Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Observing the A11's Heterogenous Cores

mikeash.com

Post reply on HN
Print view View original

41–50 of 56 posts

Re: Observing the A11's Heterogenous Cores

#41

Post by cooper12 »

dang wrote: post #30
cooper12 wrote: post #16

Earlier quoted context omitted.

It's Apple. Gotta put those iSheep in their place, know what I mean?

Would you please stop posting unsubstantive and/or inflammatory comments to HN? You've done this quite a lot, and we eventually ban such accounts. https://news.ycombinator.com/newsguidelines.html

Oh please, the other comments weren't inflammatory? My comment certainly was substantive in pointing out a long-term trend I've noticed in blind Apple-bashing on HN without substance. Sorry I care about the quality of HN commentary, which has really taken a nosedive in recent years. You're definitely doing such a great job keeping it high /s.

You're actually part of the problem by being more concerned about me stepping on toes than the actual issue at hand. Funny how I never see mods like you step in when other inflammatory bait is posted though such as the threads on women in tech. Turns out its okay to barf out any garbage you like as long as you mask your tone well enough and it doesn't offend the reactionaries.

Don't blame me, I'm only reacting to low-effort commentary in turn. Quality discourse begets quality discourse, not the opposite. If you ban me, it'll only vindicate my belief that HN is more concerned about being an echo chamber for techbros and is diametrically opposed to any dissent.

If mods like you did their jobs, low-effort crap would be hidden and the users chided, but instead you shoot the messengers who try to preserve some semblance of quality. I'm usually a civil contributor which is obvious from my post history, but I won't just sit by while the culture of HN deteriorates and others tolerate it because somehow low-effort commentary is equivalent to every other type. Sorry I didn't coddle those users when I pointed out the circlejerk :(.

If the HN you want to create is one where people just mindlessly regurgitate the latest tech meme and make low-effort pop culture references, more power to you.

Re: Observing the A11's Heterogenous Cores

#42

Post by gok »

StillBored wrote: post #39

I'm not sure all this big.little stuff is the future. The heterogenous cores make scheduling/caching/etc a nightmare particularly when someone decides that the cores have different features. Particularly, since I think its all a crutch for lack of finer grained power control within the CPU core. Want to save more power, turn off a few of the duplicate functional units, shutdown 3/4 of the ROB buffer, or fuse a few pi…

Even SMP chips today don't make scheduling particularly easy. Some cores share caches and some don't, some are on different sockets, some don't share the same memory banks, some are clock boosted or down clocked...

Re: Observing the A11's Heterogenous Cores

#43

Post by gumby »

StillBored wrote: post #39

I'm not sure all this big.little stuff is the future. The heterogenous cores make scheduling/caching/etc a nightmare particularly when someone decides that the cores have different features. Particularly, since I think its all a crutch for lack of finer grained power control within the CPU core. Want to save more power, turn off a few of the duplicate functional units, shutdown 3/4 of the ROB buffer, or fuse a few pi…

I disagree: we've been playing these games for a long time and I don't see anything on the horizon.

You know how you start a phone call in Japan with Moshi moshi‡? In the 1990s the software in some of the PHS phones would wait until there was enough energy coming through the microphone before powering up the compression hardware in order to save just a little more power. So the "Moshi moshi" would get cut off.

Basically anything to save power -- people hate recharging.

‡ It's like the linguistic equivalent of the old TTY STX (which still survives in ASCII!)

Re: Observing the A11's Heterogenous Cores

#44

Post by milcron »

jrs95 wrote: post #36
pohl wrote: post #28

Earlier quoted context omitted.

Thank you for sharing. Here's the tl;dr: 47% of Android users sampled (51 out of 107) « said things like “iPhone users are a cult,” “I hate Apple,” “only stupid people use iPhones,” or “I hate the iPhone” when explaining to me why they chose an Android device. Seriously, people actually used the word hate » 5% of iPhone users sampled (6 out of 112) « said something negative about Android or about a specific Android p…

I can sort of understand that sentiment from Android users. They tend to value control/customization. I've tried Android several times for this reason, because it is compelling. I've always gone back to iOS though, and it's basically just been for a combination of performance and polished UX. My issues with Android were basically: * Little visual stutters and other relatively minor things bother me more than they sho…

I use the Copperhead ROM with F-Droid apps, and it addresses nearly all your bullet points. Excellent battery life, surprisingly high quality (and FOSS!) apps, and zero bloatware whatsoever.

Re: Observing the A11's Heterogenous Cores

#45

Post by mikeash »

StillBored wrote: post #39

I'm not sure all this big.little stuff is the future. The heterogenous cores make scheduling/caching/etc a nightmare particularly when someone decides that the cores have different features. Particularly, since I think its all a crutch for lack of finer grained power control within the CPU core. Want to save more power, turn off a few of the duplicate functional units, shutdown 3/4 of the ROB buffer, or fuse a few pi…

That's an interesting thought. You make a lot of sense to me, although I don't know that much about hardware design.

I would like to note, though, that the A11's high-efficiency cores are still pretty fast. In this (highly artificial) benchmark, they were able to accomplish more work in the same time than the high-performance cores. Since there's twice as many of them, that means they're at least 50% as fast on this code. So I don't know if this really will be the future, but the "slow" ones aren't useless by any means.

Re: Observing the A11's Heterogenous Cores

#47

Post by leggomylibro »

pohl wrote: post #28
GeekyBear wrote: post #25

Earlier quoted context omitted.

I've always found this experiment to be revealing: >Specifically, after explaining briefly that I was the editor of a tech news site, I would ask iPhone users “why do you use an iPhone?” and Android phone users “why do you use an Android phone?” My phone was always in my pocket when I asked, and I made sure to never mention a rival platform at all. In about three months I ended up asking 219 different people this que…

Thank you for sharing. Here's the tl;dr: 47% of Android users sampled (51 out of 107) « said things like “iPhone users are a cult,” “I hate Apple,” “only stupid people use iPhones,” or “I hate the iPhone” when explaining to me why they chose an Android device. Seriously, people actually used the word hate » 5% of iPhone users sampled (6 out of 112) « said something negative about Android or about a specific Android p…

So it sounds like Android users resent Apple's tactics/platform, while Apple users don't resent Android's tactics/platform?

That is definitely interesting.

Re: Observing the A11's Heterogenous Cores

#48

Post by leggomylibro »

mikeash wrote: post #45
StillBored wrote: post #39

I'm not sure all this big.little stuff is the future. The heterogenous cores make scheduling/caching/etc a nightmare particularly when someone decides that the cores have different features. Particularly, since I think its all a crutch for lack of finer grained power control within the CPU core. Want to save more power, turn off a few of the duplicate functional units, shutdown 3/4 of the ROB buffer, or fuse a few pi…

That's an interesting thought. You make a lot of sense to me, although I don't know that much about hardware design. I would like to note, though, that the A11's high-efficiency cores are still pretty fast. In this (highly artificial) benchmark, they were able to accomplish more work in the same time than the high-performance cores. Since there's twice as many of them, that means they're at least 50% as fast on this…

I have no idea how CPUs work, but MCUs do sorta work like how they describe; want to save power? You can easily tell the chip to cut power/clock to certain peripherals, change stuff like core clock speed and memory access latencies, make some peripherals run at lower or higher clock speeds than the core, etc etc.

You get really fine-grained control that allows a powerful and beefy chip to achieve really good power efficiency if the application is well-written, and a sleek ultralight 1.8V chip to devour power if it isn't.

Re: Observing the A11's Heterogenous Cores

#49

Post by mikeash »

leggomylibro wrote: post #48
mikeash wrote: post #45

Earlier quoted context omitted.

That's an interesting thought. You make a lot of sense to me, although I don't know that much about hardware design. I would like to note, though, that the A11's high-efficiency cores are still pretty fast. In this (highly artificial) benchmark, they were able to accomplish more work in the same time than the high-performance cores. Since there's twice as many of them, that means they're at least 50% as fast on this…

I have no idea how CPUs work, but MCUs do sorta work like how they describe; want to save power? You can easily tell the chip to cut power/clock to certain peripherals, change stuff like core clock speed and memory access latencies, make some peripherals run at lower or higher clock speeds than the core, etc etc. You get really fine-grained control that allows a powerful and beefy chip to achieve really good power ef…

I believe CPUs are similar to some extent. Modern ones can clock up and down, and computing units which aren't in use can be shut down. (An example of this that's getting a lot of attention right now is Intel's AVX-512 unit, which puts out so much heat they have to clock the entire CPU down when it's in use.)

Apparently current designs can't quite do this enough, though. I'd definitely like to know the reason for that.

Re: Observing the A11's Heterogenous Cores

#50

Post by 95014_refugee »

StillBored wrote: post #39

I'm not sure all this big.little stuff is the future. The heterogenous cores make scheduling/caching/etc a nightmare particularly when someone decides that the cores have different features. Particularly, since I think its all a crutch for lack of finer grained power control within the CPU core. Want to save more power, turn off a few of the duplicate functional units, shutdown 3/4 of the ROB buffer, or fuse a few pi…

IOW you think that the folks working on these things are idiots, or lazy, or perhaps that your insights are so rich and rare that they can’t possibly have occurred to any of them?

Has it occurred to you that perhaps all of those things you’re mentioning have already been done, along with all the other easy ones, and the reason that something like this that’s harder to exploit is only just surfacing is because it got put off because it’s so hard to exploit?

Gerard &co. are a lot better at this than you give them credit for.

Post reply on HN