| |
| |
| |
|
Page: 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Comments:
<0> If you really don't understand this then you're missing something fundamental about code optimization
<1> if you think it is impossible for a human to write something a machine can do, then i think you are being silly
<1> doesnt matter how long it takes, could take infinity
<1> or close to infinity rather
<1> wouldn't matter to my argument
<0> yes, it would
<1> you are trying to bring in "time" to the equation and act like it matters over all else
<1> <1> a c compiler will never beat hand optimized ***embly, ever
<0> sigh. if you don't understand why then you're missing something really really important
<1> that's true
<0> no, its just not
<1> it never will, ever, as long as there is no timeline
<0> sure. ***ume we have processor X for 100,000,000,000,000,000,000 million years (which is how long we're talking about quite often, in the minimum)
<0> then it really doesn't matter what we do, because if we have that long a few ASM tweaks aren't going to save us *anything*
<0> we might've just ignored trying to optimize and just run it
<1> yes see the problem is you're trying to win another argument
<0> no
<0> the problem is you're trying to back out of yours
<1> what i said is true
<2> :)
<0> nope :D
<1> what you are saying only makes what i said not true IF you append "in x amount of time"
<0> Humanity would disappear before we ever knew if we had an optimal solution
<1> which i didnt, hence its true
<1> who cares
<0> um wow. What an awesome argument you have then.
<0> "Imagine I have this fairy dust that makes all relevant problems go away. then I can solve the trivial ones."
<0> Thats simply not valid
<1> you're trying to theorize humans die before it will be solved, when you don't know that
<0> You clearly don't understand NP completeness
<0> at all.
<1> it ****s being wrong doesn't it
<0> ask yourself :D
<1> NP-complete revolves around time
<1> you are bringing TIME into this, not me
<3> its time for kozmonaut
<0> Ahaha yes, I believe it is
<0> Go kozmonaut go!
<3> so whats up all
<0> Not much. Trying to explain to polestar why compiler generated code will outperform human generated code, but he doesn't understand enough of the basic concepts, so I've given up :D
<1> sure nick
<3> man can you make a machine code for me and my games??
<1> it's funny what you are talking about these NP-complete problems when who codes to solve these?
<1> 2 people in the world? if any
<0> Pole - so ... clearly you have *no* idea what you're talking about
<3> lawlz
<2> sounds like it
<4> sounds like it is ****ing you pretty hard too.
<3> haha
<0> hahahaha
<3> pole is ****iny froz!
<3> the motherland will take over teh poles!
<1> so dr^nick, is it your belief np-complete problems are unsolvable?
<0> ahahahaha
<0> again, you don't understand them at all if you're asking that question
<1> im asking what you believe
<0> Give me a specific instance of an np-complete problem and it might have a trivial solution
<2> meh who optimizes :D
<0> ie. Longest path is NP-complete. On a graph with 2 vertices and one edge connecting them, its so easy its stupid
<0> but thats such a small problem it's insignificant
<1> the theory does not suggest any are unsolvable though do they
<0> you don't understand the theory D:
<1> ok so since you do, answer the question
<0> the question is nonsense.
<0> of course a solution exists for np-complete problems
<0> and an optimal one too
<0> but thats entirely not the point *at all*
<1> so all you are suggesting is that it would take "too long" for a human to complete say the hardest one
<2> kozmonaut: anything new?
<0> im saying that if you doubled the age of the universe you wouldn't have even scratched the surface of the amount of time it would take to solve a "hard one"
<1> ...right so tell me where I was wrong in regards to the "without a timeline"
<3> hfrm
<1> see the fundamental thing you are missing here is that the problem is not unsolvable for a human
<1> it HAS to be for you to be right
<2> kozmonaut: hfrm?
<0> your statement was simply <1> a c compiler will never beat hand optimized ***embly, ever
<0> Which is ludicrous
<0> I could write a hand optimized ***embly program right now
<0> but not do as good a job as my c compiler
<0> and you would loose
<0> because it would only have to happen once.
<3> its all about anthropic computing
<1> well again its hard to work out all the rules for a statement when its in one sentence
<1> they call that context
<2> meh I say to heck with optimizing past the obvious - if you need more, buy a dual core :D
<0> basically your whole argument is nonsense and its time to give up
<2> ooh that rhymes
<1> hehe i like it when people act like they are so right, when they are wrong
<0> So, stop.
<3> pwnt
<0> oh, you like it. okay, Gotcha.
<1> either way for these "small" problems that we encounter everyday as programmers
<0> ahahah
<1> an OS would be faster in pure ***embly
<0> sigh
<1> at least with all compilers that I have seen
<1> I haven't seen them all
<0> Keep believing that :D
<1> you said you have seen one or two
<2> How much hand optimizing do you do anyways?
<1> but don't mention the names so i cant check
<0> ICC would do
<0> if you want to do x86
<1> ok let me go check that for you
<0> i can't say for *certain* because I've never used it
<1> ICC is faster than hand coded ***embly?
<0> it certainly can be
<1> lol
<1> gcc can be
<0> right, and all it takes is one instance where you're wrong - where someone didn't optimize 100% and you immediately are wrong on your entire argument
<0> because again "<1> a c compiler will never beat hand optimized ***embly, ever"
<0> if it ever *ever* happened, even once, you're wrong.
<2> Dr^Nick: It'd be just like me to prove him wrong too - I **** at optimizing :)
<0> hehehe :D
<0> So, there exists a point where the number of registers be greater than the entire humans memory capacity
<1> well nick if someone like yourself who admits you aren't that good at asm
<0> at which point humans wouldn't have a *Chance*
<1> coded an OS, im sure C might be faster
<0> PolestaR - did I say that I wasnt? :D just that I can write bad asm :D
<0> you need to start thinking critically
<1> when i mention hand-coded asm im talking about the most optimal way to code something
<0> and again you don't understand np-completeness
<1> you can do and learn the most optimal ways
<1> im talking about x86 here, not your future systems ok
<0> you *do* realize that there are situations where the "most optimal way" might require more storage while doing it than available in a humans head
<0> no, again. <1> a c compiler will never beat hand optimized ***embly, ever
<0> I'm going to hold you to that statement.
<1> i thought you said we are moving on from that
<1> you were sort of implying i had to stop or you would ban me
<1> but you keep bringing it up
<0> you said you want to "remove time" so its totally irrelevant.
<0> so if its irrelevant, we can talk about *any* architecture, past present or future
<0> your argument should hold
<1> it does
<0> well, apparently you weren't done being humiliated and I wanted to give you a break because I was feeling bad.
<0> if it does then <0> you *do* realize that there are situations where the "most optimal way" might require more storage while doing it than available in a humans head
<0> so again, you're just wrong.
<1> since you can't quantify human storage
<1> you're guessing
<0> Actually, no
<0> Having taken neuroscience in gradschool we can
<0> and its actually quite small
<1> so work on the problem in chunks
<1> make more programs which help you to solve the problem using computers
<0> and that should be possible why?
<2> Dr^Nick: How small? :D
<0> and if you're making more programs, then you're essentially using parts of a compiler to help you beat a compiler
<0> might as well just copy its output.
<1> so
Return to
#gamedev
or
Go to some related
logs:
#computers
#red
#politics
#beginner
#flash
#eggtcl
donkls
chamgagn
lcd-hacking
#windows
|
|