By: Linus Torvalds (torvalds.delete@this.linux-foundation.org), January 8, 2015 3:09 pm
Room: Moderated Discussions
juanrga (nospam.delete@this.juanrga.com) on January 7, 2015 6:03 am wrote:
>
> My point is that future applications will require the main processor to be a hundred/thousand manycore
> because multicores don't scale up and cannot provide the needed throughput
You're crazy. Manycore scales worse.
You are basically asking for a pink unicorn, stating - without anything what-so-ever to back your statement up, despite decades of history showing that you're wrong - that that pink unicorn will happen, and will be much better.
Because it's pink! And you're even willing to call something that is actually brown "pink", because just calling it that makes it magically better in your world.
You have absolutely zero technical reasons. You quote a few pie-in-the-sky PR blurbs as proof, and then you ignore actual technical arguments for why you're wrong by people who know that they are doing (which is not just me).
> I asked Linus twice how he proposes to reach the target goals
> and he twice avoided the question, he only writes ridiculous rants against manycores.
I'm ignoring your most crazy things. Pink unicorns don't exist, and they wouldn't solve any problems anyway. I've told you why you are wrong, and how your examples are pure fluff and PR-speak, and how reality is actually moving away from your proposed solutions, because your proposed solutions have been shown - over and over again - to be pure BS.
Even your one big hobby horse - knights landing - is moving away from what you state, and is more normal SMP than any kind of "sea of small weak cores". I told you so, others told you so, and you ignored it and just talk about how I'm wrong.
Please, you're just crazy clueless.
It was fun arguing with Paul DeMone, because by golly, he was technical and knew what he was talking about, even if we obviously disagreed violently about the end result.
You? Just sad.
> Reaching this goal will require novel hardware and software techniques [..]
Blah blah blah. Because mindlessly changing things will magically solve everything and make all the problems go away, and all the historical trends we've seen will be reversed.
Because "novel techniques" solves everything.
The reason you like quoting PR blurbs is that they are at exactly the same level of technical excellence as your nonsense. "Manycore is the future".
You're spouting nonsense. Those novel techniques aren't particularly new, they've been around for a long long time, and the reason they aren't in use today is that they have never really worked in the past either.
But magically, things are different, and all the knowledge we've gathered is worthless. Because pink unicorns will magically happen and make everything work.
That's literally your level of discourse. When people tell you pink unicorns don't exist, you point to horses and say that they are the new pink. And horses that have already been supplanted by horseless carriages at that.
Linus
>
> My point is that future applications will require the main processor to be a hundred/thousand manycore
> because multicores don't scale up and cannot provide the needed throughput
You're crazy. Manycore scales worse.
You are basically asking for a pink unicorn, stating - without anything what-so-ever to back your statement up, despite decades of history showing that you're wrong - that that pink unicorn will happen, and will be much better.
Because it's pink! And you're even willing to call something that is actually brown "pink", because just calling it that makes it magically better in your world.
You have absolutely zero technical reasons. You quote a few pie-in-the-sky PR blurbs as proof, and then you ignore actual technical arguments for why you're wrong by people who know that they are doing (which is not just me).
> I asked Linus twice how he proposes to reach the target goals
> and he twice avoided the question, he only writes ridiculous rants against manycores.
I'm ignoring your most crazy things. Pink unicorns don't exist, and they wouldn't solve any problems anyway. I've told you why you are wrong, and how your examples are pure fluff and PR-speak, and how reality is actually moving away from your proposed solutions, because your proposed solutions have been shown - over and over again - to be pure BS.
Even your one big hobby horse - knights landing - is moving away from what you state, and is more normal SMP than any kind of "sea of small weak cores". I told you so, others told you so, and you ignored it and just talk about how I'm wrong.
Please, you're just crazy clueless.
It was fun arguing with Paul DeMone, because by golly, he was technical and knew what he was talking about, even if we obviously disagreed violently about the end result.
You? Just sad.
> Reaching this goal will require novel hardware and software techniques [..]
Blah blah blah. Because mindlessly changing things will magically solve everything and make all the problems go away, and all the historical trends we've seen will be reversed.
Because "novel techniques" solves everything.
The reason you like quoting PR blurbs is that they are at exactly the same level of technical excellence as your nonsense. "Manycore is the future".
You're spouting nonsense. Those novel techniques aren't particularly new, they've been around for a long long time, and the reason they aren't in use today is that they have never really worked in the past either.
But magically, things are different, and all the knowledge we've gathered is worthless. Because pink unicorns will magically happen and make everything work.
That's literally your level of discourse. When people tell you pink unicorns don't exist, you point to horses and say that they are the new pink. And horses that have already been supplanted by horseless carriages at that.
Linus
