Crikey. I pretty much quit watching TV 30 years ago, because there wasn't anything worth watching... when I moved to the US I didn't even bother getting a TV set until after I married. Now it's online, and there still isn't anything worth watching. I've looked at hulu, and the content is the same kinds of shows I don't watch on regular TV.
So why is boxee vs hulu "news for nerds" or "stuff that matters"? Or do couch potatoes count as geeks these days?
Yah, I just wonder how you'd tell what the problem was before it trashed your laptop if it was a current 17" Macbook Pro... since the battery's completely internal with no way to relieve the pressure without removing 16 screws.
In a Macbook Pro? Once, so far, but it wasn't because it wasn't holding a charge... it was because it was visibly swelling in it case. If I'd had to wait until it had distorted the laptop case before I noticed it was swelling and replaced it... well, replacing that battery would have taken a week or more while I waited for Apple to repair or replace my laptop and ship it back to me.
If your Macbook Pro battery starts swelling to the point where it's likely to damage the laptop, as mine did, you'll be able to pop out the battery as soon as you notice it, and get an advance replacement from Apple overnighted to you the way I did, and run your laptop off AC in the meantime.
And then when you talk performance, x86 by having special instructions for specific functions offers far more opportunity to optimize 'hot spot' style.
And yet when you compare x86 processors with other processors using comparable process and gate counts, the larger register file on those other processors generally offers more opportunity for optimization that the special instructions on the x86, and where special instructions DO provide a benefit they have been added to other processors... so even that's not really an advantage of the x86 instruction set.
Getting back to the code size... what I'm referring to is the fact that some years back when people started talking about the x86 instruction set as being "compressed", they pointed out that it had a big advantage on the low end where memory was tight. Not it's apparently the low end where memory doesn't matter and the high end where memory is tight. This turnabout really makes it hard for me to appreciate the argument.
The x86 has simply had far more resources applied to making it go fast, that's all.
I'm not making this up, you know. The term "implied conflict of interest" shows up on the code of ethics for an awful lot of companies and professions, and I don't see any reason lawyers should be given a free ride just because your cynical (or realistic) opinion is that they're amoral mercenaries who are supposed to be above such legal niceties.
I don't care if they think they're a special case, or even if you think they're a special case. If an advertiser or a speechwriter or an engineer or a judge or a manager has to show they have no conflict of interest, lawyers get no free ride.
You're the one who brought it up, so why don't you tell us what it's about?
You're drunk or stoned.
The GGP suggested using Apple's iPod connector.
I asked a question, to wit "Don't you have to pay Apple a fee to use this connector?" I asked this connection because of the fuss a bunch of companies made a year or so back about being charged more money by Apple to make iPod-compatible peripherals.
You asked me which connector I was talking about.
I clarified that, and repeated the question.
Apple can't charge a fee for using a connector; the idea is simply ridiculous.
Have you ever heard of hardware patents? There's plenty of patented connectors, Apple has even patented some (for example the magsafe power connector and they're also licensing the mini displayport connector on the new Macbooks). In this case it appears that what Apple was charging for was the iPod trademark, but the idea that a connector is patented and requires a license fee is not "patently ridiculous", nor is the idea that Apple would have such a patent.
Don't you have to pay Apple a fee to use this connector?
Also, they seem to have decided to punt Firewire in future products, so the high speed interface isn't going to remain an advantage.
Apple is to be lauded for sticking to a consistent connector and form factor, but there's no chance of this becoming a long term standard. If there's a micro-version of USB3 available, that should alleviate your performance fears.
Also...
But just using iPhone as an example, plugging it in with the USB connector only makes it show up as a camera device to access pictures. Nothing else.
It doesn't show up as a USB Mass Storage device, like my iPod Shuffle does? o_O
Their register windows meant that despite the chip having a nice big register file, the number of available registers at any time were nearly as few as the x86. They also meant that there was very little flexibility in the calling sequence, and that you have a relatively large processor state (despite a small set of working registers) so context switches were expensive. To avoid that they built in more registers so you could keep multiple register sets live... but it was a partial mitigation at best: you could really see the effect in multitasking benchmarks: the knee in the curve when you ran out of register contexts was phenomenal.
Better RISCS with zoned register files only defined a single register set in the actual architecture, and left it to the implementation to decide how big the register file really was... the register file, then, became effectively a register cache.
Yes, compared to the x86 it looked pretty clean, but just about ANYTHING does. It was about the only RISC in the heyday of RISC that was actually slower than the P5.
The CISC x86 instruction set is a form of compression.
And the tiny register file is what?
The problem with the x86 instruction set is not that it's complex (and it's not really that complex... it's not very much like a traditional CISC, and the complex addressing modes of the VAX and 68020 made them harder to pipeline than x86), it's that it's badly designed in many ways.
The operations that have to be used to turn an x86 instruction stream into a set of microinstructions that can be efficiently pipelined are still horribly complex. They don't resemble "decompression" at all, they have to parse the instructions, reverse-engineer the registers they're using, figure out how to reschedule them using multiple register sets to avoid pipeline stalls, and so on... and it's still not efficient. It's only fast as it is because intel's able to throuw so many transistors at it and still get acceptable yeilds.
If you want a REAL "compressed" instruction set, look at the "Thumb" instructions in ARM. And for embedded systems tight code matters MORE than it does for desktops and servers: ARM doesn't dominate the low end because it's RISC, it dominates because it's well designed all the way across the board for that niche.
So, no, the x86 isn't CISC, it's not "compressed", it's just badly designed.
It's getting that way, but if the legislators get hold of it they'll probably define yet another new and unnecessary standard instead of something sensible like that.
Alpha was scaling quite nicely, thank you, all the way through EV7, and was completely on track for the original plan. Alpha didn't "run out of steam", Alpha was deliberately killed. The EV8 program had its funds and manpower cut, and then that was used as an excuse to kill it.
And little of what made Alpha good ended up in Hammer. Alpha wasn't am implementation, it was an architecture, and what made that architecture effective was in the instruction set and memory model that let the implementation change without ending up in the dead ends that claimed MIPS and SPARC... and IA64.
Were the second generation of mice raised with their environmentally enriched parents? Mice learn, that's obvious, that's how the first generation benefitted from their richer environment. Well, they learn from other mice, too. Like, their parents.
Have they corrected for nurture? Have they corrected for differences in the mother's milk, even?
... to be precise, by intel's bankroll and investment in process.
Power PC and Alpha were outcompeted by the fundamentally inferior x86 family not because of flaws in their designs, but because intel spent more on improving their process than anyone else.
Both the Power PC and Pentium turned into furnaces, the Pentium 4 and G5 were both following the "megahertz myth" into long pipelines to let the clock speed ramp up. Neither got the clock speeds they were hoping for. Both were too hot for mobile processors. In both cases the solution was going to be shorter pipelines, slower but more clock-efficient cores, and faster busses. The Freescale e700 was torpedoed when Apple went with Intel's Core Duo... because Intel had the resources to get their respin of the PIII out quicker than Freescale could get their respin of the G4 online.
So now we're still using hacks upon hack on the truly horrible x86 architecture.
Well, it could have been worse. It could have been SPARC.
Slashdot Mirror
*snort*
Don't DO that when I'm drinking coffee.
Crikey. I pretty much quit watching TV 30 years ago, because there wasn't anything worth watching... when I moved to the US I didn't even bother getting a TV set until after I married. Now it's online, and there still isn't anything worth watching. I've looked at hulu, and the content is the same kinds of shows I don't watch on regular TV.
So why is boxee vs hulu "news for nerds" or "stuff that matters"? Or do couch potatoes count as geeks these days?
Yah, I just wonder how you'd tell what the problem was before it trashed your laptop if it was a current 17" Macbook Pro... since the battery's completely internal with no way to relieve the pressure without removing 16 screws.
In a Macbook Pro? Once, so far, but it wasn't because it wasn't holding a charge... it was because it was visibly swelling in it case. If I'd had to wait until it had distorted the laptop case before I noticed it was swelling and replaced it... well, replacing that battery would have taken a week or more while I waited for Apple to repair or replace my laptop and ship it back to me.
If your Macbook Pro battery starts swelling to the point where it's likely to damage the laptop, as mine did, you'll be able to pop out the battery as soon as you notice it, and get an advance replacement from Apple overnighted to you the way I did, and run your laptop off AC in the meantime.
Right?
And then when you talk performance, x86 by having special instructions for specific functions offers far more opportunity to optimize 'hot spot' style.
And yet when you compare x86 processors with other processors using comparable process and gate counts, the larger register file on those other processors generally offers more opportunity for optimization that the special instructions on the x86, and where special instructions DO provide a benefit they have been added to other processors... so even that's not really an advantage of the x86 instruction set.
Getting back to the code size... what I'm referring to is the fact that some years back when people started talking about the x86 instruction set as being "compressed", they pointed out that it had a big advantage on the low end where memory was tight. Not it's apparently the low end where memory doesn't matter and the high end where memory is tight. This turnabout really makes it hard for me to appreciate the argument.
The x86 has simply had far more resources applied to making it go fast, that's all.
You're mixing up patents and copyrights.
I'm not making this up, you know. The term "implied conflict of interest" shows up on the code of ethics for an awful lot of companies and professions, and I don't see any reason lawyers should be given a free ride just because your cynical (or realistic) opinion is that they're amoral mercenaries who are supposed to be above such legal niceties.
I don't care if they think they're a special case, or even if you think they're a special case. If an advertiser or a speechwriter or an engineer or a judge or a manager has to show they have no conflict of interest, lawyers get no free ride.
You're the one who brought it up, so why don't you tell us what it's about?
You're drunk or stoned.
The GGP suggested using Apple's iPod connector.
I asked a question, to wit "Don't you have to pay Apple a fee to use this connector?" I asked this connection because of the fuss a bunch of companies made a year or so back about being charged more money by Apple to make iPod-compatible peripherals.
You asked me which connector I was talking about.
I clarified that, and repeated the question.
Apple can't charge a fee for using a connector; the idea is simply ridiculous.
Have you ever heard of hardware patents? There's plenty of patented connectors, Apple has even patented some (for example the magsafe power connector and they're also licensing the mini displayport connector on the new Macbooks). In this case it appears that what Apple was charging for was the iPod trademark, but the idea that a connector is patented and requires a license fee is not "patently ridiculous", nor is the idea that Apple would have such a patent.
Total stream-of-consciousness babble with no point, no reasoning, no cause and effect. What, did you do too many heavy drugs when you were 17?
Does this count the people who the RIAA sued who didn't even own a computer?
So Bush attacking Iraq and ignoring Afghanistan had nothing to do with the oil business?
Doesn't matter, there's an implied conflict of interest in intervening on behalf of former employers.
The manufacturers seem to have decided to go with micro-USB.
What Connector? Micro USB? No. Mini USB? No. The iPod connector?
The iPod connector, of course. That's the one the parent article was referring to. There's this thing called "context", you want to try using it?
Sheesh.
I thought there was a fuss a while back about Apple charging accessory manufacturers a fee to use this connector.
What was that about?
Don't you have to pay Apple a fee to use this connector?
Also, they seem to have decided to punt Firewire in future products, so the high speed interface isn't going to remain an advantage.
Apple is to be lauded for sticking to a consistent connector and form factor, but there's no chance of this becoming a long term standard. If there's a micro-version of USB3 available, that should alleviate your performance fears.
Also...
But just using iPhone as an example, plugging it in with the USB connector only makes it show up as a camera device to access pictures. Nothing else.
It doesn't show up as a USB Mass Storage device, like my iPod Shuffle does? o_O
What's wrong with SPARC?
Their register windows meant that despite the chip having a nice big register file, the number of available registers at any time were nearly as few as the x86. They also meant that there was very little flexibility in the calling sequence, and that you have a relatively large processor state (despite a small set of working registers) so context switches were expensive. To avoid that they built in more registers so you could keep multiple register sets live... but it was a partial mitigation at best: you could really see the effect in multitasking benchmarks: the knee in the curve when you ran out of register contexts was phenomenal.
Better RISCS with zoned register files only defined a single register set in the actual architecture, and left it to the implementation to decide how big the register file really was... the register file, then, became effectively a register cache.
Yes, compared to the x86 it looked pretty clean, but just about ANYTHING does. It was about the only RISC in the heyday of RISC that was actually slower than the P5.
The CISC x86 instruction set is a form of compression.
And the tiny register file is what?
The problem with the x86 instruction set is not that it's complex (and it's not really that complex... it's not very much like a traditional CISC, and the complex addressing modes of the VAX and 68020 made them harder to pipeline than x86), it's that it's badly designed in many ways.
The operations that have to be used to turn an x86 instruction stream into a set of microinstructions that can be efficiently pipelined are still horribly complex. They don't resemble "decompression" at all, they have to parse the instructions, reverse-engineer the registers they're using, figure out how to reschedule them using multiple register sets to avoid pipeline stalls, and so on... and it's still not efficient. It's only fast as it is because intel's able to throuw so many transistors at it and still get acceptable yeilds.
If you want a REAL "compressed" instruction set, look at the "Thumb" instructions in ARM. And for embedded systems tight code matters MORE than it does for desktops and servers: ARM doesn't dominate the low end because it's RISC, it dominates because it's well designed all the way across the board for that niche.
So, no, the x86 isn't CISC, it's not "compressed", it's just badly designed.
Crikey, I never was much interested in Facebook, and now I'm twice as glad. You could do all kinds of naughty things to someone like that.
It could be Mini-A instead of Mini-B, or it could be Micro-A or Micro-B, or maybe USB-OTG Mini-AB or Micro-AB.
http://en.wikipedia.org/wiki/USB#Mini_and_micro
It's getting that way, but if the legislators get hold of it they'll probably define yet another new and unnecessary standard instead of something sensible like that.
Alpha was scaling quite nicely, thank you, all the way through EV7, and was completely on track for the original plan. Alpha didn't "run out of steam", Alpha was deliberately killed. The EV8 program had its funds and manpower cut, and then that was used as an excuse to kill it.
And little of what made Alpha good ended up in Hammer. Alpha wasn't am implementation, it was an architecture, and what made that architecture effective was in the instruction set and memory model that let the implementation change without ending up in the dead ends that claimed MIPS and SPARC... and IA64.
How does Burger King know you've unfriended someone?
Were the second generation of mice raised with their environmentally enriched parents? Mice learn, that's obvious, that's how the first generation benefitted from their richer environment. Well, they learn from other mice, too. Like, their parents.
Have they corrected for nurture? Have they corrected for differences in the mother's milk, even?
... to be precise, by intel's bankroll and investment in process.
Power PC and Alpha were outcompeted by the fundamentally inferior x86 family not because of flaws in their designs, but because intel spent more on improving their process than anyone else.
Both the Power PC and Pentium turned into furnaces, the Pentium 4 and G5 were both following the "megahertz myth" into long pipelines to let the clock speed ramp up. Neither got the clock speeds they were hoping for. Both were too hot for mobile processors. In both cases the solution was going to be shorter pipelines, slower but more clock-efficient cores, and faster busses. The Freescale e700 was torpedoed when Apple went with Intel's Core Duo... because Intel had the resources to get their respin of the PIII out quicker than Freescale could get their respin of the G4 online.
So now we're still using hacks upon hack on the truly horrible x86 architecture.
Well, it could have been worse. It could have been SPARC.