Thinq: Nvidia declares that Google tablets are the future

[taltamir]

Seriously? Android OS uses a modified Linux kernel, and is closing in on 50% market share.

I know android, I thought it was not linux.
I could try to save face with a "its not really linux, its heavily modified kernal based on linux with different userspace" blah blah...

Linux compatibility
Android's kernel was derived from Linux but has been tweaked by Google outside the main Linux kernel tree.[141] Android does not have a native X Window System nor does it support the full set of standard GNU libraries, and this makes it difficult to port existing GNU/Linux applications or libraries to Android.[142] However, support for the X Window System is possible.[143]
Google no longer maintains the code they previously contributed to the Linux kernel as part of their Android effort, effectively branching kernel code in their own tree, separating their code from Linux.[144][145][146] This was due to a disagreement about new features Google felt were necessary[citation needed]. The code which is no longer maintained was deleted in January 2010 from the Linux codebase.[147] However, Google announced in April 2010 that they will employ staff to work with the Linux kernel community.[148]

but I did not recall that it is linux based when I made that statement, so I am just going to come out and say I was wrong.

 

[AnandThenMan]

I know android, I thought it was not linux.

Just out of curiosity, what did you think it was?

but I did not recall that it is linux based when I made that statement, so I am just going to come out and say I was wrong.

No shame in admitting you were wrong. :thumbsup:

BTW, there are many, many variations of the Linux kernel. Too many.

 

[taltamir]

Just out of curiosity, what did you think it was?

I thought google built it from the ground up.

 

[AnandThenMan]

I thought google built it from the ground up.

I bet Google considered it at one point. But that is a daunting task, it's easier to use Linux, the infrastructure is already in place.

I am surprised at how well Android has caught on though TBH.

 

[ModestGamer]

Linux on the phone? where?

Also, your comments are veering very close to flaming and personal insults. I don't think you are quite there but you are certainly being insulting, rude, and condescending. Please don't do that, there is no need or call for that in a debate between smart adults.

Where are these smart adults ?

Microsoft with all its might is salivating at x86 1w APU chips for phones.

You'll see it next year.

All your favoriate MS apps on your phone.

LOL

software providers are gonna support that living crap out of that. Why ?? becuase win anything will be easy to support when its on a x86 capable phone.

It'll actually be better for the consumer as having one good
Os on pretty much every port to the network and being able to easily share data will make it uber competitive.

IT departments will rejoice.

 

[GodisanAtheist]

Microsoft with all its might is salivating at x86 1w APU chips for phones.

-That's really weird cause Windows Mobile 7 is running just fine on ARM SOCs that use a couple hundred milliwatts under load, and next to nothing when idling. 1W of power use might be damn amazing for a netbook or tablet but its still 3-4 times more power than it has to be to be competitive with TODAY'S smartphones.

And its not like everyone's walking around complaining about their phones having too much battery life...

 

[ModestGamer]

-That's really weird cause Windows Mobile 7 is running just fine on ARM SOCs that use a couple hundred milliwatts under load, and next to nothing when idling. 1W of power use might be damn amazing for a netbook or tablet but its still 3-4 times more power than it has to be to be competitive with TODAY'S smartphones.

And its not like everyone's walking around complaining about their phones having too much battery life...

Yes but software must be recompiled and revalidated for windows 7 mobile. There are substantial API differences as well.

But its the other ARM spaces where they will strike first. with sub 1watt apu's in the pipeline I don't expect to see ARM be as popular.

what would drastically change the landscape would be the emergence of a good solid OS that can play ball with microsoft that is properly written so as to be portable to different architectures.

That would give arm some reverse penatration back to the dekstop/client/server market.

x86 is to domonanit but at least on support ISA is better then 40 with regards to software developement.

 

[cfedu]

Nothing last forever, Microsoft had a good run with Windows and intel had a good run with x86. 20 years ago desktops destroyed notebook shipments. today notebooks out ship desktops. Arm outsells both together. As much as I dislike nVidia's business practices tegra is a good idea, there execution may not have been optimal, but the idea is sound.

When tablets and phones are powerful enough for the average consumer to be there main computer, windows and x86 will be the niche professional market/gamers. New programs will be made for arm and counsels and then ported to x86. Prices will go up so much that gamers will be forced out of the market and have to buy a PS5 or xbox 4. Apple knows this and they are slowly disappointing there computers customers. MAC Computers make huge money and Apple does not care about it. Why? because x86 is dead.


intel will adapt as they have great technology and cash to adapt. Microsoft will be a shadow of its former self.

 

[AnandThenMan]

intel will adapt as they have great technology and cash to adapt.

How exactly will Intel adapt? Without x86 Intel has nothing. Well maybe some solid state storage sales I guess.

 

[DeathRayLoveMachine]

Yes but software must be recompiled and revalidated for windows 7 mobile. There are substantial API differences as well.

ModestGamer, do you understand the terms you're using? You were talking about CPU architectures, not APIs. APIs are a completely different level of the software stack than the executable code that gets run on your CPU.

W7M and W7 have different APIs because the W7 APIs would be slit-your-wrists terrible on a smartphone.

 

[ModestGamer]

ModestGamer, do you understand the terms you're using? You were talking about CPU architectures, not APIs. APIs are a completely different level of the software stack than the executable code that gets run on your CPU.

W7M and W7 have different APIs because the W7 APIs would be slit-your-wrists terrible on a smartphone.

you do understand the ARM is RISC and that the normal win7 x86 api's requires alot of rewriting to work and alot of x86 compiled code also needs to be rewritten to work on RISC.

The CPU can and often does effect the API if its not written architecture nuetral.

 

[ModestGamer]

Nothing last forever, Microsoft had a good run with Windows and intel had a good run with x86. 20 years ago desktops destroyed notebook shipments. today notebooks out ship desktops. Arm outsells both together. As much as I dislike nVidia's business practices tegra is a good idea, there execution may not have been optimal, but the idea is sound.

When tablets and phones are powerful enough for the average consumer to be there main computer, windows and x86 will be the niche professional market/gamers. New programs will be made for arm and counsels and then ported to x86. Prices will go up so much that gamers will be forced out of the market and have to buy a PS5 or xbox 4. Apple knows this and they are slowly disappointing there computers customers. MAC Computers make huge money and Apple does not care about it. Why? because x86 is dead.


intel will adapt as they have great technology and cash to adapt. Microsoft will be a shadow of its former self.

I really don't see that happening. What is blaringly obvious is that AMD and Intel are both building some pretty nice ultra low power APU chips which will be very very powerful.


As the process's shrink the spread in power cunsumption between devices will also shrink. The big power advantage arm has had will be very very minimal at 18nm 2014 or so that it won't matter anymore.

The big thing is going to be cross platform compatability.

don't fool yourself. Phones aren't going to replace desktops anytime soon. I doubt ever. Phones will likely plug into some larger mainframe cloud home client hybrid we have yet to see a model for.

 

[DeathRayLoveMachine]

you do understand the ARM is RISC and that the normal win7 x86 api's requires alot of rewriting to work and alot of x86 compiled code also needs to be rewritten to work on RISC. The CPU can and often does effect the API if its not written architecture nuetral.

This is word salad. RISC is not an architecture, it's a word used to describe certain kinds of architectures; ARM and x86 are source compatible, so the vast majority of code can be ported with a recompile, not a total rewrite; all APIs are "architecture neutral" by default, because APIs are several levels higher than the CPU architecture on the software stack; the only APIs that aren't "architecture neutral" are the ones that directly interface with the CPU itself; there aren't very many such APIs; a company like Microsoft easily has the resources to port those APIs to any architecture they damn well please; and those APIs aren't very important to software compatibility anyway.

PS your spelling is a sin. Spell-checks are built into all modern browsers, there is no reason not to use one.

 

[ModestGamer]

This is word salad. RISC is not an architecture, it's a word used to describe certain kinds of architectures; ARM and x86 are source compatible, so the vast majority of code can be ported with a recompile, not a total rewrite; all APIs are "architecture neutral" by default, because APIs are several levels higher than the CPU architecture on the software stack; the only APIs that aren't "architecture neutral" are the ones that directly interface with the CPU itself; there aren't very many such APIs; a company like Microsoft easily has the resources to port those APIs to any architecture they damn well please; and those APIs aren't very important to software compatibility anyway.

PS your spelling is a sin. Spell-checks are built into all modern browsers, there is no reason not to use one.

RISC and X86 are not cross platform compatable and ARM is a RISC architecture.

the API can be effected by the available instruction set. X86 code generally is far more verbose and has alot more instructions available.

API's are massively important to application compatability. API

Application Programming Interface.

ISA Instruction Set Architecture.

Some x86 code will work on risc some will not. It depends on if the code was written specifically to work with the much larger and more powerful x86_64 ISA or not.

secondly RISC stands for

Reduced Instruction Set Computing

They use less instructions and focus on only those thats are the most efficient, not nessacarily the most powerful. That means that code compiled and written for x86 may not work on ARM and vice versa. There are some differences in CPU intialization etc.

RISC is about efficiency, x86 is about powerful verbose abstraction and I don't think x86 is going anywhere anytime soon. What we may see in the future is the dropping of alot of the oklder 16 and 32 bit instructions and a more streamlined x64 type architecture.

 

[DeathRayLoveMachine]

ModestGamer, have you ever used a compiler? What programming languages do you know?

Have you ever developed on x86? On ARM? On Both?

 

[ModestGamer]

ModestGamer, have you ever used a compiler? What programming languages do you know?

Have you ever developed on x86? On ARM? On Both?

your comments withstanding are incorrect on the nature of RISC,CISC and the various ISA and software comptability as well as the x86 vrs ARM differences.

But if you want to try to draw me into some semantical argument. I will not indulge you.

Here is your hat, you may leave the discussion now.

 

[DeathRayLoveMachine]

I'm going to take that as a "no" on all questions.

 

[Voo]

Even if we ignore the facts that porting Windows to ARM really wouldn't be all that hard and porting user mode apps from Win x86 to Win ARM would be mostly trivial (because you know, the low level APIs that'd have to be rewritten aren't documented), that ignores one rather important fact:

Just because you CAN run your desktop app on a mobile phone doesn't mean it'll work terribly good. In fact you'll have to rewrite at least the whole GUI anyhow (and most probably IO as well).


Oh and someone missed the fact that modern x86 compilers don't use a big part of the x86 ISA, because those get just split up into several µops and usually are rather slow because not really optimized and also because it's much easier to write a compiler for a simple, regular ISA.


PS: And I really can't let that one go:

Reduced Instruction Set Computing
They use less instructions

I'll just quote Fred Brooks on that one:
"RISC stands for a computer with a set of reduced instructions, not with a reduced set of instructions."
And if you don't believe him, just compare some modern RISC ISAs to the VAX ISA.

 

[ModestGamer]

Even if we ignore the facts that porting Windows to ARM really wouldn't be all that hard and porting user mode apps from Win x86 to Win ARM would be mostly trivial (because you know, the low level APIs that'd have to be rewritten aren't documented), that ignores one rather important fact:

Just because you CAN run your desktop app on a mobile phone doesn't mean it'll work terribly good. In fact you'll have to rewrite at least the whole GUI anyhow (and most probably IO as well).


Oh and someone missed the fact that modern x86 compilers don't use a big part of the x86 ISA, because those get just split up into several µops and usually are rather slow because not really optimized and also because it's much easier to write a compiler for a simple, regular ISA.


PS: And I really can't let that one go:

I'll just quote Fred Brooks on that one:
"RISC stands for a computer with a set of reduced instructions, not with a reduced set of instructions."
And if you don't believe him, just compare some modern RISC ISAs to the VAX ISA.

ehhh the idea behind RISC is to take away all the extra low performance instructions and leave behind the most efficient. Your arguing semantics on the terminology they both mean the same thing. .

the x86 ISA and microsoft have a long history. There are alot of hooks deep deep in the code and the applications that have them have nothing to do with the compiler. MS and other x86 software vendors are definately going to get onboard is AMD drops a sub1w APU on the market in the next 10-12 months.



Lets see who was right at the end of 2011.

BTW alot of MS software could be made to run on phones IF they had x86 support without to much work. Thats where keeping a flat API will make that possiable.

 

[ModestGamer]

I'm going to take that as a "no" on all questions.

Incorrect. Compiler can only compile the code it is given, if you use librarys that depend on x86, your looking at a rewrite and not all code is as clean as it should be.

On another note MS did buy a ARM liscense. So now my interest it peaked. Might they cleanup the code base for nonx86 code and make it more portable ?

My thoughts are , not terriably likely.They will push forward with wince and build some BS devices. But that liscense may have been purchased without a full product protfolio comming from AMD and intel.

We shall see what the future brings.

 

[Genx87]

What phone or tablets are coming out with an x86 APU in the next 12 months? I am really curious.

 

[Voo]

Incorrect. Compiler can only compile the code it is given, if you use librarys that depend on x86, your looking at a rewrite and not all code is as clean as it should be.

Uh so you're saying that the libraries will also have to be recompiled? How horrible.. but not actually "rewriting", because you know C/C++/whatever code can easily be recompiled into ARM.
And the part that can't be easily recompiled is hidden away deep inside some kernel APIs/drivers which you know aren't publicly documented and therefore shouldn't be used in user programs.
And if you really think that there's a unreasonable amount of work involved, just look at the ported Linux kernel. Or the Windows Itanium port. Actually Win NT was written with portability in mind, see for example here

Name five publicly documented Win32 APIs that are ISA specific and can't be easily ported, because those are the ones that would break being able to easily recompile programs.

Although the problem still stands: Even if you could just recompile the program it wasn't intended to be run on a phone and you don't WANT to port the complete Win32 API to phones because lots of stuff doesn't even make sense in that context.

ehhh the idea behind RISC is to take away all the extra low performance instructions and leave behind the most efficient. Your arguing semantics on the terminology they both mean the same thing.

You didn't read the quote did you? A reduced set of instructions is something COMPLETELY different than a set of reduced instructions.
Like I said there are CISC ISAs out there that are smaller than some RISC ISAs, which means your interpretation is just plain wrong.

 

[ModestGamer]

Uh so you're saying that the libraries will also have to be recompiled? How horrible.. but not actually "rewriting", because you know C/C++/whatever code can easily be recompiled into ARM.
And the part that can't be easily recompiled is hidden away deep inside some kernel APIs/drivers which you know aren't publicly documented and therefore shouldn't be used in user programs.
And if you really think that there's a unreasonable amount of work involved, just look at the ported Linux kernel. Or the Windows Itanium port. Actually Win NT was written with portability in mind, see for example here.

Win Nt ports have had alot of compatability issues. That siad there is also alot of application specific code that needs x86. With the move to .net it may become a moot point shortly, or it may not.

Name five publicly documented Win32 APIs that are ISA specific and can't be easily ported, because those are the ones that would break being able to easily recompile programs.

Its not always about recompile. You realize MS has its own compilers which will require porting. Not to mention some of the librarys themselves.

Although the problem still stands: Even if you could just recompile the program it wasn't intended to be run on a phone and you don't WANT to port the complete Win32 API to phones because lots of stuff doesn't even make sense in that context.

As it currently stands you stance is true. As more power in terms of CPU come along it will not hold true. First sub 1w MCU style x86 device will prove this out.

You didn't read the quote did you? A reduced set of instructions is something COMPLETELY different than a set of reduced instructions.
Like I said there are CISC ISAs out there that are smaller than some RISC ISAs, which means your interpretation is just plain wrong.

there are many ISA's around RISC. However your having a semantical debate. both statements can actually be true on a ISA basis.

I am done with this topic.

Lets refresh it at this time next year and see who was correct on market direction.

 

[Voo]

Win Nt ports have had alot of compatability issues. That siad there is also alot of application specific code that needs x86. With the move to .net it may become a moot point shortly, or it may not.

Since even Win7 is still based on the Win NT kernel and that was ported to Itanium (and other kernels based on NT were ported to alpha, MIPS, powerPC,..) without much problems for a product of its size, where exactly do you get your information? Actually you do know that NT wasn't even developed on x86 machines at the beginning, but was later ported to them? 😉
Or do you have any MS statements where they claim they tried porting Win to another platform and had to give up because of compatability issues?
The only reason MS stopped porting their newest OSes to alpha, MIPS or whatever was that x86 PCs just dominate the market and there was no reason to invest money into it.

Its not always about recompile. You realize MS has its own compilers which will require porting. Not to mention some of the librarys themselves.

Yeah I do know that MS uses VC. And I also do know that writing a backend for ARM isn't really that hard, just go look at gcc, they had backends for dozens of different architectures for years, heck I wrote a small one for a fun architecture I ran on a FPGA.
And the libraries also just have to be recompiled, except for those few that are architecture dependent which are hidden behind the HAL.

As it currently stands you stance is true. As more power in terms of CPU come along it will not hold true. First sub 1w MCU style x86 device will prove this out.

Yeah, but there are basic things that are just different: The GUI for a 800x600 3" wide touch screen monitor has to be completely different than the GUI for a 2560x1600 30" LCD monitor. Just look how great those android phone apps work on tablet sized hardware. And that's a much smaller difference AND the same input variant.
Or another example: I doubt that we'll get access to a conventional FS on any modern phone without rooting it, although that's arguable.

there are many ISA's around RISC. However your having a semantical debate. both statements can actually be true on a ISA basis.

Umn I cited a guy, who was involved in the whole development of RISC architectures, from his turing award acceptance speech (iirc) and refuted your statement with a particular example (ARM vs. VAX-11; with the ARM ISA being much larger).

But yeah distinguishing between RISC/CISC isn't that simple, because both borrow from the other, but the intention of a RISC architecture was always to create a regular set of small, fast and simple instructions, while CISC guys had no problems including op codes for evaluating whole polynomials that'd take hundreds of cycles to execute.