to code a multi-process core, anyone want to help?". It would have lead to less confusion and more appropriate responses to the outcome you're looking for (help on this project).

[bugs]

can't be included the part of "crashing on part of the server" to multiprocessing ?

i'd go to multiprocessing ... but can be done many bad things by rerighting the code ...

[Derex]

Ah okay, it wasn't clear from your original post that you wanted help in something. In my opinion, you should just have said "hey, I want to code a multi-process core, anyone want to help?". It would have lead to less confusion and more appropriate responses to the outcome you're looking for (help on this project).

Help is not the correct word.

[orgrim]

Hi, you may have seen me around the forums a few times, ive been using MaNGOS for a while and have always looked for something that i could attempt to work on, i found this idea very interesting and i am a college student and know alot of languages, but basic C++ knowledge. id be willing to learn anything necessary and hope i could be of some use if you do decide to start doing multi processing

I am a NIX user and would prefer to do code in something linuxy, although if needed i could install VS in wine

i will keep up with thread for any further comments / posts

you can PM on AIM via the username metamantso, msn via [email protected] or yahoo via metamantso

my email is [email protected]

again if you think i could be of some help on this just give me a message

[alexluana]

ok...so if u choose for example multithreat?

1.how will be voicechat or croosrealm things , be done after?

2.Will be able to have a tolerance for node crashed?

3.If u do multithreat? is there a way to add multioprocessing needed part to it?...

4.Isn't there anyways implementing both?

[jorooo]

3.If u do multithreat? is there a way to add multioprocessing needed part to it?...

From 1st post:

* It wont be possible to make multiprocessing version out from the multithreaded version. ( who thinks its possible - go to hell ).

[orgrim]

any news from derex on if this is going to be attempted anytime? wondering if i should go scouring the web for anything on multi processing

[Derex]

Feel free to find me in irc.

And coding it on unix will be much easier than windows. So its good to use unix.

[orgrim]

ok, ill be on IRC named orgrimdoom

[jolan]

about voicechat : it's supposed to be a separate server software, not related to the core and perhaps not on the same server (the core and the voicechat server must communicate though) so multithread / multiprocess of the core wouldn't help for this.

[AandD]

I vote multi processing because on windows64 you can use only 4gb per process:), but only multi threading or multi processing why not to start code that will implement multi threading and multi processing .

[rilex]

I vote multi processing because on windows64 you can use only 4gb per process:)

No, that is not true.

[AandD]

No, that is not true.

can you tell me how much it supports per process?

[higi]

after changes in objectaccesor have problem to build mtmaps, anyone have working patch?)