Bay 12 Games Forum

No, I don't think your seeing what I said there. Basically thats the individual tax, not corporate.

And a sole proprietorship, grossing two hundred large a year? In the States at least, what isn't paid to employees is taxed as the owner's personal income.

Do you want to encourage him not to hire more workers by taking a larger chunk of his money? Perhaps just substituting his own labor in place of an employee's--working harder to keep the same take-home pay, while not employing someone else? I would hope not. There is a delicate balance to be struck, and erring on the side of "oh, they're rich, they can afford it" sounds rather like class envy. And as someone who very much aspires to be only marginally poorer than Bill Gates someday, that upsets me.

(That's another interesting point--higher taxes tend to drive down charitable donations. Interesting, no?)

Quote

There is no "CPU limit" on it.

There is a CPU limit. Basically it is what stops a single program from multiplying its processess until it eats up your computer.

The CPU Limit Remover is done at the same time as forcing it to a Single Core. (In fact it is the whole point to doing so)

This is also counterfactual. A program can spawn processes as often as it likes, so long as it doesn't hit a limit designed to stop processes from rabbiting and making a fork bomb.

The key thing that makes you very hard to take seriously is that we're not talking about processes, but threads. All affining a single-threaded program (or, since virtually every Windows application has at least 2-3 threads, a few-threaded program) to a single core ensures that you don't suffer the overhead of cache misses, register copyovers, etc. inherent from moving between cores (and avoids other stuff like busy-waiting on in-flight threads on whatever core the process is being dumped to). There is no magical "limit," there are just the algorithmic tendencies of the scheduler.

EDIT: damn yooooooooooou, bombcaaaaaaaaaaaaaaaaaaaaaaaaaar!

Quote from: MiamiBryce on December 09, 2009, 10:50:47 pm

Quote from: The Architect on December 09, 2009, 07:42:30 pm

From what I understand, right now DF's current format only enables it to run on a single processor, failing to utilize the power of the dual- or quad-core processors most of our computers currently have. A good emulator (something along the lines of DosBOX) could unlock DF's potential.

Your understanding is flawed and I don't think I saw it corrected in any of the posts, though I admit there was a lot of tl;dr because I tire of these same old arguments easily.  Any modern operating system will distribute the workload of DF among all cores of your processor.  I have a quad core processor.  When I run DF, all four of my cores spike to somewhere between 25 and 30 percent usage.

I think you're misunderstanding something.

Your computer is not running DF on all four cores at once. It's bouncing it AROUND the cores, so that 1/4 of the time it's on one core, 1/4 of the time it's on another, etc.

Big, big difference.

This would be why affining it to a single CPU often sees (sometimes significant) perf improvements.

Ah geez man you really, really need to chill out. No one is mad at you for shooting down ideas. Like I said: YOUR CONTRIBUTION IS CONSIDERABLE. But your tone is constantly offensive and you constantly go off-topic over your personal crusade.

Let's look at other ways to do it. I will present the idea to Toady of we have something useful, if we can't come up with anything then that's fine. Is it impossible to outsource processes much the way you outsource labor? When DF normally queries for closest resources every time the dwarves perform a job, causing massive processor delays, can't we have it query an updating separate database? We would basically be rewriting a small part of the program separately without the need to change the very thing you keep saying it is impossible to change.

Toady's cooperation is necessary. Rather than assuming we can't get that cooperation, let's just consider whether it is possible that we might be able to do something if we can. It's not going to kill us to consider the hypotheticals here in this thread.

For the last time, if you dont' want to play ball, then leave. This isn't the place for your crusade against your invisible enemies. If you want to continue to contribute without the sarcasm and off-topic defensiveness, then it would be nice to have your knowledge.

Damn you're stubborn about it.

You seem to be placing some kind of intent in my actions that is not there. I am pointing out things that are, to someone with experience and skills in this area, obvious. You are getting upset at this. That is something for you to work out, sir, not me. There is no crusade. There is me telling you what works, and me telling you what doesn't work, and you stuffing your fingers in your ears and going "LA LA LA BUT WHAT ABOUT IF WE HAD THIS STUFF THAT DOESN'T EXIST?". I am making a point of being about as polite as is possible here when I say that yeah, you could do it if you have all this stuff you don't actually have. But that doesn't make for any better discussion than "no, it cannot be done given the constraints of reality," now, does it?

Given cooperation from Toady on whatever necessary level, is it even possible to create an interface/utility that would do what I've suggested?

See what I mean? "If not for all the roadblocks in the real world, could something be done?"

You're asking the wrong question, because if not for all those roadblocks--if Toady was interested in cooperating to the degree necessary to do anything significant--it could be done just by modifying the source code. There! Done! End of problem! But in the real world, the source code is not available and as such modifying object code to parallelize application functionality (you can use whatever phrasing you want for it, it boils down to this) is a nontrivial task.

Quote from: bombcar on December 09, 2009, 10:25:07 pm

Basically, you're asking for the source, which Toady has repeatedly said he's not releasing. That's more the problem than any understanding of modern CPUs.

Though there is one way around that. If you can create an algorithm/equasion that can find Prime Numbers then you won't need the Source Code.

How does this relate? Why is prime-finding relevant to the discussion?

You don't get it, dude. What you're asking is like asking an architect how to make a thirty-mile bridge without any supports. Physics, whether on the bridge-building or the processor-architecture scale, does not work the way you think it does. You've been getting a "Whaaaaaaaaaaat?" reaction from people in-the-know (I'm not the only one in this conversation who's saying these things, I hope you've noticed) because it is not feasible.

If you have the money in your pocket to hire a dozen engineers from VMware to automatically parallelize object code, then by all means (just don't expect huge results)--but you don't, and this topic doesn't do anything. That's what I've been saying, and that's what other people have been saying, and that's what you don't want to hear. You keep saying you don't know anything about it, and people who do know about it are telling you it's not doable, and you get huffy and get mad! What the hell gives, man?

The way to solve this problem is to let Toady do it when he's ready. He has the code. It's his ballgame. Not yours, not mine. His.

I'm sorry that boundless optimism doesn't change physics and math. I really am. But getting pissed at me for pointing out the obvious is the wrong place to look.

Quote from: G-Flex on December 09, 2009, 09:22:51 pm

Asking a program to look at the compiled program (or even the source code) and do this for you relies on the same lack of understanding as asking someone "Why won't the computer do what I want?" - the computer never does what you want, only what you tell it to. Even the different parts of DF's code that could be run in parallel currently are not and will interface in ways that no program can magically figure out for you.

I think this is an oversimplification, if I'm remembering my Op Sys class correctly, which I'm not at all sure of.

Actually, he's pretty close. It is extremely difficult to figure out programmatically what code is serial (requires the results of previous code) and what code can be safely parallelized. A person can look at it and go "this, right here--this can be parallelized easily", but the heuristics to make a computer do so...not so much. There are techniques to do it, but (there be a pattern here!) they invariably require the source code.

(A pet interest of mine is compiler design, so I've been exposed to an unhealthy amount about this stuff. )

Quote from: The Architect on December 09, 2009, 09:02:46 pm

Splitting the processes DF does (if that is even necessary) might require access to code that has not previously been available to players, and we may need Toady's help. If that is so, then let's go there with it.

The point of this is to enable DF to utilize our computers' true potential without the need to rewrite the whole program. A program-specific emulator. Not a universal tool.

It seems obvious to me that I have already stated the possible need for access to the source code. You're right, I don't know a lot about what I am proposing. From what you say, access or at the very least close cooperation would be absolutely necessary. So be it, if this is in any way feasible with the source code then let's ask for it. If it's not possible either way then sure I'll forget it. But let's not dismiss it out of hand based on things that may or may not be the case (Toady/Threetoe never agreeing to such a thing).

I was actually looking for something to interpret the processes and feed DF information that it would normally need to manually cycle through. Not work the whole program on a multi-thread base. But most likely this would require non-existant output and non-existant input options. Outsourcing just the stocks database, causing DF to feed the info to a separate process and retrieve it from said process would cut the load tremendously, would it not?

DF can sequentially request info from the program rather than calculating it itself as it does now. The "emulator" would just be an interface. I do realize now that it might be the wrong term for such a program, but I dont' know a better one. The emulator would cause DF to believe xy was happening, when in fact z was performing the functions of xy separately. The same could be done with pathing, but that would be insanely problematic and likely to produce a headache just contemplating, but those are the two major drags besides job priority checks. Allowing us to set our own job priorities for dwarves and groups would probably cut lag in large forts by a ton as well. This problem could be circumvented by outsourcing the Jobs menu and feature to a separate program.

As I'm looking at it now, it definitely requires full cooperation from Toady. One thing it does not require is a rewrite of the program.

Really. Where are the credentials you're bringing to the table that say you know this? Do you tell your mechanic that no, he's wrong, just throw an extra set of wheels on it and it'll go faster? Because that's pretty much what you're doing here.

I've built monolithic single-threaded applications before and, uh, none of it's even remotely done this way, ever. The information-service architecture you're trying to suggest would almost certainly require even more restructuring than just manually parallelizing what's already there! Don't conflate this with stuff like the DF pathing library--that is being developed in a vacuum in hopes that Toady can later on integrate it himself (and they're doing fine work, I love reading that thread--I haven't posted there because algorithm grinding isn't my specialty, I don't have the head for it).

Quote

...No. Just...no

This isn't TVtropes, around here were support our ideas.

Really? You do? You, specifically, do? I couldn't tell from your half-redigested repetition of something that wasn't even relevant!

I'm not derailing this thread to indulge you and I won't reply to any silly attempts by you to rationalize what you said, but let's just go over this quickly. Vista "failed"--if an operating system that sold eight bajillion copies could be said to "fail"--because they put "ready for Vista" stickers on computers everyone knew weren't ready for it and, in doing so, poisoned the brand in the mind of Joe Public. It had nothing to do with diminishing performance returns from processor research. This is not the thread to debate this, so kindly don't. Just stop. Go make a "I think Vista failed because processors aren't getting faster no more!" thread if you really feel the need. Make sure you figure out why Win7 is a ridiculous success at the same time when it's pretty much Vista SP3, though.

And yes, I would say that Blacken's impatient tone of superiority is irritating. It would be nice to see some reasonable and well-thought, well-supported, well-written responses.

Then ask reasonable and well-thought, well-supported, well-written questions. You aren't dude. I can't be blamed for calling silliness what it is.

I asked for help and information, and got a smart-ass, "I'm tired of dealing with ignorant people" response. About half of what was said didn't even apply to my comments, but seemed more of a generic response to past irritations Blacken has experienced when dealing with irrational and ignorant people who demand the impossible.

With all due respect, I don't think you have the knowledge to determine what is relevant to your comments.

Look at everything that's been posted in this thread. I'm sorry that using blunt sentences isn't something you like, but everything I've said has been true. You are asking for what essentially amounts to the impossible. It sucks, everybody gets egg on their face once in a while--I'm sorry that your in-theory-great idea is practically not possible, but it isn't.

If we want reasonable and well-thought, well-supported, well-written responses: you start. Find some sources that suggest what you're talking about is even possible. Do some research before insisting that I'm wrong--you're asserting a positive, the burden of proof is on you.

I'll wait.

G-Flex is nicer than I am.