Jukka Larja (roskakori2006@gmail.com) on 10/5/09 wrote:
---------------------------
>AM (myname4rwt@jee.male) on 10/4/09 wrote:
>---------------------------
>
>> After three paragraphs beginning with
>> your insightful conclusion as to who
>> didn't get any of the discussion, the
>> only thing I see you asking is why
>> GPUs are in order.
>
>I thought you were talking about this (from my post):

It was several posts later. If you come to conclusion that someone misunderstands something, then why not start with clarifications, considering that you preach how easy it is to misunderstand?

>>>> 1) You admit that on some workloads an
>>>> out-of-order execution is a waste.
>>>> 2) You say out-of-order doesn't affect
>>>> clock speed.
>>>> 3) 2) can only be true if OoO doesn't
>>>> increase power consumption
>>>> 4) So if 2) and 3) are true how can 1)
>>>> be? Where is the waste?
>>>>
>>>> So, my thought was that you were just
>>>> sloppy and I shouldn't take everything
>>>> you say literally. That I should add
>>>> things like "in practise", "for
>>>> general purpose computing" and so to
>>>> your statements. But that doesn't seem
>>>> to work either, and I still don't
>>>> really understand why.
>
>Which you answered with: "Very interesting, why do you think so? Is that because
>some "theory" states that?" although I was stating what I thought you were saying.

And I believe I pointed out to you that this inference was broken. If you really wanted to clarify something, all you had to do is point out the posts you were having a problem with and ask.

>> Does "hypocrisy" have a similar
>> notion in Finnish culture?
>
>If you are so sure I'm just doing this all on purpose, why are you even replying?
>Really, if you think I'm making things up for arguments sake, then _I_ have no reason to continue and will stop.

Actually, many things you write make me seriously wonder what your real purpose is. You say how easy it's to misunderstand, yet you write in a very-easy-to-misinterpret style; you accuse me of "adding" something to your statements, yet when I bold out what you said (below), you are full of surprize etc. It just doesn't look like you're trying to resolve a misunderstanding or arrive at a meaningful answer to a question that bothers you.

>>>> Just trying to tell you that clock
>>>> rate is by far not the only way to
>>>> bring down power consumption --
>>>> both in practice and theory. (You
>>>> seem to be somehow fixed in this
>>>> understanding).
>>>
>>> Well, I never meant to say
>>> anything like that. It seems you
>>> are inserting onlys, musts and
>>> bounds to my sentences although
>>> for me it is very important that
>>> they stay away.
>>
>>>> Well, if you aren't allowed to have greater
>>>> power, and the clockspeed is the only
>>>> variable you can adjust, I don't see what
>>>> else to do.
>>
>>Did you write this? Or did I just make it up?
>
>I don't believe this. I thought we already had an understanding here. If for theory's/hypothesis'/whatever's
>sake I state that clockspeed is the only variable you can adjust, it doesn't mean
>that it is true in all theoretical or practical discussion.
>
>Really, I thought you already understood what my "everything else being equal" example was about.

I was replying to your accusation of me "inserting" onlys etc. which is apparently not true. So please don't try to find something else here.

>> If a freshman asked me something like
>> that, I would probably try to understand
>> the background of the question in the
>> first place, for it could [potentially]
>> be only loosely related to the
>> "well-worded" exercises you give them.
>
>What does this has to do with anything? Of course you try to understand the background
>of the question. You don't need to involve freshmen and universities for that statement to be true.

It has to do with the way you reply. I tell you about the holes in your arguments, and instead of staying on track, you start rambling about freshmen.

Besides, you keep deleting prior discussion (why?), and it's much easier to lose track this way.

>If you are trying to say that I didn't do enough explaining with my three statements, I can only stare in wonder.
>
>If you are trying to say that I didn't do enough explaining when I first brought
>them up, I just say that hindsight is 20-20.

Re. those, I'm still trying to figure the meaning you put in. OoOE is nothing special in that it takes extra space, and transistors occupying that space need some power to work; so what do you get if you ask these 3 questions about something else? If you honestly believe that your "big 3" is a meaningful mental exercise re. OoOE or a design feature in general, may I ask to what conclusions does it lead you? Is it supposed to show that the heavier the core, the lower it will clock, b/c any extra feature that consumes power will result in lower clock if power budget is to be kept the same? What exactly is it intended to show?

>>> It wasn't hard at all for me to
>>> choose power, clockspeed and OoO
>>> to demonstrate the issue of cost
>>> and trade-off.
>>
>> Why do you think it was necessary?
>
>Why did I wrote the three statements at all? I was trying to demonstrate that OoO
>has cost that on theoretical level can affect clockspeed.

And you don't want to consider for a minute that OoO designs tend to remove at least such well-known weak spot of in-order machines as global stall signals, which can hurt clock rate?

Apologies for repeating this to you, but by fixing power consumption you change the problem *very* significantly, and you shouldn't use a different setting as a means to "prove" that someone was wrong/misunderstanding etc. This is the problem with your argument.

>> OK, since you seem to remain convinced
>> that I keep missing your point, let me
>> ask for clarification: if you read the
>> thread and saw me readily admitting
>> OoOE has die and power costs, why
>> are these questions even relevant?
>
>Because when I tried to make basically the same statement, you said it is wrong.
>It is obvious now that you thought I was meaning something else.

Are you sure? What is the post you're talking about?

>> Power was *never* a point of contention
>> in the discussion (nor was die space).
>
>I'm pretty sure I have stated several times that I chose it to three statements.
>I know it wasn't a point in previous discussion (or if it was, only on very general
>level. I didn't just go reading through the discussion, so I don't know if it was mentioned in the passing).

Power was mentioned indeed, and not in passing; I said very clearly in the beginning that OoOE has power costs in order to concentrate on other arguments. Have you seen it?

>> You say "the power consumption has
>> limits". I agree it does, and so
>> does die space, and both of these
>> issues are way more important in
>> *practice* (it's actually theory
>> where one can choose ti design
>> anything w/o thinking much about
>> them).
>
>It's also in theory where one can set arbitary limits, like everything else being
>equal or not a nanowatt more, even though in practise you couldn't even measure that extra nanowatt.
>
>> It just goes to show how annoyingly
>> illogical your problem setting is,
>> even if looked at from "theoretical"
>> point of view as you call it. If you
>> want everything else to be the same,
>> including the same power budget, but
>> allow some die space, let me ask
>> you: how much power do you expect
>> these extra transistors to consume
>> everything else being equal?
>> Exactly zero? :)
>
>I have no idea. That's the next part of the lesson, where we learn how little theory has to do with practise :) .

Well, but I'm asking for clarification because it's *your* theory! Please, do some explanation here in case I fail to see the great academic significance of this exercise.

>> I seriously doubt that such approach
>> to judging value of a chip design
>> feature (not just OoOE) has some
>> theoretical value in it.
>
>It's pretty much standard way of teaching physics, for example. You first have
>some formulas that work on theoretical level (like with objects with no volume and
>such) and after we have learned about those, we learn where, if anywhere, in practise
>we can use them. Or we learn more theories based on those previous ones and gain wider practical solutions.
>
>Would I use those three statements if I had to teach processor design to freshmen?
>Certainly not. I would tell my superior that I have no qualifications to teach them about the subject at all.

That could well be suboptimal decision from the educational point of view. Think about it: many exercises "from the book" are simply uninteresting. Professor knows the answer from the instructor's guide, and students either have one or know there's one, or just look up equations and substitute figures (this is not true in general of course).

Now consider that you are indeed teaching CPU design, and offer these 3 questions to the class. To make things more interesting, let's say you don't make it a regular hw, but pose them for a workshop after your students have already learned what OoOE is and how it's implemented. If you have at least a bunch of fast-trackers (I mean this in a good sense), I'd bet this will do your students way more good than thought-out assignments you're proud of, no matter how well-worded they may be.

And if you are blessed with having an opportunity to teach people who are natural learners and aspiring designers, then just don't bore them with lectures or well-worded exercises. Give them good sources to work with and do your best to show the gaping abyss b/w good designs and bad ones -- I would strongly argue this is the only thing that a mentor should do, and do it well. Don't cultivate the gospel attitude to the "good" sources either; instead, try your hardest to develop a healthy critical look at whatever they come across; help them see all the pros and cons and have conflicting opinions on the matter; divide them in groups and make them argue and defend their opinions, if not for personal value, then at least for course points.

By following this way, you have a much better chance to grow thinkers and *personalities*; people who are not afraid to speak their mind, with a strong sense of pride for what they accomplish and able to admire good work done by others. Or failing all that, simply better engineers, who will recall their professor with a sense of gratitude when they are old enough.

>>> Unfortunately I don't myself own that H&P,
>>> so I can only make educated guesses
>>> about this particular 2 extra stages
>>> issue.
>>
>> But maybe you can simply ask him?
>
>Exactly why? So I can relay his answer to you? Why, in that case, would he answer to me? I don't know him.
>
>I'm not personally interested about what H&P states. If I was, I could go to university
>library and check. Or buy the book.

Well, I thought you were a fanboy of theory, so how about we come to some conclusion how his speaking from theoretical standpoint (acc. to you) is connected with real theory?

>> Since you deleted a part from one of
>> previous posts where I accentuated
>> that I asked Jouni at least once,
>> what gives you a moral right to say
>> such things? It would be somewhat
>> understandable if you apply it to
>> someone who doesn't care much about
>> clarify things (hint: Jouni and you).
>
>How can you say that I don't try to clarify things? I'm obviously not successful,
>but just look some way above. I'm trying, unsuccessfully. If you think I'm not honest
>with my efforts, then say so plainly. There's no way I'm going to convince you if
>that's the case, so I just stop trying.

To be honest, I don't know anymore. There's just too many too ambiguos things from you, very little substance.

>> That's why I prefer examples and
>> citations. Less chance to
>> misinterpret something.
>
>Yeah, just look how well I'm doing with my examples.

I wouldn't call that "examples". I addressed your 3 in more detail above.

>>> Judging from my experience, there's
>>> nothing simple in correcting and
>>> clarifying. Really, it wasn't
>>> important and you made it clear you
>>> don't think your using debating
>>> tactics. Clarification, or
>>> discussion in general, isn't zero
>>> cost.
>>
>> Don't you think a short clarification
>> can save much more time in total?
>
>It depends. If one correctly assumes that the issue isn't relevant (as Jouni seems
>to have done here), it is easier just to be quiet than start another discussion.
>
>I don't know where you are from and whether English is your native language (I
>don't recall you stating otherwise, so I would assume so), but just by looking at
>your peak posting rate, words are a lot cheaper to you than for me. There are days,
>when you post so much that I'm not sure if I could do the same even if I did nothing
>else the whole day. So we probably have quite different opinion on what counts as "short clarification" anyway.

If you look at my other posts, they are quite short, and other than posting to 2 or 3 topics the other day, it's been mostly this thread of late.

Short clarification is something like asking "Do I underst. correctly that ... ?" before joining a thread to say that one of the arguing parties doesn't get anything.

>>>> What do you find to be inappropriate(?)
>>>> about this style?
>>>
>>> As I stated, I don't think you are
>>> using inappropriate debating
>>> tactics. I don't feel comfortable
>>> speculating what exactly Jouni was
>>> talking about, but one example
>>> that has caught my eye is that you
>>> are constantly interpreting other
>>> people's words and points as you
>>> want (you may look at the other
>>> threads your participating as
>>> well).
>>
>> Sorry, you are not being consistent
>> again (on purpose?).
>>
>> Was there just one example? Or *many*
>> ("constantly misinterp")?
>
>Not misinterpreting, interpreting. Please don't misquote what I said :) .

It's shorter than "interp. as I want", for I gather you wouldn't write this if the meaning was correct interpretation as perceived by you.

>One example refers to _a_ thing you do in your postings: interpret what other people
>say in your own way.

Well, isn't "one's own way" the one and only way to interpret something?

Other people may later say what they meant, at which point
>you say that what they say was misleading or something along those lines.
>
>Constantly refers to instants when you do this thing. For example, you say that
>I make up things, it pretty much means you think your interpretation of my words
>is correct and I was lying, when I wrote those.

Made-up facts refer entirely to him as I pointed out before, by made-up theory I refer to references which are allegedly made to theory, but not backed by anything. I have already explained to you, that it's never considered good manners to try to support oneself with "theory" word w/o sources. "Theory" is simply a wrong word if you want to come up with something like your "3 basic truths".

>> Please, if you want to be understood,
>> rather misunderstood, you should
>> follow your own rules: avoid obscure
>> statements, speak the language of
>> facts.
>
>Sorry, but even though I see in what way you misunderstood what I wrote, I don't
>know how you managed that. For me, my statement looks perfectly sensible. Perhaps
>there's an important aspect in English language that I'm unaware of. Whatever it
>is, I can't just decide to suddenly come aware of it. I can only write again, with
>different words, different examples etc. and hope to be more successful.

It would probably be much more useful if you refer to actual posts to show what you're talking about.

>> Again: I don't see you asking much
>> the only question you asked in the
>> first post was OoOE in GPU. And you
>> call that a good arguing style?
>
>At that point, I had no trouble understanding what you were talking about. I just
>think you had trouble understanding Jouni's point. It was that question about OoOE
>and GPUs that in few posts time lead to me wondering why you say OoO has no (theoretical)
>cost.

I asked you already if you're sure I said that, and haven't seen anything from you. So where did I say this?

I even wrote that I think I'm just misinterpreting something and that you
>are probably just talking more in practise than I.
>
>As I stated in that question, it was mostly rhetorical. I know the basic presumption
>that graphics workload wouldn't benefit about OoO. You said the same but peppered
>you answer with what I think shows you are argueing about different thing ("if OoOE
>is such a complete and utter waste of power and die space...").

That sentence must be positioned directly below something like "If OoO is so terrific/power-efficient ...", so when someone interprets what I say in such a manner, it's just fair to reply in a similar tone.

I don't think either
>Jouni or me ever stated anything like that. It isn't about in-order being better
>than OoO. It's about in-order being better than OoO sometimes, because OoO isn't free.

Which is similar to what I said well before you joined. I disputed very specific arguments that he brought.

>>>> Usually if you want to discuss
>>>> something else (or change the
>>>> context), you don't make direct
>>>> connections to prior discussion
>>>> as if the context remains the
>>>> same.
>>
>> It appears you have misunderstood
>> me. The meaning was "... one
>> shouldn't make direct connections
>> ...". It's equivalent form, e.g.
>> you can often see "you don't say
>> things like ...".
>
>Ah, then perhaps we even have a disagreement. Contexts change (or perhaps evolve
>would be a better word) during discussion. That's natural. If one, for example,
>skips few messages, one shouldn't think the contexts has remained the same.
>
>So what's the maximum permitted change between two consecutive messages? Depends
>on how much you explain the change, how perceptive readers are etc. We cannot measure
>these things and it depends on a reader how the change is perceived. I myself am
>pretty tolerant of these, someone else may prefer a new topic on a forum instead.

A discussion with little respect to existing context (I mean you and your 3 questions, along with a few other minor occasions) is a good recipe to not arrive at anything at all. Often practiced by people who don't seek any answers in fact.

>>> Correct me if I'm wrong, but doesn't
>>> "made-up" imply that the one making
>>> the claim was intentionally making
>>> false claims? I never made such
>>> claims (can't talk for other people
>>> of course).
>>
>> Covering yourself up with theory
>> hardly deserves better attitude.
>
>It seems to me that intent doesn't matter to you. That the fact that I couldn't
>make my point in a away that you would have understood makes me almost a liar to you.

There are many ways to sound more credible and respectful than you do:

- don't mention theory simply to add weight to your opinions/views/... This is lame.
- if you want to change context, make it clear. Trying to switch may work, but more likely you'll get what you deserve.
- don't assume you're the one who understands things right. The opposite can be true.

>Is there any point going on in that case?

I don't know. On one hand, you seem to be an interesting person, on the other I'm not quite sure what your real goal is.

>> I know you didn't; it's just that if
>> you really find that I tend to
>> misunderstand/misinterpret people
>> (on purpose?), I would highly
>> appreciate if you could be a little
>> more fact-based and not limiting
>> yourself to this thread.
>
>I'm starting to think you do misunderstand me on purpose. I very, very, very, very
>clearly stated that based on this discussion, I do _NOT_ think you are misinterpreting
>me on purpose. Yet there it is, "(on purpose?)". I'm not sure if you did it on purpose
>or if you just feel natural interpreting everything I say in as hostile way as possible.
>Either way, I'm not sure what to think anymore.
>
>So, unless you are willing to change your attitude, let's just stop. There's nothing
>I can say that you can't take the wrong way.

Well, you said that I tend to constantly interpret things as I want. You surely meant that my interpretation was wrong, correct? Hence the request to be fact-based.

>-JLarja